Optimization of PLGA-CURC using central composite design
Response Surface Methodology (RSM) using the Central Composite Design (CCD) model is a well-suited experimental design strategy that offers the possibility of investigating a high number of variables at different levels with only a limited number of experiments [29
]. The methodology was originally developed by Box and Wilson and improved by Box and Hunter. This is an ideal tool for process optimization [23
], and its rotatable characteristic enables identification of optimum responses around its center point without changing the predicting variance. RSM is a collection of mathematical and statistical techniques based on the fit of a polynomial equation to the experimental data, which must describe the behavior of a data set with the objective of making statistical provisions. CCD has been successfully used to optimize the technology or production condition for drug delivery systems such as sustained release tablets, liposomes, microspheres, nanoparticles in recent years [29
The ranges for each of the variables in Table were chosen taking into account our preliminary experiments. Table shows the experimental results concerning the tested variables on mean diameter of particle size, polydispersity drug loading percentage and encapsulation efficiency. These four responses were individually fitted to a second order polynomial model. For each response, the model which generated a higher F value was identified as the best fitted model. Each obtained model was validated by ANOVA. Three dimensional response surface plots were drawn for the optimization of PLGA-CURC formulation. These types of plots are useful in studying the effects of two factors on the response at one time, when the third factor is kept constant.
Influence of formulation variables on particle size
Particle size is a critical factor for nanoparticle-based drug delivery system. It is one of the factors that control the kinetics of drug release. Generally, smaller particle size permits a faster release rate. The following second order reduced quadratic model equation was derived by the best fit method to describe the relationship between the particle size (Y1), the amount of PLGA, concentration of PVA and volume of ethyl acetate.
A positive value in regression equation for a response represents an effect that favors the optimization (synergestic effect), while a negative value indicates an inverse relationship (antagonistic effect) between the factors and the response [30
The reduced quadratic model was found to be significant with an F
value of 30.87 (p
0.0001), which indicates that response variable Y1
and the set of formulation variables were significantly related. The high R2
value indicated that 96.53% of variation in particle size was explained by the regression on formulation factors (Additional file 1
: Table S1).
The particle size values for the 20 batches show a wide variation in response i.e., the response ranges from a minimum of 77.8 nm to a maximum of 198.8 nm. The data clearly indicate that the particle size value is strongly dependent on the selected variables. The response surface plots for particle size as a function of formulation factors were constructed by holding one of the factors at a constant level. Figure A shows the response surface plot obtained for the interaction between PLGA concentration and PVA at constant value of ethyl acetate. An increase of the mean particle size was observed (Figure A) when increasing concentration of PLGA for all the amount of PVA used in the formulation (1–3%). It was reported that an increase in the amount of PVA in the formulation may lead to the smaller particle size due to tight surface that was formed from PVA macromolecular chains of high concentration [36
]. However, too much PVA is not suggested as it will hinder in vivo
In addition, PVA has been found to have a carcinogentic potential and removal of excess PVA from the particle surface is difficult [37
]. Our data support that a lower concentration of PVA (1% w/v) was suitable to obtain well controlled particle size formulations. Analyzing the response surface of interactions of PLGA and ethyl acetate at constant PVA, we found that initially, with an increase of solvent volume, the particle size does not change much and then decreases with further increase in the volume of ethyl acetate (Figure B). Formation of nanoparticles depends on the rate of diffusion of the organic solvent into the aqueous phase, which in turn influences the precipitation of polymer thereby influencing the particle size. The minimum particle size and its corresponding experimental conditions were derived from the regression model.
Figure 2 Three dimensional response surface plots showing the effect of variables on response:- particle size; (A) effect of PLGA and PVA concentration on particle size (Actual constant ethyl acetate (ml)=4.25); (B) effect of PLGA and ethyl acetate (more ...)
Influence of preparation factors on polydispersity index
After nanoparticle formation, the size population frequently follows a multimodal distribution. The polydispersity index is a very important parameter which is used to describe variation of particle size in a sample of particles. When this index is close to 1, the size range becomes wide. A desired optimal value is closer to 0. The response surfaces of polydispersity index keeping ethyl acetate and PVA constant are shown in Figure A and B respectively. The polydispersity variations are found to be in the same direction as the particle size in all the cases studied. Increasing the amount of polymer and decreasing the volume of organic phase leads to an increase of the polydispersity index. The coefficient of correlation, R2 is 0.8927 and the model gives a p-value <0.001 (ANOVA).
Figure 3 Three dimensional response surface plots showing the effect of variable on response:- polydispersity; (A) effect of PLGA and PVA concentration on polydispersity (Actual constant ethyl acetate (ml)=4.25); (B) effect of PLGA and ethyl (more ...)
The minimum polydispersity index and its corresponding experimental conditions were derived from this regression model. The values of polydispersity index predicted from this regression model are shown in Table .
Comparison of the experimental and predicted values of PLGA-CURC prepared under the predicted optimum conditions
Influence of preparation factors on encapsulation efficiency
In our study, encapsulation efficiency of PLGA-CURC reached up to 93.5% (Table ). High encapsulation efficiency is advantageous since it transports enough drug at the target site and increase the residence time of the drug. The high encapsulation efficiency in PLGA can be attributed to several factors. First the hydrophobic nature of PLGA molecules makes it relatively easy to entrap hydrophobic curcumin into PLGA-CURC. Second, the hydrophobic nature of curcumin results in a minimum loss of the drug to the external aqueous phase during the formulation process. The response surface diagrams reveal that the encapsulation efficiency first increases with increasing PLGA concentration and then decreases (Figure A and B) at constant PVA and ethyl acetate concentration. Furthermore, there is no significant change observed with variation of PVA concentration (1–3% w/v) (Figure C). The optimized variables show a good fit to the quadratic model (Eq. 9) with an F
value of 14.15 (p =0.0001), which indicates that response variable Y3
and the set of formulation variables were significantly related. The high R2
value indicated that 92.73% of variation in encapsulation efficiency was explained by the regression on formulation factors (Additional file 1
: Table S1).
Figure 4 Three dimensional response surface plots showing the effect of variable on the response:- encapsulation efficiency; (A) effect of PLGA and PVA concentration on encapsulation efficiency (Actual constant ethyl acetate (ml)=4.25); (B) effect (more ...)
The statistical analysis of the results generated a quadratic response for encapsulation efficiency is as follows
Influence of preparation factors on percentage drug loading
The response surface graphs for the most statistically significant variables on percentage drug loading are shown in Figure (A-C). The response surface diagram depicting interactions of PLGA concentration and PVA showed that increase in polymer concentration first increases the percentage drug loading and then decreases implying an optimum polymer concentration for maximum drug loading. At higher PLGA concentration, initially there was no significant change observed for drug load with respect to PVA concentration, but drug loading increased with increase in PVA concentration. The reverse was observed at low PLGA concentration.
Figure 5 Three dimensional response surface plots showing the effect of variable on response:- drug loading; (A) effect of PLGA and PVA concentration on the drug loading (Actual constant ethyl acetate (ml)=4.25); (B) effect of PLGA and ethyl (more ...)
Optimization by desirability function
Optimization process was undertaken with desirability function to optimize the four responses simultaneously. A high value of desirability coefficient δ
1) indicates that the operating point can produce acceptable formulation results. The responses: particle size (Y1
), polydispersity index (Y2
), encapsulation efficiency (Y3
) and drug loading (Y4
) were transformed into the desirability scale d1
, respectively. Among them, Y1
had to be minimized, while Y3
had to be maximized. The overall objective function (δ
) was calculated by Eq. (2). The model was fitted with a second-order quadratic expression. The higher coefficient of determination and F value in terms of the quadratic model indicated the goodness of fit. Figure shows the response surface plot for increasing desirability coefficient δ with respect to changes in variables: PLGA (X1
) and PVA (X2
) keeping the volume of ethyl acetate constant. The maximum value of desirability coefficient δ
0.716 was obtained at the conditions, PLGA amount of 85 mg, PVA concentration of 1%(w/v) and 4.25 ml of ethyl acetate.
Response surface for overall desirability (δ as a function of PLGA (mg) and PVA (%) at constant ethyl acetate at 4.25 ml.
In order to evaluate the predictive power of this model and desirability coefficient, PLGA-CURC was prepared under the optimal conditions. The results comparing the experimentally obtained and model predicted values of all four responses are presented in Table . The experimental values of the multiple batches prepared under the optimal conditions were very close to the predicted values, with low percentage bias, suggesting that the optimized formulation was reliable and reasonable. It has been shown that the highest encapsulation efficiency and drug loading with commensurate minimum mean particle size and size distribution was achieved by using the optimal conditions of 85 mg PLGA, 1% (w/w) of PVA and 4.25 ml volume of ethyl acetate.
Scale up for large batch production of PLGA-CURC
Scaling up the nanoformulation process to produce large batches of nanoparticles is the key to effective clinical use of nanoparticle based drugs. To translate this formulation into large scale production, we investigated seven critical parameters and their correlation in four sequential stages. Each stage was optimized to get the best parameter combination in terms to target response of particle size, polydispersity, encapsulation efficiency and drug loading. The parameters chosen include polymer solvent ratio, aqueous phase volume, organic and aqueous phase ratio, sonication tip diameter, sonication time, stirrer speed and stirring time. Our goal was to produce PLGA-CURC with similar physicochemical characteristics in a scaled up batch production. Results from all the different stages of scale up production are compiled in Tables and which shows the parameter variations and optimized scale up outcomes of PLGA-CURC formulation.
In the first scale-up optimization stage, polymer amount was increased from 85 mg to 500 mg but the volume of ethyl acetate was increased to only 5 mL from 4.25 mL. This was a very critical step as we needed to minimize the amount of organic solvent needed to prevent problems of solvent evaporation. This resulted in an increase in viscosity of the organic phase leading to larger particle size. To overcome this, the sonication time was increased from 60 sec to 120 sec keeping the sonication tip diameter and stirring speed and stirring time same. This resulted in average particle size of 135.4 nm, an increase of about 6 nm from the primary optimized batch. The drug loading in the first stage scale-up dropped by 0.7% while encapsulation efficiency remained almost the same. Once we achieved the first stage, next we scaled up ~10 times for producing 1 g of PLGA-CURC in the second stage. Doubling the amount of polymer and volume of solvents required increasing the sonication power to get nanoparticles in the same particle size range. For that, the sonication tip diameter was increased from 2 mm to 14 mm. Further, stirring speed was increased from 2,000 rpm to 3,000 rpm and stirring time was increased by 2 h. The resulting optimized batch had an average particle size of 142.3 nm and 11.12% of drug loading. In the third stage (~20X) of scale-up, the aqueous phase was optimized at 60 ml. To keep the nanoparticle size comparable, the sonication time was increased to 180 sec and the stirring speed was doubled to 6,000 rpm. Accounting for more than double of total volume from stage one to three, the exposed surface area for evaporation of solvents was increased by using an open mouthed vessel during stirring. With all these parameter combinations, the final optimized batch for this stage showed a 0.2% decrease in drug loading only and similar ~6 nm increase in particle size from previous stage. In the final stage, we scaled up to ~50 times to produce 5 g of PLGA-CURC. Here, the aqueous phase was increased to 150 mL with the excess water being increased to 200 mL to account for 5 g of polymer being used. Such a large volume of liquid phase needed high sonication power which was brought about by increasing the sonication time from 180 sec to 300 sec, increasing the stirring speed to 8,000 rpm and increasing the stirring time to 8 h. This resulted in nanoparticles having an average particle size of 158.5 nm, a total increase of only 29.4 nm from the primary optimized batch. Also, the drug loading decreased by 2.36% which is minimal considering ~50X scale-up. The encapsulation efficiency and polydipersity was found to be similar to the primary optimized batch. We have successfully produced PLGA-CURC in 5 g quantities through this route and identified critical parameters for scaling up the formulation process.
Characterization and evaluation of the optimized scaled-up formulation
The external morphology of lyophilized PLGA-CURC prepared at optimal conditions are shown in Figure A and B. PLGA-CURC were spherical, discrete without aggregation, and smooth in surface morphology. The size of the PLGA-CURC was found to be approximately 140 nm. The particle size determined by Differential Light Scattering (DLS) for the same batch was found to be 158.5 nm. This may be explained by the fact that particle size analyzer, based on DLS, measures the hydrodynamic diameter of the particle while the electron microscope measures exact diameter of particles in solid state. But at the same time, the amount of nanoparticles seen under SEM or TEM is a very small random sample from the bulk of nanoparticle batch produced.
(A) Transmission electron micrograph and (B) Scanning electron micrograph of PLGA-CURC formulated under optimum conditions.
In vitro release studies for PLGA-CURC prepared under optimal conditions
The drug release profile is another important criterion while formulating polymeric nanoparticles. The profile of curcumin release in PBS (pH
7.4) from the optimized formulation is illustrated in Figure . It was observed that the release consisted of an initial burst release phase corresponding to about 26% of drug release in the first hour, followed by a slow sustained release corresponding to 68% drug release in seven days and
75% in 10 days. Sustained release kinetics where 75% curcumin was released from curcumin-PBCN nanoparticles over 24 h has been reported by Sun et al. [38
]. In another study, Mohanty et al. (2010), showed 46% drug release in 24 h and 66% drug release over a period of 10 days from nanoparticlulate curcumin [39
]. Release of curcumin from PLGA-CURC was more uniform and sustained over the 10 day period of study. The burst release of curcumin may be due to the surface associated curcumin bound weakly to the surface of the nanoparticles which gets released first. The remaining amount of curcumin which is encapsulated within the structure was released in a controlled manner for the entire period of study (10 days). Dissolution diffusion of the drug from the matrices and the slow matrix erosion are the mechanisms thought responsible for the slower drug release kinetics from the nanoparticles.
Further, the release profile of curcumin from PLGA nanoparticles were investigated by using different release kinetic models: zero order, first order, Higuchi and Hixson-Crowell equations [26
], and their regression coefficient (r2
) was calculated from appropriate plots. The first order model describes the release to be concentration dependent while the Hixson– Crowell cube root model indicates a change in surface area or diameter due to erosion with progressive release of drug as a function of time. Release rate constants for burst release and sustained release are illustrated in Table . Comparing the amount of released curcumin with respect to time; for the burst release phase (first 6 h), PLGA-CURC followed the zero order model (R2
0.997). Higuchi kinetics model which states that diffusion is one of the major methods of drug release best described the controlled release phase (R2
0.996) during later part of the release which may be controlled by a combination of slow and gradual erosion and diffusion. Overall, the in vitro
release data indicates that PLGA-CURC is capable of releasing curcumin in a controlled manner over a period of 10 days.
Different kinetic models and regression coefficients of PLGA-CURC formulationsa
Cellular uptake of nanoparticle prepared under optimal conditions
In order to study the uptake of PLGA-CURC by different cancer cell lines, we investigated the ability of nanoparticles to be endocytosed by the cells. Figure A and B illustrate panels of the confocal microscope images of different cancer cell lines incubated with Nile red-labeled PLGA-CURC for 2 h. Our results depict robust uptake of the nanoparticles in all the cell lines. The cells incubated with Nile red-labeled PLGA-CURC exhibited either red (due to Nile red) or green (due to curcumin) fluorescence, depending upon the excitation wavelength.
Figure 9 Confocal images of different cancer cell lines incubated with PLGA-CURC - (A) Red :- nile red-labeled PLGA-CURC, Green:- curcumin, Bright field merged with cell nuclei stained with DAPI; (B) Merged images of nile red-labeled PLGA-CURC with DAPI; Merged (more ...)
Western blot analysis
A principal cellular target of curcumin in cancer cells is activated nuclear factor kappa B (NFκB) which is a family of five closely related proteins found in several dimeric combinations and bind to the NFκB consensus sequence on DNA [40
]. NFκB is translocated to the nucleus from cytosol, where it induces the expression of more than 200 genes that have been shown to suppress apoptosis and induce cellular transformation, proliferation, invasion and metastasis. Many of these activated target genes are critical for establishment of the early and late stages of aggressive cancers. We studied the mechanism of action of PLGA-CURC on breast cancer cell line, MDA MB 231, and compared the functional pathways affected by PLGA-CURC to what has been previously reported for free curcumin [38
]. The results of the Western blot analysis, as seen in Figure , depicts that PLGA-CURC was able to inhibit the translocation of NFκB from cytosol to nucleus in MDA MB-231 cells. The degree of inhibition with PLGA-CURC treatment was seen to be greater as compared to untreated cells, as depicted with fainter bands of NFκB corresponding to the nuclear extract from cells treated by PLGA-CURC. This result illustrates that the curcumin encapsulated within the PLGA-CURC retains its functional activity on encapsulation and subsequent release from the nanoparticles.
Figure 10 A) Western blot showing inhibition of translocation of NFκB (p65) from cytosol to nucleus in MDA MB231 cells treated with PLGA-CURC.B) Densiometric analysis of the western blot. (CC-Control cytosol extract, CN- Control nuclear extract, TC – (more ...)
PLGA-CURC were formulated to improve the bioavailability of curcumin. To evaluate this, male Sprague Dawley rats were administered PLGA-CURC nanoparticles intravenously (7.5 mg/kg equivalent curcumin nanoparticles). Blood samples were collected at predetermined time intervals and the concentration of curcumin was determined by HPLC analysis. Our results showed (Figure ) that the pharmacokinetic profile of PLGA-CURC in rats by i.v. administration followed two compartmental model. The area under the curve (AUC0-∞
) after i.v. injection of PLGA-CURC was found to be 6.139 mg/L h. This value is much higher compared to 3.16 mg/L h reported by Ma et al. (2007) for their micellar formulation of curcumin at a dose of 5 mg/kg [41
]. Duan et al. (2010) reported the area under the curve to be 3.302 mg/L h for a 5 mg/kg dose of their curcumin-PBCA nanoparticles. Both these groups reported AUC0-∞
of free curcumin at 10 mg/kg dose to be only 1.67 mg/L h and 1.92 mg/L h respectively [42
]. They also reported a higher elimination half life of curcumin when encapsulated within nanoparticles or micelles alongwith a decrease in clearance rate. Our pharmacokinetic results for PLGA-CURC showed the same trend. This is expected when the drug in circulation is restricted to the blood compartment because of being encapsulated with nanoparticles [42
]. The higher level of curcumin concentration observed in the case of PLGA-CURC nanoparticles might be explained by increased bioavailability as a function of increased aqueous dispersibility, smaller nanoparticle size and increased accumulation of nanoparticles in different organs together with sustained release of curcumin from them. Similar observations related to pharmacokinetic studies of curcumin or nanoparticles have been reported by various other groups [9
Figure 11 In vivo bioavailability of curcumin using PLGA-CURC nanoparticles. PLGA-CURC were administered intravenously to the rats at a dose of 7.5 mg/kg (n=6).
Storage stability of PLGA-CURC nanoparticles
The long term storage stability of the PLGA-CURC is an important parameter when scaling up the formulation. Nanoparticle formulations increase the surface area by many folds and this may lead to very high aggregation after long periods of storage. This poor long term stability may be due to different physical and chemical factors that may destabilize the system [43
]. Lyophilization is a promising approach for the stabilization of PLGA nanoparticles [44
]. For lyophilized nanoparticles, cryoprotectants serve as stabilizers during the freeze drying process. For our study, sucrose (5% w/v) and trehalose (2% w/v) were chosen as the cryoprotectants to prevent the hydrolytic instability, aggregation between nanoparticles, protection during processing and storage. After 6 months of storage with cryoprotectants at 4°C, the nanoparticles appear to be stable without any collapse or aggregation. Figure shows effect of long term storage on the particle size, encapsulation efficiency and drug loading of nanoparticles. We saw no major changes besides a slight increase in particle size and a slight decrease in encapsulation efficiency and drug loading. Therefore, PLGA-CURC formulated by our s-o/w emulsion solvent evaporation and diffusion technique was found to be stable for a long period of time.
The particle size, encapsulation efficiency and drug loading of PLGA-CURC against storage time at 4°C.
Gamma irradiation PLGA-CURC nanoparticles
Gamma irradiation is critical as it renders sterility to the nanoparticles before being injected into the body [45
]. There are many alternative techniques for sterilization but we chose γ-irradiation as it is known for its high penetration power and isothermal property of gamma rays that permits sterilization of even sensitive materials [46
]. However, γ-irradiation may have some effects on the nanoparticle size or drug loading. The changes in particle size and drug loading for low, medium and high exposures are graphed in Figure . Our results demonstrate that there were no statistically different changes observed between non-irradiated and different dose irradiated nanoparticles. Further, γ-irradiation did not alter the drug loading in the nanoparticles.
Figure 13 The particle size (A) and drug loading (B) of PLGA-CURC after γ-irradiation at doses 16.8 kGy for 241 minutes (Low), 25.3 kGy for 179 minutes (Medium) or 35.8 kGy for 241 minutes (High) (n=3).