Introduction

Vascular grafts prepared from expanded polytetrafluoroethylene or polyethylene terephthalate fiber (Dacron) have been used clinically for many years for the revascularization of vascular grafts with inner diameters of ≧6 mm. However, replacement of vascular grafts with these synthetic materials for smaller vascular grafts of <5 mm leads to a high incidence in graft failure mainly because of the early formation of thrombosis, especially when the distal anastomosis is below the knee.1 Therefore, in the field of surgical revascularization, the development of functional vascular grafts with inner diameters of <5 mm is strongly desired. To overcome the early formation of thrombosis for small-diameter vascular grafts, various tissue-engineered vascular grafts have been developed2 and their utility and clinical experience have also been reported.3, 4, 5, 6 However, there are still no commercially available grafts to date that fully satisfy the requirements for functional, small-diameter vascular grafts.

The silk fibroin (SF) from Bombyx mori has been used in biomedical applications, such as suturing, for centuries. SF consists of highly organized β-sheet structure in the crystalline region.3, 4 Moreover, SF also has several semi-crystalline regions, which are responsible for its elasticity compared with fibers of similar tensile integrity.5, 6 We have already reported excellent results for the development of small-diameter vascular grafts made from SF.7, 8, 9, 10 These novel vascular grafts made from SF (length: 8 mm; internal diameter: 1.5 mm) improved the patency of the graft remarkably, that is, more than 85% patency remained after 1 year of implantation in rat abdominal aorta.8, 9 Although the SF fiber is superior in patency compared with other synthetic materials used in grafts, it is difficult to use the SF alone as an artificial vascular graft because of its poor mechanical stability, such as its low yielding property and weakness against bending. In contrast, it is known that several types of polyurethanes (PUs) are satisfactory materials in solving thrombosis and operability problems because of their excellent elastic properties and their biocompatible, microporous structure.11, 12

Responsibility, or the ‘compliance’ value, for the heart cycle is very important in a vascular graft. When the compliance value for the vascular graft made from SF fiber is lower than that of a native vascular graft, it is indicative of thrombosis or intimal thickening. PU has beneficial mechanical properties, including a high toughness against bending; however, the use of vascular grafts made from PU alone is occluded by thrombosis in spite of the fact the compliance value of PU is high enough to use in grafts. Therefore, the combination of SF and PU as a material in small-diameter vascular grafts should prove successful.

The miscibility of SF and PU will be a very important factor in using the SF/PU composites in artificial vascular grafts, because if phase separation occurs, an adequate fusion of the beneficial properties of SF and PU will not occur. Thus, to combine the beneficial elasticity of PU and the superior patency of SF, it is very important to know the miscibility parameters of SF/PU composites.

In this study, the miscibility and dynamics of PU/SF composite materials were investigated using solid-state nuclear magnetic resonance (NMR) techniques, because solid-state NMR is very useful in examining the homogeneity and the local molecular motion change of the PU/SF composite.

Materials and methods

Materials

An ether-polycarbonate copolymer PU dispersion in aqueous solution (Nicca Chemical Co., Fukui, Japan) was used for the PU sample. The glass-transition temperature, Tg, of the dried PU sample was ∼223 K. Silk cocoons produced by B. mori were placed in water at 95 °C, after which the threads were reeled. The dried silk threads were placed in a mixture of sodium carbonate (0.08%, w/v) and marseille soap (0.12%, w/v) at 95 °C for 120 min.13, 14, 15 This process was repeated to completely remove the silk sericin from the raw silk fibers. The removal of silk sericin was checked by a scanning electron microscope (VE-7800, Keyence Co., Osaka, Japan). The degummed silk fibers were dissolved in a 9 M lithium bromide solution to a concentration of 20 w/v% at 60 °C for 4 h and then dialyzed against distilled water for 3 days at 4 °C using a cellulose membrane (MWCO 14 000). The final concentration of the fibroin/water solution was 6.0 w/v%.16 For the preparation of the thin films of the SF and PU composite materials, the dispersion of PU in water and aqueous SF solution were mixed together in ratios of SF/PU=1:1, 1:2 and 1:10 (w/w). The mixed solutions were then cast into polystyrene petri dishes and dried at room temperature.

NMR measurements

High-resolution solid-state 13C NMR spectra were obtained using a Varian NMR system 400 WB spectrometer by the combined use of cross-polarization (CP) and magic angle spinning (MAS) with a two-pulse phase modulation17 and a 1H decoupling of 62 kHz. The pulse delay time, 1H 90° pulse, contact time and spinning rate were 5 s, 3 μs, 2 ms and 8 kHz, respectively. The 13C chemical shifts were measured relative to tetramethylsilane using the methine carbon signal at 29.47 p.p.m. for solid adamantine as an external standard. The 1H spin-lattice relaxation times in the laboratory frame (T1H) were indirectly measured from the well-resolved 13C signals enhanced by a CP of 2 ms applied after the inversion-recovery pulse sequence for 1H. The values of the 1H-13C cross-relaxation time (TCH) and T1H in the rotating frame (T1ρH) were obtained by CP contact time-dependent experiments.

Differential-scanning calorimetry

Temperature-modulated differential-scanning calorimetry measurements were performed using a DSC8500 system (Perkin-Elmer Inc., Waltham, MA, USA) under a dry nitrogen gas flow of 50 ml min−1. The samples were heated at 108 K min−1 from 210 to 430 K with a modulation period of 60 s and temperature amplitude.

Results and Discussion

Miscibility analysis of the composite materials of SF and PU by using T1H measurements

Figure 1 shows the 13C CP/MAS NMR spectra of SF (A), PU (E) and the SF/PU=1:1 (B), 1:2 (C) and 1:10 (D) composites. All of the peaks in each spectrum were assigned as shown in Figure 1 and Table 1.

Figure 1
figure 1

13CCP/MAS NMR spectra of the SF from B. mori (a), pure PU films (e) and the composite films from SF and PU at ratios of (1:1) (b), (1:2) (c) and (1:10) (d).

Full size image
Table 1 Assignment of the 13C CP/MAS NMR spectra of PU molecules
Full size table

The primary structure of SF contains multiple repeats of (AGSGAG)n, which make up 55% of the primary structure of silk fiber and form the insoluble quasi-crystalline Cp-fraction after chymotrypsin cleavage.18 The SF has two distinct structures in the solid state: silk I before spinning and silk II after spinning.19, 20

The PU is the solid sample of the dispersion in aqueous solution. In the spectra of the composite materials (Figures 1b–d), all peaks are in the same positions as those of SF and PU (Figures 1a and c). All peaks assigned to SF showed no shift after blending with PU. This consistency suggests that the excellent S–S curve property of SF remains unchanged even when mixed with PU, because the observation of no change in the chemical shift implies an unchanged three-dimensional structure of SF, which is related to a beneficial S–S curve property at high strength. Similarly, the chemical shift values of PU were the same as those before blending with SF. The linewidth of peaks for PU, however, broadened somewhat after blending with SF, in particular, at a CH3 peak of ∼10 p.p.m. and CH peaks of ∼75 p.p.m. This indicates that the molecular motion of PU becomes slow after blending with SF.

To confirm miscibility, we performed differential-scanning calorimetry measurements (Figure 2). Figure 2 shows that every glass-transition temperature (Tg) of PU for the composite materials at different ratios is the same as that of pure PU. The observation of two Tgs suggests two individual possibilities: either both polymers are not miscible on a 10-nm scale or dynamic heterogeneity or good chain connectivity is having a role.21 As stated above, the molecular motion of PU became slightly slow (linewidth became broad) after blending with SF, so the latter possibility is the most likely candidate. Furthermore, because PU has hard and soft segmental regions, the intrinsic calorimetric transition will be difficult to detect, particularly for PU in blends towing to its diluted concentration. Therefore, we employed 1H spin-lattice relaxation time (T1H) to detect the domain size,22, 23, 24, 25 which is effective in detecting the domain size on a level of several 10 nm increments from T1H even for the current PU/SF composite case.

Figure 2
figure 2

Differential-scanning calorimetry (DSC) curves of pure PU (a) and the composite materials of PU and SF. The ratios studied are 1:10 (b), 1:2 (c) and 1:1 (d). Pure SF is also shown in (e).

Full size image

Figure 3 shows logarithmic plots of the normalized peak intensities for the carbonyl groups of SF (▴, ●) and methylene groups of the PU carbons (▵, ○) vs the interval between 180° and 90° pulses. Figure 3a is obtained from the SF/PU=1:1 blend and (b) from the 1:10 blend. The decay curves of pure SF and pure PU were single exponential, and they were depicted by dashed lines. The estimated T1H values of pure SF and pure PU were 1.059 and 0.467 s, respectively. However, their corresponding decay curves in the SF/PU composites were not single exponential and different from their original decays. The initial decay of PU is concave and is convex for SF. These decays became close to each other and showed single-exponential decay after 0.5 s elapsed. Thus, the observed decay curves are typical in the case of the existence of the effective 1H spin diffusion.22, 23

Figure 3
figure 3

The observed plots for T1H determination of the composite materials of SF/PU at ratios of 1:1 (a) and 1:10 (b). The open and filled circle symbols indicate the CH2 peak from PU molecules (67 p.p.m.) and the C=O peak from SF (170 p.p.m.), respectively. Each solid line is calculated using Equation (1) on the assumption of the two-spin system model. The dashed lines represent the relaxation curves of pure PU (T1H=0.467 s) and pure SF (T1H=1.059 s), respectively.

Full size image

The 1H spin-diffusion affects 1H spin-lattice relaxation times (T1H and T1ρH) and causes them to be non-exponential depending on both the proton densities in a system and its diffusive path length, which is the domain size of components. Therefore, if SF and PU in these composite materials are in completely separate states, the observed T1 decays of SF and PU in SF/PU composites were the same as those of pure ones. In contrast, in a miscible blend, the observed T1 decay of SF will coincide with that of PU because of the very fast spin-diffusion process between the SF and PU domains.

Figure 3 shows that the T1H curve of SF changes significantly and becomes close to that of PU although both T1H curves are not completely identical to each other. This observation indicates that SF/PU composites are not homogeneous on a molecular level, but they are partially compatible. In other words, it is possible that these molecules are interacting in these composite materials.

The observed T1H curves were successfully fitted using the assumption of the two-spin system with the insufficient 1H spin diffusion rate k.22, 26 The two proton magnetizations MA(t) and MB(t), which are indirectly observed from 13C signals, are expected as follows:22, 23

where ξA=KA+fBk and ξB=KB+fAk. Symbols A and B denote the 1H spins of SF and PU, respectively. Parameters K and f are the 1H relaxation rate (1/T1H) and 1H molar fraction in the composite, respectively. The sum of fA and fB is equal to 1. The proton mole fraction fA is approximately equal to the mixing ratio for the current SF/PU composites.

The ‘best-fit’ curves are depicted as solid lines in Figure 3. The ‘best-fit’ lines are in good agreement with the observed data points. The estimated KA and KB for SF/PU=1:1 composites are 0.89 s−1 (T1A=1.14 s) and 2.36 s−1 (T1B=0.42 s) with k=7.24 s−1, respectively. Similarly, for the SF/PU=1:10 composite, KA and KB are 1.48 s−1 (T1A=0.68 s) and 2.01 s−1 (T1B=0.50 s) with k=1.75 s−1, respectively.

The maximum diffusive path length (r) is calculated using the spin-diffusion rate k.22, 23, 24, 25 The maximum diffusive path length for three dimensions is obtained by the typical solution of the diffusion equation as We employed the average value (700 nm2 s−1) of the typical diffusion coefficient D range of 500–1000 nm2 s−1 to estimate the domain size.23, 24, 25

The estimated domain size, r, K and k values from T1H analysis for every composite are summarized in Table 2. The r values for SF/PU=1:1, 1:2, and 1:10 composites were calculated to be 24, 30 and 50 nm, respectively. The estimated domain sizes are satisfactorily small enough for 1H spin diffusion to affect T1H but not sufficient to have an identical T1H decay for both components.22, 23 The r value of 24 nm for the SF/PU=1:1 composite is smaller than those of the 1:10 and 1:2 composite materials. This estimation indicates that SF and PU are partially in close proximity, particularly in the SF/PU=1:1 composite, indicating there is a portion of the molecules that contact each other through an intermolecular interaction, such as through hydrogen bonding.

Table 2 T1H values of SF C=O and PU OCH3 of each composite
Full size table

Molecular dynamics analysis of the composite materials with SF and PU using TCH and T1ρ measurements

From the T1H observations, we deduced the possible existence of the partially interacting portions of SF and PU. Furthermore, the 13C CPMAS NMR spectra showed the linewidth change of PU with SF as stated previously. The existence of the interaction and the change in dynamics affects the cross-relaxation rate (1/TCH) between 1H and 13C for both components. To confirm the change in the molecular motion of each component, we measured the CP contact time-dependent 13C NMR spectra. The T1ρH were also estimated from the CP-dependent experiments simultaneously. The plots of peak intensities for pure PU against the CP contact time are shown in Figure 4 for two characteristic peaks of PU.

Figure 4
figure 4

The spin-lattice relaxation times for protons in the rotating flames (T1ρH) obtained by the peaks at 75 p.p.m. (a) and 67 p.p.m. (b) from PU. The solid lines are fitted curves obtained using Equation (2).

Full size image

The solid lines in Figure 4 were obtained from the least-square fit of the conventional equation for the carbon magnetization behavior MC(t) at various CP contact times as follows:22, 27

Here, MMax is the maximum value of carbon magnetization. The TCH and T1ρH values of PU and SF are tabulated in Table 3.

Table 3 Results of the TCH and T1ρH values of PU and SF molecules of each composite
Full size table

Figure 4 shows that the simulated curves are in excellent agreement with the observed data points. Figure 4a was obtained from the carbon signal of the soft segment of pure PU at 75 p.p.m. (OCH group), and Figure 4b was obtained from the hard segment of pure PU at 67 p.p.m. (OCH2 group). For the soft segment of pure PU, the obtained T1ρH value was 42.3 ms and that for the hard segment was 5.7 ms. Furthermore, the TCH values for soft and hard segments were 2.3 and 0.9 ms, respectively. These observations suggest that the large T1ρ value relates to the rapid molecular motion of the soft segment and the short T1ρH value corresponds to the slow motion of the hard segment of PU. In addition, the TCH obtained from the hard segment is shorter than that from the soft segment. This is because the cross relaxation occurs more quickly in rigid states: the rate (1/TCH) of the hard segment is greater than that of the soft segment. Figure 5 shows the comparison of the TCH and T1ρH values in each group of three types of composite materials. The plots of open and filled symbols indicate SF (○: Ala C=O; □: GlyCα; □: Ala Cα) and PU (●: 67 p.p.m. (CH2 in hard segment); ▴: 75 p.p.m. (CH in soft segment); ▪: 73 p.p.m. (CH2 in soft segment)) peaks. For the hard segment of PU, both TCH and T1ρH values did not change significantly for the entire compositions, indicating that the hard segment of PU moves individually and is not influenced by SF in SF/PU composites. However, the T1ρH values of SF slightly decreased after blending with PU. The TCH values of Ala C=O group remained unchanged, but those of the Ala and Gly Cα carbons were distributed randomly and shifted slightly.

Figure 5
figure 5

The comparison of the T1ρH values in each group of three types of composite materials. Open and filled symbols indicate SF (○: Ala C=O; □: GlyCα; □: Ala Cα) and PU (●: 67 p.p.m. (CH2 in hard segment); ▴: 75 p.p.m. (CH in soft segment); ▪: 73 p.p.m. (CH2 in soft segment)) peaks, respectively.

Full size image

In contrast, the TCH and T1ρH values for the soft segment of PU decreased dramatically upon an increase in the SF concentration. This observation indicates that the molecular motion of the soft segment of PU is significantly affected and becomes slow by its interaction with SF in the composite. Especially, the decrease in the TCH value supports the increase in rigidity and/or proton density for the soft segment of PU. Thus, it is suggested that the soft segment of PU interacts with SF to some extent.

Moreover, it is noteworthy that the T1ρH of the soft segment of PU at the SF/PU=1:1 composition increased compared with that of the 1:2 composition. According to the results of TCH, the molecular motion at the 1:1 composition is still slow, because the cross relaxation is very sensitive to its rigidity. Therefore, the increase in T1ρH for the soft segment of PU at the 1:1 composition is not only due to the mobility change but also most likely due to the difference in the domain size. The smaller domain size (24 nm) in comparison with the other compositions (30–50 nm) presumably causes a change in the proton density and the motional mode around the soft segment of PU.

Conclusions

In this study, we elucidated the molecular dynamics, domain size and intermolecular interaction of the composite materials of SF and PU for the development of biomaterials, such as artificial vascular grafts. The solid-state 13C CP/MAS NMR and relaxation-time experiments were very useful in determining the change of dynamics and interaction of the composite materials. The 13C CP/MAS NMR spectra showed that the structures of composites in several ratios were the same as that of before mixing. As for the T1H measurements, the estimated domain size for SF/PU=1:1, 1:2, and 1:10 composites were calculated to be 24, 30 and 50 nm, respectively. The domain size of 24 nm for the SF/PU=1:1 composite is much smaller than those domain sizes of the 1:10 and 1:2 composites. This result indicated that the molecular miscibility between SF and PU are partially in close proximity particularly in the SF/PU=1:1 composite, indicating there is some intermolecular interaction between SF and PU, such as a hydrogen bonding interaction. In addition, differences in the molecular motion between the hard and soft segments in the PU were detected, suggesting that the soft segment of PU interacts with SF to some extent.