Peptide‐mediated surface coatings for the release of wound‐healing cytokines

Supporting the wound healing process by sending the appropriate cytokine signals can shorten healing time and overcome chronic inflammation syndromes. Even though adhesion peptides consisting of Arg‐Gly‐Asp (RGD) are commonly used to enhance cell‐surface interactions, peptide‐mediated cytokine delivery has not been widely exploited so far. Cytokines interact with high affinity with their cognitive receptors but also with sulfated glycosaminoglycans (GAGs), both of which form a base for incorporation of cytokines into functional biomaterials. Here, we report on a mussel‐derived surface coating as a prospective cytokine delivery system using covalently bound heparin mimetics, receptor‐derived chemokine‐binding peptides, and heparin‐binding peptides (HBP). The latter enabled non‐covalent immobilization of heparin on the surface followed by chemokine binding and release, whereas the former allowed direct non‐covalent chemokine immobilization. The peptide displayed excellent binding to custom‐made polystyrene 96‐well plates, enabling convenient testing of several compounds. Released chemokine successfully induced migration in Jurkat cells, especially for the non‐covalent heparin immobilization approach using HBPs as evaluated in a transwell assay. In comparison, heparin‐mimetic coatings, comprised of sulfated peptides and GAG derivatives, proved less efficient with respect to amount of immobilized chemokine and migratory response. Thus, our study provides a roadmap for further rational optimization and translation into clinics.


| INTRODUCTION
Chemokines are a family of small 8-12 kDa proteins that mediate migration and arrest of their target cells. They are crucial players in tissue regeneration as they orchestrate the recruitment of various cell types to the wound site, modulating the immune response and directing angiogenesis. CXC chemokine ligand 12 (CXCL12) or stromal cell-derived factor 1α (SDF-1α) is a pleiotropic cytokine that signals through its G protein-coupled receptors CXCR4 and CXCR7 and acts as a chemoattractant. In adult organisms, CXCL12 promotes angiogenesis, the recruitment of hematopoietic stem and progenitor cells, immune cell trafficking, and neuronal regeneration (Guyon, 2014;Liekens, Schols, & Hatse, 2010). Thus, the chemokine is involved in various stages of wound healing including homeostasis, inflammation, and proliferation (Ratajczak et al., 2006), making it an appealing component of biomaterials in tissue engineering. However, the development of suitable delivery strategies remains an ongoing field of research. Simple adsorption to the material already promotes tissue regeneration (Ji et al., 2013), but stabilization against rapid protein clearance and a controlled release to reduce side effects is desirable. In opposition, covalent immobilization approaches require protein engineering to incorporate functional groups for surface binding and cleavable linkers for protein release (Steinhagen et al., 2014), which can alter protein activity. Thus, affinity-based delivery systems are to be exploited, where a binding ligand is immobilized to the material, which then reversibly binds the unmodified chemokine.
In the extracellular matrix, glycosaminoglycans (GAGs) like heparin and heparan sulfate represent the most important binding partners of CXCL12. Stabilization by GAGs is essential for the formation of chemotactic gradients and protects the chemokine from inactivating posttranslational modifications (Janssens, Struyf, & Proost, 2018). This heparin-binding affinity can be exploited for the delivery by covalent or non-covalent incorporation of heparin into biomaterials (Vulic & Shoichet, 2014). Heparin-binding peptides (HBPs) contain a number of basic residues to interact with the negatively charged GAGs and therefore offer the possibility to immobilize heparin by electrostatic interaction (Cardin & Weintraub, 1989;Sakiyama-Elbert & Hubbell, 2000). In these cases, the cytokine as well as the mediating GAG will be released over time. Specifically functionalized heparin or heparin analogs can mediate orthogonal conjugation to the material.
Sulfated hyaluronic acid (sHA) derivatives have been developed as heparin-like protein-binding polymers that, when applied as high molecular weight sHA or even as short oligomeric compounds, have proven to bind CXCL12 (Köhling et al., 2019;Purcell et al., 2014).
Another possibility is the complete replacement of the sugar with peptidic compounds. However, for explicit delivery of CXCL12, only receptor-derived peptides (RDPs) are known so far (Peled, Eizenberg, & Vaizel-Ohayon, 2012), some of which include also posttranslational tyrosine sulfation (Veldkamp, Seibert, Peterson, Sakmar, & Volkman, 2006). For the delivery of other cytokines, a number of heparin-mimetic peptides (HMPs) are known. Sulfates are included either by O-sulfation of serine and threonine residues or by incorporation of sulfotyrosines. Maynard and Hubbell identified the consensus sequence SY*DY*G, where Y* is a sulfotyrosine, which was confirmed to bind vascular endothelial growth factor (VEGF; Maynard & Hubbell, 2005), transforming growth factor β, and bone morphogenetic protein 4 (Hendrikse, Spaans, Meijer, & Dankers, 2018). Within another approach, sulfobenzoic acid (sba) is coupled to lysine side chains to incorporate sulfate groups into a molecular brush (Mammadov et al., 2011). The peptide EGDK(sba)S was shown to bind VEGF, hepatocyte growth factor, and fibroblast growth factor 2 (Mammadov, Mammadov, Guler, & Tekinay, 2012) to enhance angiogenesis and wound healing.
Within the present study, affinity-based delivery strategies for CXCL12 inspired by extracellular matrix-proteoglycan interactions were investigated. Here, functional peptide coatings were used to immobilize GAGs and GAG mimetics onto a polymer surface in order to provide a prospective chemokine delivery system that is easily prepared, purified, and analyzed. This was enabled by incorporation of L-3,4-dihydroxyphenylalanine (DOPA), a nonproteinogenic amino acid identified in the byssus of blue mussels (Waite & Tanzer, 1981), into surface-binding peptides. In this work, application of the catecholbased coating was shown to go beyond metal oxide surfaces, where display of cell adhesives promoted osseointegration and endothelialization (Clauder et al., 2019;Pagel et al., 2016). Using the peptide as a modifiable anchor provided a platform to compare HBPs, subsequently loaded with heparin, sHA derivatives, and sulfated peptides but also receptor-derived chemokine-binding peptides for CXCL12 binding. Finally, gradient formation was investigated by transwell migration studies to test for effective cell recruitment.

| Peptide synthesis
All peptides were synthesized by a combination of manual couplings and automated solid-phase peptide synthesis using a Syro I peptide synthesizer (MultiSynTech) under standard Fmoc/tert-butyl (tBu) conditions. The N α -Fmoc-deprotection was achieved with piperidine in DMF, and amino acids were activated with equimolar amounts of N,N-diisopropylcarbodiimide (DIC) and 1-hydroxybenzotriazole (HOBt) or hydroxyiminocyanessigsäureethylester (oxyma), respectively. The peptides HBP-N 3 and HBP 2 -N 3 were produced in the synthesizer on Rink amide resin and N-terminally modified with ε-azido-L-lysine. Heparin-mimetic peptide 1 (HMP1) and HMP2 as well as motif repetitions were likewise produced automatically, whereby Fmoc-L-Lys(Dde)-OH was incorporated at future sulfation sites. After N-terminal trityl-protection or acetylation, selective deprotection of the lysine side chain was achieved with 3% hydrazine in DMF. sba was coupled in two-fold excess with equimolar amounts of 1-[bis (dimethylamino)methylene]-1H- 1,2,3-triazolo[4,5-b] pyridinium 3-oxide hexafluorophosphate and one equivalent N,Ndiisopropylethylamine overnight. RDP was likewise produced automatically on Rink amide resin. The mussel-derived peptide (MP) derivatives were analogously synthesized as described elsewhere (Pagel et al., 2016). MP(+) and MP(−) were produced by elongation with Fmoc-L-aminohexanoic acid and subsequently biotinylated with HOBt/DIC in NMP. Conjugation of MP with the HBPs HBP-N 3 and HBP 2 -N 3 by Cu(I)-catalyzed azide alkine cycloaddition was likewise performed as described by Pagel et al. (Pagel et al., 2016). Thiol-and maleimide-functionalized compounds were combined by Michael addition reaction, which was carried out for at least 2 h in degassed sodium phosphate buffer pH 6.8. Successful conjugation was observed by a retention time shift during reversed-phase highperformance liquid chromatography (RP-HPLC) analysis, whereby in all cases, less than 2% unfunctionalized MP remained.

| Full cleavage and peptide purification
Full cleavage of the peptides was achieved by incubation with a mixture of trifluoroacetic acid (TFA) and scavenger for 3 h at room temperature, shaking. For HBPs, 90% TFA and 10% water was used.
Subsequently, the mixture was transferred into a dialysis tube (cellulose acetate, molecular weight cut-off = 1,000 Da) and excessively dialyzed against degassed water for 1 day. The product was transferred from the dialysis tube into a flask and freeze-dried to yield sHA1 as a white powder (47 mg, 96%  Table S2. Amine-functionalized sHA derivatives were initially reacted with 5 eq XLINK in 10 mM sodium phosphate buffer pH 8 at 4 C overnight. After size exclusion centrifugation (7,500 g, 15 min) with a cut-off at 10 and 30 kDa, respectively, to separate unreacted crosslinker, the compounds were subsequently coupled with 5 eq biotinylated (Bio)-MP(Ala) by Michael addition reaction as described above. Excess Bio-MP(Ala) was again removed by size exclusion chromatography. The orthogonality of the reaction was monitored on a Varian Cary® 50 UV-Vis Spectrophotometer measuring the absorption spectrum between 200 and 300 nm. The amount of peptide was determined via the absorption of the DOPA units at 280 nm and a linear standard curve obtained from MP (r 2 = 0.99). Finally, MP-XLINK-sHA2/3 were purified using affinity chromatography. Monomeric avidin agarose beads were washed and loaded with the reaction mixture as described in the manufacturer's protocol. Finally, peptide-crosslinked GAG were eluted using 2 mM biotin in PBS.

| Peptide immobilization
Serial dilutions of MP(+) and MP(−) in 10 mM Tris buffer pH 7.6 were incubated in polystyrene 96-well plates overnight at room temperature and shaking. Afterwards, unbound peptide was washed off with tris-buffered saline-tween20 (TBS-T; 50 mM Tris, 150 mM NaCl, 0.1% tween20, pH 7.6). Detection of bound MP was performed via the biotin tag in an enzyme-linked immunosorbent assay (ELISA)-like assay, as described previously (Hassert et al., 2012). The experiment was performed three times in triplicates, and data are shown as mean ± standard error of the mean (sem).
For further investigations, a 1 μM peptide dilution in 10 mM Tris buffer pH 7.6 was incubated on the microtiterplate overnight. After washing with TBS-T, incubation with heparin and/or CXCL12 continued.

| CXCL12 binding
Full-length CXCL12, carrying a C-terminal methionine due to Escherichia coli expression, was produced, refolded, and purified as described previously (Spiller et al., 2019

| Statistical analysis
Statistical analysis was implemented with Prism 5.0 (GraphPad), and significances were determined with one-way ANOVA followed by  Both peptides were found to efficiently immobilize heparin in comparison to unfunctionalized MP, whereby the elongated sequence HBP 2 gave a 2.8-fold higher signal than the single motif HBP. This loading efficiency was also observed when tested at basic conditions, as chronic wounds can have pH values up to 8.9 ( Figure S1; Tsukada, Tokunaga, Iwama, & Mishima, 1979;Wilson, Henry, Quill, & Byrne, 1979).
In comparison with the dual-affinity approach, which combines HBP and heparin for CXCL12 release, heparin-mimetic coatings were synthesized (Figure 2b). These coatings comprise small peptides as well as short and long sHA derivatives. For the generation of sulfated peptides, sba was coupled to lysine side chains of the sequence EGDKS, similar to a procedure reported by Mammadov et al. (Mammadov et al., 2011). This gave stable compounds that retained sulfation even under highly acidic conditions as during deprotection, full cleavage, and preparative HPLC. In contrast, direct incorporation of sulfotyrosine into the growth factor-binding sequence SYDYG proved to be too acid labile, especially when more than one sulfate group was incorporated. Accordingly, these residues were replaced with lysine and the peptide was analogously sulfated using sba  Table S1.
Beyond the synthesis of sulfated peptides, HA derivatives of differing size were tested: a fully synthetic nonasulfated tetrasaccharide (sHA1) and two chemically sHA derivatives of lower (sHA2) and slightly higher molecular weight (sHA3; Figure 2b

| DISCUSSION
Therapeutic chemokine delivery in wound healing can be achieved in different ways. The respective chemokine can be incorporated into hydrogels and be released upon slow degradation of the biomaterial (Baumann et al., 2012). Also, it can be immobilized covalently on a solid support using an enzyme-cleavable linker mediating release by proteolytic attack (Steinhagen et al., 2014). In contrast to that, we have developed a modular peptide system consisting of a surface binding peptide that can be decorated with different cell recruiting moieties using bioorthogonal reactions (Clauder et al., 2019;Pagel et al., 2016). The advantage apart from being modular is that the peptide components are build up by solid-phase peptide synthesis allowing site-selective incorporation of nonnatural amino acids and nonamino acid components as well as full control over identity and homogeneity of the final product using standard RP-HPLC and mass spectrometry. In this study, we aimed to apply this approach in order to create a prospective delivery system for CXCL12 using affinitybased immobilization of the chemokine. Our focus was not only based on the final outcome of CXCL12 release but also on the feasibility of the production of the release system. This led us to the design, on the one hand, of a peptide carrying covalently attached heparin mimicking moieties and, on the other hand, heparin-binding peptides allowing for non-covalent immobilization of heparin followed by chemokine binding. These approaches enabled purification by RP-HPLC and identification by mass spectrometry without the need of any further verification of product homogeneity.
A prerequisite for the stable immobilization of CXCL12-binding coatings is a strong surface anchoring. This can be achieved by a bioinspired approach adapted from mussels, which adhere to virtually all kinds of substrates even in high tide and strong waves. A derived peptide coating was now shown to bind to standard PS 96-well plates, which is convenient for fundamental investigations and high-throughput studies. Binding affinity studies demonstrated a 20-fold lower EC 50 for the DOPA-containing peptide in comparison with the tyrosine-containing control. For PS surfaces, this can be rationalized by a combination of hydrophobic, cation-π, and π-π stacking interactions (Lu et al., 2013). Although π-π stacking is based on interactions between the aromatic groups of the MP and the phenyl groups of the PS, OH-π interactions might additionally strengthen the binding and explain the observed differences between the DOPAand the tyrosine-containing peptide (Wang & Xie, 2010). The positively charged lysine residue provides additional binding properties via cation-π interactions and helps to abstract the hydration layer, enabling a tight binding of DOPA to the surface and limiting its oxidation (Maier, Rapp, Waite, Israelachvili, & Butler, 2015).
We then equipped MP with HBPs. Reported literature suggests that these peptides electrostatically interact with the negatively charged sulfate and carboxyl groups of heparin via the basic amino acids histidine, arginine, and lysine (Cardin & Weintraub, 1989). This was demonstrated for MP-HBP, and it is assumed that elongation of this sequence increases possible interaction sites and therefore retains significantly more heparin on the surface, as shown for MP-HBP 2 . The underlying sequence FHRRIKA is derived from the bone sialoprotein (Rezania & Healy, 1999), whereas others are based on antithrombin III or the heparin-interacting protein. Investigation of consensus sequences identified clusters of one to three basic amino acids with one or two interpolated hydrophobic residues (Cardin & Weintraub, 1989). This matches the natural binding partner heparan sulfate that displays regions of varying charge densities. For optimal GAG binding, basic and hydrophobic residues should likewise alternate (Fromm, Hileman, Caldwell, Weiler, & Linhardt, 1995). In addition, the secondary structure also needs to be taken into consideration, as heparin binding can induce conformational changes and align the basic residues in spatial proximity facing the GAG (Capila & Linhardt, 2002).
Subsequently, CXCL12 binding was confirmed on the peptideheparin-complexes. Notably, the higher heparin-binding capacity of MP-HBP 2 also retained significantly more CXCL12 in comparison with MP-HBP. Tight packing of the small peptides on the surface and high loading with heparin might be an explanation. Surprisingly, both coatings proved to be equal, when studying cell recruitment in a transwell assay. However, chemoattraction works in a broad range of concentrations as cells are able to compare signals reaching front and rear of the membrane, possess the ability to amplify detected differences, and perform background subtraction of uniform receptoroccupancy (Kutscher, Devreotes, & Iglesias, 2004). Another reason for the unexpected indifference between the two coatings could lie within the heparin sequestration. In addition to higher amounts of chemokine, more heparin is simultaneously released from MP-HBP 2 .
Even though GAG and receptor-binding sites are clearly separated in CXCL12 and the protein remains active in a complex (Laguri, Arenzana-Seisdedos, & Lortat-Jacob, 2008;Sweeney, Lortat-Jacob, Priestley, Nakamoto, & Papayannopoulou, 2002), soluble heparin might act as a scavenger for membrane clustering. Cell-surface proteoglycans locally enhance CXCL12 concentration toward the receptor, and free heparin would compete with their interaction, disconnecting the chemokine from the receptor binding site (Kuschert et al., 1999).
In comparison with the modular approach using non-covalently immobilized heparin for cytokine release, direct functionalization of MP with heparin-mimetic compounds was investigated. Herein, only the cytokine, but not the electrostatically bound heparin, would be released into the surrounding tissues over time. Even though heparin is extensively used as an anticoagulant, it is associated with a number of adverse effects (Gervin, 1975). In addition, the pharmaceutical production of heparin still solely relies on animal tissue. The process is highly optimized and cost-effective, but quality control is difficult and the risk of adulteration and contamination remains (Oduah, Linhardt, & Sharfstein, 2016). Potential shortages resulting from a strong dependency on porcine heparin mainly produced in China as well as animal rights and environmental concerns demand alternative manufacturing processes (Fu, Suflita, & Linhardt, 2016). Synthetic analogs might provide a solution and are subject to extensive research. Even though the F I G U R E 5 (a) CXCL12 binding to immobilized heparin-mimetic compounds (HMP), a receptor-derived peptide (RDP), and heparin-loaded heparin-binding peptides (HBP) as determined by ELISA. n ≥ 2, *** represents significance to MP, p ≤ 0.001; ## represents significance to MP-HBP, p ≤ 0.01. (b) Jurkat migration toward mussel-derived peptide coatings. The lower compartments of a transwell plate were coated with peptide (+heparin) and CXCL12, whereas cells were seeded into the upper chambers. After 2 h incubation, the number of migrated cells was evaluated. Medium containing 50 nM CXCL12 served as positive control. n ≥ 3, significance to uncoated, * p ≤ 0.05, ** p ≤ 0.01 [Colour figure can be viewed at wileyonlinelibrary.com] production can be laborious and hard to upscale, they provide better control and higher chemical versatility for the incorporation into functional materials. Therefore, we considered synthetic peptides as well as HA derivatives as heparin mimetics for the covalent coupling to MP. Previous reports had demonstrated that anionic charges solely derived from hydroxyl and carboxyl groups are insufficient to retain CXCL12, as demonstrated for HA in an NMR study (Panitz et al., 2016). Consequentially, sulfated derivatives were produced.
However, HMP coatings were generally less efficient in CXCL12 binding. None of the sulfated peptides were able to accumulate CXCL12, neither as single nor as double or triple motif, respectively.
This was unexpected, as the HMPs in principle display the three functional groups of heparin-sulfates, carboxylates, and hydroxyl groups and were confirmed to bind various cytokines in preceding studies (Hendrikse et al., 2018;Mammadov et al., 2012;Maynard & Hubbell, 2005). However, in these approaches, the peptides are presented in a three-dimensional arrangement on self-assembled nanofibers in contrast to our two-dimensional surface modification.
For the hyaluronan-based coatings, a length-dependent effect is observed. The short tetrasaccharide in MP-sHA1 was insufficient for immobilizing CXCL12, whereas the longer variants retained at least small amounts of the protein. In principle, the chemokine CXCL12 dimerizes upon heparin binding (Fermas et al., 2008) and subsequently displays two distinct GAG-binding sites. The main binding site, so called high-affinity heparin-binding region, is located at the dimer interface, where a number of basic residues create a positively charged crevice (Sadir, Baleux, Grosdidier, Imberty, & Lortat-Jacob, 2001). Residues of the consensus sequence BBXB in the first β-strand (Lys24, His25, and Lys27; Amara et al., 1999) as well as spatially adjacent Arg41 and Lys43 (Sadir et al., 2001) were found to play a major role. The second binding site, called low-affinity heparinbinding region, includes the N-terminal loop and the α-helix.
Important residues involved in this interaction include Arg20, Ala21, Asn30, and Lys64 (Murphy et al., 2007). To span the high-affinity heparin binding region, 12 to 14 monosaccharide units are required (Sadir et al., 2001). The small sulfopeptides and the tetrasaccharide are probably too short to efficiently interact with CXCL12. However, sequence elongation of the sulfated peptides did not enhance the protein-binding capacity, which suggests that the defined pattern of carboxyl and sulfate groups in heparan sulfate is not met correctly.
Studies show that especially 2-O-and N-sulfate groups contribute to CXCL12 binding (Sadir et al., 2001) as well as 6-O-sulfation to a lesser extent (Zhang et al., 2012). Also, L-iduronic acid provides conformational flexibility that is believed to be responsible for the variety of proteins that can be bound specifically (Mulloy & Forster, 2000). For the HMPs, the sulfate groups might not be displayed in the correct conformation with respect to orientation, spacing, and distance from the peptide backbone, as a rather long linker was introduced by using sba. Although heparan sulfate is only moderately sulfated with an average of one sulfate group per disaccharide, heparin is highly sulfated with an average of 2.7 sulfate groups per disaccharide. In comparison, the degree of sulfation of the tetrasaccharide is still much higher, with 4.5 sulfate groups per disaccharide, which could also be a disadvantage for the interaction. On the other hand, the difference to heparin is not as big for the higher molecular weight HA derivatives, which display 3.5 (sHA2) and 3.4 sulfate groups (sHA3), respectively. However, N-sulfation is completely absent in all of these compounds. When comparing the glycosidic linkage, the monomers of the HA derivatives are (1 ! 3)-linked and in the disaccharides (1 ! 4), whereas in heparin and heparan sulfates, both units are (1 ! 4)-linked. This additionally contributes to a discrepant arrangement of the sulfation pattern.
Taken together, the two-step coating protocol using HBPs and electrostatically bound heparin was most efficient in immobilizing the chemokine CXCL12 and recruiting cells to coated surfaces. This was rationalized by a high binding affinity toward the cytokine, owing to an unaltered charge display of the incorporated natural heparin. In addition, small HBP coatings can be prepared in a relatively easy fashion using solid-phase peptide synthesis and bioorthogonal reactions followed by simple RP-HPLC purification. As seen for sHA2 and sHA3, using larger GAGs for covalent immobilization requires more sophisticated purification protocols, and there still remains uncertainty on the homogeneity and consistent functionality of the final product. Also, HBPs reach high surface packings and are uncomplicated and more stable to pH changes during synthesis. By alteration of the HBP, the amount of heparin and heparin-binding cytokines can be tuned, catering to therapeutic purposes.

| CONCLUSION
In our study, a two-layer approach, where immobilized HBPs are subsequently loaded with heparin, was identified to be the best condition for a prospective chemokine delivery system, as the highest amount of CXCL12 was bound and the strongest effects on cell migration were detected. However, maintaining the balance between chemokine and codelivered heparin is crucial to success and multiple factors involving scavenging effects, side reactivity, and supply ethics need to be taken into consideration. Even though the synthetic compounds tested in this study could not compete with the natural product, extensive progress has been made in carbohydrate chemistry within the past years, and the development of compounds displaying the appropriate interaction sites for chemokine binding are subject to the near future (Mohamed & Coombe, 2017). This will have many benefits, as synthetic compounds provide higher structural control and allow fine-tuning of the protein binding affinity. One might imagine the creation of a more stable gradient by modulating the strength of the affinity interactions (Vulic & Shoichet, 2014), either the GAG component with higher or lower protein-binding affinities or engineered proteins displaying a stronger or weaker GAG-binding affinity (Spiller et al., 2019). In both cases, mussel-derived surfacebinding peptides provide a versatile platform that is additionally modifiable with other bioactive peptides. Combinations of cell adhesion and HBPs can cooperatively enhance tissue regeneration (Pagel et al., 2016) and provide a huge toolbox customizable to the specific needs of future applications.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article. Table S1. Analytical data of the synthesized precursor peptides and conjugates. Figure S1. Heparin binding to immobilized MP-HBP and MP-HBP 2 as investigated in a biotin ELISA-like assay at different pH conditions. n = 2, * p ≤ 0.05, ns = non significant.