Impact of the in vitro gastrointestinal passage of biopolymer-based nanoparticles on insulin absorption

Abstract

Although the oral administration of insulin is recognized as the safest and most attractive, insulin oral bioavailability is usually reduced due to the susceptibility to acidic and enzymatic degradation in the gastrointestinal (GI) tract and intrinsic low intestinal permeability. Nanoencapsulation of insulin is, thus, foreseen as a promising approach to overcome most of these drawbacks. The effect of the GI environment on the aggregation of alginate/dextran sulfate-based nanometric-sized particles, uncoated or double-coated with chitosan and albumin, and its further influence on insulin release and permeability at the cellular level, was investigated in vitro. The swelling and aggregation behavior of NPs in gastric conditions was accompanied by the prevention of insulin release. In intestinal conditions, the fast dissolution of uncoated NPs was responsible for a wide size distribution and for a burst release of insulin, while the size stability provided by albumin/chitosan-coating led to sustained release. Chitosan/albumin-coated NPs were able to significantly increase the permeability of insulin across the cell-based engineered intestinal models, further enhanced by the presence of a mucus layer and M-like cells. The influence of these models on insulin permeability was compared to the curve that better adjusted to the mathematical kinetics of insulin release from these biopolymeric-NPs. Thus, a correlation between the size behavior of NPs upon passage in the GI tract and both insulin release profile and permeation across intestinal in vitro models was addressed. These results provide proof-of-concept evidence that the GI passage of NPs has a major influence on the oral absorption of macromolecules.

---

Introduction

Diabetes mellitus is a chronic disease that has assumed epidemic dimensions around the world. Insulin administration is indispensable for the treatment of type 1 patients and is required at a later stage in patients with type 2 diabetes. However, due to enzymatic degradation in the GI tract and poor permeability through the intestinal epithelium, which results in poor oral bioavailability,¹ the most common route of insulin administration is the subcutaneous route. Oral administration is considered the safest and the most convenient alternative to deliver insulin, but it faces important challenges. In the last years, different formulation approaches have been explored to encapsulate insulin and deliver it orally. Nevertheless, these formulations have only had limited success and no oral insulin product has been made commercially available yet.

The encapsulation of insulin into polymeric NPs is a powerful tool to improve insulin oral bioavailability. Between the most commonly polymers used to produce NPs for oral delivery of insulin, natural biopolymers such as alginate, dextran sulfate and chitosan have demonstrated in vivo efficacy without toxicological effects when orally administered as NPs.^2–4 Alginate and dextran sulfate are considered a suitable choice, due to their excellent biocompatibility, drug carrying ability, adjustable controlled-release property, lower cost, abundance in nature and easier application.⁵ Previous studies revealed that the presence of a copolymer, such as dextran sulfate enhances the loading of hydrophilic drugs in alginate matrices.⁶ The recent extensive research on several chitosan-based particulate systems also demonstrates a positive role of chitosan towards oral insulin delivery.^7–9

Alginate–dextran sulfate-based NPs (ADS-NPs) coated with chitosan–polyethylene glycol (PEG)–albumin shell have shown efficiency as potential carriers for the oral delivery of insulin.^3,4 The alginate network forms an impermeable structure in acidic gastric conditions,^10,11 preventing premature insulin release in combination with dextran sulfate and increasing insulin protection in combination with albumin, which is applied as a sacrificial target to GI proteases.¹² Chitosan, as polycationic and mucoadhesive polymer is designed to significantly enhance the intestinal permeability of insulin by increasing the residence time at the site of absorption and transiently opening the tight junctions, allowing paracellular transport of insulin across intestinal epithelium.^13,14 This complexed structure acts as a synergistic effect, since at acidic pH, the high solubility of chitosan favored by the conversion of glucosamine units into soluble form R–NH₃⁺ (ref. 15) is inhibited by alginate presence due to the swelling behavior of this polymer at low pH. On the other hand, the rapid dissolution of alginate owing to the dissociation of the carboxylic groups at a higher pH is controlled by chitosan.^16,17

The use of NPs in oral insulin delivery has been increased over the last decades mainly due their smaller size, which allows a higher surface area-to-volume ratio when compared to other delivery systems. Thus, it is important to understand the factors influencing their GI uptake, more precisely their size and aggregation phenomena. The impact of these factors on GI absorption has been assessed mainly with polystyrene and latex particles but in most part of those studies neither the influence of GI tract on NPs properties was the focus nor the in vitro–in vivo correlation was a concern. Particle size is one of the most important properties of NPs related to their biodistribution behavior and can be more or less easily modulated.¹⁸ Actually, one of the main problems with NPs design remains in the fact that intrinsic properties that may help to cross GI tract may also significantly alter further behaviour in the blood.¹⁸ Some studies have quantitatively assessed the oral absorption of NPs, however a thorough understanding about the impact of size and aggregation behavior within the GI tract is still missing due to incomplete particle characterization.¹⁹

So far, very few reports about NPs properties in situ following oral administration and their relation with the percentage of NPs uptake are available. For instance, particle size may suffer in vivo variations within the GI tract due to aggregation, particularly in the gastric pH and intestinal high ionic conditions. Taking into account that the effective particle size at the absorption surface of the GI tract increases with aggregation, a size-dependent absorption can be influenced.¹⁹ Most of the studies associate their bioavailability data with characterization of the “as-dosed” material, but the in vivo aggregation state of NPs has been often ignored as a factor affecting particle uptake.¹⁹

The aim of this work was to study the in vitro effect of the coating of chitosan and albumin on ADS-NPs size and aggregation behavior across GI environment. Furthermore, the effect of these biopolymer-based NPs on enhancing insulin in vitro permeation through the intestinal membrane was also evaluated. In vitro models of Caco-2, Caco-2/HT29-MTX and Caco-2/HT29-MTX/Raji B co-cultures were applied to understand the influence of each type of cell on insulin intestinal permeability.

Materials

Low viscosity sodium alginate (viscosity of 1% solution at 25 °C, 4–12 cps), dextran sulfate, sorbitan monooleate (Span 80®), chitosan (50 kDa), bovine serum albumin, phosphotungstic acid, n-octanol, trifluoroacetic acid (TFA) 99% and acetonitrile (LiChrosolv) high-performance liquid chromatographic (HPLC) grade were purchased from Sigma-Aldrich (Steinheim, Germany). Calcium carbonate was obtained from Setacarb (Omya, France), paraffin oil from Scharlau (Spain), insulin 100 IU mL⁻¹ Actrapid® from Novo Nordisk (Bagsvaerd, Denmark) and poloxamer 188 (Lutrol®F68) from BASF (Ludwigshafen, Germany).

Human colon carcinoma Caco-2, mucus producing HT29-MTX, and human Burkitt's lymphoma Raji B cell lines were obtained from the American Type Culture Collection (ATCC) and used at passages 35–64, 8–29 and 1–13, respectively. Dulbecco's Modified Eagle Medium (DMEM), fetal bovine serum, L-glutamine, non-essential amino acids, 100 U mL⁻¹ penicillin and 100 mg mL⁻¹ streptomycin, trypsin–EDTA and Hank's Balanced Salt Solution (HBSS) were purchased from Gibco (Invitrogen Corporation, Life Technologies, 165 UK).

Methods

Preparation of ADS-NPs

ADS-NPs were prepared by emulsification/internal gelation.²⁰ Briefly, the formation of ADS matrix through emulsion dispersion was followed by triggered instantaneous particle gelation. An aqueous solution of sodium alginate (2%, w/v) and dextran sulfate (0.75%, w/v) was prepared by stirring (100 rpm) overnight. 70 mg of insulin and poloxamer 188 (0.16%, w/v) were added and dissolved. An aqueous suspension of ultrafine calcium carbonate (5%, w/v) was ultrasonicated for 30 min to break up crystal aggregates and dispersed at calcium-alginate ratio of 16.7% (w/w). The resultant dispersion was emulsified within paraffin oil (aqueous phase/oil phase 50/50, v/v) facilitated by sorbitan monooleate (1.84%, v/v) by impeller-stirring homogenization (1600 rpm) and ultrasonication (60% vibration amplitude). After 15 min, gelation was induced by addition of paraffin oil containing glacial acetic acid (molar ratio acid–calcium, 3.5) to solubilize calcium dispersed in the alginate–dextran droplets during 15 min, with continued stirring and ultrasonication. ADS-NPs were recovered through an extraction as described elsewhere.²⁰

Chitosan/PEG/albumin coating

The coatings were applied simply through polyelectrolyte complexation, as previously described.²¹ Chitosan (0.3%, w/v) and PEG solution (PEG–chitosan mass ratio of 3 with calcium level at 1.5%, w/v) dissolved in 0.5% (v/v) lactic acid at pH 4.5 was added dropwise to ADS-NPs under magnetic stirring at 800 rpm for 60 min. Then, chitosan/PEG/CaCl₂-coated ADS-NPs (CS-NPs) were coated by dropwise addition of albumin solution (1.0%, w/v) at pH 5.1 under magnetic stirring during 60 min to form albumin-coated CS-NPs (ALB-NPs). The supernatant containing unbound polymer and protein was separated through centrifugation. NPs before and after coating were submitted to ultrasonication to disaggregate.

NPs characterization

Particle size analysis was performed by laser diffractometry (LD) using a laser diffraction particle size analyzer (Beckman Coulter® LS 13 320, Miami, FL, USA) with polarization intensity differential scattering (PIDS). The real and the imaginary refractive index were set to 1.36 and 0.01, respectively. A Pasteur pipette was used as sampling instrument and the stirring speed was set to 30%. Three measurements of 90 s were used and the size distribution was represented by volume.

Zeta potential was measured by Laser Doppler Electrophoresis using Zetasizer Nano ZS (Malvern Instruments Ltd., UK). Measurements were taken in a folded capillary electrophoresis cell at pH 4.5 and 25 °C with Milli-Q-water. The equipment was routinely checked and calibrated using mobility standard (Beckman Coulter, Inc. Miami, FL, USA).

The shape, surface and mass spectroscopy of NPs were analyzed by Field Environmental (FE)-Cryo-SEM with EDS-JEOL JSM 6301 F scanning microscope (Oxford INCA Energy 350, Gatan Alto 2500, Tokyo, Japan). Phosphotungstic acid-stained particles were frozen using liquid nitrogen slush (−210 °C) under vacuum to allow their fracture in order to obtain a fresh and clean surface for examination. Then, samples were sublimated at −90 °C for 4 min to remove top layers of water molecules. Finally, samples were sputter coated with gold/palladium for 40 s, followed by image capturing.

Determination of octanol/water partition coefficients

The partition coefficients for n-octanol/water of ADS, CS and ALB-NPs were determined by a shaking flask method.²² After lyophilization, 80 mg of NPs were mixed with 9 mL n-octanol solution (the organic and aqueous phases had been mutually saturated for 24 h) and allowed to equilibrate for 3 h. The final mixture was centrifuged for 2 min at 3500g. Insulin was used as NPs probe and was extracted from NPs through dissolution in phosphate buffer (PBS) at pH 7.4 (USP 34) for 3 h in an orbital shaker. After ethanol addition (50/50, v/v) to precipitate alginate, the medium was centrifuged and insulin content in supernatant was measured by HPLC, as previously described.²³ The water partition coefficients (P_o/w) were calculated as follows: P_o/w = C_o/C_w, where C_o and C_w refer to the concentrations of insulin in n-octanol phase and water phase, respectively. Non-encapsulated insulin was used as control.

Insulin encapsulation efficiency (EE) and loading capacity (LC)

Insulin EE was calculated by the difference between the total amount of insulin used to prepare NPs and the amount of insulin that remained loaded to NPs. Insulin-loaded NPs were separated from aqueous supernatant containing free insulin by centrifugation at 12 000g for 10 min at 4 °C. Insulin was extracted from NPs through dissolution in PBS at pH 7.4, as described before. Insulin LC was calculated based on the dry mass of the NPs obtained after lyophilization. The amount of loaded insulin was determined by HPLC.

DSC analysis

To assess the interactions between polymers and confirm their presence in NPs structure, thermograms were obtained using a Shimadzu DSC-50 system (Shimadzu, Kyoto, Japan). 2 mg of lyophilized samples were crimped in a standard aluminum pan and heated from 20 to 350 °C at a heating constant rate of 10 °C min⁻¹ under constant purging of nitrogen at 20 mL min⁻¹.

Size stability in simulated GI fluids

For determination of NPs size and aggregation in the range of pH of simulated GI fluids, the incubation of NPs in simulated gastric fluid (SGF) (USP 34) was firstly done in the range of pH 1.2–4.2, followed by the incubation in simulated intestinal fluid (SIF) (USP 34) at pH between 4.7 and 8.2. The pH in each medium was gradually increased, with a maximum duration of 2 h in each medium. Before and after each pH variation, NPs size was measured by LS as already described. Results are expressed as the mean size (μm) and volume distribution (%) of the main population of NPs.

Insulin release profile in simulated GI fluids and protection against pepsin degradation

For determination of insulin retention/release profile in simulated GI tract, enzyme-free simulated digestive fluids were used to determine the response of insulin-loaded NPs, minimizing enzymatic interferences. 5 mL of NPs suspension were incubated in 10 mL of SGF at 37 °C for 2 h with shaking of 100 strokes per min using a Shaking Water Bath (SS40-D, GRANT, United Kingdom), followed by incubation in 10 mL of SIF for 6 h. Sample aliquots were collected and replaced by the same volume of fresh incubation medium at predetermined times. For determination of insulin released from NPs, samples were centrifuged at 12 000g for 10 min and insulin content in the supernatant was analyzed by HPLC.

Release kinetics of insulin from ALB-NPs was mathematically analyzed (Systat SigmaPlot version 12.1.) to better understand the impact of every coating material on the insulin release behavior.

To assess insulin protective effect of ALB-NPs from pepsin degradation in the GI tract, 5 mL of NPs suspension were incubated in 10 mL of SGF containing pepsin. This effect was determined based on the usual dose of insulin-loaded NPs orally administered (50 IU kg⁻¹) and the amount of proteases in the total fluid volume of the human stomach.^24,25 Insulin protection was determined in a ratio of 1 : 7 (insulin/pepsin) during 2 h at 37 °C with shaking of 100 strokes per min. After incubation, SIF was added to inactivate pepsin and to release insulin retained and protected in the NPs core. Insulin content was determined as described above. The effect of NPs in protecting insulin against pepsin was compared with results obtained with non-encapsulated insulin.

In vitro models

Caco-2 and HT29-MTX cells were grown separately in flasks in DMEM supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) non-essential amino acids, 1% (v/v) L-glutamine and 1% (v/v) of antibiotic/antimitotic mixture (final concentration of 100 U mL⁻¹ penicillin and 100 U mL⁻¹ of streptomycin), at 37 °C under a 5% CO₂ water saturated atmosphere. Upon confluence, cells were harvested from flasks with trypsin–EDTA. Raji B cells were cultured in flasks with DMEM supplemented and with the same conditions as described above. To establish the co-culture in vitro models, Caco-2 and HT29-MTX cells were mixed to a final density of 1 × 10⁵ cells per cm² (90 : 10 proportion). Caco-2 monolayer (1 × 10⁵ cells per cm²) was also seeded as control. Regarding the triple co-culture model, Caco-2 and HT29-MTX cells were seeded as described before. After 14 days, Raji B cells (1 × 10⁶ cells per 6-well) were added to the basolateral compartment of Caco-2/HT29-MTX co-culture and maintained for 4–6 days. All the models were established on Transwell inserts (3 μm) and maintained for 21 days.

In vitro intestinal permeability studies

All cell monolayers were used after 21 days in culture. For all permeability studies, the culture medium was removed and the cell monolayers were washed with pre-warmed HBSS. Permeability experiments were run at 37 °C during 4 h from the apical to the basolateral chambers for both insulin and insulin-loaded NPs, at a predetermined initial apical concentration. At different times, basolateral samples were collected and free-insulin amount was determined by HPLC. During permeability studies, cell monolayer integrity was monitored by transepithelial electrical resistance (TEER) measurement using a volt-ohm meter Millicell® ERS-2 (Millipore, USA).

The fraction of insulin released in SIF from ALB-NPs was plotted against the fraction of insulin that permeated across the three intestinal in vitro models along 4 h to better assess the potential and limitations of every cell culture model.

Statistical analysis

All the experiments were performed in triplicate and are represented as mean ± standard deviation. Statistical evaluation was performed with a one-way ANOVA followed by Bonferroni post hoc test. A p < 0.05 was taken as the criterion of significance. The level of significance was set at probabilities of *p < 0.05, **p < 0.01, and ***p < 0.001.

Results and discussion

NPs characterization

Particle properties such as mean size, size distribution, charge and hydrophilicity are important parameters in particle uptake or barrier crossing, since the first contact between cell membrane and particle influences their interaction and possible uptake. Therefore, particle size distribution, charge and hydrophilicity should be fully characterized.²⁶ In this study, multilayered NPs encapsulating insulin were formed by alginate and dextran sulfate nucleating around calcium ions layered with chitosan/PEG/CaCl₂, and subsequently coated with albumin. In order to evaluate the presence or nano- or microaggregates, NPs size was measured by LD. ADS-NPs had a mean diameter of 233 ± 0.31 nm and zeta potential of −28.4 ± 0.89 mV. Further coating of ADS-NPs with chitosan considerably increased NPs size and a new population higher than 1 μm appeared probably due to aggregation phenomena, which may be caused by the stickiness of the polymer.^27,28 Polycationic chitosan coating decreased the negative surface charge to −25.0 ± 0.82 mV, confirming the deposition of the positively charged polymer on the surface of NPs. Following albumin coating, it was observed a displacement of the size distribution curve to the left with a mean diameter of 166 ± 0.84 nm and zeta potential of −22.1 ± 0.76 mV. This behavior may be due to the composition of the albumin, which is mainly composed by α-helical peptides, resulting in a higher packaging of the coating membrane and also of the polymeric matrix.²¹

Chitosan layering is essential to stabilize the NPs core and interact with albumin that forms the sacrificial coating to protect insulin against proteases. NPs with reduced size are known to establish improved contact with intestinal cells than do bigger particles, facilitating permeation via different mechanisms.²⁹ Furthermore, since NPs immobilized by mucus can be cleared from the mucosal tissue, their size must be less than 500 nm (ref. 30) to avoid significant steric inhibition by the fiber mesh and adhesion to mucin fibers.³¹ The decrease of particle size after albumin coating may be propitious for NPs escape from this effect. Insulin initially added to the formulation was almost completely encapsulated (EE = 98.74 ± 0.18%), and the LC was 2.43 ± 0.04%. These values seemed to be higher compared to previous formulations prepared by this technique,^4,32 which could be justified by the use of a co-surfactant and ultrasonication assistance during emulsification, leading to a smaller and more stable nanoemulsion. Fig. 1 shows the Cryo-SEM images of all formulations. These results indicate that the particles are in the nanometer scale, according to what was obtained by LS technique. The droplet shape of the CS- and ALB-NPs was quasi-circular and had a smooth profile, while ADS-NPs presented a more elongated shape. The population of ALB-NPs represented in Fig. 1C tend to have a spherical shape with sizes ranging between 300 and 650 nm. The cellular internalization of different types of NPs has been described to be function of both size and shape curvature, wherein nanoscale cylindrical particles had a higher percentage of cellular internalization over time.^33,34

Fig. 1 Cryo-SEM images of (A) ADS-NPs, (B) CS-NPs and (C1–C3) ALB-NPs; scale bar = 1 μm.

Hydrophilicity has been described as an important parameter in determining the transport of NPs across the mucus³⁵ and M cells.³⁶ This expectation was not investigated in this study, since according to the hydrophilic nature of polymers, NPs were extremely hydrophilic before and after the coating and no insulin was detected in the octanol phase. Accordingly, all the content of non-encapsulated insulin was also detected in the aqueous phase. To date, only a few studies have correlated cellular uptake with hydrophilicity of albumin NPs. Some authors have demonstrated that higher hydrophilicity is reflected in a decrease of GI uptake.^37–39 However, the increase of hydrophilicity was also correlated with a change in the surface charge. Therefore, the results could not be conclusive, since the variation of surface charge may also have an impact at the cellular level. More recently, Gaumet et al. studied the effect of hydrophilicity independently of other parameters by using particles with the same size and charge, differing only in surface hydrophilicity.²⁶ They demonstrated that the surface hydrophilicity is a determinant factor for NPs uptake and more specifically the negatively charged 100 or 300 nm chitosan-coated PLGA particles with high hydrophilicity appear to be the best candidates to target intestinal cells. These findings highlight the potential interaction between our negatively charged and hydrophilic NPs with intestinal cells.

DSC analysis

The thermograms of alginate, dextran sulfate and chitosan, as we can see in Fig. 2A, showed initial endothermic peaks at 84.3, 85.6 and 76.8 °C and higher exothermic peaks at 245.8, 204.5 and 308.2 °C, respectively. Endothermic peaks are correlated with loss of water associated to hydrophilic groups of polymers while exothermic peaks resulted from degradation of polymers due to dehydration and depolymerization reactions most probably to the partial decarboxylation of the protonated carboxylic groups and oxidation reactions of the polymers.^40,41 Thermogram of alginate/dextran sulfate/chitosan physical mixture is also represented in Fig. 2A and showed an endothermic peak at 90.2 °C, which was sifted to the right and an exothermic peak at 296.6 °C, which probably represents the coalescence of the endothermic polymer peaks. Peaks of isolated polymers were different from those of the physical mixture probably because complexation of polymers resulted in new chemical bonds. Thermograms of NPs are represented in Fig. 2B. Shifts on endothermic and exothermic peaks of NPs represent the ionic interactions between polymers, which led to the formation of new chemical entities with different thermal and absorption properties.⁴¹ For instance, the exothermic peak of ADS-NPs was registered at 271.9 °C, broader and deviated to the left when compared with isolate polymers (alginate and dextran sulfate). After chitosan coating, exothermic peak of CS-NPs shifted to 289.3 °C, a higher peak value than ADS-NPs, which was interpreted as chitosan interaction with alginate and dextran sulfate. Further albumin coating was responsible for broader peaks, although at similar temperatures.

Fig. 2 Thermograms of (A) alginate (black solid line), dextran sulfate (grey solid line), chitosan (dashed line) and alginate/dextran sulfate/chitosan physical mixture (dotted line); (B) ADS-NPs (solid line), CS-NPs (dashed line) and ALB-NPs (dotted line).

Size stability in simulated GI fluids

Taking into account that this nanoparticulate drug delivery system is intended for oral administration, a critical study of these NPs would be the evaluation of their size stability in the different media present in the GI tract. Particle size is considered as a critical parameter⁴² but its impact on uptake has not been carefully studied with biodegradable particles.²⁶ The interactions between the NPs and the gastric and intestinal pH and ionic conditions should be evaluated to analyze their impact on NPs size, which consequently may influence the transport of NPs across the intestinal epithelium.²⁶ In this way, a study comprising a wide range of pH values was set in order to increase in vitro/in vivo correlation. Therefore, ADS-NPs were compared with coated-NPs (CS- and ALB-NPs) with regard to their size, monitored during the passage of NPs in simulated GI fluids.

As shown in Fig. 3, under simulated gastric pH there was a significant increase in the mean particle diameter for all formulations but different trends were observed with an extensive aggregation of ADS- and CS-NPs, but only moderate aggregation of ALB-NPs.

Fig. 3 Size distribution of (A) ADS-NPs, (B) CS-NPs, and (C) ALB-NPs during 2 h in contact with SGF at pH range 1.2–4.2 and with SIF at pH range 4.7–8.2. Error bars represent the mean ± standard deviation (n = 3).

When ADS-NPs were in SGF at pH 1.2, a big deviation of size distribution to the micrometer range (∼87 μm) was observed (Fig. 3A), demonstrating a strong influence of acid pH on particle size distribution. The destabilization of protein-based colloidal delivery systems under simulated gastric conditions may occur for a number of reasons, including loss of charge due to pH changes and electrostatic screening due to an increased ionic strength.⁴³ Alginate is usually in a crosslinked gel conformation with Ca²⁺ in NPs. However, with the pH decrease, the presence of monovalent cations (H⁺ ions) can displace the bound between Ca²⁺ and G-blocks of alginate.⁴⁴ Therefore, alginate chains that were previously very close and interlocked became free to take their minimum energy conformation. COO⁻ still present in alginic chains led to repulsive forces and alginate chains took a more linear conformation. Furthermore, dextran sulfate groups were predominantly only charged once (HSO₄⁻) and the presumable interaction between this anionic polymer with Ca²⁺ became weaker, leading to a decrease of NPs network compaction. The charge of insulin at such low pH is highly positive and thus electrostatic attractions with polymers could have been replaced, at least partially, by electrostatic repulsions, also contributing to the change of NPs conformation. When the pH increased to 3.2, two similar populations were observed (∼1.3 and 50 μm), as can be noticed by the decrease of volume% of the main population. The reversibility of alginate gel may have occurred, reorganizing the egg-box structure. However, there were other competitive cations in the medium, such as H⁺ and Na⁺, that can destabilize the calcium-alginate gel and therefore two populations were still present.⁴⁴ At pH 3.2, alginate gets closer to its pK_a (pK_a1 ∼ 3.38 and pK_a2 ∼ 3.65) which led to a decrease of alginate molecules in acidic conformation (COOH) towards 50% and the other 50% in basic conformation (COO⁻), resulting in two different sized populations. With alginate simultaneously in two different conformations, it was expected that the distribution profile would also be splitted in two populations. The bigger population was probably the remaining of COOH groups of alginate that tend to change the conformation of NPs and led to particles aggregation, while the smaller population may be related to alginate in COO⁻ conformation ready to interact with cations in the medium. Above this pH, the % of the small population (∼1.3 μm) started to increase while the bigger population (∼50 μm) disappeared.

In contrast to alginate behavior, chitosan is soluble at low pH due to protonation of amino groups and insoluble at higher pH values.⁴⁵ The overall behavior of CS-NPs size in gastric pH (Fig. 3B) was similar to ADS-NPs. At low pH, chitosan is strongly protonated (NH₃⁺) and soluble, therefore some polymer could be free in the medium and ADS-NPs became exposed. Furthermore, alginate is less negatively charged and thus electrostatic interactions with chitosan are weaker. As for ADS-NPs, a shift to the micrometer range was observed at pH 1.2, although less pronounced (∼25 μm). The authors explain this difference by the stabilization chitosan offers to NPs. The interaction of chitosan with negative groups of alginate prevented the interaction of those groups with H⁺ and Na⁺ of the medium, preventing the change of NPs structure into linear conformation. Furthermore, some studies have demonstrated that the presence of PEG, a component of CS-NPs, improves the stability of NPs, providing a hydrated steric barrier.^46,47 Therefore, the remained stability of PEG in gastric conditions may protect CS-NPs, preventing the rapid dissolution of chitosan. At pH 4.2, the interaction between chitosan and alginate becomes stronger, since there is more alginate in COO⁻ conformation to interact with NH₃⁺ groups of chitosan. Particle started to shrink and the smaller population (∼1.3 μm) greatly increased.

With the last coating of albumin, NPs showed a much more stable behavior. One population around 20 μm between pH 1.2 and 3.2 (Fig. 3C) was observed, which may be due to particles swelling. At pH 4.2, an increase of the smaller population of ∼1 μm was noticed (represented by the decrease of volume% of the main population), as electrostatic interactions between albumin and chitosan became stronger.

From these data, it can be concluded that the acidic pH of the gastric medium is a key factor in the aggregation process of these NPs, which are pH sensitive.

In order to provide a better insight of what happens in in vivo conditions in terms of NPs size variation, NPs were further placed in SIF with pH increments over time.

In general, NPs size distribution was more stable in intestinal pH. The size of ADS-NPs was mainly unaffected even if a bigger population (∼20 μm) was observed (data not shown). Above pH 5.7, both anionic polymers and insulin are negatively charged and repulsive forces could occur. The dissolution of the alginate nucleus as a result of calcium loss started to occur as the pH increased and the destabilization of NPs structure may be responsible for a wide size distribution.

The behavior of ADS-NPs and CS-NPs in SIF was also very alike, except for the fact that the smaller population was more pronounced in ADS- (Fig. 3A) than in CS-NPs (Fig. 3B), probably due to the effect of chitosan, preventing the burst dissolution of alginate.

Contrarily to ADS and CS-NPs, the variations of the ALB-NPs size in intestinal pH were very slight (Fig. 3C). The % of the population in the nanometer scale increased over time and the size distribution curve became narrower, which is probably due to the stronger interaction between chitosan and albumin. The helical structure of albumin might provide a higher packaging membrane coating and polymeric matrix of ALB-NPs, stabilizing their size.²¹

Insulin release profile in simulated GI fluids and protection against pepsin degradation

The in vitro insulin release tests intended to predict the release profiles of the hormone in conditions similar to the GI tract (stomach and small intestine). Firstly, the NPs were added to SGF at pH 1.2, which mimics the gastric environment. Then, after 2 h, the release medium was changed to SIF at pH 6.8, which simulates the transit of the NPs to the small intestine. Insulin release profile from NPs in gastric and intestinal conditions is shown in Fig. 4. In SGF, ADS- and CS-NPs retained insulin during 2 h, while ALB-NPs released approximately 3% of insulin. As albumin interacts with chitosan that is on its turn complexed with the matrix, this protein network is somehow destabilized, which in turn can destabilize NPs structure and allow the diffusion of insulin superficially-located to the exterior of NPs. In the acidic environment, the charge of insulin (isoelectric point pI = 5.3)⁴⁸ is highly positive, and thus, the interactions with alginate and dextran sulfate prevent insulin release. Moreover, calcium forms ionic cross-links between alginate polymer chains transforming the sol into a pre-gel state, which was responsible for the swelling behavior of NPs previously observed in gastric pH. These interactions are likely to form a stable barrier to retain insulin and limit its release from NPs.⁴⁵ When in contact with SIF, different release profiles of insulin from uncoated and coated-NPs were observed. In comparison with the dissolution profile of coated-NPs, a very rapid release behavior of insulin was observed from ADS-NPs. More than 50% of the insulin was released in the first 5 min and approximately 100% was released after 1 h. Electrostatic repulsions at intestinal pH was achieved between negatively charged ADS-NPs and insulin, which promoted insulin release. This fast release of insulin from ADS-NPs is in accordance with the rapid dissolution of ADS-NPs previously observed, that led to the destabilization of NPs structure and therefore to a wide size distribution. Compared to ADS-NPs, the release rate of insulin from coated-NPs was slower, with approximately 35% and 45% of insulin released after 1 h for CS-NPs and ALB-NPs, respectively. In the case of CS-NPs, a 3 hour sustained release in SIF was observed. The sustained release found in coated-NPs may be efficient in protecting insulin in intestinal environment minimizing insulin loss due to enzymatic attack, thereby increasing its available amount to be uptaken by epithelial cells and promoting successful systemic delivery of insulin in vivo. The impact of NPs coating on insulin release in SIF can be explained by the synergistic effect between alginate and chitosan. The high solubility of chitosan in acidic pH is inhibited by alginate presence, since alginate is insoluble at low pH.⁴⁹ On the other hand, the rapid dissolution of alginate in a higher pH is controlled by chitosan, given its stability at high pH.^16,17 The prevention of the burst dissolution of alginate by chitosan had an impact on size distribution of NPs (Fig. 3B) and was also reflected in the sustained release profile of insulin from CS-NPs (Fig. 4). These results are in agreement with other reports in which chitosan has been described to be able to decrease the burst release effect of the encapsulated drugs.^50,51

Fig. 4 Insulin release profile in SGF (pH 1.2) for 2 h followed by 6 h in SIF (pH 6.8) at 37 °C of ADS-NPs, CS-NPs and ALB-NPs. Error bars represent mean ± standard deviation (n = 3).

ALB-NPs had a different behavior, as after 15 min in contact with SIF a burst release of insulin was followed by a slow release up to 3 h, reaching 100% of total release. Insulin release mechanism from NPs depends on the solubility of insulin and also on the swelling and erosion characteristics of the NPs matrix polymers.^52,53 Initially, the release is due to the part of insulin that is located at or near the interface of NPs. The small particle size may have contributed to the faster drug release as well.

Generally, in many experimental conditions, the mechanism of drug release from swellable polymeric-based NPs follows a non-Fickian behavior.⁵⁴ Drug release mechanism from erodible hydrophilic polymeric matrices is a complex process, because several physical factors are involved, such as penetration of the dissolution medium into the NPs matrix with consequent swelling and erosion of the NPs and dissolution of the drug.⁵⁵ In such type of release mechanism, Peppas model is usually fitted. Therefore, in the present study, to determine the actual mechanism of insulin release in SIF from ALB-NPs, the parameter ‘n’ of the Peppas equation was calculated. The correlation coefficient value was found to be 0.35 (R = 0.94); this indicates diffusion controlled rather than non-Fickian diffusion mechanism⁵⁶ from ALB-NPs.

NPs at low pH collapsed forming an impermeable network structure by the presence of alginate,¹⁰ retaining and potentially protecting insulin against acidic and proteolytic degradation. The presence of albumin, which acts as a sacrificial target, provided full protection to insulin-loaded ALB-NPs in the SGF with pepsin, in contrast with the fast degradation of unprotected non-encapsulated insulin that occurred almost after 10 min (Fig. 5). The structure of ALB-NPs was able to prevent enzymatic degradation of insulin during 2 h.

Fig. 5 Remaining ratio of insulin after the incubation of free insulin and insulin-loaded ALB-NPs in SGF (pH 1.2) with pepsin at 37 °C. Error bars represent mean ± standard deviation (n = 3).

Permeability studies

To study the effect of goblet and M-cells in the permeability of insulin encapsulated into NPs, Caco-2, Caco-2/HT29-MTX and Caco-2/HT29-MTX/Raji B models were used. This last model mimics the most important features of small intestine, namely the presence of enterocytes (Caco-2 cells), the mucus-producing goblet cells (HT29-MTX) in physiological proportions (90 : 10) and the induction of M-like cells, as described elsewhere.⁵⁷ M cells have an important role in intestinal absorption of drugs since they are specialized for antigen and microorganisms uptake, providing a possible gateway for the absorption of proteins, as well as for NPs.³⁹ Together with Caco-2 and HT29-MTX cells, they form a monolayer where cells are joined to each other by tight junctions, mimicking the intestinal epithelium.⁵⁸

Permeation profiles of insulin through the intestinal in vitro models are presented as cumulative transport over time in Fig. 6. In all monolayers, Caco-2, Caco-2/HT29-MTX and Caco-2/HT29-MTX/Raji B, insulin permeation when loaded to ALB-NPs was significantly higher than insulin in solution. According to the release profile of insulin from ALB-NPs, a higher increment of insulin permeation was observed in the first 15 min. In the following 3 h in SIF, insulin was totally released from ALB-NPs in a sustained pattern, which was reflected in the permeability of insulin across monolayers. Contrarily to the results obtained for ALB-NPs, non-encapsulated insulin permeability pattern was more constant over time, almost reaching a plateau after 30 min, even if the protein was totally available to permeate the monolayers since the beginning of the experiment. The most prominent effect of ALB-NPs in insulin permeation was observed with Caco-2/HT29-MTX and triple models. The higher permeation of insulin through the Caco-2/HT29-MTX model was previously demonstrated.^27,59 The major limitations for the intestinal absorption of macromolecules, such as insulin, are the tight junctions and the protective mucus layer.^60,61 The incorporation of insulin into NPs showed an enhancement in the permeability across cell monolayers, which is probably due to chitosan properties. Chitosan is a nontoxic cationic polysaccharide exhibiting mucoadhesive properties along with transient opening of the tight junctions, thereby enhancing the permeation across the intestinal epithelium.^62–64 Furthermore, the mucus layer that may act as an enzymatic barrier to insulin absorption was overcome by albumin coating.¹² An intimate contact of nanoencapsulated insulin to the intestinal mucosa can be described by the diffusion theory of mucoadhesion as defined by Peppas and Carr,⁶⁵ related to interpenetration and entanglement of polymer chains within the mucus layer. Thus, the addition of mucus-producing cells to the monolayer provided a better simulation of natural conditions, which facilitate insulin permeation.³

Fig. 6 In vitro cumulative permeability profiles of insulin across in vitro models. The level of significance was set at probabilities of *p < 0.05, **p < 0.01, and ***p < 0.001. Error bars represent mean ± standard deviation (n = 3).

The permeability across the triple co-culture was performed in order to understand the role of M-like cells in terms of insulin transport. Similar results were obtained with the Caco-2/HT29-MTX and triple co-culture models, as both significantly enhanced insulin permeation, representing 1.3 and 1.4-fold enhancement of the insulin permeation coefficient, respectively, in comparison to Caco-2 monolayer. These results suggested that both HT29-MTX cells, and probably the mucus produced by these cells, and M-like cells were involved in the increase of insulin permeability. The positive role of goblet cells was previously demonstrated for poorly absorbed hydrophilic drugs,^59,66 designating higher paracellular permeability in the co-culture system.

The permeability of insulin-loaded ALB-NPs across the established in vitro models showed to be associated with the in vitro insulin release mechanism of the nanoparticulate system (Fig. 7). The most common models were fitted to our experimental results (zero-order, first-order, Higuchi and Weibull model). Regression analysis verified that the Higuchi kinetic model was the most appropriate for describing the kinetic of insulin release as confirmed by the correlation coefficient value (0.93). It revealed that the release of insulin from NPs occurred in a controlled manner. This in vitro correlation between the amount of insulin that is released from NPs and its permeability through models that simulate the intestinal epithelium contributes to a better understanding of the influence of the GI passage in the intestinal permeation of macromolecules and NPs. These studies are difficult to implement in vivo and according to the best of our knowledge this is the first approach of the influence of GI passage at the level of intestinal absorption of insulin NPs with antihyperglycemic properties. The results showed that the transfer speed of insulin from NPs is not limiting for the in vitro permeability of insulin, free or encapsulated. The higher permeability rates of insulin across the co-culture models during the first 2 h test, when compared to the Caco-2 model, seemed to have a higher correlation with the curve that better adjusted to the mathematical kinetics of insulin release from ALB-NPs. After 3 h, almost all insulin was available to permeate across monolayers, but no statistical differences were observed between the in vitro models. Therefore, during the first 2 h of contact with the intestinal environment, a closer relation was found between the fraction of insulin released from NPs and the fraction that permeated the monolayers, especially when HT29-MTX and M-like cells were included.

Fig. 7 Comparison between the fraction of released vs. permeated insulin for 4 h-testing for ALB-NPs. Data from insulin in vitro release in SIF (black circles) was linearized (black line) according to the mathematical model designed to fit the release profile and plotted against permeability through Caco-2 (black bars), Caco-2/HT29-MTX (red bars) and Caco-2/HT29-MTX/Raji B (green bars) models. All the data sets were compared to the permeation of insulin across Caco-2 monolayer. The level of significance was set at probabilities of *p < 0.05, **p < 0.01, and ***p < 0.001.

The gap time that exists between the moment when insulin is released from NPs and the moment it permeates through the intestinal epithelium should be the lowest possible, since the longer insulin resides in contact with the mucus layer and intestinal enzymes, the more it could degrade, lose bioactivity and escape without being absorbed. Therefore, a fast release of insulin from NPs might be desirable to ensure that a sufficient amount reaches the target site and provide therapeutic effects.⁶⁷ The values of TEER measured before and after the experiments were stable during 4 h (results not shown), demonstrating the integrity of the monolayers.

Conclusions

In the present study, a biopolymer-based nanoparticulate system was developed to enhance the oral absorption of insulin. The system is composed of an alginate/dextran sulfate-core with spherical shape, negative zeta potential values and high EE. The double coating with chitosan and albumin was successfully achieved, taking advantage of their protective, mucoadhesive and absorption-enhancing properties. Monitoring of the impact of the GI environment in NPs size and aggregation suggested the pH sensitivity of these biopolymers. ALB-NPs showed to provided better size stability in GI conditions with less aggregation and more uniformity of size distribution, preventing the release of the majority of insulin in gastric pH and sustaining the release in intestinal pH to allow insulin to be released in the sites of absorption. Such results indicated that when given orally, minimal insulin would be released from NPs in the stomach while the remaining majority could achieve a sustained and complete release in the intestine following 3 h of administration. In cellular experiments, ALB-NPs were found to significantly increase the permeability of insulin across the in vitro models, a pattern that was in accordance with the insulin release profile. The permeability of insulin across the intestinal epithelium was more prominent in the presence of the mucus layer and M-like cells, revealing their positive role on insulin absorption. Overall, our results suggest that these biopolymer-based multilayered NPs are a promising carrier towards the oral delivery of insulin.

Conflicts of interest

The author reports no conflicts of interest in this work.

Acknowledgements

The authors thank Fundação para a Ciência e a Tecnologia (FCT) of Portugal (SFRH/BD/79123/2011). This work was financed by the European Regional Development Fund (ERDF) through the Programa Operacional Factores de Competitividade – COMPETE, by Portuguese funds through FCT in the framework of the project PEst-C/SAU/LA0002/2013, and co-financed by the North Portugal Regional Operational Programme (ON.2 – O Novo Norte) in the framework of project SAESCTN-PIIC&DT/2011, under the National Strategic Reference Framework (NSRF).

References

1. G. P. Carino and E. Mathiowitz, Adv. Drug Delivery Rev., 1999, 35, 249–257.
2. C. Woitiski, R. Neufeld, F. Veiga, R. Carvalho and I. Figueiredo, Eur. J. Pharm. Sci., 2010, 41, 556–563.
3. C. B. Woitiski, B. Sarmento, R. A. Carvalho, R. J. Neufeld and F. Veiga, Int. J. Pharm., 2011, 412, 123–131.
4. C. P. Reis, F. J. Veiga, A. J. Ribeiro, R. J. Neufeld and C. Damgé, J. Pharm. Sci., 2008, 97, 5290–5305.
5. F. Stops, J. T. Fell, J. H. Collett and L. G. Martini, Int. J. Pharm., 2008, 350, 301–311.
6. M. Abbaspour, B. Makhmalzadeh, Z. Arastoo, A. Jahangiri and R. Shiralipour, Trop. J. Pharm. Res., 2013, 12, 477–482.
7. F. Cui, F. Qian, Z. Zhao, L. Yin, C. Tang and C. Yin, Biomacromolecules, 2009, 10, 1253–1258.
8. X. Li, S. Guo, C. Zhu, Q. Zhu, Y. Gan, J. Rantanen, U. L. Rahbek, L. Hovgaard and M. Yang, Biomaterials, 2013, 34, 9678–9687.
9. N. Shrestha, M.-A. Shahbazi, F. Araújo, H. Zhang, E. M. Mäkilä, J. Kauppila, B. Sarmento, J. J. Salonen, J. T. Hirvonen and H. A. Santos, Biomaterials, 2014, 35, 7172–7179.
10. M. George and T. E. Abraham, J. Controlled Release, 2006, 114, 1–14.
11. N. Kamiya and A. M. Klibanov, Biotechnol. Bioeng., 2003, 82, 590–594.
12. C. Reis, A. Ribeiro, S. Houng, F. Veiga and R. Neufeld, Eur. J. Pharm. Sci., 2007, 30, 392–397.
13. P. Fonte, F. Andrade, F. Araújo, C. Andrade, J. d. Neves and B. Sarmento, in Chapter fifteen – Chitosan-Coated Solid Lipid Nanoparticles for Insulin Delivery, ed. D. Nejat, Academic Press, 2012, pp. 295–314.
14. K. Sonaje, K.-J. Lin, M. T. Tseng, S.-P. Wey, F.-Y. Su, E.-Y. Chuang, C.-W. Hsu, C.-T. Chen and H.-W. Sung, Biomaterials, 2011, 32, 8712–8721.
15. C. Qin, H. Li, Q. Xiao, Y. Liu, J. Zhu and Y. Du, Carbohydr. Polym., 2006, 63, 367–374.
16. M. Fan, Q. Hu and K. Shen, Carbohydr. Polym., 2009, 78, 66–71.
17. W. Borchard, A. Kenning, A. Kapp and C. Mayer, Int. J. Biol. Macromol., 2005, 35, 247–256.
18. E. Roger, F. Lagarce, E. Garcion and J.-P. Benoit, Nanomedicine, 2010, 5, 287–306.
19. G. Hinkley, P. Carpinone, J. Munson, K. Powers and S. Roberts, Part. Fibre Toxicol., 2015, 12, 1–13.
20. M. A. Lopes, B. Abrahim-Vieira, C. Oliveira, P. Fonte, A. M. T. Souza, T. Lira, J. A. D. Sequeira, C. R. Rodrigues, L. M. Cabral, B. Sarmento, R. Seiça, F. Veiga and A. J. Ribeiro, Int. J. Nanomed., 2015, 10, 5865–5880.
21. M. Lopes, B. A. Abrahim, C. R. Rodrigues, F. Veiga, L. Cabral and A. J. Ribeiro, J. Chitin Chitosan Sci., 2013, 1, 173–176.
22. J. L. Vázquez, S. Merino, Ò. Doměnech, M. Berlanga, M. Viñas, M. T. Montero and J. Hernández-Borrell, Int. J. Pharm., 2001, 220, 53–62.
23. B. Sarmento, A. Ribeiro, F. Veiga and D. Ferreira, Biomed. Chromatogr., 2006, 20, 898–903.
24. M. Vertzoni, J. Dressman, J. Butler, J. Hempenstall and C. Reppas, Eur. J. Pharm. Biopharm., 2005, 60, 413–417.
25. E. L. McConnell, H. M. Fadda and A. W. Basit, Int. J. Pharm., 2008, 364, 213–226.
26. M. Gaumet, R. Gurny and F. Delie, Eur. J. Pharm. Sci., 2009, 36, 465–473.
27. P. Fonte, T. Nogueira, C. Gehm, D. Ferreira and B. Sarmento, Drug Delivery Transl. Res., 2011, 1, 299–308.
28. B. Sarmento, D. Mazzaglia, M. C. Bonferoni, A. P. Neto, M. do Céu Monteiro and V. Seabra, Carbohydr. Polym., 2011, 84, 919–925.
29. A. T. Florence, Drug Discovery Today: Technol., 2005, 2, 75–81.
30. D. A. Norris, N. Puri and P. J. Sinko, Adv. Drug Delivery Rev., 1998, 34, 135–154.
31. S. K. Lai, Y.-Y. Wang and J. Hanes, Adv. Drug Delivery Rev., 2009, 61, 158–171.
32. A. C. Santos, J. Cunha, F. Veiga, A. Cordeiro-da-Silva and A. J. Ribeiro, Carbohydr. Polym., 2013, 98, 1397–1408.
33. M. Caldorera-Moore, N. Guimard, L. Shi and K. Roy, Expert Opin. Drug Delivery, 2010, 7, 479–495.
34. S. E. A. Gratton, P. A. Ropp, P. D. Pohlhaus, J. C. Luft, V. J. Madden, M. E. Napier and J. M. DeSimone, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 11613–11618.
35. M. Verma, S. Liu, Y. Chen, A. Meerasa and F. Gu, Nano Res., 2012, 5, 49–61.
36. M. Gaumet, R. Gurny and F. Delie, Int. J. Pharm., 2010, 390, 45–52.
37. A. M. Hillery and A. T. Florence, Int. J. Pharm., 1996, 132, 123–130.
38. K. Y. Win and S.-S. Feng, Biomaterials, 2005, 26, 2713–2722.
39. A. des Rieux, E. G. E. Ragnarsson, E. Gullberg, V. Preat, Y. J. Schneider and P. Artursson, Eur. J. Pharm. Sci., 2005, 25, 455–465.
40. T. Mimmo, C. Marzadori, D. Montecchio and C. Gessa, Carbohydr. Res., 2005, 340, 2510–2519.
41. B. Sarmento, D. Ferreira, F. Veiga and A. Ribeiro, Carbohydr. Polym., 2006, 66, 1–7.
42. A. T. Florence, Pharm. Res., 1997, 14, 259–266.
43. U. Lesmes and D. J. McClements, Food Hydrocolloids, 2012, 26, 221–230.
44. X. Wang and H. Garth Spencer, Polym. J., 1998, 39, 2759–2764.
45. C. B. Woitiski, R. J. Neufeld, A. J. Ribeiro and F. Veiga, Acta Biomater., 2009, 5, 2475–2484.
46. A. des Rieux, V. Fievez, M. Momtaz, C. Detrembleur, M. Alonso-Sande, J. Van Gelder, A. Cauvin, Y. J. Schneider and V. Preat, J. Controlled Release, 2007, 118, 294–302.
47. M. Garcia-Fuentes, D. Torres and M. J. Alonso, Colloids Surf., B, 2003, 27, 159–168.
48. Y. W. Chien, Drug Dev. Ind. Pharm., 1996, 22, 753–789.
49. A. Chaudhury and S. Das, AAPS PharmSciTech, 2011, 12, 10–20.
50. S. Gupta, A. Jain, M. Chakraborty, J. K. Sahni, J. Ali and S. Dang, Drug Delivery, 2013, 20, 237–246.
51. F. Araújo, N. Shrestha, M.-A. Shahbazi, P. Fonte, E. M. Mäkilä, J. J. Salonen, J. T. Hirvonen, P. L. Granja, H. A. Santos and B. Sarmento, Biomaterials, 2014, 35, 9199–9207.
52. X. G. Zhang, D. Y. Teng, Z. M. Wu, X. Wang, Z. Wang, D. M. Yu and C. X. Li, J. Mater. Sci.: Mater. Med., 2008, 19, 3525–3533.
53. J. Liu, S. M. Zhang, P. P. Chen, L. Cheng, W. Zhou, W. X. Tang, Z. W. Chen and C. M. Ke, J. Mater. Sci.: Mater. Med., 2007, 18, 2205–2210.
54. P. Li, Y.-N. Dai, J.-P. Zhang, A.-Q. Wang and Q. Wei, Int. J. Biomed. Sci., 2008, 4, 221–228.
55. R. C. Nagarwal, D. N. Ridhurkar and J. K. Pandit, AAPS PharmSciTech, 2010, 11, 294–303.
56. P. L. Ritger and N. A. Peppas, J. Controlled Release, 1987, 5, 23–36.
57. F. Araújo and B. Sarmento, Int. J. Pharm., 2013, 458, 128–134.
58. B. Sarmento, F. Andrade, S. B. d. Silva, F. Rodrigues, J. das Neves and D. Ferreira, Expert Opin. Drug Metab. Toxicol., 2012, 8, 607–621.
59. F. Antunes, F. Andrade, F. Araújo, D. Ferreira and B. Sarmento, Eur. J. Pharm. Biopharm., 2013, 83, 427–435.
60. M.-C. Chen, K. Sonaje, K.-J. Chen and H.-W. Sung, Biomaterials, 2011, 32, 9826–9838.
61. O. Pillai and R. Panchagnula, Drug Discovery Today, 2001, 6, 1056–1061.
62. M.-C. Chen, F.-L. Mi, Z.-X. Liao, C.-W. Hsiao, K. Sonaje, M.-F. Chung, L.-W. Hsu and H.-W. Sung, Adv. Drug Delivery Rev., 2013, 65, 865–879.
63. L.-W. Hsu, P.-L. Lee, C.-T. Chen, F.-L. Mi, J.-H. Juang, S.-M. Hwang, Y.-C. Ho and H.-W. Sung, Biomaterials, 2012, 33, 6254–6263.
64. K. Sonaje, E. Y. Chuang, K. J. Lin, T. C. Yen, F. Y. Su, M. T. Tseng and H. W. Sung, Mol. Pharm., 2012, 9, 1271–1279.
65. N. A. Peppas and D. A. Carr, Chem. Eng. Sci., 2009, 64, 4553–4565.
66. E. Walter, S. Janich, B. J. Roessler, J. M. Hilfinger and G. L. Amidon, J. Pharm. Sci., 1996, 85, 1070–1076.
67. M. Dinarvand, M. Kiani, F. Mirzazadeh, A. Esmaeili, Z. Mirzaie, M. Soleimani, R. Dinarvand and F. Atyabi, Int. J. Biol. Macromol., 2015, 78, 112–121.

---