Synthesis of fish gelatin nanoparticles and their application for the drug delivery based on response surface methodology

Fish gelatin nanoparticle is produced using Tilapia fish skin for the first time by the two-step desolvation method. Fish gelatin is chosen for producing gelatin nanoparticles because no experiment have been done in using fish gelatin and to counter the problem associated with the use of mammalian gelatin, such as bovine spongiform encephalopathy disease. The effects of several factors on the particle size such as pH, acetone concentration, glutaraldehyde volume, stirring speed and stirring time are evaluated. Optimum conditions for the formation of gelatin nanoparticles are obtained using response surface method. Fish gelatin nanoparticles with optimum size of can be produced using pH of 2.45, acetone percentage of 16% (vol%), glutaraldehyde , stirring speed of 600 rpm, and stirring time for 6 h. The thermogram and molecular interaction of fish gelatin and fish gelatin nanoparticles are characterized using DSC and FTIR. In vitro drug release kinetic is examined using 5-fluorouracil as the model drug. The entrapment efficiency of 5-fluouracil as model drug is determined to be 40%. Fish gelatin could be used as a good alternative drug carrier for mammalian gelatin.

for gelatin application [14]. Under these circumstances, fish origin materials provided an alternative for producing gelatin nanoparticles. Furthermore, fish gelatins are being rarely used to produce gelatin nanoparticles. Until now, fish gelatin are used for producing gelatin film [15], and nanofibre [16]. Because of that, the experiment should be done to optimize the production of gelatin nanoparticles from fish gelatin.
Several methods have been used for producing gelatin nan oparticles such as the emulsion [17], coacervation [18], self assembly [19], and desolvation method [20]. Among those approaches the desolvation method was found to provide small particles and narrow size distribution [21]. The effects of several factors on gelatin nanoparticles size have been investigated, such as gelatin concentration, pH, and acetone volume [22][23][24][25]. However, those reports have just focused on single factor exploration, which precluded the interaction between various factors. Systematic investigations have not been carried out yet in optimizing fish gelatin nanoparticles (FGNPs) production incorporating the simultaneous effects of several significant factors. Thus, the optimization of produc tion fish gelatin nanoparticles using response surface method should be employed.
The chosen factors for producing gelatin nanoparticles depend on the gelatin characteristic, whereas the gelatin prop erties also depend on their resources [26]. Thus there are dif ferences between mammalian gelatin and fish gelatin. The specific characteristic of gelatin is the triple helical structure. The proline and hydroxyproline contents are approximately 30% for mammalian, and 2225% for warmwater fishes such as tilapia and nile perch [27]. These differences on fish gelatin lead to lower gel modulus and lower melting temperatures compared with mammalian gelatin. Hence the conditions required for the production of fish gelatin might vary with that for the production of mammalian gelatin.
In this study focus was to explore the potential of fish gel atin to produce FGNPs. Tilapia fish skin was used as a source of gelatin. Tilapia fish is commonly breaded in fish farms in Malaysia and hence tilapia fish skin is in abundance such as in filled production company. In this context, we aim to improve the desolvation method and to obtain optimum conditions for the production of FGNPs. The significant factors were chosen based on the onefactorattime method, and response surface method then was used to optimize these significant factors. The differences in thermal behavior and molecular interac tion between fish gelatin and fish gelatin nanoparticles were studied using differential scanning calorimetry (DSC) and Fouriertransform infrared (FTIR) spectroscopy. The nano particles shape was also quantified using scanning electron microscope (SEM). These experiments allowed us to deter mine a reproducible formulation of small sized FGNPs with narrow size distribution.

Materials
Tilapia fish gelatin (gelatin type A) with gel strength of 128.11 g bloom. Acetone, hydrochloric acid and sodium hydroxide were of analytical grade and were ordered from Sigma. Glutaraldehyde (25 vol%, grade I aqueous solu tion), and 5fluorouracil (5FU) were purchased from Sigma. Double distilled water was used for all the experiments. All chemicals were of analytical grade and used as received.

Production of gelatin nanoparticles
FGNPs were produced using the twostep desolvation method with slight modifications [28]. The first step is the fraction ation of low molecular weight (LMW) and high molecular weight (HMW) of gelatin. The second step is the precipita tion step to produce nanoparticles. As the second step or pre cipitation step is crucial for the formation of nanoparticles, the focus was on the optimization of this step.
The fractionation step started by dissolving 0.9 g fish gela tins in 10 ml distilled water under constant heating (45 • C) and stirring (600 rpm) until a clear solution was achieved. Acetone was used as cosolvent to precipitate the high molecular weight. The percentage composition of acetone was calculated with respect to the total volume of the mixture (100 ml). About 10 ml acetone was added to the gelatin solution and was cen trifuged at 12000 g for 5 min. The HMW fraction was obtained in the precipitate, and LMW fraction was in the supernatant solution. The supernatants containing LMW fractions were discarded. The HMW gelatin was dissolved again with 10 ml distilled water. The precipitation step begins with adjustment of pH of gelatin solution to the desired value by adding 0.1 M HCl and NaOH. FGNP were produced by adding 16 ml ace tone to HMW gelatin solution and 400 µ of glutaraldehyde solution (25 vol%). The nanoparticles were centrifuged and washed three times. The acetones were removed by evapora tion in a water bath at 45 • C temperature. The nanoparticles were stored in temperatures of 3 • C-5 • C for further research.

Design of experiment
The precipitation step in the production of FGNPs was optim ized sequentially. The screening for significant factors was done using the onefactoratatime (OFAT) method. On the basis of previous work, the effects of parameters like pH, ace tone concentration, glutaraldehyde volume, stirring speed, and stirring time on the size of FGNPs were studied. The levels of factors are summarized in table 1. The experiments were done in triplicates. The results were given as mean ± standard deviation (SD) of three independent experiments and statis tically analyzed by ANOVA, followed by the Tukey test to compare the different nanoparticle batches. The significant factors were optimized using response sur face methodology (RSM) based on three levels of face cen tered central composite design (FCCD) (table 2). The levels of parameters were selected based on the OFAT results. Mean particle sizes of FGNPs were used as the response variable. A total of 18 experimental runs including six center points were generated by designexpert version 7.0 software (State Ease Inc., Minneapolis, MN). The runs were carried out in triplicates.

Determination of isoelectric point of fish gelatin
Isoelectric point of fish gelatin was identified according to preferred method [29]. In brief, about 10 ml of 1% (w/v) fish gelatin was prepared in distilled water and pH was adjusted with 0.1 M HCl or 0.1 M NaOH. The isoelectric point of fish gelatin was determined by the measuring the turbidity as iden tified through the measured maximum intensity at 360 nm in UVvis spectrum [29,30].

Loading of fish gelatin nanoparticles with drug
5fluorouracil (5FU) was chosen as a model drug because it has been used as the major chemotherapeutic agent. 5FU loaded FGNPs were produced by adding 5FU directly to fish gelatin solution at the precipitation step.
2.6. Determination of particle size and zeta potential Mean particle size and zeta potential of FGNPs were deter mined using Zeta Sizer Nano Malvern (Zen 3600, UK). The mean diameters and polydispersity index (PI) values were spectrum FTIR over a diamond crystal. Small amount of sam ples (±1 mg) was placed in the diamond crystal and the FTIR spectra were recorded in the range of 4000-400 cm −1 . The results were plot between transmittance (%) and wavenumber (cm −1 ). The thermal behaviour of fish gelatin and fish gelatin nano particles was obtained using a differential scanning calorim eter (DSC60, Shimadzu, Japan). The samples were prepared using aluminium pans and empty aluminium pan was used as a reference. About 5 mg of samples were sealed and heated from 20 °C to 300 • C at a rate of heat flow of 10 • C min −1 .

Morphologies characterization by scanning electron microscopy
Field emission scanning electron microscopy (FESEM JEOL, JSM 6700F Model) was used to observe the size and shape of FGNPs. Briefly, a small amount of dry FGNPs was mounted on aluminum plates and, pasted with double sided copper tapes. Then the samples were sputtered with a thin layer of gold and placed on the packet chamber at an accelerating voltage of 10 kV.

Transmission electron microscopy
Transmission electron microscopy (TEM) of fish gelatin nanoparticles was performed using a Philips Tecnai F 20.2 µ of FGNP sample was dropped on the copper TEM grid and air dried for 2 h. The morphology data was carried at an acceler ating voltage of 200 kV.

2.10.Drug content and in vitro drug release
UVvis spectrophotometer was used to determine the concentra tion of 5FU in the FGNPs. About 5 mg of dried FGNP loaded with 5FU were dispersed in 5 ml of phosphatebuffered saline (PBS) (pH 7.4) at room temperature (23 • C ± 2 • C), containing 2.5 mg trypsin. After 6 h of digestion, the samples were diluted to 25 ml and filtered using 0.22 µm filters. The absorbance was measured at λ max = 265 nm using SartoriusStedim VivaSpec UVvis spectrophotometer. A calibration curve was prepared with different 5FU concentrations in PBS (the presence of gelatin and trypsin had no effect on the absorbance intensity). Unloaded FGNP were used as a blank. The entrapment effi ciency (EE) was calculated using following equation obtained at 90° angle in 10 mm diameter cells. Each measure ment was conducted in triplicates.

Physiochemical characterization
The experiment of FTIR was made on lyophilized samples of fish gelatin and fish gelatin nanoparticles using Perkin Elmer EE(%) = total amount of drug added-amount of free drug present in supernatant total amount of drug added in formulation × 100. (1) In vitro release of 5FU was determined using the method previously described [25]. A weight of 5 mg of FGNPs was dispersed in 25 ml of PBS (pH 7.4). About 2 ml sample was withdrawn at defined time intervals and was centrifuged for 20 min at 10000 g. Then 1 ml aliquots were withdrawn from the supernatant and added back to the original solution. The pellets were redispersed in 1 ml of PBS and added to the original solution to keep the particle concentration constant. The amount of 5FU in the supernatant was quantified using a UVvis spectrophotometer. All the experiments were repeated three times and the average values along with the errors were calculated.

Calculation of release kinetics
In vitro 5FU release behavior from FGNPs was studied using DDSolver utilizing data up to 60% cumulative release [31]. DDsolver is the supplement program in microsoft excel which used nonlinear leastsquares curve fitting by minimizing the sum of square differences between the observed and the pre dicted drug dissolution values at time intervals t, with the best model parameters [32]. Five models were used in this experiment namely zero order, first order, Higuchi, Weibull, and KorsmeyerPeppas. The DDsolver also calculates several parameters allowing the statistical fitness evaluation of the model. The most appropriate model to fit the dissolution data should give the highest value of adjusted coefficient of deter mination ( R 2 adjusted), smallest value of akaike information criterion (AIC), and the largest value of model selection cri terion (MSC). Exponent n of the KorsmeyerPeppas model gives infor mation about the release mechanism of the drug according to following equation where Q t is a fraction of drug released at the time t, k is the release rate constant and n is the release exponent. The n value is used to characterize different release for cylindrical shaped matrices. If n 0.45 it is a Fickian diffusion, if n = 0.85 it is a case II transport, which is related to polymer matrix relax ation and swelling. If 0.45 < n < 0.85 it corresponds to an anomalous transport, resultant from the combination of both mechanisms and if n > 0.85 it is a super case II transport [33].

Selection of significant variables
3.1.1. Effects of pH. Figure 1 shows the effects of pH on the size of FGNP. The smallest nanoparticles were found to form at pH 2.5. Decreasing the pH produced small nanoparticles until a pH of 2.5 ( p 0.05). No significant increase in the particle size was observed if the pH was below 2.5. The pH value of gelatin solution before the precipitation step is shown at pH of 4.6. In the precipitation step, pH was adjusted to be below the isoelectric point because the isoelec tric point of fish gelatin was determined at pH 6 (figure 2). As shown in figure 1, the nanoparticle size increased as the pH is near to the isoelectric point. The lowering of pH leads to protonation of the fish gelatin molecules.
The lowest pH value to produce the smallest FGNPs was at 2.5. This result is slightly different compared to most mamma lian gelatin. Previous studies on porcine gelatin have indicated that the pH value to produce nanoparticles in small sizes was 3.25 [34]. This might be due to the difference on negatively charged amino acid contents in the gelatin molecule. The fish gelatins have more negatively charged amino acids than mammalian gelatin; hence fish gelatin needs a more acidic pH for smaller nanoparticle formation compared to mammalian gelatin. Gelatin is sequences of amino acids. Fish gelatin mol ecule contains ~14% of negatively charged glutamic acid and aspartic acid, ~7% of positively charged lysine and arginine, ~8% of the chain hydrophobic amino acids (leucine, isoleu cine, methionine, and valine) and ~40% of glycine [21] while mammalian gelatin contains ~13% of positively charged, ~12% of negatively charged and ~11% of hydrophobic chain. This difference in the amino acid composition affects the solu bility characteristic of fish gelatin in water molecules and the processing parameters in the production of FGNPs [27].

Effects of acetone concentration.
Acetone is used as a cosolvent in the production of FGNPs. Figure 3 shows the  effects of acetone concentration on the FGNPs size. It can be seen that increasing acetone concentration lead to increase in nanoparticle size whereas 15% (vol%) acetone produced smallest nanoparticle fish gelatin ( p 0.05). Similar trends were observed in the previous studies [35,36].This has been attributed to the role of acetone molecules in disturbing the interaction between water and gelatin molecules resulting in gelatin molecule aggregation.
Interestingly, tilapia fish gelatin requires lesser acetone to produce nanoparticles compared to mammalian gelatin. Tilapia fish gelatin used less than 15% acetone whereas mam malian gelatin needs around 70% of acetone [37,38]. This is also due to the difference on amino acid content.
Acetone have been chosen as a desolvating agent because acetone is highly miscible in water and prevents denaturation of gelatin compared to ethanol [39]. In addition, acetone pro duced small sized gelatin nanoparticles with lower polydis persity index compared to ethanol [40].
The addition of acetone to the gelatin solution is to change the solubility character of the gelatin molecule. In the water environment, the hydrophilic amino acids existing on the sur face interact with the polar water molecules forming hydrogen bonds, while the hydrophobic amino acids remain inside that gelatin in the core. The water molecules surrounded the hydrophilic section of gelatin, and stabilized the gelatin mol ecule structures by hydrogen bonds.
When acetones are added to the gelatin solution, the solu bility of gelatin will begin to decrease and become insoluble until a certain acetone concentration is reached. This is because of the decrease in the amount of hydrogen bond that available to interact with the hydrophilic amino acids of gelatin [40]. Decreasing the hydrogen bonds destabilized the individual gelatin molecules leading to their aggregation. Hence, we can say that the aggregation is the rearrangement of the interaction between nonpolar sections of gelatin to stabilize the gelatin molecule.

Effects of glutaraldehyde as crosslinking agent.
The effect of glutaraldehyde volume on the size of the FGNPs was shown in figure 4. It can be seen that increasing the volume of glutaraldehyde during production of FGNPs decreased the size of the FGNPs initially ( p 0.05). However, further addition of glutaraldehyde volume increased the FGNPs size.
Glutaraldehyde is soluble in water and categorized as a hydrophilic crosslinking agent [21]. Glutaraldehyde works by hardening the particle through the crosslinking of amino    acid chains [41]. Initial addition of glutaraldehyde volume enhances the formation of smaller size nanoparticles by crosslinking the intra amino acid bonds. However, fur ther increase in glutaraldehyde facilitates the formation of intermolecular linkages among the nanoparticles, thereby increasing the nanoparticle size at high concentrations. Thus, the nanoparticle sizes are controlled by the delicate balance between inter and intramolecular bonds among the gelatin molecules [40]. figure 5, the nanoparticle size is found to decrease as the stirring speed increased from 150 rpm to 600 rpm which is the maximum allowable stirring speed in our experimental setting. However, no significant effect ( p 0.05) of the stirring speed on the nanoparticle size was observed. The increase in stirring speed lead to increase in shearing energy to break the large particles into smaller particles, which could prevent huge agglomera tion, but further increase of the stirring rate, does not have a significant impact in decreasing the particle size [42,43]. Thus, the stirring speed should be maintained at 600 to pro duce smaller nanoparticle sizes. Figure 6 depicts the effects of stirring time on nanoparticles size. It shows that 6 h of stirring time produced smaller nanoparticles compared to 12 h of stir ring time. This demonstrates that further increase in stirring time led to increase in particle size, which might be attributed to agglomeration due to enhancement in kinetic energy with an additional input of energy [44]. However, the variation in stirring time shows insignificant differences to the response ( p 0.05). Hence, 6 h was chosen and kept constant in the optimization design.

Effects of stirring time.
From the screening results, factors with pvalue of less than 0.05 were significant on the response, and were selected for further optimization. Results showed that pH, acetone con centration and glutaraldehyde volume have significant effects on producing smaller nanoparticle sizes, while stirring speed and stirring time gave insignificant effects. The optimum levels of these significant factors (pH, acetone concentration, The result are mean ±SD(n = 3). a The treatment were run as a random order.  and glutaraldehyde volume) were further determined by a response surface method.

Optimization by response surface methodology
The optimization process was conducted using response sur face method (RSM). Three factors namely pH solution (X 1 ), acetone concentration (X 2 ) and glutaraldehyde volume (X 3 ) were chosen as independent factors, whereas mean particles size was a dependent factor. The other two factors namely the stirring speed and stirring time were kept constant at 600 rpm and 6 h. All experimental designs were conducted in trip licates. The result of the response surface method is shown in table 3. It can be seen that the mean particles size varied from 190 nm to 276 nm. The small particle was produced at the center point of the design. A suitable polynomial equation involving the main indi vidual effects and interaction factors were selected based on estimation of several statistical parameters provided by the Design of Expert ® software. Equation (3) depicts a multiple regression analysis of the experimental data for mean parti cles size, where X 1 is pH, X 2 is acetone concentration, and X 3 is volume of glutaraldehyde. All parameters used in the polynomial equations were found to be statistically significant ( p < 0.05).

Mean particle size
The statistical significance of the model equation was eval uated by the Ftest for analysis of variance (ANOVA), which showed that regression is statistically significant at 95% ( p < 0.05) confidence level. The value of p less than 0.05 indicates that the model terms are also significant. The lack of fit pvalue larger than 0.05 implies that the lack of fit is non significant, which means the model is strong enough with less noise. In the response surface design, ANOVA was also used to determine the significant contribution of main variables and their interactions on the response variables.
It can be seen from table 4, pH (X 1 ), concentration of ace tone (X 2 ), glutaraldehyde volume (X 3 ), cross product contrib ution (X 1 X 3 , X 2 X 3 ), quadratic contribution X 1 X 2 X 3 , and X 2 1 X 3 were significant. The regression equation obtained from ANOVA depicted that the R 2 (multiple correlation coefficient) was at 0.9195 (a value > 0.75 indicates fitness of the model). This results estimates that if the fraction of overall variation in the data is accounted by the model, thus the model is capable of explaining 91% of the variation in response. The adjusted R 2 is 0.8949 and the predicted R 2 is 0.8959; these values near to 1.0 indicate that the model is good. The 'adequate precision value' at 23.266 also explains that the model is good.
The response surface plots were constructed by plotting the mean particle size as a response on the zaxis against any two independent variables, while other variables were kept at their optimal levels. This plot is to determine the optimal levels of each variable for production FGNPs ( figure 7).
The formulations of variables were then optimized for response of mean particle size. The optimum values of the variables were obtained by numerical analysis using design expert software and based on the criterion of desirability. Afterwards, a new run of production FGNPs with the pre dicted levels were conducted to confirm the validity of the   Adv. Nat. Sci.: Nanosci. Nanotechnol. 9 (2018) 045014 optimization procedure. The optimized variables were found at pH 2.45, acetone percentage at 16 %(v/v) and glutaralde hyde volume at 400 µ . Figure 7(a) represents multiple interactions between pH and acetone to FGNP size. It can be seen that around 15% concentration of acetone was required to produce small nano particles at pH 2.5, while a slightly higher concentration of 20% acetone was required to produce small nanoparticles at pH 1.5. A decrease in pH led to the increase of interactions between the positive charges in the gelatin molecule with water ions [34]. This condition requires high acetone concen tration to interrupt the water ion interaction. Thus increasing acetone concentration means increasing the effects of inter ruption between the hydrogen bond in the gelatin molecule. Because in low pH the hydration of network in the gelatin molecules were strong and more intensive [45]. Figure 7(b) shows multiple interactions between the pH and the volume of glutaraldehyde. Around 400 µ of volume of gluta raldehyde was needed to produce small sizes of FGNP at pH 2.5. As we can see at pH 2.5, the particle size decreased when glutar aldehydes were added from 300 µ to 400 µ . Then, the particle size becomes huge by increasing glutaraldehyde volume. It can be concluded that low glutaraldehyde volume was required to produce small nanoparticle sizes at low pH, moreover slightly higher volumes of glutaraldehyde were used to produce small particles at high pH. This is because at high pH, a huge volume of glutaraldehyde would harden the gelatin nanoparticles from inside the gelatin molecule instead of making new interactions outside the molecule. However, in low pH, just a little volume of glutaraldehyde was interacted to produce small size nanoparti cles, and further increments of glutaraldehyde will make a new interaction to intramolecular produce large nanoparticle.
Multiple interactions between glutaraldehyde volume and acetone concentration are shown in figure 7(c). This picture depicted that around 440 µ of glutaraldehyde was required to produce small particles when 20% of acetone concentration is used as the cosolvent, and its volume becomes less when the acetone concentrations are decreased. This phenomenon was created because as more acetone is added to gelatin solu tion, more gelatin molecules are destabilized and aggregated by intermolecular interaction.

Validation of models
The results of the statistical model and regression equa tion were validated by running five experiments under optim ized variables. The variables were pH at 2.45, acetone percentage at 16% (vol%) and glutaraldehyde volume at 400 µ . Under these optimal variables, the predicted mean particles size calculated by software was 201.8 nm, and the observed experimental value after average was 198.46 ± 6.1 nm. The results confirmed that the model is valid by the error of experiment which is quite close to about 1% error and indicated that the results are in good agreement with the predictive value.
The particle sizes of fish gelatin nanoparticles prepared in the present work are different from those of mammalian origin gelatin [46,47]. The difference in particle size can be explained by the bloom number of its source. Previous studies using mammalian gelatin showed that gelatin nanoparticles have been produced with size at 160 nm using type A gelatin bloom number 300 [46]. Different bloom numbers give dif ferent particle sizes, with larger bloom number yielding smaller particle size [47].
From the results, the small nanoparticles were produced in the interactions between pH, acetone concentration and gluta raldehyde volume. At a high pH, the intermolecular bonding is more extensive than intramolecular, while in low pH intra molecular bonding has the highest effect. The col lision of molecules increases the effects of molecular bonding between each other. In addition, in high pH conditions, high volumes of glutaraldehyde were used to produce small particles, because the high volume glutaraldehyde will enforce the bonding in the molecule (intramolecule).
The DSC thermographs of fish gelatin and fish gelatin nanoparticles are presented in figure 8. The thermograph of fish gelatin (curve a in figure 8) shows the presence of two  Adv. Nat. Sci.: Nanosci. Nanotechnol. 9 (2018) 045014 endothermic events which are glass transition temperature (T g ) and melting temperature (T m ) at 98 • C and 220 • C, consistent with the literature value for gelatin [48]. For fish gelatin nanoparticles T g and T m are observed as endothermic peak around 81 • C and 220 • C, respectively. The first of endo thermic peak represents the energy consumed during vapor izing the water present in the matric and association of the glass transition of α-amino acid blocks in the polypeptide chain [49,50]. It was found that the T g of fish gelatin is higher than its fish gelatin nanoparticles, these results suggested that thermal stability of fish gelatin is stronger than fish gelatin nanoparticles. The lower temperature of glass transition in the fish gelatin nanoparticles indicated that higher water bound in the structure because the increasing of polymer free due to the transformation of nanoparticles [51]. Figure 9 shows the FTIR spectra of fish gelatin and fish gel atin nanoparticles. The FTIR analysis was used to evaluate and to compare the chemical structural between fish gelatin and fish gelatin nanoparticles. The spectrum showed characteristic bands at approximately 3260 cm −1 (amide A, N-H stretching vibrations of NH 2 ), 2920 cm −1 (amide B, C-H stretching), 1640 cm −1 (amide I, C=O stretching), 1540 cm −1 (amide II, N-H bending), 1440 cm −1 (CH 2 bending), and 1180 cm −1 (amide III, C-N and N-H Stretching) [52,53]. The similar results on mammalian gelatins were also found by Hoseini et al [15], Dixit et al [6], and Sarmah et al [54]. In general, the result shows similar FTIR spectra for fish gelatin and fish gelatin nanoparticles but different intensity. Furthermore, the ratio of intensity of Amide I could be used to observe the loss of secondary structure and formation of random coil [55]. The expense of secondary structure and formation of random coil was related to the increasing of amide I intensity [55]. It is because of the preparation step for producing fish gelatin nanoparticles which used temperature of 45 • C to dissolve the gelatin. The protein start to lost its triple structure as the temper ature rich to 30 • C [55]. TEM image (figure 10) shows that the fish gelatin nanoparticles are spherical and have a homogeneous size distribution in the range 24 to 80 nm.
The size between FGNPs and 5FU loaded FGNPs also varied. The size of FGNP after validation showed to be around 198.46 ± 6.1 nm, while 5FU loaded FGNP was at 238.02 ± 7.4 nm. It can be concluded that the loading increases nanoparticle size. The size of drugloaded FGNPs was 20% larger than unloaded FGNPs. Figure 12 shows the release behavior of 5FU from FGNPs in PBS at pH 7.4. It can be seen at the 5 h incubation, almost 50% of 5FU was released to medium and after 12 h of incubation, 80% release was found. From the results, two stages of 5FU release from FGNPs were observed. The first stage is rapid release where around 50% of the drug was released in the ini tial 5 h. The second stage is followed by the sustained release.

In vitro drug release kinetics
The rapid release is due to the nature of the fish gelatin. As a carrier, gelatin is highly hydrophilic thus it facilitates water uptake from release medium to matrix and results in a higher initial burst release. This condition have been described by Nahar et al [47] that the release behavior of drugs from par ticles depend on several factors, such as the size of particles, type of the polymer carrier, swelling characteristic of the par ticles, nature of crosslinking agent, and nature of the drug. Another cause of initial burst release is also because of the presence of the drug attached on the surface of particles [56,57]. In vitro drug release from FGNP is expected to be sim ilar to type A mammalian gelatin that has similar isoelectric points. According to previous papers, this finding is similar to mammalian gelatin nanoparticles releasing amphotericin and resveratrol through initial burst in PBS at pH 7.4 [57,58].
This experiment also revealed that FGNP has similar release kinetics to mammalian gelatin. Mammalian gelatin nanopar ticles follow the Fickian mechanism in releasing ibuprofen and 5FU to medium release [8,12]. As shown in table 5, the goodness of fit for the various models ranked in the order: Korsmeyer-Peppas ≈ Higuchi > First-order > Zero-order . The value of exponent n from the KorsmeyerPeppas model is around 0.5 indicating that the 5FU release from FGNP fol lows the Fickian diffusion.

Conclusions
FGNPs are produced with particle sizes of 191-245 nm by the twostep desolvation method using acetone as a non solvent and glutaraldehyde as a cross linker. Three factors on production of FGNPs such as concentration of acetone, volume of glutaraldehyde and pH are found to have significant effects on the size of the gelatin nanoparticles. The effects of the significant factors have been evaluated. Increasing pH and acetone concentration leads to increase in the particles size, while increasing volume of glutaraldehyde decreases the size of FGNP. The optimum conditions of those factors are pH 2.45, acetone percentage 16% (vol%) and glutaraldehyde at 400 µ . FGNP then was produced using those optimum con ditions and was loaded with the model drug. 5FU was used as a model hydrophilic drug loaded into FGNP. The in vitro release kinetics of 5FU was investigated. The release of 5FU from FGNP followed the KorsmeyerPeppas model kinetics with Fickian mechanism. Thus it can be concluded that FGNP presented a good alternative for the delivery of hydrophilic drugs such as 5FU.