Building a microphysiological skin model from induced pluripotent stem cells

The discovery of induced pluripotent stem cells (iPSCs) in 2006 was a major breakthrough for regenerative medicine. The establishment of patient-specific iPSCs has created the opportunity to model diseases in culture systems, with the potential to rapidly advance the drug discovery field. Current methods of drug discovery are inefficient, with a high proportion of drug candidates failing during clinical trials due to low efficacy and/or high toxicity. Many drugs fail toxicity testing during clinical trials, since the cells on which they have been tested do not adequately model three-dimensional tissues or their interaction with other organs in the body. There is a need to develop microphysiological systems that reliably represent both an intact tissue and also the interaction of a particular tissue with other systems throughout the body. As the port of entry for many drugs is via topical delivery, the skin is the first line of exposure, and also one of the first organs to demonstrate a reaction after systemic drug delivery. In this review, we discuss our strategy to develop a microphysiological system using iPSCs that recapitulates human skin for analyzing the interactions of drugs with the skin.


Abstract
The discovery of induced pluripotent stem cells (iPSCs) in 2006 was a major breakthrough for regenerative medicine. The establishment of patient-specifi c iPSCs has created the opportunity to model diseases in culture systems, with the potential to rapidly advance the drug discovery fi eld. Current methods of drug discovery are ineffi cient, with a high proportion of drug candidates failing during clinical trials due to low effi cacy and/or high toxicity. Many drugs fail toxicity testing during clinical trials, since the cells on which they have been tested do not adequately model three-dimensional tissues or their interaction with other organs in the body. There is a need to develop microphysiological systems that reliably represent both an intact tissue and also the interaction of a particular tissue with other systems throughout the body. As the port of entry for many drugs is via topical delivery, the skin is the fi rst line of exposure, and also one of the fi rst organs to demonstrate a reaction after systemic drug delivery. In this review, we discuss our strategy to develop a microphysiological system using iPSCs that recapitulates human skin for analyzing the interactions of drugs with the skin.
In recent years there has been signifi cant interest in improving skin constructs, resulting in the development of pigmented [11], vascularized [12] and immunocompetent skin constructs [13]. Recently, an innervated skin model was also developed [14]. Nevertheless, most skin models, containing only one or two cell types and lack skin appendages, and therefore are insuffi cient to capture the complexity of human skin.

Skin complexity in drug response
In addition to fi broblasts and keratinocytes, the skin contains large numbers of hair follicles, immune cells, melanocytes, Merkel cell complexes, blood vessels and nerve fi bers [15]. Diseases of the skin can indicate an increased risk for separate co-morbid disorders. Two examples of this are palmoplantar keratoderma and psoriasis, both of which are associated with an increased risk for cardiovascular disease [16,17].
Moreover, the skin interacts with other organs within the body. For example, a brain-skin neuroendocrine axis has been well described that links the nervous system and the skin [18]. Th is critical axis is believed to be responsible for drug-related responses such as morphineinduced itch, where a subset of spinal neurons is activated by morphine, resulting in debilitating itching of the skin [19].
Another dramatic and adverse drug eff ect observed within the skin causes toxic epidermal necrolysis [20], which is usually observed after treatment with anticonvulsive drugs. Th e factors that govern susceptibility to this secondary reaction are unknown. However, toxic epidermal necrolysis starts as a dermal infl ammation that can progress to cover a large proportion of the body surface, resulting in epidermal detachment at the dermal-epidermal junction [21]. Toxic epidermal necrolysis is often accompanied by a toxic drug reaction in the liver [22], indicating the presence of a skin-liver axis. Evaluation of human skin equivalents alongside other organs is important to evaluate co-morbid toxic eff ects of new drugs or compounds.

Advantages of induced pluripotent stem cells as cell sources for construction of three-dimensional skin models
To generate a complex skin model interacting with other human organs would require multiple cell types, which would be impossible to accomplish with somatic cells due to their limited availability, even though some cancer cell lines may have suffi cient growth potential for 3D skin construction. Th e development of induced pluripotent stem cell (iPSC) technology enables us to obtain a large number of cells with unlimited growth potential from a limited number of somatic cells [23] and from noninvasive sources such as blood [24]. In addition, our recent construction of human skin equivalents entirely from iPSC-derived fi broblasts and keratinocytes [25] demonstrates the feasibility of construction of iPSCbased 3D skin models.
Generation and diff erentiation of iPSCs are timeconsuming and expensive procedures, but the unlimited growth potential of iPSCs allows investigators to share cell resources for diff erent diseases, and to screen more drug candidates for interactions [26]. iPSCs have the capacity to diff erentiate into diff erent cell types [27], which allows the establishment of multiorgan systems to model human physiological conditions, and perform toxicity studies [28]. In addition, a panel of iPSCs can be created from individuals with diff erent genetic variations to represent the diverse human population [26]. Taken together, iPSC technology can allow us to mimic clinical trials in vitro to obtain effi cacy and toxicity data that will reduce the drug attrition rate during patient-based clinical trials [26]. Th e improvement in iPSC generation and diff erentiation protocols will further enhance the cost-eff ectiveness of iPSCs for drug discovery.
iPSCs technology at the current time has noteworthy limitations, including retention of an epigenetic memory of their parental somatic cells [29], variability among iPSC lines [30] and genomic instability [31]. Th ese limitations may impair the ability of iPSCs to model human diseases for drug development. However, several solutions may be implemented to overcome such limitations, including using small molecules to reduce the eff ect of the residual epigenetic state [32]. Generation of several iPSC lines from diff erent donors can help to minimize variability [33], and avoiding direct targeting of p53 during reprogramming can serve to decrease iPSC genomic instability [34]. With a better understanding of the mechanisms underlying cell reprogramming, we anticipate that iPSCs will be produced and diff erentiated in a more controlled fashion in the future.

Variability issues in three-dimensional skin constructs
Construction of 3D skin equivalents is a complex procedure, involving biomaterials and diff erent cell types. Because iPSCs vary among diff erent lines [30] and iPSCdiff erentiated cells are heterogeneous with diff erent diff erentiation stages [26], inclusion of iPSC-derived cells may increase variability in 3D skin constructs. Several standards will need to be achieved to ensure the reliability of a 3D skin system for drug interactions with the skin. Variations in iPSC-derived cells such as fi broblasts and keratinocytes may be minimized by establishing a set of selection criteria, such as expression of cell-specifi c markers, assess ment of proliferation rate and lifespan, and establishment of gene expression signatures. Th e quality of 3D skin constructs can be guaranteed by criteria such as histo logical analysis, cell viability tests and barrier function assays [35]. Moreover, intended skin disease model systems will need to refl ect the corresponding disease state [36], such as the thickening of the epidermal layer in an in vitro psoriatic skin model [37]. In addition to standardization of protocols for 3D skin construction and cell reprogramming, strict quality control procedures will be required so that iPSC-based 3D skin models can be reliably incorporated into the drug development process.

Construction of skin equivalents using iPSC-derived keratinocytes and fi broblasts
A complex in vitro skin model has a higher potential to mimic human skin than its simple two-layer counterparts, and thus can be more useful in effi cacy and toxicity studies for drug discovery. To develop a complex skin equivalent, one strategy would be to construct a basic model (containing fi broblasts and keratinocytes) and then enhance the complexity by addition of more cell types. Due to the long-standing interest in 3D skin models, a variety of human skin equivalents have been developed using a range of natural or synthetic biomaterials [38], exogenous scaff olding biomaterials that are dispensable for 3D skin construction [39]. We aim to apply the most commonly used collagen scaff old, which would be benefi cial for adapting the resultant skin models by diff erent laboratories for validation.
Th e discovery of iPSCs has made it possible for a large number of cells to be obtained from a small number of somatic cells [23]. Moreover, use of patient-specifi c iPSCs allows enhanced study of a particular disease state [40]. We will establish skin constructs using collagen as a scaff old for iPSC-derived fi broblasts and keratinocytes (Figure 1), from both normal and diseased cells. We and others have developed protocols for establishment of both keratinocytes and fi broblasts from iPSCs, and have established human skin equivalents with one [41][42][43] or both [25] of these iPSC-derived cell types. Addition of more cell types to establish complex skin models from patient-specifi c iPSCs for drug testing would be especially benefi cial for complex diseases such as psoriasis.

Vascularization of skin equivalents using iPSC-derived endothelial cells
One of our current research goals is to establish a skin construct in which we can mimic diff erent modes of drug delivery (topical or systemic) [44] for studying drug interactions with the skin. While topical delivery can be easily replicated (since the epidermal surface of a skin construct is exposed to the air interface), mimicking systemic delivery requires the development of vascular channels within the dermal construct. We aim to establish vascular networks within the dermal portion of our skin constructs using a sacrifi cial hydrogel material to defi ne open microvascular channels within the dermal scaff old. Gelatin, agarose, or alginate hydrogels can be used as sacrifi cial materials, which can be gelled and solubilized into a defi ned microvascular pattern, without aff ecting cell viability (for example, by heating or cooling in physiologic temperature ranges for gelatin and agarose, and by applying calcium or calcium chelators for alginate). Fibroblast-seeded collagen (which constitutes the dermal scaff old) is then cast around this sacrifi cial hydrogel channel network, which is then dissolved, resulting in a dermal scaff old containing an open microchannel network (Figure 2). An endothelial barrier is established by fl owing in endothelial cells that attach to the open channel surfaces and form a monolayer [45]. Medium can be fl owed through the channels to feed the constructs, enabling us to avoid the drastic fl uctuation of pH and nutrients in the media, which would otherwise be encountered with traditional medium change practices. Th is microfabrication approach allows a complex, 3D tissue made of multiple cell types to be constructed. Importantly, the endothelial cells lining the channels can also be iPSC derived, following a recently published protocol [46].

Construction of pigmented skin equivalent using iPSC-derived melanocytes
Skin pigmentation is aff ected by many external and internal factors, such as ultraviolet, drugs, chemicals, infl ammation and hormones, with keratinocytes in the skin acting to control skin pigmentation through the regulation of melanocytes and their melanin production [47]. Adverse responses to certain drugs, including pain relievers and some psychoactive medications, can manifest with hyperpigmentation of the skin [47]. Inclusion of melanocytes into skin constructs will therefore enable us to utilize variations in skin pigmentation as a readout for pharmaceutical screening of drugs that may cause sunlight-induced darkening of the skin.
Pigmented skin equivalents have been previously established by seeding melanocytes along with keratino cytes onto a dermal skin construct [48]. Melanocytes and keratinocytes are able to self-organize, resulting in the production of pigmented skin. However, in keeping with the theme of developing skin constructs using cells from a single individual for patient-specifi c drug targeting, we aim to establish human skin constructs containing iPSCderived melanocytes, which originate from the same donor of the fi broblasts, endothelial and epithelial cells within the construct. Progress within the stem cell fi eld is rapid, and recently a protocol was published reporting the diff er entiation of iPSCs into melanocytes [49]. We will utilize this protocol to incorporate iPSC-derived melanocytes into our patient-specifi c skin equivalents.
Such constructs can then be exposed to UVB to mimic the environmental eff ect of sunlight.

Integration of the skin microphysiological system with other organs
Recently, there has been growing interest in human-ona-chip systems, in which diff erent human organs such as the liver, heart and lung are connected in a microphysiological system [50]. By incorporating a vascular network into our skin equivalents, the equivalents can be directly plugged into a portal that is connected to the other organs through a microfl uidic system (Figure 3), similar to a confi guration described previously [51]. Questions now relate to the longevity of skin equivalents, which usually show deterioration within the dermal compartment of the equivalent approximately 2 weeks after culturing at the air-liquid interface, leaving only a small amount of time for postproduction manipulation including drug screen ing. Th is deterioration is usually due to degradation of collagen in the construct [52]. However, the life of human skin equivalents can be prolonged up to 15 weeks, using approaches such as reinforcing the scaff olds in the dermal compartment [53] and/or inclusion of protease inhibitors in the culture medium [54]. We will take measures to ensure that 3D skin constructs have suffi cient lifespan for drug development studies. Additionally, the development of a universal medium that will fl ow through all systems and support growth of multiple organs is required.
Th e skin interacts with the peripheral sensory nervous system, the autonomic nervous system, and the central nervous system in the brain-skin neuroendocrine axis [18]. Activation of the cutaneous nervous system is also involved in various skin diseases such as psoriasis [55]. We are working to innervate our skin model so we are able to mimic the interactions between the skin and the brain, conferring an enhanced sensitivity for pharma ceutical screening. To this end, Schwann cells and peripheral neurons can be diff erentiated from iPSC-derived neural crest stem cells [56] and then integrated into 3D skin constructs [14,57].

Skin microphysiological system as a tool for studying drug interactions in the skin
Human skin equivalents have been widely used in skin disease modeling [36] and pharmacological studies [58]. Diff erent 3D skin disease models display diff erent characteristics associated with corresponding diseases [36]. Th e alleviation of disease-related symp toms, such as decreased expression of cytokeratin 16, interleukin-8 and tumor necrosis factor alpha in a psoriatic model [59], for example, may be an indication of drug effi cacy. For hyperpigmentation-related drugs, melanin content can be quantifi ed as described elsewhere [60]. Previous Human somatic cells such as skin fi broblasts can be reprogrammed into induced pluripotent stem cells (iPSCs), from which various cell types (including fi broblasts and keratinocytes) can be diff erentiated. Subsequently, the iPSC-derived cells can be combined with collagen to reconstitute three-dimensional (3D) skin constructs for drug testing.

iPSCs 3D Skin Drug testing
Reprogramming Differentiation 3D construction (with collagen) Figure 2. Schematic diagram of microfl uidic channels fabricated with sacrifi cial alginate. A sacrifi cial alginate construct will be fabricated using a polydimethylsiloxane mold and embedded in a collagen gel. Subsequent removal of alginate will leave behind a void that can serve as channels. studies with diff erent compounds on in vitro 3D skin models have revealed several sensitive readouts, including disruption of structural integrity, loss of barrier function, reduced cell viability and elevated levels of interleukin-1α in the medium [61]. Not surprisingly, release of interleukin-1α from keratinocytes is a primary event in the skin defense mechanism in response to damage [62]. We aim to evaluate these parameters for toxicity assays. Establishment of reliable drug screening protocols using skin equivalents will generate valuable effi cacy and toxicity data to aid in drug discovery and screening.

Future directions
Skin is a highly complex organ containing hair follicles, sweat glands, and nervous, lymphatic and vascular systems, among many other cell types within the dermis and epidermis. Current skin models are very basic, and only replicate the two main compartments of human skin, keratinocytes and fi broblasts. Our research plan is to gradually incorporate diff erent cell types into our human skin equivalents, in order to better replicate a functional human skin. While we are starting by introducing pigmentation and blood vessels, our longterm goals are to establish a highly complex skin construct that includes hair follicles, nerves, and immune cells to reliably mimic human skin.
We will aim to generate hair follicles in skin constructs using iPSC-derived papilla cells. Dermal papilla cells are specialized fi broblasts located at the base of the hair follicle and are crucial for hair follicle development. Cultured dermal papilla cells have previously been used to instruct hair growth in recipient tissues using both rodent and human cells [63,64]. However, to date, dermal papilla cells have not been successfully diff erentiated from iPSCs. Development of a robust procedure to diff erentiate iPSC-derived papilla cells from iPSCs will provide us with an un precedented opportunity to utilize iPSCs for regenerative medicine.
To generate skin constructs containing immune components, precursors of dendritic cells can be incorporated into constructs, to then be diff erentiated into Langerhans cells in the epidermis and into dendritic cells in the dermis [13]. T cells can be derived from iPSCs with a published protocol [65], and procedures to co-culture T cells with keratinocytes [66] or fi broblasts [67] can be adapted to integrate iPSC-derived T cells into a complex 3D skin system.
In closing, incorporation of diff erent iPSC-derived cell types into vascularized human skin equivalents will enable us to replicate complex skin disorders in vitro. Launching human-on-a-chip systems will enable us to better mimic human physiological conditions and to more effi ciently assess the effi cacy and toxicity of drug candidates to treat human disease. Figure 3. Schematic diagram of a multiorgan system. Diff erent organs can be integrated as indicated. In this system, only a single pump will be needed to recirculate a common medium through diff erent organs. Eight organs are shown in this system, but the confi guration can be modifi ed to accommodate diff erent number of human organs.

Medium reservoir
Pump Gut Marrow Kidney Heart Lung Brain Liver Skin Abbreviations 3D, three-dimensional; iPSC, induced pluripotent stem cell.

Competing interests
The authors declare that they have no competing interests.