Improved Bone Morphogenetic Protein-2 Retention in an Injectable Collagen Matrix Using Bifunctional Peptides

To promote healing of many orthopedic injuries, tissue engineering approaches are being developed that combine growth factors such as Bone Morphogenetic Proteins (BMP) with biomaterial carriers. Although these technologies have shown great promise, they still face limitations. We describe a generalized approach to create target-specific modular peptides that bind growth factors to implantable biomaterials. These bifunctional peptide coatings provide a novel way to modulate biology on the surface of an implant. Using phage display techniques, we have identified peptides that bind with high affinity to BMP-2. The peptides that bind to BMP-2 fall into two different sequence clusters. The first cluster of peptide sequences contains the motif W-X-X-F-X-X-L (where X can be any amino acid) and the second cluster contains the motif F-P-L-K-G. We have synthesized bifunctional peptide linkers that contain BMP-2 and collagen-binding domains. Using a rat ectopic bone formation model, we have injected rhBMP-2 into a collagen matrix with or without a bifunctional BMP-2: collagen peptide (BC-1). The presence of BC-1 significantly increased osteogenic cellular activity, the area of bone formed, and bone maturity at the site of injection. Our results suggest that bifunctional peptides that can simultaneously bind to a growth factor and an implantable biomaterial can be used to control the delivery and release of growth factors at the site of implantation.


Introduction
Approximately 7.9 million fractures occur each year in the United States alone, and approximately 10% of fractures exhibit delayed or impaired healing [1]. Bone morphogenetic proteins (BMPs) are osteogenic growth factors that have been shown to stimulate new bone formation and fracture healing [2,3]. In clinical trials, recombinant human BMP-2 (rhBMP-2) has been shown to accelerate healing of open tibial fractures [4], and rhBMP-7 has been used to treat tibial nonunions [5]. These clinical applications, however, require open surgical procedures to insert the BMP-loaded carrier. In addition, supraphysiological amounts of BMPs are required to promote bone formation due to the growth factor's rapid diffusion away from its carrier [6,7]. The use of high doses, however, raises concerns about bone formation away from the site and impact on nearby tissues and organs [8]; in accordance, rhBMP-2 use has been linked to a variety of serious adverse events [9].
Ideally, an injectable BMP-2 matrix carrier should have the following features: strong affinity for BMP to maintain biologically relevant concentrations over time to encourage osteoprogenitor cell migration, proliferation and differentiation; biocompatibility to minimize inflammation; sufficient porosity to allow cellular invasion and attachment; resorbability so that it will be replaced with new bone during healing; and appropriate viscosity for passage through a syringe without being washed away from the site of injection [10,11,12]. The carriers that have been explored for delivery of BMP include naturally derived polymers such as collagen, hyaluronic acid, chitosan, and fibrin; synthetic polymers such as polylactic acid (PLA), polyglycolic acid (PGA) and their copolymers (PLGA); ceramic materials including calcium phosphate cements; and various combinations of these materials [13].
For injectable BMP carriers, tested matrices include hyaluronan gels, gelatin (collagen) foams, composites of the gels and foams with tricalcium phosphate, and calcium phosphate cement [12]. Most of these injectable BMP carriers were unable to retain BMP at the site of injection; the carriers lost 50% or more of pre-loaded BMP after a few days in vivo. Calcium phosphate cement-based formulations were able to retain measurable amounts of BMP at the site of injection for 14 days [12]. These formulations, however, relied on entrapment of the BMP within the cement, raising concerns about a lack of release of BMP to the surrounding site of healing. We propose that the use of engineered peptides with high affinities for both BMP and the matrix material joined as a bifunctional peptide may provide more controlled release of BMP from the matrix and promote optimal healing in vivo.
Bifunctional peptides have been constructed by a number of groups as a way to modify materials and promote cell adhesion or growth factor binding. Gama and his colleagues [14,15] have fused cell adhesion domains such as RGD and IKVAV to a cellulose-binding domain. The 25 kD fusion proteins promote fibroblast or mesenchymal stem cell adhesion to bacterial cellulose. Murphy and colleagues have also reported the successful use of modular peptides to deliver growth factors and mesenchymal stem cells to hydroxyapatite coatings [16,17,18].
In this study, we have used phage display technology to identify a series of peptides which bind to BMPs. The BMP-binding peptides can be organized into two sequence motifs. Consensus peptides from each motif were coupled to a collagen-binding peptide to form bifunctional peptides that can bind simultaneously to BMP-2 and collagen. The bifunctional peptide therefore can bind and retain BMP-2 onto a collagen matrix, slowing the release of BMP-2 from the matrix. We tested the ability of the bifunctional peptide to improve bone formation in vivo using a rat ectopic bone formation bioassay.

Ethics statement
All procedures with animals were performed under protocols approved by Affinergy's Institutional Animal Care and Use Committee in a facility with assurance from the Office of Laboratory Animal Welfare (A4544-01).
Phage Display rhBMP-2 was biotinylated using a Sulfo-NHS-Biotin reagent (Pierce EZ-link biotinylation kit) following the manufacturer's protocol. The biotinylated rhBMP-2 was immobilized onto a streptavidin-coated 96-well microtiter plate (Immulon IV) and the plates blocked with 0.5% BSA in phosphate buffered saline, 0.05% Tween-20 (PBST). Phage display was performed as previously described [19,20]. Ten different phage display libraries were screened for peptides that bind to rhBMP-2. Each library was designed around a specific amino acid motif or amino acid bias. After 3 rounds of phage display selections, the pools of enriched phage were plated on a lawn of E. coli DH5aF' cells. Individual phage were picked and propagated on E. coli overnight. The cells were removed by centrifugation, and 10 ml of the phagecontaining supernatant was added to the wells containing BMP-2 or to a control well containing buffer. After incubation and washing, phage were detected in an ELISA-type assay using an HRP-conjugated, anti-M13 monoclonal antibody (1:1000).

Generating Focused Library for BMP-2 binding peptides
To generate the Motif 1-focused and Motif 2-focused libraries, oligonucleotides were synthesized to encode peptides that have a restricted set of possible amino acids in selected positions in the peptide ( Table 1). A short complementary primer was annealed to the 39 end of the library oligonucleotide and extended with T7 polymerase. After second strand synthesis, the DNA was digested with Xba I and Xho I and ligated into mAEK phage display vector. The ligation reaction was used to transform electrocompetent E. coli DH12S. Transformed cells were grown overnight in 26YT medium. The phage-containing culture supernatant was collected and the phage concentrated by PEG precipitation. Precipitated phage were resuspended in phosphate buffered saline containing 20% glycerol, aliquoted and stored at 220uC.

DNA sequence analysis
DNA from isolated positive phage clones was amplified using the TempliPhi DNA amplification kit and DNA analysis was performed by Sequetech (Mountain View, CA).

Peptide Synthesis
Peptides were synthesized by solid-phase peptide synthesis techniques on a Rainin Symphony Peptide Synthesizer using standard Fmoc chemistry (HBTU/HOBT activation, 20% piperidine in DMF for Fmoc removal). After all residues were coupled, simultaneous cleavage and side chain deprotection was achieved by treatment with a trifluoroacetic acid (TFA) cocktail. Crude peptide was precipitated with cold diethyl ether and purified by high-performance liquid chromatography on a Shimadzu HPLC using a Vydac C18 reversed-phase silica column (10 mm, 120 Å , 250 mm 622 mm) using a linear gradient of water/acetonitrile containing 0.1% TFA. Homogeneity of the synthetic peptides was evaluated by analytical RP-HPLC (Vydac C18 silica column, 10 mm, 120 Å , 250 mm 64.6 mm) and the identity of the peptides Table 1. Focused library design for motif 1 and motif 2 BMP-binding peptides*. was determined by MALDI-TOF-MS. All peptides were synthesized with a biotin coupled to the epsilon-amino group of the Cterminal lysine.

Bifunctional Peptide Design
A consensus BMP-binding peptide was generated for each BMP-binding motif. Peptides that bind to collagen were isolated by phage display on demineralized bone matrix [21]. Bifunctional peptides (Collagen: BMP) were generated by joining a collagen-

BMP-2 peptide binding assay
To measure the ability of a BMP-binding peptide to capture its target growth factor out of buffer or a complex biological solution such as plasma, biotinylated peptides were immobilized onto a streptavidin-coated 96 well plate. Varying low nanomolar concentrations of BMP-2 were ''spiked'' into human plasma or Tris-buffered saline (0.5 M NaCl), 0.05% Tween-20 (TBST) and added to the peptide-containing plates. After 1 hr incubation at room temperature (RT), plates were washed and incubated with an anti-BMP-2 antibody followed by a goat anti-mouse secondary antibody conjugated to alkaline phosphatase. Adding p-NPP produced a colored reactant, which was quantified using a SpectroMax (Molecular Devices) plate reader at 405 nm.

Preparation of a 4% fibrillar collagen gel
Collagen stock solution (6.4 mg/ml, Inamed cat#5413, lot1387646) was neutralized overnight at room temperature using 200 mM sodium phosphate, pH 9.4. The fibrillar collagen was pelleted by centrifugation at 17,2006g for 20 minutes at 10uC. The collagen weight percent was determined using a bicinchoninic acid protein assay (BCA assay, Pierce cat#23255).

Preparation of 1.5% injectable collagen gel for BMP-2 delivery in vivo
The stock fibrillar collagen (4%) was diluted in PBS containing rhBMP-2 with or without the collagen-BMP bifunctional peptide, to generate a final collagen wt% of 1.5%. The amount of BMP-2 and peptide were optimized using pilot experiments in vivo such that 200 ml of the injected collagen gel contained 2 mg of BMP-2 and a 50-fold molar excess of the peptide.

Injectable collagen gel binding assay
The injectable collagen gel (0.1 mL) was aliquoted into a polypropylene plate. The plate was blocked with 150 ml of 1% BSA in TBS for 30 min. After spinning for 2 min, the supernatant was removed without disrupting the collagen gel. The bifunctional peptide (30 mM) was mixed with rhBMP-2 (0.0017 to100 nM) in Binding Buffer (125 mM glutamic acid, 10% sucrose, 12.5% glycine, 10% polysorbate 80, 5 M NaCl) for 30 min with gentle agitation. The peptide: rhBMP-2 complex (50 mL) was then added to the collagen gel and incubated at room temperature for 1 h with gentle agitation. The collagen gel was washed three times with TBST and rhBMP-2 was detected with an anti-BMP antibody (R&D Systems MAB3552). The wells were then washed, and an alkaline phosphatase labeled goat anti-mouse secondary antibody was added. After incubation and washing, binding was measured using the chromogenic reagent p-NPP, and absorbance was read at 405 nm.  In vivo ectopic bone formation study Male Sprague Dawley rats (150-175 g) were obtained from Taconic farms, Charles River Laboratories (Raleigh, NC) and acclimated prior to surgery. The surgery was performed under general anesthesia by weight-adapted intraperitoneal injection of Xylazine 2% (Medistar; 12 mg/kg body weight) and Ketamine hydrochloride (Ketaset; 100 mg/mL; 80 mg/kg body weight). The thorax and abdomen were shaved and scrubbed with Betadine and alcohol. Using aseptic technique, two 5 mm incisions were made at the midline so that two subcutaneous pouches were prepared by blunt dissection. Collagen gel (200 ml) containing 2 mg rhBMP-2 with or without collagen-BMP bifunctional peptide (BC-1) was injected into the left or the right side of the subcutaneous region. Each rat received two injections and a total of 20 animals were used in the study. The BMP-2 dose (2 mg) used in this experiment had been determined from preliminary experiments (data not shown). In those experiments, rats were implanted with collagen gel containing either 1 or 5 mg of BMP-2. Collagen gel without peptide resulted in bone formation at the 5 mg dose but not at the 1 mg dose. Collagen gel with peptide resulted in bone formation at both doses (1 and 5 mg), and a dose of 2 mg BMP-2 was selected for the experiment.
The sites containing the injected material were explanted at two weeks and fixed in 10% neutral buffered formalin. The samples were placed in Formical 2000 decalcifier (American Mastertech) for 24 hours followed by Cal-arrest neutralizing solution for one hour. The tissue samples were processed in a Thermo-Electron, Shandon Excelsior Automated Tissue Processor for 14 hours. The process includes additional fixation in 10% buffered neutral formalin, dehydration through an ethanol gradient, and clearing with xylene at 40uC. Tissue was then embedded on a Leica EF 1140 H tissue embedding center in paraffin blocks for sectioning. Paraffin blocks were sectioned at 5 microns on a Reichert-Jung RM2065 microtome using Accu-Edge High Profile disposable stainless steel microtome blades. Three serial sections at intervals of 30 microns were obtained and placed on Super-frost Plus slides for hematoxylin and eosin staining (H&E). Staining was performed using a Sakura DRS-601 automatic slide stainer with a regressive Harris hematoxylin and Eosin Y Alcoholic for histological and morphological examination.

Isolation of BMP-binding peptides using phage display peptide libraries
Phage display was performed against immobilized rhBMP-2 with 10 different peptide libraries, representing over 10 billion peptide sequences. Biopanning with these libraries revealed a set of 16 peptides that bound to BMP-2 ( Figure 1). These firstgeneration BMP-binding peptides fall into two different sequence clusters. The first cluster of peptide sequences contains the motif W-X-X-F-X-X-L (single letter amino acid code, where X can be any amino acid), designated motif 1, and the second cluster contains the motif F-P-L-K-G, designated motif 2. A series of truncations in which conserved amino acids were deleted in the BMP-binding peptides results in a loss of BMP-2 binding activity (data not shown). The consensus sequence among the peptides indicates that all the peptides within a sequence cluster are binding to the same site on BMP-2, and the common amino acids that make up the motif are responsible for the specific interactions with BMP-2.
Using these sequence motifs, we designed a focused phage display library around each motif ( Table 1). Each of the focused libraries was screened for additional peptide sequences that would bind to BMP-2 using standard phage display techniques [22]. Combined, 59 peptide sequences were found positive for BMP-2   binding. Of these 59, 41 sequences were found around motif #1 ( Figure 2) and 18 represented motif #2 (Figure 3). After aligning the peptides, the frequency of amino acid identity was scored at each position (Tables 3 & 4). Positions that have only one, two or three amino acid identities represented at a given position are likely to promote BMP-binding, therefore defining a new feature for the binding motif. From the alignment, a consensus sequence was established for the binding domain revealed by each focused library. Synthetic peptides were synthesized based on each consensus sequence: B-17 is the consensus for motif #1 and B-18 is the consensus for motif #2 ( Table 5).

Relative affinity of BMP-2 binding peptides for BMP-2
To compare binding affinities between first and second generation BMP-binding peptides, we tested a range of BMP-2 concentrations (0.001 to 10 pmoles/well) for binding against immobilized peptide on streptavidin coated plates (Figure 4). The consensus peptides B-17 and B-18 bound BMP-2 with high apparent affinity with EC50 values of 1.4 and 1.9 nM, respectively. Peptide B-6 also had a low EC50 value of 1.4 nM but was not used in subsequent experiments.

Cross-reactivity of BMP-2-binding peptides with other growth factors
BMPs are members of the TGF-b superfamily and so we examined our peptides' abilities to bind other family member proteins. The two BMP-binding peptides that contain the consensus motif #1 (B-17) and consensus motif #2 (B-18) were tested for binding to BMP-2, -3, -4, -5, -6, -7, -9, -12, -14, TGF-b1, TGF-b3, and PDGF-BB. Both peptides bound to BMP-2, BMP-6 and BMP-7 but showed no binding to TGF-b1, TGF-b3, or PDGF-BB ( Figure 5; EC50, Peptide B-17: 1.4 nM, BMP-2; 1.3 nM, BMP-6; 17.0 nM, BMP-7. EC50, Peptide B-18: 2.8 nM, BMP-2; 6.5 nM, BMP-6; 16.2 nM, BMP-7). The lack of specific binding for TGF-b1, TGF-b3 or PDGF-BB, suggests that the peptides intended for BMP-2 binding harbor a specific interaction with a sequence or structural motif found in several BMPs but not in other growth factors. The two peptides, B-17 and B-18, however, do not show identical specificity among the BMP     Table 6. Histology scores (average from two observers blinded to the study) of all samples with rhBMP-2 delivered in a collagen gel with or without the bifunctional peptide. Capture of BMP-2 from a complex biological fluid To model endogenous capture of a specific protein in vitro, picomole amounts of rhBMP-2 (0.001 to 15 pmoles/well) were mixed with TBST or human plasma for 1 h. After washing the plates in TBST, an alkaline phosphatase-conjugated antibody against BMP-2 was applied, and bound BMP-2 was measured with a colorimetric assay on a plate reader. Incubating BMP-2-binding peptides in human plasma had no effect on the peptide's ability to bind and capture BMP-2 ( Figure 6). Binding reactions in TBST controls exhibited virtually identical affinity for the target growth factor relative to the biological fluids. These data show that the peptides are not ''fouled'' by other proteins in biological fluids, suggesting that bifunctional peptide-mediated capture and concentration of endogenously produced growth factors could also be a viable therapeutic strategy.

Bifunctional Peptides
Bifunctional peptides consisting of the BMP-binding sequences and the collagen-binding sequences were synthesized and tested for the ability to bind BMP-2 to a collagen gel. Both BC-1 and BC-2 bound BMP-2, but to minimize the size and cost of the animal study, only BC-1 was used in the in vivo model. BC-1 was chosen because the BMP-binding sequence in BC-1, the B-17 sequence, bound a wider range of BMPs than the B-18 sequence which might be useful in future experiments. BMP-2 was incubated with the bifunctional peptide BC-1 and then added to plates containing the injectable collagen gel. BMP-2 is known to bind collagen weakly [23], but the bifunctional peptide increased binding of BMP-2 to the collagen gel more than 10-fold (no peptide: EC50 = 5.5 nM; bifunctional peptide: EC50 = 0.41 nM; Figure 7).

In vivo model -Ectopic Bone Formation
Ectopic bone formation at two weeks was evaluated by histological scoring of the H&E stained slides by two observers. The slides were scored for infiltration of osteogenic cells (fibroblasts, osteoblasts, osteogenic progenitors and cells of the cartilage), each of which were scored on a scale of 1-4: 1-rare, 2few, 3-moderate and 4-dense. Osteogenic cellular activity was obtained by an average of the scores for each cell type for every animal. Comparison between groups was performed using medians. As shown in Figure 8, the median cellular activity of the peptide group was significantly greater than the no peptide (BMP-2 alone) group (Mann-Whitney U two tailed test, p,0.0001). Bone area was scored as percent of the implant showing new bone formation on a scale of 0-4. The bone area was significantly greater in the peptide group compared to the no peptide (BMP-2 alone) group (Mann-Whitney U two tailed test, p,0.0001). In the peptide group, approximately 25% of the implant was covered with new bone, whereas no new bone was detected in the no peptide group. The maturity of the newly formed bone was scored on a scale of 0-4, with 0-no bone, 1immature/unorganized, 2-immature, 3-mature and 4-mature/ well-organized. In the absence of peptide very little and immature/unorganized bone was formed. Bone was formed with rhBMP-2 in the presence of the peptide and it was characterized as mature. Representative images obtained at a 26 magnification are shown in Figure 9 and the histology scores (average of two observers blinded to the study) of all samples are presented in Table 6.

Discussion
Using phage display technology and biopanning on rhBMP-2, we have identified a set of peptides that bind to several forms of BMP. The peptides can be organized into two groups based on common elements in their peptide sequences. The first group shares the sequence motif W-X-X-F-X-X-L (motif 1) and the second group shares the sequence motif F-P-L-K-G (motif 2). The forty sequences that comprise motif 1 all bind to BMP at a common site and the conserved amino acids in motif 1 are largely responsible for the binding of those peptides to BMP. Similarly, the eighteen peptides that group into motif 2 bind at a common site on BMP and the binding is determined by the conserved amino acids in motif 2. Aligning all the peptides in motif 1 or motif 2 and using the most frequently identified amino acid in each position of the peptide allowed a consensus binding sequence to be generated for each motif. The consensus peptides B-17 and B-18 showed strong binding to BMP-2 with EC50 values around 1-2 nM. Although neither of these sequences was isolated in the phage display selection on BMP-2, the aggregation of the peptide sequences from the selections led to the design of peptides B-17 and B-18, which have high affinity for the target protein.
To use the BMP-binding peptides in a drug delivery system, we linked the consensus BMP-binding peptide (B-17) to a previously identified collagen-binding peptide through a short flexible four amino acid linker. The bifunctional peptide maintained the activity of each peptide domain: collagen binding and BMP-2 binding. In vitro, the bifunctional peptide BC-1 was able to increase binding of BMP-2 to a collagen gel by 10-fold. In vivo, ectopic bone formation was analyzed in rats using an injectable 1.5% collagen gel containing 2 mg of rhBMP-2 with or without BC-1. The concentration of injected collagen gel (1.5% w/w), BMP-2 dose (2 mg), peptide to BMP-2 molar ratio (50:1) and study duration (2 weeks) were optimized from preliminary experiments (data not shown). The presence of BC-1 significantly increased osteogenic cellular activity, the area of bone formed, and bone maturity.
To increase osteogenic activity of cells, BMP-2 has to interact with the BMP receptors on the cells. From our experiments, it is not possible to distinguish whether BMP-2 can bind the receptor while still bound to the peptide or if the BMP-2 is released from the peptide and then interacts with the BMP receptor. In cellbased, BMP-activity assays, when B-17 is added in excess to BMP-2 there is no inhibition of activity (data not shown). BC-1 increases the retention of BMP-2 in the collagen matrix and leads to increases in osteogenic activity and bone formation but whether the BMP-2 is free or peptide-bound is not clear at this point.
One of the challenges of local delivery of growth factors is the retention of the growth factor at the site of healing for a sufficient time to allow the growth factor to have a positive effect. An FDAapproved BMP-delivery system uses a collagen matrix and requires supra-physiologic doses of BMP to achieve the desired therapeutic effect which increases concerns about safety and cost [24]. Collagen matrices have many desirable features in growth factor delivery systems including biocompatibility, degradation to natural products and favorable interactions with cells; but when used with BMPs, collagen matrices rapidly lose BMP after implantation [23]. Using an absorbable collagen sponge (ACS), 30% of the loaded BMP-2 is lost almost immediately and 50% is lost in the first two days after implantation [23]. Incorporation of a bifunctional collagen: growth factor-binding peptide as described here could greatly enhance the performance of collagen-based drug delivery systems. In addition, the ability of these BMP-2 binding peptides to bind other members of the BMP family, such as BMP-7, extends their utility to other BMPs.