Some Mechanistic Insights into the Gelation of Regenerated Silk

Mar 11, 2009 - Complex Fluids and Polymer Engineering Group, Polymer Science and Engineering DiVision, National ... Technology, Bombay, India 400 076...
0 downloads 0 Views 4MB Size
8014

Ind. Eng. Chem. Res. 2009, 48, 8014–8023

Some Mechanistic Insights into the Gelation of Regenerated Silk Fibroin Sol Shailesh Nagarkar,†,‡ Avinash Patil,† Ashish Lele,*,† Suresh Bhat,† Jayesh Bellare,‡ and R. A. Mashelkar*,† Complex Fluids and Polymer Engineering Group, Polymer Science and Engineering DiVision, National Chemical Laboratory, India 411 008, and School of Biosciences and Bioengineering, Indian Institute of Technology, Bombay, India 400 076

We provide some new insights into the kinetics and mechanism of sol-gel transition as it pertains to regenerated silk fibroin, which is the principle structural protein of silkworm silk fiber. Silk fibroin was dissolved in lithium bromide and dialyzed against deionized water to prepare a regenerated fibroin solution. This solution was found to be unstable at lower pH and transformed into a colloidal gel. The kinetics and mechanism of the sol-gel transition were investigated using rheology and light scattering. We show that gelation proceeds in two steps. In the first step, a weak gel is formed almost immediately upon lowering the pH, while in the second step further gelation proceeds rapidly after a long induction time to form a self-similar structure. Introduction Silk fiber produced by the silkworm Bombyx mori (B. mori) possesses remarkable mechanical properties in addition to its natural luster, feel, and comfort.1,2 As a result, it has been widely used as a textile fiber since antiquity. Recent studies have shown that the tensile strength of a B. mori silk fiber can be further improved by controlling the spinning speed using an artificial take-up device and that such a fiber can potentially meet the needs of demanding engineering applications such as high strength ropes and ballistic armors.3 Another emerging application of silk fibroin is in the area of biomaterials.4-8 Properties of silk fibroin such as its biocompatibility, environmental stability, proteolytic degradability, and possibility of attaching growth factors make it a suitable candidate for creating new materials for biomedical applications. Of particular interest are three-dimensional microporous fibroin hydrogels, which can be used as scaffolds for tissue engineering applications. Silk fibers are composed of two different proteins: fibroin and sericin.8 The latter is the glue that binds fibroin fibrils to form the silk fiber. Fibroin is a fibrous protein and is the main structural component of the silk fiber. It is formed by two different peptide chains of molecular weights 350 kDa (heavy fraction) and 25 kDa (light fraction) linked together by a disulfide bond.2,9 The heavy chain is composed of alternating blocks of hydrophobic and hydrophilic oligopeptides. The peptide linkages are made up from 18 amino acids, but among these the Gly-Ala sequence predominates (∼80%) with characteristic repeat units of GAGAGSGAGAGY and GAGAGVGY.3,10 On the whole, a fibroin molecule has an overall negative charge in neutral pH and has an isoelectric point at pH ) 3.8-3.9.11 The fibroin content of naturally spun silk fibers can be separated from sericin in vitro. Complete removal of sericin is important for biomedical applications because sericin is known to stimulate adverse immunological response in vivo. Silk fibroin offers versality in processing it into foams, films, meshes, and hydrogels, which can be used as scaffolds for tissue engineering.12-18 The mechanical properties of silk-based scaffolds * To whom correspondence should be addressed. E-mail: ak.lele@ ncl.res.in (A.L.); [email protected] (R.A.M.). † National Chemical Laboratory. ‡ Indian Institute of Technology.

and its interactions with cells control the generation of cartilage, muscle, or bone in tissue engineering. Pure fibroin can be dissolved in water using a variety of salts, leading to a transparent solution of so-called regenerated silk fibroin (RSF).19 RSF is unstable especially at lower pH and higher temperatures where it forms a hydogel after a delay time that depends on the conditions. During the gelation process, fibroin molecules rearrange from a random coil conformation in the sol state to an antiparallel β sheet conformation in the gelled state.19,20 The presence of a large amount of Gly-Ala repeat units in fibroin favors the formation of β sheets. RSF hydrogels are formed with strong physical cross-links that cannot be broken by dilution or changes of the pH, although they may be reversible just after formation.21 The mechanical properties and pore size of silk hydrogels can be controlled by modulating fibroin concentration.21 Osteoblast cells have been successfully cultured using silk hydrogel.22 Another demonstrated application of fibroin hydrogel combined with gelatine and elastin has been in controlled drug delivery.23,24 Sol-gel transition of fibroin solution is also important in the natural fiber forming process practiced by silkworms and spiders.25 In this sophisticated spinning technology, a 27 wt % high viscosity aqueous silk solution present in the middle gland of silkworm is extruded out of a gradually narrowing spinning duct in which it experiences flow-induced orientation coupled with lowering of pH and changes in ionic concentration. As a result, the silk dope is first transformed into a liquid crystalline state in the anterior part of the silk gland, and then into an isotropic sol of low viscosity. Just before the entrance to the spinneret, the sol transforms into a gel, which undergoes a unique drawdown process26 to form the silk filament. The drawdown phenomenon results in significant orientation of the microstructure that ultimately results in excellent mechanical properties of the silk. The sol to gel transition is critical to the success of the drawdown process. In this work, we investigate the gelation of RSF at pH lower than the isoelectric point of fibroin using a combination of rheology and light scattering techniques. We show that the evolutions of creep compliance, static light scattering intensity, and dynamic light scattering intensity as a function of gelation

10.1021/ie801723f CCC: $40.75  2009 American Chemical Society Published on Web 03/11/2009

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

time provide useful insights into the kinetic and mechanistic aspects of the gelation process. Experimental Section Materials. Cocoons were obtained from B. mori silkworms reared in controlled conditions at the Central Silk Research and Training Institute, Mysore. NaHCO3 (S. D. Fine chemicals) and LiBr (>99% purity, Sigma Aldrich, Germany) were used as such without further purification. Cellulose acetate dialysis bags of molecular weight cutoff equal to 12 400 (Sigma Aldrich) were used for dialysis. 0.1 N HCL (S. D. Fine chemicals) was used for the adjustment of pH of the RFS. Deionized water (MilliQ, Millipore Inc.) of pH 6.9, resistivity 18.2 MΩ cm, and TOC less than 10 ppb was used for preparing the RFS. Preparation and Characterization of RFS. Silk cocoons were boiled in 0.5 wt % of NaHCO3 for 1 h to remove sericin. The boiled fibroin fibers were washed thoroughly with excess water to remove NaHCO3. The silk fibroin so obtained was then dissolved in 9.3 M LiBr to obtain a 10 wt % solution. The solution was extensively dialyzed against deionized water to yield the regenerated fibroin solution (RFS). The pH of freshly dialyzed RFS was found to be 8.2. The protein concentration (C) of the RFS was determined by measuring the absorption at 272 nm in a UV spectrophotometer (Shimadzu Scientific Instruments, Japan) and using the molar extinction coefficient 11.8 mol/L/cm.28 Typical fibroin concentration of a freshly dialysed RFS was found to be about 40 g/L, and the same was used to prepare sols for all sol-gel phase transition studies. Molecular weight and size of fibroin in the RFS were measured using light scattering as discussed below. Molecular weights of the heavy and light fractions were also measured using gel electrophoresis (SDS-PAGE). Two polyacrylamide gels prepared by using 6% and 12% bisacrylamide were used for gel electrophoresis. Preparation of Sol. Sols of desired pH were prepared by slowly adding 0.1 N HCl to freshly dialyzed RFS under constant stirring. At pH lower than 4.0, a part of the RFS gelled instantly on addition of HCl. The solution was then centrifuged at 10 000 rpm for 10 min to obtain a clear supernatant. The fibroin concentration of the supernatant was measured, and further dilutions to desired concentrations (0.01-7.5 mg/mL) were performed by adding deionized water of same pH. Detailed light scattering studies during gelation were carried out for a fewer number of sols, of 0.5, 1, and 1.5 g/L concentration. Rheological measurements were also done on the same sols, and additionally on a 7.5 g/L sol. Phase Diagram. Phase behavior studies were carried out on sols containing 0.01-10 g/L fibroin at four different temperatures in the range 5-70 °C and over a pH range of 2-8. Two milliliters of sol was gently pipetted in clean and sterile 5 mL flat bottomed vials (diameter 10 mm, Borosil Glass Work Ltd., India). The vials were sealed to avoid water loss and kept at the desired temperature (5-70 °C) for 30 days to monitor the state of the sample. The sol was considered to have gelled if it resisted flow upon significantly tilting the vials. Light Scattering. Light scattering measurements were performed on a 3D-DLS equipment (LS Instruments) employing a JDU Uniface laser of 628 nm wavelength. The intensity of the laser was controlled by neutral density filters (Thorlabs, Newton, NJ). Samples were held in a 4 mm diameter quartz cylindrical sample cell (Sigma Aldrich), which was placed in a toluene bath (viscosity ) 1.496 Cp) at 25 °C. Measurements were made for angles between 15° and 115° in steps of 5°. At each angle, data were collected for 1 min, and two such

8015

measurements were made. The time-average intensity from these measurements was used for static light scattering data analysis, while the time series data were used for dynamic light scattering analysis. For static light scattering measurements, the relative excess scattering intensity (Ir) was determined as the total intensity minus the solvent scattering divided by the scattering of toluene. In dilute solutions, Ir is related to the weight average molar mass (Mw) and the z-average structure factor (S(q))29 by Ir ) K·C·MwS(q)

(1)

where C is the solute concentration, and K is an optical constant that depends on the refractive index increment. S(q) describes the dependence of Ir on the scattering wave vector: q ) (4πn/ λ) · sin(θ/2), with θ as the angle of observation. For dynamic light scattering measurements, the normalized electric field autocorrelation function, g1(t), was calculated from the measured intensity correlation function, G2(t), using the socalled Siegert relation.30 g1(t) was analyzed in terms of a relaxation time distribution using the REPES routine:31 g1(t) )

∫ A(τ

LS)

exp(-t/τLS) dτLS

(2)

Here, A(τLS) is the amplitude of a mode having relaxation time τLS. For cases where a fast q2-dependent relaxation mode was observed, the cooperative diffusion coefficient (Dc) was calculated from the average relaxation rate as -1 Dc ) 〈τLS 〉/q2

(3)

At low concentrations when interparticle interactions become negligible, the z-average hydrodynamic radius (Rh) can be calculated from the diffusion coefficient using the so-called Stokes-Einstein relation: Rh )

kT 6πηD

(4)

In addition, a slow diffusional mode was observed in some cases caused by the presence of large spurious scatterers, probably protein aggregates. Ir was corrected for the contribution of the spurious scattering, which could be as much as 50% by multiplying the total scattering intensity by the relative amplitude of the fast mode. Rheology. Rheological measurements were performed with MCR 301 rheometer (Anton Paar, Austria) using a cup and bob fixture (bob 16.66 mm outside diameter, cup 18.01 mm inside diameter, and sample volume of 4.7 mL) inserted in a peltier environmental system maintained at 25 °C. A thin film of silicon oil of 1000 cP viscosity was used to cover the sample surface to prevent evaporation of water from the sample during the test. Creep tests were performed by applying a small stress (0.2 Pa for 0.5, 1.0, and 1.5 g/L sols, and 2.0 Pa for 7.5 g/L sol) for a period of 215 s, and repeating this test at every 3 h interval until the creep compliance of the sample showed very little change between consecutive creep tests, which took approximately 24 h. Results and Discussion Freshly dialyzed RFS was characterized by gel electrophoresis to determine the molecular weight of fibroin fractions in the sample (Figure 1a). Gels of two different cross-linking densities were required to resolve the bands corresponding to the light

8016

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

Figure 1. (a) SDS-PAGE of freshly dialyzed RFS showing the light and heavy fractions. (b) DLS of freshly dialyzed RFS of pH 8.2 at ambient temperature.

and heavy chain fractions of the fibroin molecule. Also shown on each gel are distinct bands of molecular weight markers. The 12% gel showed a clear band for the light chain fraction of fibroin. The molecular weight, determined qualitatively from comparison with marker bands, was approximately 25 kDa, which is in agreement with the reported molar mass of the light chain fraction of fibroin.2 This indicates that the light chain fraction remained intact during the regeneration process. The 6% gel, however, showed a smear starting from the well and stretching down to a significant distance along the gel. A relatively darker band at a location corresponding approximately to 350 kDa was barely visible in the smear. This molecular weight is in agreement with the reported molar mass for the heavy chain fraction.2 Smear patterns similar to ours have also been reported earlier for regenerated fibroin solutions.31 Smear can be caused by stretching of proteins trapped in gel pores under the influence of applied potential. Alternatively, the presence of smear could also imply a broad distribution of molecular weight. In our case, the molecular weight distribution corresponding to the observed smear length ranged from values much higher than 300 kDa to about 150 kDa. The lower molecular weight species could be formed due to degradation of the fibroin during the regeneration process. It may be noted that the cocoons were boiled in alkaline water for 60 min, and some degradation can happen during this process. The higher molecular weight species could form due to aggregation as discussed below.

Static and dynamic light scattering experiments were performed on a freshly dialyzed RFS at 25 °C. The refractive index increment for this solution is known to be dn/dC ) 0.18 mL g-1.27 The relaxation time distribution obtained from dynamic light scattering data using eq 2 displayed two diffusive modes of relaxation as shown in Figure 1b. The slow mode was attributed to the presence of large aggregates (Rh ) 50-60 nm), whereas the fast mode of relaxation was attributed to individual fibroin molecules (Rh ) 9 nm). The size of the individual fibroin molecules determined here is in good agreement with the size determined by Hossain et al.32 in 6 M LiBr solutions and by us33 in 4.7 and 2.3 M LiBr solutions. The relative amplitude of the slow mode, and thus the relative scattering intensity of the aggregates, was at most 0.8 in pure water. However, the concentration of the aggregates was negligible. The large relative amplitude of the slow mode results from the fact that the scattering intensity is proportional to the molar mass of the solute that is orders of magnitude larger for the aggregates than that of the individual fibroin molecules. The intensity scattered by the fibroin proteins was calculated by multiplying the total scattering intensity by the relative amplitude of the fast mode. After correction, the intensity was found to be independent of q. The molar mass of the fibroin proteins was determined by extrapolation to zero concentration and found to be (3 ( 1) × 105 g/mol in pure water. This value is in reasonable agreement with that reported for isolated native fibroin (about 4 × 105 g/mol)34,35 and with our gel electrophoresis data discussed

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

8017

Figure 2. (a) Physical states of RFS of different concentrations upon gelation at pH 3 and 50 °C. Numbers at the bottom of vials denote concentration of RSF in g/L. (b) Phase diagram of RFS as a function of pH, temperature, and concentration.

earlier. The light scattering data also provide a possible explanation for the observed smear in the SDS-PAGE using 12% polyacrylamide gel. It is likely that the aggregates in the freshly dialyzed RFS appear as the “high molecular weight” species in the SDS-PAGE. The molecular characterization data for freshly dialyzed RFS presented above suggest that the solutions are intrinsically unstable even at pH near 7. Fibroin molecules were found to aggregate to a small extent even during the dialysis process. The stability of RFS sols of different protein concentrations at various pH and temperatures was investigated further by monitoring the samples for a period of 30 days, at the end of which the state of the samples was categorized into four distinctive groups, opaque gel, translucent/transparent gel, precipitate, and sol. Figure 2a shows a representative set of samples obtained after keeping sols of different fibroin concentrations and pH 3 at 50 °C for 30 days. Going from left to right in this picture, the first three samples are opaque gels, the next three are transparent/translucent gels, and the last three are sols containing small amounts of precipitates. Figure 2b shows the phase diagram obtained from such observations for all concentrations, pH, and temperatures studied in this work. It can be seen from Figure 2b that for C e 0.3 g/L, the RFS remains in sol state independent of pH and temperature. For C

g 5 g/L, the RFS sol always forms opaque gels irrespective of pH and temperature. In the intermediate concentration regime, the sol forms translucent or transparent gels. The tendency to remain in sol state was found to be higher at pH greater than the isoelectric state. It may, however, be noted that the sol state is metastable at all pH and might gel or form precipitates over a period of time much longer than the 30 days waiting period arbitrarily chosen in this study. In this context, Figure 2b should not be considered as a true equilibrium phase diagram for aqueous regenerated fibroin sols. To investigate the kinetics and mechanism of gelation, we have chosen in the present work to study only a representative set of fibroin sols of 0.5, 1.0, and 1.5 g/L concentrations. For rheological studies, we have additionally used sols of higher concentrations up to 7.5 g/L. Further, only sols of pH 2 were studied in this work. Similarly, the gelation temperature was fixed at 25 °C. This choice of this set of concentrations, pH, and temperature is based on our previous work in which we have shown that the gelation time of fibroin sol is more or less independent of pH at values below the isoelectric point (pH 4.0) of fibroin33 and that the gelation time was found to depend only weakly on the fibroin concentration and temperature. We first present rheological data, which describe the macroscopic mechanical response of RFS sols as they undergo

8018

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

η2i , and Gi are, respectively, the viscosities of the two dashpots and the elasticity of the spring comprising the ith mode as shown in Figure 3. In general, the solution to the coupled eqs 5 and 6 is not analytically tractable. However, there is an analytical solution available for a single mode,37,38 and it is given by J(t) ) γ˙ Jt - BJ + e-AJt[BJ cos(ωJt) + CJ sin(ωJt)]

(7)

Here, σ0 η2

(8)

GJ + η1η2b/I 2(η1 + η2)

(9)

γ˙ J )

AJ ) Figure 3. Schematic of Jeffery model.

BJ )

gelation. In a separate paper, we have discussed isothermal dynamic oscillatory response of RFS sols during gelation,33 and we have shown that the dynamic moduli of the sol increased relatively slowly at first until a gelation time (tg), after which a rapid increase in the moduli was observed accompanied by an increase in turbidity. In the present work, we have used creep experiments to probe rheological changes of RFS sols during their gelation. While creep experiments are indeed complimentary to previously used dynamic oscillatory experiments, they provide additional insights into the mechanistic aspects of the gelation process as will be described below. The response of a viscoelastic fluid to an imposed step stress is dominated by viscous character at long time and elastic character at short time. Consequently, at times greater than a characteristic response time of the material, its creep compliance typically increases linearly with time. At short times, the inertia of the system (measuring fixture plus rheometer drive) couples with the elasticity of the material, and for certain fluids this results in a damped oscillatory response of the creep compliance. This phenomenon, called creep ringing, has been observed for biopolymer gels36 and colloidal gels37 and has been reviewed recently.38 The creep behavior of a viscoelastic fluid can be effectively described by a generalized Jeffery model, which is essentially a combination of multiple Jeffery modes organized in series as shown in Figure 3. Each mode contains a viscous element (dashpot) arranged in series with a Voigt-Kelvin mode, which consists of an elastic element (a Hookean spring) in parallel with another dashpot. The governing equations that describe the creep response of the generalized Jeffery’s model on application of a stress (σ0) are38 I b

∑ dγdt˙

σ + τi1

i

) σ0 - σ

(5)

i

(

dσ dγ˙ i ) ηi2 γ˙ i + τi2 dt dt

)

(6)

Here, I is the moment of inertia of the system, b is a parameter related to the geometry of the measuring fixture, σ is the stress in each Jeffery mode (note that all modes share the same stress when arranged in series), and γ˙ i is the strain rate of the ith mode. τ1i ) (η1i + η2i )/(Gi) and τ2i ) (η1i )/(Gi) are, respectively, the relaxation time and retardation time of the ith mode, and η1i ,

CJ )

ωJ )

[

σ0 η1 + η2 2AJI -1 GJ η2 η2b



(

AJ γ˙ J B ωJ J AJ

)

]

GJb η2 - A2J I η1 + η 2

(10)

(11)

(12)

It can be shown from eq 3 that at very short times, J(t) ≈ (b)/ (2I)t2, and at long times, J(t) ≈ γ˙ Jt. At intermediate times, the last term in eq 3 produces an oscillatory response when GJ > A2J (I)/(b)((η1 + η2)/(η2)). Figures 4a shows the creep response of 0.5 g/L RFS sol of pH 2 at temperature 25 °C during gelation. The zero hour data shown in the figure were collected immediately upon loading the sample in the cup and bob fixture. The sol was prepared as explained in the Experimental Section. Note that when the pH of a freshly dialyzed RFS was lowered to 2, a part of the fibroin content was lost by instantaneous precipitation, and this was centrifuged to leave behind a transparent sol. This sol was used for the creep experiments after adjusting its concentration to 0.5 g/L with an aqueous solution of pH 2. The remaining creep data shown in Figure 4a were collected at 3 h intervals after the zero time. Figure 4a also shows creep data for a freshly dialyzed RFS sol of pH 8.2 and diluted to 0.5 g/L. The creep compliance of this solution is clearly seen to be dominated by its viscous character. From the long time slope of the compliance data, we calculated the viscosity of this freely flowing sol to be approximately 9 cP. In contrast, the sol of pH 2 at zero hour shows a significantly lower compliance relative to the RFS of pH 8.2, and also shows creep ringing. This is indicative of a weak gel nature of the sol of pH 2 even at zero time. Thus, the lowering of pH of the freshly dialyzed RFS seems to change its rheological response from being dominantly viscous at pH 8.2 to distinctly elastic at pH 2. With time, the creep compliance reduced further and creep ringing became increasingly predominant in amplitude and frequency. Data for 12 and 15 h show a distinct jump in the frequency of ringing, which suggests a rapid increase in the elasticity of the sample in this time interval. The elasticity of sample can be estimated using Struick’s formula:38 G' ≈

Iω2* ∆ 1+ b 2π

2

[ ( )]

(13)

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

under the assumptions of small logarithmic decrement ∆ and negligible sample inertia. Here, ω* is the experimentally determined creep ringing frequency. The logarithmic decrement was calculated for all data shown in Figure 4b using38 ∆ ) 2 ln[(J1 - J2)/(J3 - J2)]

(14)

and was found to be ∆ < 2π. In eq 14, J1, J2, and J3 are the compliance values of the first peak, first valley, and second peak of the creep ringing data. The elastic modulus of the sample calculated from eq 13 for the 0.5 g/L sol can be seen to increase rapidly between 12 and 15 h (Figure 4b). This behavior agrees with our previously reported data on the dynamic moduli of RFS sols measured during gelation by using SAOS tests.33 The lines going through each of the data sets in Figure 4a represent calculations of the Jeffery’s model given by eqs 5 and 6. For creep data above 3 h, only a single Jeffery mode was required to fit the data (eqs 7-12), whereas below this time at least two modes were required. The data for zero time could not be satisfactorily modeled by even a two-mode fit. Increasing the number of modes might have provided a better fit to the data, albeit at the cost of requiring a large number of fitting parameters. A single mode Jeffery model has three material parameters η1, η2, and GJ and a rheometer-related parameter I/b that are required to be determined to fit the creep data. Consequently, an n-mode Jeffery model would require 3n + 1 model parameters. In view of the difficulties in determining the values of a large number of fitting parameters, we have restricted the number of modes to a maximum of two in our model fitting exercise. The determination of model parameters in the case of a single mode was done as follows. The value of GJ was determined from GJ ≈ 1/J(tring), where J(tring) is the compliance at the end of the oscillations, and η2 was determined from the slope of the plot of long time creep compliance versus time. The value of I/b was obtained by fitting the very short time data to J(t) ≈ (b)/(2I)t2. This leaves only one parameter, η1, for fitting the model to the data. Data fitting was done by visually comparing the model calculations and experimental data. For fitting the data shown in Figure 4a, the value of I/b was taken as 0.5 N · s2/m2, and the values of the material parameters are shown in Figure 4c in the form of the two time constants, the relaxation time τ1 ) (η1 + η2)/(GJ) and the retardation time τ2 ) (η1)/(GJ). The values of the two viscosities were found to be such that η2 . η1 for all times starting from zero time. We may infer that η1 represents a microscopic viscosity experienced by the structure in the sample, while η2 represents the macroscopic viscosity of the sample. The plot in Figure 4c shows values of the two time constants for times starting from 3 h after zero time because the data can be reasonably well described by a single mode Jeffery model only after this time. We find that while the relaxation time decreases slightly during the gelation process, the retardation time changes significantly between 12 and 15 h. This happens because both η2 and GJ increase during gelation, whereas η1 remains essentially constant. The creep data for 1.0, 1.5, and 7.5 g/L RFS sols of pH 2, not shown here, were qualitatively similar to the data for 0.5 g/L RFS sol shown in Figure 4a. The value of creep compliance decreased with increasing fibroin concentration. For all concentrations, the creep compliance at zero time showed a weak gel structure. The data at initial times could be modeled only by assuming more than one mode in the Jeffery model. After about 3 h from zero time, the ringing data could be modeled nearly quantitatively using a single mode Jeffery model. In the course of gelation the ringing became increasingly predominant, and for all concentrations a jump in ringing frequency was

8019

observed between 9 and 15 h after zero time. This suggests that the time at which rapid gelation starts (tg) is weakly dependent on the fibroin concentration of the RFS sol, which is in agreement with our previous work.33 For all concentrations, the relaxation time decreased gradually with gelation time, whereas the retardation time changed to a greater extent and rapidly near tg. We now discuss the results of our light scattering studies on fibroin sols during their gelation process. The sols were prepared as described earlier and were quickly transferred into sample cells. The first set of scattered intensity readings at different scattering angles (15-135°) was obtained immediately after placing the sample cell in the beam path, and this was denoted as the 0 h data. Further measurements were done by maintaining a waiting period of 3 h between consecutive data sets. In the cases where dynamic light scattering revealed the presence of a slow mode in addition to the dominant fast mode, the amplitudes of the two modes were calculated at each angle as explained earlier. The slow mode was observed to be significant in amplitude at small q for 0 and 3 h, indicating that the slow mode corresponds to larger aggregates (which have longer relaxation time), and that in the initial stage of gelation the sol has at least two modes. Note, however, that the concentration of larger aggregates is expected to be negligible for reasons argued earlier. At 6 h and higher, the amplitude of the slow mode was negligible, and the sol contained predominantly a single mode. These observations qualitatively agree with the rheological data presented earlier wherein modeling of the creep ringing at initial times (up to 3 h) required more than one Jeffery’s mode, while at 6 h and above a single mode Jeffery model was found to be adequate to describe ringing. The static scattered intensity of the fast mode was obtained by multiplying the measured total time-averaged scattering intensity by the amplitude of the fast mode. The relative scattered intensity of the fast mode is plotted versus the wave vector in Figure 5a for 1.0 g/L RFS sol of pH 2 as a function of time during the gelation process. The relative scattered intensity was normalized by the product of protein concentration (C) and optical constant (K) so that, as suggested by eq 1, the Y-axis represents M(w)S(q) of the sol at any given time. It can be seen from Figure 5a that the relative scattered intensity showed a q-independent region in the low q limit for gelation times between 0 and 12 h. During this time span, the intensity increased gradually with time, but after 12 h a rapid rise in the scattered intensity was observed and the sol became visually turbid. A strong q-dependence of the scattered intensity was observed after this time, indicating the development of heterogeneous structures. Beyond 18 h the relative scattered intensity did not evolve further with time. The time at which a rapid rise in scattering intensity was observed agrees well with the time at which rheological parameters, the frequency of creep ringing and the dynamic moduli of the sample, also increased rapidly. The apparent molecular weight of aggregates representing the fast mode at various times between 0 and 12 h was obtained from the q-independent relative scattered intensity data shown in Figure 5a. The apparent molecular weight increased from about 106 g/gmol at 0 h to about 3 × 107 g/gmol at 12 h. The corresponding hydrodynamic radii of these aggregates at various times were calculated from the dynamic light scattering data as follows. The relaxation time of the fast mode 〈τLS,fast-1〉, obtained using eq 2, was plotted versus q2 for various times between 0 and 15 h and is shown in Figure 5b. A straight line equation passing through the origin provided an excellent fit (R2 > 0.99) to each of the 〈τLS,fast-1〉 versus q2 data up to 12 h, indicating

8020

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

Figure 4. (a) Creep behavior of 0.5 g/L RSF sol of pH 2 at 250 °C. (b) Evolution of storage modulus calculated from Struick formula [eq 13]. (c) Relaxation and retardation times estimated by fitting single mode Jeffery model to the creep data shown in (a).

diffusive behavior of the aggregates. For the data collected at 15 h after zero time and above, the error in fitting a straight line to the 〈τLS,fast-1〉 versus q2 data was higher (R2 < 0.75) so that eq 3 may be considered to be inapplicable. The diffusion coefficients of aggregates at various times between 0 and 12 h

were calculated from the slopes of the straight line fits shown in Figure 5b. The hydrodynamic radii of aggregates at various times between 0 and 12 h were calculated using eq 4, assuming that interaggregate interactions can be neglected. Reliable estimates of the hydrodynamic radius may be obtained by

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

8021

Figure 5. (a) Time course static light scattering on 1.0 g/L RSF of pH 2 at 250 °C. (b) Frequency versus q2 for RFS sol of pH 2 at 250 °C as a function of time. (c) Evolution of hydrodynamic radius and apparent molecular weight of aggregates as a function of gelation time.

diluting the sol at each time during gelation and measuring their scattering in the dynamic mode. The assumption here is that the dilution process would not alter the structure of the aggregates. Alternatively, we can try to estimate the hydrodynamic radius of aggregates from the nondiluted sols. To do this, we need to make assumptions about the value of viscosity appearing in eq 4. If the viscosity of the medium containing the aggregates is assumed to be the microscopic viscosity η1 ) 0.2 Pa s obtained from fitting the creep ringing data to a single mode Jeffery model, then the hydrodynamic radius of aggregates in the sol is estimated to be about 0.1 nm at 0 h and 0.4 nm at 12 h. This value of viscosity is considerably larger than the viscosity of water and would therefore imply a concentrated

suspension of aggregates. Consequently, the calculation of Rh using eq 4 would be incorrect. If, however, the viscosity in eq 4 is considered to be that of water, thereby implying a dilute suspension of the aggregates, then their hydrodynamic radius is estimated to be about 20 nm at 0 h and 80 nm at 12 h. We have separately viewed the 1 g/L sol of pH 2 during its gelation under a confocal scanning light microscope (CSLM), and we observed that no structure was visible until 9 h, whereas at 12 h and above a network of protein aggregates became clearly visible. Given that the resolution of our CSLM was about 200 nm, it is clear that aggregates of 80 nm radius that are formed at 12 h would become visible. Thus, the assumption of a dilute suspension of aggregates that essentially experiences an aqueous

8022

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

medium of 1 cP viscosity seems reasonable. On the basis of this discussion, we have calculated the increase in apparent molecular weight and hydrodynamic radius of aggregates as a function of the gelation time until 12 h for the 1.0 g/L RFS sol of pH 2, and the same is shown in Figure 5c. From the light scattering and rheology data, we can now try to reason the gelation process as follows. The data in Figure 5 show that in the initial sol state (zero time) the aggregates were 20 nm in size and had a molecular weight of ∼106 g/gmol. Based on the measured molecular weight of individual fibroin molecules, this implies that the aggregates are made, on an average, by the association of three to four protein molecules. The association of fibroin molecules is driven by two factors: a decrease in intermolecular electrostatic repulsion due to protonation of carboxyl groups on the fibroin molecules at pH lower than the isoelectric point, and a simultaneous increase in intermolecular attractions caused perhaps by hydrophobic interactions. The protein aggregates seem to form a weakly percolating network as evidenced by the significantly lower creep compliance and distinct creep ringing of the sol of pH 2 at 0 h as compared to that of a freshly dialyzed RFS sol of pH 8.2. This appears to be a surprising result because in a 1 g/L RFS sol the volume fraction of aggregates, as calculated from their mass, size, and protein concentration, is only about 0.002. That such a dilute suspension of aggregates is able to form a percolating network seems unreasonable. It is possible that fibroin molecules might form a transient network in the sol as suggested by Jin et al.16 However, the clear evidence of aggregates in our light scattering data does not support this hypothesis entirely. Alternatively, the elasticity of the sol might arise out of some interfacial effects caused by adsorption of proteins on surfaces of the rheometer fixtures and the air-water interface. We have not probed the exact origin of the surprising elasticity of the sol at zero time. With time, the aggregates gradually increase both in size and in mass as seen from Figure 5c. After about 6 h, the initial bimodal aggregate distribution in the sol appears to evolve into only one type of aggregate. The creep ringing data can now be fitted by a single mode Jeffery model, and the viscosity and elasticity of the sol during this time increase gradually as evidenced by the gradual increase in η2 and GJ. At gel time (tg), the aggregates rapidly agglomerate to form a heterogeneous structure. After 18 h, the relative scattering intensity of the gel shows a power law dependence Ir(q) ≈ q- 2.1 on wave vector (line in Figure 5a), suggesting a self-similar structure. Interestingly, the hydrodynamic radius and molecular weight of aggregates between 0 and 12 h also show a power law relation given by Ma ≈ Rh2.35 (inset in Figure 5d). The similarity of the two exponents suggests that the initial aggregates are also fractal objects and that they aggregate in a self-similar manner to produce the gel. At tg, the elasticity of the gel also increases rapidly following the buildup of the fractal structure. This is evidenced by a rapid increase in G′ and ω*. We cannot comment upon the exact microstructural origin of the model parameters GJ, η2, and η1. While η2 should mean the viscosity of the percolating network, GJ and η1 might correspond to the elasticity and local viscosity of substructures of the fractal network. The pattern in which the scattered intensity and the rheological parameters increase, that is, an initial phase of gradual increase until tg followed by a phase of rapid increase after tg, is perhaps suggestive of a nucleation and growth phenomenon. At a molecular level, the gelation process is accompanied by, or perhaps driven by, conformation transition of the protein from a dominantly random coil state in the sol to a predominantly

Figure 6. Circular dichroism of freshly dialyzed RSF solution, RSF sol of pH 2 at 0 h and RSF gel at 16 h. The protein concentration was 0.5 g/L. Gelation was done at pH 2 and 25 °C.

β-sheet state in the gel. Figure 6 shows circular dichroism data for the as-dialyzed RSF solution (pH 8.2), the initial RSF sol (pH 2, zero time), and the final RSF gel. The protein concentration was C ) 0.5 g/L, and gelation was done at pH 2 and 25 °C. Analysis of the data showed that in the as-dialyzed state the fibroin molecules have about 75% random coil conformation and 25% β-sheet conformation. This composition remained essentially unaltered in the sol state when measured immediately upon pH adjustment followed by centrifugation and filtration. However, the β-sheet content increased with time during the gelation process, and in the final gel the β-sheet content was approximately 52%. The formation of β-sheet structures during gelation is in agreement with previous work.15 A further detailed study of events at the molecular length scale is presently under progress. Conclusions We have investigated the mechanism of gelation of aqueous sols of regenerated silk fibroin at room temperature and pH 2, which is lower than the isoelectric pH of fibroin molecule. Time evolution of the creep compliance of sols of different fibroin concentrations was tracked. Similarly, the change in relative scattering intensity of the same sols was measured as a function of gelation time. We find that immediately upon lowering the pH of a freshly dialyzed regenerated fibroin solution from 8.2 to 2, a weak gel was formed as evidenced by low compliance and a creep ringing effect. The sol consisted of aggregates formed by the association of collapsed fibroin molecules. The aggregates were on the average made from three to four fibroin molecules and had a size of about 20 nm. These are most likely the primary aggregates of the final fibroin gel. With time, the elasticity of the gel increased gradually with time until gel time tg. Simultaneously, the size and mass of the aggregates increased gradually while preserving a self-similar structure. At tg, gelation was observed to progress rapidly. The elasticity of the gel increased rapidly as evidenced by increased creep ringing effect. During this time, the fibroin aggregates rapidly agglomerated to create a hydrogel with a self-similar microstructure. After complete gelation, the scattering intensity did not further change with time. However, the creep compliance kept decreasing slowly, indicating an aging process. The sol-gel transition appears to be driven by a nucleation and growth process, which involves at a molecular level the formation of β-sheets in the aggregated proteins.

Ind. Eng. Chem. Res., Vol. 48, No. 17, 2009

In addition to elucidating the mechanism and kinetics of gelation, the present study has provided useful information on two important attributes of silk fibroin hydrogels: its mechanical strength and microstructure, which can potentially be used to design gels for drug delivery and tissue engineering scaffold applications. Acknowledgment This Article is dedicated to Prof. J. B. Joshi on the occasion of his 60th birthday. S.N. would like to acknowledge financial support from the Council of Scientific and Industrial Research, India. We are grateful to Dr. Nirmal Kumar and Dr. Kanika Trivedy of the Central Sericulture Training and Research Institute, Mysore, for providing us with silk cocoons. We thank Dr. Taco Nicolai and Dr. Christophe Chassenieux at the University du Maine for many useful discussions. We would also like to thank Dr. Asmita Prabhune of NCL for helping with gel electrophoresis. Finally, we are grateful to Dr. Shyamalava Mazumdar at the Tata Institute of Fundamental Research, Mumbai, for helping us with circular dichroism data. Literature Cited (1) Gosline, J. M.; Guerette, P. A.; Ortlepp, C. S.; Savage, K. N. The mechanical design of spider silks: From fibroin sequence to mechanical function. J. Exp. Biol. 1999, 202, 3295–3303. (2) Vollrath, F.; Knight, D. P. Liquid crystalline spinning of spider silk. Nature 2001, 410, 541–548. (3) Shao, Z.; Vollrath, F. Surprising strength of silkworm silk. Nature 2002, 418, 741. (4) Altman, G. H.; Diaz, F.; Jakuba, C.; Calabro, T.; Horan, R. L.; Chen, J.; Lu, H.; Richmond, J.; Kaplan, D. L. Silk-based biomaterials. Biomaterials 2003, 24, 401–416. (5) Foo, C. W. P.; Kaplan, D. L. Genetic engineering of fibrous proteins: spider dragline silk and collagen. AdV. Drug DeliVery ReV. 2002, 54, 1131– 1143. (6) Megeed, Z.; Cappello, J.; Ghandehari, H. Genetically engineered silk-elastinlike protein polymers for controlled drug delivery. AdV. Drug DeliVery ReV. 2002, 54, 1075–1091. (7) Sofia, S.; McCarthy, M. B.; Gronowicz, G.; Kaplan, D. L. Functionalized silk-based biomaterials for bone formation. J. Biomed. Mater. Res. 2001, 54, 139–148. (8) Perez-Rigueiro, J.; Viney, C.; Llorca, J.; Elices, M. Mechanical properties of single- brin silkworm silk. J. Appl. Polym. Sci. 1998, 70, 2439– 2447. (9) Jin, H. J.; Kaplan, D. L. Mechanism of silk processing in insects and spiders. Nature 2003, 424, 1057–1061. (10) Zhou, C. Z.; Confalonieri, F.; Medina, N.; Zivanovic, Y.; Esnault, C.; Yang, T.; Jacquet, M.; Janin, J.; Duguet, M.; Perasso, R.; Li, Z. G. Fine organization of Bombyx mori fibroin heavy chain gene. Nucleic Acids Res. 2000, 28, 2413–2419. (11) Asakura, T.; Sugino, R.; Yao, J.; Takashima, H.; Kishore, R. Comparative structure analysis of tyrosine and valine residue in unprocessed silk fibroin (Silk I) and in processed silk fiber (Silk II) from bombyx-mori solid state 13C, 15N, and 2H NMR. Biochemistry 2002, 41, 4415–4424. (12) Yeo, J. H.; Lee, K. G.; Kim, H. C.; Oh, Y. L.; Kim, A.-J.; Kim, S. Y. The effects of PVA/Chitosan/Fibroin (PCF)-blended spongy sheets on wound healing in rats. Biol. Pharm. Bull. 2000, 23, 1220–23. (13) Megeed, Z.; Haider, M.; Li, D.; O’Malley, B. W., Jr.; Cappello, J. In vitro and in vivo evaluation of recombinant silk-elastinlike hydrogels for cancer gene therapy. J. Controlled Release 2004, 94, 433–445. (14) Nazarov, R.; Jin, H. J.; Kaplan, D. L. Porous 3D scaffold from regenerated silk fibroin. Biomacromolecules 2004, 5, 718–726. (15) Tamada, Y. New process to form a silk fibroin porous 3-D structure. Biomacromolecules 2005, 6, 3100–3106.

8023

(16) Jin, H. J.; Park, J.; Valluzzi, R.; Cebe, P.; Kaplan, D. L. Structure and properties of silk hydrogels. Biomacromolecules 2004, 5, 711–717. (17) Unger, R. E.; Peters, K.; Wolf, M.; Motta, A.; Migliaresi, C.; Kirkpatrick, C. J. Endothelialization of a non-woven silk fibroin net for use in tissue engineering: Growth and gene regulation of human endothelial cells. Biomaterials 2004, 25, 5137–5146. (18) Li, C.; Vepari, C.; Jin, H. J.; Kim, H. J.; Kaplan, D. L. Electrospun silk-BMP-2 scaffolds for bone tissue engineering. Biomaterials 2006, 27, 3115–3124. (19) Ayub, Z. H.; Arai, M.; Hirabayashi, K. Mechanism of the gelation of fibroin solution. Biosci. Biotechnol. Biochem. 1993, 57, 1910–1912. (20) Chen, X.; Knight, D. P.; Shao, Z.; Vollrath, F. Regenerated Bombyx silk solutions studied with rheometry and FTIR. Polymer 2001, 42, 9969– 9974. (21) Kim, U.-J.; Park, J.; Li, C.; Jin, H.-J.; Valluzi, R.; Kaplan, D. L. Structure and properties of silk hydrogels. Biomacromolecules 2004, 5, 786– 792. (22) Motta, A.; Migliaresi, C.; Faccioni, F.; Torricelli, P.; Fini, M.; Giardino, R. Fibroin hydrogels for biomedical applications: Preparation, characterization and in vitro cell culture studies. J. Biomater. Sci., Polym. Ed. 2004, 15, 851–864. (23) Gil, E. S.; Spontak, R. J.; Hudson, S. M. Effect of β-sheet crystals on the thermal and rheological behavior of protein-based hydrogels derived from gelatin and silk fibroin. Macromol. Biosci. 2005, 5, 702–709. (24) Gil, E. S.; Frankowski, D. J.; Spontak, R. J.; Hudson, S. M. Swelling behavior and morphological evolution of mixed gelatin/silk fibroin hydrogels. Biomacromolecules 2005, 6, 3079–3087. (25) Asakura, T.; Umemura, K.; Nakazawa, Y.; Hirose, H.; Higham, J.; Knight, D. Some observations on the structure and function of the spinning apparatus in the silkworm Bombyx mori. Biomacromolecules 2007, 8, 175–181. (26) Vollrath, F.; Knight, D. Liquid crystalline spinning of spider silk. Nature 2001, 410, 541–548. (27) Iizuka, E.; Yang, J. T. The disordered and β conformations of silk fibroin solution. Biochemistry 1968, 7, 2218–2228. (28) Brown, W. Light Scattering. Principles and DeVelopments; Clarendon Press: Oxford, 1996. (29) Berne, B. J.; Pecora, P. Dynamic Light Scattering with Applications to Chemistry, Biology and Physics; Wiley: New York, 1976. (30) Stepanek, P. In Dynamic Light Scattering; Brown, W., Ed.; Oxford University Press: Oxford, 1993. (31) Horan, R. L.; Antle, K.; Collette, A. L.; Wang, Y.; Huang, J.; Moreau, J. E.; Volloch, V.; Kaplan, D. L.; Altman, G. H. In vitro degradation of silk fibroin. Biomaterials 2005, 26, 3385–3393. (32) Hossain, K. S.; Ohyama, E.; Ochi, A.; Magoshi, J.; Nemoto, N. Dilute-solution properties of regenerated silk fibroin. J. Phys. Chem. B 2003, 107, 8066–8073. (33) Nagarkar, S.; Nicolai, T.; Chassenieux, C.; Lele, A., in preparation. (34) Tashiro, Y.; Otsuki, E. Studies on the posterior silk gland of the silkworm Bombix mori. IV. Ultracentrifugal analyses of native silk proteins, especially fibroin extracted from the middle silk gland of the mature silkworm. J. Cell Biol. 1970, 46, 1–16. (35) Hossain, K. S.; Ochi, A.; Ooyama, E.; Magoshi, J.; Nemoto, N. Dynamic light scattering of native silk fibroin solution extracted from different parts of the middle division of the silk gland of the bombyx mori silkworm. Biomacromolecules 2003, 4, 350–359. (36) Baravian, C.; Quemada, D. Using instrumental inertia in controlled stressrheometry. Rheol. Acta 1998, 37, 223–233. (37) Joshi, Y.; Reddy, G. R.; Kulkarni, A.; Kumar, N.; Chhabra, R. Rheological behaviour of aqueous suspensions of laponite: new insight into the aging phenomena. Proc. R. Soc. A 2008, 464, 469–489. (38) Ewoldt, R. H.; McKinley, G. H. Creep ringing or how to deal with oft-discarded data in step stress test. Rheol. Bull. 2007, 76.

ReceiVed for reView November 12, 2008 ReVised manuscript receiVed February 10, 2009 Accepted February 12, 2009 IE801723F