Laminin heparin-binding peptides promiscuously bind growth factors and enhance diabetic wound healing

Laminin, as a key component of the basement membrane extracellular matrix (ECM), regulates tissue morphogenesis. We show that multiple laminin isoforms promiscuously bind to growth factors (GFs) with high affinity, through their heparin binding domains (HBDs) located in the a chain LG domains. Interestingly, these domains also bind to syndecan cell-surface receptors, promoting attachment of fibroblasts and endothelial cells. We next explore application of these multifunctional laminin HBDs in skin healing in the type 2 diabetic mouse. We demonstrate that covalent incorporation of laminin HBDs into fibrin matrix enables the slow-release of GFs. Incorporation of the α33043-3067 laminin HBD significantly enhances in vivo wound-healing efficacy of vascular endothelial cell growth factor (VEGF)-A165 and platelet-derived growth factor (PDGF)-BB, under conditions where the GFs alone in fibrin are inefficacious. This laminin HBD peptide may be clinically useful by improving biomaterials as both GF reservoirs and cell scaffolds, leading to effective tissue regeneration.


Introduction
Laminins are the most abundant glycoproteins of the basement membrane extracellular matrix (ECM) and can be found in almost all tissues of the body.
They play essential roles in establishment of tissue architecture and stability, and provide cells with a structural scaffold. As such, laminins are involved in a variety of biological processes ranging from tissue survival, angiogenesis 1 and neural development 2 , to skin re-epithelialization and wound healing [3][4][5] , and even cancer metastasis 6,7 . Laminins have been shown to regulate core cellular activities, such as adhesion, apoptosis, proliferation, migration and differentiation. Laminin is structured as a heterotrimer comprising three chains, a, b, and g, that assemble into a cross shape 3,8,9 . At least 16 different isoforms of laminin exist, made by various combinations of the five a (LAMA1-5), three b (LAMB1-3), and three g (LAMC1-3) chains that have been identified. Isoforms are accordingly named by their -αβγ chains: for instance, laminin-332 contains α3, β3 and γ2 chains. The differential expression of laminin isoforms depends on tissue type and state 7,10 .
In skin, for example, the epithelial basement membranes contain laminin-111 and laminin-211 during embryogenesis, but predominantly laminin-332 and laminin-511 in adults, the latter isoform expression seeming to diminish with age.
Interestingly, dermal fibroblasts can transiently re-express laminin-211 after wounding 11 . Moreover, dermal blood vessels specifically express laminin-511 and laminin-311, in addition to the laminin-411 commonly found in endothelial basement membrane 3 .
In skin wound healing, laminin has been shown to have a critical role in re-epithelialization and angiogenesis 1,3 . Indeed, laminin a chain possesses 5 laminin-type G domain (LG) modules at the C-terminus, arranged in a tandem array, that are differentially processed under homeostatic conditions or during tissue repair 8 . In fact, laminin a3, a4, and a5 chains are physiologically cleaved by proteases, such as plasmin and elastase, in the linker sequence between LG3 and LG4 domains [12][13][14][15] , and processing in this region has been shown to correlate with the speed of wound closure 3 . As an example, laminin-332 is present in a cleaved form under homeostatic conditions; however, the expression of laminin-332 is upregulated after injury and the LG4 and LG5 domains are subsequently more present in wounds 3,12 . The release of LG4-LG5 module has been demonstrated to undergo further processing that releases bioactive peptides 12 , which promote blood vessel formation and keratinocyte migration, notably via syndecan binding 1,[16][17][18] . In addition, laminin LG modules have been shown to bind to heparin sulfate, perlecan and fibulin-1 19 , as well as to a number of integrins, e.g. a3b1, a6b1, a7b1 and a6b4 20 .
In the past, ECM glycoproteins, including laminin, have been mainly considered for their biomechanical role in providing substrates for cell adhesion and migration, via direct interactions with cell-surface receptors 21 . Later, some ECM proteins have raised interest for their ability to regulate the partitioning and bioavailability of soluble signaling molecules within tissues, thus highlighting a new role for the ECM in coordinating the spatio-temporal release of these molecules. Examples of such soluble signals are growth factors, which are key morphogenetic proteins broadly involved in the control of core cellular behaviors, and which have been shown to be crucial for wound healing [22][23][24] . Particularly, fibronectin 25,26 , vitronectin 27 , and fibrinogen 28 have been reported to directly bind to GFs, which control their release kinetics in vivo, acting as a GF reservoir.
Importantly, it has been noticed that GF binding occurs at the heparin-binding domains (HBDs) of ECM glycoproteins, especially in fibronectin 25,26 , vitronectin 27 , fibrinogen 28 and osteopontin 29 . Sustained release of GFs from the ECM eventually enhances and prolongs GF-receptor signaling. Furthermore, the proximity of GF-binding sites and integrin-binding sites in some ECM glycoprotein chains, as in fibronectin or vitronectin, can induce synergistic GF signaling by clustering GF receptors and integrin at the cell surface.
Despite the importance of laminin in the ECM composition and its role in tissue morphogenesis, interactions between laminin and GFs have been poorly investigated to date. Therefore, in this study, we aim to elucidate direct interactions between laminin and GFs; we hypothesized that laminin could bind to several GFs, as observed in other ECM glycoproteins, and that these interactions could take place at laminin HBDs, some of which having been already identified in LG domains 30-32 . We then sought to demonstrate the GF reservoir role of laminin in vivo, and its potential to control the delivery of GF from fibrin materials in chronic wound healing, as a clinically-relevant model of tissue repair.

Multiple GFs bind to multiple isoforms of laminin
We first examined the capacity of a variety of full-length laminin isoforms (-111, -211, -332, -411, -421, -511, and -521) to bind GFs from the VEGF/PDGF, FGF, BMP, NT, IGF, EGF and CXCL chemokine families. Binding of laminin to absorbed GFs was detected using an antibody against laminin, and signals greater than 0.1 were considered to be indicative of a binding event. Overall, we found that multiple GFs strongly bound to all tested laminin isoforms (Fig. 1A).
SPR chips were functionalized with laminin-521, and growth factors were flowed over the surface. The obtained binding curves were fitted with Langmuir binding kinetics to calculate specific dissociation constants (K D ) (Fig. 1B). K D values were 5.8 nM for VEGF-A165, 12.9 nM for PlGF-2, and 24.2 nM for PDGF-BB.
The nM range of K D values demonstrated the strong binding affinities of laminin-521 to the selected GFs.

GFs bind to the HBDs of laminin
Because the GFs that bound to laminins have also been previously reported to bind to other ECM glycoproteins through HBDs 25,28,33 , we hypothesized that HBDs of laminins might be responsible for the interactions between GFs and laminin. To address this hypothesis, ELISA assays were repeated for VEGF-A165, PlGF-2 or FGF-2 in the presence of heparin added in excess (10 µM). As a result, we observed that heparin inhibited GFs binding to laminin ( Fig. 2A-C), supporting that laminin HBDs mediated interactions with GFs. To further confirm this, we tested direct GF binding to the LG domains from human laminin a3, a4 and a5, within which HBDs of laminin were localized 30 . We found that VEGF-A165, PlGF-2, PDGF-BB, and FGF-2 bound to laminin LG domains a3 2928-3150 , a4 826-1816 and a5 3026-3482 , in contrast to VEGF-A121 and PlGF-1 which did not show any binding ( Fig. 3A-C), as tested by ELISA. The binding affinities between a3 2928-3150 and VEGF-A165 or PDGF-BB were then measured by SPR, and K D values were 1.2 nM for VEGF-A165, and 10.2 nM for PDGF-BB (Fig. 3D). These data again demonstrated the strong affinities of the laminin LG domain to the tested GFs.

Retention of VEGF-A165 and PDGF-BB in fibrin matrix is increased by the incorporation of laminin HBDs peptides
We then sought to determine whether laminin HBD peptides, which showed binding to GFs, were able to improve the retention of VEGF-A165 and PDGF-BB within fibrin matrix. These GFs are known to be quickly released from fibrin matrices upon delivery, which limits their wound healing efficacy in vivo 26,28,33 .

Laminin HBD-functionalized fibrin matrices potentiate GFs and promote wound healing in vivo
We further evaluated whether fibrin matrices engineered with laminin-HBD peptides could enhance skin repair in a model of delayed wound healing, by controlling the release of VEGF-A165 and PDGF-BB in vivo. More precisely, VEGF-A165 (100 ng/wound) and PDGF-BB (50 ng/wound) were co-delivered from fibrin matrix onto full-thickness back-skin wounds in type 2 diabetic db/db mice, which provides a well-established and clinically-relevant model of impaired wound healing 26 . Here, we particularly functionalized fibrin with the laminin peptide a3 3043-3067 , since it bound to GFs and syndecans, and promoted fibroblast and endothelial cells adhesion in vitro ( Fig. 4-6). Four groups were tested: fibrin only, fibrin functionalized with a 2 PI 1-8 -a3 3043-3067 , fibrin containing GFs, and fibrin functionalized with a 2 PI 1-8 -a3 3043-3067 and containing GFs.
Wound histology was analyzed after 7 days, considering that wounds are normally fully closed after 15 days when treated with fibrin matrix 26 . As a result, wounds that received fibrin matrices containing GFs or a 2 PI 1-8 -a3 3043-3067 peptide only did not differ from wounds treated with fibrin alone, neither in amount of granulation tissue nor in extent of wound closure (the latter indicated by re-epithelialization) (Fig. 7C, D). In contrast, the co-delivery of VEGF-A165 and PDGF-BB in fibrin functionalized with a 2 PI 1-8 -a3 3043-3067 led to a significantly faster wound closure after 7 days, as well as a significant increase in granulation tissue formation (Fig. 7C, D). Representative wound morphology for all four treatments is presented in Fig. 7E. Clear differences in granulation tissue thickness and extent of re-epithelialization can be visualized when GFs were delivered within the a 2 PI 1-8 -a3 3043-3067 peptide-functionalized fibrin matrix compared to the other conditions. This set of data indicates that a 2 PI 1-

Discussion
As a cell scaffold protein, laminin tightly regulates cell adhesion, motility, survival and differentiation, thus playing a critical role in tissue homeostasis and wound healing 8 . In fact, although the expression of particular laminin isoforms depends on the tissue type and state, laminins reportedly promote tissue repair in muscle, nerve, liver, and skin 34 . In this study, we uncovered a novel property of laminin, showing that multiple laminin isoforms promiscuously bind to heparin-binding GFs with high affinities (Fig. 1). Interactions between GFs and the ECM are known to be essential for controlling GFs release kinetics in vivo, which strongly modulates tissue morphogenesis 26 . While sequestration of GFs within basement membranes was previously reported to be mediated through binding to glycosaminoglycans 35 , our data demonstrates that laminin can serve this GF reservoir function as well.
Interestingly, most of the GFs that bind to laminin have also been shown to bind to other ECM proteins, such as fibrinogen and fibronectin, suggesting similar binding mechanisms 25,26,28,33 . However, a few exceptions can be noticed; for example, PDGF-CC showed specific binding to laminin, but neither to fibrinogen nor fibronectin 28,33 . In contrast, BMP-7, which binds specifically to fibronectin 25 but not fibrinogen 28 , did not show binding to laminin. Moreover, kinetic analysis of laminin isoform-521 binding towards VEGF-A165, PDGF-BB and PlGF-2 revealed K D in nM range, highlighting the broad yet high affinity interactions between laminin and GFs.
Furthermore, we identified GF binding to laminin to mainly occur at heparin-binding sites, by showing that heparin directly compete with GF-laminin interactions and dramatically reduced them when added at high concentration ( Fig. 2). Additionally, all laminin HBDs tested in this study were able to bind to GFs (Fig. 3, 4). Yet, unexpectedly, GFs binding to laminin does not seem to be limited to HBDs, as a few non-heparin binding peptides also bound to some GFs, notably α3 3031-3043 and a5 3539-3550 . These peptides are human alignments of reported mouse HBD peptides, called A3G75 and A5G94 respectively 30 .
Although they do not show heparin binding under our experimental conditions, they may bind to GFs via another mechanism. Thus, the mechanism of GF-binding to laminin still remains incompletely clarified, and may be resolved by further crystallography studies of GF-laminin complex.
Physiologically, proteolytic cleavage of LG4 and LG5 domains is crucial for the deposition of laminin in the native ECM 12,14,36 . Upon tissue injury, laminin is overexpressed, and LG4-LG5 domains accumulate in wounds 3,12 , wherein they promote tissue healing mechanisms 37 . In this study, we particularly characterized laminin-derived peptides that are located just before the proteolytic cleavage site, in the linker between the LG3 and LG4 domains, or within the LG4-LG5 domains (Table. 2, Fig. 4A). On one side, we discovered 3 novel heparin-, GF-and syndecan-binding peptides within the LG3-LG4 linker regions of a3, a4, and a5 chains, namely a3 2932-2951 , a4 1408-1434 , and a5 3300-3330 , identifiable through their highly cationic sequences (Fig. 4). Since a3, a4 and a5 chains are known to be predominantly present in their processed form (i.e. lacking LG4-LG5) in mature, unwounded skin 3,12-15 , it is likely that these peptides are exposed in vivo under homeostatic conditions, thus providing both GF ligands and cell adhesion sites in basement membranes. Interestingly, laminin a1 chain, which is not proteolytically processed 13 , and a2 chain do not contain such cationic sequences in the LG3-LG4 linker region, which might reflect functional differences between α chain isoforms. On the other side, we identified 5 peptides in the LG4 and LG5 domains of a3, a4 and a5 chains that displayed specific binding to GFs, in particular to VEGF-A165. Among them, a3 3043-3067 , a5 3539-3550 , and a5 3417-3436 additionally bound to PDGF-BB, FGF-2 and PlGF-2 with high affinities (Fig. 4). These growth factors are well-known as key regulators of the wound healing cascade, and are particularly involved in wound angiogenesis. Therefore, we propose that the reported positive effects of LG4-LG5 domains during wound healing might be related to promiscuous interactions with GFs, in addition to via binding to syndecans and release of laminin-derived pro-angiogenic peptides 1,16,17 .
Previous studies have shown that the formation of ECM protein:GF complexes can synergistically enhance GF receptor signaling 26,27 . For example, simultaneous presentation of GFs and integrin-binding sites by an engineered fibronectin fragment (namely FN III9-10/12-14) incorporated into fibrin drastically improved the effect of VEGF-A165 and PDGF-BB on skin repair 26 . Here, we identified 5 laminin HBDs that are able to bind to both GFs and syndecan cell-receptors ( Fig. 4 and 5), among which α3 3043-3067 , α4 1521-1543 and α5 3417-3446 further promoted cell attachment (Fig. 6). Although syndecans are not known to directly activate major signaling pathways, they support cell adhesion and integrin signaling 18 . Moreover, direct binding of laminin peptides from LG domains to integrins has also been reported; for example, the integrin a3b1 binds to a3 2932-2943 38 . Nevertheless, in our assays, EDTA did not abolish cell adhesion, suggesting that initial cell attachment was mediated by syndecans rather than integrins (the binding of which is Ca 2+ -dependent). Consequently, and considering the short length of the laminin HBD peptides, it is unlikely that laminin HBD peptides can enhance GF signaling via synergy with integrins.
However, we believe that both GF-and cell-binding properties of laminin HBDs substantially contribute to the promotion of wound healing, and so propose a potential application for laminin HBDs in the treatment of chronic diabetic ulcers ( Fig. 7).
Although GFs are promising drugs for tissue regeneration, their uncontrolled delivery upon application on wounded tissue had limited their clinical efficacy and safety to date 39,40 (Fig. 7). In contrast, wounds treated with fibrin matrix containing GFs only, in which PDGF-BB and VEGF-A165 were not specifically retained in the fibrin matrices, had no detectable effect on wound healing at the tested dose (Fig. 7). We have previously shown that HBDs derived from fibrinogen (Fg b15-66 (2) ) and fibronectin (FN III9-10/12-14) promote wound healing in combination with GFs when incorporated into fibrin matrix 26,28 . A main advantage of using the laminin HBD peptide for this purpose, compared to the previously described GF-binding domains, is production simplicity: the laminin HBD peptide is short enough to be chemically synthesized in large scale, rather than requiring recombinant expression. Furthermore, we showed that a laminin HBD can functionalize fibrin matrix in both aspects as a GF reservoir and an adhesion-promoting cell scaffold ( Fig. 6 and 7).
In conclusion, we found that multiple isoforms of laminin promiscuously bind to a number of GFs from the VEGF/PDGF, FGF, BMP, and NT families, in addition to HB-EGF and CXCL12g, through their HBDs. By engineering a fibrin matrix displaying the a3 3043-3067 laminin HBD, as a proof example, we have demonstrated that the laminin HBD peptide promotes skin wound closure in the db/db mouse, as a model of delayed wound healing, when applied with VEGF-A165 and PDGF-BB. In addition to highlighting a GF-modulating function for laminin, an important tissue repair protein, we show that both GF-and cell-binding character promotes tissue repair when incorporated within fibrin matrix, which may be clinically useful.  and rate constants (K on and K off )] determined from the fitted curves are shown.    LG2 LG3 Linker LG4 LG5 LG4 LG3 LG2 LG1 LG5 α3 chain LG4 LG3 LG2 LG1 LG5 α4 chain LG4 LG3 LG2 LG1 LG5       supplemented with 1% fetal bovine serum (FBS) and 100 µg/mL VEGF-A165, with or without 5 mM EDTA (Sigma-Aldrich). Cells were plated at 3000 cells/well on laminin peptide pre-coated plates and incubated for 30 min at 37°C, 5% CO 2 . Then, the medium was removed, and wells were quickly washed three times with PBS. Cell numbers were quantified using a CyQUANT assay, according to the manufacturer's instructions (Invitrogen). All cell lines were checked for mycoplasma contamination and used in passages from 5 to 8.

Release of GF from fibrin matrix
Fibrin matrices were generated with human fibrinogen (VWF and fibronectin depleted, Enzyme Research Laboratories) as described previously 1

Mouse skin chronic wound healing model
Skin wound healing assays were performed as previously reported 1

Histomorphometric analysis of wound tissue sections
Histomorphometric analyses were performed as previously reported 1 . Briefly, an area of 8 mm in diameter, which includes the complete epithelial margins, was excised. Wounds were cut in the center into two and embedded into paraffin.
Histological analysis was performed on 5 μm serial sections. Images were captured with an EVOS FL Auto microscope (Life Technologies). The extent of re-epithelialization and granulation tissue formation was measured by histomorphometric analysis of tissue sections (H&E stain) using ImageJ software. For analysis of re-epithelialization, the distance that the epithelium had traveled across the wound was measured; the muscle edges of the panniculus carnosus were used as indicator for the initial wound edges; and reepithelialization was calculated as the percentage of the distance of edges of the panniculus carnosus muscle. For granulation tissue quantification, the area covered by a highly cellular tissue was determined.

Statistical analysis
Statistical methods were not used to predetermine necessary sample size, but into treatment groups within a cage immediately before the wound surgery and treated in the same way. GF-laminin binding ELISA assays were repeated 4 times. Wound healing assays were repeated 5 times. The P values less than 0.05 are considered to be significantly different. The P values less than 0.05 and 0.01 indicate symbols * and **, respectively.

Data availability
The data that support the findings of this study are available from the authors on reasonable request.