Improved percutaneous delivery of azelaic acid employing microemulsion as nanocarrier: formulation optimization, in vitro and in vivo evaluation

Abstract

The present study aimed to develop and optimize a microemulsion (ME) nanocarrier system as a topical vehicle for azelaic acid (AZA) to improve its skin location and therapeutic efficacy. A D-optimal mixture experimental design was utilized to optimize ME for realizing maximum skin retention and appropriate droplet size. Three formulation variables, Smix X₁ (a mixture of Span 20/ethanol, 1 : 9, w/w), water X₂ and Oil X₃ (Capryol 90), were included in the design, while the three responses were skin retention (Y₁), amount of AZA in collection medium after 24 h (Y₂) and mean particle size (Y₃). The values of the formulation components (X₁, X₂ and X₃) were 50.3%, 13.5% and 36.2%, respectively. In in vitro studies, the optimal ME revealed a much higher release rate, and enhanced skin targeting and penetration effects of AZA relative to control formulations (ethanol solution based gel and commercial cream). An attenuated total reflectance Fourier-transform infrared spectroscopy study further confirmed to us that the vehicles could transport the active agents across the stratum corneum (SC) layer by changing the amount and arrangement of lipid within the SC. In addition, a skin irritation test and pharmacodynamics studies were conducted, and the results suggested that the optimal ME exhibited a prominent therapeutic effect compared to control formulations, without any irritant response.

---

1. Introduction

Rosacea is a common chronic inflammatory dermatosis, characterized by transient or persistent central facial erythema, visible blood vessels, and often papules and pustules.¹ The cause of rosacea is still unknown; however, the available evidence supports that genetic and environmental factors (such as sun exposure, drinking alcohol, and use of cosmetics) are responsible for the etiology of this skin disorder.² Rosacea affects mostly facial skin, which leads to sufferers having difficulties with social contact.³ Moreover, the current treatment of rosacea has been claimed to be empirical and imperfect.

Azelaic acid (1,7-heptanedicarboxylic acid, AZA) is a saturated, straight-chained C9-dicarboxylic acid that has been reported to be the active pharmaceutical ingredient in a number of prescription drugs for the treatment of rosacea.⁴ However, AZA, with its commercial formulations of 15% gel (FINACEA®) and 20% cream (Skinoren®), has limited penetration across the stratum corneum (SC) due to poor bioavailability, mainly caused by a low fraction of dissolved drug and poor skin permeability.^2,5 Theoretically, suitable percutaneous permeation is an essential factor for pharmaceutical agents to achieve satisfactory therapeutic effect. Topical delivery systems aiming to promote cutaneous penetration of AZA are necessary to maximize its biological efficacy. Meanwhile, considering the local nature of skin disorders, it is advisable to restrict the drug to the site of application for localized delivery.

Effective penetration of active agents through the SC is a major challenge in topical drug delivery.⁶ On this matter, a number of research works have been carried out to increase penetration through the SC, including work on chemical modification,⁷ addition of a penetration enhancer or retardant,⁸ micro-needles,⁹ and microwaves.¹⁰ Recently, nanoscale vehicles have attracted significant attention as delivery agents for active molecules, e.g. liposomes,¹¹ solid lipid nanoparticles,¹² and microemulsions.^13–16

Microemulsions (MEs) have been proven to have significant potential for increasing the penetration of lipophilic, hydrophilic, and amphiphilic substances into and through the skin, compared to conventional vehicles.^17,18 MEs are optically isotropic and thermodynamically stable nanosized structure-mixtures of an aqueous phase, oil phase and amphiphile(s).^19,20 Several mechanisms have been proposed to explain the advantages of ME over conventional vehicles. First, the ingredients of a ME could interfere with the diffusional barrier of the SC and improve cutaneous permeation of a drug by acting as permeation enhancers.¹³ Second, the increased thermodynamic activity of drugs incorporated in ME formulations is a significant driving force for drug release and skin penetration.²¹ Third, small sized droplets could settle into close contact with the skin, which leads to a considerable increase in surface area and hence improves absorption.²² Also, the continuously and spontaneously fluctuating interface of MEs enables high drug mobility and subsequently enhances the drug diffusion process.²³

The aim of this work was to optimize a ME nanocarrier system for AZA that provided a skin targeting effect and maximum dermal therapeutic effect. ME formulations were developed by constructing pseudo-ternary phase diagrams and were optimized by D-optimal design based on the maximum drug amount in skin layers, appropriate skin penetration and small particle size. The optimized formulation was characterized according to its droplet size, size distribution and pH value. Attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) study was carried out to elucidate the interaction mechanism between the ME and skin. In addition, the optimized ME formulation, ethanolic solution based gel and commercial cream were evaluated for their in vitro skin permeation, skin sensitivity, and pharmacodynamics for comparison purposes.

2. Materials and methods

2.1. Materials

Azelaic acid (MW 188, 99% purity) was purchased from Huabei reagent Co., Ltd (Tianjin, China). Capryol® 90 with a purity of 98% was a kind gift from Gattefossé (Saint-Priest, France). Ethanol (99% purity) and Span 20 (with a purity of 98%) were obtained from Jiangtian pharmaceutical reagent Co., Ltd (Tianjin, China). Klucel® MF was purchased from Hercules, Inc. (Wilmington, DE, USA). 20% AZA commercial cream (Skinoren®) was obtained from Bayer Co., Ltd (Taiwan). All other reagents were of analytical grade.

2.2. Skin membranes and animals

Abdominal porcine skin was obtained from a pig less than one month old. After removing the hair and the subcutaneous tissue, the skin was washed with normal saline, divided into small pieces and stored at −20 °C until use. Wistar rats (of about 200 ± 20 g) and male Kun-Ming mice (weighing 20 ± 2 g) were purchased from the Chinese Academy of Medical Sciences (Tianjin, China) and used for skin irritation tests and to determine the in vivo anti-rosacea therapeutic effects, respectively. All work performed with animals was in accordance with and approved by the Institutional Animal Care and Use of Tianjin University.

2.3. Construction of pseudo-ternary phase diagram and preparation of formulations

The pseudo-ternary phase diagram was constructed based on the oil phase (Capryol 90), surfactant (Span 20), cosurfactant (ethanol) and water. The mass ratio of surfactant to cosurfactant (Smix) was fixed at 1 : 9. Then, the oil phase was mixed with Smix at w/w ratios ranging from 1 : 9 to 9 : 1. Finally, 1 g of oil/Smix mixture in an appropriate ratio was titrated with water drop by drop under magnetic stirring at ambient temperature. The resultant mixtures were examined according to their visual appearance. Usually, the system that was a transparent and low viscous solution was defined as the ME region. In contrast, the turbid sample was identified as a conventional emulsion. The boundary point between the ME region and emulsion region was determined and the corresponding component ratio was recorded to plot the pseudo-ternary phase diagram.

When preparing drug-loaded ME formulations, 10% (w/w) AZA was dissolved in the oil/Smix mixture. Then, an appropriate amount of water was added under magnetic stirring to prepare the ME formulations.

An ethanolic solution-based gel (ESBG) containing the same AZA concentration (10%, w/w) was prepared and utilized as a control formulation. Klucel® MF was added to bidistilled water under stirring until complete integration. The obtained gel was diluted with an equal amount of ethanol solution followed by the addition of AZA, resulting in a final AZA concentration of 10% (w/w).

2.4. Formulation optimization of AZA-loaded MEs

A D-optimal mixture experimental study was designed based on a three component system: Smix X₁ (a mixture of Span 20/ethanol, 1 : 9, w/w), aqueous phase X₂ (water) and oil phase X₃ (Capryol 90). The total concentration of the three components was 100%. Based on the obtained ME region in the phase diagram, the range of each component was selected as follows: X₁ (40–80%), X₂ (0–30%), and X₃ (20–60%) (this region is shown in Fig. 1). The amount of skin retention of AZA at 24 h (Y₁), amount of AZA in the collection medium after 24 h (Y₂) and mean particle size (Y₃) were used as the responses (dependent variables). The responses of all model formulations were treated with Design-Expert software (version 7; Stat-Ease, Inc, Minneapolis, MN). Suitable models for D-optimal design included linear, quadratic, special cubic and cubic models. The best fitting mathematical model was selected by comparing statistical parameters including the standard deviation (SD), multiple correlation coefficient (R²), adjusted multiple correlation coefficient (adjusted R²) and the predicated residual sum of square (PRESS), provided by Design-Expert software. Since the PRESS value indicates how well the model fits the data, the value of the selected model should be the smallest among the models.²⁴ The base design consisted of 16 runs (Table 1).

Fig. 1 Pseudo-ternary phase diagram showing a W/O microemulsion region (area surrounded by pink line) made up of Capryol 90 (oil phase), water, and the mixture of Span 20 (surfactant) and ethanol (cosurfactant) at a fixed mass ratio of 1 : 9. The area surrounded by the blue line was used for D-optimal design.

Table 1 The formulations of mixture design and their characterization results

No.	Smix (X1)	Water (X2)	Oil (X3)	Skin retention (μg) (Y1)	Amount of AZA in collection medium at 24 h (μg) (Y2)	Particle size (nm) (Y3)
1	68.27	3.49	28.24	317.86	353.42	0
2	80	0	20	202.91	261.67	0
3	65.37	14.63	20	403.56	663.65	9
4	40	9.81	50.19	289.17	472.65	1.48
5	80	0	20	202.91	161.67	0
6	53.98	13.59	32.43	593.26	573.95	4.36
7	50.49	0	49.51	120.67	193.38	0
8	53.98	13.59	32.43	591.47	583.93	4.36
9	65.37	14.63	20	503.56	663.65	8.55
10	50	30	20	543.89	792.74	12
11	40	0	60	111.49	222.33	0
12	40	20.28	39.72	385.43	520.91	7.75
13	40	0	60	123.67	246.30	0
14	40	30	30	514.58	699.58	12
15	60.32	0	39.68	214.40	327.17	0
16	60.32	0	39.68	234.45	359.23	0

---

2.5. Evaluation of prepared formulations

A Malvern Mastersizer (Nano ZS90, Malvern Instruments, Malvern, UK) was used for determining the droplet size for D-optimal design. The optimized ME formulation was characterized according to its droplet size, size distribution profile and zeta potential. The pH values of the optimized ME formulation, ESBG and marketed cream were determined using a digital pH-meter (PHS-3C, Shenbang Instrument Corporation, Shanghai, China) at 25 ± 2 °C.

2.6. Stability assay

The optimal AZA-loaded ME was preserved in a glass vial with a sealing cap and was kept long-term under conditions of 25 ± 2 °C/60 ± 5% RH. The physical stability of the ME formulation was assessed in terms of its appearance, droplet size and polydispersity index (PDI) at predetermined time intervals of 0, 1, 2, and 3 months. For chemical stability, the concentration of AZA in the ME was determined by HPLC analysis at each predetermined time interval.

2.7. In vitro skin permeation studies

Porcine skin samples were mounted on Franz Diffusion Cells with the SC side facing the donor chamber (diffusion area = 1.77 cm²). The receptor medium was 17.6 mL of normal saline under constant magnetic stirring at 500 rpm. After equilibration of the skin samples with normal saline for 1 h at 37 ± 0.5 °C, finite doses (25 mg cream and 50 mg of the optimal ME formulation and ESBG, which correspond to 5 mg AZA, respectively) were applied to the skin surface (n = 6). At predetermined time intervals (4, 6, 8, 10, 12 and 24 h), approximately 0.5 mL of the receptor medium was withdrawn for HPLC analysis and was compensated for with an equal volume of fresh normal saline. The remaining formulation on the skin surface was wiped with a cotton ball soaked with methanol–water (40/60, v/v) after incubation for 24 h. The tape-stripping method was employed to remove the SC layer.²² The skin was stripped with 15 pieces of adhesive tape and all the tapes except for the first one were digested with methanol–water (40/60, v/v), then filtered for analysis. After removal of the SC, the remaining skin samples were minced, vortexed with 5 mL of methanol and centrifuged to extract residual AZA in the epidermis and dermis. The supernatants were collected and filtered for analysis. The permeation rate of AZA (flux, μg cm⁻² h⁻¹) through porcine skin was calculated from the slope of the linear portion of the plot of cumulative amount permeated through the skin per unit area versus time.

2.8. HPLC method

A Waters e2695 series HPLC with a UV 2489 detector (Waters, USA) was used for AZA method validation. The optimized chromatographic conditions were as follows: 250 mm × 4.6 mm stainless steel C18 column (I.D., 5 μm, Thermo, USA); column temperature of 35 °C; 20 μL injection volume; detection wavelength of 215 nm; mobile phase of acetonitrile and phosphate buffer (pH 3.0, 50 mM) at 25 : 75 (v/v); flow rate of 1.0 mL min⁻¹.

For in vitro studies, the peak area (y) correlated linearly with AZA concentration (x, μg mL⁻¹) in the range of 5.0–100.0 μg mL⁻¹ with a mean correlation coefficient of 0.9999. The regression equation of the calibration curve was y = 586.54x − 317.69, with a recovery of 99.36%.

2.9. Attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) study

To prepare the SC sample for ATR-FTIR study, the SC was firstly separated by placing the skin sample in 0.5% trypsin (type I, Sigma Aldrich) in phosphate-buffered saline, pH 7.4, for 4 h.²⁵ The obtained SC sheet was cleaned with deionized water and dried in a desiccator for 12 h. Then, the SC samples were incubated with different formulations for 24 h using diffusion cells as described in Section 2.6. All experiments were performed in triplicate. The samples were mixed with KBr to make pellets and were measured on an FTIR spectrometer (Bruker EQUINOX, Germany) with a spectral resolution of 4 cm⁻¹. The absorbance was measured in the region from 400 cm⁻¹ to 4000 cm⁻¹ at 37 °C.

2.10. Skin irritation test

To determine the skin compatibility of the developed formulations, a skin irritation test was carried out based on histopathological examination. The hair on the dorsal side (2 cm × 3 cm) of Wistar rats was carefully removed without damaging the skin.²⁶ The control group was treated with normal saline while other groups were treated with the optimized ME formulation, ESBG and commercial cream (containing 5 mg of AZA), respectively, three times a day for three consecutive days (n = 3). These formulations were uniformly spread within an area of 1.77 cm². After 3 days, the animals were observed for any signs of itching or change in skin such as erythema, papules, and dryness. Then, the rats were sacrificed using carbon dioxide gas. The test skin was removed, fixed and stored in formaldehyde (10%, v/v). Tissue specimens were processed routinely and embedded in paraffin wax. The paraffin blocks were cut serially at 10 μm. Sections were stained with hematoxylin and eosin (H&E) and examined using a light microscope (Olympus BX-51, Japan).

2.11. Pharmacodynamics studies

The croton oil inflammation model was used to induce a rosacea model.^27,28 Briefly, 10 μL of croton oil in acetone (5% v/v) was painted on the inner surface of the right ears in groups a–e, while the left ears were used as the control. Fifteen minutes later, 60 μL of blank ME, AZA-loaded ESBG (10%, w/w) and AZA-loaded ME (10%, w/w) were topically applied to groups b, c, and d, respectively. Commercial cream (20%, 25 mg) was administrated to group e. At 4, 8 and 24 h, ear thickness was measured near the top of the ear distal to the cartilaginous ridges. The change in ear thickness from the control was taken as an edema index. The ear tissue samples were collected after 24 h and submitted to histopathological analysis.

2.12. Data analysis

At least three to six replicates of each experiment were carried out. All results were reported as mean ± SD. A paired two-tailed Student’s t-test was employed to calculate the statistical significance. The level of significance was set as p < 0.05.

3. Results and discussion

3.1. Construction of pseudo-ternary phase diagram

Pseudo-ternary phase diagrams were constructed to determine the components and concentration ranges for the ME. Based on the optimization study of pseudo-ternary phase diagrams in our lab (unpublished data), the optimized pseudo-ternary phase diagram is shown in Fig. 1. The ME region was observed near the surfactant vertex characterized by high surfactant content and low water content. In other words, water-in-oil (W/O) MEs were easily formed at high Smix content. The maximal water solubilization capacity of this W/O system was nearly 40%, which might be ascribed to the excellent intersolubility of water and ethanol. Different ME formulations in the area designated by the blue lines were prepared and optimized based on D-optimal design.

3.2. Formulation optimization of MEs using D-optimal design

D-optimal design is an efficient method for the optimization of pharmaceutical formulations, which could clarify the relationship between independent variables and dependent variables in a formulation. In our study, a D-optimal mixture experimental design was conducted to rapidly obtain the optimal ME formulation. Smix (a mixture of Span 20/ethanol, 1 : 9, w/w) (X₁), water (X₂) and Capryol 90 (X₃) were chosen as formulation variables, and the mean time of skin retention (Y₁), amount of AZA in the collection medium after 24 h (Y₂) and mean particle size (Y₃) were selected as responses (dependent variables). The responses of these formulations were summarized in Table 1.

The independent and response variables were related using a polynomial equation with statistical analysis using the Design-Expert software (version 7; Stat-Ease, Inc, Minneapolis, MN). The equation that fitted the data was as follows:

Y = b₁X₁ + b₂X₂ + b₃X₃ + b₄X₁X₂ + b₅X₁X₃ + b₆X₂X₃ + b₇X₁X₂X₃ + b₈X₁X₂(X₁ − X₂) + b₉X₁X₃(X₁ − X₃) + b₁₀X₂X₃(X₂ − X₃) (1)

where b₁ to b₁₀ are the coefficients computed from the observed experimental values of Y. Coefficients with one factor represent the effect of that particular factor while the coefficients with more than one factor represent the interaction between those factors. A positive sign in front of the factors indicates synergistic effects while a negative sign indicates an antagonistic effect of the factors.²⁹

In our study, drug accumulation in the skin layers was considered to be the most significant factor for evaluating the efficiency of the formulations. As shown in Table 1, the amount of skin retention of AZA released from ME formulations varied from 111.49 to 593.26 μg, inferring that the three independent factors had a profound effect on the amount of skin retention of AZA. The approximation of the response values of Y₁ based on the Sp. cubic model was the most suitable due to its smallest PRESS value (Table 2). The related regression equation was:

Y₁ = 204.20X₁ + 546.05X₂ + 111.33X₃ + 510.31X₁X₂ + 170.11X₁X₃ + 314.46X₂X₃ + 5007.79X₁X₂X₃ (2)

Table 2 Model summary statistics of the measured response

Response	Model	SD	R^2	Adjusted R^2	PRESS
Y1	Linear	95.18	0.7362	0.6957	161 000
	Quadratic	69.83	0.8908	0.8362	104 600
	Special cubic	34.37	0.9762	0.9603	31320.77
	Cubic	40.45	0.9780	0.9450	575 200
Y2	Linear	83.69	0.8305	0.7655	145 400
	Quadratic	51.00	0.9580	0.9371	65780.87
	Special cubic	52.97	0.9593	0.9321	71972.51
	Cubic	57.44	0.9681	0.9202	1 726 000
Y3	Linear	1.34	0.9274	0.9162	32.66
	Quadratic	1.19	0.9556	0.9335	48.09
	Special cubic	1.01	0.9714	0.9523	39.18
	Cubic	0.32	0.9980	0.9951	64.77

---

The positive values of all coefficients confirmed the synergistic effect of the three independent variables on Y₁. In addition, it was obvious that the term X₁X₂X₃ had the greatest effect on this response with the largest coefficient of 5007.79, which could be confirmed from the 2D contour diagram that illustrated the effect of varying the ratios of X₁, X₂, and X₃ on the skin retention of MEs (Fig. 2a). As was indicated by the central solid portion of the plot, the ME formulations at moderate levels of oil, Smix and water could exhibit higher skin retention, which represented higher therapeutic efficacy.

Fig. 2 2D contour plots for the effects of variables on the skin retention after 24 h (μg) (a), amount of AZA in the collection medium after 24 h (μg) (b) and particle size (nm) (c) of the W/O ME.

The amount of AZA in the collection medium after 24 h (Y₂) of the different ME formulations ranged from 161.67 to 792.74 μg (Table 1). As presented in Table 2, the quadratic model was the most appropriate mathematical model for Y₂, with the obtained regression equation:

Y = 217.90X₁ + 736.56X₂ + 230.15X₃ + 1048.47X₁X₂ + 290.91X₁X₃ + 337.94X₂X₃ (3)

The coefficient of X₁X₂ for this response was the largest one, indicating the positive effect of the combination of Smix and water content on the drug penetration into the receptor medium. From the 2D contour plot (Fig. 2b), we could observe that moderate levels of the three factors indicated a lower cumulative amount of AZA in the receptor medium, which represented less systemic side effects. For the mean particle size, the cubic model was the most suitable model based on the largest R² value (R² = 0.9980, Table 2). The regression equation was presented as follows:

Y = −0.034X₁ + 6.26X₂ + 0.020X₃ + 31.91X₁X₂ − 0.69X₁X₃ + 18.79X₂X₃ − 97.55X₁X₂X₃ − 15.26X₁X₂(X₁ − X₂) − 5.20X₁X₃(X₁ − X₃) + 38.89X₂X₃(X₂ − X₃) (4)

According to the 2D contour plot (Fig. 2c), the water content provided the largest contribution to the mean droplet size. In other words, an increasing amount of water resulted in nonlinear escalations in particle size. In fact, the mean droplet size of formulations containing less than 4% water, as well as water-free systems (mixtures of the surfactant, cosurfactant and oil, S/COS/O-mix), was not measurable. As water content increased, the droplet size of the ME formulations also increased, indicating a swelling process taking place within the droplets at high aqueous contents.³⁰

In order to obtain optimal ME formulations with maximum skin targeting effect and minimum skin permeation, the response Y₁ should be maximized (>600 μg) while Y₂ should be minimized (<600 μg). The S/COS/O-mixtures resulted in significantly lower AZA permeation relative to ME droplets (Fig. 2a and b), which demonstrated that the presence of nanosized droplets makes a prominent contribution to the percutaneous penetration of drugs.³¹ Thus, the response Y₃ should have an optimal intermediate range (5–10 nm) to ensure the formation of ME droplets, resulting in maximum skin retention with less systemic side effects. Based on these conditions, the three responses were then combined to determine an overall optimum region (Fig. 3). According to the selection criteria, an ME with optimal drug skin retention, appropriate permeated amount of drug and droplet size was considered to be the optimal formulation. An optimal response was found with Y₁, Y₂ and Y₃ of 571.64 μg, 573.97 μg and 3.78 nm at X₁, X₂ and X₃ value of 50.3%, 13.5% and 36.2%, respectively (Table 3). In order to assess the reliability of the developed mathematical model, a microemulsion formulation was formed corresponding to the above mentioned factor levels. The experimental values of Y₁, Y₂ and Y₃ were 593.57 μg, 584.69 μg, and 3.83 nm, respectively. The predicted and experimental values demonstrated a small percentage error of 3.69%, 1.83% and 1.32%, respectively. In addition, a good agreement was obtained between the model prediction and experimental observation. The optimal ME formulation was used for next steps, while the ethanolic solution based gel (ESBG) and 20% AZA commercial cream (Skinoren®, Bayer Co., Ltd, Taiwan) were used as control formulations.

Fig. 3 Overlay plot for the effect of different variables on the three responses: skin retention after 24 h (μg) (Y₁), amount of AZA in the collection medium after 24 h (μg) (Y₂) and particle size (nm) (Y₃).

Table 3 Predicted and experimental values for the optimized microemulsion

Response	Predicted value	Experimental values	Error^a%
a Error% was calculated using the formula [(experimental value − predicted value)/experimental value] × 100.
Y1	571.64	593.57	3.69
Y2	573.97	584.69	1.83
Y3	3.78	3.83	1.32

---

3.3. Evaluation of prepared formulations

The appearance of the optimal ME by visual observation was clear and transparent (Fig. 4). The particle size, PDI value and zeta potential of the optimal ME were 3.83 nm, 0.216 and −4.99, respectively, demonstrating its excellent homogeneity and stability. The pH values were determined as 3.44, 3.13 and 4.15 for the cream, ESBG and ME, respectively (Table 4). Among the three formulations, the pH value of ME was consistent with that of human skin surface (typically slightly above pH = 5), resulting in less skin irritation potential, to a certain degree.⁹

Fig. 4 Typical appearance and particle size distribution of the optimal ME formulation.

Table 4 pH value and the permeation parameters of commercial cream, ESBG and the optimized ME formulation

Formulation	pH value	Flux (μg cm^−2 h^−1)	Amount in collection medium at 24 h (μg)	ER^a	Total skin retention (%)
a ER: enhancement ratio for drug permeation = flux in ESBG or ME/flux in cream. *P < 0.05, when compared to control. **P < 0.01, when compared to control.
Cream	3.44 ± 0.052	4.29 ± 0.23	129.48 ± 14.56	—	3.41 ± 0.31
ESBG	3.13 ± 0.058	10.73 ± 2.68*	401.875 ± 26.99*	3.10	4.74 ± 0.35
ME	4.15 ± 0.071	15.64 ± 2.49**	584.69 ± 40.87**	4.52	11.87 ± 0.76**

---

3.4. Stability assay

The optimized ME formulation was stable when stored at 25 ± 2 °C/60 ± 5% RH for three months, during which there was no obvious change in visual appearance (Table 5). In addition, no changes in droplet size and PDI were observed during the 3 months. The concentration of AZA in the optimal ME was above 98.69% ± 3.96 during the 3 months, which demonstrated that there was no degradation.

Table 5 Long-term storage stability of AZA ME. Data represent mean ± SD for three batches

Parameters	M^a 0	M^a 1	M^a 2	M^a 3
a M stands for month.
Appearance	Transparent	Transparent	Transparent	Transparent
Droplet size (nm)	3.83 ± 0.071	3.92 ± 0.095	3.98 ± 0.083	4.15 ± 0.075
PDI	0.216 ± 0.002	0.198 ± 0.003	0.295 ± 0.004	0.167 ± 0.003

---

3.5. In vitro skin permeation studies

The penetration behaviors of AZA from the optimized ME, ESBG and commercial cream were evaluated for comparison purposes. The cumulative amount of AZA permeated through porcine skin after 24 h was calculated and plotted against time. As shown in Fig. 5, the cumulative amount of AZA in the receptor chambers steadily increased over time. The optimal ME and ESBG presented comparable penetration behavior through the skin, which was significantly higher than that of the marketed AZA cream (P < 0.05 for ESBG; P < 0.01 for ME). Moreover, the ME formulation and ESBG provided a higher permeation rate than the cream, which represented a possible rapid therapeutic effect. The results demonstrated that the tested ME formulation and ESBG had a potent enhancement effect for the topical administration of AZA.

Fig. 5 Permeated amount of AZA in the receptor medium at various time points: comparison of commercial cream, ESBG and the optimal ME. Results are expressed as mean ± SD, n = 6. *p < 0.05, **p < 0.01.

The drug accumulation in different skin layers (SC and viable skin layers) after 24 h application of the three formulations was determined (Fig. 6). The total skin retention was defined as the sum of the amounts in the SC and viable skin layers (epidermis and dermis, ED). The three formulations could be arranged in descending order in relation to the percentage of total skin retention after 24 h as follows: ME (11.87%) > ESBG (4.74%) > cream (3.41%) (Table 4). As shown in Fig. 6, there was no significant difference between ESBG and AZA cream after 24 h application (P > 0.5 for both SC data and ED data). However, the drug content in the skin layers (both in SC and viable skin layers) treated with the optimized ME was significantly higher compared to the cream suspension (P < 0.01) and ESBG (P < 0.05), which was inconsistent with the permeation tendency through the skin.

Fig. 6 Percentage amount of azelaic acid distributed in stratum corneum (SC), epidermis and dermis (ED) after 24 h exposure to commercial cream, ESBG and the optimal ME. Results are expressed as mean ± SD, n = 6. *p < 0.05, **p < 0.01.

As AZA was water insoluble, it could not completely dissolve in the cream and was mainly suspended in this dosage form.⁵ However, in the optimal ME and ESBG, AZA mainly existed in dissolved form due to their co-solvent and physicochemical properties. In general, only the dissolved fraction of an active agent in a vehicle can enter the skin.³² Therefore, both the optimal ME and ESBG resulted in significantly higher skin permeability than the commercial cream. For AZA retention in the skin layers, however, both ESBG and the cream resulted in significantly lower amounts than the optimal ME, which can be ascribed to the microstructure of the ME.

3.6. Attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) study

An ATR-FTIR study was conducted to study the skin–vehicle interaction and reveal the mechanism of enhanced cutaneous penetration based on various vehicles. In the IR spectra of skin treated with different formulations and that which was untreated (control), the changes in peak position and intensity (peak height) of bands were compared, including CH₂ stretching (around 2924 cm⁻¹, representing the asymmetric stretching CH₂ vibrations) and amide 1 stretching (C=O stretching, around 1653 cm⁻¹, sensitive to H-bond changes in the SC).^25,33

The change in peak intensity of the bands was considered to be important because it provided information on the amount of lipid present in the SC. As shown in Fig. 7, after the treatment with ME, the peak height of CH₂ stretching (around 2924 cm⁻¹, 0.82%) was significantly increased compared to the control (0.77%), implying lipid extraction in the SC and enhancement in drug cutaneous permeation;²⁵ while the SC treated with ESBG and cream displayed decreased peak intensity (0.73% for ESBG and 0.42% for cream), suggesting lipid strengthening in the SC and a subsequent significant retardation effect on percutaneous transport. Analysis of the amide 1 model pointed to a shift to higher wavenumber when SC was treated with ME (from 1653.40 to 1657.32 cm⁻¹) and ESBG (from 1653.40 to 1656.32 cm⁻¹) relative to untreated SC. The shift indicated a weakening of the H-bonds between the amide linkages within the SC, which favored substance penetration into the skin.²⁵ The region corresponding to CH₂ asymmetric vibration (around 2924 cm⁻¹) provided information on the conformational order of the SC lipid chains.³⁴ After treatment with the ME formulation, the band shifted to higher values in comparison with the control, suggesting permeation enhancement due to disorder in the lipid arrangement. However, the band of the SC treated with ESBG showed a shift to lower values, supporting the enhancement of a stable arrangement of lipids.

Fig. 7 Representative ATR-FTIR spectra of untreated porcine skin SC (control) and SC treated with the optimal ME, ESBG and commercial cream.

3.7. Skin irritation test

Although most of the ingredients used in the ME preparation were pharmaceutically approved, they might also irritate the skin at higher concentrations.³⁵ As a result, histopathological examination was performed to evaluate any irritant potential of the optimized ME compared to control formulations.³⁶ After 3 days, rats in all the groups showed no apparent edema, erythema or other irritant response. Microscope images of rat skin treated with various formulations are shown in Fig. 8. Compared to normal skin (Fig. 8a), the SC layer of rat skin treated with ME and ESBG became thinner, but without any apparent change in epidermis and dermis (Fig. 8c and d), and the SC, epidermis and dermis layers were normal following cream application (Fig. 8b). In addition, the skin treated with cream, ESBG and ME showed no sign of inflammation cells. The results suggested that the optimized ME might be safe to use for topical AZA delivery.

Fig. 8 Microscope images of rat skin treated with (a) normal saline, (b) commercial cream, (c) ESBG and (d) the optimal ME.

3.8. Pharmacodynamics studies

Cutaneous polymorphonuclear leukocyte inflammation was induced with croton oil to evaluate the therapeutic effect of ME formulation on rosacea, based on reduced ear redness, edema, etc.^27,28 Untreated mice (group f) and those treated with croton oil only (group a) were used as negative and positive controls, respectively. Croton oil could produce intense redness, accompanied by a large number of infiltrated inflammatory cells in the viable skin layers, edema, and even severe skin ulcers (Fig. 9a). The application of blank ME4 could not improve inflammation compared to the positive control ear (Fig. 9a and b). On the contrary, AZA-loaded ME exhibited a significant inhibitory effect on the inflammation response based on the significantly reduced number of inflammatory cells in the whole skin layers (Fig. 9d), which was superior to the ESBG and commercial cream treated ears (Fig. 9c and e). It seemed that there was no significant difference between the AZA-loaded ME treated ear and the negative control ear (Fig. 9f) in both macrophotos and microphotos. In addition, the application of AZA-loaded ME (group d) also significantly inhibited the increase in the ear thickness (ear edema) compared to the other formulations (P < 0.01 for group a, b; P < 0.05 for group c, e) (Fig. 10). The results indicated that the optimized ME formulation significantly improved the therapeutic effect compared to market cream.

Fig. 9 Photos of mice ears (upper panel), and photomicrographs of H&E-stained mouse-ear tissue at a magnification of 10× (middle panel) and magnification of 20× (bottom panel), sensitized with various formulations. Mouse ears were treated with (a) croton oil, (b) croton oil and blank ME, (c) croton oil and drug loaded ESBG, (d) croton oil and drug-loaded ME, (e) commercial cream. Untreated mouse ears were used as a control (f). The number 1 indicates inflammatory cells. The number 2 and number 3 indicate skin ulcers and edema, respectively.

Fig. 10 Ear thickness differences between left (untreated) and right (treated) ears of mice treated with different formulations.

AZA, a bioactive molecule used for many skin disorders, has limited penetration across the stratum corneum due to poor bioavailability, mainly caused by low drug solubility and poor skin permeability. In order to enhance AZA solubility in the vehicles, ionization and the monosodium salt of AZA has been investigated.^5,37 In our study, however, AZA was completely solubilized in the optimal ME without any physical or chemical treatment. In addition, gel,³ liquid crystal³⁸ and nanoscale vehicles (including microemulsion,³⁷ ethosomes and liposomes³⁹) have been developed as alternative topical formulations of AZA. In these studies, the effect of the developed vehicles on the cutaneous permeation of AZA was investigated using only an excised skin model in an in vitro study. However, the therapeutic efficacy of AZA based on topical vehicles has to be proven, since there are many other variables that could affect the efficacy when used in vivo. Thus, in our study, in vivo pharmacodynamics studies were also conducted. The results indicated that the optimized ME formulation containing AZA significantly improved the therapeutic effect on rosacea.

Both metronidazole (MTZ) and AZA are considered to be the first-line treatment of rosacea. In our previous work, we have developed and optimized a ME to enhance targeted localization of MTZ in skin layers and improve the therapeutic efficacy of MTZ.²⁷ However, some comparative research demonstrated that AZA was superior to MTZ in improving inflammatory lesions and erythema of rosacea.² Generally speaking, there were three obvious differences between these two research articles. Firstly, MTZ, with a log P value of −0.18, shows highly hydrophilic properties, resulting in limited permeation into and through the skin as a result of the lipophilic barrier of the SC; however, AZA is a lipophilic drug (log P value is 1.45) and skin penetration is restrained mainly due to its poor solubility. Secondly, considering the significantly different properties between MTZ and AZA, we developed an oil-in-water (O/W) ME for MTZ and water-in-oil (W/O) ME for AZA. The components of the W/O ME in this study were also different from that of the O/W ME in our previous study. Last but not least, in addition to the similar optimization and pharmacodynamics studies in both of our studies, in the current study ATR-FTIR was also carried out to investigate the molecular vibrations of the SC components and reveal the mechanism of enhanced cutaneous penetration based on the ME vehicle.

In addition, porcine skin was chosen as an in vitro penetration model in our study, not only due to its physiological, biochemical and histological similarities to human skin, but also because of it having less variability than other skin models.^40,41 In contrast, the skin model from mice exhibits an extremely high density of hair follicles which might affect percutaneous absorption of molecules. Thus, hairy rodent skin is usually used in in vivo studies rather than other in vitro studies.⁴⁰ Nevertheless, in vivo studies are still performed on this species. We realize the potential limitations caused by the different animal models used for in vitro (porcine skin) and in vivo (mice) studies. However, the optimal therapeutic effect of the optimal ME on rosacea might indicate improved AZA retention in mouse skin, demonstrating that these two models may have good correlations for the permeation of AZA, to a certain degree.

4. Conclusion

In the current study, the application of ME systems for the topical delivery of AZA was investigated. A D-optimal mixture experimental design was applied to rapidly obtain the optimal AZA-loaded ME formulation, realizing maximum skin accumulation, appropriate penetration into the receptor medium and optimal globule size. The optimal ME was composed of 50.3% Smix (a mixture of Span 20/ethanol, 1 : 9, w/w), 13.5% water and 36.2% Capryol 90. Contrary to ESBG and commercial cream, the optimized ME significantly enhanced AZA retention in the skin and penetration through the skin in in vitro permeation studies. The ATR-FTIR study indicated that the improved AZA release from the optimal ME was mainly due to the disturbed SC barrier function via lipid extraction, weakening H-bonding between the amide linkages and disordering the lipid arrangement of the SC. Additionally, the results of the skin irritation test and pharmacodynamics study inferred that the AZA-loaded optimized ME formulation was safe and more effective in the treatment of croton oil-induced rosacea than commercial cream and ESBG. Taken together, the optimal W/O ME might be a promising topical vehicle of AZA for improved anti-rosacea therapeutic effect.

Declaration of conflict of interest

The authors report no conflicts of interest.

Acknowledgements

We acknowledge the National Basic Research Project (2014CB932200) of the MOST for financial support.

References

1. G. H. Crawford, M. T. Pelle and W. D. James, J. Am. Acad. Dermatol., 2004, 51, 327.
2. Y. Tuzun, R. Wolf, Z. Kutlubay, O. Karakus and B. Engin, Clin. Dermatol., 2014, 32, 35.
3. K. Yamasaki and R. Gallo, J. Dermatol. Sci., 2009, 55, 77.
4. N. Li, Q. Su, F. P. Tan and J. Zhang, Int. J. Pharm., 2010, 387, 167.
5. N. Li, X. H. Wu, W. B. Jia, M. C. Zhang, F. P. Tan and J. Zhang, Drug Dev. Ind. Pharm., 2012, 38, 985–994.
6. S. A. Fouad, E. B. Basalious, M. A. El-Nabarawi and S. A. Tayel, Int. J. Pharm., 2013, 453, 569.
7. Y. Yang, R. M. Pearson, O. Lee, C. W. Lee, R. T. Chatterton, S. A. Khan and S. Hong, Adv. Funct. Mater., 2014, 27, 2442.
8. N. Li, W. B. Jia, Y. Zhang, F. P. Tan and J. Zhang, Int. J. Pharm., 2011, 415, 169.
9. G. Cevc and U. Vierl, J. Controlled Release, 2010, 141, 277.
10. T. Bhuvaneswari, M. Thiyagarajan, N. Geetha and P. Venkatachalam, Acta Trop., 2014, 135, 55.
11. W. W. Gao, D. Vecchio, J. M. Li, J. Y. Zhu, Q. Z. Zhang, V. Fu, J. Y. Li, S. Thamphiwatana, D. N. Lu and L. F. Zhang, ACS Nano, 2014, 8, 2900.
12. G. Gainza, M. Pastor, J. J. Aguirre, S. Villullas, J. L. Pedraz, R. M. Hernandez and M. Igartua, J. Controlled Release, 2014, 85, 51.
13. S. Sahoo, N. R. Pani and S. K. Sahoo, Colloids Surf., B, 2014, 120, 193.
14. H. X. Ma, M. Yu, M. Z. Lei, F. P. Tan and N. Li, AAPS PharmSciTech DOI:10.1208/s12249-014-0278-5.
15. D. Pepe, M. Mccall, H. Zheng and L. B. Lopes, J. Pharm. Sci. DOI:10.1002/jps.23482.
16. F. T. M. C. Vicentini, T. R. M. Simi, J. O. Del Ciampo, N. O. Wolga, D. L. Pitol, M. M. Iyomasa, M. V. L. B. Bentley and M. J. V. Fonseca, Eur. J. Pharm. Biopharm., 2008, 69, 948.
17. L. M. P. D. Araujo, J. A. Thomazine and R. F. V. Lopez, Eur. J. Pharm. Biopharm., 2010, 75, 48.
18. M. Kreilgaard, Adv. Drug Delivery Rev., 2002, 54, S77.
19. M. J. Lawrence and G. D. Rees, Adv. Drug Delivery Rev., 2012, 64, 175.
20. M. Fanun, Curr. Opin. Colloid Interface Sci., 2012, 17, 306.
21. Y. H. Tsai, F. K. Lee, Y. B. Huang, C. T. Huang and P. C. Wu, Int. J. Pharm., 2010, 388, 257.
22. F. T. M. Vicentini, T. R. M. Simi, J. O. Del Ciampo, N. O. Wolga, D. L. Pitol, M. M. Iyomasa, M. V. L. B. Bentley and M. J. V. Fonseca, Eur. J. Pharm. Biopharm., 2008, 69, 948.
23. A. S. B. Goebel, U. Knie, C. Abels, J. Wohlrab and R. H. H. Neubert, Eur. J. Pharm. Biopharm., 2010, 75, 162.
24. Y. B. Huang, Y. H. Tsai, W. C. Yang, J. S. Chang, P. C. Wu and K. Takayama, Eur. J. Pharm. Biopharm., 2004, 58, 607.
25. D. Kaushik and B. Michniak-Kohn, AAPS PharmSciTech, 2010, 11, 1068.
26. N. Aggarwal, S. Goindi and R. Khurana, Colloids Surf., B, 2013, 105, 158.
27. M. Yu, H. X. Ma, M. Z. Lei, N. Li and F. P. Tan, Eur. J. Pharm. Biopharm., 2014, 88, 92.
28. J. X. Zhang, X. Y. Xu, N. V. Rao, B. Argyle, L. McCord, W. J. Rusho, T. P. Kennedy, G. D. Prestwich and G. Krueger, PLoS One, 2011, 6, e16658.
29. H. K. Patel, B. S. Barot, P. K. Shelat and A. Shukla, Colloids Surf., B, 2013, 102, 86.
30. A. C. Sintov and I. Greenberg, Int. J. Pharm., 2014, 471, 516.
31. A. C. Sintov and L. Shapiro, J. Controlled Release, 2004, 95, 173.
32. A. H. Elshafeey, A. O. Kamel and M. M. Fathallah, Pharm. Res., 2009, 26, 2446.
33. G. Rodriguez, L. Barbosa-Barros, L. Rubio, M. Cocera, A. Diez, J. Estelrich, R. Pons, J. Caelles, A. D. L. Maza and O. Lopez, Langmuir, 2009, 25, 10595.
34. S. M. Ge, Y. Y. Lin, H. Y. Lu, Q. Li, J. He, B. Chen, C. B. Wu and Y. H. Xu, Int. J. Pharm., 2014, 465, 120.
35. F. F. Sahle, J. Wohlrab and R. H. H. Neubert, Eur. J. Pharm. Biopharm., 2014, 86, 244.
36. N. Ü. Okur, Ş. Apaydın, N. Ü. K. Yavaşoğlu, A. Yavaşoğlu and H. Y. Karasulu, Int. J. Pharm., 2011, 46, 136.
37. E. Peira, M. E. Carlotti, R. Cavalli and M. Trotta, J. Drug Delivery Sci. Technol., 2006, 16, 375.
38. F. Aytekin, R. N. Gursoy, S. Ide, E. H. Soylu and S. Hekimoglu, Drug Dev. Ind. Pharm., 2013, 39, 228.
39. E. Esposito, E. Menegatti and R. Cortesi, Int. J. Cosmet. Sci., 2004, 26, 270.
40. B. Godin and E. Touitou, Adv. Drug Delivery Rev., 2007, 59, 1152.
41. F. C. Rossetti, L. B. Lopes, A. R. H. Carollo, J. A. Thomazini, A. C. Tedesco and M. V. L. B. Bentley, J. Controlled Release, 2011, 55, 400.

---