Introduction

Silk fibroin fibers produced from the silk of the mulberry silkworm Bombyx mori has outstanding mechanical properties, including high strength and toughness, despite being spun from an aqueous solution at ambient temperature. A fibroin molecule consists of a heavy chain of 391 kDa and a light chain of 26 kDa connected by a disulfide bond.1, 2, 3, 4 The primary structure of B. mori fibroin has an unusual repeat sequence in the heavy chain, as shown in Figure 1. This structure comprises a highly repetitive Gly-Ala-Gly-Ala-Gly-Ser (GAGAGS) sequence, relatively fewer repetitive sequences of hydrophobic and/or aromatic residues and amorphous regions containing negatively charged, polar, bulky hydrophobic and/or aromatic residues.

Figure 1
figure 1

Schematic of the organization of 12 repetitive domains and 11 amorphous repeat motifs in the B. mori silk fibroin heavy chain primary structure. The repetitive domains comprise 1–8 repeats of repetitive unit R, which consists of three repeated elements of sequence motifs. A full color version of this figure is available at the Polymer Journal online.

Full size image

Silk proteins are secreted and stored in silk glands before they are processed into fibers. Silkworms produce fibroin fibers from an aqueous fibroin solution by applying shear stress within the spinneret and applying tension, which is induced by the repeated drawing back of the silkworm’s head, at ambient temperature. As such, silk fiber processing is an attractive model for developing sustainable fiber processing technology. However, a complete understanding of the fibroin processing mechanism requires the determination of fibroin structures before and after spinning.

Herein I report the structural nuclear magnetic resonance (NMR) analyses of fibroin before and after spinning. NMR is a suitable method to study the silk fibroin structure because it can determine the atomic-level molecular structure of various sample forms ranging from solutions to solids. In this study, fibroin stored in the middle silk gland, called ‘liquid silk,’ was characterized to determine the molecular structures before spinning into B. mori and Samia cynthia ricini (S. c. ricini) silks. Second, the precise structure of fibroin after spinning was determined using a combination of site-specific stable isotope-labeled peptides and solid-state NMR. Third, the fibroin structure in fluorinated solvents and the interaction between model peptides and solvent molecules were studied to determine the relation between the molecular structure and the physical properties of regenerated silk fibers. Finally, the development of silk-based small-diameter vascular grafts by electrospinning was demonstrated as an example of the application of silks as biomedical materials.

Solution structures of repeat sequences in B. mori liquid silk

The structural features of B. mori fibroin can be conveniently studied by using the synthetic peptide (AG)n as a model for the crystalline region.5, 6 Through solid-state NMR, the backbone structure of (AG)n before spinning was determined to be a repeated type II β-turn structure. Next we determined the solution structure of fibroin molecules in native liquid silk extracted from B. mori silkworm larvae. The solution structure of native liquid silk, especially for tandem repeat sequences with (GAGXGA)n (X=Ser, Tyr or Val) and GAASGA motifs, was then determined by solution NMR.7

The sample preparation procedures for non-labeled and 13C-labeled liquid silks were as follows: the silk glands containing liquid silk were extracted from B. mori larvae from the sixth to eighth days of the fifth instar and soaked in a container filled with distilled H2O so that the liquid silk could be gently removed from the glands. The liquid silk was then soaked six times in freshly distilled H2O for 5 min to remove sericin from the silk’s surface.8 Next the liquid silk was placed in a 5-mm-diameter NMR tube together with a sealed capillary containing D2O for NMR locking. Uniformly 13C-labeled liquid silk from B. mori was biosynthetically prepared by feeding the silkworm larvae with U-13C D-glucose added to their artificial diet. By comparing 13C-13C coupled peak intensities with corresponding 12C-13C non-coupled peak intensities for a given nucleus, the 13C labeling was determined to be approximately 80%.

The 1H-15N heteronuclear single-quantum correlation spectrum of natural abundant liquid silk is shown in Figure 2a; it exhibits relatively sharp and fairly well-separated cross peaks despite the very high molecular weight of fibroin. The 1H, 15N and 13C resonances for residues in the repetitive motifs GAGXGAG (X=S, Y or V) and GAASGA were assigned by multidimensional NMR analyses. The results of the chemical shifts for Ala 13Cα and 13Cβ indicated that these motifs have neither typical α-helix nor β-sheet structures.9, 10

Figure 2
figure 2

(a) 1H-15N HSQC spectrum and peak assignments of natural abundant B. mori liquid silk. Peak assignments represent amide protons and nitrogen resonances of corresponding residues (upper case) and neighboring residues (lower case); (b) (φ, ψ) maps for typical type I, type II, type I′, type II′ and type VIII β-turns from the literature12 and an illustration of the type II β-turn conformation; (c) 25 best matches for torsion angles (φ, ψ) for GAGXGAG (X=S, Y or V) and GAASG motifs obtained using the TALOS-N program; and (d) structural models constructed using averaged (φ, ψ) (in the red circle) for each motif. Hydrogen bonds are assumed to exist between the HN of the i-th and CO of the (i+3)th residues for GAGXGAG (X=S, Y or V) motifs.

Full size image

The peaks of liquid silk were relatively broad because of the very condensed (approximately 25% w/v) sample solution, and they overlapped because of tandem repeat sequences. Therefore, the standard three-dimensional structure determination technique based on the distance geometry method was not applicable. The TALOS-N11 program is a database system used for the empirical prediction of backbone torsion angles (φ, ψ) through a combination of six types of backbone chemical shifts (1Hα, 13Cα, 13Cβ, 13CO, 15N and 1HN) and sequence information. This program was used to obtain structural information about the four motifs. Figure 2b shows the (φ, ψ) database for (i+1) and (i+2) residues in typical β-turn structures formed with four amino acids (i−i+3) for five types of β-turns in well-defined crystal structures.12 Figure 2c shows the (φ, ψ) maps of the 25 closest database matches predicted for the four motifs. As evident in Figure 2c, the 25 database heptapeptides matched to the target motif GAGXGAG (X=S, Y or V) show that the torsion angles of the residues (A2, X4 and A6) fall into narrow regions of the (i+1) residue in typical type II β-turn structures. In contrast, the Gly residues (G3, G5 and G7) fall into regions of the (i+2) residue in typical type II β-turn structures as well as other regions. Notably, for the G5 residues in the GAGYGAG and GAGVGAG motifs, a larger number of the matches of torsion angles outside the regions of typical type II β-turn structures were predicted by TALOS-N than for the corresponding G5 residue in the GAGSGAG motif. This observation suggests that the repeated type II β-turn structures for the former two motifs are less stable than the latter GAGSGAG motif.

In contrast to the three aforementioned motifs, the predicted (φ, ψ) for residues A3S4 in the GAASGA motif were scattered and inconsistent with the (φ, ψ) maps for any typical secondary structure, thus indicating that the motif assumes a random-coil structure. A replacement of Gly by Ala at the third position in the G1A2G3S4G5A6 motif destabilizes the type II β-turn structure because a Gly residue without a bulky side chain is favored at this position with a positive (left-handed) φ value in the β-turn, whereas a positive value of φ is energetically unfavorable for Ala. Figure 2d shows structural models constructed using average torsion angles for the 25 best matches (φ, ψ) for each motif. The models for GAGXGAG (X=S, Y or V) show hydrogen bond formation between the HN of the i-th and CO of the (i+3)th residues, which characterizes the β-turn structure. Furthermore, inter-residue HN-Hα nuclear Overhauser effect (NOE) cross peaks between the i-th and (i+2)th residues in GAGXGA (X=S, Y or V) motifs were observed, supporting the repeated type II β-turn structure. Thus the presence of a repeated type II β-turn structure in B. mori liquid silk was identified.

Solution structures of repeat sequences in S. c. ricini liquid silk

In addition to B. mori, various wild silkworms produce fibroin with a unique primary and higher-order structure. Recently, silk from these wild silkworms has received attention as a potentially valuable material.13, 14, 15 The diverse mechanical properties and morphologies of various wild silks are advantageous for developing novel biomaterials with appropriate properties for specific applications. S. c. ricini is one such wild silkworm and the primary structure of its fibroin consists of tandem repeat sequences composed of polyalanine and Gly-rich regions.16 The primary structure of fibroin from S. c. ricini resembles that of spider dragline silk.17, 18, 19 The fibroin structure of S. c. ricini before spinning was examined in the same manner as that of B. mori.

A sequential assignment of the most common S. c. ricini fibroin repeat motif Y1GGDGG6(A)12G19GAG22 was determined by solution NMR for liquid silk extracted from S. c. ricini larvae. Using the 1H, 13C and 15N chemical shifts and TALOS-N, we developed a model consisting of a typical α-helix structure for the polyalanine region and capping structures for terminal regions (Figure 3a). In addition, we analyzed amide proton temperature coefficients to verify the structure. A change in the amide proton chemical shift as a function of temperature can be used as an indicator of hydrogen bond formation. In general, the coefficients of backbone amide protons in a random coil configuration are rather large and negative (that is, <−6.5 p.p.b. K−1). Examination of amide temperature coefficients strongly suggested that the polyalanine region forms an α-helix (Figure 3b). Moreover, N- and C-terminal regions form capping motifs, which adopt a Schellman C-cap structure at the C terminus. The terminal cap motifs stabilize the polyalanine α-helix. Capping motifs may have a role in preventing both the structural transition from α-helix to β-sheet and fibril formation inside the silkworm body, which would be fatal to the silkworm.20

Figure 3
figure 3

(a) An S. c. ricini larva, a depiction of the fibroin molecule in a silk gland and an energy-minimized model structure of YGGDGG(A)12GGAG constructed from estimates of backbone dihedral angles using the TALOS-N program; and (b) a plot of amide proton temperature coefficients for the S. c. ricini repeat motif in liquid silk with those of Ala and Gly in B. mori as random-coil references (shown as dashed lines).

Full size image

Structure of the crystalline domain of B. mori silk fibroin

The fibroin structure after spinning was first proposed to be an antiparallel β-sheet structure by Marsh et al.21 on the basis of a fiber diffraction study of native B. mori silk fiber. Later, Fraser et al.,22 Lotz et al.23 and Takahashi et al.24 reported some intrinsic structural disorders in this antiparallel β-sheet model of fibroin. Moreover, the Ala Cβ peak in the solid-state NMR spectrum of [3-13C]Ala B. mori silk fiber is broad and asymmetric, reflecting the heterogeneous structure of silk fiber.25, 26 Conformation-dependent NMR chemical shifts can be used to easily and selectively determine the fraction of mixed structures. Furthermore, the atomic-level structure can be obtained by using NMR with a stable isotope labeling of model peptides. The sequential peptide model of the crystalline region (AGSGAG)5 was evaluated by solid-state NMR.27, 28

The fractions of random-coil and/or distorted β-turn components of each Ala residue were determined by using 13C CP/MAS NMR spectra of versions of [3-13C]Ala-(AGSGAG)5 with 10 different 13C labeling positions. In addition, the fractions of random-coil and/or distorted β-turn components of each Ser residue were determined from the 13C-15N atomic distances of five versions of the peptide with different [1-13C]Gly-Ser-[15N]Gly positions evaluated by rotational-echo double-resonance experiments. In this manner, the fractions of distorted β-turn and/or random-coil components were determined. Figure 4a shows that these data plotted against the residue position within (AGSGAG)5; the result suggests the appearance of a folded lamellar structure with two β-turns, one at approximately residue 11 and the other at approximately residue 19. By combining the structural information of Ala and Ser residues from solid-state NMR and statistical mechanical calculations, we proposed probable lamellar structures of (AGSGAG)5 with two turns at the central part of the sequence consisting of 8–12 amino acids (Figure 4b).

Figure 4
figure 4

(a) Observed relative intensities of the distorted β-turn component of (AGSGAG)5 obtained from the peak at 16.7 p.p.m. of Ala Cβ and REDOR bimodal fitting for Ser residues shown for different positions; and (b) a probable lamellar structure and turn-placed residue number of (AGSGAG)5 obtained by combining solid-state NMR and statistical mechanical calculations. A full color version of this figure is available at the Polymer Journal online.

Full size image

Interactions between fibroin and fluorinated alcohols for preparation of regenerated silk fiber

Fibroin has been developed into a wide range of forms such as microparticles, films, foams, sponges and composites.29 The processing of such wide range of forms requires solvents that dissolve fibroin. The fluorinated organic solvents 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) and hexafluoroacetone trihydrate (HFA) have been successfully used to produce regenerated fibroin fibers with high strength.30, 31, 32, 33 To understand the reasons for the difference in strength between regenerated silk fibers prepared from these two solvents, we compared the properties of native fibroin and (AGSGAG)2, a model for the crystalline part of fibroin, in the two solvents. The 13C and 1H chemical shifts obtained from 1H-13C heteronuclear single-quantum correlation spectra of fibroin in HFIP, HFA and water indicate that fibroin forms a helix-like structure in the fluorinated alcohols and a random-coil structure in water (Figure 5a). The intramolecular 1H-1H NOE data for (AGSGAG)2 imply the presence of helical structures in the middle part of the peptide in HFIP but not in HFA. Moreover, 1H-19F NOE experiments indicate that peptide–solvent interactions in HFA persist longer than those in HFIP (Figure 5b).34

Figure 5
figure 5

(a) Overlaid 1H-13C HSQC spectra of the Ala methyl region for fibroin in HFIP, HFA and water; and (b) 1H-19F intermolecular NOE spectra of (AGSGAG)2 in HFIP and HFA. The upper spectra are the observed 1D 1H spectra (control) and the lower spectra are the 1H-19F NOE spectra. A full color version of this figure is available at the Polymer Journal online.

Full size image

In processes that produce regenerated silk fibers from a silk solution, the solution typically travels through a spinneret into methanol to achieve coagulation. Coagulation involves the removal of solvents around dissolved fibroin, thereby producing a conformational transition of the crystalline regions of fibroin to β-sheet structures. Regenerated fibers prepared from HFIP solution with an appropriate draw ratio exhibit a tensile strength slightly greater than that of native silk fibers; however, the strength of regenerated fibers prepared from the HFA solution was approximately 40% less than that of native silk fiber.35 The X-ray diffraction patterns indicate that this difference in the tensile strength of regenerated silk fibers between the two solvents arises from the difference in the long-range orientation of crystalline regions.35 According to these results, the displacement of HFA molecules during coagulation may be less complete because of the relatively stronger interactions of HFA with fibroin molecules, as revealed by 1H-19F NOE experiments. A less extensive β-sheet aggregation during coagulation can occur, leading to a relatively lower tensile strength of fibers from HFA solutions. Furthermore, the silk model peptide conformation possesses a helical structure in HFIP compared with a loosed helical structure in HFA, which suggests that fibroin with more extensive helical conformations in HFIP tend to align more like a native silk solution in a silk gland and favor extensive β-sheet aggregation.

Development of silk/polyurethane small-diameter vascular graft by electrospinning

Silk fibroin fiber has a long history of use as a suture in surgery because of its biocompatibility and diverse mechanical properties. Currently, silk fibroin remains an attractive candidate for use as biomedical material. Fibroin has high strength and toughness but less elasticity than collagen and elastin, which are components of the extracellular matrix. Appropriate elasticity is critical for a material to be used as an artificial vascular graft.36 Thus a small-diameter vascular graft composed of fibroin and polyurethane was developed by electrospinning.37 The parameters of electrospinning, particularly the target-syringe distance and collection time, were optimized to provide desired characteristics for vascular grafts. The target-syringe distance influenced the porosity of a graft and the collection time influenced its compressive elastic modulus and permeability. A fibroin/polyurethane graft was prepared by using the optimized parameters of a target-syringe distance of 9 cm and a collection time of 60 min. An electrospun graft with no fray on its cutting surface (Figure 6a) and exhibiting porosity but low permeability was successfully prepared. The graft was evaluated with a rat implantation test. A patency of 100% was obtained after implantation for 2 weeks (n=2) and 4 weeks (n=4). The formation of thrombus and intimal hyperplasia, which cause occlusion of the graft, was not observed. Moreover, as shown in Figure 6b, histological staining revealed that the organization of the blood vessel walls, which are composed of collagen, smooth muscle cells and endothelial cells, progressed on both sides of the graft surface. These results indicate that the biocompatibility of silk fibroin is conserved for fibroin/polyurethane complex grafts.

Figure 6
figure 6

(a) A fibroin/polyurethane graft with an inner diameter of 3.5 mm. No fraying is observed at the cutting surfaces; and (b) histological MTC staining of the graft implanted into a rat for 2 weeks. The organization of the blood-vessel wall composed of collagen, smooth muscle cells and endothelial cells was observed on the inner side of the graft surface.

Full size image

Summary

The structures of silk fibroin before and after spinning were investigated by solution and solid-state NMR. For the fibroin structure before spinning, liquid silk extracted from silkworm larvae was measured, and sequential assignments of 1H, 13C and 15N for repeat motifs were achieved. The molecular structures were determined for B. mori and S. c. ricini liquid silks. The precise lamellar structure of the repeat motif of B. mori fibroin after spinning was determined using a combination of a stable isotope labeling of peptides and solid-state NMR. In addition, a small diameter vascular graft was successfully developed by electrospinning, and the resulting graft exhibited high patency and engraftment. The results of these studies demonstrate that fibroin has good potential for use in novel processing techniques and biomedical materials.