Adapted from (Flaten et al., 2015).

In order to overcome the disadvantages of ex vivo human and animal models, mainly related with the availability and the ethical problems, numerous efforts have been done to develop alternative skin equivalents based on in vitro approaches. In vitro skin alternative models may consist either on mimetic membranes (non-lipid and lipid systems) or cell cultures (Abd et al., 2016, Flaten et al., 2015, Berben et al., 2018, Sinkó et al., 2014, Faller, 2008). The existent models and their applications will be reviewed in the following sections.

The first studies regarding the development of in vitro skin equivalents developed in mid 70s included the use of normal human keratinocytes (NHKs) as a model for skin irritancy (Rheinwald, 1989). The cell culture of NHKs begins with a small piece of human skin (about 0.5-1 cm²) obtained from surgery after separation and specific treatment, that grows easily in culture medium. It allows the use of a large number of cells, leading to the opportunity for wide ranging toxicity screening tests with many substances. This model shows good ability for testing hydrophilic compounds however presenting less capability for the evaluation of poorly water soluble compounds and complex formulations (reviewed in (Ponec, 1992)).

The model was improved with its application in membranes, which support the NHKs during growing and forming the called reconstructed human epidermis. More complex models mimicking the full thickness skin consisting of the fibroblast populated collagen matrices (dermis equivalent) and an epidermal cover composed by NHKs (Van Gele et al., 2011).

Nowadays there are different commercially available models of the human epidermis or full thickness skin. These models will be reviewed with more detail in section 3.2.3.

Additionally, other acellular in vitro models have been described: a) non-lipid-based models like the silicone membranes or poly(dimethylsiloxane) (PDMS) models (Oliveira et al., 2011) and b) lipid-based models like Parallel Artificial Membrane Permeability Assay (PAMPA) (Sinko et al., 2012, Sinkó et al., 2014) and Phospholipid Vesicle-based Permeation Assay (PVPA) systems, as simpler alternative approaches (Engesland et al., 2013).

The first studies regarding the use of non-lipid-based skin models date from 1970 and report the use of silicone membranes to study the release of salicylic acid (Nakano and Patel, 1970). Later, some studies reported the use of different microporous membranes based on pure cellulose acetate, cellulose and polysulfone to investigate the permeation of hydrocortisone from two commercial creams (Shah et al., 1989). Other synthetic membranes considering polysulfone, cellulose mixed esters, polytetrafluoroethylene and polypropylene in their composition were described and used to study the nitroglycerin release from commercial ointments (Wu et al., 1992). More recently, some studies describe the interaction of many drugs and vehicles with skin models of PDMS or with silicone membranes skin mimetic models (Dias et al., 2007, Watkinson et al., 2009a, Watkinson et al., 2009b, Watkinson et al., 2011, Oliveira et al., 2011, Oliveira et al., 2010, Oliveira et al., 2012).

These skin equivalents are simple models with great applicability to test a basic diffusion mechanisms however they present some disadvantages namely the lack of similarity with the human skin and they are not much useful in the study of the permeability of hydrophilic compounds despite the good results obtained for lipophilic drugs (Abd et al., 2016, Miki et al., 2015).

Another study reports a new and improved model based on a membrane impregnated with a polymer of PDMS and PEG 6000 and some preliminary results elucidating the permeation of drugs in the aqueous solutions were described by Miki and co-workers (Miki et al., 2015).

The lipid-based skin equivalents appeared as valuable alternatives to non-lipid-based models, improving its complexity and the ability to mimic the human skin, namely the SC layer.

A first report dates from 1989 and describes the preparation of model membranes, using SC lipids, to study the permeability of drugs in the skin. The authors consider ceramides, cholesterol, FFA and cholesteryl sulphate in its composition (Abraham and Downing, 1989). The selected lipid composition agrees to those found in the human SClayer and thus it constitutes an advantage for this model since it better represents the layer in terms of its lipid composition. However, the most consistent studies started in 1998 when Kansy and co-workers reported the first PAMPA model as a tool for rapid determination of passive membrane permeability of drugs (Kansy et al., 1998). This system included a mimetic membrane from hydrophobic filter coated with phosphatidylcholine dissolved in n -dodecane as a membrane barrier, which differentiates the donor and acceptor parts. Even though this model was oriented for testing the transcellular intestinal permeability, it was the precursor of the following models, especially those developed to mimic the human skin.

The first PAMPA model for skin penetration estimation was published by Ottaviani and co-workers in 2006 (Ottaviani et al., 2006) however this proposal is, in fact, a non-lipid-based model which incorporates silicone oil and isopropyl myristate as membrane components. This model was used to study the permeation of a large variety of drugs and vehicles (Dobričić et al., 2014, Markovic et al., 2012, Karadzovska and Riviere, 2013, Karadzovska et al., 2013).

In the same year, Loftsson and co-workers described a novel skin substitute membrane based on a hydrated semi-permeable cellophane membrane and a lipophilic membrane of pure n -octanol in a nitrocellulose matrix. This model was used to study the permeation of different cyclodextrin formulations and the results have shown that the drug permeation patterns of the different formulations were similar to those previously observed for biological membranes (Loftsson et al., 2006).

In 2012, Sinko et al . designed the skin-PAMPA (Sinko et al., 2012), which consist of ceramides analogues called the synthetic ceramides or certramides, as substitution for the native ceramides found in SC . Certramides are cheaper alternatives to natural ceramides with the potential to prolong the storage time. This skin-PAMPA model was used in the skin permeability studies and exhibited poor correlation with skin epidermis, however it presented a good correlation with full thickness skin (Sinko et al., 2012). Even though its composition intents to simulate mainly the SC , in some cases, skin-PAMPA can even be a valuable alternative to replace models that use real human skin. Moreover, it constitutes an easy, quick and cost-effective research tool to study the permeation of pharmaceutical and cosmetic ingredients.

More recently, a modified version of the skin-PAMPA was reported (Tsinman and Sinkó, 2013) and used to evaluate the skin permeation of different some ibuprofen-containing formulations. The developed system was able to distinguish between the different types of formulations the results correlate well with those found in permeation studies using human epidermis as a mimetic system.

Several other reports can be found in the literature reporting the use of PAMPA models to investigate drugs’ permeability in the skin (Markovic et al., 2012, Dobričić et al., 2014, Faller, 2008, Köllmer et al., 2019, Alvarez-Figueroa et al., 2011, Vizserálek et al., 2015, Wu et al., 2019, Luo et al., 2016, Zhang et al., 2019, Balázs et al., 2016). The vast number of reports considering the use of this type of mimetic models to study the interaction and permeation of many different bioactive ingredients highlights the large spectra of applicability of PAMPA approaches and points out the ability of these barriers to predict the permeation of drugs in the human skin, despite of the more simplistic composition of these systems.

The original PVPA (PVPA_o) was introduced as a model for screening the intestinal permeability and is composed of a consistent coat of liposomes deposited on a filter support acting as biological barrier (Flaten et al., 2006b, Flaten et al., 2006a). Later, by changing the lipid composition of the liposomes used to produce the permeation barrier, a new PVPA model was developed aiming to mimic the SCbarrier of the skin (Engesland et al., 2013). In this model, liposomes are located within the pores and on the surface of the membranes (Flaten et al., 2006a). Thereafter, other modified versions have been reported (Berben et al., 2018, Engesland et al., 2013, Engesland et al., 2015, Engesland et al., 2016, Ma et al., 2017, Palac et al., 2014, Zhang et al., 2016).

Mainly, two skin PVPA models, presenting different lipid composition, for estimating skin penetration have been described: a) PVPAc - liposomes made of egg phosphatidylcholine (EPC) (77.1%, w/w) and cholesterol (22.9%, w/w) (Engesland et al., 2013) and b) PVPAs - lipid mixture of EPC (50%, w/w), ceramide (27.5%, w/w), cholesterol (12.5%, w/w), FFA (7.5%, w/w) and cholesteryl sulfate (2.5%, w/w) (Engesland et al., 2013).

The permeability of different compounds was evaluated in these PVPA skin models and the results were compared with reported permeabilities using animal skin models (rat, cattle, dog and pig) and with estimatedin silico values. The PVPA permeability data mainly corresponded with the literature values of the animal skin penetration assays and thein silico values, with the exception of flufenamic acid that showed a relatively lower permeation (Engesland et al., 2013).

Later, the PVPAs model was examined with diverse formulations made of different types of liposomes containing diclofenac sodium salt. The results showed a rising permeation ranking of diclofenac sodium from liposomal formulations correlating with the physicochemical parameters of the liposomal vehicle. The permeability of diclofenac increased in the availability of the penetration enhancers (Palac et al., 2014). The PVPA model was further optimized considering a complex skin PVPA containing all the classes of lipids found in the SC , and the penetration enhancing effect of menthol was investigated for a set of active compounds with different physicochemical properties (Ma et al., 2017).

The previously described PVPA skin models were studied in comparison with a reconstructed human skin model (EpiSkin^®). The permeability results indicate that the PVPA model has the ability to distinguish between the liposomal formulations and drug solutions, as opposed to EpiSkin^®. PVPA models were better than EpiSkin^® concerning their potential to determinate the influence of the formulations on the drug permeability which could be used in drug development at early stage. Moreover, PVPA barriers revealed straightforward, effectiveness, economical and long storable properties (Engesland et al., 2015).

More recently, other two lipid-based models have been reported as skin mimetic systems and they were used for the study of the effect of a set of synthetic surfactants on the skin. The new models proposed contain DPPC and cholesterol in a molar ratio of 7:3 or a mixture of ceramide, stearic acid and cholesterol in a molar ratio of 14:14:10 (Jurek et al., 2019).

Recently, our research group recently developed and characterized a cheaper and simpler alternative SC mimetic model (Shakel et al., 2019) that simulates this human skin layer and can allow the screening of drug candidates. The design of this new model was inspired on PVPA approaches and comprises a lipid composition which closer resembles that found in the human SC layer, namely in the percentage of ceramides considered in its constitution. Thus, the novel human SC PVPA model is made of liposomes composed by ceramide (50% w/w), EPC (25% w/w), cholesterol (12.5% w/w), free fatty acids (10% w/w) and cholesteryl sulphate (2.5% w/w). This model presents some advantages since it can be stored up to 2 weeks at -20 °C, without losing their integrity.

The complexity of the skin mimetic models has been increasing along the time, starting from in vitro non-lipid-based to lipid models and later to models comprising simple or more complex cell cultures. Single or multi-cell type 2D cultures, in which cells are grown as a monolayer disposed on solid flat surfaces, such as polystyrene or glass, are some of the mostly used cell-based approaches due to their relative simplicity and cost-effectiveness (reviewed in (Randall et al., 2018)).

The first studies regarding 2D skin cell mimetics were performed by Rheinwald and Green and consist in the growing of human keratinocytes monolayer culture (obtained from foreskin of newborns) deposited on plastic culture plates (Rheinwald and Green, 1977). As a further advance, the authors included a primary fibroblast cell line (3T3) in the keratinocyte cultures in order to close resemble the skin. Furthermore, fibroblasts are important for the growth of keratinocyte cultures, namely by the secretion of extracellular matrix components, like collagen (Rheinwald and Green, 1975).

The successfully use of 2D skin mimetic models has been reported in many studies for skin irritation and drug development, as referred in (Ponec, 1992, Amelian et al., 2017). Silva and co-workers established anin vitro epidermal monolayer using human keratinocytes (HaCaT cells) differentiated in a high calcium concentration medium, and applied it in the evaluation of drug delivery systems, namely lipid-based nanocarriers (Silva et al., 2017).

Despite the good performance of 2D skin mimetic models, 3D cellular models represent more precisely the human skin due to complexity and stiffness. Bioengineered skin can be obtained by the building of reconstructed skin models which constitutes artificially fabricated skin mimetics. 3D (single or multi-cell type) reconstructed skin models consist in the deposition of different layers of human cells in culture on a polymeric matrix, thus conducing to the production of a more complex and interactive system. In some recent studies, the incorporation of melanocytes (Min et al., 2018, Dai et al., 2018a), adipocytes (Klar et al., 2017) and endothelial cells (Dai et al., 2018b) in 3D skin mimetic models has already been reported.

The main differences between 2D and 3D skin models and the advantages of each class of models are depicted in Table 2.

Table 2 – Main characteristics of 2D and 3D cell-based skin mimetic models.