Silk fibroin solution was prepared following the method previously published (de Moraes et al., 2014). Briefly, Bombyx mori silkworm cocoons (Bratac, Brazil) were degummed using 1 g.L^–1 Na₂CO₃ solution in a thermostatic bath at 85°C (Nova Analitica, Brazil) for 30 min, the process was repeated other two times. The fibers obtained were washed with deionized water and maintained at room temperature until completely dry. They were milled using a sieve with 10 mesh of diameter and then solubilized in a solution of CaCl₂: Ethanol: water (1:2:8 molar ratio). To remove the salts, the solution was dialyzed using a dialysis membrane (MWCO 3.5 kDa, Thermo Fisher Scientific, EUA) in ultrapure water (1:10 ratio). To remove the insoluble parts that are formed during dialysis, the solution was centrifuged for 30 min and 2300 relative centrifugal force (RCF). The final concentration was calculated by gravimetry and was around 4% (w/v).

Sodium alginate (Sigma-Aldrich, United States) extracted from Macrocystis pyrifera were dissolved in deionized water with a concentration of 2% (w/v). The solution was kept 3 days, so the water could solvate the powder, then the solution was homogenized by stirring.

The SF and SA blends were prepared in seven different mass proportions, as shown in Table 1. These compositions were chosen following previous studies (Lopes et al., 2018) which showed that for more concentrated systems, i.e., with less water, hydrogel formation would occur. Thus, it would not be possible to make a quantitative analysis of each phase. To prepare the blends, SF was slowly added to SA solution under stirring, the system was kept in a thermostatic bath at 25°C (Nova Analitica, Brazil). After 8 days, each phase was collected and the components were quantified. The notation used to refer the blends was SA_x%/SF_y%/WA_z% where x, y, and z are the mass fraction of each component in the blend.

Zeta potential was performed using Malvern Zetasizer Nano ZS (Malvern, United Kingdom). It was measured using a zeta universal dip cell and polystyrene cuvettes. The polymers concentration was 1 g.L^–1 and the samples were measured three times. The values of the refractive index and absorbance used were 1.33 and 0.001 a.u for SA and 1.45 and 0.001 a.u for SF. The absorbance was measured using the Varioskan LUX multimode microplate reader (Thermo Fisher Scientific, United States) and the refractive index was measured using a refractometer (Instrutherm, Brazil). The pH of SF and SA solution was 6.77 and 7.63, respectively. Zeta potential was performed to analyze the blends by mixing the SF and SA solution with the same concentration in different volume proportions, 75–25, 50–50 and 75–25 (SA-SF).

The optical microscopy was used to analyze the morphology of microparticles formed after blend preparation. The optical microscopy was performed using the E200 optical microscope (Nikon, Japan). The image was analyzed using ImageJ software.

After phase separation, the blends were centrifuged for 15 min and 2300 RCF, the final state of SA/SF blends is a liquid phase and a “solid” one that looks like a hydrogel and it is mostly composed by SF with water trapped in the structure. The analysis used to quantify each phase is described in Figure 1.

Water was quantified gravimetrically by solvent evaporation in the liquid and solid phase in an oven until the weight was constant. SF in the liquid phase was quantified by Bradford assay (Bradford, 1976), a consolidated method to quantify proteins using the Coomassie Blue G-250 stain with maximum absorbance in the range between 465 and 595 nm. This method was chosen because carbohydrates do not interfere with the results. It was prepared a calibration curve from 0 to 100 μg/mL using Bradford reagent (Amresco, United States) that was analyzed at 590 nm using a 24 plate-well in a spectrophotometer (Tecan M200 Pro, Switzerland). In the solid-like phase, it was not possible to use the same method because SF was insoluble, so it was quantified by mass balance using Eq. 3.

where m_msp is the total mass of solid phase, m_sfsp is the SF mass in the solid-like phase, m_wasp is the water mass in the solid-like phase and m_sasp is the SA mass in the solid-like phase.

The SA was quantified by mass balance in both phases using Eqs 4 and 5.

where m_salp is the SA mass in the liquid phase, m_mlp is the total mass of liquid phase, m_sflp is the SF mass in the liquid phase (determinate by Bradford assay) and m_walp is the water mass in the liquid phase.

where m_sasp is the SA mass in the solid phase, m_t is the blend mass, α is the SA mass fraction in the blend, and m_salp is the SA mass in the liquid phase.

The Flory-Huggins model is a classic mathematical approach for polymer systems. It uses the balance between the entropic and enthalpic contributions (Eqs 6 and 7), regarding the size and form effect and the intermolecular interactions.

where R is the gas constant, T is the temperature, x_i is the molar fraction of component i, V_i e V_s are the volume per amount of matter of pure component i and for solvent, respectively; χ_ij is the Flory-Huggins parameter of interaction for component i and j, φ_i and φ_j are the volumetric fractions of component i and j.

By replacing Eqs 6 and 7 in Eq. 1, the extended expression of the Flory-Huggins equation is obtained, which is usually simplified for use in a polymer-solvent system. The Flory-Huggins parameter can be, also, calculated using the activity value of each component in each phase. This is useful because in the equilibrium the chemical potential of each component in each phase is equal, and using the data of equilibrium is possible to calculated Flory-Huggins parameter solving a non-linear equation system, described by Eqs 8–10 for a ternary system (Favre et al., 1996).

where a₁ and a₂ are the activity of component 1 and 2, respectively; V₁, V₂ and V₃ are the volume per amount of matter of component 1, 2, and 3, respectively; χ₁₂, χ₁₃, and χ₂₃ are the Flory-Huggins parameter of component 1–2, 1–3, and 2–3, respectively.

The Flory-Huggins model is used to describe various polymeric systems, due to its simplicity and the possibility of its interaction parameter being determined experimentally by calorimetric techniques or, osmotic pressure, for example (Safronov et al., 2009; Kim et al., 2010). The extended equation can be used as an initial model to describe and understand the forces involved in polymer-polymer and polymer-solvent interactions.

After the addition of silk fibroin in sodium alginate solution, for blend preparation, it was possible to notice that the solution became cloudy immediately, even in blends with low polymer concentration, for instance, SF₀.₅SA₀.₅WA₉₉. When observing the blend in optical microscopy, microstructures with a diameter of 0.247 ± 0.120 mm were detected (Figure 2A). The formation of these structures, like drops, was also verified in other protein-polysaccharide systems, such as alginate-casein and alginate-pea protein (Pacek et al., 2000; Antonov and Moldenaers, 2011; Mession et al., 2012). After 1 day, it was possible to notice two phases, shown in Figure 2B. The proportion of each phase changes over the days, the solid-like phase becomes denser, indicating that even if phase separation started after 1 day (macroscopically), the system was not in thermodynamic equilibrium. The evolution of this phenomenon was investigated by static light scattering (Lopes et al., 2018) and the system was considered in equilibrium when changes in the scattering profile were not observed anymore, which happened after the eighth day. It is possible that over time, the microstructures observed in the microscope, collapse and precipitate forming the solid-like phase (Figure 3).

The hydrogel formation, here called the solid-like phase, is a spontaneous phenomenon. It could happen due to the hydrophobic interaction and intra and intermolecular hydrogen bonds, changing the SF conformation from α-helix to β-sheet, precipitating the protein and turning it stable in water (Matsumoto et al., 2006). Some parameters can contribute to the hydrogel formation, such as protein concentration, decreasing pH, chemical crosslinking and temperature (Nogueira et al., 2011; Yin et al., 2017). In previous studies, it was observed that the increase in temperature could accelerate the hydrogel formation from SF solution, as well as, the presence of poly(ethylene oxide) (PEO), a water-soluble polymer (Kim et al., 2004; Nogueira et al., 2011; Zhong et al., 2015; Yin et al., 2017). Based on the literature, SA may have the same influence, accelerating the formation of the solid-like phase.

Table 2 shows the values of zeta potential. that was negative for both polymers. For SF this parameter is close to zero, showing a balance of charges between the amino and carboxyl groups. The pI for SF is around 4, so in neutral pH, there is a predominance of negative charges, which justify the value obtained (Wenk et al., 2011). This zeta potential value also indicates that the SF solution is unstable, based on DLVO theory (Derjaguin and Landau, 1941), which could explains the spontaneous hydrogel formation after some days. The SA zeta potential value was −57.9 mV. At neutral pH, the carboxyl groups are deprotonated (pKa = 3.4–3.6) (Draget et al., 1997), so the polymer chain will be negatively charged.

Analyzing the zeta potential values of each polymer is reasonable to conclude that there will be an electrostatic repulsion between the polymers. At the same time, there will be also an interaction between the amino groups from SF and the carboxyl groups of SA. The results of blends zeta potential show that for the ones with more SF in the composition, the zeta potential is higher showing charge compensation phenomenon.

After phase separation, a phase rich in SA (liquid) and another rich in SF (solid-like) are observed, the experimental results for the composition of each phase are shown in Table 3 and plotted in the ternary graph shown at Figure 4. Water is the most predominant component of the phases, even in the solid-like one, where water is probably entrapped between the polymer chains, and then the hydrogel formation occurs, as proposed in Figure 3. The tie lines are very close to each other and it was observed that each blend presents some variability, that is expected due to the characteristics of SF and SA solutions, which form a dispersion and may have a small variation in the same sample. Besides, there is inherent variability in the measurements, therefore, the establishment of this region is important.

The phase diagram can be analyzed by the method presented by Popova et al. (2008) that is based on two major principles. The first one is that the protein-rich phase is not a single phase but a mixture of two: a supernatant and a “true” precipitate, which cannot be separated. The other is that the extension of the tie lines would converge to one point, which represents the “true” precipitate that in this case would be composed mostly by SF. The phase diagram shows that the tie-lines converge to a point with a high concentration of SF (5.4 wt% of SF, 1.25 wt% of SA, and 93.35 wt% of water). In practice, the “true” precipitate could not be obtained, because, at this concentration, the system would be in the “hydrogel” zone, therefore, it would not be possible to observe two distinguished phases.

It was observed that, for blends with the same SF initial concentration, the increase of SA initial concentration resulted in a higher concentration of this polymer in solid-like phase (Figure 5). It was expected that an increase in the polysaccharide concentration would lead to a more concentrated SF solid-like phase since the SA has more affinity to the solvent and would stay in solution. Thus, probably part of the SA rich phase could be entrapped within the SF rich phase, as observed in Figure 5, as a consequence of SA viscosity (Mession et al., 2012). This behavior is observed in other protein-polysaccharide systems, such as alginate-pea protein and soy protein-κ- carrageenan (Li et al., 2008; Mession et al., 2012).

The three Flory-Huggins parameters of interaction were estimated by solving the non-linear equation system described by Eqs 8–10, the values are reported in Table 4. Using those values and the equilibrium data, new mass fractions were calculated by fixing the composition of two values and using Eq. 8 for both phases. The theoretical mass fractions for each component in liquid and solid-like phases are shown in Table 5, and the ternary phase diagram with the calculated and experimental data is shown in Figure 6. The binodal curve was plotted using the Flory-Huggins model. It was not possible to obtain the experimental binodal curve, because this is, usually, obtained by cloudy point experiments. As previously described, when mixing the silk fibroin solution in sodium alginate solution, the formation of microstructures (like drops) is observed, even in blend with low polymer concentration, which makes it difficult to determinate the cloudy point.