Skin Cell Heterogeneity in Development, Wound Healing, and Cancer

1The skin consists of two layers, the upper epidermis and the lower dermis, which are separated by a basement membrane, and harbours specialised structures such as hair follicles (HFs) and sweat glands (Figure 1). The epidermis is a multilayered stratified epithelium, which is constantly renewed throughout life; homeostasis is maintained by a balance between basal cell proliferation and suprabasal cell differentiation/stratification. At the onset of differentiation, basal cells become detached from the basement membrane, stop proliferating, and once located in the suprabasal cell layer, start executing terminal cell differentiation programmes culminating in loss of the nucleus. The precise sequence of events for cell commitment and differentiation on the single cell and tissue scale is still being investigated (reviewed in ).

2The dermis is composed of different sublayers that are distinguished by cell type, cell density, and extracellular matrix (ECM) composition (reviewed in, ). The papillary layer is located close to the basement membrane and shows a high fibroblast density and dense meshwork of thin, poorly oriented collagen fibres. The reticular dermis is the central and largest layer of the dermis, consisting of thick, highly organised collagen fibre bundles and lower cellular density. Under the reticular dermis lies the dermal white adipose tissue (DWAT), also referred to as hypodermis, which harbours pre- and mature adipocytes. Mature adipocytes are filled with lipids (Figure 1B). In addition there are specialized fibroblast subsets in the skin that form the dermal papilla (DP) (see Glossary) at the base of HFs, the dermal sheath that is wrapped around HFs and the arrector pili muscle (APM), which is connected close to the HF bulge and is responsible for piloerection (reviewed in ).

3In this review we highlight recent advances in dissecting different cell subpopulations in the epidermis and dermis. We discuss how epidermal and dermal cells interact with each other during homeostasis, wound healing, and cancer.

4Multiple epithelial stem cell (SC) populations have been shown to contribute to skin homeostasis. In mice a highly diverse pool of SCs has been identified in the HF, ranging from the lower HF (Lgr5+ and CD34+/Krt15+) bulge to the upper HF (Gli1+ and Lgr6+) and junctional zone (Lrig1+) (Figure 1A). In undamaged skin the different cells populate distinct and restricted areas in the HF, whereas upon tissue injury these cells have the intrinsic ability to give rise to all epidermal cells, including the interfollicular epidermis (IFE), which is located between HFs and comprises the largest pool of keratinocytes in the skin,. How these differential cellular behaviours are regulated at the molecular level is currently being investigated (Box 1).In the IFE and HF maintenance of distinct SC populations relies on a tight interplay of intrinsic factors and extrinsic factors that define the cellular microenvironment or niche (Figure I). The composition of the niche – neighbouring cells, ECM, GF, and physical parameters – is highly location and cell specific. Key signalling pathways involved in cell microenvironment regulation in the skin include WNT/β-catenin, TGFβ, BMP, fibroblast growth factor, SHH, and Notch, which can act in an autocrine and paracrine manner,,.Key intrinsic factors include SC-specific regulation of metabolism, gene expression, and chromatin arrangement. For example, HFSCs produce more lactate than other epidermal cells, suggesting that differences in cellular glycolytic metabolism influence SC activity. In the HF bulge the transcription factor Foxc1 is specifically induced in HFSC upon activation of the hair cycle, and regulates the expression of BMP and nuclear factor of activated T cells 1 to maintain HFSC quiescence during HF growth,.Within the niche interactions between neighbouring cells are important regulators of cell behaviour. Upon epithelial SC ablation neighbouring cells are able to replace niche SC even if they are committed progenitors. In addition, loss of single ECM components can be sufficient to dramatically impact SC maintenance in the niche. DNA damage induced proteolysis of collagen 17a1 in the HF bulge with age leads to a cyclic elimination of HFSCs through epidermal terminal differentiation. Maintaining or replenishing collagen 17a1 in the skin inhibits the ageing phenotypes, suggesting a possible therapeutic application,.Extrinsic niche factors can act on a single cell level and are either produced by the SCs themselves or provided by surrounding cells such as progenitor cells, fibroblasts, resident immune cells, or sensory neurons,,,,. Sensory neurons in contact with the HF create a perineural niche microenvironment in the bulge and isthmus. The secretion of SHH results in a specialised subset of bulge cells characterised by high Gli1 expression and activated Hedgehog signalling. A subset of resident regulatory T cells located close to the HF bulge and expressing high levels of Jag1 promotes HFSC proliferation and differentiation through direct activation of Notch signalling in HFSCs. Impaired crosstalk between epidermal and dendritic T cells severely affects wound repair in aged skin, highlighting the importance of different resident immune cell types in homeostasis as well as tissue repair.Physical parameters such as tension, pressure, or temperature contribute to the cellular microenvironment, are sensed by distinct cell populations, and influence cellular fate. For example, in epidermal SCs a mechanosensory complex consisting of emergin, nonmyosin-IIA, and actin affects gene transcription and lineage commitment. Extrinsic force-induced relocation of emergin and a change in cellular G-actin levels leads to a switch in histone methylation and loss of heterochromatin anchoring to the nuclear lamina, thereby inhibiting global gene transcription. It has also been noted in single cells that there is a potential interplay between the actin cytoskeleton and epigenetic modifications.It remains to be determined whether force-induced changes in chromatin structure are transient or if SCs functionally adapt and memorise the stress response. It was recently shown that epithelial SCs exhibit a memory of previous inflammation upon tissue damage by maintaining chromosomal accessibility at key stress response genes, promoting faster transcription. This memory does not require the presence of resident immune cells but is critically dependent on Aim2 (absent in melanoma 2), which is an activator of the inflammasome upstream of caspase-1 and interleukin-1β. A side effect of this increased responsiveness to stressors in epithelial SCs might be increased susceptibility to autoimmune and hyperproliferative disorders including cancer.Epidermal β-catenin activation not only results in cell-intrinsic effects but also alters the behaviour of neighbouring cells. Recent single cell gene expression profiling has shown on cell intrinsic β-catenin activation, gene expression heterogeneity is reduced in neighbouring cells and the effect is most dramatic for genes associated with protein synthesis. The effect is dependent on cell–cell contact and the changes in gene expression are accompanied by a shift to a more proliferative SC state.It is conceivable that many more specialized SC subpopulations, maintained and defined by a complex array of niche factors, will be discovered; indeed that all epidermal cells in contact with ECM have SC properties.Alt-text: Box 1

5In the IFE and HF maintenance of distinct SC populations relies on a tight interplay of intrinsic factors and extrinsic factors that define the cellular microenvironment or niche (Figure I). The composition of the niche – neighbouring cells, ECM, GF, and physical parameters – is highly location and cell specific. Key signalling pathways involved in cell microenvironment regulation in the skin include WNT/β-catenin, TGFβ, BMP, fibroblast growth factor, SHH, and Notch, which can act in an autocrine and paracrine manner,,.

6Key intrinsic factors include SC-specific regulation of metabolism, gene expression, and chromatin arrangement. For example, HFSCs produce more lactate than other epidermal cells, suggesting that differences in cellular glycolytic metabolism influence SC activity. In the HF bulge the transcription factor Foxc1 is specifically induced in HFSC upon activation of the hair cycle, and regulates the expression of BMP and nuclear factor of activated T cells 1 to maintain HFSC quiescence during HF growth,.

7Within the niche interactions between neighbouring cells are important regulators of cell behaviour. Upon epithelial SC ablation neighbouring cells are able to replace niche SC even if they are committed progenitors. In addition, loss of single ECM components can be sufficient to dramatically impact SC maintenance in the niche. DNA damage induced proteolysis of collagen 17a1 in the HF bulge with age leads to a cyclic elimination of HFSCs through epidermal terminal differentiation. Maintaining or replenishing collagen 17a1 in the skin inhibits the ageing phenotypes, suggesting a possible therapeutic application,.

8Extrinsic niche factors can act on a single cell level and are either produced by the SCs themselves or provided by surrounding cells such as progenitor cells, fibroblasts, resident immune cells, or sensory neurons,,,,. Sensory neurons in contact with the HF create a perineural niche microenvironment in the bulge and isthmus. The secretion of SHH results in a specialised subset of bulge cells characterised by high Gli1 expression and activated Hedgehog signalling. A subset of resident regulatory T cells located close to the HF bulge and expressing high levels of Jag1 promotes HFSC proliferation and differentiation through direct activation of Notch signalling in HFSCs. Impaired crosstalk between epidermal and dendritic T cells severely affects wound repair in aged skin, highlighting the importance of different resident immune cell types in homeostasis as well as tissue repair.

9Physical parameters such as tension, pressure, or temperature contribute to the cellular microenvironment, are sensed by distinct cell populations, and influence cellular fate. For example, in epidermal SCs a mechanosensory complex consisting of emergin, nonmyosin-IIA, and actin affects gene transcription and lineage commitment. Extrinsic force-induced relocation of emergin and a change in cellular G-actin levels leads to a switch in histone methylation and loss of heterochromatin anchoring to the nuclear lamina, thereby inhibiting global gene transcription. It has also been noted in single cells that there is a potential interplay between the actin cytoskeleton and epigenetic modifications.

10It remains to be determined whether force-induced changes in chromatin structure are transient or if SCs functionally adapt and memorise the stress response. It was recently shown that epithelial SCs exhibit a memory of previous inflammation upon tissue damage by maintaining chromosomal accessibility at key stress response genes, promoting faster transcription. This memory does not require the presence of resident immune cells but is critically dependent on Aim2 (absent in melanoma 2), which is an activator of the inflammasome upstream of caspase-1 and interleukin-1β. A side effect of this increased responsiveness to stressors in epithelial SCs might be increased susceptibility to autoimmune and hyperproliferative disorders including cancer.

11Epidermal β-catenin activation not only results in cell-intrinsic effects but also alters the behaviour of neighbouring cells. Recent single cell gene expression profiling has shown on cell intrinsic β-catenin activation, gene expression heterogeneity is reduced in neighbouring cells and the effect is most dramatic for genes associated with protein synthesis. The effect is dependent on cell–cell contact and the changes in gene expression are accompanied by a shift to a more proliferative SC state.

12It is conceivable that many more specialized SC subpopulations, maintained and defined by a complex array of niche factors, will be discovered; indeed that all epidermal cells in contact with ECM have SC properties.

13Alt-text: Box 1

14During development, the epidermis is formed by a flat single layered epithelium known as the surface ectoderm. Local induction of Wnt signalling in the epidermis and subsequently in the dermis leads to the formation of HF placodes, which are characterised by expression of adult SC markers such as Sox9. While adult skin SC maintenance and differentiation depend on signals from local niches such as the HF bulge, during hair bud development, SC specification is achieved by asymmetric cell division, differential levels of Wnt signalling, and the response to sonic hedgehog (SHH) of basal and suprabasal cells. During further maturation of the HF placode, the SC markers start to segregate into the distinct HF structures including the HF bulge isthmus and sebaceous gland (reviewed in ). When the HFs form they further provide niche structure and signals to induce the specification of Merkel cells, a specialised subpopulation of keratinocytes that become organised into touch domes around primary HFs. These innervated mechanoreceptors mediate light touch sensation.

15Beside epithelial cells (keratinocytes and Merkel cells), the adult IFE harbours melanocytes and immune cells (Langerhans cells and γδT-cells). Until recently the organisation of epithelial SCs within adult mouse IFE was a matter of some debate. Initial in vivo clonal analysis in mouse tail and ear epidermis indicated that a single cell population is responsible for epidermal homeostasis, and variation in clone size could be explained by stochastic (random) cell division of a homogeneous population of keratinocytes, referred as the neutral drift model,,. However, it is now clear that IFE SCs are heterogeneous. More detailed characterisation of tail IFE, in which clonal growth studies supporting the neutral drift model were carried out, revealed that there are two distinct pathways of terminal differentiation, one corresponding to the parakeratotic scale IFE that is not associated with HFs, and the other to the orthokeratotic interscale IFE located close to the HFs, each being generated and maintained by a different pool of basal cells,. The size of the scale and interscale regions is controlled by epidermal Eda and Wnt/β-catenin signalling, and there is a corresponding patterning of melanocytes and papillary dermal fibroblasts. Lgr6+ cells contribute to the interscale but not to the scale IFE. The scale and interscale IFE can also be distinguished by expression of Slc1a3 and Dlx1, respectively. The two tail IFE SC compartments differ in their proliferative dynamics, gene-expression profiles and ability to repair the epidermis after injury,.

16Single cell transcriptomic analysis of mouse dorsal epidermis and cultured human epidermis has identified at least two distinct IFE SC transcriptional signatures, even though there appears to be a single terminal differentiation programme,. It is not known at present whether the cellular heterogeneity in the IFE reflects differential susceptibility to initiating keratinocyte differentiation. In addition the proliferative properties of cells in the IFE basal layer are influenced by the HF cycle. Lineage tracing experiments have revealed that while cell clones associated with HF show a rapid increase in size during the HF growth phase, distant clones cycle more slowly, yet can be mobilised upon tissue injury. Thus, while in mouse tail IFE, distinct SC populations are associated with unique differentiation programmes, SC heterogeneity in mouse back skin IFE underlies a single differentiation programme and could reflect different cellular states.

17To gain further insights into the proliferative dynamics of epidermal cells with age, in recent yearsclonal analysis has been applied to human epidermis by making use of sunlight induced mutations in cancer-associated genes, such as p53, as markers,. This has led to conflicting conclusions about the relative importance of positive selection and neutral drift in clonal evolution. Recently, by sequencing larger areas of skin than previously and focusing on skin from patients who had previously developed a skin tumour, it has been possible to establish that some human mutant clones are too large to be accounted for solely by neutral drift. Rather, secondary mutations arising at the edge of a mutant clone have a selective growth advantage.

18Beside its role as an ECM-rich scaffold, the dermis harbours highly diverse fibroblast, pericyte, and immune and endothelial cell populations that dynamically change with age and influence the properties and cellular behaviour of the overlying epidermis,, (Figure 1B). Although the dermal layers can be easily distinguished by collagen structure and cellular density, the cellular events generating and maintaining dermal architecture have not been explored in detail until recently. During mouse embryonic development, dermal fibroblasts arise from at least two spatially and functionally distinct cell lineages that differentiate into distinct subpopulations and contribute to the dermal layers,. Neonatal dermis fibroblasts of the papillary layer are characterised by active Wnt signalling and proliferation, whereas populations in the reticular layer show increased expression of ECM and immune cell associated genes,,. Whether bone-marrow-derived mesenchymal stromal cells (MSCs) contribute to the resident fibroblasts of mouse dorsal skin under homeostatic conditions or following wounding is controversial (reviewed in ).

19During development, gene expression in dermal fibroblasts is highly dynamic and there is a swift change in dermal fibroblast behaviour on the tissue scale during dermal maturation,. While fibroblasts are highly proliferative during embryonic development they rapidly stop proliferating in the postnatal dermal growth phase, which is characterised by extensive ECM deposition and remodelling. Clonal analysis reveals that individual fibroblasts within clones start to be segregated by increased ECM deposition, leading to a dramatic reduction in fibroblast density postnatally. In addition there is an increase in the adipocyte layer with age. The different dermal fibroblast lineages are spatially segregated in P2 mouse skin, whereas mixing of lineages occurs during subsequent dermal maturation. Recent single cell RNA sequencing studies of human dermal fibroblasts identified several transcriptionally distinct subpopulations, some of which are spatially segregated whereas others are not,. Further studies are needed to clarify whether the transcriptionally heterogeneous fibroblast subpopulations in human dermis represent different cellular states or functionally distinct fibroblast lineages.

20While the nature of the switch between fibroblast proliferation and ECM production is unknown, gene expression analysis of neonatal (proliferative) and aged (nondividing) dermal fibroblasts suggests that it is controlled by epigenetic changes at the chromatin level,. In vitro studies suggest that there is not a common nonproliferative state in fibroblasts but that it is rather an accumulation of different states that are actively maintained,. Besides actively reinforcing the nondividing cell cycle state and repressing the transition into senescence or terminal differentiation, aged fibroblasts remain highly metabolically active. They increase expression of ECM proteins such as collagen I and III, which is partly due to changes in expression of miRNAs such as miR-29.

21Reciprocal signalling between epidermis and dermis plays a key role in skin development, homeostasis, wound repair, and cancer. A prime example of tight temporal and spatial regulation involves the DP fibroblasts at the base of the HF and the overlying HF bulb (reviewed in, ). The transcriptional repressor Blimp1 is a key target and mediator of this interaction. Upon HF induction during development epidermal Wnt/β-catenin signalling induces Blimp1 expression in the DP via transforming growth factor (TGF)β signalling. In the DP, Blimp1 promotes Wnt/β-catenin signalling activity and HF growth.

22When the DP matures during HF growth, heterogeneity of mesenchymal signals at the single cell level involving gradients of Wnt ligands and bone morphogenetic protein (BMP) inhibitors creates distinct microniches along the epithelial–mesenchymal interface in the DP. By secreting distinct combinations of factors these microniches coordinate the hierarchy of self-renewal and differentiation states of epithelial cells in the HF matrix, enabling formation of the different HF layers.

23The HF bulge SC niche harbours a distinct combination of ECM components. Bulge cells have been shown to create microniches for APM cells by depositing specific ECM components including nephronectin in the basement membrane. Nephronectin is specifically recognised by α8β1-integrin positive mesenchymal cells of the APM, which is essential for proper anchorage and function. Interestingly, regenerated HFs in wounds lack APMs, indicating that additional factors or specific mesenchymal subpopulations are required for APM formation,.

24In contrast to the HF, in the IFE the spatial and temporal coordination of epidermal–dermal interactions has not been defined at single cell resolution. However, in tail skin, leucine-rich repeats and immunoglobulin-like domains protein 1 (Lrig1) expression is selectively upregulated in dermal fibroblasts that underlie the interscale epidermal compartment, pointing to different dermal niches in the interscale and scale.

25One of the key pathways of epidermal–dermal communication is via canonical Wnt signalling. Activation of Wnt/β-catenin in basal keratinocytes of adult mouse epidermis induces a rapid increase in fibroblast proliferation and stimulates ECM remodelling,.

26It seems likely that, similar to the epidermis, dermal subpopulations are maintained by a complex combination of extrinsic and intrinsic factors (Box 1). On the one hand, reciprocal transplant experiments involving fibroblasts of different developmental origins (oral cavity vs skin) have revealed that their differential ECM deposition and migratory behaviour is maintained and thus are cell intrinsic features. Furthermore Sox2+ DP fibroblasts maintain their identity in cell culture. On the other hand, different fibroblast subpopulations respond to distinct paracrine signals from the epidermis. Epidermal SHH signalling stimulates proliferation and ECM remodelling of the upper dermis, while TGFβ-2 influences proliferation, ECM deposition and differentiation of the lower dermis. In addition, secreted factors of Wnt/β-catenin activated keratinocytes promote adipocyte differentiation and expansion of the hypodermis whereas PDGF expressed by preadipocytes stimulates HF SC activity.

27Wound healing and tissue regeneration are coordinated processes that involve epidermal, dermal, endothelial and immune cells, and can be divided into distinct phases (reviewed in, ). Upon wounding, a blood clot forms and immune cells infiltrate the wound site (inflammatory phase). The cells from the epidermis and dermis start proliferating and migrate into the wound bed to close the wound (proliferative phase). Then, dermal cells deposit and restructure the ECM in the wound bed (resolution phase). Of note, wound healing responses differ between species: for example, in mice wounds close mainly through tissue contraction as a result of the action of the panniculus carnosus muscle, a structure that is absent in human skin (reviewed in ). In mouse skin, several epidermal cell populations contribute to the skin wound repair process, leading to the picture that not only cells close to the wound edge but also HF epithelial cells are recruited to the wound site (Figure 2A,B). How distinct cell populations from different niches coordinate on the tissue scale and move into the wound bed, ensuring rapid wound closure and tissue regeneration, has only recently been addressed.

28In the past, two distinct mechanisms for epidermal cell movement into the wound bed have been proposed. One involves a smooth flow of epidermis following a homeostatic rule of unidirectional basal to suprabasal transfer of cells with front edge movement that is achieved by basal cells migrating across the wound,. The second, referred to as leapfrogging, involves suprabasal cells sliding over the leading basal cells to become basal cells themselves,. Two recent studies, combining in vivo live cell imaging, lineage tracing, and transcriptomic analysis, have mapped the anatomy and spatiotemporal dynamics of the wound healing response and identified two concentric zones of cellular activity and differential gene expression,. The cell population closest to the wound edge is characterised by rapid migration and differentiation. Further from the wound there is a zone with high epidermal proliferation and little migration along the basement membrane. In the first (migratory) zone both basal and suprabasal cells actively migrate with increasing speed towards the wound centre and upregulate genes involved in ECM remodelling and cell adhesion. These include integrin α5β1, a fibronectin receptor, which enables keratinocytes to migrate on the provisional ECM deposited by immune and fibroblast cells. Epidermal cell migration and differentiation rates are coupled, inducing a coordinated tissue thickening over time in the leading edge. The second (proliferative) zone of keratinocytes not only supplies new cells but also controls the involvement of the surrounding unwounded epithelium during re-epithelialisation. During wound healing, the direction of cell movement within the migratory epidermal zone influences cell division orientation towards the wound bed centre, demonstrating a spatial interplay between proliferation and migration. Clonal lineage tracing reveals that most of the committed progenitor cells become highly proliferative and rapidly differentiate in the early wound healing phase, whereas basal (SC-enriched) cells become activated and increase the pool of SCs later. These studies suggest a model whereby the nonproliferating leading edge functions as a scaffold, preparing the wound bed for efficient repopulation towards the wound bed centre and protecting SCs during tissue repair. The clonal dynamics of different SC populations from the HF and IFE are similar, indicating that the behaviour balancing proliferation and differentiation within the wound healing zones is independent of the original cell. Furthermore, the leading edge cell gene signature is specified independently of cell division or inflammation. This indicates that both intrinsic and extrinsic factors establish the wound healing zones.

29Epidermal wound healing induces plasticity in differentiated sebaceous duct cells such that they dedifferentiate, proliferate, and contribute to long-term maintenance of the IFE. These cells are located in the suprabasal layer of the sebaceous duct that connects the HF junctional zone with the sebaceous gland and are defined by high expression of the transcription factor GATA6, a key regulator of the sebaceous duct lineage during homeostasis. Upon wounding, differentiated GATA6 lineage cells in the HF duct become mobilised, migrate to the wound site suprabasally as individual cells, reattach to the basement membrane, and start proliferating to regenerate the IFE in the wound bed. Mitogen-activated protein kinase signalling is a key pathway that regulates keratinocyte differentiation, while the signals regulating dedifferentiation remain unclear,. In vivo live imaging reveals that downward cell migration is more pronounced at the wound edge than distally. This is in contrast to the behaviour of basal Lrig1+ cells of the HF junctional zone, which enter the wound bed as a cohesive basal cell population,,. GATA6 lineage cells not only contribute to wound repair but also populate the lower HF and sebaceous gland of HF close to the wound. Similar cell displacement has been observed in regenerating tail IFE. Upon wounding, SC populations in the interscale and scale migrate into neighbouring territories and contribute to tissue repair outside their normal niche. This observation indicates a loss of lineage restriction during tissue repair, which is controlled by the transcription factors KLF5 and Sox9.

30Our understanding of epidermal wound healing has rapidly advanced, while the contribution of different dermal cell populations has been less clear (Figure 2A,C). Fibroblasts in the vicinity of a wound are known to become rapidly activated, are referred to as myofibroblasts, and change their behaviour and transcriptional program. TGFβ signalling is the key pathway for fibroblast activation,,,. During the whole wound repair process there is an intensive and highly dynamic crosstalk with multiple immune cell populations, which influences mesenchymal cell behaviour (reviewed in, ). We and others have shown that distinct fibroblast populations show differential recruitment and contribution to wound healing,,. Fibroblasts residing in the reticular dermis mediate the first phase of wound repair and are the major source of myofibroblasts, whereas fibroblasts in the papillary dermis enter the wound at a later stage and are essential for HF regeneration. Moreover, while fibroblasts from the DP do not contribute to wound repair, preadipocytes and adipocytes do contribute,. Intriguingly, depletion or manipulation of specific mesenchymal subpopulations in the dermis has been shown to significantly reduce tissue fibrosis or scar formation upon tissue injury,.

31Fibroblast-specific inhibition of Wnt/β-catenin signalling in mice enhances HF regeneration by preventing the early expansion of lower dermal cells. In contrast, sustained Wnt/β-catenin signalling in dermal fibroblasts impairs regeneration, induces fibrotic lesions in adult skin, and inhibits adipocyte differentiation,. Lineage tracing experiments reveal that this is due to a TGFβ-mediated conversion of adipocytes to myofibroblasts. Conversely, during wound healing, BMP signalling from regenerating HFs induces myofibroblasts to convert to adipocytes in the wound bed. The signals promoting myofibroblast conversion to other mesenchymal cell populations are currently unclear.

32Similar to wound healing, different epidermal and dermal populations contribute to cancer in different ways (Figure 3). This is well illustrated in the case of oncogenic β-catenin signalling, where depending on the epidermal SC type in which stabilised β-catenin is expressed, different tumours are formed. For example, targeting the Lgr5+ population promotes formation of pilomatricomas (benign HF skin tumours), while Lrig1+ cells develop trichoadenomas (a rare benign follicular tumour with cornifying cysts) and the Lgr6+ population gives rise to dermatofibromas within the IFE. Similarly, activation of Hedgehog signalling demonstrates that only basal Krt14+ cells in the IFE and HF infundibulum can initiate basal cell carcinoma (BCC) formation,. BCC initiation and progression are highly dependent on expression of the transcription factor Sox9. In contrast, squamous cell carcinoma (SCC) can originate from more than one epidermal population, including HF bulge SCs, as overexpression of a hyperactive Kras mutant in different epidermal lineages induces tumours with comparable efficiency. Consistent with this, it has been shown that SCC cells harbour a distinct open chromatin landscape combining active gene signatures from distinct SC lineages. This state of lineage infidelity can be transiently observed in wounds, but persists during malignant progression, promoting uncontrolled growth and heterogeneous tumour cell behaviour,.

33During tumour initiation, mutant epidermal cells have the ability to engage nontransformed (healthy) cells via paracrine signalling, such as Wnt ligand secretion, to induce aberrant growth of the whole tissue. Intriguingly, the converse is also observed: a tumour protective, ‘neighbourhood watch like’ mechanism, contributes to neoplastic tumour suppression. Using an in vivo live imaging approach, it was observed that healthy epithelial cells routinely recognise, surround, and eliminate mutant cells to restore tissue homeostasis, revealing an innate cellular ability to prevent over proliferation and tumour initiation. So far the molecular mechanism of cell recognition and elimination and whether it is restricted to specific cell types remains unclear. The tumour-promoting as well as protective effects upon oncogenic β-catenin signalling rely on Wnt ligand secretion, suggesting that distinct Wnt ligand combinations balance cellular plasticity and behaviour.

34The tumour microenvironment, including the tumour stroma, comprising all nontransformed tissue components associated with a tumour, can have both tumour-promoting and -inhibitory effects. Besides endothelial and immune cells a major component of the microenvironment are cancer-associated fibroblasts (CAFs), which play an important role in the evolution of solid tumours. Similar to myofibroblasts, CAFs seem to originate from different mesenchymal populations, ranging from normal fibroblasts and MSCs to transdifferentiated epithelial and endothelial cells. In contrast to normal fibroblasts, CAFs either reside within the tumour margin or infiltrate the tumour mass and show increased proliferation, migration, ECM deposition, and secretion of growth factors and other ECM modulators (reviewed in, ). Functionally, while CAFs are highly heterogeneous in terms of gene expression, they show enrichment of similar gene ontology classes such as cell adhesion, immune response and ECM modulation, suggesting that different cell types under similar conditions perform similar tasks. Thus it has been proposed that CAFs represent a dynamic cellular state of fibroblast-like cells in the vicinity of the tumour rather than a specific cell lineage. This state could be maintained by a combination of genetic mutations, epigenetic alterations, and persistent environmental effects.

35Similar to wound healing it seems likely that distinct fibroblast populations give rise to CAFs. Indeed, one study identified that in adult mice CD26+ fibroblasts are the main contributor to ECM deposition in a skin melanoma xenograft model. Depletion of this fibroblast subpopulation significantly reduced tumour growth, revealing that targeting distinct fibroblast subpopulations impacts tumour development. Inhibition of CD26 activity reduces the growth of wound-induced epidermal tumours, although it must be noted that CD26 is expressed by both epidermal and dermal cells and that during tumour progression there are dynamic changes in dermal CD26 expression.

36To date there have been few studies of how different fibroblast lineages contribute to tumour stroma formation. Intriguingly, the different tumours induced by stabilizing β-catenin in Lgr5+, Lgr6+, and Lrig1+ epithelial cells exhibit both similarities and differences in stromal composition. In all three cases there are local increases in fibroblast proliferation, ECM remodelling, and expression of CD44, a major hyaluronic acid surface receptor. However, dermal CD26 expression is upregulated in the stroma of Lgr6 but not Lgr5 or Lrig1 tumours. Conversely Lgr6 and Lrig1 tumours have associated stromal inflammation while Lgr5 tumours do not. It will be of interest to discover whether or not these differences reflect differences in the lineages of fibroblasts associated with each type of tumour or whether fibroblasts respond to tumour-specific signals independent of lineage.

37In recent years our appreciation of epidermal and dermal cell heterogeneity has grown, together with a realisation that cells are capable of exhibiting plasticity and changing fate through dedifferentiation and transdifferentiation. It is now clear that the responses of the skin to the challenges of wounding or tumorigenesis reflect a combination of changes in cell intrinsic properties and responses to different microenvironments. Microniches and cell memory are exciting new concepts that warrant further investigation, together with unravelling the distinction between cell types and states, for example in tumour stroma (see Outstanding Questions).

38As more single cell gene expression profiles become publically available and tools to make those datasets readily accessible to researchers with a biological background improve,, we anticipate greater appreciation of the significance of cellular heterogeneity. The combination of experimental data with computational modelling,,,, not only allows rigorous evaluation of data quality but also fosters hypothesis generation. Ultimately we anticipate major benefits in terms of understanding tissue scale behaviour, dynamics, and the interplay of distinct cell populations during tissue development, homeostasis and disease.How is the behaviour of cell populations in the epidermis and dermis coordinated during wound repair and tumour development on the tissue scale?What defines and distinguishes cellular state and cellular type in the skin?What are cell type specific and universal mechanisms for establishing and maintaining cellular identity and plasticity in the skin?What are the key changes that create a pro-oncogenic environment?How do distinct epidermal and dermal cell populations contribute to tumour heterogeneity and progression?What are the dynamics and organisation of microniches in the dermis and epidermis?

39How is the behaviour of cell populations in the epidermis and dermis coordinated during wound repair and tumour development on the tissue scale?

40What defines and distinguishes cellular state and cellular type in the skin?

41What are cell type specific and universal mechanisms for establishing and maintaining cellular identity and plasticity in the skin?

42What are the key changes that create a pro-oncogenic environment?

43How do distinct epidermal and dermal cell populations contribute to tumour heterogeneity and progression?

44What are the dynamics and organisation of microniches in the dermis and epidermis?