Osteopontin: a key cytokine in cell-mediated and granulomatous inflammation

Source

Opn is synthesized by a variety of immune and non-immune cells (Table 1). It was identified as a phosphoprotein secreted by malignantly transformed epithelial cell lines, but was also independently isolated from bone and T cells (see below) (Senger et al. 1983; Franzen & Heinegard 1985; Patarca et al. 1989). In all cells studied Opn protein is expressed rapidly after cellular activation. Non-immune sources of Opn include epithelial cells, endothelial cells, smooth muscle cells, fibroblasts, osteoclasts/osteoblasts and tumour cells (Senger et al. 1983; Nagata et al. 1991; Brown et al. 1992; Merry et al. 1993; O'Brien et al. 1994; Qu et al. 1994; Liaw et al. 1995a; Ashizawa et al. 1996; Nau et al. 1997; Rittling & Denhardt 1999). Opn is expressed by T cells, macrophages (including macrophage-derived cells such as osteoclasts and Kupffer cells), and NK cells (including NK-derived granulated metrial cells found in the endometrium) (Franzen & Heinegard 1985; Nomura et al. 1988; Patarca et al. 1989; Waterhouse et al. 1992; Pollack et al. 1994; Nau et al. 1997; Kawashima et al. 1999). Opn is not produced by monocytes but its expression is rapidly up-regulated when monocytes are stimulated to differentiate into macrophages (Atkins et al. 1998).

Tissue Expression

Tissue expression reflects the cellular sources of Opn noted above. In normal tissue, it is expressed prominently in bone and at epithelial surfaces (Brown et al. 1992). In bone Opn is found mainly at the cement (renewal) lines and the lamina limitans (McKee et al. 1990; McKee et al. 1993). Epithelial expression is found in the respiratory (bronchial), reproductive (secretory endometrium and fallopian tubes), urinary (long loop of henle and distal convoluted tubules) and gastrointestinal (stomach, colon, gall bladder) tracts as well as mammary epithelium where it localizes predominantly to the lumenal surface (Brown et al. 1992). Significant concentrations of Opn have been detected in plasma, urine and breast milk (Senger et al. 1988; Senger et al. 1989; Bautista et al. 1996). During pathologic processes Opn is highly produced by injured and inflamed epithelium, endothelium, smooth muscle cells and certain tumour cells as well as inflammatory infiltrates involving T cells and macrophages (Patarca et al. 1989; Murry et al. 1994; O'Brien et al. 1994; Liaw et al. 1995a; Chambers et al. 1996; Nau et al. 1997; Giachelli et al. 1998; Tuck et al. 1998; O'Regan et al. 1999). Although elevated levels of Opn are present in the serum of patients with sepsis, it is not detectable in neutrophil-mediated lung diseases such as bronchopneumonia or acute respiratory distress syndrome (ARDS) (Senger et al. 1988; Nau et al. 1997; O'Regan et al. 1999).

Matrix Association of Opn

Opn has structural similarities to non-collagenous matrix proteins. In bone, Opn is indeed associated with the bone matrix. Opn has been found to associate with other matrix proteins, including fibronectins and collagen (Singh et al. 1990; Kaartinen et al. 1999). Opn is a substrate for tissue transglutaminase, which induces a conformational change, significantly increasing binding of Opn to collagen in vitro (Kaartinen et al. 1999). In other tissues, including inflammatory lesions, Opn is cell associated. Structural studies have suggested that multiple forms of Opn are expressed, varying at the level of phosphorylation and glycosylation. Singh et al. showed that phosphorylated Opn formed heat-stable complexes with cell-surface fibronectins, while non-phosphorylated, glycosylated Opn formed heat-dissociable complexes with plasma fibronectins (Singh et al. 1990). These data suggest that different forms of Opn may associate either with matrix or cells. Apart from bone matrix, there is no convincing evidence to date showing matrix association of Opn in vivo.

Biological activity

Opn is associated with a variety of pathological processes involving many different tissues and organs. The second part of this review will focus on the role of Opn in immune responses, specifically cell-mediated and granulomatous inflammation. Other work has demonstrated that Opn is associated with bone resorption, malignancy, atheromatous plaque formation and dystrophic calcification of inflamed and damaged tissues. While the precise role of Opn in these diverse responses is unknown, a common theme is its association with inflammatory stages of disease.

Opn was isolated as a marker of malignant transformation in epithelial cells (Senger et al. 1983). Further studies indicated that Opn expression was associated with cellular phenotypes exhibiting increased metastatic behaviour and it was hypothesized that the expression of Opn may facilitate tumour cell homing to lymph nodes via its ability to interact with CD44 (Behrend et al. 1994; Weber et al. 1996; Tuck et al. 1997). In vivo, Opn is expressed by a number of tumours, and recently Opn expression has been shown to be associated with a poorer prognosis in patients with breast and lung cancers (Brown et al. 1994; Chambers et al. 1996; Tuck et al. 1997; Tuck et al. 1998). Although Opn is expressed by only a subset of tumour cells, it is widely expressed by macrophages which infiltrate tumour tissue (Brown et al. 1994; Chambers et al. 1996). Liaw et al. demonstrated that Opn produced by tumour cells and macrophages fulfilled different roles (Crawford et al. 1998). Macrophage-derived Opn functioned as a chemoattractant and was associated with reduced tumour burden while tumour-derived Opn appeared to inhibit macrophage function and enhance tumour growth. The exact role of Opn in tumour biology remains unclear.

Opn is produced by both osteoclasts and osteoblasts and is one of the more abundant non-collagenous proteins in bone matrix (Merry et al. 1993; Rittling & Denhardt 1999). Its expression at cement lines and lamina limitans suggest that it may regulate bone resorption and repair (McKee et al. 1990; McKee et al. 1993). Osteoclasts adhere to Opn via the αVβ3 integrin and it is thought that this ligand–receptor interaction is central to the attachment of osteoclasts to bone and the subsequent formation of a clear zone of bone resorption (Ross et al. 1993). Although Opn-deficient knockout (KO) mice have normal bone structure and cellular composition, recent studies show that Opn deficiency protects from ovariectomy-induced bone loss (Liaw et al. 1998; Rittling et al. 1998; Yoshitake et al. 1999). Cultured bone marrow cells from Opn KO mice develop increased numbers of osteoclasts (Rittling et al. 1998). The significance of this observation is unclear but it is notable that Opn also inhibits macrophage fusion to form giant cells in vitro (Sterling et al. 1998). Overall Opn appears to augment bone loss in vivo.

Opn is also found in a variety of other pathological processes. It is expressed by smooth muscle cells and macrophages during vascular injury including atherosclerosis, stroke and pulmonary artery hypertension (Giachelli et al. 1995; Yoshitake et al. 1999; Cowan et al. 2000). It is found in macrophages and renal epithelial cells during tubulo‐interstitial inflammation of the kidney resulting from a variety of inciting agents ranging from ischemia and angiotensin II induced injury to glomerulonephritis (Giachelli et al. 1994; Pichler et al. 1994; Nambi et al. 1997). While the function, anti- or pro-inflammatory, of Opn in these processes is unclear, there is evidence that it may regulate macrophage recruitment and function, and act as a survival factor for epithelial cells, endothelial cells, and pro-B cells (Ophascharoensuk et al. 1999; Lin et al. 2000).

Liaw et al. studied the role of Opn during skin incisional wound repair using Opn KO mice (Liaw et al. 1998). Compared to wild-type mice, Opn mutants developed abnormal matrix organization and collagen fibrillogenesis with wounds containing homogeneous collagen fibrils of small diameter. Although a number of hypotheses were proposed to explain these findings, it was notable that wounds in Opn KO mice exhibited decreased levels of debridement consistent with abnormal macrophage function. Furthermore, a higher proportion of macrophages were mannose-receptor positive, a finding consistent with a resting, unactivated state. Other studies have shown that Opn may negatively regulate metalloproteinase expression (Cowan et al. 2000; Nemir et al. 2000). As such, defective macrophage recruitment and activation, and abnormal expression of matrix-degrading enzymes, are possible explanations for both aberrant wound debridement and abnormal collagen fibrillogenesis seen in the setting of Opn deficiency.

Finally, perhaps not surprisingly in view of its association with bone matrix, Opn is associated with dystrophic calcification of inflamed and damaged tissues including atherosclerosis and cardiac valve calcification (Srivatsa et al. 1997; Wada et al. 1999). In general, Opn appears to negatively influence tissue calcification, as well as the formation of renal stones in vivo, by inhibiting hydroxyapatite crystal formation (Bautista et al. 1996; Wada et al. 1999). The expression of Opn during dystrophic calcification correlates with macrophage and T-cell infiltration (Srivatsa et al. 1997). Once again Opn appears to function at the inflammatory phase of a pathological response.

In summary, although these data suggest that Opn is a multifunctional protein, in general Opn appears to regulate aspects of inflammation and tissue repair. In particular it is associated with responses characterized by the presence of macrophages and T cells. While the precise role of Opn in vivo remains unclear, studies using Opn gene-deficient mice will provide further insights into the physiologic and pathological role of Opn in vivo.

Opn receptor

Both RGD-dependent and -independent Opn cell receptor interactions have been demonstrated (Table 2). The best characterized Opn receptor is the αVβ3 integrin (vitronectin receptor). This receptor facilitates Opn adhesion of B cells, platelets, osteoclasts and smooth muscle cells (Reinholt et al. 1990; Liaw et al. 1994; Liaw et al. 1995b; Bennett et al. 1997). Opn can also interact with αVβ1, αVβ5, α9β1, α8β1, α4β1 and α5β1 integrins (Liaw et al. 1995b; Smith et al. 1996; Bayless et al. 1998; Denda et al. 1998; Barry et al. 2000). All integrins bind to the N-terminal thrombin cleaved fragment of Opn which contains the RGD domain. Unlike other integrins which recognize both the intact protein and the N-terminal fragment, α9β1 can only bind to the cleaved N-terminal fragment, presumably due to the liberation of a cryptic binding domain by thrombin digestion (Smith et al. 1996). Recently α9β1 adhesion domain has been identified as a seven amino acid sequence (SVVYGLR, human residues 162–168) located between the RGD site and the thrombin cleavage site (RS 168–169) (Yokosaki et al. 1999). Tyrosine (Y) 165 and Leucine (L) 167–Arginine (R) 168 were critical to α9β1 binding. The α4β1 binding site has not been identified but other integrin binding involves the RGD domain(Liaw et al. 1995b). CD44 (hyaluronic acid receptor) is an Opn receptor that also appears to have multiple non-RGD binding sites (Weber et al. 1996; Katagiri et al. 1999). Opn mediates chemotaxis and adhesion of fibroblast and T-cell lines through CD44 variant isoforms (v7–10) (Weber et al. 1996; Katagiri et al. 1999). Using cellular transfection, a variety of CD44 variant isoforms (in particular v6 and v7) but not the standard form of CD44, were shown to bind to Opn (Katagiri et al. 1999). These CD44 variants bind to both the N-terminal and C-terminal regions of Opn independently of the RGD sequence, suggesting multiple CD44 binding domains (Katagiri et al. 1999). In one study, CD44 binding of Opn was inhibited by antiβ1 integrin antibodies implying the interaction of CD44 and integrins at the level of cellular adhesion (Katagiri et al. 1999). The intracellular signalling pathways activated by Opn are unknown.

Expression of Opn by immune cells in vitro

The association of Opn with cell-mediated and granulomatous responses was first recognized in studies analysing gene expression during macrophage infection with mycobacteria (Nau et al. 1997). Using differential display, Opn was identified as the most prominent gene upregulated by murine peritoneal macrophages following infection with Mycobacterium (M.) bovis bacillus Calmette-Guerin (BCG) compared with intracellular challenge with Escherichia coli or Latex beads. Similar results were seen when human alveolar macrophages were infected with M. tuberculosis. Other studies have confirmed that Opn is highly expressed by activated but not resting mononuclear inflammatory cells including not only macrophages but also T cells and NK cells (Patarca et al. 1989; Pollack et al. 1994). In fact, Opn was independently described as an early T-cell activation gene (Eta-1), and its product was found to be the most abundant early RNA transcript in conA activated murine T cells (Patarca et al. 1989; Patarca et al. 1993). Murine T cells have also been shown to express Opn in response to mycobacterial challenge (Patarca et al. 1989).

The regulation of Opn expression is complex and tissue specific. However, analysis of the Opn promoter sheds some light onto its cell-specific expression. Among other motifs, the murine Opn promoter contains both a purine-rich sequence (PU Box) and an ets-like sequence, which may bind to the macrophage-specific, Pu.1, and lymphocyte-associated, ets, family of transcription factors, respectively (Young et al. 1990; Denhardt & Guo 1993; Patarca et al. 1993). The functional significance of these sites in Opn regulation has not been studied. Multiple regulatory sequences have been identified in the 5′ flanking sequences of both human and murine Opn. These include vitamin D response elements and interferon-inducible elements (Patarca et al. 1993). Using deletion analysis of promoter constructs a dominant positive regulatory site was identified at residues −124 to −80 in the human gene (Hijiya et al. 1994). The impact of these sites on the regulation of Opn expression is not known.

With specific reference to cell-mediated/granulomatous inflammation it is notable that Opn production is induced by a number of mediators that are abundantly expressed during these responses. Tumour necrosis factor (TNF) α is a critical cytokine expressed in early cell-mediated immunity and is essential for adequate granuloma formation. Macrophages stimulated with TNFα upregulate Opn mRNA and protein expression (Denhardt & Guo 1993; Patarca et al. 1993). In addition, a murine model of pulmonary fibrosis induced by transgenic over-expression of TNFα targeted to alveolar type II pneumocytes is associated with prominent early production of Opn (Nakama et al. 1998). In this model Opn expression correlated with macrophage and T-cell infiltration of the lung and waned as infiltrates receded and fibrosis ensued. Interleukin-(IL-) 1β and IL-2 are also associated with granuloma formation and can induce Opn expression in macrophages and NK cells, respectively (Patarca et al. 1993; Pollack et al. 1994). Finally, 1,2S-dihydroxycholecalciferol (vitamin D3) and angiotensin-II are characteristically expressed by macrophage-derived cells within granuloma lesions (Oldberg et al. 1986). The Opn promoter contains a vitamin D3 response element and simulation of macrophage cell lines with vitamin D3 induces Opn transcription and protein secretion (Denhardt & Guo 1993; Patarca et al. 1993). Angiotensin II can also induce Opn expression. Cardiac myocytes stimulated with angiotensin II upregulate expression of Opn mRNA. In addition, the effects of angiotensin II on rat cardiac fibroblasts, which include DNA synthesis and collagen gel contraction, appear to be mediated through angiotensin II dependent induction of Opn production by fibroblasts (Ashizawa et al. 1996). Taken together, these observations may explain the high level of Opn expression at early stages of cell-mediated immunity in vivo.

Expression of Opn in cell-mediated responses in vivo

Opn has been identified in a variety of granulomatous diseases including sarcoidosis, tuberculosis, silicosis, histoplasmosis, temporal arteritis, rheumatoid nodules and foreign body granulomas (Carlson et al. 1997; Nau et al. 1997; O'Regan et al. 1999). Using immunohistochemistry, Opn was found in T cells and macrophages as well as macrophage-derived epithelioid cells and multinucleated giant cells within granulomas (Figure 2) (Nau et al. 1997; O'Regan et al. 1999). Unlike other matrix proteins such as fibronectin and collagen, Opn was not expressed in the granuloma matrix (Nau et al. 1997; O'Regan et al. 1999). In fact, in a series of patients with different stages of sarcoidosis, immunoreactivity for Opn correlated directly with granuloma cellularity and inversely with granuloma fibrosis (O'Regan et al. 1999). In situ hybridization confirmed the expression of Opn during granulomatous responses, but identified Opn message only in macrophages and not in T cells (Carlson et al. 1997). At present it is not known whether T cells directly express Opn during granuloma formation or respond to macrophage-derived Opn. In contrast to granulomatous inflammation, Opn is not found in normal lung parenchyma or in neutrophil-mediated pulmonary diseases such as ARDS and bacterial bronchopneumonia (Nau et al. 1997; O'Regan et al. 1999)·

Opn is also expressed during experimental cell-mediated immune responses. In mice, Opn production marks an early T-lymphocyte dependent response to intracellular infection with Listeria (L.) monocytogenes and Orientia (Rickettsia) tsutsugamushi, or challenge with extracts of mycobacteria (Patarca et al. 1989; Patarca et al. 1993). In the latter studies, mouse strains resistant to Rickettsial challenge developed an early (within 48 h, peaking at day 3) surge of Opn gene induction, while sensitive strains had a lesser induction at later time points (6–7 days) (Patarca et al. 1989). These studies provide evidence of Opn expression in cell-mediated diseases in vivo.

Opn regulates inflammatory cell function

Critical events in cell-mediated and granulomatous inflammation include leukocyte recruitment and activation with cytokine production. Recent studies have defined a role for Opn in these processes. Opn binds to activated T cells and macrophages (Patarca et al. 1989; O'Regan et al. 1999). The biological consequences of these interactions are discussed below.

Cellular recruitment

Opn is chemoattractant to a variety of cell types (Patarca et al. 1989; Patarca et al. 1993; Senger et al. 1994; Liaw et al. 1995b; Yamamoto et al. 1995; O'Regan et al. 1999). As mentioned previously above, studies using both primary cells and cell lines have demonstrated that Opn can induce migration of both T cells and macrophages in vitro (Yamamoto et al. 1995; Weber et al. 1996; Giachelli et al. 1998; O'Regan et al. 1999). The migratory response is dose dependent (range 1 nm to 300 nm) and exhibits high-dose inhibition. The dose range may reflect the receptor involved in mediating migration; Opn-dependent αVβ3 migration peaks in the 100–300 nm range while CD44-mediated migration occurs at 1–3 nm doses (Liaw et al. 1994; Weber et al. 1996). In the case of T cells, checkerboard analysis confirmed that Opn-dependent migration was chemotactic (gradient dependent) rather than chemokinetic (random migration) in nature (Weber et al. 1996; O'Regan et al. 1999). Checkerboard analysis has not been described for monocyte-macrophage migration. The ability of Opn to induce B cell, NK cell or dendritic cell migration has not been studied.

In vivo studies confirm the ability of Opn to induce cellular migration. Macrophages accumulate at sites of subcutaneous injection of native or recombinant Opn (300 ng) in mice and rats (Singh et al. 1990; Giachelli et al. 1998). Inflammatory cells were present within 3 hours of injection and at 24 h showed a 10-fold increase compared to control (in rats 2-fold increase). The majority of infiltrating cells were macrophages (25-fold increase over control at 24 h) with a less pronounced accumulation of neutrophils (5-fold increase). At the doses of Opn studied, T cells, NK cells and B cells did not accumulate at sites of subcutaneous Opn injection. Another study demonstrated that the cellular response induced by subcutaneous injection of polyvinyl pyrrolidone (PVP) was deficient in C57BL/6 nu/nu mice and that co-injection of native Opn (purified from a T-cell line) retrieved the cellular migratory response. Adding back Opn (10 µg) resulted in cellular accumulation equivalent to that seen in C57BL/6 mice injected with PVP and was comprised of 85% macrophages. (Ashkar et al. 2000). In keeping with the hypothesis that Opn is critical for cellular recruitment, inflammatory cell accumulation and granuloma formation after injection of PVP, collagen or latex was markedly reduced in Opn knock-out (KO) mice (Ashkar et al. 2000).

In addition to direct receptor-mediated chemotactic effects, Opn may modulate cellular migration by inducing the expression of other chemotactic cytokines (chemokines), or by facilitating cellular migration in response to other chemotactic agents. In regard to the last possibility, Opn has been shown to facilitate the chemotactic response of macrophages to the chemotactic peptide n-formyl-met-leu-phe (fMLP) (Giachelli et al. 1998). The cutaneous inflammatory response elicited by injection of fMLP is characterized by high macrophage expression of Opn and it is inhibited by intravenous treatment with an antibody to Opn. Thus, Opn may facilitate macrophage migration to other chemoattractants. The effect of Opn on induction of chemokines has not been studied.

In vivo studies of pathological responses also support a role for Opn in macrophage recruitment. Renal injury after ureteral obstruction is characterized by prominent Opn expression and the accumulation of macrophages (Ophascharoensuk et al. 1999). Similar injury in Opn-KO mice is associated with a 5-fold reduction in macrophage infiltration compared to wild-type controls (Ophascharoensuk et al. 1999). Other studies in Opn-KO mice have demonstrated normal numbers of macrophages at sites of inflammation suggesting that migratory defects are insult and/or tissue specific (Liaw et al. 1998; Nau et al. 1999).

The in vitro migratory and adhesive effects of Opn are both RGD- and non RGD-dependent. Singh et al. demonstrated that binding of resident peritoneal murine macrophages and a myelomonocytic cell line (Wehi 3B) to Opn was inhibited by pre-incubation with an RGD peptide but not with the mutated peptide LGL (Singh et al. 1990). Similarly, both the adhesion and migration of another murine macrophage cell line (P388D1) to Opn was abrogated by mutation of the Opn RGD binding domain or by pre-incubation with RGD peptide (Yamamoto et al. 1995). Migration but not adhesion was inhibited by an anti‐αv integrin monoclonal antibody (Yamamoto et al. 1995). A recent study showed that WEHI 3B binding of Opn was inhibited by rat monoclonal antibodies to CD44 (Weber et al. 1996).

Receptors mediating T-cell interactions with Opn are poorly defined. Weber et al. have demonstrated that Opn induces chemotactic migration to a murine T-cell hybridoma that naturally expresses CD44 and that this migration is inhibited by pre-incubation with anti-CD44 monoclonal antibodies (Weber et al. 1996). αVβ3 dependent T-cell interactions have not been demonstrated but αVβ3 is expressed on human and murine T cells (Klingemann & Dedhar 1989; Gerber et al. 1996). Lymphocytic and myeloid cell lines have been shown to adhere to Opn via the α4β1 integrin (Bayless et al. 1998). Finally, both B cells and platelets can bind Opn via the αVβ3 integrin (Bennett et al. 1997). These results demonstrate the Opn can interact with immunocytes through both αVβ3 integrin and CD44.

Cytokine expression

The Th1/Th2 cytokine paradigm of cellular immunity has been broadly applied to granuloma formation. The Th-cytokine pattern impacts the type of granuloma formed, its intensity, cellular composition and fibrogenic potential (Chensue et al. 1994; Modlin 1994). In hypersensitivity granulomatous responses like sarcoidosis, Th1 cytokines predominate and lack of functional interferon (IFN) γ, TNFα and IL-12 results in defective granuloma formation (Moller 1999). The presence of cytokines such as IL-12 and IL-10 during priming of uncommitted T cells plays an important role in subsequent Th1 or Th2 differentiation, respectively (Hsieh et al. 1993; Chensue et al. 1996; Agostini et al. 1998; Moller 1999).

Opn enhances Th1 and inhibits Th2 cytokine expression. In one study, soluble phosphorylated Opn (5 nm) directly induced murine peritoneal macrophages to produce IL-12. In contrast, soluble de-phosphorylated Opn (5 nm) inhibited IL-10 expression by LPS-stimulated murine peritoneal macrophages (Ashkar et al. 2000). Endotoxin contamination was not responsible for Opn-dependent IL-12 production because macrophages from C3H.Hej mice which are defective in endotoxin signalling produced similar amounts of IL-12 in response to Opn stimulation. The induction of IL-12 was inhibited by GRGDS peptide (but not GRADS) and by antibody to the integrin β3 subunit. In contrast it was not dependent on CD44, since anti-CD44 antibody did not inhibit Opn-dependent IL-12 production and macrophages from CD44 KO mice displayed an unimpaired IL-12 response when stimulated with Opn. A 10-kDa fragment of Opn (NK10) obtained by Lys-C digest of the NH2-terminal portion of Opn was sufficient to induce IL-12 expression from macrophages. The NK10 peptide runs from the Lys-Gln (residues 69–70) to the thrombin cleavage site (Arg-Ser 154–155) containing the RGD integrin binding domain. The presence and function of such a peptide in vivo has not been demonstrated. In contrast to IL-12 production, inhibition of IL-10 expression by Opn depended on engagement of the CD44 receptor: IL-10 inhibition was blocked by anti-CD44 antibody (but not by antibody to the β3 integrin), and macrophages from CD44 KO mice were resistant to Opn inhibition of the IL-10 response. In summary Opn augments IL-12 expression from macrophages in a β3 integrin-dependent manner and inhibits IL-10 expression from macrophages via an interaction with CD44.

In Opn-KO mice, infections characterized by Th1 cytokine expression (herpes simplex virus [HSV] type 1; L. monocytogenes) were associated with reduced IL-12 and increased IL-10 expression (Ashkar et al. 2000). Herpes simplex virus [HSV] type 1 infection can lead to a destructive type 1 autoimmune inflammatory reaction called HSV keratitis. This response depends on the production of IL-12 and is inhibited by IL-10. Within 10–14 days of HSV-1 infection, 65% of Opn+/+mice developed HSV keratitis while Opn KO did not develop this disease. Draining lymph node cells from HSV virally infected Opn KO mice produced exaggerated amounts of IL-10 and IL-4 and reduced IL-12 compared to Opn+/+controls. Th1 responses were also defective in Opn KO mice infected with L. monocytogenes. The murine response to L monocytogenes is critically dependent on early macrophage production of IL-12 and downstream expression of IFNγ. Opn KO mice were defective in their ability to clear L monocytogenes after systemic infection, similar to the defect seen in IL-12 deficient mice, and stimulation of splenocytes obtained from infected Opn KO mice exhibited reduced IFNγ expression compared to Opn+/+mice. These reciprocal effects of Opn on IL-12 and IL-10 expression suggest that Opn may polarize early Th1 cytokine responses.

Opn also co-stimulates human T-cell proliferation and augments early Th1 cytokine expression from human mononuclear cells. In situ proliferation accounts in part for the accumulation of inflammatory cells within granulomas. Opn (15 nm) increases CD3-mediated T-cell production of IL-2 and the p55 component of the IL-2 receptor (CD25) (O'Regan et al. unpublished observation), and T-cell proliferation (O'Regan et al. 1999). In addition, Opn co-stimulation of T cells enhances CD3-dependent induction of IFNγ and CD40 ligand (L) and via the induction of these molecules, augments T-cell dependent IL-12 production by human monocytes in vitro (O'Regan et al., 2000). Enhanced T cell expression of CD40L may also explain the reported ability of Opn to induce B-cell proliferation and antibody production (Patarca et al. 1990; Lampe et al. 1991; Patarca et al. 1993). Opn did not augment Th2 cytokine (IL-4) expression by T cells in vitro.

Macrophage differentiation

Opn has been shown to modulate a number of other cellular functions that have relevance to granulomatous inflammation and cell-mediated immunity. Opn is associated with monocyte–macrophage differentiation. Monocytes do not express Opn, even upon activation, until they assume characteristics of macrophages (Atkins et al. 1998). Cell lines (e.g. HL-60) that are induced toward macrophage differentiation by treatment with phorbol ester express Opn mRNA and protein within 6 and 24 h, respectively (Atkins et al. 1998). In these experiments the integrin subunits, αV and β3, were similarly induced and the β1 subunit and CD44 were up-regulated coincident with macrophage differentiation. In contrast, induction of granulocytic differentiation with retinoic acid does not affect Opn, αV, β1, β3 or CD44 expression.

Opn has also been associated with giant cell formation (Nau et al. 1997; Sterling et al. 1998; O'Regan et al. 1999). Opn was independently isolated from osteoclasts, which are macrophage-derived multinucleated giant cells associated with bone resorption (Oldberg et al. 1986; Merry et al. 1993). Giant cell formation is also characteristic of granulomatous inflammation. Recent studies have shown that CD44 is involved in mediating the fusion of macrophages to form giant cells (Sterling et al. 1998). Opn is a CD44 ligand and treatment of rat macrophages with Opn inhibits giant cell formation in vitro (Sterling et al. 1998). The in vivo significance of these studies has not been studied.

Apoptosis

Opn may also function as a cell survival factor. Recent studies have shown that apoptosis is an important part of the early host response to mycobacterial and other intracellular infections (Keane et al. 1997). Furthermore, murine tuberculosis is characterized by the expression of FAS ligand on macrophages at the lymphocyte–macrophage interface within granulomas (Mustafa et al. 1999). This suggests that FAS ligand-mediated apoptosis is a potential mechanism of immune evasion by M. tuberculosis in granulomatous inflammation. In sarcoidosis, extensive macrophage apoptosis and T-cell FAS ligand expression within granulomas has been detected. These authors suggest that apoptosis of epithelioid histiocytes and inflammatory cells may participate in the course of granulomatous inflammation (Kunitake et al. 1999). Opn has been shown to inhibit apoptosis in smooth muscle, endothelial cells, epithelial cells and pro-B cells (Ophascharoensuk et al. 1999; Rittling & Denhardt 1999; Cowan et al. 2000; Lin et al. 2000). Its role in the regulation of inflammatory cell survival and death signal expression has not been studied.

Inhibition of nitric oxide

Although the previous discussion outlines the role of Opn as a pro-inflammatory cytokine, an anti-inflammatory function is suggested by Opn-dependent inhibition of nitric oxide production by macrophages in vitro (Rollo et al. 1996; Rittling & Denhardt 1999).· Macrophages stimulated with LPS and IFNγ up-regulate iNOS and NO production (Rollo et al. 1996; Rittling & Denhardt 1999). Treatment with Opn (pM-nM range) substantially inhibits (by 40–50%) both iNOS and NO production by murine macrophage (RAW 264.7 macrophage cell line). This response was dose-sensitive such that 100-fold higher or lower concentrations had no effect on iNOS expression. In various pathological responses, there are indications that reactive nitrogen intermediates cause damage to tissues. For example, renal ischemia results in tissue damage which is in part mediated by nitrogen metabolites (Noiri et al. 1996). In Opn KO mice, clamping of the renal artery results in increased structural and functional damage to the kidney than in Opn+/+mice (Noiri et al. 1999). Increased levels of iNOS and nitrotyrosine, an indicator of NO levels in vivo, were dramatically elevated in the post-ischemic Opn deficient kidneys compared to post-ischemic Opn+/+kidneys. While the role of NO production is important in cell-mediated responses to intracellular infection in mice, such a role has not been clearly demonstrated in humans and its role in granuloma formation is unknown. Nevertheless, these data provide evidence that Opn may regulate NO production in vivo.

Tissue repair, fibrosis and calcification

As outlined above, Opn is associated with tissue repair, and may regulate fibrosis and dystrophic calcification after immunological injury (Miyazaki et al. 1995; Liaw et al. 1998; Nakama et al. 1998; Ophascharoensuk et al. 1999)· The initiation and maintenance of granulomatous inflammation is followed by a phase of repair and resolution. During this phase granulomas may completely resolve or may undergo fibrosis with significant tissue destruction. Determinants of granulomatous fibrosis are poorly defined but include host (genetic, racial) and aetiologic (type of antigen) factors. Opn is expressed in a number of pathological responses characterized by significant tissue fibrosis including idiopathic pulmonary fibrosis (IPF) and two murine models of pulmonary fibrosis (bleomycin injury and TNFα transgenic models) (Nakama et al. 1998). In these models Opn is prominently expressed in early stages characterized by interstitial pneumonitis. In fact, a recent study using cluster analysis showed that Opn was a member of a small group of genes dramatically up-regulated by bleomycin treatment in fibrosis-sensitive mice but not in mice rendered fibrosis-resistant by a null mutation of the integrin subunit β6 (Kaminski et al. 2000). The functional significance of Opn in lung fibrosis is unknown.

Opn is also expressed during fibrotic responses in other organs. In a model of renal interstitial fibrosis in mice, expression of Opn was demonstrated in renal epithelial cells and infiltrating macrophages (Ophascharoensuk et al. 1999; Rittling & Denhardt 1999). In Opn KO mice there was less interstitial fibrosis (deposition of type I and IV collagen) in comparison to control mice and a reduction in macrophage infiltration and TGF-β expression (Ophascharoensuk et al. 1999). Therefore Opn may modulate fibrosis by regulating fibrogenic cytokine expression either directly at the cellular level or indirectly by recruiting fibrogenic inflammatory cells.

Tissue repair can also result in dystrophic calcification. Many, although not all, granulomatous diseases culminate in granuloma calcification. Again, the mechanism of granuloma calcification is unknown. As Opn is a bone matrix protein associated with dystrophic calcification in other inflammatory responses and is prominently expressed by granuloma reactions, it is possible that Opn may also regulate aspects of granuloma calcification. Recent studies have suggested that dystrophic calcification is an actively regulated process of cellular inflammatory responses and is, at least in part, reversible in a manner similar to bone deposition and resorption (Demer & Tintut 1999). Opn has been associated with bone loss and inhibition of tissue calcification (see above). The role of Opn in granuloma fibrosis and calcification merits further investigation.

Opn receptors and granuloma formation

The Opn receptors αVβ3 integrin and the CD44v6–10 possess properties relevant to early granuloma formation. As well as supporting adhesion and migration, ligation of αVβ3 induces IFNγ production by NK cells (Rabinowich et al. 1995). CD44 exerts a complex role on cell trafficking. Neutralization of CD44 abrogates DTH reactions in vivo and inhibits giant cell formation in vitro (Rosel et al. 1997; Sterling et al. 1998). The Opn receptor(s) on primary T cells and monocytes have not been defined. However, it has been demonstrated that the interaction of Opn with CD44v7–10 supports chemotaxis and adhesion of a murine T-cell hybridoma and monocyte cell line (Weber et al. 1996). Murine peritoneal macrophages can bind Opn in an RGD-dependent manner and Opn adheres to B cells, platelets and osteoclasts via the αVβ3 integrin (Singh et al. 1990; Bennett et al. 1997; Bennett et al. 1997). As outlined above, the induction of IL-12 from macrophages by Opn is β3 integrin dependent while its ability to inhibit IL-10 expression is CD44 dependent (Ashkar et al. 2000).

Regulation of Opn receptor binding by thrombin

RGD-dependent adhesive activity of Opn is modulated by thrombin cleavage (Senger et al. 1994; Xuan et al. 1995; O'Regan et al. 1999). Opn contains a thrombin cleavage site in close proximity to its RGD-binding domain (Patarca et al. 1993; Young et al. 1990). Opn is actively and specifically cleaved at this site both in vitro and in vivo (Senger et al. 1988; Senger et al. 1994). Studies evaluating the effect of thrombin cleavage on Opn activity have reported conflicting results. One report demonstrated that thrombin treatment reduced RGD-mediated adhesion to recombinant human and native bovine osteopontin (Xuan et al. 1995). Conversely, it has been shown that cleavage of native human osteopontin augments adhesion to primarily human T cells as well as a variety of human cell lines including fibroblasts, smooth muscle cells and tumour cells (Senger et al. 1994; O'Regan et al. 1999). By generating peptides corresponding to the cleaved fragments of osteopontin, it has also been shown that the N-terminal fragment, which contains the RGD domain, but not the C-terminal fragment of Opn, can support adhesion of a human melanoma cell line (Smith et al. 1996). These conflicting studies do not represent mutually exclusive results. Osteopontin contains another potential thrombin cleavage site within the RGD domain and it is possible that different thrombin cleavage conditions or variable access to this thrombin cleavage site could disrupt the RGD sequence and RGD-mediated adhesion (Patarca et al. 1993; Smith et al. 1996). Furthermore, since Opn binds a number of receptors which are differentially expressed on different cells and cell lines, it is possible that some interactions are augmented while others are inhibited by thrombin cleavage. In general, thrombin cleavage appears to up-regulate Opn-dependent cellular adhesion in vitro.

The role of thrombin regulation of osteopontin adhesion has interesting implications in the pathogenesis of granuloma formation. There is ample data suggesting that active thrombin and procoagulant activity is present during granuloma formation. Increased bronchoalveolar lavage procoagulant activity and the presence of circulating d-dimers have been demonstrated in pulmonary sarcoidosis (Hasday et al. 1988; Perez et al. 1993). Moreover, Perez reported that mice susceptible to granuloma-inducing antigens from M. tuberculosis have increased procoagulant activity, while non-susceptible mice have increased plasminogen activator activity (Perez et al. 1994). This suggests that activation of the T-cell adhesive properties of Opn represents a novel role for thrombin in granuloma formation and that the immunomodulatory effects of the coagulation system may, in part, be delivered through its interaction with Opn. The impact of inhibition of coagulation on granulomatous inflammation is unknown. However, heparin treatment has been shown to inhibit delayed type hypersensitivity reactions and subcutaneous heparin has been used successfully in the treatment of ulcerative colitis (Cohen et al. 1967).

Allelic variation in Opn expression

Defective Opn expression is associated with abnormal cell-mediated and granulomatous immunity in mice. The Opn gene maps to the murine rickettsial resistance (ric^r) locus on chromosome 5 (Fet et al. 1989; Patarca et al. 1989). Alleles of the ric^r gene (alleles b and c) that result in deficient early (24–48 h) production of Opn are associated with susceptibility to Orientia tsutsugamushi infection, while allele a, associated with an early high Opn response, results in disease resistance (Patarca et al. 1989; Patarca et al. 1993). The cDNA's of alleles a and b have been sequenced (Ono et al. 1995). At least 10 amino acids were different between the alleles, though the principal active sites of the molecule were not involved except for an amino acid substitution in a calcium binding site (K for R at residue 208) and in one heparin binding site (Y for H at 279).

Opn allelic variants have been identified in humans using RFLP of restriction enzyme digests (Young et al. 1990). Using 29 normal individuals, digestion with bglII resulted in a simple two allele RFLP. The larger variant (11kb) was predominant over the smaller 9.1 kb allele, 64 : 36. A similar two allele result was found with MspI, with 6.4 kb and 5.1 kb alleles with a frequency of 12 : 88. The phenotypic effects of such alleles and their role in disease pathogenesis has not been studied.

Opn knock-out mice

Opn-KO mice develop deficient Th1 antimicrobial immunity to a broad range of pathogens. Specifically, Opn-KO mice are defective in their ability to clear L. monocytogenes after systemic infection (Ashkar et al. 2000). Likewise, when infected with HSV via the cornea, Opn-KO mice do not develop delayed-type hypersensitivity to HSV or HSV-associated Th1 autoimmune keratitis (see above) (Ashkar et al. 2000).

Opn-KO mice display defective antimicrobial control when challenged intraperitoneally with M. bovis BCG (Nau et al. 1999). The granulomas that develop demonstrate defective mycobacterial killing resulting in increased bacterial burden in liver and spleen (Nau et al. 1999). This was manifested by increased numbers of organisms seen on tissue sections, and in the number of organisms cultured from tissue. The ability of Opn to facilitate the clearance of mycobacteria was most pronounced early after infection (4 weeks) and appeared to be independent of known mediators of resistance to infection by mycobacteria (antigen-specific T-cell immunity, IFNγ production), and nitric oxide production. However, an intrinsic macrophage defect was suggested by demonstrating that BCG grew more rapidly in peritoneal exudate elicited macrophages derived from Opn mutant mice compared to wild-type mice. Despite these findings showing defective control of the infection, the hepatic and splenic granulomas that developed in response to BCG in these mice were of similar size and cellular composition to those in wild type littermate controls. These data suggest that while Opn-KO animals could form granulomatous cellular infiltrates in this model of mycobacterial infection, these granulomas were functionally defective.

Data derived from the PVP skin granuloma model has suggested that granuloma formation may also be deficient in the Opn-KO animal (Ashkar et al. 2000). This model is dependent on Th1 cytokine expression as illustrated by a diminished granuloma formation in IL-12-KO mice and an enhanced response in IL-10-KO mice (Ashkar et al. 2000). Opn-KO mice completely fail to form granulomas when challenged intradermally with PVP, latex or collagen (Ashkar et al. 2000). Together these data suggest an important role for Opn in the formation and function of Th1 granulomas.

Primary Opn deficiency has not been described in humans but secondary Opn deficiency may be concomitant with other immune defects. Disseminated mycobacterial infection and poor granuloma formation occur in humans with inherited defects of the IFNγ receptor. In contrast to the abundant expression in granulomas of immune competent hosts, Opn is not present in the granulomatous response of these patients to M. bovis BCG (Nau et al. 2000). Opn expression correlated inversely with dissemination of infection and death. These data suggest that the IFNγR1 may be involved in the regulation of Opn expression in human mycobacterial infection. In summary, in human and murine hosts, defective Opn expression is associated with alterations in granuloma and cell-mediated pathology.