Human Skin Equivalents Demonstrate Need for Neuro-Immuno-Cutaneous System

2.1 System Design, Timing, and Sample Collection

Model construction methods are described in Cell Culture Conditions, Adipose Surgery, and Tissue Model Construction sections. The most complex HSE, SC3, contains all model components: hypodermis, dermis, epidermis, and a neuronal component using only primary, human cells or tissue (Figure 1). To assemble the model, first the hypodermis is created (described in Adipose Surgery section). We have previously defined that this lipoaspirate contains many cell populations, including adipocytes, pre-adipocytes, endothelial cells, smooth muscle pericytes, and immune cells (likely macrophages/monocytes).6, 21 Next, a silk-collagen protein composite hydrogel was used as the dermal material, containing primary human fibroblasts. The silk-collagen hydrogel is resistant to degradation and contraction, and combines the Arg-Gly-Asp (RGD) domains of collagen with the biocompatibility and tunability of silk.6, 22-24 Primary human keratinocytes are then seeded on the dermal gel and allowed to epithelialize for one week prior to the experiment start date. On the experiment start date, the entire model is constructed (as described in Tissue Model Construction section), with the hiNSC-containing collagen gel coating the hypodermis, which is placed underneath the dermal/epidermal gels, raised to air–liquid interface (ALI), and maintained over the experiment duration. Samples were collected for protein analysis (cell culture supernatant, Cell-Specific Assays section) at D1, D5, D7; samples for imaging were fixed (Histology and Immunohistochemistry section) at D1 and D7; samples for RNASeq were collected (RNA Extraction section) at D1 and D7.

2.2 Histological and Immunohistological Imaging

Of the four HSE models constructed, only one study group contains all four cell/tissue components (epidermis, dermis, hypodermis, neural): SC3. All layers of SC3 are presented with hematoxylin and eosin (H&E) (Figure 2A,B) or histology and immunohistochemistry (IHC) staining (Figure 2C). IHC was performed to identify key cell type(s) of each layer. The epidermis was characterized by K10, a keratinocyte marker. The dermis was characterized with vimentin for fibroblasts. The hypodermis was characterized in four ways: to identify adipose tissue morphology (Figure 2B), adiponectin production by adipocytes in the hypodermis (Figure 2C), inherent-immune cells from the adipose tissue by CD68 staining (for macrophages/monocytes), and to visualize hiNSCs which surrounded the hypodermis with TUJ1 and Vybrant DiD Cell-Labeling Solution (DiD) staining.

2.3 Cell-Specific Protein Secretion

To elucidate cell-specific effects with respect to sample group, key markers were selected for further analyses which are related to specific cell functions (Table 1). Proteins produced by keratinocytes, fibroblasts, neurons, adipocytes/adipose tissue, and immune cells (macrophages, monocytes), or a combination thereof, were selected.

Differences in protein secretion were observed within and between groups (Figure 3A–F, Data Availability contains “between groups” significance information). TSLP secretion was highest from the skin explant and the fat (in vitro hypodermis). TSLP secretion from HSE groups was lower than the skin explant and fat, and was similar with time across all HSEs (Figure 3A). TSLP, FAP, and glycerol secretion significantly increased with time in the skin explant (Figure 3A,B,D). FAP secretion was highest in the skin explant, but lower and similar with time in the Fat and all HSE study groups (Figure 3B). Nerve growth factor (NGF) secretion was minimal from the skin explants and fat group (Figure 3C). Differences were noted between groups with and without neurons added (i.e., SC1 vs SC2, SC3), where groups with neurons at D1 secreted significantly more NGF than SC1 (Figure 3C) at all timepoints (P < 0.001). SC4, which does not have neurons added, but does contain hypodermis, had similar secretion levels to SC2 and SC3.

Glycerol secretion was highest in the skin explant, and increased with time, was negligible in SC1 and SC2 (without hypodermis) and was low but detectable in the fat, SC3, and SC4 study groups, which all contain an in vitro hypodermis (Figure 3D). MCSF1 secretion was lowest from the skin explant and fat group but was highest in SC1 and significantly increased with time (Figure 3E). All other HSE groups (SC2, SC3, and SC4) also secreted MCSF1, which increased with time, but SC1 secretion was almost three times higher than all other HSE groups. GMCSF secretion was detected in the skin explant, not in the fat, and was highest in SC1 (increasing with time), and low but stable secretion from all other HSE groups (Figure 3F).

2.4 Multigroup Analysis of Gene Expression via RNA-Seq

To further characterize the differences between the HSE groups, RNA-Seq was performed on all four study groups. As a visualization tool, Reactome pathway browser was used to view the most relevant genes from the RNA-Seq dataset (q-value, false discovery rate (FDR) = 0.1) (Figure 4). Of note, the immune system, signal transduction, protein metabolism, and gene expression (transcription) were among the most dominant pathways visualized by Reactome for the HSE systems overall.

Through Gene Set Enrichment Analysis (GSEA), specific pathways and the related genes were identified through application of the hallmark gene set with respect to HSE sample type (Figure 5A).40, 41 The complement pathway was the most upregulated in SC4, then SC3, and downregulated in SC2 and SC1 (both without hypodermis). Similar trends were observed for coagulation, glycolysis, adipogenesis, epithelial–mesenchymal transition, and apoptosis pathways, which overall were upregulated in SC3 and SC4 but downregulated in SC1 and SC2.

From the entire dataset at q = 0.1, all hallmark pathways with respect to number of matches per pathway were presented (Figure 5B). Hypoxia, complement pathway, and the epithelial–mesenchymal transition had the highest number of matches via this analysis tool. Other key pathways included the inflammatory response, angiogenesis, and Notch, KRAS, Interleukin-2 (IL2)/Signal transducer and activator of transcription 5 (STAT5), and mammalian target of rapamycin complex 1 (MTORC1) signaling.

GO gene sets of specific interest to the HSE systems (hallmark adipogenesis, immune system process, inflammatory response, nervous system development, and sensory perception) were applied to the dataset at q = 0.1 (Figure 6A). Using this approach, the HSE groups and their key processes could be visualized. For example, SC3 and SC4 (both containing hypodermis) upregulate several adipogenesis-specific genes Enoyl-CoA Hydratase, Short Chain 1 (ECHS1), Perilipin 2 (PLIN2), and Adiponectin, C1Q And Collagen Domain Containing (ADIPOQ), whereas these genes are downregulated in SC1 and SC2 (both without hypodermis). SC1, which contains only keratinocytes and fibroblasts, highly upregulates several genes related to skin development Hornerin (HRNR), Loricrin (LOR), and Keratin-71 (KRT71) which were downregulated in all other study groups. Other genes related to skin development SRY-Box 18 (SOX18) and Insulin Like Growth Factor Binding Protein 5 (IGFBP5) were upregulated in SC3 and SC4 but downregulated in SC1 and SC2. Certain genes related to sensory perception Myosin IIIA (MYO3A), ATP Binding Cassette Subfamily A Member 4 (ABCA4), and Myosin VIIA (MYO7A) were only upregulated in SC1. Two genes related to immune system processes and inflammatory responses were upregulated in SC3 and SC4, but downregulated in SC1 and SC2 (BCL2 Interacting Protein 3 (BNIP3) and Complement-2 (C2)), and Orosomucoid 1 (ORM1) and Forkhead Box P3 (FOXP3) were upregulated only in SC4. SC1 upregulated three genes related to immune system processes and inflammatory responses: Apolipoprotein A4 (APOA4), Growth Hormone Secretagogue Receptor (GHSR), and Interleukin 31 Receptor A (IL31RA). The only gene for nervous system development upregulated in both SC2 and SC3 (both groups contain hiNSCs) was SRY-Box 11 (SOX11). SC3 also upregulated Hes Related Family BHLH Transcription Factor with YRPW Motif-Like (HEYL) and Contactin 4 (CNTN4), downregulated Cholinergic Receptor Muscarinic 2 (CHRM2) (which SC1 and SC2 upregulated), and downregulated Protocadherin Alpha 5 (PCDHA5) and Nuclear Receptor Subfamily 2 Group E Member 1 (NR2E1) which only SC1 upregulated.

With the GO sets applied, principal component analysis (PCA) confirms in 2D (Figure 6B) and 3D (Figure 6C) that all groups were distinct from one another. SC3 and SC4 are the most similar groups in this instance, and SC1 is the least similar from all other groups. SC3 and SC2 are more similar than SC2 and SC4. SC1 is more related to SC2 than any other group. Time points for all samples aligned with their respective group (Figure 6B,C).

2.5 Two-Group Analysis of Gene Expression via RNA-Seq

Two-group analysis was performed to identify key genes of interest with respect to sample group (Figure 7A). SC1 had 222 genes uniquely up/downregulated, SC2 had 31, and both SC3 and SC4 had 23 genes of interest total. Selected genes from each sample group were selected for further analysis (Figure 7B), which are defined with further detail in Table 2: Bet1 Golgi Vesicular Membrane Trafficking Protein Like (BET1L), Contactin 4 (CNTN4), Chemokine (C-X-C Motif) Receptor 4 (CXCR4), Fibroblast Growth Factor-19 (FGF19), Keratin-71 (KRT71), Macrophage Stimulating 1 Like (MST1P9), Orosomucoid 1 (ORM1), and Sp4 Transcription Factor (SP4).

Additional genes identified through two-group comparison are described in further detail in Table 2.

3.1 Distinctions between HSE Groups

In our previous paper, we discussed the utility of a HSE with neurons and immune system components as a physiologically relevant, yet complex in vitro tissue model system.6 In the present work, the cell culture conditions were modified, and the experimental groups were expanded to better discern the contributions of specific cell types to the skin tissue functions. Of note, keratinocytes, fibroblasts, and hiNSC cell seeding densities were increased to enhance cell–cell communication, and the HSE model size was reduced to compensate for the lower contraction of silk-collagen hydrogels (vs collagen-only hydrogels).6

A parametric system was designed to analyze the HSEs (SC1, SC2, SC3, and SC4) (Figure 1, Table 3). SC1, a keratinocyte and fibroblast model, is most similar to HSEs commonly used today, containing only these cell types, but the biomaterial scaffold is a silk fibroin-collagen hydrogel. Tissue models using only fibroblasts and keratinocytes are used for numerous applications: briefly, as an emerging technique, they can be 3D printed layer-by-layer with precision as reconstructed skin,1-3 can support the study of 3D printing of large-scale objects,4 added to biomaterial substrates as tissue-engineered alternatives to human skin for study of pharmacological, biological, or mechanical processes,5 utilized as a tissue-engineered substitutes to compare suitability versus human skin for permeability applications which necessitates consideration to the complexity of the barrier layer.6-8 The application of these systems extends to many areas of study, and while there have been many promising results to date, they rely on two cell types, and therefore may be omitting key factors relevant to skin equivalent models.

SC2, which includes hiNSCs in addition to keratinocytes and fibroblasts, is more similar to other innervation tissue models that include neurons.1, 2 Through research of tissue models of this type, keratinocytes were found to play a key role in neural outgrowth, which can be used to understand skin conditions such as eczema.1 Further, neural cells (porcine dorsal root ganglia) influenced epidermal morphology, a key to skin homeostasis. This tissue model could potentially serve as a tool to observe interactions between nerve and skin in normal versus disease states.2 However, there are known interactions between neurons with other cell types, especially immune cells, with the nervous system in skin that are involved in numerous conditions, including psoriasis and atopic dermatitis.14 Tissue models that only include keratinocytes/fibroblasts and neural cells may therefore be lacking important components to understand these diseases.

SC4 is composed of keratinocytes, fibroblasts, and a hypodermis, which is similar to previously reported constructs, however the hypodermis in this study was composed of human adipose-derived stem cells (hASCs).3 The objective of that model was to use the HSE as an in vitro tool to study obesity or diabetes.3 In the current study, exchanging hASCs for lipoaspirate alters the tissue model and may be better suited to study obesity, for example, because the entire adipose tissue is exploited, including the immune component. The silk scaffold with lipoaspirate was used as a patient-specific model.16 The main advantage of using lipoaspirate, as opposed to a single-cell source such as hASCs, is that numerous primary, human cell types are contributed from the lipoaspirate (adipocytes, preadipocytes, endothelial cells, and smooth muscle cells, immune cells including macrophages)17, 18 which results in a more physiologically relevant system than a stem-cell-only model.43 However, the disadvantages include patient variability16-18 and variability in cell-type composition which may differ from patient-to-patient, or even in different samples from the same patient source.43

We previously reported that this system (SC4) had complex secretion of pro-inflammatory cytokines, several of which have been implicated in obesity research.6, 44, 45 However, as SC4 lacked a nervous component it may be missing key factors to serve as an obesity or skin tissue model, as the nervous system has a key role not only in cutaneous inflammation but also in adipose tissue functions.46 For instance, adipocytes are known to secrete pigment epithelium-derived factor, an adipokine with neurotrophic function, at very high amounts, as a potential protein of interest for diabetes and obesity research. However, the link to the nervous system for this protein has not been fully explored.47 Also, it has been reported that sympathetic neurons can influence adipocytes by neuropeptide Y (NPY) secretion, and that neurons are in turn also influenced by adipocyte-derived soluble factor or insulin secretion from adipocytes in coculture.48

SC3 is composed of keratinocytes, fibroblasts, hiNSCs, and a hypodermis to contribute inherent immune cells from the lipoaspirate6, 49 (Figure 2C, and Figure S1, Supporting Information). To our knowledge, no HSE model exists that combines all of these components beyond our initial publication.6 We hypothesize that this model is the most relevant to studying the neuro-immuno-cutaneous system, which is known to be important for skin homeostasis and many key functions of the skin.12, 13 Of all the in vitro tissue models explored in this work, SC3 would be most representative of the cell types and tissue layers of natural human skin, including the epidermis, dermis, hypodermis/immune, and neural components.8

Each HSE group studied (SC1, SC2, SC3, SC4) has potential to contribute useful information to skin research, as each has a focus: skin only (SC1), skin/nerve interactions (SC2), skin/nerve/adipose/immune interactions (SC3), and skin/adipose/immune interactions (SC4), respectively. When directly compared, the differences in functions between the study groups were elucidated based on mRNA and protein analyses. Through the improvements in HSEs presented here, new information and insights into the respective components of the skin (e.g., neural, adipose, and immune) can be gained to better understand the neuro-immuno-cutaneous system in the context of homeostasis, damage, disease, and regeneration.

3.2 Protein Analysis

TSLP is primarily produced by epithelial and dendritic cells, and is understood to be increased during inflammation.25-27 The skin explant secreted the highest amount of TSLP which increased with time. The fat scaffold had higher levels of TSLP compared to the HSEs, and this also increased with time, whereas the HSEs secreted low but stable levels of the protein suggesting that the keratinocytes in the HSEs had some activity but may have been less active than other groups.

FAP, a fibroblast-associated protein,50 is typically expressed by stromal fibroblasts of inflamed tissue when compared to normal tissue.28 The skin explant group had elevated levels of both TSLP and FAP compared to the in vitro HSEs. The FAP level in the HSEs overall increased with time, suggesting that while levels were lower compared to an explant, that the fibroblasts within the HSEs were increasingly active with time.

Glycerol concentration was also highest from the skin explant. Glycerol is known to exist on the skin surface for hydration and barrier properties,51, 52 and is an important molecule for adipose metabolism.33 SC3 and SC4, both with an in vitro hypodermis, had a glycerol concentration equivalent to the fat group (in vitro hypodermis alone), suggesting that these groups can maintain the function of the adipose component. SC1 and SC2, which do not have a hypodermis component, do not secrete glycerol.

NGF can be produced by a variety of skin cells, including keratinocytes,31, 53 fibroblasts,32, 54 and neurons;29 adipocytes can also produce NGF.55 NGF was injected into the subepidermal space of all HSEs to encourage upward migration of hiNSCs. However, from imaging (Figure 2C), hiNSCs remained in the location where they were initially seeded. Directing migration of the hiNSCs should be considered in subsequent tissue models. Despite equal amounts of NGF being injected into the HSEs, there were differences in secretion between groups. The half-life of human recombinant NGF has been reported as ≈45 min or less; we therefore speculate that the NGF was produced by the HSEs, and was not represented by residual NGF following the initial injection.56, 57 Samples SC2 and SC3 both have added hiNSCs, but SC4 does not, however each group has equivalent starting concentrations of NGF. The NGF concentration decreased with time from all HSE groups, with SC4 having the highest remaining concentration by D7, which may be due to contributions from the adipose tissue. SC1, with no hiNSCs or adipose tissue, had the lowest concentration of NGF with time, initially about half that of the other HSE groups. Release of NGF from the skin explant and fat group was negligible. While HSEs were injected with NGF at D0, the differences noted between groups suggested that there was an added contribution from both hiNSCs and the adipose tissue, but only when interfaced with in vitro skin tissue.

Proliferating fibroblasts can produce MCSF1,36 which could be the main reason that SC1, composed of only fibroblasts and keratinocytes, had the highest MCSF1 concentration of all sample groups, and especially for HSE groups. Since SC1 has only two cell types and layers, compared to up to four tissue layers with numerous cell types (from the adipose tissue), the fibroblasts in SC1's dermal layer may be more proliferative, and in turn, secrete more MCSF1.

Keratinocytes are known to produce GM-CSF, which is multifunctional and known to be involved in dendritic cell formation, inflammation, and production/differentiation of immune cells including macrophages and neutrophils.58-62 Fibroblasts can also produce GM-CSF.34 SC1 had the highest concentration of GM-CSF, which may be an effect from the fibroblasts and keratinocytes combined. SC1 having the highest signal could also be a result of lack of signaling from other skin cell types that are present in the other HSEs which may be regulating their GM-CSF levels.

Overall, from all proteins studied with respect to the HSEs, all groups functioned as expected (e.g., all HSEs expressed similar levels of TSLP and FAP, SC1 had the lowest NGF level, samples without hypodermis had negligible glycerol levels) except for SC1, which had elevated levels of GM-CSF and MCSF1, unlike any other HSE study group. As previously stated, many of the proteins explored can be secreted by multiple cell types of the skin utilized in this study. This approach has limitations that emerging experimental techniques (such as combinations of proteomics/genomics readouts from single cells) are attempting to address by providing fully integrated information on cell populations.63, 64 Further, it is possible that differences between sample groups could be due to cell count or tissue density, especially in the case of the human explant compared to HSEs. This presents a limitation in direct comparison to human explants and to the understanding of whether the HSE tissues were activated (in the case of MCSF1, for example) or not. However, we speculate that these HSE tissue models could be a very useful system with which to study inflammation. For example, the high values of MCSF1 from SC1 (the model most similar to many commercial skin models), when compared to the other study groups, could identify potential overexpression of inflammatory signals from simpler HSEs versus more complex ones since MCSF1 secretion decreases as more cell types and layers are added. As a drug screening tool, these models could provide an upper and lower bound for inflammatory signals from in vitro models which could be translated for downstream applications.

3.3 Multigroup Analysis

Numerous key pathways were first visualized (Figure 4) then identified (Figure 5) of the HSEs, including: adipogenesis, complement, coagulation, glycolysis, epithelial–mesenchymal transition, hypoxia, apoptosis, and UV response. Overall, GSEA pathway analysis identified a split in the HSEs, where SC1/SC2 behaved similarly as did SC3/SC4. The SC3/SC4 group upregulated genes of several of the groups mentioned (adipogenesis, glycolysis, for example) whereas SC1/SC2 downregulated them. The addition of the hypodermis to HSEs appears to greatly influence the pathways activated or deactivated. Without a hypodermis (SC1, SC2), the HSEs downregulated pathways that are known to be important to skin function, such as glycolysis, whereas it was upregulated in the study groups containing hypodermis (SC3, SC4).65

Several biological processes GO sets were selected for their pertinence to skin homeostasis and function (Table 4): hallmark adipogenesis, immune system process, inflammatory response, nervous system development, sensory perception, skin development.13, 15, 55, 62, 66-68 The HSE groups remained distinct both with (Figure 5B,C) and without (Figure S2, Supporting Information) GO sets applied.

Only groups with a hypodermis (SC3, SC4) upregulated adipogenesis genes (ECHS1, PLIN2, ADIPOQ), and only SC1 upregulated sensory development genes at q-value (0.1); SC4 upregulated Guanylate Cyclase 2D, Retinal (GUCY2D), a sensory gene, at q-value (0.2) (Figure S3, Supporting Information). However, for all other GO sets studied there was no clear distinction between which group up or downregulated genes belonging to a certain set (i.e., SC2 was not singularly upregulating all nervous system processes). Certain genes, such as LOR (a skin development gene), were upregulated only in SC1, in agreement with research that classifies positive LOR expression as a key developmental marker of healthy skin.19, 69 In comparison to another cultured keratinocyte-only HSE model, SC1 had several overlapping genes (including KRT71, and several similar gene classes between both datasets including C#orf# genes, TRIM#, and LOC# genes), but numerous others which were not mentioned in that analysis.70 It would be difficult to directly compare these models, as many components are different: SC1 is a 3D HSE with a dermis composed of a silk-collagen hydrogel material, and also contains dermal fibroblasts, which could influence the RNASeq results. However, SC1 may be a more accurate representation of skin than the cultured keratinocyte model due to its 3D architecture and inclusion of dermal fibroblasts. Comparisons between the HSE systems (SC1–SC4) with similar models, or commercial models, are presented in Table 5.

Samples with hiNSCs added, SC2 and SC3, both upregulate SOX11 for nervous system development which regulates neuronal proliferation, axon growth, and promote peripheral nerve regeneration. This result was consistent with previous findings of hiNSCs gene expression.72 The SC3 group also upregulates HEYL and CNTN4. HEYL has been implicated in tumor suppression and the p53 pathway but is also a neuronal differentiation mediator. CNTN4 is an axon-related gene relevant to the development of the nervous system.

SC3 is the most complex group with function in every GO set studied (with exception of sensory development); SC4 similarly has upregulation of every GO set (with exception of sensory development), but SC3 upregulates nervous system development genes (CNTN4, SOX11) that are downregulated in SC4. The other main difference between SC3 and SC4 is FOXP3 (immune) and ORM1 (inflammatory) expression in SC4. We speculate that the addition of hiNSCs could be implicated in why there were these differences between groups. For ORM1, overexpression has been linked to inflammatory skin conditions like atopic dermatitis.73 This could explain why SC3, which has most of the components of human skin, did not overexpress this gene. It could be possible to utilize these HSE systems to investigate ORM1 in “healthy” (perhaps SC3) versus “diseased” (perhaps SC4) human skin.

In terms of the immune and inflammatory response of the HSEs, SC1 and SC4 uniquely upregulated several genes (APOA4, IL31RA, GHSR; FOXP3 and ORM1, respectively). Complement-2 (C2) was downregulated in SC1/SC2 but upregulated in SC3/SC4, likely as a result of the presence of the hypodermis. However, SC3 downregulates FOXP3 (immune) and ORM1 (inflammatory) which were upregulated by SC4, which may be due to either the addition of hiNSCs in SC3 or because of the total interaction of cells/components within SC3.

3.4 Two-Group Analysis

Through two-group analysis, several uniquely expressed genes with respect to sample type were identified. Multiple genes were selected for each sample type to classify groups further. One emerging difference between groups was that SC1 had numerous genes relevant to epithelialization or skin development uniquely expressed. Descriptions of genes and their related functions are in Table 2. The classifications of these genes (e.g., several neuronal system related genes were present in SC2 and SC3 groups), aligned to the relevant groups, but the approach was not exhaustive.

3.5 Future Directions of HSE Assessment

Tissue-engineered systems, including HSEs, are in the age of an information renaissance due to the advancements in genomics and proteomics. RNASeq is uniquely suited to study novel genes and identify pathways that microarrays alone may not; the larger datasets from RNASeq provide a means to identify specific genes and relevant pathways for certain skin diseases such as psoriasis.74, 75 However, standard RNASeq of a tissue sample or model cannot selectively identify gene expression from individual cell types.17 Alternative techniques which potentially can pair differential expression with cell type include single-cell RNASeq, or fluorescent-activated cell sorting (FACS) followed by RNASeq (FACS-Seq). FACS-Seq has been used on human skin cells, by first isolating only certain skin cells of interest and then performing RNASeq with respect to individual cell populations. This methodology allowed for analysis of individual cell contributions of the skin, which solidifies the technique as a powerful tool.17

For future application, the HSEs could be utilized as an in vitro diagnostic to connect specific cell types of the skin with expressed genes and pathways and implicate them for therapeutics, or to simply better understand the full complexity and interconnected nature of cells of the human skin.

3.6 Limitations

A major limitation of the current HSEs presented herein is the epidermal layer, and lack of similarity to native human skin in terms of stratification of epidermis, and accurate differentiation of the keratinocytes. Questions remain on the necessary level of differentiation of the epidermis, and the barrier function, that are needed for an optimal clinical HSE for patient use.9, 76 In native skin, both keratinocytes and fibroblasts maintain the barrier layer of the skin, specifically the stratum corneum of the skin, and when it is broken the wound healing process initiates and transepidermal moisture loss and percutaneous absorption are compromised.10, 11, 77, 78 With age, and as the skin has reduced ability to replace lipids, the barrier function of the skin is compromised.12, 79 Most in vitro HSEs do not recapitulate the barrier function of human skin,13 as they are typically composed only of keratinocytes and fibroblasts. Barrier function may not be a relevant concern since the model does not include blood vessels, hair follicles, or sweat glands of the native skin, and as a result, the barrier function reported would be impaired and not necessarily transferable to native skin.13 Other groups have suggested that including blood vessels and/or providing a method to allow for transfer of immune cells (i.e., microfluidic device or embedded channels in tissue-based HSEs) would facilitate more relevant barrier function results.9, 13 Therefore, another aspect to make this model more relevant to native skin and improve barrier function simultaneously could be to include such devices to mimic functional components of the native skin.

We establish a methodology for a complex in vitro HSE that includes numerous cell types and layers, including components of the neuro-immuno-cutaneous system. This model provides the ability to study complex skin pathologies using only primary, human cells, in an in vitro setting. However, beyond the epidermal layer, there are other considerations. Due to the complexity of the tissue model, optimizing a coculture medium is important, currently, the model uses a skin-specific media formulation. Through additional characterization of the lipoaspirate and the specific immune cell type(s) and ratio of cell populations, a new media composition could be developed to improve other aspects of the tissue model including epidermal morphology. Histological analysis also suggests that the layers peel apart, which is due to the complex construction of the model and different biomaterial interfaces. Improvements on construction and composition could also be considered to develop a model that is structurally cohesive. Through improvements the tissue model could become useful as a complete in vitro model for interpretation of the skin.