Rufinamide (Rufi) was obtained from Glenmark Pharmaceuticals, India. Piribedil (Internal Standard, IS) was a gift sample from Reddy’s Laboratories, Hyderabad, India. Tamarind seed xyloglucan (TSX) was a gift sample from Encore Natural Polymers Pvt. Ltd., Ahmedabad, India. β-Galactosidase from Aspergillus oryzae was purchased from Sigma Aldrich, Mumbai, India. Medium molecular weight chitosan (190,000–310,000 Da; 75–85% deacetylated), sodium tri polyphosphate (STPP), mannitol, poloxamer 407, polyethylene glycol 400, and N-methyl-2–pyrrolidone were obtained from Sigma Aldrich, Mumbai, India. HPLC grade methanol, acetonitrile (ACN), glacial acetic acid (GAA), ammonium acetate, and thiomersal were procured from SRL Chemicals Pvt. Ltd., Mumbai, India. For all experimental processes and analysis, HPLC grade Milli Q water from our in-house water purification unit (Millipore^®, MA, United States ) was used. Male Wistar rats were procured from Veeba Biosciences Pvt. Ltd., Hyderabad, India.

Rufi-loaded chitosan nanoparticles (Rufi-Ch-NPs) were prepared using the ionic gelation method. First, the solutions of Rufi, STPP, and chitosan were prepared separately. Chitosan solution (1.0%w/v) was prepared by medium molecular weight chitosan in 1.5%v/v of GAA in water. For the preparation of Rufi solution, 5 mg of Rufi and poloxamer 407 were dissolved in 300 µL of NMP followed by the addition of PEG 400 (volume as per the design). STPP solution (30 mg/ml) was prepared by dissolving STPP in water. STPP solution (volume as per the design) was added to the Rufi solution to form a mixture which was then added to the chitosan solution under high-speed homogenization (Polytron PT 3100D, Kinemetica, Lucerne, Switzerland) at room temperature, using a syringe in a dropwise manner. The amount of chitosan (25 mg) taken was kept the same in all the experimental runs. The volume of STPP solution added in the preparation of nanoparticles varied based on the level of chitosan:STPP mass ratio required in the experimental run. Following the homogenization process, the dispersion was subjected to ultrasonication (Vibra cell, Sonics, Connecticut, United States). In the optimization runs, the amplitude (20%) and pulse (3 s on followed by 3 s off) of ultrasonication and homogenization time were kept at fixed values while the homogenization speed was varied according to the design. The dispersion containing Rufi-Ch-NPs obtained after processing was subsequently centrifuged and washed three times to remove the free drug. The nanoparticle pellet was re-dispersed in Milli Q water containing 3.0%w/v mannitol as a cryoprotectant (3.0%w/v of the total volume used to re-disperse the nanoparticles) and freeze-dried (Coolsafe 110–4, Scanvac, Lynge, Denmark). The freeze-dried formulation was stored in air-tight glass vials at 2–8°C until further use.

Rufi-Ch-NPs were designed based on the principles of DoE. Particle size (PS) (nm), zeta potential (mV), and entrapment efficiency (EE) (%) of the nanoparticles were identified as critical quality attributes or responses. Critical factors/variables affecting the PS and EE of nanoparticles were first identified using a minimum resolution (Minires) screening design. A total of six factors were screened at two levels [minimum (−1) and maximum (+1)]: A-chitosan: STPP mass ratio (1 and 10), B-volume of PEG 400 (200 and 1,200 µL), C-amount of poloxamer 407 (2 and 10 mg), D-homogenization speed (5,000 and 15,000 rpm), E-homogenization time (5 and 15 min), and F-ultrasonication time (2 and 8 min). A few preliminary trials combined with knowledge from previously reported literature helped in the selection of the polymers, solvents, processing conditions, and their respective high and low levels used in the screening design. The Minires design consisted of 17 runs including 3 center points. From the results of the Minires design, four factors viz. A-chitosan: STPP mass ratio, B-volume of PEG 400, C-amount of poloxamer 407, and D-homogenization speed, were found to be significantly affecting the responses.

Further, a high-resolution (Box Behnken Design) BBD was applied to understand how these critical factors and their interactions impacted the responses, and to optimize the preparation of Rufi-Ch-NPs. The BBD is a type of response surface design, used to obtain a second-order polynomial equation to optimize a formulation by performing very few experiments. The BBD consisted of 27 runs (inclusive of 5 center runs to check reproducibility) for four factors. The second-order polynomial equation generated from the BBD is of the following form: Y = β 0 + β 1 X 1 + β 2 X 2 + β 3 X 3 + β 12 X 1 X 2 + β 23 X 2 X 3 + β 13 X 1 X 3 + β 11 X 1 2 + β 22 X 2 2 + β 33 X 3 2 (1) Where, Y is the dependent variable, β 0 is the arithmetic mean response of the 27 runs, β ′ i s and β ′ i i s ( i   =   1 – 3 ) are coefficients of individual linear and quadratic effects of the factors, respectively, and β ′ i j s ( i , j   =   1 – 3 ;   i   <   j ) are coefficients of the effect of interaction between the i t h and j t h factors.

The optimum values for all factors were given by the Design-Expert version 11 software (Stat-Ease Inc., Minneapolis, MN) based on the desirability function criteria. The desirability function was calculated based on the constraints set for the dependent variables: Maximizing the EE % and zeta potential, and minimizing the PS. In order to validate the model, n = 6 repetitions were performed based on the experimental conditions predicted by the optimization model. PS, zeta potential, and EE % were determined for all the repetitions. Wilcoxon signed-rank test was performed to check if there was any significant difference between observed values and values predicted by the model.

Reacted xyloglucan gel (RXG) was prepared using a method published previously by our research group (Dalvi et al., 2021). Briefly, a 3.0% w/v solution of tamarind seed xyloglucan (TSX) prepared in 10 mM of sodium acetate buffer (pH 5.0) was treated with β galactosidase enzyme obtained from Aspergillus oryzae to partially remove galactoside residues. The reaction was kept at 35°C for 18 h. To terminate the reaction, the entire reaction mixture was heated to 90°C for 30 min. RXG was obtained by precipitation using 95 %v/v ethanol followed by drying in an oven at 50°C. The formulation development and optimization were based on several parameters which are discussed in detail in our previously published work (Dalvi et al., 2021). The optimized, freeze-dried Rufi-Ch-NPs were dispersed in a 2.0 %w/v solution of RXG using a magnetic stirrer to form a Rufi-Ch-NPs-loaded reacted xyloglucan-based in situ gelling formulation (Rufi-NP-RXG). To the Rufi-NP-RXG formulation, thiomersal was added at a level of 0.01% v/v as a preservative.

Rheological evaluation (Anton Paar MCR 302, Graz, Austria) was carried out to determine the T_sol→gel (solution to gel transition temperature) for Rufi-NP-RXG. The LVER (linear viscoelastic region) was identified using an amplitude sweep. Measurements were performed in the oscillatory mode using a temperature sweep within the LVER of the sample. The change in storage modulus (G′) with change in temperature was plotted for both Blank-RXG and Rufi-NP-RXG.

An in vitro drug release study was carried out by a membrane-less sample and a separate method for the aqueous suspension of Rufi-Ch-NPs (Rufi-NP-Susp) and Rufi-NP-RXG formulations. Dissolution was carried out in 250 ml beakers, all of the same dimensions. At the base of the beaker, an aluminum pan was glued at the center. The beakers were pre-equilibrated at 34 ± 1°C for around 30 min. Formulation quantity equivalent to 2.5 mg of Rufi was carefully dropped into the aluminum pan which was glued to the beaker. Within 2 min of addition of the formulations, 125 ml of dissolution medium, [simulated nasal electrolyte solution (SNES)] pre-equilibrated at 34 ± 1°C, was carefully poured into the beakers. The beakers were left in the incubator orbital shaker at 34°C, 100 rpm. Samples of 2 ml were withdrawn at predetermined time points (60, 120, 240, 360, 480, 720, and 1,440 min) from each beaker at different time points. The samples were centrifuged at 11,269 × g for 30 min at 10°C. Samples were analyzed using a validated RP-HPLC analytical method (as given in chapter 2). The results from the in vitro study were fitted into different mathematical models viz. zero order, first order, Higuchi, and Korsmeyer-Peppas models. The mechanism of drug release was determined based on the value of ‘n’ obtained from the Korsmeyer-Peppas model. Similarity factor (f2) was used for pair-wise comparison of the dissolution profiles of Rufi-NP-Susp and Rufi-NP-RXG formulations.

Stability was checked for both Rufi-Ch-NPs and Rufi-NP-RXG. Freeze-dried Rufi-Ch-NPs were stored separately in airtight vials (n = 3) at room temperature conditions (25 ± 2°C and relative humidity of 60 ± 5%), and Rufi-NP-RXG was stored separately in refrigerated conditions (2–8°C). Samples were collected after every 15 days and evaluated for their PS, EE %, zeta potential, and drug content.

In vivo studies included the assessment of nasal mucociliary transit time and pharmacokinetic (PK) evaluation of Rufi-NP-RXG and an aqueous suspension of Rufi-Ch-NPs (Rufi-NP-Susp). Adult male Wistar rats weighing 240–260 g were used for these studies. Animals were housed in our institute’s animal housing facility which is maintained at controlled temperature, humidity, and light-dark cycle (22 ± 1°C, 55 ± 10% relative humidity, and 12 h light-dark cycle). After procuring the animals, they were allowed to acclimatize to the new environment for at least 10 days before using them for experimentation. Food and water were provided ad libitum; except during experimentation, food was not provided until 4 h post-dosing; water was provided during experimentation. All animals were kept for overnight fasting before carrying out PK studies. Prior approval was obtained from the institute’s animal ethics committee (approval number: BITS: Hyd/IAEC/2017/19) for all the procedures carried out during animal experimentation.

To determine statistically significant/critical variables affecting the response variables, analysis of variance (ANOVA) was performed for the data obtained from the experimental runs conducted based on the Minires screening design. In the case of the optimization design using the BBD, the regression model between each of the response variables and their corresponding critical factors were tested and validated based on various diagnostic plots. In addition, the significance of the regression model for each of the response variables PS ( Y 1 ) , zeta potential ( Y 2 ) , and EE % ( Y 3 ) were evaluated based on the results obtained from ANOVA, adjusted R ², and predicted R ² values.

In the case of the in vivo PK studies and in vitro characterization, the data were expressed as mean ± SD. One-way ANOVA was used to compare PK data obtained from different experimental groups at a 5% level of significance. If the ANOVA results showed a statistically significant difference, a suitable post hoc test was applied to further compare the groups.

From the preliminary trials, a total of six independent factors/variables were identified with their upper and lower limits. A screening design was performed to select the statistically significant critical factors affecting the response variables (PS, EE %, and zeta potential).

Out of the six factors, the main effect of four factors viz. X₁-chitosan: STPP mass ratio, X₂-volume of PEG, X₃-amount of poloxamer 407, and X₄-homogenization speed was found to be statistically significant. Statistical analysis of the model revealed that the regression models obtained in the screening design were significant for all three responses. The ‘F_Cal’ values for regression models of the responses were 8.33 (p < 0.05) for PS, 312.2 (p < 0.001) for zeta potential, and 934.15 (p < 0.001) for EE %. In addition, the ANOVA results revealed that the curvature was significant in all three regression models.

The BBD, being one of the two popular quadratic response surface methods, was selected for the optimization of the response variable as a function of the four critical factors. The sum of squares, ‘F_Cal’ value, and ‘P’ value for the factorial term are given in Table 1. The factors with ‘P’ values less than 0.05 were considered to have a statistically insignificant effect on the response. Table 1 also shows the ‘F_Cal’ values for lack of fit, pure error, and model terms for each response. Factors affecting individual response variable and their regression equation are discussed in the subsequent sections.

The DSC thermograms are shown in Figure 2A. The thermogram for pure Rufi showed a sharp endothermic peak at 240°C. Pure chitosan showed a broad endothermic peak at 80°C which is attributed to the loss of water molecules from chitosan. There was no change in the characteristic peak of Rufi in the physical mixture, which indicates an absence of incompatibility between Rufi and other excipients used in the preparation of the Rufi-Ch-NPs. The thermogram of freeze-dried Rufi-NP-RXG showed a sharp endothermic peak at 160°C which corresponds to the melting point of mannitol, which was used as the cryoprotectant in the freeze-drying process of Rufi-Ch-NPs. Apparently, the absence of the characteristic peak for Rufi in Rufi-Ch-NPs might be due to its presence in amorphous form in the nanoparticle matrix. The SEM image of optimized Rufi-Ch-NPs is shown in Figure 2B. From the image, it can be seen that the nanoparticles are spherical in nature.

The T_sol→gel was determined in the LVER of the formulations. The T_sol→gel for Rufi-NP-RXG was compared with the T_sol→gel of Blank-RXG (Dalvi et al., 2021). The G′ values for both formulations at different temperatures are given in Figure 3. The plots for G′ values vs temperature for Rufi-NP-RXG and Blank-RXG are shown in Figure 3. The G′ values for Rufi-NP-RXG were consistently higher than Blank-RXG at all temperatures.

The in vitro release studies were performed to evaluate the kinetics and mechanism of drug release from Rufi-NP-Susp and Rufi-NP-RXG formulations. The release profile is shown in Figure 4. We have modeled the release of Rufi only from the Rufi-NP-Susp formulation, because in the case of the Rufi-NP-RXG formulation, the release of Rufi is affected at two different stages—release of Rufi from the Rufi-Ch-NPs and release of free Rufi from the RXG gel formulation in the dissolution medium. The release profile of Rufi from the Rufi-NP-Susp formulation was fit to different release kinetics models and it was observed that the Higuchi kinetics model (R ² = 0.9267) was the best fit model when compared to the zero order kinetics model (R ² = 0.7313) and first order kinetics model (R ² = 0.8691). The ‘n’ value for the Korsmeyer-Peppas model was 0.64 for Rufi-NP-Susp indicating that the mechanism of drug release followed non-Fickian diffusion.

Rufi-Ch-NPs and Rufi-NP-RXG formulations did not show a significant difference in their PS, PDI, zeta potential, and %EE for a period of 60 days. The % bias calculated for all the parameters at each interval for both the formulations was found to be not more than 5.0%. This is indicative of the physical and chemical stability of both the formulations. The stability data are given in Table 2.

Prior to the identification of critical factors using a screening design, a few preliminary trials were performed to choose the molecular weight of chitosan and the stabilizer, and also to set the limits for various factors used in the screening design. A few trials were performed with low, medium, and high molecular weight chitosan with the same degree of deacetylation (75–85%). It was observed that at same processing conditions and same concentrations, high molecular weight chitosan consistently yielded nanoparticles with greater PS than medium and low molecular weight chitosan. This may be because of higher viscosity of high molecular weight chitosan than the other two grades, which impeded the diminution of the particles beyond a certain size (12). It was observed that the EE % values with low molecular chitosan were consistently less than 50%. A similar observation was reported by Mohammadreza Abbaspour et al. (13). Out of several stabilizers tested, poloxamer 407 was chosen because it did not increase the solubility of Rufi in the dispersion medium (used in preparation of nanoparticles). PEG 400 was used as a co-solvent to maintain Rufi in the solubilized form during the addition of STPP solution into the Rufi solution during the preparation of nanoparticles.

For the screening of factors identified from the preliminary trials, a Minires screening design with resolution IV was selected. In a resolution IV design, the main effects are not confounded with other main effects or even two factor interactions. The screening design consisted of 17 runs (including 3 center point runs) with each factor at two different levels. Center point runs were performed to determine if the curvature (or the quadratic terms) was significant in the regression model. The information provided by the center point runs (significance of curvature) can help in selecting the appropriate optimization design.

The chitosan:STPP mass ratio greatly affected PS and zeta potential, while the %EE was only slightly affected by it. With increases in chitosan:STPP mass ratio from 1 to 7 units, PS of the nanoparticles decreased significantly. However, as the chitosan:STPP mass ratio is increased beyond 7–10 units, the PS of nanoparticles did not change significantly. This observation was found to be in line with the observations reported by F. Rázga et al. in their review article (Rázga et al., 2016). As the chitosan:STPP ratio decreases below 6, the balance between NH₃ ⁺ ions and O⁻ ions shifts toward STPP. For the same amount of STPP, a greater number of NH₃ ⁺ groups can bind to the O⁻ ions rapidly. Such a strong ionic interaction hinders further breakdown of particles at the same amount of energy input. No significant change was observed in PS with the change in volume of PEG 400 used in the formulation. Similarly, change in the amount of poloxamer 407 did not have a significant effect on the PS of the nanoparticles (Figure 1B). The effect of homogenization speed on the PS of nanoparticles was also insignificant.

In the case of zeta potential, as the chitosan:STPP mass ratio increased, the zeta potential also increased. This increase in positive zeta potential can be attributed to NH₃ ⁺ groups on chitosan which were left unbound from the O⁻ ions of STPP. The interaction term X₁X₂ showed a statistically significant effect on the zeta potential. Hence, to maintain hierarchy of the model, the factor X₂ (volume of PEG 400), although not significant, had to be included. Overall, the zeta potential was profoundly affected by chitosan:STPP mass ratio. The optimized formulations showed a high zeta potential which resulted in strong repulsion, no particle aggregation, and therefore good stability for the formulation.

Thermal analysis of freeze-dried Rufi-Ch-NPs showed that there were no interactions that could indicate incompatibility between the components of the formulations. Rheological studies showed that the G′ values for Rufi-NP-RXG were consistently higher than Blank-RXG at all temperatures. This can be attributed to hydrophobic interactions, and some ionic interactions between chitosan and xyloglucan molecules (Martínez-Ibarra et al., 2019). Although the strength of Rufi-NP-RXG (indicated by G′ values) was higher than Blank-RXG even at temperatures below the intranasal temperature, it did not affect the dosing precision of the formulation as indicated in section 3.6.1. Greater strength of Rufi-NP-RXG than Blank-RXG in the gelled state was found to be beneficial in order to retain the formulation in the nasal cavity for a longer time. From the in vitro study, the ‘n’ value for the Korsmeyer-Peppas model was 0.64 for Rufi-NP-Susp indicating that the mechanism of drug release followed non-Fickian diffusion, more specifically, a special case of non-Fickian transport–‘anomalous transport’. In anomalous transport, the velocity of solvent diffusion and the polymeric relaxation have similar magnitudes. Around 50% of the drug was released at 360 and 480 min in Rufi-NP-Susp and Rufi-NP-RXG, respectively. At the time point of 1,440 min, Rufi-NP-Susp and Rufi-NP-RXG had released 78 and 70% of Rufi, respectively. A slightly delayed release in the case of the Rufi-NP-RXG formulation may be attributed to a strong cohesive interaction between chitosan chains and RXG chains. The similarity factor (f2) value obtained for the comparison of the drug release profiles of Rufi-NP-RXG and Rufi-NP-Susp (f2 = 52) indicated that their drug release profiles were similar. Stability analysis of both formulations showed that they were stable for a period of 60 days.

I.n. dosing in rats requires optimization of two aspects viz. site of delivery of formulation in the nasal cavity and dose volume. Site specific deposition of formulation into the nose affects the pathway via which the drug gets taken up. One of the main strategies to enhance direct nose to brain drug delivery is to deposit the drug at the olfactory region of the nose. Hence a special delivery set-up with a soft and flexible cannula of 1.3 cm attached to a microtip was used to administer formulations. Further, once deposited into the nasal cavity, the formulation should withstand clearance from the nose by the mucociliary clearance (MCC) process. The MCC is a defense mechanism of the nose to push away foreign particles entering the nose. In order to have higher residence time in the nasal cavity, a formulation must resist the MCC process. Rufi-NP-RXG formulations showed a significantly higher MTT compared to other formulations. This can be attributed to the mucoadhesive properties of chitosan arising from an interaction between negatively charged sialic acid residues in mucin and NH₃ ⁺ groups of chitosan (Ways et al., 2018).

From the brain pharmacokinetic studies, it was evident that the concentrations of Rufi in the brain were significantly higher in the case of both nanoparticulate formulations compared to the brain concentrations achieved with the Rufi-Susp formulation. Also, the %DTE and % DTP values for both nanoparticulate formulations were significantly higher than Rufi-Susp. This could be attributed to longer residence time of the nanoparticulate formulations inside the nasal cavity. The brain concentrations of Rufi-NP-Susp and Rufi-NP-RXG were not significantly different at any time point, the % DTE and % DTP values were slightly greater for Rufi-NP-RXG formulation. This could be attributed to a slightly higher brain AUC_0→tlast value and lower plasma AUC_0→tlast value for Rufi-NP-RXG when compared to Rufi-NP-Susp. With the nanoparticles suspended in RXG gel, we expected to observe a significant increase in the brain uptake of nanoparticles. However, the observed PK data (%DTP values and brain concentrations at different time points) of Rufi-NP-Susp and Rufi-NP-RXG formulations did not show a significant advantage of the RXG in situ gel in enhancing the brain uptake of Rufinamide. Chitosan shows good mucoadhesive properties, and has been shown to enhance paracellular transport of small molecules, biomolecules, and nanoparticles by altering tight junction proteins. Consequently, a significant amount of Rufinamide brain uptake was observed only with Rufi-NP-Susp alone.