Engineered dual affinity protein fragments to bind collagen and capture growth factors

Abstract

Collagen type I lacks affinity for growth factors (GFs) and yet it is clinically used to deliver bone morphogenic protein 2 (BMP-2), a potent osteogenic growth factor. To mitigate this lack of affinity, supra-physiological concentrations of BMP-2 are loaded in collagen sponges leading to uncontrolled BMP-2 leakage out of the material. This has led to important adverse side effects such as carcinogenesis. Here, we design recombinant dual affinity protein fragments, produced in E. Coli, which contain two regions, one that spontaneously binds to collagen and a second one that binds BMP-2. By adding the fragment to collagen sponges, BMP-2 is sequestered enabling solid phase presentation of BMP-2. We demonstrate osteogenesis in vivo with ultra-low doses of BMP-2. Our protein technology enhances the biological activity of collagen without using complex chemistries or changing the manufacturing of the base material and so opens a pathway to clinical translation.

Keywords: Bacteria; Biomaterials; Bone regeneration; Collagen; Fibronectin; Human mesenchymal stem cells; Recombinant protein fragment.

Graphical abstract

Fig. 1

A) Outlines the proposed system of using a protein fragment to create a link between a collagen substrate and GFs such as rhBMP-2. To achieve this a collagen-binding domain (CBD) is linked to a growth factor binding domain (GFBD) (Scale bar 1 ​μm). B) The CBDs that were explored were from collagenase G (S3a-S3b), collagenase H (S3) and PLGF (PLGF(123–144)). The GFBD explored was the FNIII(12–14) domain as it is easy to produce in bacteria and has been extensively studied. C) Shows how the protein fragments with varying CBDs (namely the collagenase G-FNIII12-14 (ColG-FNIII12-14), collagenase H–FNIII12-14 (ColH-FNIII12-14), PLGF-FNIII12-14) performed when collagen substrates were treated with the respective protein fragment. The protein fragment with collagenase G as the CBD was retained the most by the collagen substrate and was chosen for further evaluation. D) Different concentrations of the ColG-FNIII12-14protein fragment were retained similarly in terms of percentage (%) by the collagen scaffold but the mass amount of the protein fragment retained was higher the higher the concentration of ColG-FNIII12-14 initially loaded. E) An immunostaining image of the collagen scaffold with and without the protein fragment can be seen which clearly shows how the protein fragment decorates the collagen scaffold (Scale bar 50 ​μm). Statistical significance (∗p ​< ​0.05).

Fig. 2

A) Shows the chemistry used to crosslink the collagen scaffolds before they are treated with the protein fragment. B) Shows the mechanical properties, pore size and swelling characteristics of the collagen scaffolds. Scaffolds were 2.5±0.47 ​kPa in wet conditions and 74.25±30 ​kPa in dry conditions, the pore size ranged from 57 ​μm to 407 ​μm with a mean of 198 ​μm, total scaffold swelling (swelling A) was 2743±300 (%), fiber swelling (swelling B) 286±30 (%) and the water in pores was measured to be 85±4 (%). The mechanical properties are consistent with collagen scaffolds in literature, the pore size is sufficient for osteogenesis, and the swelling suggests great scaffold wettability and water retention important for cell culture. C) The viability of hMSC on collagen scaffolds was found to be greater than 99% (Scale bar 100 ​μm) even after the addition of the protein fragment and D) cells were spindle-like and spread along the pores of the scaffolds (Scale bar top: 100 ​μm bottom: 50 ​μm). E) Vinculin staining revealed that hMSCs formed focal adhesions when interacting with collagen (right image) similar to the ones seen on glass (left image) suggesting integrin-mediated interaction with the substrate despite the addition of the protein fragment (Scale bar 50 ​μm).

Fig. 3

A) Shows the % rhBMP-2 (1 ​μg ​mL⁻¹) absorbed on collagen scaffolds treated with different concentrations of ColG-FNIII12-14 protein fragment and compared to rhBMP-2 absorption onto non-protein fragment treated collagen scaffolds. The release of the rhBMP-2 absorbed onto the scaffolds was also monitored for a 120 ​h period. The data suggests that there is significantly higher rhBMP-2 absorption and retention in the conditions where the collagen scaffolds were treated with protein fragments compared to the no protein fragment condition. There was no significant difference between using different amounts of protein fragment. B) Keeping the amount of protein fragment the same (250 ​μg ​mL⁻¹) and varying the amount of rhBMP-2 from 0.5 to 2 ​μg ​mL⁻¹ also did not affect the percentage absorption and release but the absolute amount in μg of protein fragment absorbed and retained did increase with increasing concentrations. C) An image visualising rhBMP-2 was taken after the conclusion of the absorption-release experiments and the protein fragment treated collagen substrate was compared to the no treatment condition. It can be clearly seen that the substrate with the protein fragment retained more BMP-2 compared to the collagen scaffold without protein fragment (Scale bar 100 ​μm). D) The interaction strength between the protein fragment and rhBMP-2 was also assessed using MST. The Kd of the interaction was found to be around 604 ​nM which suggest a medium high affinity for rhBMP-2 and the protein fragment. Statistical significance (∗∗p ​< ​0.01).

Fig. 4

Shows hMSC osteogenic differentiation on a collagen scaffold (neg. control), a collagen scaffold loaded with 2 ​μg ​mL⁻¹ of rhBMP-2 (rhBMP-2 (no protein fragment)), a protein fragment treated collagen scaffold loaded with 2 ​μg ​mL⁻¹ of rhBMP-2 (protein fragment ​+ ​rhBMP-2) and collagen scaffold cultured in osteogenic media (pos. control) A) Runx2 translocation to the nucleus was investigated after 6 days using immunostaining and measuring the mean intensity of Runx2 staining within the nucleus. The protein fragment treated collagen scaffolds with rhBMP-2 showed significantly higher Runx2 translocation to the nucleus compared to the neg. control and the collagen scaffold with rhBMP-2 but no protein fragment. The protein fragment ​+ ​rhBMP-2 condition was not statistically different to the osteogenic media condition which also showed increased translocation to the nucleus (Scale bar 50 ​μm). B) The same trend was observed when culturing cells for 21 days and staining for osteopontin and C) mineral deposition. The protein fragment ​+ ​rhBMP-2 condition performed most similar to the osteogenic media but better than both the neg. control and the rhBMP-2 (no protein fragment) conditions (Scale bar 100 ​μm). Statistical significance (∗∗∗p ​< ​0.001).

Fig. 5

Protein fragment-functionalized ACSs promote new bone growth in vivo. A) Picture of the implant tube filled will a ACS and an X-ray image of the critical size (2.5 ​mm) murine radial defect model used. B) SEM images of the ACSs implanted. C) The new bone volume (NBV) and BV/TV (NBV/Total Volume) formation was quantified in the positive control (ACS with 75 ​μg ​mL⁻¹ rhBMP-2 with no added ColG-FNIII12-14 protein fragment), 250 ​μg ​mL⁻¹ ColG-FNIII12-14 ​+ ​75 ​μg ​mL⁻¹ rhBMP-2, 250 ​μg ​mL⁻¹ ColG-FNIII12-14 ​+ ​5 ​μg ​mL⁻¹ rhBMP-2, 250 ​μg ​mL⁻¹ ColG-FNIII12-14 ​+ ​2.5 ​μg ​mL⁻¹ rhBMP-2, and plain ACS (negative control) condition (mean±SEM, n ​= ​6, p ​< ​0.05). D) 3D reconstruction of micro-computerized tomography (μCT) scans of the area where the implant was placed in the critical size murine radial defect. The samples closest representing the mean are depicted. E) Histological images of the same samples depicted in (D) showing hematoxylin and eosin staining. Black arrows point to fibrotic tissue, white arrows point to the formation of a bone marrow cavity/cancellous bone, while maroon arrows point to new cortical bone. Statistical significance (∗p ​< ​0.05).