Sensing extracellular matrix: An update on discoidin domain receptor function

Abstract

Discoidin Domain Receptors (DDRs) have recently emerged as non-integrin-type receptors for collagen. The two mammalian gene products Discoidin Domain Receptor 1 and -2 constitute a subfamily of tyrosine kinase receptors that are selectively expressed in a number of different cell types and organs. Upon collagen activation, DDRs regulate cell adhesion, proliferation and extracellular matrix remodeling. Here we review the various signaling pathways and cellular responses evoked by activated DDRs. Additionally, we give an overview of the more recent advances in understanding the role of DDRs in various human diseases, in particular during tumor progression, atherosclerosis, inflammation and tissue fibrosis. Furthermore, we discuss potential roles of genes homologous to mammalian DDRs identified in flies, worms and sponges. We show that the structural organization of these DDR-related genes is highly conserved throughout evolution suggesting that invertebrate DDRs may also function as receptors for collagen. By highlighting current questions about these unusual collagen receptors, we hope to attract new research on DDRs from a variety of different fields.

Introduction

Living cells must integrate a myriad of extracellular stimuli into highly cohesive responses. To manage this wealth of information, a diverse array of specialized cell surface receptors exists that binds extrinsic factors such as mitogens, differentiation factors, cell membrane-bound molecules or extracellular matrix (ECM) proteins, and then transmit signals through the plasma membrane. Many of these receptors belong to the family of receptor tyrosine kinases (RTKs) characterized by an extracellular ligand binding domain, a single transmembrane domain and a catalytic tyrosine kinase domain. RTKs have been grouped into 18 subfamilies according to the domain structure of their extracellular region, which defines ligand specificity [1]. This review focuses on one subfamily of RTKs, the Discoidin Domain Receptors (DDRs). Two members of this subfamily are present in the human genome, DDR1 and DDR2. DDRs are unique due to their ligand-specificity and remarkable conservation throughout evolution.

Unlike most other RTKs, DDRs are not activated by soluble growth factors. Instead, various types of collagen act as ligands for DDRs. DDR1 is activated by all collagens tested so far, including collagens type I to type VI and type VIII, while DDR2 is only activated by fibrillar collagens, in particular collagens type I and type III [2], [3]. DDRs are activated only when collagen is in its native, triple-helical form, as heat-denatured collagen (gelatin), which lacks triple-helical structure, fails to induce kinase activity. While most other RTKs are fully activated in minutes, maximal activation of DDRs occurs several hours after the initial stimulation with collagen [2]. Some attempts have been made to further define the molecular interaction between collagen and DDRs, but the precise location of the DDR-binding site within triple-helical collagen is yet unknown. Recent work however suggested the second quarter of type II collagen has been as a possible binding site for DDR2 [4], [5].

Four integrin receptors, formed between the β1 subunit and the α1, α2, α10 or α11 subunit, also act as functional collagen receptors, but do not require DDRs as co-receptors [6]. Conversely, binding of collagen to integrins results in non-DDR-dependent tyrosine phosphorylation events, which are mainly driven by integrin-associated kinases of the Src- and Fak-family. An important and well-described outcome of integrin activation is the alteration in cytoskeletal tension and cell migration, which is mediated by the actomyosin network. In contrast to the integrins, a potential role of DDRs in transmitting these kinds of mechanical stimuli within or between cells has not been explored.

Structurally, DDRs are distinguished from other RTKs by a discoidin domain, an approximately 160 amino acid long homology region first identified in the protein discoidin I from the slime mold Dictyostelium discoideum, where it functions as galactose-binding lectin [7], [8], [9]. Aside from DDRs, the discoidin I-homology repeat is also present in more than a dozen other mammalian transmembrane as well as secreted proteins. Utilizing the crystal structures of the discoidin domains found in the coagulation factors V and VIII, molecular models of the domains in DDR1 and DDR2 were generated [10], [11]. Hallmark of these models is a central eight-stranded beta-barrel, which is stabilized by two intramolecular disulfide bridges, and four finger-like loops protruding from one side of the beta-barrel. The position of these loops is well conserved between discoidin domains of DDRs, blood clotting factors V and VIII or neuropilin [12]. Work with a recombinant preparation of the DDR1 discoidin domain led to the identification of loops 1 and 3 being essential for collagen binding and receptor activation [11]. However, one will have to await a detailed structural analysis of DDRs to draw more definitive conclusions on the architecture of the ligand-binding pocket.

In several cell lines and tissues, DDR1 is partially processed into a 62 kDa membrane-anchored beta-subunit and a 54 kDa soluble extracellular domain-containing alpha-subunit [13]. This process, also termed shedding, is significantly enhanced upon DDR1 activation. Proteases belonging to the ADAM or MT-MMP family could potentially be responsible for DDR1 shedding, since they have been pinpointed as sheddases for a number of other receptors involved in cell-adhesion, including Eph receptors, selectins and the heparin-binding epidermal growth factor [14], [15]. In near future, more experimental work will hopefully “shed” better light on the mechanism of DDR1 processing.

Compared to other RTKs, the juxtamembrane regions of DDR1 and DDR2 are much longer (176 and 147 amino acids, respectively). As observed for members of the platelet derived growth factor receptor or Eph receptor subfamilies, we speculate that the juxtamembrane region of DDRs also has an auto-inhibitory function [16]. For Eph receptors, it was found that sequences within the juxtamembrane region block the ATP binding site in the kinase domain and, upon ligand binding, need to be displaced prior to activation of the catalytic function. Potentially, the protracted kinetics of DDR activation are the result of a similar rate-limiting structural re-arrangement in the juxtamembrane region that is necessary to overcome an intrinsic auto-inhibition.

Thus far, five isoforms of DDR1 have been identified, all of which are generated by alternative splicing in the cytoplasmic region [17]. The longest DDR1 transcript encodes the c-isoform with 919 amino acids. The a- and b-receptor isoforms lack 37 or 6 amino acids in the juxtamembrane or kinase domain respectively [9]. DDR1d and DDR1e are truncated variants that lack either the entire kinase region or parts of the juxtamembrane region and the ATP binding site [17]. In contrast, no isoforms have been identified for DDR2 yet.

The relative expression ratios and the post-translational modifications of the DDR1 a- and b-isoforms appear to be controlled by complex regulatory mechanisms. The DDR1b protein is the predominant isoform expressed during embryogenesis, whereas the a-isoform is commonly found in several human mammary carcinoma cell lines [18]. Furthermore, DDR1a is the predominant isoform during rat neuronal development, but DDR1b is induced following irradiation of astrocytes [19]. Importantly, the alternatively spliced insert in DDR1b contains the motif LLXNPXY that associates with the phosphotyrosine-binding domain of the ShcA adapter protein upon collagen-induced tyrosine phosphorylation [2]. In contrast, the juxtamembrane region of the a-isoform binds to fibroblast-growth factor receptor substrate-2 and appears to be unique in its ability to trigger the migration and pseudopod extension of leukocytes [20], [21]. Most likely, the selective availability of juxtamembrane binding sites is critical for differential downstream responses during DDR1 signaling.

In a three-dimensional (3D) tissue culture model with type I collagen gels, epithelial cells overexpressing DDR1a or -b grew slower and formed fewer branches than the parental cell line [22]. In contrast, a truncated DDR1 that resembles the DDR1d isoform had the opposite effects. Possibly, high levels of full length DDR1 can restrict cell migration and proliferation leading to a transient cell cycle arrest. This notion is supported by recent findings in melanoma cells, which undergo DDR2-mediated transient cell cycle arrest when grown within 3D collagen [23]. We conclude that further work is required to fully understand this putative dichotomy of DDR isoform signaling: promoting cell growth in monolayer culture, versus inhibiting it in a 3D, collagen-rich environment. One possibility is that cells in 3D not only sense the lack of tension provided by a rigid 2D substratum, but also are more receptive to diverse cues from the ECM, such as the formation of neo-epitopes during matrix remodeling. Conceivably, DDR-induced signals are converted into different cellular outcomes depending on the nature of the stimulus by collagens, which could either be acid-solubilized, monomeric collagen coated onto tissue culture plastic versus a thick layer of neutralized, and fiber-associated collagen.

DDR1 and DDR2 have a total of 15 and 13 tyrosine residues in their cytoplasmic regions respectively, which serve as potential phosphorylation sites upon receptor activation by collagen. Using phosphopeptide mapping, 3 major and 5 minor phosphorylation sites were recently identified in DDR1 [24]. Sustained phosphorylation of these tyrosines will potentially allow binding of a number of different Src-homology-2 (SH2)- and phosphotyrosine binding (PTB) domain-containing molecules (Fig. 1). Previously, we showed that the alternatively spliced tyrosine-513 of DDR1b directly associates with the PTB domain of ShcA upon receptor activation [2]. Moreover, Shc was also found tyrosine phosphorylated in several cell lines and tissues, such as monocytes or bronchial lavage cells, presumably by activated DDR1 [25], [26]. In macrophages, DDR1 induced ShcA phosphorylation led to activation of the TRAF6 complex, which triggers the p38 mitogen-activated protein kinase and NFκB pathways [26], [27], [28]. However, Shc-mediated downstream events may be highly cell-type dependent, since activation of DDR1b in human breast cancer T-47D cells results in strong ShcA binding to the receptor but not in protein phosphorylation [2]. Several other molecules were found to directly interact with DDR1, such as Shp-2, an SH2 domain-containing phosphotyrosine phosphatase, and Nck2, an SH2- and SH3 domain containing adapter protein [24], [29]. The binding site for Shp-2 was mapped to tyrosine-740 of DDR1 while binding of the p85 subunit of phosphatidyl-inositol-3 kinase is mediated by tyrosine-881 [30]. To summarize these interactions an overview of the DDR1 and DDR2 signaling events is given in Fig. 1.

In a recent approach, Stat5 was found to be a tissue-specific molecule downstream of activated DDR1 [31]. Specifically, mammary epithelial cells that express DDR1 respond with extended Stat5 activation upon prolactin stimulation compared to control cells. These results suggest that DDR1 plays a pivotal role in maintaining lactation, which is also supported by data from knockout mice (see below). In human breast cancer cells, it was suggested that DDR1 receives lateral input from other transmembrane receptor/ligand complexes, such as Frizzled and Wnt5a, however the integration of these signals within a cellular context requires further investigation [32], [33], [34].

In contrast to DDR1, signaling of DDR2 has mainly been studied in hepatic stellate cells and skin fibroblasts. Intriguingly, it was found that full activation of DDR2 by fibrillar collagen requires the presence of ShcA and Src-like tyrosine kinases [35]. Src appears to be an obligate DDR2 partner to allow full autophosphorylation. Like DDR1, ShcA binds to the juxtamembrane region of DDR2 (tyrosine-471), but unlike DDR1, the SH2 domain rather than the PTB domain of ShcA is mediating this interaction. Furthermore, Src-mediated phosphorylation of some DDR2 sites, including tyrosine-740, appears to be auto-inhibitory, at least in a direct binding experiment with purified proteins [36].

Little is yet known about specific transcriptional targets of active DDRs. Using a microarray specifically designed towards matrix genes and their modifiers, it was found that DDR signaling enhances the expression of P-selectin glycoprotein ligand and represses the levels of matrix proteins such as agrin or syndecan-1 [37]. However, large scale arrays and functional protein assays are necessary to connect the signaling routes described above with any specific transcriptional response.