Integrins and their ligands in rheumatoid arthritis

Integrins play an important role in cell adhesion to the extracellular matrix and other cells. Upon ligand binding, signaling is initiated and several intracellular pathways are activated. This leads to a wide variety of effects, depending on cell type. Integrin activation has been linked to proliferation, secretion of matrix-degrading enzymes, cytokine production, migration, and invasion. Dysregulated integrin expression is often found in malignant disease. Tumors use integrins to evade apoptosis or metastasize, indicating that integrin signaling has to be tightly controlled. During the course of rheumatoid arthritis, the synovial tissue is infiltrated by immune cells that secrete large amounts of cytokines. This pro-inflammatory milieu leads to an upregulation of integrin receptors and their ligands in the synovial tissue. As a consequence, integrin signaling is enhanced, leading to enhanced production of matrix-degrading enzymes and cytokines. Furthermore, in analogy to invading tumors, synovial fibroblasts start invading and degrading cartilage, thereby generating extracellular matrix debris that can further activate integrins.

Integrins are a large group of transmembrane proteins that anchor the cell to the extracellular matrix (ECM) or other cells. Upon binding, integrins remodel the ECM by inducing the expression of certain proteases. Integrins control cellular proliferation, migration, and invasion. Whereas in some cases integrins can mediate these eff ects on their own, they usually signal in the context of growth factor or cytokine receptors [1]. Ligand binding to integrin controls adhesion-dependent proliferation, whereas unligated integrins enhance apoptotic pathways [2]. Th is is one reason why dysregulated integrin expression or downstream signaling pathways can be observed in almost all forms of cancer, and integrin levels often determine the aggressiveness or propensity to metastasize.
Migration, invasion, and proliferation of synovial cells are major problems in rheumatoid arthritis (RA) [3]. Th is debilitating disease is characterized by an infl amed synovial tissue with a massive infl ux of immune cells and an infl ated synovial lining consisting mainly of synovial fi broblasts (SFs) and macrophages that adhere to the adjacent ECM. Th ese SFs are highly proliferative and contribute signifi cantly to cartilage and bone destruction. In some aspects, they can be considered 'tumor-like' as SFs are resistant to many apoptotic pathways, show increased proliferation, and produce high amounts of matrix metalloproteinases (MMPs) because of dysregu lation of the transcription factors AP-1, p53, and nuclear factor-kappa-B (NF-κB) [4]. Fibroblasts, macrophages, but also endothelial cells in synovial tissue show high levels of integrin expression [5]. In RA, not only integrins but also their ligands (for example, fi bronectin and collagen as well as their degradation products) are upregulated [6]. Th is 'overstimulation' of cells by integrinmediated signals increases basal secretion of proinfl ammatory cytokines like interleukin-6 (IL-6) and IL-8 but also levels of various MMPs [7][8][9].
Th is review introduces the predominant integrins expressed on synovial cells along with their binding partners and demonstrates their signifi cance in disease progression. We focused mainly on the β1, β3, and β5 integrins because altered levels of their ligands and respective integrins themselves are one major factor that fuels the destruction of cartilage in RA. Although integrins infl uence RA signifi cantly, only a few reviews that address integrins in RA have been published. A Medline search for 'integrins' and 'arthritis' yielded 108 (compared with 'integrins' and 'cancer' with over 1,200 reviews) reviews, most of which outlined single integrins or cell types. Two reviews in 1993 and 1995 -by Postigo and colleagues [10] and Liao and Haynes [11], respectively -were the last comprehensive articles describing the infl uence of integrins on the pathogenesis of RA. Both articles, however, failed to describe the intricacies in integrin biology, from the generation of integrin ligands from the ECM to the subsequent triggering of integrin signaling cascades that fi nally lead to the generation of pro-infl ammatory factors and cartilage degrada tion. To our knowledge, this is the fi rst review that demonstrates the whole integrin machinery in arthritis, including ligands, receptors, and eff ector cells.

Integrin biology
A cell needs to sense the composition of the extracellular environment to elicit appropriate responses. Integrins, a family of heterodimeric proteins, serve as sensors of the ECM. Integrins consist of an α subunit that determines ligand specifi city and a β subunit that initiates intracellular signaling events. Whereas most integrins bind almost exclusively to ECM molecules, β2 hetero dimers bind other adhesion molecules, mediate cell-cell contacts, and are important components of the immune system. In contrast to the β2 integrin, which is expressed only on immune cells, other β subunits have a much more diverse expression. In the synovium, fi broblasts, macrophages, and endothelial cells are just a few examples of cells that express the β1, β3, and β5 subunits. Th e ECM-binding integrins mediate cellular processes such as adhesion, migration, tissue invasion, and chemotaxis but also cytokine and MMP production.
Within the protein sequences of fi bronectin, collagen, and vitronectin, integrins bind to arginine-glycineaspartic acid (RGD) motives. Interaction with the RGD sequence is an important switch for integrin activation and subsequent cellular changes. Integrins can transduce signals from outside the cell to the interior and vice versa, allowing the cell to respond dynamically to a wide variety of stimuli. Growth factors, hormones, and cytokines as well as the composition of the ECM are able to modulate integrin signaling and affi nity [12]. When integrins are activated by ligands, several signaling events are initiated. Besides the focal adhesion kinase (FAK), the mitogenactivated protein kinase (MAPK) pathway, the phosphatidyl inositol 3-kinase (PI3K), and c-Jun N-terminal kinases (JNKs) are activated [13]. Growth factors and cytokines employ the same signaling pathways as integrins, demonstrating an intricate interplay between the two. Integrins are linked to the cytoskeleton by coupling to adaptor proteins such as vinculin and talin. Th erefore, changes in integrin activation lead to changes in the actin cytoskeleton, a prerequisite for adhesion, migration, and invasion on ECMs.

Integrin knockouts and related pathologies
Given that various integrin combinations recognize the same ECM molecule (for example, fi bronectin is recognized by α4β1, α5β1, ανβ1, ανβ3, and α3β1 and weakly by others), one would expect some redundancy in integrin function. In fact, this is the case when collagen-binding integrin subunits α1 or α2 are knocked out. When α1 is knocked out, corresponding mice show decreased vascula ri zation, and isolated endothelial cells from these animals show reduced proliferation whereas α2 knockouts demonstrate only a subtle phenotype. Th e picture is diff erent when regarding the knockout of fi bronectinbinding integrin subunits. Knockout of either integrin α3, α4, α5 or αν is lethal during embryogenesis and respective mice die shortly after birth. Also, the fi bronectinbinding integrins have a more tissue-specifi c distribution than collagen-binding integrins, making target cells more susceptible to specifi c knockouts.
Besides having a role in development, integrin mutations lead to a variety of other defects. Leukocyte adhesion defi ciency type 1 (LAD-1) is an autosomal recessive disorder based on point mutations within the β2 integrin, resulting in either its absence on the cell surface or the inability to associate with the corresponding α subunit. Th is leads to reoccurring bacterial infections, suggesting an impaired function of macrophages and neutrophils.
Another autosomal recessive (although an autoimmune form exists) disease is the Glanzmann thrombasthenia, in which the platelet integrin αIIbβ3 fails to bind to fi brinogen and fi brin. Th is step is important to form blood clots after injury, and failure to do so leads to prolonged bleeding.
Integrins are also involved in the generation and perpetuation of cancer. In several tumors, expression of specifi c integrin subunits is associated with malignancy. In melanoma, increased expression of ανβ3 and α5β1 promotes vertical growth and lymph node metastasis. In non-small-cell lung carcinoma, high α5β1 levels predict decreased patient survival. In general, alterations in integrin activation or expression promote tumor growth and metastasis. However, in some cases (for example, α2β1 in breast cancer), integrins can be protective by decreasing cell dissemination [14]. Th is integrin increases diff erentiation of epithelium and enhances ECM binding, thus decreasing migration.

Integrins in synovial tissue
When the integrin distribution in the synovial tissue is analyzed, several factors need to be considered. First, integrin expression and function diff er depending on cell type. Second, the location of the cell is important as expression levels diff er at the synovial lining compared with the sublining. Every cell in the synovial tissue expresses a specifi c subset of integrins, depending on lineage and origin. Several groups have investigated synovial tissue samples and isolated cells to study integrin expression. Th e integrin subunit β1 is most widely expressed since it can be found on macrophages, lymphocytes, endothelial cells, and fi broblasts [5,15,16]. Th e fi bronectin receptor α5β1 and the laminin receptor α3β1 were also found on every synovial cell type, whereas α4β1, which besides fi bronectin also binds VCAM-1, was found only on lymphocytes. Endothelial cells and fi bro blasts also express the collagen receptor α1β1 and the vitronectin receptor ανβ3 [17]. In synovial tissue, integrin expression also depends on the location of the particular cell. Expression of most integrins is similar throughout the synovial tissue but is diff erent at the synovial lining layer [5]. Th is is the area where synovial cells adhere to cartilage and activated fi broblasts and macrophages degrade ECM and invade cartilage. Th e laminin receptor α6β1 is expressed by fi broblasts at the lining layer but not by macrophages, which usually express this receptor [18]. At the synovial lining, we found an increase in α5, αν, and β1 integrin levels compared with the sublining area. Th is clearly indicates that increased integrin expression is associated with a more 'destructive' phenotype of the corresponding cell. Figure 1 depicts the integrin-ligand interactions and their consequences (for example, proliferation and secretion of pro-infl ammatory mediators) described in the next three sections. Th e most important cell types (fi broblasts, endothelium, T cells, B cells, macrophages, and Th 17 cells), including their individual reaction to integrin engagement, are depicted ( Figure 1). Th e individual integrin levels on each cell type are presented in Table 1. Th is table refl ects our own observations of tissue samples stained for individual integrins and fl ow cytometric analyses as well as literature data.

The collagen-binding integrins α1β1 and α2β1
Collagens are widely expressed in synovial tissue, and, during the course of RA, a signifi cant proportion of patients develop antibodies against at least one collagen subtype. Th e collagens II, IX, X, and XI are restricted to cartilage, whereas all other forms can also be found within the ECM. Th e classical collagen-binding integrins rely on the triple-helical structure for binding [19]. However, if collagen is proteolytically degraded, RGD sequences are exposed and other RGD-binding integrins can bind. In this case, ανβ3, αIIbβ3, and α5β1 can be activated by collagen-derived RGD sequences [20].
Four diff erent integrin combinations recognize native collagen: α1β1, α2β1, α10β1, and α11β1. Integrin α10β1 is important for chondrogenesis, whereas α11β1 has a role in the recognition and organization of interstitial collagen matrices during development. During the course of arthritis, only α1β1 and a2β1 seem to play a role; this is because these are the two major collagen-binding integrins in the synovial tissue. Although all of the abovementioned subunits bind to collagen, their respective specifi cities diff er for each collagen subtype; for example, integrin α1 preferentially binds collagen type IV.
Th e integrin α1 was found to be expressed at the synovial lining and in blood vessels of arthritic joints. Interestingly, expression was increased compared with osteoarthritis and trauma controls, suggesting that the pro-infl ammatory milieu in RA promotes integrin α1 synthesis [21]. However, our group also found an increase of this adhesion receptor on synovial fi broblasts after treat ment with cortisol [5]. An increase of α1 -in analogy to integrin α5, which is also upregulated by cortisolpoten tially increases adhesion to collagen. Owing to cortisol action, the increase in integrin expres sion is uncoupled from integrin signaling because intra cellular signaling pathways are silenced by glucocorticoid treatment.
Lymphocytes also express integrin α1, but only synovial lymphocytes show increased expression of this receptor [22]. One important modulator for α1 levels is tumor necrosis factor (TNF) because anti-TNF therapy reduces the number of α1-positive eff ector memory T cells that contribute signifi cantly to infl ammation by producing large amounts of interferon-gamma (IFN-γ). Besides adhesion to collagen, α1 mediates several other eff ects, depending on cell type. On endothelial cells, vascular endothelial growth factor (VEGF) strongly upregulates α1, which is a prerequisite for the formation of new blood vessels. Inhibition of the integrin/collagen binding by specifi c antibodies reduced the formation of new blood vessels but did not alter pre-existing vasculature [23]. As angiogenesis is important for the progression of arthritis, its inhibition could be an attractive target in the therapy of RA. Engagement of integrin α1 on macrophages and T cells stimulates migra tion and cytokine production [24]. Th e role of α1 on synovial fi broblasts has not been elucidated yet, but studies from dermal fi broblasts suggest that α1 controls collagen synthesis, migration, and proliferation on colla gen substrates. In a murine anti-collagen II antibody-induced arthritis model, inhibition of integrin α1 reduced cartilage degradation and leukocyte infi ltration, demon strating an important role for this integrin in infl amma tion [25].
Integrin α2β1 has functions similar to those of α1β1 (for example, VEGF-induced proliferation of endothelial cells) but is much weaker in expression in the synovial tissue. Ligation of α2β1 by collagen type I or II, but not collagen type IV, augmented IFN-γ production in T cells. In contrast to CD4 + and CD8 + T cells, which upregulate α1β1 after activation, Th 17 cells upregulate α2β1, which is thought to be a co-stimulatory molecule necessary for IL-17 production [26]. In a model of experimental autoimmune encephalomyelitis, an antibody against α2 suppressed clinical signs and central nervous system infl ammation because of reduced extravasation of immune cells. Since leukocyte infi ltrates also fuel synovial infl ammation, one would expect that α2 inhibition might also be benefi cial in the joint.

Laminin-binding integrins α3β1 and α6β1
Laminins are a major component of the basal lamina and usually are secreted by epithelial cells, endothelium, and organogenic fi broblasts. Laminin is a trimeric molecule consisting of an α-chain, a β-chain, and a γ-chain that can combine to form 15 diff erent laminin isoforms [27]. Several integrin hetero dimers recognize laminin as a ligand, but the affi nity for each laminin subtype diff ers. Laminin binds to integrins α3β1, α6β1, α7β1, and α6β4, but only the fi rst two are present in the synovial tissue. In the synovial tissue, α6β1 is strongly expressed in the synovial lining layer by fi broblasts, whereas α3β1 is also expressed by lymphocytes, macrophages, and endothelial cells [18]. Not only the number of laminin receptors but also the production of laminin itself is upregulated in the synovial lining. Besides endothelial cells, SFs produce and secrete laminin, and its deposition at the synovial lining is increased in RA [28]. Along with collagen and fi bronectin, laminin promotes adhesion and proliferation of synovial cells and lymphocytes. Furthermore, integrin binding to laminin increases the expression of MMP-3 and MMP-10, which is further augmented when transforming growth factor-beta (TGF-β) is present [29]. Although no studies have focused on the contribution of laminin and its respective integrin receptors on the progres sion of RA, results from tumor research indicate that upregulation of α6β1 or α3β1, at least in most cases, increases cell invasion and metastasis. As a consequence, increased laminin deposition and increased integrin expression on synovial lining cells increase MMP secretion by these cells, resulting in greater joint damage and invasion into cartilage.

Fibronectin-binding integrins α4β1, α5β1, aνβ3, and αvβ5
Fibronectin is a large glycoprotein consisting of two mono mers linked by disulfi de bonds. Owing to alternative splicing, fi bronectin exists in 20 diff erent isoforms. Th e molecule is folded into globular domains, each of which has a diff erent function. In these domains, repeating amino acid sequences form the so-called type I, II, and III repeats. Th ese sequences interact with other proteins; for example, type I repeats bind to fi brin, and regions in type III repeats contain the RGD motive that binds integrins. Fibronectin can be divided, on the basis of solubility, into two classes: soluble plasma fi bronectin and insoluble cellular fi bronectin, the latter of which is organized into fi brils as a scaff old for the ECM [30]. Fibronectin levels are elevated in cartilage and synovial fl uid of patients with osteo arthritis, whereas in RA, fi bronectin is highly expressed at the synovial lining and invading pannus [31].
About 12 integrin heterodimers bind to fi bronectin but only some combinations are found in the synovial tissue. Th e classical fi bronectin receptor α5β1 recognizes only fi bronectin, whereas ανβ3 and ανβ5 also bind to other ECM molecules or cellular proteins. Th e α5 integrin is expressed by all cells in the synovial tissue, and α5 levels are increased in the synovial lining layer [5]. Lymphocytes have been shown to attach to other synovial cells by binding to the fi bronectin coating on their cell surface by α4β1 and α5β1 and to some extent by ανβ3 integrins [32]. Adhesion to fi bronectin increases proliferation and survival of chondrocytes and protects synovial cells from Fas-induced apoptosis [33]. In addition, Zeisel and colleagues [34] found an increase of MMP-3 production by SFs after stimulation with a bacterial ligand of α5β1, suggesting that enhanced integrin ligation increases joint destruction in RA. Matrix-degrading enzymes in synovial fi broblasts are induced by small integrin-binding peptides derived from the degradation of fi bronectin and other ECM molecules. Werb and colleagues [7] demonstrated that only fi bronectin-derived fragments, but not intact fi bronectin, increased the expression of MMP-1 and MMP-3. Besides having a role in cell survival and prolifera tion, α5β1 also regulates cytokine and growth factor production. In SFs, ligation of α5β1 increased synthesis of B-lymphocyte activating factor (BAFF) [35]. Increased BAFF synthesis not only promotes the proliferation of B cells and immunglobulin classswitching but possibly acts as an autocrine mechanism of SFs to stimulate their own NF-κB activity. Besides BAFF, direct ligation of α5β1 induces the expression of NF-κBcontrolled genes in fi broblasts and endothelial cells. Most of these genes -like IL-8 and heparin-bound epidermal growth factor (HB-EGF) -or endothelin-1 control angiogenesis or infl ammation and their enhanced expression aggravate RA.
Naïve and memory T cells depend on binding of a4β1 and α5β1 to fi bronectin to promote proliferation [32]. In this context, SFs assist in T-cell activation, as T cells bind to the fi bronectin coating on their surface and α4β1 transmits co-stimulatory signals. In RA, infi ltrated T cells in the synovial tissue display higher surface levels of α4β1 compared with peripheral T cells and localize to sites of tissue injury by binding to proteolytically degraded cartilage fragments. Th e most important function of α4β1 on lymphocytes is adhesion to endothelium and subsequent transmigration to sites of injury. For most integrins, expression level data in the literature are given as 'high' or 'strong' . This makes it hard to estimate integrin levels on diff erent cell populations. We consulted several publications and, albeit subjectively, categorized integrin expression levels on a range from strong to weak expression (indicated by one, two, or three plus signs). ?, There is no available data in the literature about the expression of respective integrins.
Unfor tu nately, extravasation of lymphocytes into the synovial tissue is also the mechanism that further supports the infl am matory process. In RA, endothelium expresses high levels of VCAM-1, which acts as a binding partner for α4β1. After chemokine-induced activation and clus ter ing of α4β1, lymphocytes are immobilized on endo thelium and transmigrate to sites of infl ammation [36]. Th e important role for integrin α4β1 in RA is further underlined by a study by Raychaudhuri and colleagues [37], who show that antagonism of this receptor prevents infl ammation and MMP production in a murine model of accelerated collagen-induced arthritis. Heterodimers containing αν are another family of fi bronectin-binding integrins. A polymorphism in the gene coding for integrin αν is associated with RA in the European Caucasian population [38]. Th e integrin ανβ3 binds fi bronectin with high affi nity and acts in concert with α5β1 to arrange fi bronectin matrices [39]. Besides binding fi bronectin, ανβ3 binds vitronectin, osteopontin, and bone sialoprotein with high affi nity. Th is integrin heterodimer is expressed on SFs, endothelial cells, and osteoclasts [40]. Not much is known about the specifi c role of ανβ3 on SFs, but studies from other cell systems and from tumor research indicate that this integrin promotes tissue invasion and MMP-2 production. Th e major function of ανβ3 (and also of ανβ5) together with vascular endothelial growth factor receptor 2 (VEGF2R) is angiogenesis promotion [41]. Integrin ανβ3 does so by phosphorylating intracellular residues of the VEGR2R, thereby enhancing signaling of this receptor. Whereas the expression of ανβ3 is low on resting endothelium, levels increase with infl ammation and tumor angiogenesis. In RA, the pro-infl ammatory milieu promotes angiogenesis, and its inhi bi tion by a small-molecule αν antagonist was effi cacious in a rabbit model of arthritis. Whether this is applicable to humans remains to be seen as the development of vitaxin, a monoclonal antibody against ανβ3, was dis continued in 2004 because the drug lacked effi cacy in RA. It could well be that vitaxin is only minimally eff ective by itself but potentiates the eff ect of other therapies. Th is phenomenon is evident with cilengitide, a small-peptide αν antagonist that is being clinically tested against a variety of cancers. Th is peptide not only inhibits angiogenesis but also increases the permeability of endothelium. Th is could be exploited to increase the delivery of other drugs [42].
Another cell type that responds to αν antagonists is the osteoclast. In RA, osteoclast activity is enhanced because of increased RANKL (receptor activator of NF-κB ligand) expression, and osteoclasts are res pon sible for bone erosions in RA [43]. Th e integrin ανβ3 increases the bone resorptive capacity of osteoclasts by activating several intracellular signaling pathways such as FAK and c-Src. Blockage of this integrin was protective in vitro.

The β2 integrins (CD18)
In contrast to all of the abovementioned integrin heterodimers, β2 integrins bind other adhesion receptors and are expressed exclusively on immune cells. Th e focus of this section is the αLβ2 integrin (lymphocyte functionassociated antigen 1, or LFA-1). Th is is the major integrin of leukocytes and is important for tissue extravasation, chemotaxis, and formation of immunological synapses. Th e ligand for LFA-1 is intracellular adhesion molecule (ICAM), which is expressed in three diff erent isoforms. ICAM-1 is expressed in the synovial tissue by endothelial cells and fi broblasts, and cell surface levels of this glycoprotein are increased by the action of cytokines like IL-1β. As the pro-infl ammatory milieu in RA also leads to the upregulation of αLβ2 on leukocytes, the net eff ect is an enhanced infl ux of immune cells in the synovial tissue, contributing signifi cantly to infl ammation. Increased cell-cell contacts between synovial fi broblasts and T lymphocytes lead to the activation of fi broblasts and result in an increased cytokine production by both T-cells and fi broblasts. Th is eff ect was dependent on T-cell adhesion on fi broblasts as T-lymphocyte cell culture supernatants failed to elicit appropriate responses [44]. Furthermore, αLβ2 is part of the immunological synapse between antigen-presenting cells and lymphocytes, and disruption of this interaction leads to immune impairments. Owing to its importance, LFA-1 is a suitable target for therapeutic intervention, and inhibition of this integrin by a mononclonal antibody was effi ca cious in a rabbit model of RA [45]. Th e positive eff ect of LFA-1 inhibition was also shown in mice treated with a small-molecule antagonist against this integrin. Animals showed reduced adhesion of T-cells to endothelial cells, reduced T-cell proliferation, and decreased Th 1 cytokine production [46]. In humans, efalizumab, a monoclonal antibody against the αL subunit, was withdrawn from the market as serious side eff ects occurred because of the immunosuppressive action of the drug [47]. A more feasible approach to dampen the activity of lymphocytes in RA is the inhibition of ICAM-1. Mice defi cient in ICAM-1 showed only minor disease activity in a collagen-induced arthritis model, and clinical studies in humans also demonstrated benefi cial eff ects when ICAM-1 was blocked by an anti-ICAM-1 monoclonal antibody in early RA [48,49]. Unfortunately, this antibody could not be administered repeatedly, as its mouse origin makes this antibody immunogenic, resulting in side eff ects such as fever and leukopenia [50].

Conclusions
During the course of RA, immune cells are recruited into the synovial tissue, where they produce large amounts of pro-infl ammatory cytokines and interact with residential fi broblasts and macrophages. Th e rate of migration of immune cells into the synovium is controlled by the expression of ICAM-1 on endothelial cells and integrin αLβ2 on immune cells. Th e interaction of T lymphocytes (Th 1 type) with synovial fi broblasts, endothelial cells, and macrophages activates those cells and they start to produce ECM proteins, cytokines, adhesion molecules, and matrix-degrading enzymes. Th e pro-infl ammatory milieu also increases the expression of most integrin subunits. Th e combination of increased integrin expression, increased ECM deposition, and degradation along with high cytokine and growth factor levels in the synovial tissue make a perfect microenvironment for the perpetuation of RA. Th e enhanced integrin signaling in RA is probably one cause of the high basal secretion of cytokines from fi broblasts and macrophages. Furthermore, increased integrin ligation enhances growth factor and cytokine signaling and induces the expression of MMPs. MMPs degrade ECM molecules, thereby generating small RGD peptides, which further activate integrins. When all the facts are taken together, it can be said that RA is, at least in part, an integrin-driven disease that can be modulated with specifi c integrin antagonists.