Engineering human skin model innervated with itch sensory neuron‐like cells differentiated from induced pluripotent stem cells

Abstract Atopic dermatitis (AD), driven by interleukins (IL‐4/IL‐13), is a chronic inflammatory skin disease characterized by intensive pruritus. However, it is unclear how immune signaling and sensory response pathways cross talk with each other. We differentiated itch sensory neuron‐like cells (ISNLCs) from iPSC lines. These ISNLCs displayed neural markers and action potentials and responded specifically to itch‐specific stimuli. These ISNLCs expressed receptors specific for IL‐4/IL‐13 and were activated directly by the two cytokines. We successfully innervated these ISNLCs into full thickness human skin constructs. These innervated skin grafts can be used in clinical applications such as wound healing. Moreover, the availability of such innervated skin models will be valuable to develop drugs to treat skin diseases such as AD.


| INTRODUCTION
Human skin is a complex structure containing many cell types including sensory neurons, which mediate various sensation responses, such as pain and itch. Cutaneous sensory neurons have their cell bodies located in the dorsal root ganglia (DRG), with the nerve fibers branching into the dermis and moving upward into the epidermis. 1 Sensory neurons are involved in neurologic disorders, immune responses and the regulation of aging and metabolism. 2 Understanding the molecular mechanisms underlying sensation will benefit the development of strategies to treat skin diseases characterized by pain and itch.
Atopic dermatitis (AD) is the most common chronic skin disease, afflicting 2%-3% people in the adult population and 10%-30% in infants, thus causing a global disease burden. 3 AD is both highly pruritic (itchy) and inflammatory, 4 and skin inflammation can cause itch. 5 Type 2 cytokines are well-established mediators of skin inflammation in AD, 6,7 specifically the cytokine (IL-4/IL-13) driven AD pathogenesis. 8 Notably, itch sensory neurons express receptors (IL-4Rα and IL-13Rα1) shared by IL-4/IL-13, and can be directly activated by the two type 2 cytokines, 9 indicating a link between itch sensation and immune response. However, it is not known how the neuroimmune interactions are mediated by sensory neurons.
Human neurons are not readily available, which necessitates the use of animal models as surrogates. 10 An organotypic skin model was developed that contained fibroblasts, keratinocytes, and neuronal cells, 11 reminiscent of in vivo innervated skin with neurites branching through the dermis in proximity to keratinocytes. However, these neurons were derived from porcine DRG and the mechanisms of human nocioception and prunitoception are different from those of animals. 12,13 Several groups have attempted to circumvent the limited availability of human sensory neurons by conversion of fibroblasts to neurons using defined factors, 14,15 or by differentiation of human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs) to sensory neurons. [16][17][18][19][20][21][22] However, no itch-specific sensory neurons were produced. Thus, technical challenges in obtaining a sufficient number of sensory neurons precluded the development of a reliable model to investigate neuroimmune interactions with human cells.
To overcome this obstacle, here, we obtained iPSC-derived sensory neurons (iNCs) which displayed action potentials, responded to a subset of itch-specific stimuli, and demonstrated the ability to innervate human skin. The availability of functional itch-specific iNCs will be of great value for delineating neuroimmune crosstalk as well as modeling AD to develop pharmaceutical agents.

| Neurons differentiated from iPSCs exhibit molecular and morphological hallmarks of sensory neurons
To obtain human sensory neurons, we sought to use iPSC technology because iPSCs have the potential to differentiate into many cell lineages, including sensory neurons. To this end, we obtained iPSCs using different metrologies, including a retroviral system, 23 episomal vectors to reprogram human fibroblasts into integration-free iPSCs, 24 as well as feeder-free culture conditions. 25 Next, we screened different protocols to differentiate iPSCs into sensory neurons. 2.2 | Human iNCs exhibit physiological properties of mature neurons Next, we performed whole cell patch-clamp recordings to determine whether human iNCs were sufficiently mature to display the physiological properties (action potentials) of mature neurons. We found that the iNCs differentiated from our iPSCs with this method rarely showed any action potentials (Figure 3a) (= 6), suggesting that further optimization was needed.
Recently, a neuronal medium was developed to support activity of human neurons. 26 We sought to improve the physiological properties with this approach by substituting the DMEM/F12 with this medium. Patch-clamp recordings showed that some of the iNCs that were more matured with the modified medium but displayed no action potential or an abortive action potential (no action potential in 9 cells, abortive action potentials in 14 cells as shown in Figure 3b).
These cells showed no response to high KCl, mustard oil, capsaicin, BAM8-22, or menthol, but displayed small responses to chloroquine.
F I G U R E 3 Whole cell patch-clamp recordings of iNCs. iNCs were placed on coverslips and the current changes of the cells were recorded while the voltage remained constant. The cells were allowed to mature in a previously described protocol (a), 17 modified media (b), or keratinocyte coculture (c and d). The modified media was formulated by using BrainPhis (Stem Cell Technologies) 26 to replace the DMEM/F12 in the published medium 17 and keratinocyte coculture was conducted in the modified medium for at least 2 weeks F I G U R E 4 iNC calcium responses to sensational reagents. iNCs were treated with 10 μM of mouse (mPAR-2) or human (hPAR-2) protease-activated receptor 2 peptides: SLIGRL-NH 3 (mPAR-2) and SLIGKV-NH 3 (hPAR-2). Also, histamine (100 μM) and capsaicin (1 μM) were administered at different time points as indicated and calcium responses were monitored by Fura2 Some of the cells showed strong response to histamine, mouse, and human PAR-2, all of which are itch specific agonists (data not shown).

| iNCs are itch-specific sensory neuronlike cells
Human sensory neurons have distinct subsets that mediate different sensory responses such as heat, cold, pain and itch. We tested whether the iNCs responded to non-pruritogenic compounds (such as capsaicin, menthol, and mustard oil) as well as pruritogenes (such as histamine, receptors. Proteases like trypsin or tryptase can activate neuronal and/or keratinocyte-derived PAR-2, which subsequently leads to itch. 27

| iNCs migrate though the dermis toward the epidermis
Human skin contains many types of innervated sensory neurons. We postulated that it might be possible to construct an innervated 3D skin model with our iPSC-derived sensory neurons. To accomplish this, we dissociated iNCs and placed the cells beneath mature 3D constructs that were generated as we described previously, 28,29 containing fibroblasts in the dermis and keratinocytes in the epidermis. We placed iNCS outside the dermis and opposite to the epidermis. Two weeks after coculture, we observed neuron-like cells in the dermis, indicating iNCs migrated into the dermal compartment toward the epidermis. The results showed that iNCs successfully migrated across the dermal compartment to reside in close contact of keratinocytes ( Figure 5) (N = 6 dishes).

| iNCs respond to IL-4/IL-13 treatment
After verifying that the iNCs were physiologically reminiscent of itch sensory neurons, we next tested the potential of the cells to recapitulate features of AD. AD is an inflammatory disease of the skin (dermatitis), which causes the skin to itch, swell and crack. Recently, it was reported that thymic stromal lymphopoietin (TSLP), a keratinocytesecreted cytokine, can link inflammation with itch. 30 In fact, both human and mouse DRG contain skin-innervating sensory neurons, express IL-4/IL-13 specific receptors (Il4rα and Il13rα1), and can be directly activated by the two AD driver cytokines on itch-sensory pathways. 9 Therefore, we examined whether our itch-specific iNCs would exhibit the same properties. We detected expression of Il4rα and Il13rα1, but not Il5rα (Figure 6a) (N = 3), which is similar to the expression pattern previously reported in cultured human DRG. 9 However, we did not detect expression of Il31rα (Figure 6a), which is

| DISCUSSION
We successfully differentiated human iPSCs into cells that recapitulated the morphology and characteristic markers of peripheral sensory neurons. These iPSC-derived neurons (iNCs) fired action potentials and responded specifically to two types of chemical mimics of pruritogens, histamine and PAR2, indicative of functional properties of mature itchspecific sensory neurons. These itch-specific iNCs were responsive to AD driver cytokines (IL4/IL-13) as indicated by calcium imaging.
The iNCs displayed the expression of the receptors shared by the AD driver cytokines (IL-4/IL-13) and could be directly activated to produce a calcium response after the treatment with the two cytokines, which agrees with previous findings. 9 However, our RT-PCR did not detect any expression of the receptor (il31rα). It is likely that il31rα is only expressed in a subset of sensory neurons that was not present in our iNCs, but in the DRG used for the previous study. 9 Moreover, the iNCs are capable of innervating an in vitro 3D skin model. Recently sensory neurons were differentiated from iPSCs with a procedure similar to the one we used here. Their neurons were able to innervate 3D skin constructs, only in the presence of iPSC-derived Schwann cells. 22 Our iNCs had the ability to innervate 3D skin without the help of Schwann cells. It is likely that the subtle discrepancy in 3D skin construction and/or in iNC differentiation procedure between ours and the published protocol 22 led to the difference in iNC innervation capability.
Protease-activated receptors (PARs) are a group of four unique G-protein-coupled receptors. These receptors are self-activated following specific proteolytic cleavage of their extracellular domains to expose a new amino terminus that acts as a ligand to activate PARs. 31 Various extracellular proteases such as serine and cysteine proteases mediate PAR-2 proteolysis. Microbial agents such as house dust mites, cockroaches, and bacteria secrete proteases to activate PAR-2 and thus elicit itch. 27 PAR-2 activation likely stimulates itch pathways implicated in the pathophysiology of pruritus in AD. 32,33 Moreover, PAR-2 antagonists and antibodies can alleviate itch-related behavior in mice. 34 These findings indicate that PAR-2 may be a target molecule for the therapeutic treatment of AD. 35 Our sensory neurons are responsive to PAR-2 as well as to the AD driver cytokines (IL4/IL-13).
Organotypic skin models are responsive to IL-4 and IL-13, 36 The in vivo dynamics of [Ca 2+ ] i in epidermal keratinocytes can be assessed with two-photon microscopy. 37 We postulate that IL-4/IL-13 treatment can perturb keratinocyte calcium dynamics. Thus, further IL-4/ IL-13 experiments can be performed on organotypic skin models. AD skin displays increased innervation and pruritus is a prevalent symptom of the disease. 36 Therefore, a 3D skin model innervated with itch-specific sensory neurons would be advantageous to recapitulate AD pathophysiology and develop drugs to treat the disease. To this end, itch-specific iNCs can be innervated into 3D skin and calcium imaging and/or AD-specific markers can be used to assess AD pathogenesis induced by IL-4/IL-13 or Th2 cells. Drugs can be selected based on their ability to reduce AD markers on the innervated skin model. Thus, the availability of such cells will be valuable to model AD and develop therapeutic agents to treat the disease.

| Generation of iPS cells
We initially used a retroviral system to generate iPSCs with a feeder system. 23 Then, to avoid the detrimental effect of random genomic F I G U R E 6 iNC expression of receptors for selected cytokines and calcium response to IL-4/IL-13 treatment. RNA was isolated from iNCs and subjected for RT-PCR analysis (a) using published primer sequences 9 ( Table 2). Cells were treated with a combination of IL-4/IL-13 (20 μM/ml each) and imaged at different time points (as indicated) with a calcium indicator (Fura4) (b) integration by retrovirus, we evolved to Integration-free iPSC reprogramming using the transfection of integration-free episomal plasmids containing OCT3/4, SOX2, KLF4, and L-Myc (Addgene plasmids 27077, 27078, 27080, 37624). Vectors were electroporated into fibroblasts with the Amaxa Nucleofector 2 Device (Lonza) with a NHDF Nucleofector kit. 24,38,39 Recently, we combined our integration-free iPSC reprograming methodology with a feeder-free culture system to avoid the variability associated with feeders. 25 The pluripotency of iPSCs was confirmed by immunostaining of stem cell markers and the capacity to form teratomas in nude mice. [23][24][25]

| Construction of innervated 3D skin models
3D skin was constructed following a previously described protocol. 25 Briefly, a type I collagen matrix was used to embed normal human fibroblasts and the polymerized cell-containing gel was incubated on polyethylene terephthalate membranes (BD Biosciences) for 5-7 days. Then, normal human keratinocytes were seeded onto the matrix, and incubated for additional 6 days, before the composite culture was raised to the air-liquid interface for 7 days to induce epidermal differentiation. For innervation, iNCs were seeded on a Matrigel-coated transwell membrane before a skin construct was placed atop. Medium was fed from below for 2 weeks before the constructs were harvested for analysis.

| Immunofluorescent imaging
For cultured iNCs, cells on cover slips were fixed with 4% paraformaldehyde at room temperature for 20 min before blocking with 1.5% fish skin gelatin (Sigma) in 0.1% Triton X-100/phosphate-buffered saline (PBS) for 1 h. Then, the fixed cells were incubated with primary antibodies at room temperature for additional 1 h. After rinsing with PBS, secondary antibodies were applied to the cells and incubated at room temperature for 1 h. VECTASHIELD Mounting Medium containing DAPI (Vector Laboratories) was used for nuclear staining before visualization with a confocal microscopy (Carl Zeiss LSM 5 EXCITER). All antibodies are listed in Table 3. For formalin-fixed, paraffin wax-embedded tissues, samples were sliced onto poly-Llysine-coated slides and dried overnight at 55 C. Subsequently, the slides were dewaxed in xylene and rehydrated through a graduated ethanol series (100%, 95%, 70%) and distilled water (dH 2 O) before antigen retrieval at 97 C in 10 mM sodium citrate buffer (pH 6.0) for 30 min. The blocking and incubations with primary and secondary antibodies were similar to the procedure described above.

DATA AVAILABILITY STATEMENT
Data available on request from the authors.