Non-Animal Models in Dermatological Research

Despite being widely used for basic and preclinical studies in dermatology, available animal models only partly recapitulate human skin features, often leading to disappointing outputs when preclinical results are translated to the clinic. Therefore, the need to develop alternative, non-animal models is widely recognized to more closely recapitulate human skin pathophysiology and to address the pressing ethical demand to reduce the number of animals used for research purposes, following the globally accepted 3Rs principle (Replacement, Reduction, and Refinement). Skin is the outermost organ of the body, and is, as such, easily accessible. Different skin cell types can be propagated in vitro, and skin can be reconstructed for therapeutic transplantation as well as for in vitro modelling of pathophysiological conditions. Bioengineered skin substitutes have been developed and have evolved from elementary to complex systems, more and more closely resembling complete skin architecture and biological responses. In silico analyses take advantage of the huge amount of data already available from human studies to identify and model molecular pathways involved in skin pathophysiology without further animal testing. The present review recapitulates the available non-animal models for dermatological research and sheds light on their prospective technological evolution. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is appropriately cited. # contributed equally


Introduction
The development of alternative models has gained a high priority following the European ban on animal testing for cosmetic ingredients (2009/1223/EU), the REACH guideline for chemicals (2006( /1907/EC) /EC) and the recommendation to follow the 3R principle (Replacement, Reduction, Refinement) for research usage (2010/63/EU).To date, there is an extraordinary variety of nonanimal models based on human cells and tissues representing a suitable alternative to replace animal studies for cosmetic industries.Non-animal models also represent a huge opportunity for pharmaceutical industry since the animal testing ban, while not mandatory at present, is foreseen soon.However, to formulate pharmaceutical industrial applications, non-animal models do require further development and adaptation.Investments toward improvement of alternative models to animal testing would be profitable for several reasons, including: the relevant reduction of R&D related costs; the efficacy issue, since animal models not always appropriately reproduce the human setup; a worldwide visibility related to the cruelty-free and impact-reduction policy distinctive of such strategies.Interestingly, the research world, involved as main actor in non-animal model set-up and development, has in general only partial knowledge of the possibilities of alternative methods and does not always apply them as the first choice for scientific investigation, still preferring the use of animal models.
In 1981 a first attempt was made of in vitro reconstituting human skin for dermatological research.Since then, great efforts have been constantly put into developing models more closely recapitulating in vivo skin complexity, as tools for functional studies and to test therapeutic approaches.In this review, improvements in non-animal models for dermatological research, especially skin equivalent techniques and in silico tools, will be discussed together with an overview of their application to main fields of dermatological interest.Moreover, a prospective view of further alternative method development will be given.

Human skin structure
Human skin is the first line of defense from environmental insults, covers the whole body and consists of three layered tissues differing in cell composition, thickness, strength and flexibility: the epidermis, the dermis and the hypodermis (Murphree, 2017)(Fig.1).

The epidermis
The epidermis is the outermost layer, mainly composed of keratinocytes, organized into four major layers at progressive differentiation stages: basal, spinous, granular and cornified layers (Candi et al., 2005;Bouwstra and Ponec, 2006;Dellambra and Dimri, 2009;Fuchs, 2016).The basal layer contains proliferating and stem keratinocytes, the latter playing a pivotal role in maintaining epidermal homeostasis and promoting epithelial repair.Keratinocytes of the spinous layer form desmosomes that reinforce intercellular contacts and induce the spinous-like morphology.The granular layer is composed by keratinocytes with an 2013).Animal models have the great advantage to manifest the complexity of an organ, and its interactions with other body districts.However, besides the ethical concerns related to the use of animals, animal skin only in part recapitulates human skin features making it often difficult to translate basic and preclinical findings into the clinic.Such differences underline the need of other, more reliable models.

Fig. 2: Features of human and animal model skin
Upper panels: histological comparison of human skin and mouse back and tail skin.Human skin is much thicker as compared to mouse skin, in the epidermal (yellow bars), dermal and subcutaneous (fat layer) compartments.Moreover, human epidermis has projections (arrows) that protrude into the dermis, unlike mouse skin.In the mouse, hair follicles (hf) are found all over the body, whereas in human skin are generally scarce.The presence of a fur has a protective function and mouse skin has less developed differentiated cell layers forming the epidermal barrier.Mouse tail skin, which has a strongly reduced number of hairs has a thicker epidermis with increased cell layers compared to back skin and a more developed external stratum corneum.e: epidermis; d: dermis; hf: hair follicle.Bars: 50 µm.
Lower panels: simplified schematic representation of pig, zebrafish and drosophila skin.Pig skin is very similar to human skin, in terms of cell composition, physiology and thickness, apart from the subcutaneous fat layer which is thicker in the pig (not shown).The dermis has collagenous bundles in the extracellular matrix (orange strings) released by fibroblasts (yellow cells), is vascularized, and sparse immune cells are present: macrophages (blue cells) and lymphocytes (green cells).Ascending arterioles (red) and descending venules (blue) are also shown.Zebrafish skin is formed by epidermis and dermis separated by a basement membrane.Zebrafish midterm embryo has a bi-layered epidermis containing mucus cells (green) and ionocytes (violet), responsible for ionic reabsorption and acidbase regulation, besides keratinocytes, and a dermis with collagenous fibers (blue strings) released by fibroblasts (blue cells).Pigment cells (chromophores) are melanophores, similar to melanocytes (black), xanthophores (yellow), and iridophores (light grey).
Melanophores contribute to the longitudinal dark stripes.In the adult fish, the epidermis is multilayered and covered by scales (not shown).Drosophila skin (upper part) has a single epidermal layer covered by a cuticle.Epidermal cells are attached to a thin basement membrane (basal lamina) overlying muscle cells.The lower part of Drosophila skin panel represents the larval epidermis observed from outside.Larval epidermis is used for studying molecular mechanisms of wound re-epithelialization. e: epidermis; d: dermis; bm: basement membrane; hf: hair follicle; m: muscular layer; c: cuticle.

Mouse and other rodents
Mice represent the most widely used animal model for pathophysiological studies.They are available and easy to handle, their skin shows similarities with the human counterpart, they can be genetically engineered and have a short lifespan as compared to larger mammals.Pathological features recapitulating dermatological diseases have been reproduced in the mouse by gene knock-out or knock-in, transgene over-expression, by using mouse strains with spontaneous mutations, or immunodeficient humanized mice engrafted with human skin (Garcia et al., 2007;Chen and Roop, 2008;Schneider, 2012;Avci et al., 2013).Mice are used in all areas of dermatological interest, from immunological diseases to cancer, skin repair, genetic diseases, and hair disorders (Table 1).However, data interpretation is limited by structural, physiological, and molecular differences (Fig. 2).In fact, in silico studies have identified genetic networks that underline species differences in human skin versus mouse skin (Wu et al., 2017).Such differences explain, at least in part, why results obtained in mouse models cannot often be reproduced in the human context.The most obvious difference is that mice and other rodents, such as rats, guinea pigs and rabbits, have a fur covering most of the body, while in human the ratio between haired and nude skin is inversed.This feature contributes to make mouse skin more permeable than human skin.Therefore, the skin of nude rodents is more similar to human skin, for structure and percutaneous absorption (Jung and Maibach, 2015).Moreover, in mouse skin all three layers, epidermis, dermis and hypodermis, are thinner and the epidermis does not deepen into the dermis with projections as it does in human skin.Again, the body areas not covered by hairs (ears, tail) manifest a thicker epidermis with a higher cell layer number, like human skin (Fig. 2).Rodent skin is detached from the underlying structures and it includes a muscle layer underneath the adipose tissue, not present in human skin.The muscular layer has an impact in wound closure that in mice is mainly due to early contraction, while in human it relies on keratinocyte proliferation and migration (reepithelialization).Other differences regard: 1) timing of epidermal renewal and hair follicle cycle, which are much faster in mouse than in human (8-10 days against 26-28 days for epidermal renewal, respectively); 2) immune cell composition; 3) melanocyte distribution.Immune cell differences comprise dendritic cell subtypes with a higher number of antigen presenting dendritic cells expressing γδ receptor in mouse skin (MacLeod and Havran, 2011).As for melanocytes, they are distributed in the hair follicle and in the epidermal basal layer in human skin where they interact with keratinocytes, while they are mainly confined to the hair follicles in mouse.This can be a limit for studies on melanoma development, particularly for UV-inducible melanoma genesis that starts in the epidermis.
Tab. 1: Different animal models and their use in main fields of dermatological interest Species Melanoma Wound healing

Pig and large mammals
Pig skin shares highest similarity to human skin, in terms of anatomy and physiology (Fig. 2).Epidermal and dermal thickness is very similar, while the fat layer is thicker in pig skin.Horny layer lipid composition has common features in pig and human skin; epidermis interdigits with the upper dermal compartment; hair follicles are sparse in both species; collagen and elastin dermal content is analogous.Moreover, immune cell composition and melanocyte distribution are similar, and pig skin adheres to underlying tissue as human skin does.For all these similarities, pig is a model for studies in all dermatological fields, such as skin surgery, melanoma research, wound healing and therapies, microbiology, vitiligo and depigmentation (Table 1) (Bourneuf, 2017;Irons et al., 2018;Klein et al., 2018).In particular, the porcine model represents the golden standard for wound healing preclinical studies, as the concordance rate of results obtained in human and pig wound therapy is the highest compared with other animal models (Sullivan et al., 2001).The main limitation in working with pigs and other big mammals is their dimension that impacts on costs, numbers of animals to be used for experimental group, difficulty in manipulation and in genetic engineering.Miniature pigs can be used as an alternative model, as they maintain the benefits of structural and physiological similarities with human skin but minimize seizure limitations.

Zebrafish
The small freshwater fish Danio rerio, known as zebrafish for its stripped tegument, is commonly used for basic research and disease model studies.Several features make zebrafish a valuable research system, such as genome similarity to humans, easy genetic manipulation, embryonic transparency, rapid embryonic development and great offspring size.Another advantage is the lower neurological development compared to mammals, likely implying reduced pain perception and stress susceptibility.Zebrafish is being used as a model system for both skin biology and pathology (Li and Uitto, 2014).Its skin is composed by an epidermis and a collagenous dermis, separated by a basement membrane (Fig. 2).Zebrafish is used for wound healing studies, as major steps found in mammalian skin repair are present (Richardson et al., 2013), although the surrounding liquid environment represents an objective major difference with humans.Zebrafish is extensively used also in studies regarding melanogenesis, pigment disorders and melanoma (Choi et al., 2007;van Rooijen et al., 2017) (Table 1).Indeed, black pigment cells originating from the neural crest (melanophores) are present in the dermis and behave similarly to human melanocytes.As for melanoma studies, genetically modified zebrafish expressing melanoma oncogenes recapitulates melanoma genesis and development.Heritable skin diseases affecting cell adhesion and keratinization disorders were also reproduced in zebrafish (Li et al., 2011;Li and Uitto, 2014).Limitations in using zebrafish as a skin animal model regard the lack of an epidermal barrier and of appendages, and the presence of scales in the adult.However, lack of epidermal barrier allows for systemic drug screening by simply adding compounds to the water.

Drosophila
Despite so far phylogenetically, the fruit fly Drosophila melanogaster is used as an animal model to elucidate molecular networks underlying skin physiology (Table 1).Drosophila has a single-layered epidermis (Fig. 2) and it has been widely used for studies on wound closure.Indeed, several insights in the molecular mechanisms underlying epithelial cell behavior in skin repair were obtained using the Drosophila larval dorsal closure model (Matsubayashi and Millard, 2016).Other dermatological fields using fruit fly models are aging (Scherfer et al., 2013) and immune disorders (Bergman et al., 2017).Moreover, Drosophila was also used to identify molecular mechanisms underlining human genetic skin diseases (Bohnekamp et al., 2015).Drosophila models are convenient as they are genetically tractable, and, due to the large number of offspring generated, allow high-throughput analyses.However, to transfer findings to human dermatological diseases is limited by the structural and physiological differences with human skin.

Skin equivalents
Tissue-engineered human skin equivalents have been developed to reproduce skin key structural and functional properties in vitro.

In Vitro Reconstituted Epidermis
Epidermal equivalents composed of keratinocytes were the first in vitro tool to be generated (O'Connor et al., 1981).Human keratinocytes can be isolated from a small skin biopsy and serially propagated in vitro (Rheinwald and Green, 1975).They reconstitute sheets of stratified epithelia that retain biochemical and histological characteristics, as well as differentiation features of the original donor site (Rheinwald and Green, 1975;Gallico and O'Connor, 1985).Seeding keratinocytes on biological matrices enhances the performance of in vitro reconstituted epidermis.For instance, a fibrin substrate allows keratinocyte stem cell preservation (Pellegrini et al., 1999;Ronfard et al., 2000;Panacchia et al., 2010) and growth factors delivery (Guerra et al., 2009;Panacchia et al., 2010).However, an authentic dermis is required for reproducible cell take and improvement of histological quality of the regenerated skin, as demonstrated by several clinical trials (Cuono et al., 1986;Pellegrini et al., 1999;Ronfard et al., 2000).
Fig. 3: Evolution of in vitro skin models: from reconstituted epidermis to "skin on a chip" Epidermis reconstituted by seeding primary keratinocytes on a biological matrix has been the first in vitro three-dimensional model to be generated.Full-thickness skin equivalents have then been obtained by seeding keratinocytes on a de-epidermized dermis or on biodegradable polymer substrates containing human dermal fibroblasts.Melanocytes have been also added by seeding them with keratinocytes.Last generation skin equivalents incorporate additional skin cell types.Microfluidic platforms have been developed to create "skin on a chip" models: perfusable skin equivalents with endothelial cell-coated wires within the dermal compartment attached to silicone tubes and a peristaltic pump.These skin equivalents within microfluidic device allow to develop "skin-on-a-chip" models of growing complexity with other skin cell types.
Besides their application in the clinical setting as wound coverage, these skin equivalents have been widely used to investigate cellular and molecular mechanisms regulating cutaneous pathophysiology, as discussed below.Nowadays, commercially available full-thickness skin equivalents have become the alternative to animal models for research and development applications in the cosmetic industry (Böttcher-Haberzeth et al., 2010;Brohem et al., 2011;Groeber et al., 2011;Zhang and Michniak-Kohn, 2012)

Last generation skin equivalents
Last generation skin equivalents have been produced by incorporating additional cell types or skin appendages into full-thickness skin equivalents.Mature adipocytes are successfully co-cultured with fibroblasts and keratinocytes to improve balance between epidermal proliferation and differentiation, and to develop a more efficient epidermal barrier (Trottier et al., 2008;Bellas et al., 2012;Delort et al., 2013;Monfort et al., 2013;Huber et al., 2016).Skin equivalents containing adipocytes are also a useful tool to insert hair follicles that are partly hosted in the adipose tissue in vivo.New hair follicles can be generated by bulge-derived epithelial stem cells cultured in proximity of "inductive" follicular dermal papilla cells (Stenn and Cotsarelis, 2005).Three-dimensional spheroid cultures of dermal papilla implanted in human skin and grafted on SCID mice are able to induce de novo hair follicle generation in vivo (Higgins et al., 2013).The same procedure could be reasonably applied in vitro to induce hair follicle formation by inserting spheroids into bulge cell-populated skin equivalents.Recently, an "endogenous" human skin equivalent has been generated by using a dynamic spinner culture of fibroblasts embedded in a temporary scaffold made of gelatin microspheres.The gelatin microscaffold is slowly degraded while fibroblasts assemble their own extracellular matrix.This new microenvironment promotes the morphogenesis of follicle-like structures in vitro, in the absence of dermal papilla cells, and it also improves epidermal barrier function (Casale et al., 2016).
As for immunological cells, CD34-positive hematopoietic progenitor cells have been cultured in skin equivalents.Following treatments with defined cytokine cocktails, these cells are able to differentiate into Langerhans cells and to properly localize in suprabasal layers (Régnier et al., 1997;Facy et al., 2005).
The major weakness of skin equivalents is their lack of a functional vascular system.Nutrients, oxygen supply, and metabolic by-product removal is achieved by manual media changes, leading to periodic fluctuations in culture medium composition (Bauwens et al., 2005;Zhao et al., 2008).Avascular skin equivalents cannot be used as models for studying key features of diseased skin, such as trafficking of leukocytes across vessel endothelium, or for evaluating transdermal drug penetration into the blood stream and/or skin localization of an intravenous administered substance.In 1998, the first skin equivalent with a capillary-like network was obtained by co-culturing keratinocytes, fibroblasts and umbilical vein endothelial cells (HUVEC) in a collagen biopolymer (Black et al., 1998).Skin equivalents with endothelial cells have been mainly used in a clinical setting to ameliorate graft uptake, or employed to test angiogenic and angiostatic drugs (Hudon et al., 2003;Schechner et al., 2003).To allow long-term analyses, modifications have been introduced to this basal setting.Networks with both lateral and vertical vessel expansion were obtained using an extracellular matrix of human origin with dermal microvascular endothelial cells (HDMEC) isolated from the same skin biopsy from which keratinocytes and fibroblasts were derived (Ponec et al., 2004).A fibroblast-derived extracellular matrix produced after vitamin-C treatment has been found to be essential for correct capillary development (Berthod et al., 2006).Moreover, co-culturing endothelial cells and adipose-derived stem cells in fibrin gels induces formation of a stabilized vascular network.Notably, adiposederived stem cells acquire typical features of pericytes, the smooth muscle cells surrounding vessels (Rohringer et al., 2014).Lymphatic capillaries have been generated by using three-dimensional hydro gels (Marino et al., 2014).To generate vascularized tissues, decellularized porcine small bowl segments have been also used, as they preserve a collagen matrix scaffold with structures of the native vascular network.This scaffold is repopulated with allogeneic porcine endothelial progenitor cells that achieve typical endothelial differentiation features, are vital up to 3 weeks post reconstitution and form a vascular network (Mertsching et al., 2005;Schanz et al., 2010).Decellularized plant tissues have been also used as scaffolds to be colonized by endothelial cells (Gershlak et al., 2017).
To reproduce the peripheral skin nerve system, fibroblasts have been cultured with endothelial cells and dorsal root ganglion neurons in a collagen-chitosan sponge, with or without keratinocytes seeded above (Gingras et al., 2003a).Similarly, to study motor neuron axonal migration and myelination, mouse spinal cord motor neurons are seeded on a collagen sponge populated with Schwann cells and fibroblasts (Gingras et al., 2008).Very recently, neural stem cells, generated through direct reprogramming of dermal fibroblasts, have been incorporated into silk-collagen-based skin equivalents with immune and adipose cells to obtain a neuroimmune-cutaneous system (Lightfoot et al., 2018).
A limitation of these skin equivalents is the limited availability and the low proliferative potential of cultured skin primary cells.Induced pluripotent stem cells (iPSCs) can differentiate in all skin cell types, including sensory neurons and dermal papilla cells (Abaci et al., 2017;van den Broek et al., 2017), and have been already used to generate full-thickness skin equivalents.Thus, they may be a valid source for large scale production of different skin cell types, with enhanced reproducibility.

Next generation skin equivalents: towards fully reconstituted and functional skin
Despite all efforts toward improved skin equivalents, an organotypic model recapitulating the entire complexity of human skin is not yet available.Developing skin equivalents within microfluidic devices represent a step forward for models of growing complexity, and for better mimicking skin function (van den Broek et al., 2017).Microfluidic platforms are an ensemble of controllable chambers and channels in the µm range, combining microsystems and cell biology techniques.Early-developed microfluidic devices were produced using silicone microfabrication and micromachining techniques (Chiu et al., 2001).Less expensive and easier to fabricate devices have then been developed by employing the biocompatible silicone rubber poly(dimethyl siloxane) (PDMS) (Whitesides, 2006).Microfluidic platforms allow to obtain perfused vascular networks, thus mimicking the in vivo physical force exerted by flowing blood (shear stress), essential for modulating endothelial cell gene expression, morphology, proliferation and apoptosis (D'Arcangelo et al., 2006(D'Arcangelo et al., , 2016a;;Song and Munn, 2011).A dynamically perfused chip-based bioreactor chamber able to apply variable mechanical shear stress to skin equivalents was first described in 2013 (Ataç et al., 2013).This platform has been used for culturing skin equivalents or skin biopsies or explants of single hair follicular unit.It allows long-term maintenance of composite skin equivalents, and to perform multiple tests without devise disassembly or tissue disruption.Moreover, the tissue can be easily removed from the device for histological processing and further analyses.To obtain perfused vascular networks, Groeber and coworkers employed a model in which the decellularized porcine jejunum segment, physiologically supplied by a single artery-veinpair, was reseeded with HDMEC and perfused by using a bioreactor system generating a physiological pulsatile pressure (Groeber et al., 2013).The same porcine scaffold was used to generate a vascularized skin equivalent by seeding human fibroblasts and keratinocytes on its surface.The bioreactor system supports the skin equivalent under submerged conditions or at the air-liquidinterface, allowing the formation of typical histological skin architecture and of a functional barrier (Groeber et al., 2016).In general, microfluidic systems, by allowing a dynamic perfusion and a finely tuned control of the airflow and gas composition of the airexposed region, improve epidermal stratification, differentiation and barrier functions (Sriram et al., 2018).
A microfluidic devise has been recently used to generate a fully innervated human skin tissue recapitulating and replicating skin sensory function in vitro by supplementing skin equivalents with rat dorsal root ganglion (Martorina et al., 2017).However, the current unavailability of human sensory neurons in culture limits the development of such a model.Application of microfluidic devices is also useful to create skin immuno-competent models.To this end, a bi-channel device has been used to culture the keratinocyte cell line HaCaT as a model of the epidermis barrier, on one side, and a human leukemic monocyte lymphoma cell line (U937) to represent human dendritic cells, on the other side (Ramadan and Ting, 2016).The effects of chemical and physical stimulations, such as ultraviolet irradiation, are assessed in this system by measurement of the trans-epithelial electrical resistance and with a magnetic-bead immune assay integrated into the device.Similarly, to mimic skin inflammation and edema for drug testing, a model based on three microfluidic channels has been created (Wufuer et al., 2016).The HaCaT cell line, mimicking the epidermis, dermal fibroblasts and HUVEC, recreating the endothelial layer, are separated by porous membranes to allow interlayer communication.Skin inflammation and edema are induced by applying pro-inflammatory cytokines on the dermal layer and drug efficacy is evaluated by analyzing epidermal barrier functionality and endothelium permeability.
To further improve skin equivalent complexity, 3D bioprinting technology has been employed.Indeed, this fully automated system permits deposition of several cell types and biomaterials to build a structure similar to native human skin (Ng et al., 2016;Lee and Dai, 2017).By using 3D bioprinting, Abaci and co-workers developed a skin equivalent with perfusable, pre-organized vasculature with both primary and iPSC-derived endothelial cells.Sacrificial alginate channels have been suspended in transwell inserts on which skin equivalents are generated by adding fibroblasts in a collagen matrix, as dermal compartment, and keratinocytes for the epidermal one.After the cornification process, the sacrificial channels are dissolved and the alginate removed to leave a hollow network.Endothelial cells are then added to obtain a vascularized skin equivalent (Abaci et al., 2016).Notably, the same authors developed a pumpless "skin-on-a-chip" model in which the culture medium recirculates by placing the devices on a rocking platform (Abaci et al., 2015).Via 3D bioprinting, Mori and co-workers created a perfusable skin equivalent inserting HUVEC-coated nylon wires within the dermal compartment.The device is attached to silicone tubes and a peristaltic pump.Beyond skin barrier function, this model displays an efficient percutaneous penetration in endothelialized tubes following drug application (Mori et al., 2017).Recently, a 3D printed niche matrix has been created which directly and restrictively guides in vitro epidermal progenitor differentiation into sweat gland morphogenesis (Liu et al., 2016).3D bioprinting technology may be useful to reconstruct miniorganoids reproducing the organization of skin appendage niches.Addition of skin appendages will likely further enhance skin equivalent barrier properties.
Native dermis is composed of multiple fibroblast subpopulations that differ in density, anatomic distribution and functions.Extracellular matrices also exhibit differences in composition and organization (Sriram et al., 2015).3D bioprinting technology may be helpful to recapitulate the dermal native composition by combining different fibroblast subpopulations with diverse extracellular matrix components.
Altogether, these progresses pave the way to connect the skin with in vitro models of other tissues to create a "human-on a chip" system for drug screening.Currently, skin has been connected with organs such as liver, intestine and kidney (Ataç et al., 2013;Abaci et al., 2017;van den Broek et al., 2017).

5
In silico models In silico-based approaches represent innovative strategies to reduce the use of animal models and to contain expenses on both research materials and time efforts.Given the extreme complexity of the field, appropriate bioinformatic tools to manage the enormous number of biological variables are still under development.In the last years, increasing effort has been put to set up in silico strategies based on clinical data sets and novel analytical algorithms for usage in evidence-based medicine and biomarkers identification for both diagnostic and prognostic purposes (Ozerov et al., 2016;Pinto et al., 2017).Such strategies can also be exploited in drug re-positioning, metabolic pathway analyses (Scafuri et al., 2016;Abbas et al., 2017) or toxicological investigations (Dumont et al., 2015;Kranthi Kumar et al., 2017).In the dermatological field, in silico strategies have been applied to 3-D tissue engineering (Vijayavenkataraman et al., 2017), analysis of drug delivery to the skin (Gajula et al., 2017), and prediction of skin permeability (Rocco et al., 2017).Several public databases are available hosting large clinical data that refer to gene and protein expression of healthy or disease-affected individuals.Samples are stratified according to different pathologies; however, the most numerous datasets are in the cancer field.Analyzing such large data collection allows researchers to set and test clinical hypotheses directly on hundreds or thousands human specimens.Therefore, mechanistic hypotheses can be tested and somehow refined in silico before being moved toward more "experimental" stages.Upload of new datasets is daily occurring, further increasing the number of available samples to be tested.Thus, the possibility to in silico compare disease conditions, such as stage, with patient gender or age or survival, is every day more concrete.However, efforts should be paid in recruiting more numerous healthy controls, since data from healthy individuals are less available than those from affected ones.
In the present review, we focus on the most represented databases with an overview of their application in dermatological issues.These databases contain datasets with variable case numbers, ranging from very few to hundreds of cases.Statistical evaluation largely depends on the quality of the reported expression data.In fact, samples quality is routinely evaluated before any further analysis.The actual sample numerosity is also crucial, since larger homogenous experimental datasets better represent the general population.In silico applications to melanoma and skin wound healing will be reported in the dedicated paragraphs.

Gene Expression Omnibus
Gene Expression Omnibus -GEO2 is a searchable database comprising datasets (named "GDS" as GEO Data Set) or series (named "GSE" as GEO SEries), all referring to published studies.Data can be directly accessed and transferred to statistical packages for evaluation.Data from human biopsies as well as from cultured cell lines, both of human and non-human origin, are available.Several studies have been reported exploiting GEO to address in silico clinical issues, such as identification of genes involved in breast cancer aggressiveness (Guo et al., 2017), key genes in hepatocellular adenoma (Liu, et al., 2017a), lung cancer (Liu et al., 2017b) or colorectal cancer (Geng et al., 2017;Wu et al., 2017;Zong et al., 2017).Gene expression data from GEO have been used to validate autoantibody signature in meningioma patients (Gupta et al., 2017a).Relevant up-or down-regulation of key genes identified with GEO analyses were usefully exploited to propose novel prognostic markers in pancreatic ductal adenocarcinoma (Ma et al., 2017), thyroid cancer (Chang et al., 2017) and glioblastoma (Bo et al., 2017).For instance, high levels of interleukin-6 receptor mRNA have been related to a better prognosis in ovarian cancer patients (Chen et al., 2017).Expression profile data from this database have been also investigated to identify genes involved in rheumatoid arthritis pathogenesis (Zhang et al., 2017a), endometrial function (Young et al., 2017) and Japanese encephalitis virus-infected cells (Gupta et al., 2017b).GEO has been extensively used to investigate dermatological pathologies.Inkeles and colleagues (Inkeles et al., 2015) analyzed gene expression in 16 different skin-inflammatory, skin-infectious and skin-neoplastic conditions, such as leprosy or psoriasis, using Affimetrix data from 311 human skin biopsies, and their analysis led to identification of molecular classifiers able to discriminate among different diseases with high accuracy, sensitivity and specificity.Such an approach opens new ways toward the molecular characterization of skin diseases and represents a relevant diagnostic improvement.Ghosh and colleagues published a study exploiting GEO to identify genes discriminating between normal skin and atopic dermatitis; 89 genes involved in immune response regulation, keratinocyte differentiation, development, inflammation and lipid metabolism were detected (Ghosh et al., 2015).

Oncomine database
Oncomine3 is a database that allows the analysis of gene expression in cancer versus normal as well as cancer versus cancer samples, either in human biopsies or cultured cells.Investigations in Oncomine have led to published studies since 2007.Long non-coding RNA PVT1 expression has been studied in hepatocellular carcinoma, identifying the DLC1 gene as negatively correlated to PVT1 expression (Zhang et al., 2017b).The melanoma antigen A12 has been investigated in prostate and colorectal cancer cells demonstrating its role as cell cycle regulator via p21 (Yanagi et al., 2017).Other genes such as RAD51, GINS1, TRIP13 and MCM2, involved in DNA-damage control, have been related to a worse outcome in breast cancer (Nieto-Jiménez et al., 2017), while FRG1, a component of the human spliceosome, has been related to tumor progression and tumor angiogenesis (Tiwari et al., 2017).Oncomine analysis was also useful to reveal that cyclin D1 over-expression correlates with poor prognosis in gastric cancer (Shan et al., 2017).
We investigated the expression of 90 ion channels genes in a total of 3673 patients reported in Oncomine (namely, 2999 samples of bladder cancer, glioblastoma, melanoma, breast invasive-ductal cancer, lung carcinoma patients and 674 controls), demonstrating that their expression was significantly modified in tumor samples and hypothesizing a relationship between such modifications and altered angiogenesis (Biasiotta et al., 2016).In this study, the in-silico analysis of publicly available data allowed to test a working hypothesis on thousands human specimens from different cancer types, an investigation not doable with a conventional experimental approach.

Other available databases
IST online database (ist.medisapiens.com)comprises transcriptomic data and allows interesting correlation plots of the genes of interest in several hundred human specimens, from cancer to healthy tissues.In the last 10 years the Human Protein Atlas4 (HPA) has become available, representing a large collection of protein expression and protein localization data in several healthy and diseased tissues as well as cell lines (Thul and Lindskog, 2018).HPA can be searched for any given protein.Several studies exploited HPA in diverse pathological contexts, such as diabetes (Lindskog et al., 2012), hearth (Li et al., 2013) and kidney diseases (Wang et al., 2017a).Skin models and diseases have been largely investigated with HPA, allowing to identify skin-enriched proteins (Edqvist et al., 2015), proteins distinctive of squamous cell carcinoma (Azimi et al., 2016) or expression of transcription factors in several cancers including skin ones (Kaufhold et al., 2016).
Besides human databases, a mouse gene expression atlas5 is also available.It has been used to evaluate gene expression in Huntington disease (Mazarei et al., 2010) and to investigate mouse tissue development and functions (Hoffman et al., 2006).This kind of databases may be relevant to understand aspects of mouse biology without directly performing experiments in the animal models.

6
Non-animal models for the study of skin pathophysiology 6.1 Immunological diseases 6.1.1Irritant and allergic contact dermatitis One of the critical fields in which alternative methods would be commendably employed is the detection of adverse drug reactions.
For in vitro modelling of irritant dermatitis, where a cosmetic or an allergen is directly toxic to the keratinocytes, skin equivalents are already in use.Efforts to replace the in vivo Draize skin irritation test led in 2007 to validation of the reconstructed human epidermis model EpiSkin TM as the only stand-alone test showing sufficient sensitivity and specificity (Eskes et al., 2007).With the completion of the validation study, performance standards have been defined (ECVAM SIVS6 , May 2007), and then used to evaluate the accuracy and reliability of other analogous systems.For example, the reference test method is used by Alépée et al. to demonstrate the equivalence of the in vitro SkinEthic TM RHE skin irritation assay in a formal inter-laboratory study (Alépée et al., 2010).
Efforts have been also made to develop reliable in vitro assays able to distinguish sensitizers from non-sensitizers for the safety evaluation of substances and for eliminating animal studies, such as the guinea pig test or the murine local lymph node assay.The pathway that leads to allergic contact dermatitis is well known: 1) allergen penetrates the skin and can react towards endogenous proteins as it is (hapten), or can be metabolically activated in order to react (pro-hapten); 2) allergen activates a cytoprotective pathway in keratinocytes and, in parallel, keratinocytes elaborate an inflammatory response by secreting pro-inflammatory mediators; 3) pro-inflammatory cytokines, in turn, induce activation, maturation, mobilization and migration to proximal lymph node of cutaneous Langerhans cells, mature dendritic cells; 4) in the lymph node, these cells present antigen to naïve T lymphocytes that start proliferating and differentiate into specific T-cells.
Considering the complexity of this pathology, developing alternative methods has proceeded by evaluating the abovedescribed four key events that lead to the adverse outcome pathway (AOP), i.e. inflammation upon challenge with an allergen.Today, the first three key events are assessable by non-animal methods approved by the Organization for Economic Cooperation and Development (OECD).To mimic the cascade first step, the direct peptide reactivity assay (DPRA), or peptide binding assay, has been developed (OECD TG442C).This is a cell-free assay using artificial model peptides, developed by Gerberick and co-workers in 2004.It is based on the ability of a chemical allergen to react with proteins prior to induction of skin sensitization (Gerberick et al., 2004).Since most chemical allergens are electrophilic, they react with neutrophilic amino acids, such as cysteine or lysine.Test substances are diluted in an appropriate buffer, depending on the amino acid model peptides, incubated with the peptides, and free peptides analyzed by high-pressure liquid chromatography (HPLC).A skin sensitizer molecule will induce a high depletion of free cysteine and lysine peptides.To address the induction of a cyto-protective pathway in skin keratinocytes, the KeratinoSens TM assay has been developed (OECD TG442D).It utilizes genetically-modified HaCaT carrying a reporter gene for luciferase under the control of an antioxidant-response element, thus monitoring the activity of Nrf2 transcription factor that accumulates in the nucleus of cells following protein modification by sensitizers (Emter et al., 2010;Andreas et al., 2011).The human-Cell Line Activation (h-CLAT) test (OECD TG442E) aims at testing the third key event, dendritic cell maturation following exposure to skin sensitizers, evaluating changes in mature dendritic cell biomarker expression.The assay uses THP-1 monocytic leukemia cell line as a surrogate for human dendritic cells (Ashikaga et al., 2006;Sakaguchi et al., 2006).These cells are exposed to different concentrations of substances under examination and cell viability is assessed through a propidium iodide cytotoxicity assay.Membrane expression of CD54/ICAM-1 and/or CD86/BT.2 as differentiation markers is then analyzed as final readout.A variant is represented by the U-SENS TM assay, formerly known as Myeloid U937 Skin Sensitation Test (MUSST) that uses U937 cells and measures CD86/BT.2 induction by flow cytometry (Python et al., 2007)(OECD TG442E), and another variant by the IL-8 Luc assay (Kimura et al., 2015)(OECD TG442E), where the effects of chemicals on IL-8 promoter activity are evaluated by an IL-8 reporter cell line, THP-G8 cells.A validated assay for the forth key event, the immunological response in the draining lymph node, is still lacking.In this field, the human T-cell priming assay (hTCPA) has been proposed (Dietz et al., 2010;Richter et al., 2013).The method uses naïve T lymphocytes and monocyte-derived dendritic cells exposed to the chemical and evaluates by flow cytometry the frequency of antigen-specific T-cells and production of the cytokines IFN-γ and TNF-α.While simple in principle, it is quite complicated from an experimental point of view and not much informative.Additional work is needed to further clarify this aspect (van Vliet, 2017).
It is now clear that, to reach a good predictivity and improve on animal-based methods, different in vitro tests should be performed in sequence and only combination of methods in integrated approaches to testing and assessment (IATA) would be effective (Bauch et al., 2012;Roberts and Patlewicz, 2018).

Atopic dermatitis
Atopic dermatitis is a chronic, eczematous skin disease caused by a disequilibrium in the interaction between immune and skin barrier systems, and resulting in the formation of erythematous plaques, eruption, lichenification accompanied by pruritus and cutaneous hypersensitivity (Otsuka et al., 2017).Acute skin lesions show T-helper type 2 (Th2) lymphocyte inflammation with CD4+ T-cell and eosinophil infiltration.The subsequent chronic phase shows local Th1 interferon-γ responses, augmented number of mast cells and dendritic cells, tissue remodeling with increased collagen deposition.Moreover, atopic dermatitis manifests changes in the chain length of the stratum corneum lipids.Both environmental and genetic factors have been linked to the development of atopic dermatitis.In particular, mutations in the genes encoding filaggrin, the serine peptidase inhibitor Kazal-type 5 and corneodesmosin, three components of the epithelial barrier, are associated with atopic dermatitis.Several mouse models for atopic dermatitis have been developed and have been essential for a deep understanding of disease pathogenesis.They comprise: 1) models induced by epicutaneous application of sensitizers; 2) transgenic mice that over-express or lack selective molecules, including the components of the skin barrier that are mutated in human patients; 3) mice that spontaneously develop atopic dermatitis-like lesions in the skin.These models display many but not all features of the human disease (Kabashima and Nomura, 2017).To give answer to specific molecular questions on the development of atopic dermatitis, skin equivalents have been proven relevant as well, enabling the controlled induction of atopic dermatitis-associated features by the selective addition of proinflammatory cytokines (Kamsteeg et al., 2011).Such models can be only used for the study of epidermal pathologic features and for identification of drugs that act at the epidermal level.In fact, most of the known atopic dermatitis therapeutics that target the immune system are not able to reverse the atopic phenotype of these skin equivalents (Kamsteeg et al., 2011).Danso et al. found that addition of IL-4, IL-13, IL-31 and TNF-α during generation of the skin equivalents induces intercellular edema (spongiosis), altered expression of epidermal differentiation proteins and secretion of thymic stromal lymphopoietin, a master regulator of Th2-driven inflammation by keratinocytes (Danso et al., 2014).Moreover, such a treatment decreases long-chain fatty acid and ceramide levels, thus affecting lipid organization, all hallmarks of atopic dermatitis.More recently, a skin equivalent has been obtained with keratinocytes that do not express filaggrin cultured in the presence of the Th2-derived cytokines IL-4 and IL-13 (Hönzke et al., 2016) or of activated CD4+ T-cells (Wallmeyer et al., 2017).These new skin equivalents represent a valid alternative to animal models for identifying the relative contribution of filaggrin deficiency and inflammation in the development of atopic dermatitis pathological manifestations.

Psoriasis
Psoriasis is a chronic inflammatory disease characterized by erythematous-squamous plaques with epidermal hyperplasia, increased vascularity, and a dermal mixed leukocyte infiltrate (Liang et al., 2017).Both skin innate and adaptive immune systems are causally involved in disease pathogenesis.To develop a relevant in vitro model of psoriasis, multiple interactions between different cell types, such as epidermal keratinocytes, endothelial cells, neutrophils and T-lymphocytes, should be considered.So far, such a cell complexity has not been achieved in vitro.Skin equivalents have been produced with keratinocytes and/or fibroblasts derived from psoriatic patients in order to reproduce the cytokine and chemokine expression pattern of psoriasis (Barker et al., 2004;Jean et al., 2009), or with keratinocytes treated with transglutaminase inhibitors to induce typical features of psoriatic keratinocytes (hyperproliferation and parakeratosis, i.e. retention of nuclei in the keratinocytes of the horny layer) (Harrison et al., 2007), or in which lymphocytes are included (van den Bogaard et al., 2014).Angiogenesis is another important hallmark of psoriasis.Using the self-assembly methods without any scaffold-like biomaterials, Ayata et al. proposed a skin equivalent composed either of psoriatic fibroblasts and keratinocytes or healthy cells, with the addition of normal skin microvascular endothelial cells (Ayata et al., 2014).In this model, capillary-like structures are more numerous, complex and branched in the psoriatic substitutes than in the non-psoriatic ones, clearly indicating an effect of psoriatic keratinocytes and fibroblasts on endothelial cells.Although diseased cells have the intrinsic ability to give rise to a psoriatic phenotype, their use is limited by difficulties in obtaining skin biopsies from patients, and by heterogeneity of patient-derived cells.Therefore, a different approach has been used by other authors.Starting from a reconstructed skin model with normal adult keratinocytes, they induced a psoriatic phenotype through addition of pro-inflammatory cytokines (Sa et al., 2007;Tjabringa et al., 2008).Depending on the type of cytokines used, these models may be useful to study the psoriasis pathogenesis without the need to incorporate neutrophils or T-lymphocytes into the skin equivalent.

Cutaneous T-cell lymphomas
Cutaneous T-cell lymphomas are a heterogeneous group of malignant lympho-proliferative disorders of which the most frequent form is mycosis fungoides (MF) (Dummer et al., 2000).From a primary involvement of the skin with patches and plaques, tumor cells may spread to lymph nodes and visceral sites.In MF, cancer cells have a mature CD4+ T-helper cell phenotype.Despite MF tumorigenic nature, these cells hardly grow in vitro, and cell line establishment is extremely difficult.At the same time, a limited number of animal models are available, as MF cell lines rarely grow when engrafted in nude mice (Charley et al., 1990;Krejsgaard et al., 2010), making challenging the study of pathogenic mechanisms and preclinical analyses of potential novel therapeutic agents.Considering the difficulties of obtaining tumors by subcutaneous inoculation of cancer cells, other routes have been considered.An example is the human MF cell intra-hepatic xenograft in nude mice (Andrique et al., 2016).This approach leads to quick tumor growth and kidney and lung metastasis formation.However, no skin infiltration is achieved.Taking into consideration the problems met with animal models, cutaneous T-cell lymphomas represent an interesting research area for alternative experimental models.By using organotypic systems, Thode et al. have provided insights into the interplay between MF T-cells and skin keratinocytes and into the MF potential in remodeling the skin barrier (Thode et al., 2015).In this model, cells from established MF cell lines are added to skin equivalents 24 hours after seeding keratinocytes onto the collagen/fibroblasts matrix.Alternatively, devitalized human acellular dermis is used as the supporting framework for 3D cultures (Ridky et al., 2010).Fibroblasts are introduced into the stromal side of the tissue, keratinocytes and MF cells seeded onto the basement-membrane side, and MF persistence as well as migration in the dermis compartment are monitored.These models have led to some informative results, but no data has been reported so far on the action of therapeutic compounds on embedded MF cells, and no trial has been made to establish similar models for other, rarer cutaneous T-cell lymphoma forms.

Aging
Two types of skin aging may be distinguished: intrinsic or chronological aging, and extrinsic or photo-aging.Chronological aging is time-dependent, and it is mainly influenced by genetic factors.Photo-aging depends on UV irradiation.These two aging types are superimposed in sun-exposed skin (Fisher et al., 2002;Kim and Park, 2016).Clinical signs of chronological aging include thinning, wrinkle formation, dryness, loss of elasticity and fragility (Kim and Park, 2016).Histological manifestations comprise epidermal atrophy associated with reduced amounts of melanocytes, Langerhans cells and fibroblasts in the dermis where collagen fibers are loose, thin, and disorganized (Yaar and Gilchrest, 2007).Epidermal atrophy mainly affects the stratum spinosum and it is a consequence of lower epidermal turnover rate (Fuchs, 2016).A skin equivalent model displaying epidermal atrophy has been obtained by over-expressing the senescence effector p16 in keratinocytes.We and others have demonstrated that atrophy severity depends on p16 expression levels in the epidermal component of the skin equivalent (Adamus et al., 2014), and it can be reverted by down-regulating p16 expression (Maurelli et al., 2006;Adamus et al., 2014).Advanced glycation end products (AGEs), resulting from non-enzymatic glycation of proteins and lipids, accumulate during chronological aging (Ahmed, 2005).In the skin, this reaction creates cross-links between collagen fibers, leading to accumulation of insoluble collagen, loss of elasticity and wrinkle development (Corstjens et al., 2008).Moreover, AGEs interact with AGE receptors and other receptors on fibroblasts that become activated and release either growth factors or cytokines involved in inflammatory responses or matrix metalloproteinases (MMPs) that remodel the dermal matrix (Okano et al., 2002;Molinari et al., 2008;Pageon, 2010).Taking this into account, skin equivalents have been developed using artificially-glycated collagen.These models closely mimics features of chronologically aged skin, thus strengthening the pathogenic role of glycation in skin aging (Pageon and Asselineau, 2005;Pageon et al., 2015).Additionally, since aged skin is characterized by inflammatory processes, skin equivalents comprising AGE receptor-expressing monocytes have been developed (Pageon et al., 2017).
Acute or repeated sun exposures induce short-term cutaneous damages, such as skin erythema and sunburn, or long-term effects, such as photo-aging or UV-induced skin cancers (Fisher et al., 2002;Kim and Park, 2016).UV radiations affect both the epidermis and the dermis, and skin equivalent irradiation with defined UVA or UVB doses is able to simulate acute and chronic sun exposure.Thus, skin equivalents are an excellent model to study cell and extracellular matrix modifications induced by acute and chronic sun exposure, and a reliable system to evaluate sunscreen efficacy (Bernerd and Asselineau, 2008;Bernerd et al., 2012).Twenty-four hours after acute exposure to UVB, sunburn cells with DNA lesions appear in suprabasal layers (Bernerd and Asselineau, 1997).Superficial fibroblasts undergo apoptosis after a 6-hour acute exposure to UVA, and are eliminated 48 hours later (Bernerd and Asselineau, 1998).UV exposure at a suberythemal dose potently stimulates production and release of fibroblast-derived elastase that is implicated in the early events of photo-aging leading to drastic alterations of dermal structure and to "solar elastosis" (Bernerd et al., 2000;Bernerd and Asselineau, 2008;Imokawa and Ishida, 2015).UV radiation also increases the production of MMP-1 that hydrolyses type I collagen, thus playing a crucial role in the disorganization and progressive degeneration of dermal extracellular matrix.Under UVA exposure, MMP-1 production is directly induced in dermal fibroblasts.On the other hand, UVB radiation stimulates keratinocytes to release soluble factors that induce MMP-1 secretion by a paracrine mechanism (Fagot et al., 2004;Bernerd et al., 2012).UV radiation also contributes to AGE accumulation, likely through induction of an oxidative environment, thus exacerbating alterations caused by chronological aging (Pageon, 2010).A hallmark of photo-aged skin is the presence of "senilis lentigines" that are associated to increased melanogenesis and melanogenic cytokines in the upper dermis.The influence of dermal fibroblasts on skin pigmentation levels has been demonstrated using a functional pigmented reconstructed skin equivalent composed of a melanocyte-containing epidermis.Notably, natural photo-aged fibroblasts contribute to skin hyperpigmentation associated with photo-aging (Duval et al., 2014).
The protecting effects of sunscreens against the described UV-induced damages may be evaluated in skin equivalents.The presence of a horny layer allows topical application of sunscreen formulations on skin equivalent surface, mimicking in vivo conditions (Bernerd et al., 2000(Bernerd et al., , 2012;;Bernerd and Asselineau, 2008).
Sunscreens and other cosmetics contain a variety of chemicals able to induce genotoxic stress.Notably, several products are applied daily for a long time, thus cell sensitivity to genotoxic stress increases with age.Human skin equivalents overcome the limits of in vitro classical genotoxicity tests, such as the analysis of difference in DNA-repair competence or of the xenobiotic metabolism, due to the fact that these tests are performed on species different from human.Two in vitro genotoxicity tests have been developed using commercially available skin equivalents to investigate DNA damage after chemical topical application.The Reconstructed Skin Micronucleus (RSMN) assay analyzes chromosomal damage by detection of micronuclei in EpiDerm™ tissues (Curren et al., 2006;Hu et al., 2009;Aardema et al., 2010).The 3D Skin Comet assay detects both chromosomal damage and DNA lesions in the EpiDerm™ and in the Phenion® Full-Thickness skin model (Reisinger et al., 2018).These assays are able to address different types of DNA damage with high predictive capability and good intra-and inter-laboratory reproducibility (Hu et al., 2009;Aardema et al., 2010;Reisinger et al., 2018).Thus, the use of both the RSMN and the 3D Skin Comet assay is recommended by the Scientific Committee on Consumer Safety (SCCS), an independent expert panel of the European Commission, as follow-up systems to be applied on suspected misleading results from in vitro standard genotoxicity testing batteries (Reisinger et al., 2018).

Non-melanoma skin cancers (NMSCs)
NMSCs are the most frequently diagnosed cancer in humans (Didona et al., 2018).The majority of NMSCs are basal cell carcinomas (BCC) (70%) and squamous-cell carcinomas (SCC) (25%) arising from the basal layer and the squamous layers, respectively.BCC rarely metastasizes whereas SCC show a variable metastatic rate, a potential for recurrence, and an overall higher risk for human health.BCC cells hardly grow in vitro, and cell line establishment is extremely difficult.Therefore, most studies were carried out on SCC-derived cell lines (Rheinwald and Beckett, 1981;Patel et al., 2012;Adhikary et al., 2013).Remarkably, these cells not always can drive tumor formation and recreate the original SCC histological organization when implanted in mice.Monolayer SCC cultures were used to characterize different steps of cancer progression, but they cannot comprise the effects of cancer-associated fibroblasts, which are crucial for the development of an invasive state in vivo.On the contrary, skin equivalent models allow the study of tumor development dynamics (Obrigkeit et al., 2009;Commandeur et al., 2012;Brauchle et al., 2013).Addition of SCC cell lines onto skin equivalents generates an organotypic model with histological and immunohistochemical similarities to in vivo SCC (Fig. 4C and D).Twenty-four hours after seeding, SCC cells are present in different epidermal layers of the skin equivalent and, in turn, the epidermal compartment displays an altered morphology, resembling SCC in situ.SCC cells can penetrate through the basement membrane into the dermal component.Indeed, clusters of tumor cells may be found in the dermis following additional 5-8 days in culture.However, these skin equivalents do not exhibit either differentiated keratinocytes within invasive cells or the presence of the typical roundshaped foci of hyperkeratosis named 'horny pearls' (Obrigkeit et al., 2009;Commandeur et al., 2012).Recently, two model systems mimicking SCC early and late stage have been established.As a model of pre-invasive stage, SCC cells have been co-seeded with keratinocytes onto the fibroblast-containing substrate.The obtained skin equivalent is characterized by encapsulated keratin pearls in the epidermis.As a model of invasive stage, SCC cells were seeded together with fibroblasts directly into the dermal layer.SCC cells formed hyper-keratinized tumor cell nests resembling "horny pearls" (Brauchle et al., 2013).These models are excellent tools to study dermal-epidermal interactions in SCC initiation and progression, and are reliable in vitro systems to study the effects of therapeutics (Obrigkeit et al., 2009;Commandeur et al., 2012).In fact, to evaluate the biological effects observed in vivo by the combined administration of the two chemotherapeutic drugs gemcitabine and a small inhibitor of Wee1, we used a skin equivalent model and found that modulation of both CDC2 phosphorylation and of a previously-identified gene signature was clearly detectable in human epidermis, suggesting that this system represents a promising tool for identification and validation of novel pharmacodynamic biomarkers (Paolini et al., 2013).

Alternative technologies for melanoma studies
Melanoma has been largely investigated using human and non-human melanoma cell lines, as well as animal models.Human melanoma tissue arrays have been used in early 2000s (Baldi et al., 2002) and employed to validate molecular mechanisms underlying melanoma development (Kerkar et al., 2016).As an alternative, 3D models better mimicking tumor microenvironment have been developed to study melanoma progression and for preclinical drug screening.

Spheroids
Cells dissociated from an organ and put in a dish under appropriate culture conditions have the intrinsic capability to reorganize and reform a smaller, simpler version of the originating organ, named organoids or spheroids.In the first innovative study with spheroids, a mini-gut was developed starting from intestinal epithelial stem cells (Sato et al., 2011).Subsequently, spheroids have been obtained from stem cells of other epithelial organs (Duarte et al., 2017), and tumors (Fatehullah et al., 2016).Spheroids can be formed from freshly tumor-derived cells, from established tumor cell lines, or from single cancer stem cells after clonal expansion (Ishiguro et al., 2017).Spheroids represent a methodology to preserve and store clinical tumor samples (van de Wetering et al., 2015) and have been extensively used to get insight into tumor growth and metastasis formation (Gao and Chen, 2015) or for drug discovery (Weeber et al., 2017).As spheroids can be derived from a single cell and be ready for drug testing within 2-3 months, they represent a step towards personalized medicine, even if the high amount of cytokines and factors in the culture medium can influence the responses to various compounds (Broutier et al., 2017).Spheroids are likely better representative of early tumors, however they maintain most of the feature of the originating cells (Sachs and Clevers, 2014).Melanoma spheroids have been developed by culturing human melanoma cell lines alone or together with stromal cells (Marconi et al., 2018).Melanoma spheroid models have been used to assess several aspects of melanoma biology including growth regulation by cyclin-dependent kinases (LaRue et al., 2004), involvement of laminins in melanoma differentiation (Larson et al., 2014), melanoma drug screening (Müller and Kulms, 2018;Stevenson et al., 2018).

Skin equivalent-based models
Skin equivalents for melanoma studies have been produced by culturing primary and metastatic melanoma cells from established cell lines plated, together with keratinocytes, on fibroblast-contracted collagen gels.Interestingly, cells from a radial growth phase melanoma nestle within the epidermis and at the epidermal-dermal junction, whereas cells from a vertical growth phase melanoma or from metastatic melanoma colonize the epidermis and infiltrate deeply into the dermis (Meier et al., 2000).Using both spheroids and melanoma skin equivalents models, Ostyn and colleagues showed that the pro-inflammatory cytokine tumor necrosis factor-alpha (TNF-α) expands the cancer stem cell compartment (Ostyn et al., 2014).To identify the best matrix for melanoma growth, comparisons have been carried out among bovine or rat-tail collagen matrix, human fibroblast-derived matrix, non-cellular deepidermized dermis and a fully-cellular de-epidermized dermis with an intact basement membrane.Notably, melanoma cells of nonmetastatic origin grow in the dermis of skin equivalents only with fibroblast-derived matrix (Commandeur et al., 2014).Therefore, additional analyses aimed at investigating early melanoma invasion have been carried out with fully humanized skin equivalents in which fibroblasts are stimulated to produce their own extracellular matrix.Melanoma cells are then seeded onto the dermal equivalent prior to the incorporation of human keratinocytes (Hill et al., 2015).Finally, with the aim of recapitulating metastasis formation, the two above described techniques have been merged: melanoma spheroids have been prepared and then settled into the dermal fibroblast/collagen scaffold of the skin equivalent prior to the addition of keratinocytes and their differentiation (Vörsmann et al., 2013).

In silico models
Different melanoma studies have been carried out by exploiting in silico investigations with the above-mentioned databases.Wang et al. have identified in GEO long non-coding RNAs differently expressed in melanoma biopsies as compared to normal skin (Wang et al., 2017b).GEO gene-expression data have been analyzed to investigate specific molecular immunoglobulin pathways (Saul et al., 2016), as well as vascular endothelial growth factor (VEGF)-A and melanogenesis associated transcription factor (MITF) expression (Puujalka et al., 2016) in human melanoma.According to data reported in four different GEO datasets (namely GDS1375, GDS1989, GDS2200, GDS1965) referring to 94 patients, we have evaluated the expression of different isoforms of platelet-derived growth factor receptors (PDGF-R) in human biopsies from melanoma versus nevi, demonstrating that PDGFR-α is strongly reduced in melanoma samples (Faraone et al., 2009).Such data suggested a possible onco-suppression function of PDGFR-α in melanoma pathogenesis and represented a preliminary indication for future experimental studies about the over-expression of PDGFR-α to inhibit melanoma proliferation.Other studies have confirmed the in silico data, further addressing the mechanisms underlying PDGFR-α anti-tumor activity (D'Arcangelo et al., 2016b).Using the IST online database we demonstrated that expression of PDGFR-α is almost completely lost in 208 melanomas as compared to 147 normal skin biopsies, and experimental data showed that PDGFR-α acts as a potential onco-suppressor gene, negatively selected in melanoma (D'Arcangelo et al., 2016b).Integration of expression data from the NCI-60 panel of cell lines available in GEO with expression data in human patients available in HPA, have led to the clarification of oxygen consumption and other metabolic strategies in melanoma (Aurich et al., 2017).With regards to Oncomine database, a negative correlation has been found, indicating the possibility to use Annexin V as a novel angiogenesis inhibitor in melanoma (Zhang et al., 2017c).Moreover, in silico analysis carried out onto Oncomine have been used to validate skin cancer proteomic data (Azimi et al., 2016).

Non-animal methods in wound healing
Cutaneous wound healing is a complex process aimed at the restoration of skin integrity and function following injury.It has been conventionally divided into sequential and partly overlapping phases: hemostasis, inflammation, cell proliferation and tissue remodeling.All resident cell types, as well as cell populations recruited via the blood stream, take active part in this process (Eming et al., 2014).At present, such a complexity cannot be recapitulated in vitro, and different animal models are used to investigate cellular and molecular events taking part in physiological or altered wound healing, as well as for testing therapeutic approaches (see chapter on animal models).However, a continuous effort has been put in the development of alternative methods for wound healing to partly replace animals for these analyses.

Two-dimensional models
The simplest in vitro method to investigate cell behavior and molecular signaling following injury is by mechanically performing a wound in confluent monolayers of a single cell type.Cell migration can be analyzed by monitoring cell gap closure with a time-lapse microscopy; molecular events can be also investigated.Diverse approaches are used for interrupting cell monolayers.The scratch assay uses pipette tips (Liang et al., 2007), toothpicks (Klettner et al., 2014), special cell scraper scratching devices (Doyle et al., 2012;Zhang et al., 2013a), or metallic micro-indenter (Topman et al., 2012).Modifications of these methods have been developed to obtain automation for high throughput studies (Yue et al., 2010;Liu et al., 2014;Brönneke et al., 2015).Another possible way to mechanically remove cells is by destroying them with a stamp that has the advantage of not removing the matrix coating that is often damaged with scratches.For most stamping assays, a mould made of rubber (Lan et al., 2010) or polymers (Lee et al., 2010) is used.Other methods employ temperature-controlled stamps (Hettler et al., 2013).Cell layer wounding can be also obtained with electrical (pulses of high voltage current) or optical (laser beam) devises; both methods allow better reproducibility and high throughput process (Stamm et al., 2016).
Keratinocytes, fibroblasts, and endothelial cells have singularly been analyzed in two-dimensional assays.Despite useful for investigating some aspects of the wound healing process, confluent cell monolayers are far from reproducing the in vivo situation, particularly for dermal fibroblasts that are normally embedded in the extracellular matrix with few direct cell-cell contacts.As for keratinocytes, lack of differentiation and barrier formation can limit the information got from two-dimensional assays.An in vitro assay better recapitulating the re-epithelialization process with barrier function restoration was set up on corneal stratified epithelial cells in which wounds were performed with a metal circular punch (Gonzalez-Andrades et al., 2016).Lack of epithelialmesenchymal crosstalk is another major limit of 2D in vitro wound models.

Three-dimensional assays
Three-dimensional wound healing models have been developed to better mimic the in vivo cell-cell and cell-matrix interactions.Different types of skin equivalents containing fully differentiated epidermis and fibroblasts embedded within a matrix have been used (Egles et al., 2010).A wound is performed with a scalpel, a punch, or a thermal devise, or following laser irradiation, and it involves either the epidermis only or the skin equivalent in its full thickness.Epithelial closure can be monitored with time, and cellular and molecular events occurring during the healing process analysed on tissue sections or on protein and RNA extracts (Breetveld et al., 2006;Safferling et al., 2013;Marquardt et al., 2015).To overcome the limitation of getting donor skin biopsies, skin equivalents containing TERT-immortalized fibroblasts and keratinocytes have been produced and tested for wound healing responses (Reijnders et al., 2015).Besides recapitulating skin features like primary cell-derived skin equivalents, they can also re-epithelialize and secrete wound healing inflammatory mediators.As a further development for analysis of altered wound healing mechanisms, skin equivalents assembled with cells isolated from fibrotic scars or keloids, or from the skin of individuals with chronic ulcers could be theoretically employed.They have been already set up with keloid cells and used in wound healing studies after grafting onto nude mice (Lee et al., 2016).
Analysis of mechanisms underlying wound healing and of drug effects is limited in conventional skin equivalents due to the lack of cell populations playing pivotal roles in skin repair, such as endothelial cells for vessel formation, immune cell types, adipocytes, hematopoietic and endothelial precursors cells, nerve endings.Attempts have been made to include at least endothelial cells forming capillary-like tubes and activate an angiogenic response following skin equivalent injury (Herman and Leung, 2009).Hair follicles, containing epidermal and mesenchymal precursors have been cultured in skin equivalents to study their contribution to wound healing, particularly for re-epithelialization (Ojeh et al., 2017).A skin model composed of an injured epidermis and a dermis innervated with sensory neurons has been also established to study the re-epithelialization process (Gingras et al., 2003b(Gingras et al., , 2008;;Blais et al., 2009Blais et al., , 2014)).
Alternative 3D methods for evaluating novel therapies to promote healing include the use of human skin biopsies taken from abdominal or breast reduction surgery, which are cut into pieces, cultured within matrices with the epithelium at the air-liquid interface, and wounded (Heilborn et al., 2003;Mendoza-Garcia et al., 2015).This ex-vivo wound healing model has the advantage of maintaining skin characteristics, with keratinocyte proliferation and dermal matrix deposition after wounding.However, cell apoptosis and loss of tissue architecture occur starting from 7 days of culture, not allowing long-term wound healing analyses (Xu et al., 2012;Mendoza-Garcia et al., 2015).

Microfluidic models
Microfluidic models represent a further step for recapitulating a wound environment more closely resembling the in vivo situation.Soluble factors, gradients, gas concentration, fluid flow, extracellular matrix composition, mechanical properties, geometry are some of the parameters which microfluidic technology allows to more finely control (Meyvantsson and Beebe, 2008).A lab-on-a-chip system capable of mechanically inducing circular cell-free areas within confluent cell layers in microfluidic environment has been developed to allow automated, miniaturized and integrated induction of defined wounds (Sticker et al., 2017).Aspects investigated by using microfluidic devices include the control of biofilm infection for chronic wounds (Terry and Neethirajan, 2014), a quantitative investigation of cell migration and proliferation (Zhang et al., 2013b), vascular smooth muscle cell migration (Wei et al., 2015), cell-cell and cell-extracellular matrix interactions (An et al., 2015), endothelial cell migration or sprouting in the presence of growth factor gradients and shear stress (van der Meer et al., 2010;Verbridge et al., 2013).

In silico models
In silico models have been used to reconstruct wound healing mechanisms and to predict the effect of therapeutic approaches.Many different aspects of skin repair have been analyzed by using mathematical models, including the activity of growth factors (Sun et al., 2009), re-epithelialization (Safferling et al., 2013), and angiogenesis (Schugart et al., 2008;Flegg et al., 2012).Collagen bundle distribution and orientation and its effect in wound contraction has been analyzed (Koppenol et al., 2017), as well as impaired healing in hypoxic wounds with varying levels of contamination, the effect of fibroblast death and recruitment rate (Menke et al., 2010), the construction of biomaterials able to support stem cell infiltration and differentiation (Tartarini and Mele, 2015) or the application of hyperbaric therapy to chronic wounds (Flegg et al., 2012).Moreover, agents most involved in skin fibrosis following injury or in delayed healing of diabetic foot ulcers, and best targets to counteract these pathological conditions have been identified (Mi et al., 2007;Nagaraja et al., 2017).

Conclusions
Studies directly referring to non-animal methods were 212 in 1997, 628 in 2007, 1219 in 2017, according to a PubMed search for "alternative to animal model" in ALL filed, reflecting the large and increasing interest towards animal studies and setup of suitable alternatives.Alternative methods for dermatological research include different models, from 2D and 3D culture systems to in silico techniques.At present, the most promising tool is certainly represented by the skin equivalent approach that has evolved from systems containing only keratinocytes seeded on a matrix to more complex cell and matrix aggregates.3D bioprinting and microfluidic devise represent an improvement to reproduce the skin structural, functional and molecular network properties.They more accurately reconstitute skin architecture and mimic blood flow effects, respectively.Further development of skin in vitro systems is a key part of the challenging task toward a more accurate knowledge of the biological systems, properly addressing costs-, time-and ethics-issues.Ideally, future alternative technologies will be able to reproduce skin within an artificial body context, thus being able to recapitulate at least some of the interactions with other organs.A lot of work must be done but it will be highly rewarding.

Fig. 4 :
Fig. 4: Histological comparison of human skin and fullthickness skin equivalents Histological staining of human skin specimen (A) and SCC sample (C), and in vitro skin equivalents generated by seeding normal human keratinocytes (B) or SCC keratinocytes (D) onto fibroblast-containing collagen gel.Arrow indicate a "horny pearl" in the skin dermis.Arrowheads indicate clusters of tumor cells in the dermal compartment of the skin equivalents.Bars: 200 µm.