3.3.1. Droplet Size

The droplet size ranged from 58.1 nm to 507.2 nm with the smallest droplet sizes observed when Tween® 80 was used at its maximum level of 0.9 in proportion in the surfactant-mixture design space whereas, the droplet size increased as Span® 20 content increased. A sharp increase in particle size was observed as ethanol content is increased in the region of the Span® 20 vertex as can be observed in the contour plot depicted in Figure 6. The model F-value of 8.91 indicated that the special quartic model for droplet size was significant and that there is only a 0.46% chance that a model F-value this large is due to noise. The *p*-value < 0.05 implied that the coefficients of the model terms were significantly different from zero, i.e., the effect of the model terms or combination of terms exerts an effect than can be estimated in the formulation composition. The model terms for the mixture A, B, C and A2BC were significant. The terms AB, AC, BC, AB2C and ABC<sup>2</sup> were not significant and the ANOVA data and results are summarized in the supplementary data. The lack of fit F-value of 3.59 implied that lack of fit was not significant. The predicted R<sup>2</sup> for droplet size was negative which implied that the mean data may be a better predictor for this response than using the model. However, adequate precision was >4 and was desirable showing that an adequate signal was observed and that the model could be used to navigate the design space and the model equation in terms of coded factors for droplet size is reported as Equation (1).

$$\begin{aligned} \text{Droplet size} &= 342.00 \text{A} + 90.75 \text{B} + 3328.66 \text{C} - 530.59 \text{AB-2941.83} \text{AC} - 3530.63 \text{BC} + 353.4 \text{B} \\ &\quad 52,488.71 \text{A}^2 \text{BC} - 11,742.82 \text{AB}^2 \text{C} - 32,873.13 \text{ABC}^2 \end{aligned} \tag{1}$$

*Pharmaceutics* **2020**, *12*, x FOR PEER REVIEW 12 of 22

**Figure 6.** Contour plot for droplet size showing the characterized region where ethanol content is ≤20%. **Figure 6.** Contour plot for droplet size showing the characterized region where ethanol content is ≤20%.

#### 3.3.2. Polydispersity Index 3.3.2. Polydispersity Index

to noise.

The PDI ranged between 0.119 to 0.75 in the design space with the highest PDI observed at the center-point for Tween® 80 and Span® 20 content with ethanol content used closest to the lower limit. As the ethanol content was increased in the upper region of the contour plot towards the Span® 20 The PDI ranged between 0.119 to 0.75 in the design space with the highest PDI observed at the center-point for Tween® 80 and Span® 20 content with ethanol content used closest to the lower limit.

effect and relationship of the model terms A, B, C, AB, AC, BC, A2BC, AB2C and ABC2 were significantly different from zero and could be estimated and the ANOVA data and results for this special quartic model are given in Table S4 of the Supplementary Material. The model equation in terms of coded factors for PDI is reported as Equation (2). The predicted R2 0.8335 is in reasonable agreement with the adjusted R2 0.9789. The lack of fit F-value 0.08 implies that lack of fit was not significant for the model and that there is a 56.02% chance that the lack of fit F-value this large is due

Polydispersity index = 0.2868A + 0.1155B + 10.85C + 2.19AB − 12.93AC − 10.78BC –

43.34A2BC + 1.016AB2C + 37.74ABC2 (2)

As the ethanol content was increased in the upper region of the contour plot towards the Span® 20 vertex, a general decrease in PDI is observed as depicted in (Figure 7). The model F-value of 86.11 indicates that the special quartic model for PDI was significant and that there is only a 0.01% chance that a model F-value this large was due to noise. A *p*-value of <0.05 implied that the coefficients of effect and relationship of the model terms A, B, C, AB, AC, BC, A2BC, AB2C and ABC<sup>2</sup> were significantly different from zero and could be estimated and the ANOVA data and results for this special quartic model are given in Table S4 of the Supplementary Material. The model equation in terms of coded factors for PDI is reported as Equation (2). The predicted R<sup>2</sup> 0.8335 is in reasonable agreement with the adjusted R<sup>2</sup> 0.9789. The lack of fit F-value 0.08 implies that lack of fit was not significant for the model and that there is a 56.02% chance that the lack of fit F-value this large is due to noise.

Polydispersity index = 0.2868A + 0.1155B + 10.85C + 2.19AB − 12.93AC − 10.78BC − 43.34*A* <sup>2</sup>*BC* + 1.016*AB*2*C* + 37.74*ABC*<sup>2</sup> (2)

*Pharmaceutics* **2020**, *12*, x FOR PEER REVIEW 13 of 22

**Figure 7.** Contour plot for polydispersity index (PDI) showing point prediction when ethanol content is ≤20%. **Figure 7.** Contour plot for polydispersity index (PDI) showing point prediction when ethanol content is ≤20%.

#### 3.3.3. Zeta Potential 3.3.3. Zeta Potential

is due to noise.

All components of the surfactant mixture exhibited a combined effect on the zeta potential and a linear relationship for this model is reported in Equation (3). The ethanol content exhibited the greatest effect on the zeta potential as observed by the magnitude of the coefficient for the term C and in the contour plot where 20% ethanol (C) was used. The contour plot for ZP is depicted in Figure 8 in which a significant region (in blue) revealed that as the concentration of Span® 20 in the surfactant mixture was increased, the ZP decreased and became more negative with the lowest negative point occurring at the upper limit of ethanol and lower limit of Tween® 80 content. The model F-value of 5.09 for ZP indicated that the linear model for ZP was significant and that there was only a 2.33% chance that a model F-value this large was due to noise. The probability *p*-value of < 0.05 implied the model terms were significant, the coefficients of the effect of each model term was significantly different from zero and could be estimated for this linear model, thus therefore only model terms A, B and C were significant within the formulation composition of these nanoemulsions. The equation All components of the surfactant mixture exhibited a combined effect on the zeta potential and a linear relationship for this model is reported in Equation (3). The ethanol content exhibited the greatest effect on the zeta potential as observed by the magnitude of the coefficient for the term C and in the contour plot where 20% ethanol (C) was used. The contour plot for ZP is depicted in Figure 8 in which a significant region (in blue) revealed that as the concentration of Span® 20 in the surfactant mixture was increased, the ZP decreased and became more negative with the lowest negative point occurring at the upper limit of ethanol and lower limit of Tween® 80 content. The model F-value of 5.09 for ZP indicated that the linear model for ZP was significant and that there was only a 2.33% chance that a model F-value this large was due to noise. The probability *p*-value of < 0.05 implied the model terms were significant, the coefficients of the effect of each model term was significantly different from zero and could be estimated for this linear model, thus therefore only model terms A, B and C were significant within the formulation composition of these nanoemulsions. The equation

in terms of coded factors for ZP is reported as equation 3. The predicted R2 of 0.1152 is not in close

the model could be used to navigate the design space. The lack of fit F-value 0.49 implies that lack of fit was not significant for the model and there is an 82.64% chance that a lack of fit F-value this large

Zeta potential = −23.10A – 17.08B − 29.48C (3)

in terms of coded factors for ZP is reported as equation 3. The predicted R<sup>2</sup> of 0.1152 is not in close agreement with the adjusted R<sup>2</sup> of 0.3530 suggesting a large block effect or possible problem with the model although, the adequate precision > 4 which is desirable and the adequate signal indicates that the model could be used to navigate the design space. The lack of fit F-value 0.49 implies that lack of fit was not significant for the model and there is an 82.64% chance that a lack of fit F-value this large is due to noise.

*Pharmaceutics* **2020**, *12*, x FOR PEER REVIEW 14 of 22

$$\text{Zeta potential} = -23.10 \text{A} - 17.08 \text{B} - 29.48 \text{C} \tag{3}$$

**Figure 8.** Contour plot for zeta potential (ZP) showing the point prediction when the ethanol content is ≤20%. **Figure 8.** Contour plot for zeta potential (ZP) showing the point prediction when the ethanol content is ≤20%.

#### *3.4. Optimization of Surfactant Mixtures and Assessment of Optimized Nanoemulsions 3.4. Optimization of Surfactant Mixtures and Assessment of Optimized Nanoemulsions*

nanoemulsions and assessed in in vitro release and characterization studies.

**Ethanol%**  *m/m*

the Supplementary Materials.

**20%** *m/m*

**Tween® 80%** *m/m*

F2 33.3 33.3 33.3 190.3 ±

F3 28.5 28.5 42.8 156.8 ±

F4 58.1 36.0 6.0 225.6 ±

F5 32.2 58.3 9.5 146.7 ±

**Formulation Span®**

The optimization function was used to predict optimum levels for each for the components of the surfactant mixture. The primary criterion used was that the ethanol content should be minimized in the surfactant mixture. The second criterion required that a droplet size of between 100 and 200 nm was desired. Finally, the third and fourth criteria required minimization of the PDI and ZP. Two optimized solutions were produced based on the desirability function and the proportions of the formulation composition for the two solutions are reported as batches F4 and F5. Batches F1, F2 and F3 were nanoemulsion formulations made using arbitrary surfactant mixtures when assessing the phase behavior for the S1, S2 and S3 mixtures, respectively. The five formulations that were manufactured and assessed are listed in Table 6 with their respective compositions. The specified optimization criteria (constraints) in Table S6 and solutions produced are shown Table S7 with their associated desirability in the Supplementary Material. The prediction error was calculated against the experimental values obtained to give the prediction error of the D-optimal design. An overlay plot that depicts the area in which the desired optimization criteria is met is shown on Figure S6 in The optimization function was used to predict optimum levels for each for the components of the surfactant mixture. The primary criterion used was that the ethanol content should be minimized in the surfactant mixture. The second criterion required that a droplet size of between 100 and 200 nm was desired. Finally, the third and fourth criteria required minimization of the PDI and ZP. Two optimized solutions were produced based on the desirability function and the proportions of the formulation composition for the two solutions are reported as batches F4 and F5. Batches F1, F2 and F3 were nanoemulsion formulations made using arbitrary surfactant mixtures when assessing the phase behavior for the S1, S2 and S3 mixtures, respectively. The five formulations that were manufactured and assessed are listed in Table 6 with their respective compositions. The specified optimization criteria (constraints) in Table S6 and solutions produced are shown Table S7 with their associated desirability in the Supplementary Material. The prediction error was calculated against the experimental values obtained to give the prediction error of the D-optimal design. An overlay plot that depicts the area in which the desired optimization criteria is met is shown on Figure S6 in the Supplementary Materials.

**PDI** 

0.444 ±

0.387 ±

0.342 ±

0.487 ± 0.003

0.402 ± 0.012

**Zeta Potential mV** 

−31.9 ±

−24.1 ±

**Efavirenz Content mg/mL** 

0.003 −35.4 ± 0.9 377 ± 4.9 0.989 ± 0.006

0.016 −34.4 ± 0.7 437 ± 13.1 1.040 ± 0.009

0.048 −41.0 ± 0.9 571 ± 18.7 1.040 ± 0.012

3.12 329 ± 9.45 0.817 ± 0.007

2.33 334 ± 11.2 0.833 ± 0.009

**Mass of 1 mL of Nanoemulsion g** 

**Droplet Size nm** 

0.7

2.0

23.4

16.8

25.3

**Table 6.** Composition of the surfactant mixtures used for the manufacture of 10% flaxseed


**Table 6.** Composition of the surfactant mixtures used for the manufacture of 10% flaxseed nanoemulsions and assessed in in vitro release and characterization studies.

#### *3.5. Characterization and Assessment of Nanoemulsions 3.5. Characterization and Assessment of Nanoemulsions*

Nanoemulsion F3 was able to incorporate the largest amount of EFV of 571 mg/mL suggesting that the high ethanol content improved the solubility of EFV and miscibility of flaxseed oil, Tween® 80 and Span® 20 in the nanoemulsion. A decrease in EFV loading was observed as the proportion of ethanol in the composition was decreased. The highest PDI of 0.487, although arguably monodisperse for batch F4, is bimodal (Figure 9). The droplet size distribution of F4 could be attributed to the high Span® 20 content due to a shift in HLB for these surfactant mixtures and the effect to thermodynamic stability [51]. A general decrease in the ZP was observed as the proportion of ethanol used was increased for batches F1 to F3 suggesting that the composition of F1, F2 and F3 would produce stable nanoemulsions over the long term. All five formulations exhibited ZP values < −20 mV suggesting the nanoemulsions that were produced were likely to be stable. The negative charge of nanoemulsion droplets may be useful for macrophage targeting since macrophages identify and take up negatively charged particles [52]. Macrophages are key factors in HIV infection and are significant cellular reservoirs of the HIV [53]. Emulsion droplets with a ZP of approximately ± 20 mV exhibit only short-term stability, with the tendency for the droplets to flocculate and coalesce [54]. To administer the recommended maximum adult dose of 600 mg formulation F3 would require a mass of 1.09 g of the nanoemulsion to be administered. The total mass is considerably lower and more convenient than the commercially available 600-mg EFV tablets manufactured by Cipla Medpro, Aurobindo, Adcock Ingram and Aspen Pharmacare for which tablets mass measured was 1.34 ± 0.09, 1.25 ± 0.11 g, 1.20 ± 0.08 g and 1.106 ± 0.045 g, respectively for (*n* = 20) suggesting that the nanoemulsion may produce a more convenient dosage form size for patients to use. Nanoemulsion F3 was able to incorporate the largest amount of EFV of 571 mg/mL suggesting that the high ethanol content improved the solubility of EFV and miscibility of flaxseed oil, Tween® 80 and Span® 20 in the nanoemulsion. A decrease in EFV loading was observed as the proportion of ethanol in the composition was decreased. The highest PDI of 0.487, although arguably monodisperse for batch F4, is bimodal (Figure 9). The droplet size distribution of F4 could be attributed to the high Span® 20 content due to a shift in HLB for these surfactant mixtures and the effect to thermodynamic stability [51]. A general decrease in the ZP was observed as the proportion of ethanol used was increased for batches F1 to F3 suggesting that the composition of F1, F2 and F3 would produce stable nanoemulsions over the long term. All five formulations exhibited ZP values < −20 mV suggesting the nanoemulsions that were produced were likely to be stable. The negative charge of nanoemulsion droplets may be useful for macrophage targeting since macrophages identify and take up negatively charged particles [52]. Macrophages are key factors in HIV infection and are significant cellular reservoirs of the HIV [53]. Emulsion droplets with a ZP of approximately ± 20 mV exhibit only shortterm stability, with the tendency for the droplets to flocculate and coalesce [54]. To administer the recommended maximum adult dose of 600 mg formulation F3 would require a mass of 1.09 g of the nanoemulsion to be administered. The total mass is considerably lower and more convenient than the commercially available 600-mg EFV tablets manufactured by Cipla Medpro, Aurobindo, Adcock Ingram and Aspen Pharmacare for which tablets mass measured was 1.34 ± 0.09, 1.25 ± 0.11 g, 1.20 ± 0.08 g and 1.106 ± 0.045 g, respectively for (*n* = 20) suggesting that the nanoemulsion may produce a more convenient dosage form size for patients to use.

**Figure 9.** Particle-size distribution for batches F1, F2, F3, F4 and F5. **Figure 9.** Particle-size distribution for batches F1, F2, F3, F4 and F5.

#### *3.6. Transmission Electron Microscopy 3.6. Transmission Electron Microscopy*

Transmission electron microscopy revealed the presence of largely spherical droplets of lipid as depicted in Figure 10 with an average droplet size in close agreement with that determined using Transmission electron microscopy revealed the presence of largely spherical droplets of lipid as depicted in Figure 10 with an average droplet size in close agreement with that determined

associated with CNS tissue by distributing and perfusion into such tissues [55,56]. The bimodal distribution observed for droplet size distribution for F4 can be explained by the differences in the surfactant composition of the formulation. Ethanol exhibited the largest effect on solvent capacity and miscibility of the mixture for both surfactants, the oil phase and EFV. Rigid interfacial films give rise to bimodal distributions [57] and the greater the ethanol content, the more flexible the interfacial film of the immiscible phase. The bimodal distribution for F4 can be explained by the solvent capacity of the system for the lipophilic phase, Tween® 80 and Span® 20 which become miscible with addition of ethanol. As formulation F4 includes only a small proportion of ethanol, some proportion of the using dynamic light scattering. Smaller droplets in the size range 20–100 nm were present in the nanoemulsion dispersion in lower numbers as reflected in the droplet size distribution in Figure 8. These small droplet sizes may add to therapeutic possibilities by reducing the viral load in reservoirs associated with CNS tissue by distributing and perfusion into such tissues [55,56]. The bimodal distribution observed for droplet size distribution for F4 can be explained by the differences in the surfactant composition of the formulation. Ethanol exhibited the largest effect on solvent capacity and miscibility of the mixture for both surfactants, the oil phase and EFV. Rigid interfacial films give rise to bimodal distributions [57] and the greater the ethanol content, the more flexible the interfacial film of the immiscible phase. The bimodal distribution for F4 can be explained by the solvent capacity of the system for the lipophilic phase, Tween® 80 and Span® 20 which become miscible with addition of ethanol. As formulation F4 includes only a small proportion of ethanol, some proportion of the lipophilic phase (span® 20 + 10% flaxseed oil) could possibly have been dispersed in agglomerates of a different size within the nanoemulsion mixture and larger than the small peaks of smaller particle sizes of F1, F2 and F3. Given the large concentration of ethanol in F3, the bimodal effects are seen to produce agglomerates within the nanoemulsion mixture of a smaller droplet size. *Pharmaceutics* **2020**, *12*, x FOR PEER REVIEW 16 of 22 lipophilic phase (span® 20 + 10% flaxseed oil) could possibly have been dispersed in agglomerates of a different size within the nanoemulsion mixture and larger than the small peaks of smaller particle sizes of F1, F2 and F3. Given the large concentration of ethanol in F3, the bimodal effects are seen to produce agglomerates within the nanoemulsion mixture of a smaller droplet size.

**Figure 10.** Transmission electron micrograph of nanoemulsion F5. The images for F1, F2, F3 and F4 are reported in the Supplementary Materials. **Figure 10.** Transmission electron micrograph of nanoemulsion F5. The images for F1, F2, F3 and F4 are reported in the Supplementary Materials.

#### *3.7. Raman Spectroscopy 3.7. Raman Spectroscopy*

The Raman spectra for EFV and EFV loaded nanoemulsions is reported in the Supplementary Materials (Figure S7). The EFV skeleton stretching vibrations were not affected by encapsulation of EFV in the nanoemulsion implying no interaction between EFV and lipids used in the formulation. The functional groups for EFV detected using Raman spectroscopy revealed that all expected signals observed for pure efavirenz and blank nanoemulsions were present and were in agreement with previously reported spectra [58]. The signal for the CH2 (A) functional group at 3093 cm−1, the C≡C (B) bond at 2250 cm−1, the (C=O) (C) bond at 1750 cm−1 and the C–H stretch at approximately 1000 cm−1 reflect the presence of EFV [59]. A comparison of experimentally determined vibrational wavenumbers for EFV is listed in the Supplementary Materials (Table S8). The Raman spectra for EFV and EFV loaded nanoemulsions is reported in the Supplementary Materials (Figure S7). The EFV skeleton stretching vibrations were not affected by encapsulation of EFV in the nanoemulsion implying no interaction between EFV and lipids used in the formulation. The functional groups for EFV detected using Raman spectroscopy revealed that all expected signals observed for pure efavirenz and blank nanoemulsions were present and were in agreement with previously reported spectra [58]. The signal for the CH<sup>2</sup> (A) functional group at 3093 cm−<sup>1</sup> , the C≡C (B) bond at 2250 cm−<sup>1</sup> , the (C=O) (C) bond at 1750 cm−<sup>1</sup> and the C–H stretch at approximately 1000 cm−<sup>1</sup> reflect the presence of EFV [59]. A comparison of experimentally determined vibrational wavenumbers for EFV is listed in the Supplementary Materials (Table S8).

#### *3.8. Fourier-Transform Infrared Spectroscopy (FT-IR) 3.8. Fourier-Transform Infrared Spectroscopy (FT-IR)*

crystallization of EFV in solution.

*3.9. In Vitro Efavirenz Release* 

To understand the possibility of chemical interactions between the drug and nanoemulsion mixture, blank nanoemulsion, pure EFV, EFV loaded nanoemulsions and EFV nanoemulsions after dissolution testing were harvested left to dry in an open petri dish under room temperature over 72 h and characterized by FT-IR. The FT-IR spectrum of EFV loaded nanoemulsion in Figure S8 (in the Supplementary Material) showed the characteristic peaks of alkyne at 2247.63 cm−1, C–H stretch at 3000 cm−1, C–F stretch at 1400 cm−1, tertiary amide at 1602 cm−1 and C=O (D) at 1750 cm−1 [60]. A table To understand the possibility of chemical interactions between the drug and nanoemulsion mixture, blank nanoemulsion, pure EFV, EFV loaded nanoemulsions and EFV nanoemulsions after dissolution testing were harvested left to dry in an open petri dish under room temperature over 72 h and characterized by FT-IR. The FT-IR spectrum of EFV loaded nanoemulsion in Figure S8 (in the Supplementary Material) showed the characteristic peaks of alkyne at 2247.63 cm−<sup>1</sup> , C–H stretch at 3000 cm−<sup>1</sup> , C–F stretch at 1400 cm−<sup>1</sup> , tertiary amide at 1602 cm−<sup>1</sup> and C=O (D) at 1750 cm−<sup>1</sup> [60].

peaks is diminished in the crystalline product harvested following dissolution testing and interaction of the EFV nanoemulsion with aqueous media resulting in conversion of electrons on carbon atoms changing from sp2 to sp3 [61]. The decrease in intensity of the O–H stretch detected as a broad band at 3500 cm−1 following dissolution testing may be due to the loss of ethanol from the nanoemulsion into the aqueous dissolution medium which results in an increase in the rate of nucleation and

In vitro release testing revealed burst release of EFV for batches F1, F2, F4, F5 and for pure EFV within the first two hours of commencement of testing. Dissolution testing was conducted for the A table of the comparison of the peaks reported in literature and those experimentally found are given in Table S8 in the Supplementary Material. The absorption band at approximately 3000 cm−<sup>1</sup> was assigned to C–H stretching of the methylene group for Tween® 80 and Span® 20, the intensity of these peaks is diminished in the crystalline product harvested following dissolution testing and interaction of the EFV nanoemulsion with aqueous media resulting in conversion of electrons on carbon atoms changing from sp2 to sp3 [61]. The decrease in intensity of the O–H stretch detected as a broad band at 3500 cm−<sup>1</sup> following dissolution testing may be due to the loss of ethanol from the nanoemulsion into the aqueous dissolution medium which results in an increase in the rate of nucleation and crystallization of EFV in solution.
