Impaired dermal wound healing in discoidin domain receptor 2-deficient mice associated with defective extracellular matrix remodeling

Collagen receptors mediate fibroblast responses during tissue regeneration and healing. Integrins are the most studied collagen receptors (for review, see [1]). Integrins signaling results in fibroblasts recruitment to the wounded area and synthesis of extracellular matrix (ECM) components to conform the healing tissue. Integrins also function as mechanotransducers of the tensile strength exerted by the healing matrix [2, 3] to further activate downstream signaling that drives fibroblast contraction of the wound [4–6]. In addition to integrins, the discoidin domain receptor (DDR) family of receptors (DDR1 and DDR2) also interact with the collagenous ECM. As opposed to integrins, DDRs are tyrosine kinase receptors, and thus become phosphorylated in response to collagen [7]. The molecular details of collagen recognition by DDRs are starting to become known [8]. DDR1 preferentially binds to collagen type I. DDR2 has higher specificity for fibrillar type I collagen than DDR1, but also for collagen type II through a specific DDR2 recognition site in the D2 period of collagen II [9]. DDR overexpression is associated with fibrotic diseases of the lung, kidney and liver [10], atherosclerosis, osteoarthritis [11], as well as several tumors of epithelial origin [12, 13]. DDR2 mRNA is also upregulated in dermal burn wounds [14]. Mice deficient in DDR2 present a dwarfed phenotype with reduced proliferative response of experimentally wounded skin compared to wild-type littermates [15].

Collagen receptor signaling results in matrix metalloproteinase (MMP) release [16], and dysregulation of MMP activity is a key feature of defective wounding response. MMP2 is the predominant protease in dermal wound healing [17] and MMP2 activity is reduced in DDR2^-/- skin and cultured DDR2^-/- skin fibroblasts, with lower MMP2-dependent cell proliferation and chemotactic invasion [18].

The role of collagen receptor signaling in response to tissue stiffness is becoming an area of intensive research [4, 19]. Stiffness of the healing tissue depends on fibrillar collagen formation and its covalent crosslinking. Deficient crosslinking machinery results in hyperelastic tissue that scars easily, and heals slowly and poorly [20], whereas excessive collagen crosslinking is a hallmark of tissue fibrosis [21]. Fibroblasts synthesize ECM components such as the matricellular glycoprotein 'secreted protein, acidic and rich in cysteine' (SPARC; also known as osteonectin) that modulates collagen fibril growth and deposition [22], and synthesize two of the enzymes that mainly mediate collagen crosslinking: lysyl oxidase (LOX) and lysyl hydroxylases (LH) [23].

In the present study, we investigated the ability of DDR2^-/- mouse skin to heal. The skin wounds of DDR2^-/- mice heal more slowly than DDR2^+/+ mice, and display diminished tensile strength, reduced numbers of recruited α smooth muscle actin (αSMA)-expressing fibroblasts and decreased MMP2 and collagen levels, with reduced crosslinking and modified mRNA expression of several proteoglycans, collagens, integrins and MMPs. These findings are explained by reduced expression of the crosslinking molecules LOX, LH1 and SPARC in dermal DDR2^-/- wounds and altered Src-dependent fibroblast contractile ability. The data implicate DDR2 in modulation of collagen processing and tensile wound strength during tissue healing.

DDR2^-/- mice and DDR2^+/+ littermates were subjected to dermal incisions, and allowed to heal for up to 12 days. Whereas hematoxylin and eosin (H&E) staining of DDR2^-/- dermal scar sections 4.5 days post wounding were comparable to those of DDR2^+/+ scars [15], closure of DDR2^-/- wounds was significantly delayed (Figure 1a), with maximal differences between DDR2^-/- and DDR2^+/+ mice at day 4. The wounds of DDR2^-/- mice remained significantly larger than DDR2^+/+ mice for the following 7 days.

The rate of tissue healing depends largely on efficient recruitment of fibroblasts to the wound. Delayed DDR2^-/- mouse wound healing might be ascribed to delayed fibroblast recruitment to the granulation tissue. Western blot analysis of DDR2^-/- skin extracts 7 days post wounding revealed an average of 45% reduction in expression of αSMA, a marker of contractile wound myofibroblasts (Figure 1b). Using gel zymography, we previously detected a 60% reduction in MMP2 activity in unwounded DDR2^-/- mouse skin extracts [18]. Gelatin zymography analysis of skin extracts 7 days post wounding revealed two bands of gelatin degradation at 68 kDa and 62 kDa, representing latent and active MMP2, respectively (Figure 1c). MMP2 activity was reduced by an average of 75% in the DDR2^-/- wound extracts compared to DDR2^+/+ ones. These data are in accordance with our previous in vitro observations demonstrating reduced MMP2-dependent chemotactic migration of isolated αSMA-expressing DDR2^-/- skin fibroblasts [18].

We next evaluated the tensile strength of skin wound closure. At 7 days after wounding, the maximal load before rupture in DDR2^-/- wounds was an average 65% less than in DDR2^+/+ wounds. The differences remained significant at day 10, where maximal load before rupture was 52% less in DDR2^-/- mice (Figure 2a).

Collagen is a key regulator of tissue tensile strength. Using a colorimetric assay, we observed that the collagen concentration in DDR2^-/- wounds was an average of 25% less than in DDR2^+/+ wounds 7 days after wounding, and an average of 40% less after 10 days (Figure 2b). These data correlate with our previous description of reduced collagen type I mRNA expression in cultured DDR2^-/- skin fibroblasts [18]. There were no significant differences in collagen content between unwounded DDR2^-/- and DDR2^+/+ mouse skin (data not shown).

Collagen fibrils must be crosslinked for proper alignment and function. Reverse high-performance liquid chromatography (HPLC) analysis of pyridoline allowed us to measure trifunctional collagen crosslinks in the wounded skin extracts. At 7 days after incision, total pyridoline content (lysylpyridinoline + hydroxylysylpyridinoline) was 60% of that in the wild-type samples and levels of total pyridoline remained significantly lower 10 days after wounding (Figure 3a). Collagen crosslinking is mediated by LOX and LH. Collagen crosslink formation is initially catalyzed by LOX. There are two subsequent crosslinking pathways, an allysine and a hydroxyallysine route [24]. Pyridinolines, lysylpyridinolines (LPs) and hydroxylysylpyridinolines (HPs), are trifunctional crosslinks, derived from the hydroxyallysine crosslink route. Finally, LH catalyzes the conversion of lysine into a hydroxylysine in the collagen molecule. Further collagen fibril growth and deposition is mediated by SPARC [22]. Quantitative real-time reverse transcription polymerase chain reaction (RT-PCR) experiments demonstrated reduced mRNA expression of LOX, LH1 and SPARC in DDR2^-/- skin tissue, with an average reduction of mRNA expression of 75%, 81% and 45%, respectively, compared to wild-type tissue (Figure 3b).

Next, we investigated if defective collagen crosslinking in DDR2^-/- skin correlates with reduced expression of collagen crosslinking molecules in DDR2^-/- fibroblasts. As shown in Figure 4, DDR2^-/- fibroblasts express significantly less LOX, LH1 and SPARC mRNAs (Figure 4a), and reduced LOX, LH1 and SPARC proteins (Figure 4b). We also analyzed DDR2-/- skin fibroblasts in which wild type DDR2 expression was reconstituted by retroviral infection (WTR2). We utilized WTR2 fibroblasts that expressed roughly similar levels of DDR2 wild-type cells (data not shown)[18]. LOX, LH1 and SPARC mRNA expression levels were restored in WTR2 cells and remained significantly higher than in the DDR2^-/- cells (Figure 4a, dashed columns). As shown in Figure 4b, LOX, LH1 and SPARC protein expression in WTR2 cells was similar to that of DDR2^+/+ cells. DDR2^-/- skin fibroblasts retrovirally infected with control vector (internal ribosomal entry site and green fluorescent protein (GFP) cDNA) behaved as non-infected cells (data not shown).

Wound closure reflects contraction of the newly formed matrix by myofibroblasts. To assess the importance of DDR2 in cell contraction, skin fibroblasts were assayed for their capacity to reduce a collagen gel in response to a contractile stimulus (keratinocyte culture supernatants) and in the presence of DDR2 ligand (soluble collagen type I) (Figure 5a). DDR2^-/- skin fibroblasts activated by keratinocyte supernatants contracted collagen gels 40% less than DDR2^+/+ fibroblasts.

Defective matrix metalloproteinase expression is associated with in vitro fibroblast contraction [26] and in vivo impaired wound contraction in murine wound models [27, 28]. Using the specific MMP inhibitor N-[(2R)-2-(hydroxamido carbonylmethyl)-4-methylpentanoyl]-l-tryptophan methylamide (GM6001), we previously reported that DDR2-dependent migration of skin fibroblast is also MMP dependent [18]. Herein, pretreatment of DDR2^+/+ and DDR2^-/- skin fibroblasts with 1 μM GM6001 completely blocked DDR2^-/- skin fibroblast contraction, but retained DDR2^+/+ skin's ability to contract 30% of the initial collagen gel area.

LOX activity also promotes fibroblast contraction [29]. Interestingly, under experimental conditions, pretreatment with the LOX inhibitor N-bromoacetylpyridoxamine (BAPM) allowed DDR2^+/+ skin fibroblasts to contract 20% of the initial collagen gel area, DDR2^-/- skin fibroblast contraction was completely blocked, indicating reduced LOX activity dependent fibroblast contraction in the DDR2^-/- cells. Under the assayed conditions, neither GM6001 nor β-aminoproprionitrile (BAPN) affected fibroblast survival, as demonstrated by trypan blue exclusion test (data not shown).

Fibroblast contraction is mediated by focal adhesion kinase (FAK)/Src signaling [30]. Skin fibroblasts were recovered from contracted collagen gels and analyzed by western blot (Figure 5b). While there were no significant differences in total Src, DDR2^-/- fibroblasts expressed an average of 30% less phosphorylated Src^Y418 than DDR2^+/+ ones, which may correlate with their reduced contraction ability.

Fibroblasts are major effectors of wound repair and regenerative healing. They coordinate the deposition and further organization of the collagenous extracellular matrix in the wound bed, and thereby determine the final outcome of the scarring process. We have previously shown that fibroblasts from DDR2 knockout mice are defective in key aspects of cell proliferation and ECM remodeling [18]. In this work, we further explore DDR2^-/- skin fibroblasts in the context of experimental cutaneous wound healing. Furthermore, we identify new molecules tightly related to ECM assembly whose expression is blunted in fibroblasts from wounded DDR2^-/- skin. Because reconstitution of DDR2 expression by retroviral infection of DDR2^-/- skin fibroblasts resulted in almost full recovery of their mRNA levels to that of DDR2^+/+ cells, we conclude that the observed effects were attributable to DDR2.

While there is considerable information about DDR1 signaling [31, 32], DDR2 signaling is poorly characterized [33]. We have previously shown that Src kinase and Shc recruitment are key signaling intermediates in DDR2 signal transduction following collagen type I induction of tyrosine 740 phosphorylation [34]. Src phosphorylation at tyrosine 418 is required for maximal DDR2 phosphorylation [35], resulting in increased MMP2 promoter activity. Furthermore, DDR2 is involved in the regulation of FAK levels in vascular smooth muscle cells adherent to type I collagen matrices [36].

DDR2 deficiency in skin fibroblasts correlates with reduced mRNA levels of collagen type I and MMP2 [18]. In this work, we compared mRNA and protein expression between DDR2^+/+ and DDR2^-/- skin fibroblasts and describe a cluster of ECM-related genes regulated by DDR2 signaling, including collagen type VI, collagen type VIII, syndecan 1, SPARC, LOX, LH1 and MMP10.

DDR2^-/- skin wounds show delayed closure and impaired recovery of tensile strength, which may be ascribed to a defective ability of DDR2^-/- fibroblasts to synthesize the ECM in the newly formed healing tissue. The observed reduced content of fibrillar collagen is the result of downregulation of collagen expression, alteration in matrix remodeling through variation in MMP composition, and reduced collagen crosslinking by LOX, LH1 and SPARC. Interestingly, the extracellular domain (ECD) of DDR2, when used as a purified, soluble protein, inhibits collagen fibrillogenesis in vitro[37].

Our work demonstrates the delayed closure of DDR2^-/- dermal wounds, and impaired contraction on collagen type I gels by DDR2^-/- skin fibroblasts. Fibroblast contraction is a key step in dermal wound healing, where mechanical signals derived from the extracellular matrix induce mechanoreceptor signaling and activation of the contractile machinery. While integrins were the first mechanoreceptors described, increasing evidence demonstrates a role for receptor tyrosine kinases (RTKs) in cell motility that may relay on interactions between integrins and RTK signaling pathways, and activation of the FAK/Src signaling pathway [38–40]. While no evidence exists for integrin-DDR interaction at the plasma membrane, convergence of their downstream signaling may occur. In this regard, Src phosphorylation is a downstream event following both integrin α2β1 and DDR2 mediated cell contraction, as DDR2^-/- fibroblasts contain less phosphorylated Src^Y418 than wild-type cells. Interestingly, collagen stimulation of DDR1 mediates migration of smooth muscle cells through Src [41].

As described in retinal pigment epithelium (RPE) cells [31], skin fibroblasts require the FAK/Src complex for collagen gel contraction. DDR2 is involved in the regulation of FAK levels [36] but its activation does not modify cellular adhesion [18]. Furthermore, only anti-integrin α2 antibody-treated RPE cells showed a statistically significant, albeit modest decrease in adhesion as compared to wild-type cells.

DDR2^-/- skin fibroblasts express higher mRNA levels of other receptors for fibrillar collagen such as integrin α2. However, mRNA expression of DDR1, the other member of the DDR family is not significantly modified in skin fibroblasts. Interestingly, α2 and DDR1 expression are increased in DDR2^-/- stellate cells from carbon-tetrachloride intoxicated livers (E. Olaso, unpublished results). Thus, upregulation of α2 and/or DDR1 may compensate for DDR2 deficiency during cell adhesion.

The motogenic phenotype of skin fibroblasts is regulated by MMP-mediated remodeling of its associated ECM, and the MMP inhibitor GM6001, specifically impairs wound contraction [26]. GM6001 also inhibits DDR2-mediated invasion into basement membrane matrix by hepatic stellate cells [10]. In this work we demonstrate impaired contraction of collagen gels in skin fibroblasts treated with GM6001.

DDR2^-/- skin fibroblasts express reduced levels of LOX and deficient cell contraction, and blockade of LOX activity by BAPN results in reduced skin fibroblast contraction. A growing body of evidence shows that LOX plays a key role in intracellular signaling underlying cell motogenicity. LOX activity promotes focal adhesions through FAK/Src signaling [13] and crosslinked collagen induces integrin clustering that enhances PI3K signaling, and induces invasion of premalignant epithelium. Ongoing studies on the interaction of LOX and DDR2 signaling in the tumor microenvironment may further explain the relationship between collagen dysregulation and cancer invasion and metastasis.

Based on our findings to date we propose multiple interrelated features are altered by DDR2 signaling in skin fibroblasts, which may affect skin wound healing and tensile strength: (1) chemotactic migration to the wounded area and proliferation, (2) synthesis and remodeling of the wound matrix, (3) three-dimensional organization of the wound matrix by collagen cross linking and collagen fibrillogenesis, and (4) fibroblast contraction of the healing wound.