Research ArticleRegenerative Medicine

Engineering the Growth Factor Microenvironment with Fibronectin Domains to Promote Wound and Bone Tissue Healing

See allHide authors and affiliations

Science Translational Medicine  14 Sep 2011:
Vol. 3, Issue 100, pp. 100ra89
DOI: 10.1126/scitranslmed.3002614


Although growth factors naturally exert their morphogenetic influences within the context of the extracellular matrix microenvironment, the interactions among growth factors, their receptors, and other extracellular matrix components are typically ignored in clinical delivery of growth factors. We present an approach for engineering the cellular microenvironment to greatly accentuate the effects of vascular endothelial growth factor–A (VEGF-A) and platelet-derived growth factor–BB (PDGF-BB) for skin repair, and of bone morphogenetic protein–2 (BMP-2) and PDGF-BB for bone repair. A multifunctional recombinant fragment of fibronectin (FN) was engineered to comprise (i) a factor XIIIa substrate fibrin-binding sequence, (ii) the 9th to 10th type III FN repeat (FN III9-10) containing the major integrin-binding domain, and (iii) the 12th to 14th type III FN repeat (FN III12-14), which binds growth factors promiscuously, including VEGF-A165, PDGF-BB, and BMP-2. We show potent synergistic signaling and morphogenesis between α5β1 integrin and the growth factor receptors, but only when FN III9-10 and FN III12-14 are proximally presented in the same polypeptide chain (FN III9-10/12-14). The multifunctional FN III9-10/12-14 greatly enhanced the regenerative effects of the growth factors in vivo in a diabetic mouse model of chronic wounds (primarily through an angiogenic mechanism) and in a rat model of critical-size bone defects (through a mesenchymal stem cell recruitment mechanism) at doses where the growth factors delivered within fibrin only had no significant effects.


Although tremendous knowledge has emerged on the roles of growth factors (GFs) in tissue repair, progress in translating this knowledge to clinical impact has been more limited (15). One of the reasons for this poor clinical translation might lie in the rapid breakdown and clearance of GFs from tissue sites in vivo. Controlled-release (6, 7) and protein engineering strategies (6) have been explored to provide retention of GFs within matrices. For example, we developed a protein engineering approach that enzymatically links GFs to fibrin as a matrix for tissue repair. We have shown that this can considerably accentuate GF function (8, 9). This and other biomolecular engineering approaches have been inspired by the natural interactions existing between GFs and the extracellular matrix (ECM), because the ECM naturally binds and releases GFs (10, 11). In addition to GF sequestration, ECM-GF interactions also directly modulate GF signaling through co-association of integrins with GF receptors. In this context, complexes between ECM proteins and GFs can mediate enhanced GF receptor–integrin signaling by the formation of clusters between GF receptors and integrins (1214).

Fibronectin (FN) is a ubiquitous ECM protein that is a ligand for several integrins. We recently found that a domain including its 12th to 14th type III repeats (FN III12-14) acts as a highly promiscuous GF-binding site (15), which suggests that this domain could potentially be used as a generic approach to deliver GFs; for instance, GFs from the vascular endothelial growth factor (VEGF)/platelet-derived growth factor (PDGF), fibroblast growth factor (FGF), transforming growth factor–β (TGF-β), and neurotrophin families have shown strong affinity for FN III12-14 (15). Furthermore, recent evidence suggests that the close proximity of the major integrin-binding domain of FN, namely, the 9th to 10th type III repeats (FN III9-10), with FN III12-14 allows strong cross talk between ligated integrins and GF receptors (16). The two domains fused in a single polypeptide chain (FN III9-10/12-14) enhanced the proliferation and migration of endothelial cells (ECs) induced by VEGF-A165 (17).

We therefore hypothesized that GF-induced tissue healing could be strongly enhanced when GFs are delivered within a hydrogel comprising a well-defined microenvironment designed to trigger synergistic signaling between GF receptors and integrins. Accordingly, we generated recombinant FN fragments: FN III9-10 alone (to allow signaling via the integrin-binding domain), FN III12-14 alone (the GF-binding domain), or the two fragments fused in a single polypeptide chain (FN III9-10/12-14). The FN fragments were further engineered to include a substrate sequence for a coagulation transglutaminase, factor XIIIa (18), to provide covalent immobilization of the FN fragments in a clinically relevant fibrin matrix for tissue regeneration (Fig. 1). Here, we engineered a microenvironment that allowed sequestration of multiple GFs while promoting synergistic signaling between GF receptors and integrins. In a model of impaired wound healing in a diabetic mouse (db/db) and in a critical-size bone defect model in rats, we could drastically improve GF efficiency in tissue repair, with direct implications for use in humans.

Fig. 1

A multifunctional recombinant FN fragment is engineered to display the integrin-binding domain (FN III9-10) linked to the GF-binding domain (FN III12-14) and to comprise the substrate sequence α2PI1–8 for factor XIIIa. The fragment is covalently cross-linked into a fibrin matrix during the natural polymerization process of fibrin via the transglutaminase activity of factor XIIIa. The engineered matrix allows sequestration of GFs and joint integrin–GF receptor signaling, thus leading to cell recruitment, proliferation, and differentiation.


GF-induced cell proliferation and migration are enhanced via FN domains

To evaluate the extent to which FN fragments can modulate GF activities, we first focused on cell proliferation induced by VEGF-A165, PDGF-BB, and BMP-2 (bone morphogenetic protein–2) as GFs of interest in skin and bone repair. Human cells were stimulated with solutions of GFs co-delivered with soluble FN fragments (FN III9-10, FN III12-14, and FN III9-10/12-14) or full-length FN. We stimulated ECs with VEGF-A165, smooth muscle cells (SMCs) with PDGF-BB, and mesenchymal stem cells (MSCs) with PDGF-BB or BMP-2. Co-delivery of FN III9-10/12-14 significantly enhanced GF-induced proliferation of all cell types, as did co-delivery of full-length FN, when compared to stimulation with the soluble GFs only (no FN) and co-delivery of FN III9-10 or FN III12-14 (separate, not linked) (Fig. 2A). BMP-2, being a differentiation and migration factor rather than a growth-inducing factor, did not induce MSC proliferation (Fig. 2A). Only a slight increase in proliferation (<20%) was observed when full-length FN was co-delivered with BMP-2.

Fig. 2

Proliferation and migration induced by GFs is enhanced by co-delivering FN III9-10/12-14. Human ECs, SMCs, and MSCs were stimulated in vitro with soluble FN proteins only (white bars) or with FN proteins and GFs (black bars); VEGF-A165 for ECs, PDGF-BB for SMCs, and PDGF-BB or BMP-2 for MSCs. Functional blocking antibodies for α5β1 or αvβ3 integrins were delivered with FN III9-10/12-14 to parse out signaling mechanism. (A) Cell proliferation after 72 hours in the presence of GFs and FN proteins. Graphs show proliferation changes (cell number increase) over baseline. Data are means ± SEM (n = 5, in triplicate). (B) Number of cells per square millimeter that passed through a migration transwell after 6 hours. Data are means ± SEM (n = 4, in triplicate). For (A) and (B), the first statistical comparison is between the GF-only condition (No FN, white bars) and the GFs with FN protein conditions (black bars). The second statistical comparison raises the effect of integrin blocking. *P < 0.05; **P < 0.01; ***P < 0.001, Student’s t test.

Another key feature of VEGF-A165, PDGF-BB, and BMP-2 is their ability to induce cell migration; therefore, we also evaluated this cellular process. We stimulated ECs, SMCs, and MSCs with their respective factors (VEGF-A165, PDGF-BB, or BMP-2) co-delivered with FN fragments or with full-length FN. Compared to stimulation with GFs only and co-delivery with either FN III9-10 or FN III12-14, co-delivery of GFs with FN III9-10/12-14 or full-length FN significantly enhanced the GF effects on cell migration (Fig. 2B). Thus, in terms of both proliferation and migration, the activities of the very large and complex protein can be essentially mimicked in the smaller and recombinantly linked FN domains FN III9-10/12-14.

Proliferation and migration experiments suggested synergy between various GF- and FN III9-10/12-14–mediated signaling, because the sum of their effects alone is much lower than when they are delivered together. Without GFs, full-length FN or FN fragments had only minor effects on all cell types, which suggests that engaging integrins alone is insufficient to enhance cell proliferation or migration. Moreover, co-delivering GFs with [FN III9-10 + FN III12-14] (that is, the two domains delivered in concert, but not linked in the same polypeptide chain) did not significantly increase GF-induced proliferation or migration (Fig. 2, A and B). By contrast, the marked effects observed with GFs in the presence of FN III9-10/12-14 demonstrate that the two domains must be proximal in the same polypeptide chain.

GF signaling is enhanced by FN domains via integrins

To test which integrins were involved in the synergy between GFs and FN III9-10/12-14, we selectively blocked integrins with antibodies. Functional blocking of α5β1 and αvβ3 integrins was chosen because they are the two major receptors for FN III9-10. We found that GF synergies with FN III9-10/12-14 were mostly dependent on α5β1 integrin. Blocking α5β1 integrin during co-delivery decreased proliferation and migration of all cell types to the levels of stimulation with the GFs only, whereas blocking αvβ3 integrin had a smaller impact (Fig. 2, A and B). Functional blocking of integrins α5β1 or αvβ3 had no significant effects on basal proliferation (fig. S1).

We then investigated in vitro whether the enhanced cell proliferation and migration observed in the presence of FN III9-10/12-14 and GFs was the result of GF receptor phosphorylation. Co-delivery of GFs with FN III12-14 does not influence GF receptor phosphorylation (15), but we tested here whether fusing the integrin-binding domain FN III9-10 to FN III12-14 would have any effect. We found that phosphorylation of GF receptors (pVEGF-R2 and pPDGF-Rβ) was significantly higher when FN III9-10/12-14 or full-length FN was co-delivered compared to co-delivery with FN III9-10, collagen type I (a control ECM protein that does not bind α5β1 integrin), or without ECM proteins (fig. S2). Similarly, downstream in the GF signaling pathways, extracellular signal–regulated kinase 1/2 (ERK1/2) phosphorylation (pERK1/2) was also significantly higher. Co-delivering FN III9-10/12-14 or full-length FN also resulted in prolonged signaling; 40 min after stimulation, phosphorylation of GF receptors and pERK1/2 was still elevated compared to co-delivery with FN III9-10 or collagen type I, or without ECM (fig. S2).

Microenvironments containing FN domains enhance GF-induced morphogenesis

To translate the synergistic effects observed with soluble FN III9-10/12-14 (Fig. 2) to a more physiologically relevant form, we enzymatically conjugated the FN fragments to a fibrin matrix (Fig. 1). Fibrin matrices functionalized with FN III9-10/12-14 or FN III12-14 could effectively sequester VEGF-A165, PDGF-BB, and BMP-2, as evidenced by delayed release of GFs from the modified matrices (15) (fig. S3). Fibrin matrices were then functionalized with equimolar concentrations (2 μM) of FN III9-10, FN III12-14, [FN III9-10 + FN III12-14], FN III9-10/12-14, or full-length FN. We found that GF-induced morphogenesis was improved drastically in FN III9-10/12-14–functionalized matrices compared with all other matrices (Fig. 3).

Fig. 3

Morphogenesis induced by GFs is enhanced within FN III9-10/12-14– and FN III9*-10/12-14–functionalized fibrin matrices. Fibrin matrices were functionalized with FN proteins only (white bars) or with FN proteins and GF (black bars); VEGF-A165 for ECs, PDGF-BB for SMCs, and BMP-2 for MSCs. Functional blocking antibodies for α5β1 or αvβ3 integrins were delivered with FN III9-10/12-14 to parse out signaling mechanism. (A) EC tube-like structure formation in response to VEGF-A165 in fibrin-only or in FN protein–functionalized fibrin matrices. Representative tube-like structures (VEGF-A165 only or VEGF-A165 with FN III9-10/12-14) are shown in fluorescent images (actin is in green; nuclei are in blue). Scale bar, 50 μm. Data are means ± SEM (n = 3, in triplicate). (B) SMC sprouting in response to PDGF-BB in fibrin only or in FN protein–functionalized fibrin matrices. Representative sprouting induced by PDGF-BB only and by PDGF-BB within FN III9-10/12-14–functionalized matrix is shown in phase images. Scale bar, 50 μm. Data are means ± SEM (n = 3, with 5 to 10 spheroids per matrix). (C) MSC osteoblastic differentiation induced by BMP-2 after 2 weeks within FN protein–functionalized fibrin matrices. Expression of the osteoblastic genes ALPL, RUNX2, IBSP, and BGLAP was determined by qPCR and is shown as a fold expression over their respective baseline levels in a nonfunctionalized fibrin matrix (n = 3, in duplicate; data are mean fold expression ± SEM). For (A) and (B), the first statistical comparison is between the condition GF only and the conditions GF with FN proteins. The second statistical comparison raises the effect of integrin blocking. *P < 0.05; **P < 0.01; ***P < 0.001, Student’s t test. For (C), differences in relative expression over baseline and the effect of functional blocking antibodies were analyzed by pairwise fixed reallocation randomization test built in relative expression software tool (45). *P < 0.05; **P < 0.01.

Using an in vitro model of angiogenesis, we explored the tube-like organization of ECs in response to VEGF-A165 within the various microenvironments. Co-delivery of VEGF-A165 within FN III9-10/12-14– and full-length FN-functionalized matrices significantly accentuated the tube length compared to nonfunctionalized fibrin matrices (Fig. 3A). By contrast, FN III9-10–, FN III12-14–, or [FN III9-10 + FN III12-14]–functionalized matrices did not significantly improve the effects of VEGF-A165 on tube length.

Because concurrent vascular SMC and EC migration is directly linked to angiogenesis (19), we characterized sprouting of SMCs from spheroids (cluster of 300 to 400 SMCs) in response to PDGF-BB. With PDGF-BB, all matrices functionalized with FN proteins were able to enhance SMC sprouting significantly. However, the sprouting induced by PDGF-BB in matrices functionalized with FN III9-10/12-14 or full-length FN was much higher compared to that induced within matrices functionalized with the other domains (Fig. 3B). Thus, when enzymatically conjugated to fibrin matrices, FN III9-10/12-14 enhanced morphogenesis of both cell types (ECs and SMCs) directly involved in angiogenesis.

Considering an in vitro model that is relevant to bone repair, we explored the extent to which osteoblastic differentiation of MSCs induced by BMP-2 (a key step in bone regeneration) could be modulated through the FN domains. For these experiments, we engineered the fragments containing FN III9-10 with a point mutation on the 9th repeat (Leu1408 to Pro), which were named FN III9*-10 and FN III9*-10/12-14. This mutation was included to enhance FN III9-10 specificity for α5β1 integrin (20), because we and others have shown that osteoblastic differentiation of MSCs can be increased by engagement of this integrin (20, 21). We cultured human MSCs within the various FN-functionalized fibrin matrices, always in the presence of BMP-2. After 2 weeks, the expression of the osteoblastic markers alkaline phosphatase (ALPL), runt-related transcription factor 2 (RUNX2), integrin-binding sialoprotein (IBSP), and bone γ-carboxyglutamic acid–containing protein (BGLAP) was measured using quantitative polymerase chain reaction (qPCR). The effect of BMP-2 in a nonfunctionalized fibrin matrix served as a baseline. Expression levels of these osteoblastic markers were not up-regulated in matrices functionalized with full-length FN and FN III12-14, whereas MSCs cultured in matrices functionalized with FN III9*-10/12-14 differentiated along the osteoblastic lineage more extensively, as shown by a significant up-regulation of all osteoblastic markers (Fig. 3C). Full-length FN could not improve differentiation of MSCs, even if it binds BMP-2 and α5β1 integrin; however, the fragment FN III9*-10/12-14 synergistically enhanced it with BMP-2. It may be that other domains within full-length FN inhibit osteoblastic differentiation of MSCs.

Functional blocking antibodies were also used to assess the role of α5β1 and αvβ3 integrins in synergistic signaling within the fibrin matrices. For all morphogenesis models, blocking α5β1 integrin when GFs were co-delivered with FN III9(*)-10/12-14 reduced morphogenesis significantly to the levels induced by the GFs only, suggesting that the synergy was principally mediated through α5β1. In contrast, blocking αvβ3 integrin did not reduce morphogenesis significantly, suggesting that this integrin is less involved in the synergy (Fig. 3, A to C).

Wound healing in vivo

We evaluated whether this microenvironment engineering approach could be used to enhance skin healing. We used the db/db mouse, which is a genetic mouse model of diabetes mellitus that provides a well-established and widely used experimental system of impaired wound healing (22, 23). Compared to wild-type mice, this strain heals wounds principally by the formation of granulation tissue rather than by contraction, and its impairment is due to lower levels of several GFs and receptors (23). We treated full-thickness back-skin wounds of these mice (four wounds per mouse, n = 24 mice) with combined VEGF-A165 (100 ng) and PDGF-BB (50 ng) delivered by a fibrin matrix. Four groups were tested: fibrin only, fibrin functionalized with FN III9-10/12-14 only, fibrin containing GFs only, and fibrin functionalized with FN III9-10/12-14 and containing both GFs. Wound histology was analyzed after 7, 10, or 15 days. The wounds that received fibrin matrices containing GFs or FN III9-10/12-14 only did not differ from wounds treated with fibrin alone, in either amount of granulation tissue or extent of wound closure (the latter indicated by reepithelialization) (Fig. 4A). In contrast, whereas all wounds closed after 15 days, the delivery of both VEGF-A165 and PDGF-BB in the FN III9-10/12-14–modified fibrin microenvironment led to significantly faster wound closure at 7- and 10-day time points, as well as twofold greater development of granulation tissue by day 10 (Fig. 4A). Representative wound histology for all four treatment conditions is shown at the intermediate time point (day 10) in Fig. 4B. Clear differences between the granulation tissue thickness and the extent of reepithelialization can be seen between GFs delivered within the FN III9-10/12-14–functionalized matrix and all other conditions.

Fig. 4

Delivering GFs within the FN III9-10/12-14–functionalized fibrin matrices enhances skin wound healing in diabetic mice. Full-thickness back-skin wounds were treated with combined VEGF-A165 and PDGF-BB. Four groups were tested: fibrin only, fibrin functionalized with FN III9-10/12-14 only, fibrin containing GFs only, and fibrin functionalized with FN III9-10/12-14 containing GFs. (A) After 7, 10, and 15 days, wound closure and granulation tissue area were evaluated by histology. All points are means ± SEM (n = 6 per matrix per time point). *P < 0.05; **P < 0.01, Student’s t test. (B) Wound histology (hematoxylin and eosin staining) at 10 days. Black arrows indicate wound edges; red arrows indicate tips of epithelium tongue. The granulation tissue (pink-violet) is characterized by a large number of granulocytes with nuclei that stain in dark-violet or black. Muscle under the wounds is stained in red. Fat tissue appears as transparent bubbles. Scale bar, 1 mm. Higher magnification (×5) of the granulation tissue is shown on the right. (C) Five days after treatment, the percentage of ECs (CD31+ cells) was determined by flow cytometry. Data are means ± SEM (n = 6). *P < 0.05, Student’s t test. (D) Angiogenesis within the granulation tissue was assessed by staining for ECs (CD31+) and SMCs (desmin+) in the wound tissue at 10 days. E, epidermis; D, dermis; hashed line, basement membrane. Scale bar, 0.2 mm.

Angiogenesis is a crucial step in sustaining newly formed granulation tissue within the wound bed (24). Hence, we focused on the extent to which angiogenesis (represented by EC recruitment) differed between the treatments. After only 5 days, higher percentages of CD31+ ECs (>6%) were present in the wounds treated with GFs delivered by FN III9-10/12-14–functionalized matrices compared to wounds treated with the GFs only (<2%) (Fig. 4C). As a control, the percentage of granulocytes remained constant at about 80% for all experimental matrices (fig. S4). Thus, the increase in the number of ECs is indicative of more pronounced angiogenesis and not merely an increase in the amount of granulation tissue formed. Immunohistological analysis for CD31 and desmin (an SMC marker) after 7 and 10 days confirmed that angiogenesis within the granulation tissues was much more pronounced when GFs were delivered within FN III9-10/12-14–functionalized matrices (Fig. 4D). SMCs were observed to be associated with ECs, which suggests that blood vessels that developed after 10 days were mature and stable (Fig. 4D and fig. S5). Within the same condition, the degree of angiogenesis was not different between day 7 and day 10 (fig. S5).

Bone regeneration in vivo

We explored whether the FN III9*-10/12-14 fragment could be useful in engineering a microenvironment for bone healing. Because both BMP-2 and PDGF-BB are beneficial for bone repair (25), we tested whether fibrin matrices containing combined BMP-2 (100 ng) and PDGF-BB (50 ng) could enhance bone formation, both ectopically in nude mice and in critical-size calvarial defects in skeletally mature rats. Four groups were tested: fibrin only; fibrin functionalized with FN III9*-10/12-14 only; fibrin containing combined BMP-2 and PDGF-BB; and fibrin functionalized with FN III9*-10/12-14, with the addition of both GFs.

In the subcutaneous implant model (n = 6 mice per matrix formulation), matrices that contained the GFs did not completely degrade. Histology and immunostaining for RUNX2 and BGLAP, which are early and late markers of osteoblastic differentiation, respectively, showed that organized bone tissue was formed only when the GFs were delivered within FNIII9*-10/12-14–functionalized matrices (fig. S6). In contrast, although marginal expression of BGLAP was observed when the GFs were delivered alone, the formation of mature bone tissue was negligible (fig. S6).

A relevant model to illustrate human translational potential is the critical-size calvarial defect in a skeletally mature rat, which is a standard model for non-union bone healing (26, 27). We treated two defects measuring 6 mm each. After 4 weeks, bone healing—characterized by bone tissue deposition and coverage of the defects—was analyzed using microcomputed tomography (μCT) (Fig. 5, A and B) and histology (fig. S7). The delivery of GFs or FN III9*-10/12-14 alone in fibrin did not increase bone healing when compared to the lesions treated with fibrin only as a control (Fig. 5A). In contrast, GFs delivered within the FN III9*-10/12-14–functionalized matrix led to a marked increase of bone tissue deposition, and the defects were close to more than 96% compared with ~50% for the other matrices tested (Fig. 5, A and B, and fig. S7).

Fig. 5

Delivering GFs within the FN III9*-10/12-14–functionalized fibrin matrix greatly enhances bone regeneration in rats. Critical-size calvarial defect was treated with combined BMP-2 and PDGF-BB. Four groups were tested: fibrin only, fibrin functionalized with FN III9*-10/12-14 only, fibrin containing GFs only, and fibrin functionalized with FN III9*-10/12-14 and containing GFs. (A) Four weeks after treatment, bone regeneration was measured by μCT as bone volume (mm3). (B) Representative skull reconstitution is shown for FN III9*-10/12-14–functionalized matrix with GFs and for the fibrin matrix with GFs. The defect area is shaded. (C) Five days after treatment, the percentage of MSCs (CD45CD90+) within the defects that were also positive for CD54 (CD45CD90+CD54+) or for CD29 (CD45CD90+CD29+) was analyzed by flow cytometry. Representative plots are shown. Data are means ± SEM (n = 5 per condition). *P < 0.05; ***P < 0.001, Student’s t test.

Because recruitment of bone progenitor cells is a critical step for bone healing, we determined whether there was any difference in the extent of MSC recruitment by the matrices. After 5 days, we noted that a much higher number of rat MSCs (CD45CD90+CD54+CD29+ cells) were recruited when BMP-2 and PDGF-BB were delivered within the FN III9*-10/12-14–functionalized matrix (5.6 ± 2.3% of total cells) (Fig. 5C). Recruitment of ECs in the bone defects was not affected by the composition of the matrices, thus suggesting that enhanced neovascularization was not the driver behind improved bone healing in FN III9*-10/12-14–functionalized matrices laden with GFs (fig. S8).


Although GFs are crucial biomolecules for regenerative medicine, they are often ineffective during clinical trials (15). Among the many GFs that have entered into clinical trials to date, only PDGF-BB and BMP-2/7 have shown clinical efficacy (3, 28), but those factors still present safety (29, 30) and cost-effectiveness (1) issues. For example, PDGF-BB in Regranex, which is used for the treatment of chronic diabetic foot ulcers, bears a warning on its product insert from the U.S. Food and Drug Administration (FDA) about cancer risk (31). BMP-2 in Amplify, a product tested clinically in posteriolateral spinal fusion, showed clinical incidences of cancer that were of substantial concern to the FDA as well (32). A recent review of the safety of BMP-2 in spinal surgery concluded an incidence of adverse events from 10 to 50% depending on the approach (30). Thus, there exists strong motivation to engineer smart systems to reduce GF doses and potentially improve safety and cost-effectiveness. We have achieved this in the current study by considerably lowering the GF dose. We show here that the effect of VEGF-A165, PDGF-BB, and BMP-2 on tissue healing is considerably enhanced in combination with the recombinant-linked FN domains FN III9(*)-10/12-14.

Several in vitro assays demonstrated synergistic signaling only when the GF-binding domain (FN III12-14) and the integrin-binding domain (FN III9-10) of FN were in the same polypeptide chain (FN III9-10/12-14), not when the two domains were merely co-immobilized within the fibrin matrix ([FN III9-10 + FN III12-14]). Toward a mechanistic understanding, we found that α5β1 integrin was the main integrin involved in the synergistic effects of GFs and FN fragments. However, owing to the integrin-binding promiscuity of FN III9-10, it is possible that other integrins are involved in the synergy. It is likely that the physical proximity of integrins and GF receptors facilitates their reciprocal activation and their ability to jointly control the signaling pathways (25).

Preclinical evaluations of GFs for chronic skin wound healing are generally performed in rodents and most commonly in the db/db diabetic mouse (3336), even though the optimal disease model does not yet exist for human chronic wounds. Nevertheless, the genetically modified db/db mouse represents a clinically relevant model for diabetes-impaired skin wound healing (22, 23). Success in the db/db mouse model directly opens the way for clinical trials (33, 35). For example, in a clinical study of a fibrin-binding PDGF-AB variant, clinical investigation was launched on the basis of studies in rodents (37); thus, movement from the db/db mouse model to the clinic is a reality. Here, we used the db/db diabetic mouse model to demonstrate wound healing by low doses of VEGF-A165 and PDGF-BB in FN III9-10/12-14–functionalized fibrin matrices. This healing process was driven by enhanced angiogenesis in the wounds compared to various control microenvironments (Fig. 4). Improved vascular cell recruitment and neovessel formation, which sustain the newly formed granulation tissue, resulted from effective sequestration of VEGF-A165 and PDGF-BB and their synergistic signaling with integrins templated via the FN fragment. Granulation tissue morphogenesis translated to improved morphogenesis at the level of the dermal epithelium, as reflected by faster wound reepithelialization and closure.

Although GFs in combination with full-length FN showed similar angiogenic activity to FN III9-10/12-14 in vitro, the use of short, recombinant FN fragments that can be easily produced would be more beneficial for clinical translation, because only small quantities of intact FN can be purified from human plasma (38), making this a cost-prohibitive source. Furthermore, it has been shown that in vitro and in vivo use of full-length FN is much less potent than FN fragments for bone repair applications (Fig. 3C) (20, 39, 40).

Preclinical evaluations of bone repair materials and osteoinductive proteins commonly include critical-size bone defect models, such as the critical-size calvarial defect in the rat (27), although before human clinical trials, studies are often carried out in large animals. For example, in clinical development of a fibrin-binding parathyroid hormone(1–34) variant, studies were first carried out in the sheep before entering clinical evaluation in humans (41). Still, calvarial defects can answer questions about the biocompatibility and the biological functions of bone repair materials and morphogens before putting them into a clinical setting; the healing process in this model has been well characterized, and, the defect being too large to heal without intervention, the lesion represents a nonunion defect (27). In the critical-size calvarial bone defect model in the skeletally mature rat, we showed that a combination of BMP-2 and PDGF-BB delivered within the FN III9*-10/12-14–functionalized fibrin microenvironment leads to much better regeneration than obtained with the GFs delivered in fibrin only (Fig. 5, A and B, and fig. S7). We reason that the higher efficiency of BMP-2 and PDGF-BB delivered within the FN III9*-10/12-14 microenvironment was a result of their sequestration (fig. S3), as well as their synergistic signaling with the FN III9*-10/12-14 fragment (Figs. 2, A and B, and 3C). Moreover, BMP-2 and PDGF-BB delivered within the FN III9*-10/12-14 microenvironment led to recruitment of MSCs in vivo (Fig. 5C)—a phenomenon probably owing to PDGF-BB rather than to BMP-2, because PDGF-BB more strongly promotes MSC migration (Fig. 2B).

Using FN III9(*)-10/12-14, we could heal tissues with much lower doses of GFs than the ones previously reported. The fact that GFs delivered in fibrin only were ineffective is not surprising, given the low dose administered and the lack of smart delivery system. We used 100 ng per wound of VEGF-A165 and 50 ng per wound of PDGF-BB to treat full-thickness wounds, whereas 20 μg per wound of VEGF-A165 or 10 μg per wound of PDGF-BB (Regranex gel) applied topically for 5 consecutive days has been reported to be efficient in the db/db mouse (42, 43). For comparison, the Regranex gel used in humans to treat diabetic foot ulcers contains 100 μg of PDGF-BB/g of gel (28). We used 100 ng of BMP-2 and 50 ng of PDGF-BB to treat calvarial defect in the rat, whereas 1 μg of the GF alone is usually insufficient in rats (9). Typically, milligram quantities of BMP-2 are needed to treat tibial fractures in humans (3).

Here, we developed a simple and efficient system that greatly improves GF-induced wound healing and bone repair. This fibrin-, FN III9(*)-10/12-14–, and GF-based system can be translated to human application, because all of the components except the engineered FN fragment are already used in the clinic. Furthermore, such a strategy can be adapted into other biomaterial matrices and applied to other GFs as well, considering that FN III12-14 promiscuously binds GFs from several families (15). This study establishes integrin/GF receptor synergistic activities as a key parameter for GF translation into regenerative medicine treatments and demonstrates that GF therapies will not have optimal benefit without the support of functional components of the ECM. This work also highlights the crucial role of the microenvironment in modulating signaling of GFs and the importance of driving these biomolecules and biological matrices toward more widespread clinical use.

Materials and Methods

FN fragments, GFs, and cells

FN domains were synthesized as previously described (15, 20). Complete sequences are provided in the Supplementary Material (table S1). Because all FN fragments contain the α2PI1–8 sequence at their N terminus (table S1), they cross-link equally to fibrin matrix (>80%) (8, 20, 41). Full-length FN from human plasma was purchased from Sigma-Aldrich; VEGF-A165 and BMP-2 from R&D Systems; PDGF-BB from Invitrogen; human umbilical vein ECs from Promocell; human umbilical artery SMCs from Lonza; and human bone marrow–derived MSCs from Biopredic International.

In vitro assays

Methods of all in vitro assays, including proliferation, migration, and cell morphogenesis assays, are provided in the Supplementary Material.

Mouse skin wound model

Twenty-four C57BLKS/J-m/Lepr db (db/db) male mice were 10 to 12 weeks old at the start of the experiments. Their backs were shaved and four full-thickness punch biopsy wounds (6 mm in diameter) were created in each mouse. Directly after, fibrin matrices [80 μl total, fibrinogen (10 mg/ml), 2 μM FN III9-10/12-14, 100 ng of VEGF-A165, and 50 ng of PDGF-BB] were polymerized on the wounds. To avoid drying of the matrices, we covered the wounds with nonadhering dressing (Adaptic, Johnson & Johnson) and then with adhesive film dressing (Hydrofilm, Hartmann). After 5, 7, 10, or 15 days, animals were killed and the wounds were harvested for histological analysis. Details of the histomorphometric analysis and immunohistochemistry of the wound tissue sections are provided in the Supplementary Material.

Rat calvarial defect model

Eighteen male Sprague-Dawley albino rats, 8 weeks old, were used. For cell recruitment experiments, a craniotomy of 8-mm diameter in the junction of the two ossa frontalia with the two ossa paritealia was created with a drill. A preformed matrix [80 μl total, fibrinogen (10 mg/ml), 2 μM FN III9*-10/12-14, 100 ng of BMP-2, and 50 ng of PDGF-BB] was implanted into the defect. The periosteum was closed with sutures and the soft tissues were closed with skin staples. For bone regeneration, two 6-mm-diameter craniotomy defects were created in the parietal bones of the skull on each side of the sagittal suture line. Matrices were then directly polymerized into the defects, and the periosteum and soft tissues were closed. Rats were killed 4 weeks after surgery, and skulls were analyzed by μCT as described previously (44). The detailed μCT analysis procedure is provided in the Supplementary Material.

Flow cytometric analysis

For the skin wound model, wound tissues were harvested 5 days after treatment with biopsy punch. For the calvarial model, implanted matrices were recovered after 5 days by means of forceps. Samples were digested twice (45 min at 37°C) with type I collagenase (100 U/ml) and 0.1% trypsin (Sigma-Aldrich) in phosphate-buffered saline (PBS). Next, enzymes were inactivated with Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum, and recovered cells were washed with PBS. The phenotype of recruited cells was analyzed by flow cytometry with monoclonal antibodies against CD31, GR-1, CD29, CD45, CD54, and CD90 (BioLegend).


Results are presented as means ± SEM. Statistical comparisons were based on Student’s t test or analysis of variance (ANOVA) with Tukey post hoc test for pairwise comparisons. A confidence level of 95% was considered significant. In vitro assays were conducted in at least duplicate per experiment and replicated in three separate experiments.

Supplementary Material

Materials and Methods

Fig. S1. Functional blocking antibodies for α5β1 and αvβ3 integrins do not influence basal proliferation of ECs, SMCs, and MSCs.

Fig. S2. Growth factor signaling is enhanced with FN III9-10/12-14.

Fig. S3. Prolonged retention of VEGF-A165, PDGF-BB, and BMP-2 in fibrin matrices functionalized with FN III9-10/12-14.

Fig. S4. The percentage of granulocytes within skin wounds is not affected by matrix composition.

Fig. S5. The presence of blood vessels within wound granulation tissue when VEGF-A165 and PDGF-BB are delivered within the FN III9-10/12-14 microenvironment.

Fig. S6. Formation of subcutaneous ectopic bone induced by BMP-2 and PDGF-BB within the FN III9*-10/12-14 microenvironment.

Fig. S7. Bone healing in the rat calvarial model is enhanced by delivering BMP-2 and PDGF-BB within the FN III9*-10/12-14 microenvironment.

Fig. S8. EC recruitment to cranial defect is not affected by matrix composition.

Table S1. Protein sequences of the FN fragments.


References and Notes

  1. Acknowledgments: We thank C. Dessibourg for help with the production of FN domains, X. Quaglia for help with animal surgery and histology, S. Srouji for implant surgery, and the Proteomic Core Facility, the Histology Core Facility, and the Microscopy Core Facility of the Ecole Polytechnique Fédérale de Lausanne. Funding: This work was funded by the European Community’s Seventh Framework Programme in the project Angioscaff NMP-LA-2008-214402 and by the Swiss National Science Foundation. Author contributions: M.M.M. (in vitro, in vivo), F.T. (in vivo, fluorescence-activated cell sorting), M.M. (in vitro), S.T. (in vivo skin healing), D.B.-D. (in vivo implants), and G.A.K. (μCT) contributed to collection of experimental data. M.M.M., F.T., E.L., and J.A.H. analyzed the data. M.M.M., F.T., R.M., S.A.E., and J.A.H. contributed to the writing of the paper. Competing interests: J.A.H. is an inventor on patents for the factor XIIIa–mediated method for linking proteins to fibrin (US 6,331,422; US 6,607,740; US 7,601,685; US 2004/0082513; EP 1 124 590; and EP 1 280 566). These patents have been licensed to Kuros Biosurgery AG, in which J.A.H. holds equity.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article