Skin fibroblast functional heterogeneity in health and disease

Fibroblasts are the major cell population of connective tissue, including the skin dermis, and are best known for their function in depositing and remodelling the extracellular matrix. Besides their role in extracellular matrix homeostasis, fibroblasts have emerged as key players in many biological processes ranging from tissue immunity and wound healing to hair follicle development. Recent advances in single‐cell RNA‐sequencing technologies have revealed an astonishing transcriptional fibroblast heterogeneity in the skin and other organs. A key challenge in the field is to understand the functional relevance and significance of the identified new cell clusters in health and disease. Here, we discuss the functionally distinct fibroblast subtypes identified in skin homeostasis and repair and how they evolve in fibrotic disease conditions, in particular keloid scars and cancer. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Dermal fibroblasts are the major constituent cells of the dermis and have a pivotal role in maintaining skin physiology, mainly by providing structural integrity. They produce structural components of the extracellular matrix (ECM), including collagens, elastic fibres, proteoglycans (aggrecan, versican, perlecan, decorin), adhesive glycoproteins (laminins, fibronectins, thrombospondins, tenascins), and ECM remodelling enzymes (metalloproteinases, lysyl oxidases) [1]. Single-cell sequencing and lineage tracing studies have identified numerous fibroblast subtypes that exhibit distinct gene expression patterns and modulate different cutaneous functions [2]. In the skin, fibroblasts interact with epidermal, immune, and other cells, orchestrating homeostatic, fibrotic, or wound healing conditions [3]. Besides cell-cell and cell-ECM interactions, fibroblasts secrete an array of paracrine and autocrine signals, such as cytokines and growth factors controlling ECM production, cellular states, and signalling interactions [3,4]. Fibroblasts also serve as progenitors for mesenchymal cell types such as pericytes, osteoblasts, or adipocytes and are instrumental for skin immunity and appendages, in particular hair follicles. During these processes, fibroblasts not only show an astonishing plasticity but also acquire a memory of their positional identity, as well as change their mechanical, inflammatory, and metabolic environment [5].
Most dermal fibroblasts arise from the primary mesenchyme and thus are referred to as mesenchymal cells; however, they can have different origins depending on the body site. While fibroblasts in face skin also arise from the neural crest, ventral skin fibroblasts can derive from the lateral plate mesoderm, and dorsal skin originates from the dermomyotome of the somites [6]. Lineage tracking experiments in mice have revealed that at embryonic (E) day E16.5 multipotent mesenchymal progenitors give rise to Blimp1+, Lrig1+, CD26+ papillary fibroblasts in the upper dermis that contribute to dermal papilla (DP), dermal sheath (DS), and arrector pili muscle cells and Dlk1+/À, Sca1+ reticular fibroblast progenitors in the lower dermis that can further differentiate into preadipocytes and mature adipocytes of the dermal white adipose tissue (DWAT) [7,8].
On the tissue scale, we recently showed that there was a coordinated switch in fibroblast behaviour, from highly proliferative in embryonic to quiescent in postnatal skin, allowing for efficient ECM deposition/remodelling, which is promoted and maintained by accumulating ECM [9]. While fibroblast quiescence can last a long time, upon wounding fibroblasts rapidly become activated, proliferate, and migrate into the wound bed before depositing ECM. After normal tissue repair, wound bed fibroblasts re-establish quiescence to maintain skin homeostasis; however, in pathological scars, such as hypertrophic scars or keloids and in cancer, fibroblasts become 'locked' in a pro-fibrotic/pro-inflammatory state, promoting an ongoing tissue repair response. How this tight balance of fibroblast proliferation, quiescence, and ECM deposition is regulated at the molecular level during dermal development, repair, and disease and how it differs between fibroblast subpopulations remains unclear.
Here we discuss the latest advances in our understanding of fibroblast functional heterogeneity, the key changes during tissue repair and fibrotic disease conditions, and how it can contribute to the development of therapeutic approaches in fibrosis and cancer.

Fibroblast functional heterogeneity in skin homeostasis
Fibroblasts are regarded as one of the most difficult cell types to classify given their heterogeneity and plasticity. Markers such as vimentin (VIM), FSP1/S100A4, Twist2, COL1A1, and PDGFRα have been proposed as discrete markers but are often highly expressed in other cell types, such as immune, smooth muscle, or endothelial cells [10][11][12]. Therefore, it comes as no surprise that single-cell RNA-sequencing (RNA-seq) studies use different marker combinations to identify skin fibroblast clusters [11][12][13][14][15]. While fibroblast marker expression between mice and humans differs significantly, the enrichment of subpopulation-specific gene ontology (GO) terms and signalling pathways are largely conserved, indicating functional conservation across species [14]. Ascensi on et al. integrated human singlecell RNA-seq datasets from different anatomical locations and age, identifying three main fibroblast classes with multiple subclusters. The first class is responsible for dermal cell and ECM homeostasis, the second seems to have a role in immune surveillance and promoting inflammation, and the last subset includes more specialised subpopulations such as DP and dermohypodermal junction fibroblasts [16].

Secretory papillary fibroblast
Papillary fibroblasts are the primary source of de novo DP cells during skin development, which are required for hair formation [7]. This subgroup is generally located in the upper papillary dermis and in human skin expresses known papillary markers including COL6A5, COL18A1, WIF, APCDD1, AXIN2, PTGDS, and COLEC12 associated with collagen or ECM production/organisation, Wnt/βcatenin signalling, and mesenchymal functions [11,14]. In neonatal mouse skin, several papillary fibroblast subpopulations, including a secretory papillary population, have been characterised by expressing markers such as Blimp1, Lrig1, and CD26 linked to embryonic morphogenesis, regulation of cell development, tissue morphogenesis, and negative regulation of chemotaxis [14,17]. A sharp decrease in papillary fibroblast heterogeneity has been observed during dermal maturation, suggesting that most cell clusters in neonatal mouse skin are transient cellular states rather than true subpopulations [17]. Besides providing progenitor cells for hair follicle-associated mesenchymal cells, papillary fibroblasts also regulate hair growth by modulating the function of perifollicular macrophages via Icam1 juxtacrine interactions and generate the hyaluronan-rich papillary dermis stroma that adheres to Tregs that are responsible for immunotolerance [18]. Notably, the cell surface ectonucleotidase CD39 was identified as a conserved marker between papillary fibroblasts in mouse and humans, which has a role in ATP hydrolysis in the extracellular space [14].

Secretory reticular fibroblasts
Secretory reticular fibroblasts located in the reticular dermis show a prominent ECM gene signature (COL1A1, COL1A2, ELN, FN1, FBN1, MFAP4, MFAP5, LOX), indicating that this subpopulation is responsible for ECM assembly, fibre organisation, wound healing, and angiogenesis [13,14]. Within the secretory reticular fibroblasts there seems to be further functional heterogeneity that may be due to differences in anatomical location, tissue, and cellular state. For example, in human trunk skin, a subpopulation with prominent secretory reticular fibroblast characteristics expressing COL14A1, FN1, HAS1, and MMP14 and contributing to matrix assembly and ECM organisation was also associated with IFN-γ response, NF-κB, and TNF-α signalling, suggesting an additional proinflammatory or immunomodulatory role [13]. Reticular fibroblasts were also shown to recruit monocytes and dendritic progenitors by secreting Ccl2 [19]. Interestingly, a skin single-cell transcriptomic study comparing resting (telogen) and growing (anagen) hair follicles discovered dynamic changes in reticular fibroblast subpopulations during hair cycle progression [20]. The subtype enriched in telogen skin was characterised by distinct matrix proteoglycans (Dcn, Lum, Mfap4) and genes involved in cell migration and cytoskeleton (Tmsb4x, Gsn, Cd9). In contrast, the subpopulation enriched in anagen skin expressed high levels of profibrotic markers (Loxl2, Sparc, Col1a1), transcription factors (Creb3, Creb3l3, Mxd4) and proteins involved in vesicular transport (Tmed9, Sec31a) and energy metabolism/mitochondria (Ndufs4). Currently, their functional implication for skin and hair follicle homeostasis remains largely unclear. a reticular gene signature, it is characterised by proinflammatory genes such as CCL19, APOE, CXCL2, CXCL3, and EFEMP1, which are related to inflammatory response, cell chemotaxis, or negative regulation of cell proliferation. Spatial tissue analysis showed a broader dermal distribution with preferential vasculature association [11]. Vorstandlechner et al. identified two pro-inflammatory fibroblast subpopulations in human trunk skin [13]. The first subset (APOE+, CXCL12+) was associated with cell migration, antimicrobial response, regulation of keratinocytes proliferation, lipid transport/homeostasis, and growth factor binding; the second (GEM+, CXCL1+, CXCL2+, CXCL3+) revealed many distinct functional properties, including MAPK and NF-κB signalling, immune response, regulation of apoptosis, and response to cytokines. While a pro-inflammatory population has not been functionally described in mouse skin, fibroblasts in reindeer dorsal skin uniquely expressed 'pro-inflammatory' genes (CXCL1, CXCL3, and CCL2), indicating functional conservation across species [21].

Mesenchymal fibroblasts
This fibroblast subpopulation is predicted to be localised in the reticular dermis and shows a strong reticular signature of markers including ASPN, POSTN, GPC3, TNN, and SFRP1 [11]. GO term analysis indicates a Fibroblast functional heterogeneity 611 potential involvement of anatomical structure formation and muscle development [10]. The distinguishing feature is prominent mesenchymal cell characteristics with high COL11A1 (associated with cartilage development) and COL24A1 (associated with bone development) [11,15]. Interestingly, mesenchymal fibroblasts become enriched in multiple fibrotic disease conditions and appear to have an increased plasticity for chondrocyte, osteocyte, and adipocyte differentiation [22,23]. Although this subpopulation has not been defined in other species, a reticular fibroblast subcluster of postnatal mouse skin shares most of these markers [17], and similar genes enriched in reindeer velvet fibroblasts are associated with regenerative competence (CRABP1, MDK, and PTN) [21].

Cellular heterogeneity in specialised fibroblast compartments
Besides interfollicular fibroblasts, there is accumulating evidence for functionally distinct subpopulations residing in the DP, DS, DWAT layer, and around blood vessels.
Dermal papilla cells DP cells located in the hair bulb are essential for hair follicle development and homeostasis. Functional cellular heterogeneity has been observed within the DP as well as between different hair follicle types. During hair follicle growth Yang et al. revealed that in the DP distinct micro-niches are formed along the epithelialmesenchymal interface through single-cell level heterogeneity of mesenchymal signals involving gradients of Wnt ligands and bone morphogenetic protein inhibitors [24]. By secreting distinct combinations of factors, these mesenchymal micro-niches coordinate the hierarchy of epithelial cell self-renewal and differentiation, enabling the formation of the different hair follicle layers. We showed that the zinc finger transcription factor Blimp1 becomes transiently expressed in DP cells, promoting hair follicle growth and highlighting the temporal heterogeneity of DP cells during hair cycle progression [25]. Distinct DP cells also modulate hair follicle types. For example, the transcription factor Sox2 is highly expressed in the DP of guard/awl/auchene but not zigzag hair follicles [26]. DP specific ablation of Sox2 prevents awl/auchene hair follicles and promotes zigzag hairs in skin grafting experiments.

Dermal sheath cells
DS cells are a specialised α-smooth muscle actin (αSMA)+ mesenchymal cell type that tightly wraps around hair follicles [27]. The intensive DS contraction and expansion during hair cycling suggest a hierarchal organisation of a heterogenous cell population including long-lived stem cells and short-lived progenitor cells, which can contribute to the DP. Similar to the DP, DS cells interact closely with neighbouring epithelial progenitors and may adopt a micro-niche organisation during hair follicle growth. With age, DS number and potentially heterogeneity decrease, impacting hair follicle homeostasis (smaller DPs and thinner pelage coat) and survival [28].

Pericytes and perivascular adventitial fibroblasts
Pericytes are a specialised mesenchymal cell population embedded in the basement membrane of capillaries, contributing to blood vessel development, homeostasis, and blood flow regulation. Transcriptionally, pericytes can be identified and distinguished from perivascular adventitial fibroblasts by high expression of ACTA2, RGS5, and PDGFRB [29][30][31]. More detailed analysis of human atopic dermatitis skin samples revealed two distinct pericyte subpopulations, one with high CD36 and FABP4 expression, suggesting involvement in fatty acid transport across blood vessels [15]. In mice, Ng2+ pericytes can be defined in four distinct subpopulations based on Pdgfrα and Pdgfrβ expression [32]. While their functional difference remained unclear, lineage tracing of different fibroblast subpopulations during skin development and wound repair uncovered that multiple fibroblast subsets could give rise to Ng2+ pericytes and that their origin was highly dependent on the tissue location of the forming blood vessel. It is tempting to speculate that pericytes retain a memory of their cell of origin, which may account for their functional heterogeneity. More recently, Gli1 was shown to mark two small subsets of pericytes and perivascular adventitial cells, and only Gli1+ perivascular adventitial fibroblasts were able to contribute to tissue repair upon wounding [33]. Likewise, perivascular adventitial fibroblasts have been shown to expand in multiple inflammatory human skin diseases, such as systemic sclerosis or cutaneous lupus, and acquire functionally distinct activated stages with differential VCAM1 and PDPN expression supporting T cell retention [34].

DWAT cells
The DWAT layer in mice and, presumably, humans develops from the lower dermis without influence from other subcutaneous adipose depots [35]. In vitro, two distinct cell clusters have been characterised in human differentiating preadipocytes, which differ in cell cycle, protein synthesis, growth-associated pathways, and glycolytic metabolism [36]. A recent single-cell RNA-seq study comparing different fat depots in mice revealed at least three white adipocyte subpopulations that differed in their abundance, metabolism, and response to inflammatory cytokines, insulin, and growth hormones. However, whether a similar cellular heterogeneity can be observed within the DWAT layer is currently unknown [37].

Fibroblast heterogeneity in wound healing
Fibroblasts are one of the most active players in restoring tissue homeostasis during wound repair. Besides rebuilding the ECM, fibroblasts support wound epithelialisation, 612 T Bensa, S Tekkela et al hair follicle regeneration, angiogenesis, DWAT layer formation, and modulation of the inflammatory response ( Figure 2). Lineage tracing of distinct fibroblasts revealed that in full-thickness wounds, fibroblasts in the lower dermis, including reticular and fascia fibroblasts, are the first to repopulate the damaged tissue and mediate ECM deposition, whereas fibroblasts in the papillary dermis move in later and may have a role in remodelling and appendage regeneration [7,38]. In vivo live imaging showed distinct modes of migration between fibroblast subtypes. While interfollicular fibroblasts migrate individually [9], fascia cells increase cell-cell contacts by upregulating N-cadherin and connexin Cx43 and show a swarm-like movement into the wound bed [39,40]. This coordinated migration in the fascia enables existing ECM to be pulled into the wound to accelerate wound closure. Ablation of Cx43 prevents fascia fibroblast migration and collagen deposition in wounds, inhibiting scarring. Within the wound bed, fibroblasts of both the papillary and the reticular dermis randomly distribute and start to expand during the proliferation phase [9,41]. In contrast to DP cells and Ng2+ pericytes, adipocytes, DS, and Gli1+ perivascular fibroblasts significantly contribute to wound repair [32,33,42,43]. In addition to differentiating into myofibroblasts, mature dermal adipocytes undergo lipolysis, promoting macrophage recruitment and efficient wound repair [43]. Besides depositing/remodelling ECM in the wound bed, fibroblasts are able to acquire a DP or adipocyte fate in response to distinct signals promoting hair follicle and DWAT regeneration [22,44,45]. Lineage tracing of papillary (Lrig1+) and reticular fibroblasts (Dlk1+) revealed that during wound-induced angiogenesis the location of the regenerating blood vessel will determine which fibroblast subpopulation differentiates into blood vessel associated pericytes [32]. If and how pericytes from different progenitor wound bed fibroblasts influence blood vessel function is currently unclear. This remarkable fibroblast functional heterogeneity could be due to their diversity. Multiple single-cell and spatial transcriptomic studies have started to dissect the complex dynamic and signalling crosstalk of activated fibroblasts in regenerating and scarring wounds [22,44,45]. Guerrero-Juarez et al. identified 12 subpopulations, which not only differ in transcription factor and signalling pathway marker/receptor/ligand expression but also in cell cycle state, origin, and spatial distribution [46]. Although the functional significance of this cellular diversity requires further study, there is accumulating evidence that the surrounding wound bed microenvironment has a key role in controlling these multiple activated fibroblast states.
Combining lineage tracing, spatial transcriptomics, single-cell RNA-seq, and assay for transposaseaccessible chromatin using sequencing (ATAC-seq) analysis of a wound healing time course identified four spatially and temporally distinct wound healing fibroblast phenotypes: a mechanofibrotic, an 'activatedresponder', a 'remodelling', and a proliferator [41]. While in the early wound healing phase 'activated responder' fibroblasts dominate the wound bed centre, a mechanofibrotic phenotype is observed at the wound edges. These then further migrate into the wound, acquiring a proliferative state, and an ECM remodelling fibroblast subpopulation starts to appear in the deeper dermis. Interestingly, inner wound bed fibroblasts showed greater functional heterogeneity, and changes Figure 2. Fibroblast functional heterogeneity in murine regenerating and scarring wounds. Note that papillary fibroblasts are highly abundant in the upper dermis of regenerating wounds, creating a regenerative zone (blue), whereas pro-fibrotic (reticular) fibroblast populations accumulate in surrounding scarring zones (red). Arrow thickness indicates different abundance of fibroblast subpopulation recruitment into the wound beds where they become activated (yellow framing). While fascia fibroblasts show a collective migration and pull existing ECM into the wound bed, adipocytes can either transdifferentiate into myofibroblasts or undergo lipolysis. At later wound healing stages, myofibroblasts may transdifferentiate into adipocytes to regenerate the DWAT layer. Also, epidermal cell signalling differs in regenerating (white) and scarring (red) wounds. Fb, fibroblast.

Fibroblast functional heterogeneity 613
in the local mechanical microenvironment appear to modulate chromatin accessibility of key transcriptional regulators and ECM components, which frequently precede transcriptional activation [41]. In regenerating wounds, the appearance of hair follicle neogenesis predominantly in the central wound bed area further suggests the presence of tissue repair zones with differential regenerative potential [47,48]. Activated fibroblasts in the upper centre preferentially express cellular retinoidbinding proteins (i.e., Crabp1 and Fabp5), Prss35, and Runx1, whereas wound periphery and deeper dermal fibroblasts express scar-associated markers such as Dlk1, Sca1, and Mest [48]. Pharmacological modulation of retinoic acid and Runx1 ameliorates the regenerative potential in wounds. Key epidermal signals promoting a regenerative niche in the upper dermis include Wnt/β-catenin and Sonic hedgehog (Shh), and selective targeting of these signalling pathways may hold therapeutic benefits [45,49]. Notably, sustained Wnt/β-catenin signalling activation in dermal fibroblasts is insufficient for hair follicle neogenesis, inhibits adipocyte differentiation, and induces fibrotic lesions [38,50,51]. Fibroblast-specific inhibition of β-catenin signalling enhanced hair follicle regeneration by preventing the early expansion of lower (pro-fibrotic) dermal cells [38]. In contrast, constitutive overexpression of the canonical Wnt/β-catenin cotranscription factor Lef1 in dermal cells primed wound bed cells towards a regenerative trajectory and dermal Lef1 depletion impaired hair follicle formation [17]. Here dermal Hedgehog (Hh) activation was shown to enable hair follicle neogenesis by inducing DP signature genes including Lef1 in Wnt/β-catenin signalling high (pro-fibrotic) wound fibroblasts [45,52]. Intriguingly, single-cell RNA-seq revealed heterogenous expression of these DP signature genes within the Hh active wound bed fibroblasts, suggesting that additional factors, such as spatial distribution, cell-cell/ECM interactions, and other signalling pathways, are essential for controlling cell plasticity.
In addition to epidermal interactions, myeloid cells appear to play a critical role in determining the wound healing fate [53]. In scarring wounds, prolonged Wnt/βcatenin signalling was enhanced by macrophages that phagocytose and degrade the Wnt inhibitor SFRP4 in the lower wound bed dermis [54]. A macrophage subset expressing CD301b was shown to induce proliferation and activation of a pro-fibrotic adipocyte precursor population, distinguished by Sca1, CD34, and CD29 [55]. This provides a prominent example of how fibroblast functional heterogeneity is regulated at the cellular level within a tissue. Similarly, a human single-cell RNA-seq study comparing healing and non-healing diabetic foot ulcerations revealed enrichment of a unique fibroblast subpopulation with a prominent ECM remodelling and inflammatory gene signature that may be influenced by a higher M1 to M2 macrophage ratio in healing wounds [56]. Intriguingly, comparison of regenerating (velvet) and scarring (dorsal skin) wounds in reindeer revealed that fibroblasts in scar-prone skin exhibited a pro-inflammatory priming inducing a site-specific inflammatory response upon injury [21]. While velvet fibroblasts showed a foetal-like gene signature and immunosuppressive phenotype, dorsal skin fibroblasts promoted rapid myeloid cell recruitment and maturation during wound repair. Targeting these pathological fibroblast inflammatory signals and associated immune cell interactions enhanced skin regeneration.
In addition to cell interactions, local mechanical cues also influence fibroblast behaviour and diversity [57]. The increased mechanical tension in the lower wound bed was recently shown to promote scar formation by reactivating expression of the homeobox gene Engrailed-1 (En1) in reticular fibroblasts [58], which is known to induce a pro-fibrotic fibroblast fate during embryonic skin development [8]. Notably, pharmacological inhibition or depletion of mechanical tensioninduced YAP signalling prevented En1 reactivation and supported a regenerative wound healing response by activating the Wnt pathway regulator Trps1 in fibroblasts.
Skin regeneration is also significantly influenced by aging-induced changes in fibroblast density, composition, and microenvironment [43,59]. Environmental stress factors, such as chronic UVB irradiation penetrating the epidermis and upper dermis, were shown to selectively deplete fibroblasts in the papillary dermis impacting fibroblast heterogeneity and, potentially, wound-healing ability [60]. Recent single-cell analyses comparing wound bed fibroblasts from young and aged mice revealed distinct fibroblast subpopulations with different (pro-inflammatory) cytokine signatures, which could offer new therapeutic targets to improve wound healing in aged skin [61].

Fibroblast heterogeneity in fibrosis
Fibrosis is a non-physiological scarring process that affects parenchymal organs (e.g. skin, liver, lung, or heart) and plays an important role in many chronic diseases, including scleroderma and cancer. Excessive ECM deposition caused by fibroblasts in fibrotic lesions results in progressive destruction of tissue architecture and organ function, representing one of today's major healthcare burdens with no curative treatment. Similar to wound repair, multiple single-cell transcriptomic and lineage tracing studies are hunting the key fibroblast subpopulations and molecular regulators promoting fibrosis in the skin and other organs. In an irradiationinduced dorsal skin fibrosis model, the CD26-expressing fibroblast subpopulation that developmentally derived from an En1-positive cell population has been shown to be a key driver for fibrotic ECM deposition [8]. Conversely, homeobox 1 (Prrx1)-expressing fibroblasts in the ventral dermis [62] and Adam12+ myofibroblast progenitors in the perivascular space are largely responsible for fibrosis [63], indicating the importance of distinct fibroblast subsets in different anatomical locations.

T Bensa, S Tekkela et al
Notably, targeted depletion of these identified pathological fibroblasts using a diphtheria toxin strategy was able to suppress fibrosis. Fibroblast heterogeneity in localised skin fibrotic lesions has been explored in several recent single-cell RNA-seq studies, in particular in keloid scars [23,64,65]. Comparing keloid scar tissue to normal scar identified five major fibroblast clusters and mapped the four functional fibroblast subsets (secretory-papillary, secretory-reticular, mesenchymal, and pro-inflammatory fibroblasts) observed in healthy human skin [11,23] ( Figure 1). The mesenchymal fibroblast population expressing high levels of POSTN, ADAM12, COMP, and COL1A11 was increased in keloid scars and primarily responsible for the pathological collagen deposition, which could be partially blocked by a POSTNneutralising antibody. A subsequent study comparing keloid and adjacent skin revealed a prominent expansion of fibroblasts and vasculature endothelial cells in keloid scars [64]. Subclustering of the two major keloid fibroblast clusters identified a similar functional subset distribution with a mesenchymal-like fibroblast population highly expressing POSTN and COL1A11.
Intriguingly, an increased mesenchymal fibroblast subpopulation was also observed in systemic sclerosis (SSc), suggesting a universal role in fibrotic skin conditions [23]. While fibroblast heterogeneity appears generally preserved in SSc, single-cell RNA-seq of healthy and SSc fibroblasts revealed a prominent shift in cell clusters associated with reticular fibroblasts towards pro-fibrotic gene signatures [66]. Further bioinformatic analysis predicted that a specific fibroblast progenitor subpopulation expressing SFRP2 and DPP4 would give rise to myofibroblasts by a sequential upregulation of pro-fibrotic transcription factors, including FOSL2, RUNX1, STAT1 and SMAD3. Combining single-cell RNA-seq and ATAC-seq, a first large-scale genomic study including almost 100 SSc patients with different disease stages uncovered that disease severity correlated with changes in an LGR5-positive fibroblast subpopulation involved in ECM assembly/remodelling and immune/endothelial cell interactions [67]. While this fibroblast subpopulation resides deep in the reticular dermis in healthy skin, with increasing SSc disease severity LGR5-positive fibroblasts decreased, lost their elongated morphology, and acquired a pro-fibrotic gene signature, which could be modulated by pharmacological intervention.
Also in Dupuytren's syndrome, a progressive fibroproliferative disease condition affecting the hands, key pathological fibroblast subpopulations could be identified, including an immunomodulatory ICAM1+ IL6+ subset and four myofibroblast populations with distinct fibrotic potential [68]. Here, the cell surface marker tetraspanin (CD82) marked one of the highly contractile and activated pro-fibrotic myofibroblast subpopulations and helped to maintain their proliferative potential, which may offer a promising opportunity for selectively targeting this pathological subset.

Fibroblast heterogeneity in cancer
Fibroblasts located in the vicinity of cancer cells, referred to as cancer-associated fibroblasts (CAFs), are a key component of the tumour microenvironment with diverse pro-fibrotic functions that can promote or inhibit tumour progression [69]. CAFs either reside within the tumour margin or infiltrate the tumour mass and show increased proliferation, migration, connective tissue deposition, and secretion of growth factors and other tissue remodelling and immunomodulatory factors. CAFs are recognised as key drivers of malignancy in squamous cell carcinomas and other tumours [70], and distinct subsets have been shown to correlate with patient prognosis [71]. Dissecting and mapping the heterogeneity, origin, and fate of CAFs is an increasing area of research and a prerequisite for the development of new cancer therapies in skin and other organs.
Multiple single-cell studies have started to characterise the expression profiles and spatial organisation of CAFs in different tumour types (Figure 3) [72]. While we still await a comprehensive CAF analysis in skin tumours, here we will summarise the major functionally distinct CAF subpopulations identified in other tumours that may be present in skin as well. A recent multiomic analysis comparing breast, pancreas, and skin tumours in mice and humans defined three major CAF subpopulations across tissue types: matrix-producing mechanoresponsive Postn+ Mgp+ myofibroblastic CAFs (myCAFs), immunomodulating Il1r1+ Cxcl12+ inflammatory CAFs (iCAFs), and Pi16+ Dpp4+ steady-state-like CAFs [73]. Within the tumour, iCAFs co-localise with lymphoid immune cells and myCAFs co-localise with epithelial cells. myCAFs have increased pro-fibrotic markers such as αSMA, TAGLN, MMP11, and POSTN but generally lack expression of inflammatory cytokines. Pancreatic tumour co-culture experiments revealed that this population becomes activated when in proximity with the epithelium via juxtacrine signalling [74]. Gene set enrichment analysis and transgenic cancer models indicate a prominent role in contraction, ECM deposition/remodelling, and immunosuppression [74][75][76]. In most cancers, myCAFs have been shown to suppress tumour growth by restraining the tumour, inhibiting vascularisation, and recruiting immunosuppressive cell populations [73,77]. Conversely, iCAFs are located more distantly from the epithelium, become activated by paracrine tumour signals, and secrete cytokines (IL6, IL11, LIF) and chemokines (CXCL1, CXCL2), which are known for their tumour-promoting abilities [78]. In addition, murine cancer studies have shown that stromal cells expressing iCAF markers increase angiogenesis, chemoresistance, metastases, and immune suppression [79,80]. There is increasing evidence that the iCAF secretome shares substantial overlapping functions with senescent fibroblasts while maintaining iCAF proliferative capacity [74].
More recently, a distinct subpopulation of antigenpresenting CAFs (apCAFs) was identified in pancreatic ductal adenocarcinoma (PDAC), characterised by MHC Fibroblast functional heterogeneity 615 class II and CD74 expression [75]. While lacking classical co-stimulatory molecules expressed by professional antigen-presenting cells, apCAFs are able to activate CD4+ T cells in an antigen-specific manner. Gene set enrichment analysis highlights their prominent immunomodulatory and tumour-promoting role, in addition to increased fatty acid metabolism, MYC targets, and mTOR complex 1 signalling [75]. Besides pancreatic cancer, a similar CAF subtype has been observed in head and neck squamous cell carcinoma and non-small cell lung cancer [81,82]. Furthermore, metabolic CAFs (meCAFs) expressing PLA2G2A, CREB3L1, and CRABP2 are characterised by a highly active metabolic state and were the major CAF subpopulation identified in loose-type PDAC [83]. Biological processes enriched in meCAFs include mitochondrial translation, glycolysis, and gluconeogenesis, and they may promote tumour progression by producing metabolic intermediates supporting cancer and immune cell activity. PDAC patients with abundant meCAFs had a poor prognosis and higher risk for metastasis but showed a dramatically better response to immunotherapy, highlighting how phenotyping CAF subsets can inform cancer prognosis and treatments.
In addition to these major CAF subpopulations, there is an increasing number of more specialised CAF subtypes emerging. For example, complement-secreting CAFs (csCAFs), located in the tissue stroma adjacent to malignant ductal cells, have been identified only in early PDAC [84]. While sharing some similarities with iCAFs, csCAFs appear to influence ECM organisation, angiogenesis, and complement/coagulation cascades. In breast cancer, a developmental (dCAF) and a vascular CAF (vCAF) subpopulation have been described [85]. dCAFs showed a distinct ECM and stem cell-associated gene signature (Scrg1, Sox9, and Sox10), whereas nidogen-2-positive vCAFs are associated with vascular development and may originate from a pool of tumourinvading perivascular cells. Furthermore, a subpopulation of vCAFs, named cycling CAFs (cCAFs), was highly proliferative, indicating differential cellular states within subpopulations, warranting further investigation.

Conclusion and outlook
Functionally distinct fibroblast subpopulations have now been identified in almost every organ and significantly advanced our understanding of how fibroblasts are able to orchestrate highly diverse biological processes within tissues. This observed fibroblast heterogeneity reflects a combination of differences in tissue state, regional/anatomical variations, local heterogeneity (microenvironment), and cellular state [2]. While some of these cell clusters may be highly transient and/or context dependent, relationships between heterogenous populations and their abundance often accompany and

616
T Bensa, S Tekkela et al influence disease progression and response to therapy in fibrosis and cancer. In skin, it is becoming clear that there are at least four major fibroblast subpopulations ( Figure 1); however, the plasticity and dynamic interplay between functionally distinct subpopulations remain poorly understood. In addition, specialised fibroblast compartments, such as the DP or DS, appear to harbour an astonishingly diverse and highly organised heterogeneity. The subcutaneous fascia may also show a more complex organisation, with three distinct fascia fibroblast clusters that differ in abundance during skin development and wound healing [17].
Single-cell sequencing studies of aged skin revealed that fibroblasts possess the highest level of transcriptional variability of all skin cell types, resulting in a partial loss of fibroblast subset identity (in particular papillary phenotype) and increased expression of genes encoding pro-fibrotic regulators, adipocyte markers, and skin aging-associated secreted proteins [11,59,86].
Characterisation of specific subsets in fibrotic disease conditions is now paving the way for the development of highly selective targeting strategies of pathological subpopulations. In particular, comparison of regenerative and scarring wounds has enabled identifying key subpopulations, signalling interactions and molecular regulators that determine tissue repair fate and may lead to promising anti-scarring therapies. However, the cell type/subpopulation specific interplay between Wnt and Hh signalling during wound repair highlights the challenges faced when developing pathway-/subpopulationspecific treatment strategies [45,52].
Detailed time course analysis of skin development, wound healing, and cancer are emphasising the importance of the tissue microenvironment in determining fibroblast identity, fate, and plasticity [41,87,88]. Besides cell-cell interactions, mechanical cues are a major regulator of fibroblast behaviour, and modulation of their local mechanical environment could offer more selective targeting strategies. Disruption of stromal mechanical forces by FAK depletion during wound healing inhibited scarring [89] and shifted CAF subpopulation distributions in cancer, affecting tumour growth [73].
During wound healing, adipocytes differentiate into myofibroblasts, which can themselves become adipocytes. Similarly, the perivascular derived Gli1+ cells are able to contribute to wound repair long-term, and multiple fibroblast subsets can differentiate into pericytes [32,33]. Whether these cells maintain a memory of their origin and how this impacts their functional heterogeneity warrants further investigation. There is accumulating evidence that fibroblast memory influences their response to therapy. For example, in addition to eliminating or modulating pathological fibroblast subsets in wounds, local transplantation of non-scarring fibroblasts prevented tissue scarring [90]. Pharmacological inhibition of inflammatory priming signals in reindeer skin is able to inhibit pro-fibrotic fibroblast behaviour during tissue repair [21].
Like wound bed fibroblasts, CAFs are highly dynamic, and their phenotype can change depending on their spatial and biochemical niche within the tumour microenvironment. Identified CAF subtypes share many similarities with disease-associated fibroblasts in fibrosis [91]. The common expansion of a mesenchymal fibroblast subpopulation associated with high plasticity may actively support the increased cellular heterogeneity observed in fibrosis and cancer by promoting transdifferentiation between subsets during disease progression. In a transgenic breast cancer progression model, single-cell RNAseq revealed that the transcriptional composition of CAFs changes dynamically from a 'wound healing' and 'immunoregulatory' signature in the early stage to an 'ECM remodelling', 'antigen presentation', and 'protein folding function' in advanced tumours, with an additional inflammatory subset appearing in lung metastasis [88]. Notably, this conversion towards pro-tumorigenic CAFs appears to be reversible, so CAF subtype reprogramming, switching pro-tumoral CAFs to tumour constraining CAFs, could offer a therapeutic opportunity for cancer treatment. Indeed, pharmacological JAK/STAT signalling inhibition was able to shift the iCAF subtype towards an ECMproducing myCAF population, leading to an increase in the myCAF/iCAF ratio and ECM deposition and to a reduction in tumour cell proliferation and growth in a murine PDAC model [92]. Metabolic reprogramming of myofibroblasts in fibrosis through glycolytic inhibition is another exciting new treatment approach [93,94].
While most fibroblast research has focused on fibrosis and cancer, functionally distinct subsets are also being increasingly recognised as key players for disease progression in other conditions. A recent single-cell RNA-seq study of lesional atopic dermatitis identified a distinct fibroblast-immune cell interaction between a previously undefined pro-inflammatory COL6A5+, COL18A1+ fibroblast subpopulation and LAMP3+ dendritic cells [15]. With more and more disease-targeted single-cell transcriptomic studies emerging, it is becoming apparent that we might still be at the beginning of appreciating the implications of fibroblast heterogeneity and associated cell interactions in tissue homeostasis and disease. Therefore, the challenge remains to identify the relevant fibroblast subpopulations and develop selective targeting strategies for therapeutic applications.