In Silico Estimation of Skin Concentration of Dermally Metabolized Chemicals

Background: A great deal of in silico estimation methods were proposed for skin concentration and permeation of drugs by many researchers including us. The aim of the present study was to expand our in silico estimation method of skin concentration to dermally metabolized chemicals. Materials and Methods: A three-layered diffusion model consisting of stratum corneum, viable epidermis and dermis was constructed based on Fick’s second law of diffusion incorporated with Michaelis–Menten equation and plasma clearance in the viable epidermis and dermis, respectively. Ethyl nicotinate was used as a model chemical, and the in vivo skin concentration of the ester and its metabolite, nicotinic acid were measured after topical application to hairless rats. Permeation parameters were determined from the in vitro permeation data through full-thickness skin and stripped skin after application of the ester or acid with and without esterase inhibitor treatment. Metabolic parameters were obtained from the metabolic profi le of the ester using skin homogenate. Results and Conclusion: The skin concentrations calculated from our improved model using the permeation and metabolic parameters obtained beforehand were similar to the observed values. Infl uence of cutaneous enzyme distribution and plasma clearance on the skin concentrations were also estimated using appropriately modifi ed models, resulting in higher infl uence on the acid than the ester. This estimation method will become an effective tool to assess the effi cacy and safety of dermally metabolized chemicals. Research Article In Silico Estimation of Skin Concentration of Dermally Metabolized Chemicals Tomomi Hatanaka1,2,#, Saki Yamamoto1, Mayuko Kamei1, Wesam R Kadhum1, Hiroaki Todo1 and Kenji Sugibayashi1*,# 1Faculty of Pharmaceutical Sciences, Josai University, Sakado, 350-0295, Japan 2Tokai University School of Medicine, Isehara, 2591193, Japan #These authors contributed equally to this work. Dates: Received: 08 December, 2016; Accepted: 20 January, 2017; Published: 23 January, 2017 *Corresponding author: Kenji Sugibayashi, Professor, Faculty of Pharmaceutical Sciences, Josai University, Keyakidai 1-1, Sakado, Saitama 350-0295, Japan, Tel/Fax: +81-492-71-7943; E-mail:


Introduction
Skin has been the administration site for therapeutic agents for skin diseases and disorders since ancient times [1].
Because skin is the direct interface with the environment, it also undergoes daily exposure to external substances including toxicants [2]. The pharmacological and toxicological activities of chemical compounds depend on their concentrations at the site of action, so many methods for estimating skin concentration have been developed by utilizing a variety of techniques, including tape-stripping [3], microdialysis [4], Raman spectrophotometry [5] and in silico modeling [6,7].
In silico modeling is a promising method for evaluating effi cacy and safety after dermal exposure to topically active compounds. Once the modeling of the permeation process is completed, the skin concentration of chemicals can be estimated noninvasively, rapidly, and at a low cost. Mammalian skin is composed of two primary layers, the epidermis and subjacent dermis; the former consists of the stratum corneum, stratum granulosum, stratum spinosum and stratum basale [8]. The stratum corneum, outermost epidermal layer, consists of 10-25 layers of dead, elongated, and fully keratinized corneocytes and the intercellular lipid lamellae linked to the cornifi ed envelope of corneocytes, so that this layer is frequently described by "brick and mortar" model [9]. Because the lipid matrix in the stratum corneum greatly contributes to the skin's barrier function [10] and only a small area of skin surface is occupied by appendages such as hair follicles, sebaceous glands and sweat glands [9], the two-layered membrane consisting of the stratum corneum and its lower layers has been often assumed in in silico models. We previously demonstrated that the mean skin concentration and concentration at each depth in the skin layers can be estimated based on the two-layered diffusion model for drugs and chemicals with a wide range of lipophilicity [11,12].
Although in silico modeling provides suffi ciently satisfactory predictability of skin concentration, this is not the case for compounds that are easily metabolized in the skin. These Citation: Hatanaka  compounds applied via the stratum corneum diffuse through the viable epidermis, which is the most metabolically active layer in the skin and has many enzymes similar to the liver [13]. Biotransformation in the viable epidermis may change the biological activity of compounds after passage through this layer. In fact, most steroids are applied to the skin as antedrugs to avoid systematic adverse effects; they are metabolically inactivated in the skin as well as in the blood circulation [14].
On the other hand, there have been numerous attempts at transdermal drug delivery via the prodrug derivation [15]. After having overcome the fi rst critical step in skin permeation, i.e. permeation via the stratum corneum, the prodrugs are metabolized to drugs prior to exhibiting their biological effects.
Understanding of the skin concentrations of both the applied chemical compound and its metabolite is required to correctly evaluate the effi cacy and safety of the drugs.
Although our previous estimation for skin concentration used in vitro data after permeation experiments using the excised skin, the fi nal goal is to estimate the in vivo skin concentration. The papillary layer of dermis directly under the epidermis contains blood microcirculation consisting of permeable capillaries and thus the plasma clearance plays an important role in the determination of skin concentration after dermal application of compounds [16]. It is expected that the infl uence of plasma clearance makes the estimation of skin concentration of dermally metabolized chemicals more complicated: skin concentrations of both applied chemical and its metabolite will be dramatically changed by the combination of distribution of metabolic enzyme in skin and location of cutaneous microvasculature.
The aim of the present study was to expand our in silico estimation method of skin concentration to chemicals that are easily metabolized in the skin. A three-layered diffusion model consisting of stratum corneum, viable epidermis and dermis was constructed based on the Fick's second law of diffusion, with Michaelis-Menten equation and plasma clearance in the viable epidermis and dermis, respectively. Ethyl nicotinate (EN) was used as a model compound, because we have already confi rmed that EN is metabolized to nicotinic acid (NA) in the human and rat skin [17,18]. The skin permeation experiments were carried out using excised hairless rat skin.
The permeation parameters, such as the diffusion coeffi cient and partition coeffi cient, were determined from the amount permeated through full-thickness skin and stratum corneum- The in vivo skin concentrations were separately measured 6 h after dermal application of the ester compound to hairless rat, and the observed values were compared with the calculated values. Infl uence of cutaneous enzyme distribution and plasma clearance on the skin concentration of the dermally metabolized chemical were also simulated using appropriately modifi ed three-layered diffusion models.

Hydrolysis experiment in skin homogenate
Rat skin was freshly excised by the same method as for the skin permeation experiment and homogenized directly with PBS at 4°C for 2 min using a homogenizer (Polytron PT 1200 E, Kinematica AG, Littau-Lucerne, Switzerland) to make 10% skin homogenate. The homogenate was centrifuged at 9,000 × g and 4°C for 20 min, and the supernatant, except for the fl oating lipid layers, was stored at -80°C until the following experiments. The supernatant was thawed on ice and preincubated at 37°C for 10 min together with PBS containing EN at various concentrations. Two tubes of solutions were mixed, resulting in a fi nal concentration of EN of 0.098-4.9 mM and in the skin homogenate of 0.2%. The reaction solution was kept at 37°C and sampled at 0, 5, and 10 min. The sample was mixed with the same volume of acetonitrile containing 16% trichloroacetic acid and 1 mM methyl 4-hydroxy benzoate, the mixture was centrifuged (21,500 rpm, 4°C and 5 min), and the concentration of NA in the supernatant was determined.

Protein content in the reaction solution was determined by
Lowry's method [20].

High-performance liquid chromatography analysis
The HPLC system (Shimadzu; Kyoto, Japan) consisted

Data analysis
where C and D are the concentration and diffusion coeffi cients, and the subscripts p and sc refer to the parent compound and stratum corneum, respectively. In the viable epidermis where parent compound is transformed to metabolized compound, the rate of change in the concentrations of chemicals with time at a position are described as follows: Citation: Hatanaka  where Q and f are the plasma volume fl ow and unbound fraction, and the subscript de denotes the dermis.
Initial and boundary conditions of the skin concentrations of two compounds are defi ned by: where L and K are the thickness and partition coeffi cients from the donor vehicle, and the subscript do refers to donor solution, respectively. The concentration of parent and where ,  (16) and (17) for equations (1)-(5) and arranging, we fi nd: Skin permeation parameters, which were required to predict the skin concentrations of the parent compound and its metabolite, were obtained by fi tting permeation data under various conditions to the diffusion models [22]. In order to obtain the , m ved D value, the permeation data for NA applied to stripped skin were fi tted to the one-layered diffusion model.   (21) and (22).  Table   1. NA permeation through the full-thickness skin was analyzed by the two-layered diffusion model using the previously    Table 1 were determined from the permeation data after application EN to the skin treated with DFP, a serine protease inhibitor (Figures 2c and d). The permeation profi les Citation: Hatanaka  of EN through the stripped skin and full-thickness skin could also be described by the one-layered and two-layered diffusion models, respectively. Almost no fl ux of NA was observed after DFP treatment. Figure 3 shows the relation between the hydrolysis rate of EN to NA and initial EN concentration in 0.2% skin homogenate.

Determination of skin metabolic parameters
The hydrolysis of EN followed Michaelis-Menten kinetics, and the kinetic parameters ( max V and m K ) estimated by data-fi tting are listed in Table 1.

Estimation of skin concentrations
In vivo skin concentrations of both NA and EN after application of EN to skin surface without DFP treatment were estimated from equations (18) (Figure 5b). In contrast, the concentration-distance profi les of NA showed convex curves in the viable epidermis followed by concave curves in the dermis as shown in Figure   5d. The concentration of NA was raised with time until 1 h and remained unchanged until 6 h.
In order to more precisely assess the infl uence of cutaneous enzyme distribution and plasma clearance on  Table 1. Each color is assigned to a certain time after EN application.
the concentration of EN and NA, the concentration-distance profi les were simulated based on additional three permeation models. Because the profi les of EN in the stratum corneum were not so different among models, the profi les of both EN and NA in the viable epidermis and dermis are shown in Figure   6. Homogeneous enzyme distribution in the viable epidermis and dermis caused a monolithic decrease of EN concentration with an increase of distance from skin surface (Figure 6a,b), whereas enzyme distribution only in the viable epidermis produced a biphasic decline (Figures 5b,6c). Existence of plasma clearance reduced EN concentrations throughout the profi les (Figures 5b,6b). NA concentration was raised to be about 10 times by heterogeneous enzyme distribution, although the shapes of NA concentration-distance profi les were not changed (Figures 5d,6f). The concentration was reduced by plasma clearance (Figures 5d,6e), and the effect was higher than EN concentration.

Discussion
We previously proposed a two-layered diffusion model to estimate the mean skin concentration and concentration at each position in the skin layers [11,12]. However, the use of this model may result in an inaccurate estimation of skin concentration for chemicals extensively metabolized in the skin. Because a large part of steroids contained in the topical formulations are ante-drugs [14] and because transdermal drug delivery via prodrug derivation has been much attempted [15], further improvement of our method is desirable for expanding to enable the estimation of skin concentrations after dermal exposure of chemicals extensively metabolized in the skin. The skin, especially epidermis, is metabolically active [13], and any permeant is subjected to the metabolic properties of the living layer. Moreover, both applied chemical compound and its metabolite are cleared by blood microcirculation of permeable capillaries in the dermis after passing through the viable epidermis [16]. and near hair follicles than the dermis by a microphotographic study using fl uorescein-5-isothiocyanate diacetate [25].
In contrast, an immunohistological study showed that nonspecifi c -naphthylacetate-esterase, by which nicotinates are metabolized to nicotinic acid, was also located in the dermis [26]. Carboxylesterases are present in cytosol as well as microsomal fractions of tissues [27], so that such dissolved enzymes may be washed out during histological preparation.
The concentration of NA was more affected by the cutaneous enzyme distribution than that of EN. It is necessary to construct the skin permeation model accompanying an accurate enzyme location to estimate the skin concentration of metabolites.
The high metabolizing potential in the viable epidermis was attributed by the fact that barrier lipids in the stratum corneum are synthesized in the living layer together with keratinocyte differentiation [28]. The barrier lipids, their precursors and metabolizing enzymes are packaged into the epidermal lamellar body in the stratum spinosum and stratum granulosum, and are secreted at the interface between the stratum granulosum and stratum corneum by fusing the lamellar body with the apical plasma membrane of the outermost stratum granulosum cell layer [29]. The secreted enzymes include many lysosomal hydrolases [30], and the esterases present in the cytosol fraction of the skin were mainly responsible for the hydrolysis of propranolol prodrug [31]. Thus, some of the lysosomal enzymes must contribute to the metabolism of EN to NA in the skin. Carboxylesterases have a wide substrate specifi city and play an important role in the hydrolytic biotransformation of many chemicals [32]. About 70% of the hydrolysis activity of prednisolone 21-acetate in skin homogenate from hairless mice were not deactivated by 3,4-dichloroisocoumarin, which completely inhibited the activity of carboxylesterase [33].
Hydrolysis of EN to NA must be mediated by many enzymes in the skin.
Two assumptions were given when plasma clearance was incorporated in the skin permeation model. One assumption is that the concentrations of both parent and metabolized chemicals in the plasma are negligible low compared with those in the dermis. In fact, indomethacin concentrations in the underlying tissues was 1000-10000 fold higher than those in the plasma following topical application [24]. and EN concentration-distance profi les in stratum coneum were almost the same among the skin permeation models.
Recently, in silico prediction of transdermal and systemic drug disposition by a physiologically based pharmacokinetic model, in which the stratum corneum was characterized by a brickand-mortar structure, was proposed [7,35]. The method is superior to overcome the numerical complexity of model by utilizing the quantitative structure-property relationships.
Although the fi nal goal of our skin permeation model is also a physiological model incorporating skin metabolism, a high predictability for the concentrations of extensively metabolized chemicals in skin was obtained at the present stage.
The in vivo skin concentrations of both EN and NA after EN application could be estimated by our improved model, a three-layered diffusion model based on the Fick's second law of diffusion, and incorporated Michaelis-Menten equation and plasma clearance term. Although skin permeation and metabolic parameters were obtained from in vitro skin permeation experiments in the present study, these parameters can be also predicted using the physicochemical properties of chemicals [36][37][38]. Only metabolic parameters that are determined from the distinct experiments are needed. There are many skin models, e.g. subcellular fractions of skin, tissue-cultured skin, monolayer cultures of keratinocytes or fi broblasts, and reconstructed human skin equivalents, other than skin homogenates and excised skin for assessing skin metabolism [39,40]. It is desirable to develop a simple model which mimicks accurate in vivo skin metabolism.

Conclusion
We proposed a three-layered diffusion model consisting of stratum corneum, viable epidermis and dermis, which was constructed based on Fick's second law of diffusion incorporated with Michaelis-Menten equation and plasma clearance in the viable epidermis and dermis, respectively. The skin concentrations calculated from our improved model using the permeation and metabolic parameters obtained beforehand were similar to the observed values. Infl uence of cutaneous enzyme distribution and plasma clearance on the skin concentrations were also estimated using appropriately modifi ed models, resulting in higher infl uence on the acid than the ester. Our in silico estimation method of skin concentration of chemical and its metabolite gives important information about their effi cacy and safety, and thus becomes a powerful tool to promote the development of pharmaceutical and cosmeceutical products.