Characterisation of human hair follicle development Authors: Zoe R. Sudderick, James D. Glover, Cameron Batho‐Samblas, Barbara Bo‐Ju Shih, Denis J. Headon Published: 2026 ABSTRACT AbstractHumans have a characteristic distribution of hair across the body. Visible, relatively long and thick terminal hair fibres are present on the scalp and eyebrows in childhood, and are stimulated to grow on other parts of the body, such as the beard and armpits, by hormones during puberty. The short and fine vellus hairs, in contrast, are not readily visible and cover most of the body, including the face. Here we report quantification of the timing and characteristics of hair follicle development in human embryogenesis, from gestational Weeks 8–19, and compare this to mouse hair follicle development. We find that human hair follicles develop first on the head, where we identify several distinct initiation sites, followed by the torso. Clustered follicular units are produced by secondary hair follicles budding off the epidermis from sites close to earlier‐forming primary follicles. Arrector pili muscles are detected from gestational Week 17. Although terminal and vellus hair follicles have clear differences in the adult, both hair types initially develop from placodes and dermal condensates of similar size. Once their development is initiated, we find that human hair follicles grow and mature at the same rate, regardless of anatomical location, but have different density at different body sites. These findings demonstrate that regional differences in hair characteristics of human skin, such as the distinction between scalp and forehead, are largely caused by processes that act after the initial stages of hair follicle morphogenesis. Efforts to understand the evolution of human ‘hairlessness’ should, therefore, focus on genetic and cellular events that take place after hair follicle morphogenesis. We compared human skin appendages, including eccrine sweat glands, with those in mouse. We found that molecular markers, such as EDA, EDAR, SOX2 and WNT pathway components, are broadly similar in expression between both species, although species‐specific differences do exist. Notably, SOX2 expression, which distinguishes different hair follicle types in the mouse, is detected in all human hair follicle types, though with a later onset of expression than that observed in mouse. Together with comparison of morphology and gene expression, these results support the use of embryonic mouse primary hair follicles as a model for human hair follicle development. INTRODUCTION Hair follicle development initiates prior to birth in both mouse and human. Hair follicles begin as an epidermal thickening, termed a placode, which acquires a dermal condensate (DC) composed of mesenchymal fibroblasts. This composite structure grows down into the dermis, driven by rapid epithelial proliferation and develops through defined stages to become a mature, fibre‐producing follicle (Lee & Holbrook, 2019; Schneider et al., 2009). This initial morphogenesis phase is followed by lifelong cycling through three different stages of variable duration: regression (catagen), resting (telogen), and regeneration (anagen), each cycle producing a new hair fibre (Botchkarev & Paus, 2003). Human and mouse skin both carry multiple different types of hair. In human development, fine unpigmented lanugo hair fibres are produced first and cover the fetus, usually being shed prior to or around the time of birth (Barman et al., 1967; Domagala et al., 2017; Verhave et al., 2022). The adult human body carries two types of hair: long, thick terminal hair fibres and tiny unpigmented vellus hairs, each present in clearly demarcated body regions, often with a sharp boundary or ‘hairline’ between them. The scalp and eyebrows produce terminal hair in children, with beard and axillary terminal hair produced under hormonal influences beginning at puberty (Randall, 2007). In contrast, in the mouse coat, the different hair types of straight guard (also called monotrich or tylotrich), awl, kinked auchene and zigzag are intermingled across the body (Dry, 1926). Mouse hair types develop in sequence through the last third of gestation (Mann, 1962), with primary guard hairs developing first, followed by secondary awl and auchene, and finally tertiary zigzag hairs (Duverger & Morasso, 2009). There is no equivalent of a lanugo hair cycle in mouse. All hair follicle types in mouse rely on active WNT/β‐catenin signalling in both epithelium and mesenchyme to form (Saxena et al., 2019). However, the primary hair follicles selectively require EDA/EDAR signalling (Headon & Overbeek, 1999), while secondary hair follicles rely instead on LEF1 (van Genderen et al., 1994) and Noggin (Botchkarev et al., 2002). FGF20 is required by both the primary and secondary waves of hair follicles and has a particular role in the recruitment of the DC (Biggs et al., 2018). The transcription factor SOX2 is a marker of the DC and dermal papilla in primary and secondary hair follicles, but is not expressed in the last‐forming and smallest tertiary hair follicles (Driskell et al., 2009), thus serving as a molecular marker to distinguish between these hair types in mouse skin. Here we describe hair follicle formation during human development at molecular and morphological levels, and compare these processes to those operating during development of the primary hair follicles in the mouse. METHODS Tissues Human fetal tissue was obtained following elective termination with informed maternal consent from the Royal Infirmary of Edinburgh. Approval for the collection of tissue was granted by the Lothian Research Ethics Committee (ref: 08/S1101/1). Gestational age was estimated by an ultrasound scan and foot length measurements. Maintenance of transgenic mouse lines was approved by the Roslin Institute Animal Welfare and Ethical Review Board and carried out under UK Home Office license (P682B81E4). TCF/Lef::H2B‐GFP (Ferrer‐Vaquer et al., 2010) mice were maintained on the FVB background. For timed matings, noon on the day of detection of a copulatory plug was assigned E0.5. Sample preparation and histology Dissected skin samples stored at 4°C for up to 24 h were fixed overnight in 4% PFA or NBF, then dehydrated through an ethanol (EtOH) series from 25% to 100% with additional clearing steps in 1:1 chloroform: 100% EtOH and 100% chloroform. Samples were then incubated in paraffin at 65°C overnight and embedded in paraffin blocks. 6 μm sections were cut on a Thermo Microtome HM325 for histology and immunofluorescence. For histology, slides were dewaxed and stained with haematoxylin and eosin (H&E) using a Leica Autostainer XL and imaged on a Nikon Ni brightfield microscope or Hamamatsu Nanozoomer XR slide scanner. Whole unprocessed fixed samples were imaged with an Olympus SZX10 stereo microscope. Immunofluorescence Paraffin sections on slides were dewaxed using a Leica Autostainer XL and immersed in citrate buffer for antigen retrieval using an Antigen Retriever 2100 (Aptum Biologics). Slides were washed with Tris‐buffered saline (TBS) containing 0.1% Tween (TBST) and 0.1% Triton, and blocked for 1 h at room temperature in 5% goat serum/TBST before primary antibodies were added and slides incubated overnight at 4°C. Sections were washed in TBST, then secondary antibody (1:500, Life Technologies A‐11035, or A‐32723) was applied for 1 h at room temperature. Slides were washed twice in TBST and once in TBS (no detergent) then stained for 30 s with TrueBlack (Biotium). Slides were rinsed with TBS and counterstained with DAPI (1:5000, Sigma) for 3 min before being mounted with Prolong Gold (Life Technologies). Imaging was carried out on a Zeiss LSM 880 confocal microscope or Leica DMLB upright microscope. For EdU staining the Click‐iT™ EdU Cell Proliferation Kit for Imaging, Alexa Fluor™ 555 dye (Thermo Fisher) was used according to the manufacturer's instructions. Following incubation with the EdU cocktail, sections were washed with phosphate‐buffered saline (PBS)/3% BSA before undergoing immunofluorescence performed as detailed above from the blocking step. For wholemount immunofluorescence, fixed tissue samples were washed in PBS with 0.5% Triton X100 (PBTx) three times for 30 min at RT then incubated in 10% goat serum/PBTx for 2 h at RT. Samples were incubated with primary antibodies (Table 1) in 10% goat serum/PBTx for 24 h at 4°C. Samples were washed three times for 45 min in PBTx at RT before incubation with secondary antibodies and/or phalloidin (Table 1) in 10% goat serum/PBTx for 24 h at 4°C. Samples were washed again three times for 45 min in 10% goat serum/PBTx then incubated in DAPI/PBTx for 30 min and mounted in Prolong Gold. Samples were imaged using a Zeiss LSM 880 microscope. TABLE 1Reagents used for protein and RNA detection in tissues.ReagentSpeciesSupplierCat. No.DilutionRetrieval Primary antibodies Anti‐Cytokeratin 14 [LL002]MouseAbcamab78001:1000CitrateRecombinant Anti‐Ki67 [SP6]RabbitAbcamab166671:200CitrateAnti‐Cytokeratin 10 [RKSE60]MouseAbcamab90251:600CitrateRecombinant Anti‐beta catenin [E247]RabbitAbcamab325721:500CitrateAnti‐α E‐catenin (G‐11)MouseSanta Cruzsc‐99881:100CitrateRecombinant Anti‐LEF1 [EPR2029Y]RabbitAbcamab1378721:500CitratePhospho‐SMAD1/SMAD5 (Ser463, Ser465) Recombinant Rabbit (31H14L11)RabbitThermo Fisher700,0471:200CitrateAnti‐SOX2RabbitAbcamab979591:500CitrateAnti‐Cytokeratin 8RabbitAbcamab594001:200CitrateAnti‐P‐cadherinRabbitInvitrogen1:1000CitrateAnti‐α‐Smooth Muscle Actin (ACTA2)MouseSigma‐AldrichA52281:500Citrate Secondary antibodies, dyes and detection reagents Goat anti‐Rabbit IgG (H + L) Highly Cross‐Adsorbed Secondary Antibody, Alexa Fluor 546GoatThermo FisherA110351:500Goat anti‐Rabbit IgG (H + L) Highly Cross‐Adsorbed Secondary Antibody, Alexa Fluor 647GoatThermo FisherA212451:500Goat anti‐Mouse IgG (H + L) Highly Cross‐Adsorbed Secondary Antibody, Alexa Fluor 488GoatThermo FisherA110291:5004′,6‐Diamidino‐2‐phenylindole dihydrochloride (DAPI)Sigma‐AldrichD95421:5000Alexa Fluor™ 647 PhalloidinThermo FisherA222871:200Opal Dye 520Akoya biosciencesOP‐0010011:1500Opal Dye 570Akoya biosciencesOP‐0010031:1500Opal Dye 620Akoya biosciencesOP‐0010041:1500Opal Dye 690Akoya biosciencesOP‐0010061:1500Click‐iT™ EdU Cell Proliferation Kit for Imaging, Alexa Fluor™ 555 dyeInvitrogenC10338 RNAscope probes Hs‐BMP2ACD430641Hs‐FGF20‐C2ACD460511‐C2Hs‐EDARACD494601Hs‐DKK1ACD421411Hs‐DKK2‐C2ACD531131‐C2Hs‐SHH‐C2ACD600951‐C2Hs‐EDA‐C2ACD535971‐C2Hs‐DKK4‐C3ACD429371‐C3Hs‐BMP7ACD424401Mm‐DKK4ACD404851Hs‐DKK4ACD429371 Reagents used for protein and RNA detection in tissues. In situ hybridisation RNA in situ hybridisation on tissue sections was performed using the RNAscope Fluorescent Multiplex Assay V2 kit (ACD biosciences) following the manufacturer's protocol. Hybridisation was done on 6 μm formalin fixed paraffin sections and detected using Opal dyes (Akoya Biosciences) (Table 1). After in situ hybridisation, samples were washed in TBST and immunofluorescence performed as described above, from the initial blocking step, but excluding TrueBlack treatment. For RNA in situ hybridisation on whole, unsectioned, specimens, tissue was fixed in 4% PFA at 4°C for 48 h then dehydrated through a methanol series and bleached with 5% H2O2 in 100% methanol. After rehydration into PBS + 0.1% Tween (PBT), samples were permeabilised with 20 μg/mL proteinase K for 20 min, washed with 2 mg/mL glycine, and post‐fixed with 0.2% glutaraldehyde. Pre‐hybridisation was carried out over 2 days at 65°C followed by hybridisation with DIG‐labelled human EDAR riboprobe (consisting of the entire open reading frame; nucleotides 280–1626 of NCBI Reference Sequence NM_022336.4) for a further 2 days. Samples were washed, blocked in 10% heat inactivated sheep serum (HISS) and incubated at 4°C with anti‐DIG antibody (1:2000, Roche) in TBST with 1% HISS overnight. Staining was detected with 5‐bromo‐4‐chloro‐3‐indolylphosphate/nitro‐blue‐tetrazolium (BCIP/NBT, Sigma). Quantification and statistics Unless otherwise stated, Fiji software was used for measurements and image analysis. Graphpad Prism 9 was used for statistical analysis and graph production. Statistical tests are named in the text and supplementary statistical information. Single‐nucleus RNA sequencing and data processing snRNAseq of the human 14 week back skin samples and the data processing of both the human and mouse datasets were carried out as in Glover et al. (2023). Human 14 week back skin snRNAseq data GEO: GSE195657. Mouse E14.5 back skin scRNAseq data GEO: GSE198487. A U score = avg_log2FC × pct.1/pct.2 was used to determine genes that were up/downregulated in the placode and DC, while U score = avg_log2FC × pct.2/pct.1 was used for the interfollicular epidermis and dermis, where pct.1 and pct.2 refer to the percentage of cells in cluster 1 and cluster 2 with detection of the transcript. Placode measurements Images from wholemount EDAR in situ hybridisations, P‐CADHERIN wholemount immunofluorescence, and H&E stained tissue sections were analysed using ImagePro Plus 6.2, Fiji, Zen 2 (Blue edition), and Hamamatsu NDP.view 2 software to measure placode diameter and density. EDAR wholemount in situ hybridisation images were processed with a linear local equalisation filter to remove background signal before measurement. Mouse measurements were taken from E14.5 TCF/Lef::H2B‐GFP (Ferrer‐Vaquer et al., 2010) mouse back skins imaged on a Zeiss Axiozoom V16 microscope, or H&E stained sections. Cell density and proliferation Cell counts were carried out using Fiji software on immunofluorescent images of sectioned skin stained with Ki67 and counterstained with DAPI. Images were taken using Zeiss LSM 880 Confocal, Zeiss LSM 710 Confocal, or Leica DMLB upright fluorescent microscopes. Images were taken at ×20 magnification and scaled appropriately. Counts were taken from the entire field of view encompassing 300–600 μm linear skin length. Damaged tissue or hair follicle structures were disregarded. To assess dermal cell density, DAPI and Ki67 positive nuclei were counted to a depth of 40 μm from the basal epidermis. Epidermal counts of DAPI and Ki67 positive nuclei were taken from the basal layer only. Skin culture and quantitative RT‐PCR Dissected skin samples were incubated on methylcellulose filters (Millipore) at 37°C in DMEM containing 5% FBS and either control (vehicle), recombinant human WNT3A (250 ng/mL, R&D Systems) + recombinant human RSPO2 (500 ng/mL, R&D Systems), or recombinant human/mouse/rat BMP2 (500 ng/mL, R&D Systems). Samples were collected after 24 h and RNA was extracted using TriReagent (Sigma‐Aldrich) according to the manufacturer's instructions. cDNA was synthesised using Superscript III Reverse Transcriptase (Invitrogen) with random primers. All qRT‐PCR reactions were carried out in 20 μL using SYBR Green Universal Master Mix (Roche) including Rox reference dye. Each reaction was performed in triplicate.Oligonucleotide sequences for quantitative RT‐PCRGeneForwardReverseBMP2TGCTTCTTAGACGGACTGCGGACCTGGGGAAGCAGCAACBMP7ACCATCGAGAGTTCCGGTTTAAGAGATCCGATTCCCTGCCEDARGTTCCTCGGTCTGTTCTCCACGGCGGAGAAGTTTTCCAAAFGF20ACAGCCTCTTCGGTATCTTGTCTTGGAGTTCCGTCTTTGTGAPDHGGTGGTCTCCTCTGACTTCAACAGTTGCTGTAGCCAAATTCGTTGTID3GCTCACTCCGGAACTTGTCATGCGTTCTGGAGGTGTCAGLGR6TTAGCTGGAACGCCATCCGGTGCATCAAGCCCCCAAGTCCLRP4CGAGTTTCCCTGCCAGAATGCCATCTTTGCAGTCGGTGTCWNT16CCCCCTCTTTGGCTACGAGCGTTGCCTGCACTGCATGACC Oligonucleotide sequences for quantitative RT‐PCRGeneForwardReverseBMP2TGCTTCTTAGACGGACTGCGGACCTGGGGAAGCAGCAACBMP7ACCATCGAGAGTTCCGGTTTAAGAGATCCGATTCCCTGCCEDARGTTCCTCGGTCTGTTCTCCACGGCGGAGAAGTTTTCCAAAFGF20ACAGCCTCTTCGGTATCTTGTCTTGGAGTTCCGTCTTTGTGAPDHGGTGGTCTCCTCTGACTTCAACAGTTGCTGTAGCCAAATTCGTTGTID3GCTCACTCCGGAACTTGTCATGCGTTCTGGAGGTGTCAGLGR6TTAGCTGGAACGCCATCCGGTGCATCAAGCCCCCAAGTCCLRP4CGAGTTTCCCTGCCAGAATGCCATCTTTGCAGTCGGTGTCWNT16CCCCCTCTTTGGCTACGAGCGTTGCCTGCACTGCATGACC RESULTS Human hair follicle initiation sites and growth Through analysis of human fetal skin sections from weeks 10–19 estimated gestational age (EGA), we identified stages of hair follicle morphogenesis in humans that match those reported in mice (Figure 1a) (Paus et al., 1999; Saxena et al., 2019), as expected (Lee & Holbrook, 2019). To define the locations on the body at which hair follicle development initiates we used whole mount RNA in situ hybridisation to visualise expression of the hair placode marker EDAR, together with histological assessment of the skin. We first observed punctate, regular, and bilaterally symmetrical spots of expression of EDAR at 10 weeks EGA on the medial part of each eyebrow, close to the nose, and the lateral edges of the upper lip. At 11 weeks, the organised array of spots on the eyebrows had extended both laterally and medially, filling the space between the eyebrows and the formation of placodes on the upper lip had also extended medially, towards the philtrum. New areas of punctate EDAR expression were present bilaterally below the eyes on the upper cheeks, and at the lateral edges of the lower lip. By 12 weeks, hair placodes had extended to cover the cheeks and filled the upper lip. New areas of placode initiation were also observed on the scalp, temples, forehead and chin, thereby covering most of the face (Figure 1b). The early initiation sites of supraorbital eyebrow, the chin and upper and lower lips correspond to clearly demarcated sites of eyebrows and beard in adults. The observed discrete placode initiation sites below each eye, on the other hand, are not associated with any obvious feature present in adult skin. Areas of punctate EDAR expression were always symmetric on the right and left of the head (Figure 1c). Hair follicle initiation and development in human skin. (a) Developmental stages of human hair follicles from H&E stained fetal skin sections. (b) Detection of EDAR transcripts by wholemount RNA in situ hybridisation of human fetal skin showing sites of hair follicle placode initiation. (c) Schematic of hair follicle placode initiation sites and corresponding estimated gestational age (EGA) on the human head. (d) Wholemount immunofluorescence detecting P‐CADHERIN (PCAD) on week 13 back and front skin. Hair placodes are indicated by white arrowheads. (e) Quantification of the most mature hair follicle morphogenesis stage observed by EGA for eight areas of skin. (f) Brightfield images of hair fibre pigmentation. (g) Brightfield images showing local hair fibre orientation. (h) H&E‐stained section showing follicles of different developmental stage in close proximity. (i) Wholemount human scalp skin stained with phalloidin (to detect F‐Actin) showing formation of follicular units with groups of hair follicles of different developmental stages present. More mature central hair follicles are indicated with arrows, younger accessory follicles with arrowheads. Each follicular unit is indicated by different coloured arrows/arrowheads (green, purple, and white). (j) Transverse section of 19 week EGA scalp skin showing variable grouping of hair follicles forming follicular units. Inset shows a primary follicle flanked by two smaller accessory follicles. (k) Longitudinal section of a hair follicle unit in week 17 scalp skin. Inset shows the separate attachment of the two adjacent follicles of different sizes to the epidermis. DC, dermal condensate; PC, Placode. Scale bars: a, h, i, j, k = 100 μm; b = 250 μm; d = 50 μm; f, g = 500 μm. The first hair follicles produced in mouse are the sensory vibrissae on the face. Primary hair placode formation then initiates at the flanks, adjacent to the mammary gland rudiments under the forelimbs, and spreads as a rapid wave dorsally and posteriorly (Painter et al., 2021). In contrast, in human development we find that on the torso the placodes form on the front (ventrum) before the back (dorsum) (Figure 1d). By quantifying the most advanced hair follicle morphogenesis stage observed for each area of skin we find that all regions undergo downgrowth and maturation at the same rate, with follicles taking approximately 6 weeks to reach stage 10 and display hair fibres emerging from the skin surface. As hair follicles on all skin regions grow and mature at the same rate, differences in hair follicle stage arise only from earlier or later initiation of hair follicle formation in a given region (Figure 1e). Pigmentation of hair fibres was first observed at 19 weeks in the eyebrow, upper lip and forehead, which is expected as these follicles are the earliest to appear and thus the most mature at this age (Figure 1f). Orientation of lanugo hair fibres varies between areas of the face but appears to be related to the specific topography of a region, such as the whorls often present at the corners of the lower lip and the changes in orientation across the medial cleft of the upper lip (Figure 1g). We observed continuous formation of placodes in all areas of human fetal skin analysed, with some new placodes observed at all ages, resulting in different stages of follicles being observed in close proximity (Figure 1h). In human skin, hair follicles are arranged into groups called follicular units (Headington, 1984). This arrangement of a mature central follicle and flanking less mature follicles can be observed from Week 17 on the scalp (Figure 1i–k) with a large, presumably earlier‐forming, central follicle flanked by a variable number of 1–3 smaller, less developed secondary follicles (Figure 1j). Such follicular units also form on other parts of the body, and the secondary follicles most frequently appear to arise independently of the structure of the primary follicle, as sections show that they are attached to the epidermis close to the primary follicle rather than budding directly off it (Figures 1k and S1). Follicles that comprise a follicular unit share a single arrector pili muscle (Poblet et al., 2002), a smooth muscle that can draw the hairs into an erect state and cause ‘goosebumps’ in the skin. We find that arrector pili muscles (APMs) are clearly present in scalp skin by Week 17 of development, visible on sections as a long trunk of Smooth Muscle Actin (SMA)‐positive cells, with branched, splayed fibres at both ends (Figure S2). At Week 16 we did not detect coherent APMs, though there were scattered SMA positive cells in the dermis that could serve as precursors or contributors to these structures (Figure S2d). Skin regional differences in human hair follicle characteristics In mice, different hair follicle types can be distinguished by several features. Primary hair follicle placodes have a larger diameter than those of the later‐forming secondary and tertiary follicles, as well as a greater hair fibre width, lower placode density, and a longer mature follicle (Duverger & Morasso, 2009). Whisker follicle placodes are larger at inception than the primary hair placodes, and the whiskers are longer and thicker than the guard hairs that the primary hair follicles produce (Mou et al., 2008). Thus, in mouse, the size of the placode upon its formation is closely related to the size of the hair follicle and hair fibre that it produces. This could occur through inductive templating, whereby placodes recruit an underlying DC in proportion to the former's area via epithelial signals such as FGF20 and SHH (Glover et al., 2017; Huh et al., 2013; Qu et al., 2022). To determine whether human hair follicle primordia differ in size in different regions of the skin, potentially corresponding to the distinct sizes of the hair fibres at these locations in postnatal skin, we assessed early hair placode diameter from H&E sections and compared these to mouse back skin primary follicles. We found that these range from 26 to 94 μm, with a mean placode diameter for each region of approximately 50 μm. No statistically significant difference in placode size was detected between human skin regions or between the sizes of human and mouse primary follicles (Ordinary one‐way ANOVA, p = 0.5163, F (8,141) = 0.9026) (Figure 2a). Thus, the size of the hair placode does not predict skin region differences in hair characteristics in human development. Morphometrics of the developing human hair follicle and skin. (a) H&E stained early hair placodes from different developmental stages and skin regions (left). Placode diameter quantification in mouse and human skin (right). Embryonic stages as indicated. (b) Wholemount immunofluorescent detection of P‐CADHERIN (PCAD) on 14 week EGA front and chin skin (left). Quantification of placode density from eight human skin areas and mouse back skin (right). (c) H&E stained section of 16 week eyebrow skin showing hair follicle cross‐sections. Arrowheads indicate hair fibres (left). Hair fibre width on different skin areas coloured by age (right). (d) DAPI staining showing a dermal papilla (DP) and a dermal condensate (DC); dotted lines outline the area measured (left). Quantification of dermal condensate area on tissue sections by hair follicle stage for eight areas of skin; each dot represents the mean for each skin area (right). (e) H&E stained eyebrow sections showing sebaceous glands attached to a hair follicle (above). Mean sebaceous gland area from tissue sections of eight areas of skin at different ages (middle). Number of sebaceous glands detectable per mature hair follicle on different areas of human skin (below). (f) Quantification of cell density and proliferation (defined by Ki67 immunofluorescence) in the epidermis and dermis at different ages and across anatomical sites. (g) Immunofluorescent detection of Ki67 in unpatterned skin, an early hair follicle, and a later stage follicle (left). Quantification of proliferation in early hair follicles and in the upper (distal) or lower (proximal) half of the epithelial component of follicles at stages 5–7. Each dot represents an individual hair follicle (right). (h) Fluorescent detection of incorporated EdU and of Ki67 protein showing proliferating cells in hair follicle DCs (arrowheads). A dotted line indicates the epidermal‐dermal boundary. For graphs each dot represents an individual specimen unless otherwise stated. Scale bars: a, c, d, e, g, h = 50 μm; b = 100 μm. We find that the density of human hair placodes varies across the body (Ordinary one‐way ANOVA, p = <0.0001, F (8,56) = 8.872), with the eyebrow and chin showing the highest densities and the trunk the lowest. We did not identify a consistent relationship between developmental stage and placode density, suggesting continued insertion of new hair placodes as the skin expands with growth across the developmental ages examined (Figure 2b). We found that mouse back skin primary hair follicles are laid out at approximately the same density as follicle primordia in human fetal back skin (Tukey's multiple comparisons test, p = >0.9999). Unlike placode size, hair fibre diameter does display some significant differences between skin regions (Ordinary one‐way ANOVA, p = 0.0127, F (7,28) = 3.203). Hair fibres on the eyebrow and upper lip are the largest, potentially as these are the first areas to form follicles and so are the most developed at the stages of analysis. The chin, however, is the next area to form follicles and shows a smaller fibre width than the scalp and forehead, which develop follicles slightly later. Follicles on the torso show the lowest hair fibre width, with back similar to that of cheek and chin (Figure 2c). Smaller fibre widths in the back, front and cheek compared with the coarser fibres of the eyebrow and scalp are consistent with these areas forming only vellus hairs in adult life. The larger fibre widths observed in the forehead are, however, not consistent with this relationship as the forehead forms only vellus hairs in adult skin. Given the importance of the dermal papilla in defining hair follicle characteristics (Jahoda, 1992; Van Scott & Ekel, 1958), differences in the size of its embryonic precursor, the DC, could underlie later‐emerging differences in size between human hair follicle types. However, we find that DC size on tissue sections increases steadily as the follicles grow through stages of morphogenesis, with significant differences only observed between the regions with the smallest (front skin) and two largest (eyebrow and scalp) DC areas (Tukey's multiple comparisons test, Front vs. Eyebrow p = 0.0471, Front vs. Scalp p = 0.0408) (Figure 2d). These modest differences in DC and placode size suggest that neither component of the early hair follicle anlagen can predict whether a follicle will later produce a terminal or a vellus hair, but instead that the stark distinctions in follicle size between vellus and terminal hairs emerge later in development, perhaps related to the first hair cycle, which occurs in utero (Barman et al., 1967). The epithelial placode undergoes downgrowth into the dermis to produce the entire pilosebaceous unit of hair follicle and associated sebaceous gland. Postnatally, sebaceous glands are larger and more densely packed on the face and scalp than the torso or other body sites, making skin at these sites oilier (Tsatsou & Zouboulis, 2014). At the developmental stages examined we did not detect a statistically significant difference in the size or density of sebaceous glands between different areas of skin (ordinary one‐way ANOVA, size p = 0.5260, F (7,34) = 0.8887, and density p = 0.1393, F (7,34) = 1.712), when measured as mean sebaceous gland area, or as the number of sebaceous glands in relation to the number of mature hair follicles observed per unit length of skin (Figure 2e). The placode and DC are formed from cells recruited from the basal epidermis and upper dermis, respectively. We quantified basal epidermal and upper dermal cell density and proliferation in different body sites and ages (Figure 2f) using the proliferation marker Ki67 to determine whether there are distinct patterns of cell division associated with hair follicle initiation and development. When comparing areas of skin, we detected significant differences in dermal and epidermal cell density and in the proliferation of the dermis (ordinary one‐way ANOVA: epidermal cell density p = <0.0001, F (7,68) = 5.777, dermal cell density p = <0.0001, F (7,68) = 6.275, dermal proliferation p = 0.0004, F (7,67) = 2.170) while no significant difference was found between regions in the proliferation of the epidermis (Ordinary one‐way ANOVA, p = 0.6982, F (7,67) = 0.2011). Although significant differences in proliferation and cell density of the epidermis and in proliferation of the dermis were detected when comparing across different ages, no simple patterns related to hair follicle formation could be discerned (ordinary one‐way ANOVA: epidermal cell density p = 0.0204, F (10,68) = 2.322, epidermal proliferation p = <0.0001, F (10,67) = 5.623, dermal cell density p = 0.0525, F (10,68) = 1.954, dermal proliferation p = <0.0001, F (10,67) = 4.749). In general, basal epidermal cell density increases while proliferation decreases as the skin matures, but we did not observe a discrete change coincident with the onset of hair follicle formation in different skin regions. Upper dermal cell density remains quite constant across skin development while proliferation decreases slightly, but, as observed for the epidermis, there is no discrete change in gross cell density at the time of follicle formation. This suggests that, unlike the events during feather formation (Ho et al., 2019), there is no clear demarcation of dense and loose dermis into tracts that might define the characteristics of different human hair follicle types. However, the eyebrow and beard (chin and upper lip) regions do have a high density of dermal cells at the time of hair follicle formation, compared with other skin regions (Figure 2f), and these areas may be comparable to the feather tracts in avian skin. More broadly, the epidermal and dermal cell density of skin of facial regions (cheek, chin, eyebrow and upper lip) appears to be higher than that of the trunk. Within developing hair follicles, the number of proliferating cells is greater in the lower half of the epithelial component of the follicle (stage 4+) than the upper half (Tukey's multiple comparisons test: p = 0.0038). The percentage in the lower half of these (stage 4+) follicles is also significantly greater than that of stage 3 and below follicles (Tukey's multiple comparisons test: p = 0.0050) (Figure 2g). Proliferation in developing human hair follicles thus occurs mainly in the epithelial cells at the leading edge of the growing follicle closest to the DC, as in mouse (Saxena et al., 2019). In the DC itself, approximately 40% of follicles examined in sections showed at least one Ki67‐positive cell, most of which were present at the periphery of the condensate (Figure 2h). Epidermal differentiation in relation to hair follicle formation At the time of hair placode formation the epidermis is multilayered, with periderm as the most superficial layer, at least one suprabasal keratinocyte layer beneath it, and the more densely packed basal keratinocytes as the lowest layer. The periderm is a continuous layer of nucleated cells at Week 14, these cells being larger than the suprabasal or basal keratinocytes. Periderm cells have complex actin cytoskeletal structures at this stage, focalised in particular on the basal side of the cells, but, unlike the basal and suprabasal layers of the epidermis, the periderm cells do not express α‐catenin (Figure 3a). Timing of epidermal differentiation relative to hair follicle development. (a) Wholemount immunofluorescent detection of α‐catenin and F‐Actin (stained with phalloidin) in different layers of 14 week back skin. Orthogonal projection, 3 μm z‐plane intervals. (b) Immunofluorescent detection of KRT14 and β‐catenin, and H&E staining, showing apical localisation of nuclei in the basal epithelium, with arrowheads indicating the cytoplasm in the basal side of the epithelial cells. (c) Immunofluorescent detection of Keratins (KRTs) 8, 14 and 10 in human back and scalp skin at 10–19 weeks EGA. Arrowheads indicate isolated positive cells; asterisks indicate autofluorescent blood cells. (d) Immunofluorescent detection of Keratins 8, 10 and 14 in early and late human hair follicles, and KRT14 in mouse early and late primary hair follicles. Ap, apical; Ba, basal. Scale bars: a, c, d = 50 μm; b = 25 μm. Some sections of human skin displayed a clear gap between epidermis and dermis, visible on H&E stained and immunofluorescent tissue sections (Figure 3b). This apparent gap arose from basal epidermal cells being columnar, with apically located nuclei and a large basal cytoplasm. We observed this arrangement of basal epidermal cells at many skin sites at weeks 10 and 12 EGA, including around early hair placodes, but very rarely at later ages, and never on volar skin. We did not observe this configuration of epidermal cells in mouse skin. We assessed the first appearance of specific epidermal keratins across the face, head and torso to assess relative rates of epidermal maturation and their timing relative to hair placode development. Keratins 8, 14 and 10 (KRT8, KRT14, KRT10) were assessed as markers of simple epidermis/periderm, basal epidermis and intermediate/suprabasal layers, respectively. KRT8 was detected as a continuous suprabasal layer in the back and scalp from 10 weeks EGA, becoming restricted to the periderm, and then lost as the periderm regresses. KRT14 is first detected in occasional isolated individual cells beneath the KRT8 expressing layer, at 10 weeks in the scalp (Figure 3c) and volar skin on the distal digit (Glover et al., 2023). Similar sporadic, isolated KRT14 + ve cells are present at 11 weeks in the front skin (Figure S3) and 12 weeks in the back skin. The number of KRT14‐expressing cells steadily increases in the basal layer until Week 14, when all regions examined had a continuous KRT14‐expressing basal layer. KRT10 expression is also detected in suprabasal epidermis of the scalp and back skin from 12 weeks, with more layers of KRT10‐positive cells observed as the epidermis thickens (Figure 3c). Thus, there are relatively small differences in epidermal maturation between different regions of hair‐bearing skin, but KRT14 expression broadly follows the same order as the development of hair follicle placodes, first the scalp, followed by the front, then the back skin. In the vicinity of the hair follicles, KRT8 is not found within the follicle epidermis while KRT10 is found in the suprabasal layers above the developing placode and later marks an internal core of follicle epidermis in the infundibulum near the exit point of the shaft (Figure 3d). In human skin there is a significant decrease in the intensity of KRT14 expression in placodes and hair follicles compared with the interfollicular epidermis. This contrasts with mouse skin development, where KRT14 expression remains high in the developing follicles (Figure 3d). Intercellular signalling across the period of follicle development In areas of skin where hair follicles develop, immunoreactivity for LEF1, a component and a readout of the canonical WNT pathway (Gupta et al., 2019), was stronger in epidermis than in the dermis. In general, as skin development proceeded across stages the dermal LEF1 signal reduced while the epidermal signal intensified. Dermal LEF1 was greater in facial skin, particularly in the upper dermis, than in skin of the scalp and the torso (Figure 4a). Of all body sites examined, the torso skin consistently had the least dermal LEF1 expression. The future beard and eyebrow regions had a high density of LEF1‐positive, presumably WNT‐activated, dermal cells, and these areas may be comparable to the high cell density dermal tracts of avian skin, which are also identifiable by increased dermal LEF1 expression (Figure S4). LEF1 expression is strong in the early forming placode before it becomes restricted to the leading epithelial edge of downgrowing follicles. The DC exhibits LEF1 expression at levels greater than that of the surrounding dermis (Figure 4a). Expression of intercellular signalling factors, signal transducers, and transcriptional regulators during human hair follicle development. (a–d) Immunofluorescent detection of (a) LEF1, (b) β‐catenin, (c) phospho‐SMAD1/5, and (d) phospho‐ERK1/2 in unpatterned skin, early hair placodes, and later stage follicles. LEF1 is detected in developing hair follicles and there is strong dermal LEF1 immunoreactivity in eyebrow, upper lip, and chin, but not in scalp, forehead, or back skin prior to and during early follicle formation. (e‐i) RNA in situ hybridisation detecting expression of (e) EDA and EDAR, (f) EDAR and FGF20, (g) BMP2 and SHH, (h) DKK1 and DKK2, and (i) BMP7, in unpatterned skin, early hair follicles, and later stage follicles. (j) Immunofluorescent detection of SOX2‐positive dermal condensates and dermal papillae in human terminal and vellus hair areas, and mouse primary hair follicles. (k) Immunofluorescent detection of mouse and human later stage follicles showing SOX2 localisation in dermal condensates. (l) Immunofluorescent detection of SOX2 in human and mouse early hair follicle primordia, showing earlier SOX2 expression in cells that are forming mouse dermal condensates as compared with human, where expression is detected only after condensate compaction. (m) Wholemount immunofluorescent detection of SOX2 in TCF/Lef::H2B‐GFP mouse back skin and fetal human back skin showing earlier SOX2 expression in mouse dermal condensates than human. Arrow indicates advanced hair follicle primordium with SOX2 expression in the dermal condensate. Arrowhead and boxed area indicate region shown in lower panels, where an incipient hair follicle primordium has a condensate lacking SOX2 expression. a, b, e, f, g + i = asterisks indicate appearance of autofluorescent blood cells. Scale bars: a–i = 25 μm; j‐m = 50 μm. β‐catenin expression is patchy in the epidermis of both trunk and scalp at 10 weeks EGA. This becomes continuous across the entire epidermis by 12 weeks, with strong signal in the basal layer and weaker signal in the suprabasal layers. Expression is consistently strong in the hair follicle epidermis, with evidence of nuclear localisation in the placodes and DC of early follicles on the torso (Figure 4b). BMP signalling inhibits hair follicle formation in mouse (Botchkarev et al., 1999; Cheng et al., 2014; Mou et al., 2006). We visualised BMP signalling locations by detecting the active, phosphorylated form of the BMP signal transducer SMAD1/5. This was present in both the epidermis and dermis at all areas and locations assessed. Phospo‐SMAD1/5 (pSMAD1/5) was decreased in the developing placode, with low epidermal detection persisting as the follicle develops, while it was lost from the DC as it becomes a dermal papilla (DP) (Figure 4c). Phospho‐ERK1/2, which is stimulated in response to several growth factors including EGF and FGF, was detected in the developing epidermis and in mesenchymal cells surrounding early follicles on the scalp, but in more mature follicles phospho‐ERK1/2 was detected in the upper/distal basal epithelium likely to be the developing outer root sheath. No signal was detected in the DC at any follicle stage examined, though dermal phospho‐ERK1/2 was detected surrounding the condensates of late stage follicles (Figure 4d). Prior to placode formation, EDAR transcript is detected at low levels throughout the epidermis, before becoming highly enriched in the developing placode. As hair follicle development continues, EDAR expression becomes restricted to the leading edge of the downgrowing follicle (Figure 4e,f). EDA transcript was detected within the placodes, with limited expression in the remainder of the epidermis and in the dermis both before and during placode formation (Figure 4e). Expression increases in the later stage follicles with the most intense signal present at the leading edge of the epidermal component. No expression was observed in the initial DC, but EDA transcript was detected in the condensate of more mature stage follicles as well as in the developing dermal sheath. For EDAR, these results match those reported during mouse primary hair follicle development (Headon & Overbeek, 1999), but EDA expression differs in that it is reduced in expression in mouse primary hair placodes (Laurikkala et al., 2002) and in feather primordia (Ho et al., 2019), but we find it to be intensified in expression in human hair follicle placodes. FGF20 expression was rarely detected in the skin before the formation of follicles and is expressed selectively in the hair placodes as these emerge. FGF20 expression decreased as follicle downgrowth progressed (Figure 4f). BMP2 and SHH expression were detected in a similar pattern in the initiating placodes, and these genes remain expressed in the more proximal epithelial component of the downgrowing hair follicles (Figure 4g). At the time of hair follicle formation DKK2 is detected broadly at low levels throughout the dermis, but not in the epidermis. DKK1 is observed more sparsely in the dermis, increasing in the DCs and then the DPs of the developing hair follicles (Figure 4h). BMP7 was detected in the epidermis before follicle formation, with expression intensifying in both epidermal placode and DC. In later follicles, very strong expression is seen in the DC with slightly less in the epidermal compartment of the follicle (Figure 4i). These results are generally consistent with the expression profiles of these genes during mouse primary hair follicle formation (Saxena et al., 2019). The transcription factor SOX2 is a marker of the DC and DP of primary and secondary hair follicles, but not tertiary follicles, in mouse (Saxena et al., 2019). We identified SOX2 immunofluorescence in the DCs of hair follicles in all areas of skin and all ages examined in human (Figure 4j), therefore, SOX2 does not mark different follicle types in human as it does in mouse. SOX2 expression is uniformly distributed across the entirety of mouse primary DCs, whereas in human hair follicles, SOX2 was expressed in the core of the DC, with a layer of SOX2‐negative cells at the periphery (Figures 4j,k and S5). When compared with developing mouse skin, the DCs emerging under human hair follicle placodes appear to be more loosely arranged, and SOX2 was not detected in the very early stages of condensate development (Figures 4l,m and S6). Occasional SOX2‐positive Merkel cells are observed in the interfollicular epidermis, appearing first in the facial regions at Week 12 and the torso by Week 16 (not shown). Comparison of gene expression in developing hair follicles of mouse and human fetal skin To compare early human hair follicle development to that of the primary hair follicles in mouse, we analysed existing single‐nucleus RNA sequencing (snRNAseq) data of Week 14 human back skin (Glover et al., 2023) and single‐cell RNA sequencing (scRNAseq) of E14.5 mouse back skin (Qu et al., 2022). Cell/nucleus clusters representing the placode, interfollicular epidermis (IFE), DC and general dermis were identified in both datasets. Expression of BMP7 and INHBA identified the DCs; PTCH1 the general dermis; EDAR and FGF20 the placode; and KRT14 (without high EDAR or FGF20) the basal interfollicular epidermis (Figure 5a–d). Using the same thresholds as reported in Sulic et al. (2023) and Sennett et al. (2015) to define significantly upregulated genes (Fold change > 1.5, adjusted p‐value < 0.05), we identify many known markers of placodes and DCs in both mouse and human. However, key markers such as Shh in the placode and Sox2 in the DC were not found to be significantly upregulated in mouse placode or human DC using this simple threshold (Figure 5e–h). We therefore assessed whether genes were significantly up or downregulated in placode/condensates relative to the IFE/general dermis, taking into account both the level of expression and the proportion of cells in which the gene is expressed, which we refer to as uniqueness (U) score (Glover et al., 2023). We assessed genes with a U score > 1.5 in both mouse and human, and identified 18 genes as placode‐enriched and 39 as DC‐enriched (Supporting Information Tables), with the top 5 enriched genes and two known markers shown in Figure 5i,j. All mouse placode‐enriched genes we identified in the mouse dataset were also identified as placode‐upregulated in mouse by Sulic et al. (2023) and some by Sennett et al. (2015), with many of those in the DC also identified by Sennett et al. (2015). Transcriptome profiling of early mouse and human hair follicles. (a) Unbiased clustering of snRNAseq from human 14 week back skin. Identified clusters are placode, interfollicular epidermis (IFE), dermal condensate and dermis. Here dermis is defined as those fibroblasts closest in expression profile to the dermal condensate (Figure S7). (b) Feature plots from human data of signature genes used to identify the placode, IFE, dermal condensate and dermis. (c) Unbiased clustering of scRNAseq from published E14.5 mouse data (Qu et al., 2022). Identified clusters are placode, IFE, dermal condensate and dermis. Here dermis is defined as those fibroblasts closest in expression profile to the dermal condensate (Figure S7). (d) Feature plots from mouse data of signature genes used to identify the placode, IFE, dermal condensate and dermis. (e–h) Volcano plots of genes expressed in the epidermis and dermis of human and mouse. The top five genes upregulated in the placode/dermal condensate (dark blue) or downregulated (light blue), and known marker genes of placode, basal epidermis, and dermal condensate, are indicated. Significance thresholds correspond to the fold change >1.5 or <−1.5 and adjusted p‐value <0.05. (i, j) Tables of enriched genes shared between mouse and human during early hair follicle development as determined by U score (U = avg_log2foldchange × proportion of detection in cluster1/proportion of detection in cluster2). Shared upregulated genes were defined as those with a U score > 1.5 in both species. (i) Genes with high expression in placodes compared with basal epidermal cells. Results for E14.5 mouse skin from other studies shown in right of table. (j) Genes with high expression in dermal condensate cells compared with general dermal cells. Results for E14.5 mouse skin from other studies shown in right of table. (k–n) RNA in situ hybridisation detecting DKK4 transcripts in early mouse and human hair follicles; (k) a primary hair follicle from uncultured mouse embryonic day 17 (E17) back skin, (l) early hair follicles found on uncultured 14 week human back and front of torso skin, (m) an early placode (boxed area) on untreated cultured human 12 week scalp skin, and (n) an early placode on human 12 week scalp skin cultured with recombinant WNT3A and RSPO2. Asterisks indicate autofluorescent blood cells. (o) RNA in situ hybridisation to detect DKK4 transcripts in dematoglyph ridges found on the ventral digit of E17 mouse and 16 week human volar skin. (p, q) Immunofluorescent detection of KRT17 in stage 2/3 placodes from (p) E14 mouse back skin, and (q) 14 week human back skin. Arrows indicate the basal epidermis and placodes, arrowheads indicate periderm cells, and chevrons indicate suprabasal cells above the basal placode. (r) RT‐qPCR of hair follicle related genes in human back and scalp skin treated for 24 h with recombinant WNT3A + RSPO2 or BMP2 proteins. Each point represents an individual piece of skin from back or scalp from 11 week (triangles) or 12 week (dots) embryos. GAPDH was used for normalisation. *p < 0.05; **p < 0.01; ***p < 0.001. Student's paired t‐test. Scale bars: k–n, p, q = 25 μm; o = 20 μm. We identified genes specifically upregulated in the human placode and DC using a two‐step process. First, with thresholds of U score > 1.5 in the human and <0.5 in the mouse, and secondly removing any genes that had been found to be upregulated in the mouse placode or DC by Sulic et al. (2023) or Sennett et al. (2015). This approach identified genes that may be expressed specifically in human, but not mouse, early hair follicle development (Tables S2 and S5). However, the human skin analysed may carry hair follicle primordia that are more developmentally advanced than the tissue from which the mouse datasets were generated. Thus, further characterisation of this candidate gene set would be required to assess whether any might have species‐specific expression patterns. The observed differences in gene expression between mouse and human could be influenced by the different analyses used; the mouse study being conducted using single‐cell RNAseq whereas the human study used single‐nucleus RNAseq. Thus, we assessed by in situ hybridisation and immunofluorescent detection selected genes with notably different U scores between mouse and human, to assess the distinctiveness of their expression patterns. Dkk4 is a widely used marker of hair placode identity in mouse (Bazzi et al., 2007) and has been proposed to be a key regulator of hair follicle spacing (Sick et al., 2006). From snRNAseq using Week 14 EGA back skin, we readily identified a placode cell cluster with high expression of known placode marker genes such as EDAR, SHH, and FGF20 (Figure 5a,b, Supporting Information Tables). However, from a total of roughly 10,000 nuclei, only a single read of the DKK4 transcript was identified. This single detection event was in a mesenchymal cell cluster, while Dkk4 expression reported in mouse skin is strictly epithelial (Bazzi et al., 2007). To assess the expression of DKK4 in developing human skin in more detail, we performed RT‐PCR experiments using freshly isolated, cultured, and WNT‐stimulated cultured human fetal skin. We found DKK4 to be detectable by RT‐PCR in all samples of human skin tested (Figure S7). The majority of these samples (from 12 and 13 weeks back and scalp) were cultured skin maintained for 20 h. Only one sample was uncultured skin (13 week front). When we used RNAscope in situ hybridisation to explore the localisation of DKK4 expression, we found abundant expression in mouse hair placodes (Figure 5k) but detected essentially no signal in developing human epidermis or hair follicle placodes at any age when uncultured (Figure 5l). When human fetal scalp skin was cultured for 20 h, a weak signal was detected in control (Figure 5m), with much higher expression in the sample treated with recombinant WNT3A and RSPO2 proteins (Figure 5n). Thus, in the absence of supplemented WNT stimulation, there was no detectable DKK4 expression in intact native human fetal hair‐bearing skin. Similar to the hair‐bearing skin, Dkk4 is also readily detected in mouse volar digit ridges, but not in human primary fingerprint ridges (Figure 5o). Unlike the mouse scRNAseq, in human hair placodes, the expression of KRT14 was downregulated compared with that of the IFE (Figures 5e,f and S7) based on snRNAseq, and also apparent by immunofluorescent detection of the protein (Figure 3d). It is possible that the KRT14 transcript and encoded protein have a long half‐life, and the slower rate of human development may permit its disappearance from placodes more completely than in the faster‐developing mouse skin. KRT17 was also differentially expressed in mouse and human epidermal placodes (Figure 5e,f), with increased abundance in mouse placode vs. IFE but not in human placode vs. IFE. Immunofluorescent detection agreed with this finding, with significantly stronger KRT17 expression in the mouse placode and suprabasal cells above the placode than in the IFE, but very slightly higher expression in the IFE than the placode of human skin (Figure 5p,q). In embryonic day 13 and 14 mouse skin EDAR, FGF20 and BMP2 have been reported to be rapidly upregulated by WNT signalling and downregulated by BMP signalling (Glover et al., 2017; Mou et al., 2006). We assessed the response of these genes to stimulation of the WNT pathway (using recombinant WNT3A + RSPO2 proteins) and BMP pathway (using recombinant human BMP2 protein). We find that WNT pathway stimulation increases abundance of EDAR, FGF20 and BMP2 transcripts on back and scalp skin at 11 and 12 weeks EGA, confirming conservation of WNT targets between species. We did not detect any decrease in EDAR expression elicited by BMP signalling, though the positive control for BMP treatment, ID3, did increase in abundance (Figure 5r). This suggests that, unlike the situation in mouse, EDAR is not a BMP target gene in cultured developing human skin. EDA, EDAR expression and BMP and WNT activity in eccrine sweat gland development Formation of eccrine sweat glands relies on EDAR signalling, causing hypohidrotic ectodermal dysplasia when mutations in EDA, EDAR or EDARADD abolish pathway function (Sadier et al., 2014). We assessed the expression of key components and transducers of the WNT, BMP and EDA/EDAR pathways in the development of human eccrine sweat glands of the volar skin. Sweat glands begin development on the volar skin (that of the ventral side of the hands and feet) from approximately 16 weeks EGA, budding off from the primary ridges. Morphologically, the eccrine sweat gland rudiments appear similar to an early hair follicle placode but are not associated with condensed dermal cells (Figure S8). Like the hair follicles, the sweat glands show strong expression of LEF1, β‐CATENIN, EDAR, and BMP2 (Figure S8), with the majority of these signals concentrated at the leading edge of the downgrowing sweat gland epithelial cord. Unlike hair follicles, there is very little expression of FGF20, increased pSMAD1/5, and no expression of SHH or SOX2 in developing sweat glands at any stage of their development. Expression of EDA is very low in the downgrowing sweat gland bud (Figure S8), unlike its prominent expression in the hair placode. Since EDA‐EDAR signalling is required for sweat gland formation in mouse and human, and sustained EDA signalling is required to generate an eccrine sweat gland in mouse (Cui et al., 2009), it is possible that the extended growth of the sweat gland in human skin draws EDA ligand from non‐epithelial sources. It is notable that sweat gland development can be stimulated in EDA‐deficient individuals through the administration of recombinant EDA protein starting at Week 26 of gestation (Schneider et al., 2018), demonstrating that the competence of skin to respond to EDA by producing sweat glands, and thus presumably EDAR expression, continues until at least this developmental stage. DISCUSSION This comparison of early human and mouse hair follicles reveals that their morphology, physical size, signalling pathway activity and gene expression are broadly comparable, supporting the use of mouse primary hair follicles as a model for human hair follicle development. The spacing and size of the hair follicle rudiments are strikingly similar between the species, suggesting similar mechanisms of embryonic pattern formation. However, the distinction between the large terminal and small vellus hair follicles in human is not apparent at early stages of hair follicle formation, and so appears to arise after their morphogenesis and the production of lanugo hair fibres (Barman et al., 1967). Expression of SOX2, which distinguishes the last‐forming zigzag hair follicles from other types in mouse (Driskell et al., 2009), does not identify future vellus or terminal hair follicles in human, suggesting that these hair types in mouse do not have analogues in human skin. Thus, the study of early embryonic hair follicle formation in mouse appears to be limited as a model to understand the evolution of the human hair types and their anatomical distribution. In humans, hormonal regulation of hair follicle size and cycling behaviour in some regions of the skin occurs throughout life, with axillary and beard hairs transitioning from vellus to terminal, and scalp hair follicles transitioning from terminal to vellus, under the influence of androgens (Randall, 2007). This illustrates the ability of human hair follicles to undergo dramatic changes of size through hair cycles. It may be that the poorly understood cellular processes that drive hair follicle enlargement or miniaturisation, which from puberty are regulated by androgen stimulation, are the same events that mediate the vellus‐terminal transition during prenatal development (Barman et al., 1967; Whiting, 2001), though at this early stage of life operating without an androgen trigger. Formation of feather buds in chicken skin has long been studied as a model for skin appendage development in general. In chicken skin, a densely cellular dermis forms tracts, which will bear feathers, while loose dermis will not (Olivera‐Martinez et al., 2004). Thus, the structure of the avian skin influences which regions will and will not be feather bearing. In mouse, however, the onset of primary hair follicle formation is not accompanied by an increase in dermal cell density (Makela & Mikkola, 2023), and we find that the density of dermal cells in human skin development does not closely align with the onset of hair follicle formation nor define areas with future terminal or vellus hair follicles. The closest parallel to the feather tracts of avian embryos that we observed during human development was the definition of a high density LEF1‐positive dermis at the very first sites of hair follicle formation: the upper lip, chin and eyebrow. These could potentially reflect an influence of dermal structure and signalling on specific hair‐bearing regions of human skin. DKK4 has been proposed to be a central factor in defining hair follicle density and spacing (Sick et al., 2006), though no Dkk4 mutant mouse hair patterning phenotype has yet been reported. In cats, a mutation in DKK4 is associated with altered coat colour patterns, while the distribution of the hair follicles themselves is not altered by this mutation (Kaelin et al., 2021). In our investigations, we did not detect DKK4 expression in native human skin undergoing hair follicle development by in situ hybridisation or single‐nucleus RNA sequencing, though we did detect the transcript by RT‐PCR and upon WNT stimulation of cultured skin. Further assessment of DKK4 expression in human skin is warranted, though the current findings suggest that this widely employed placode marker gene in mouse may not be expressed or functionally relevant in human skin development. Edar regulation by BMP signalling has also been proposed as a driver of primary hair follicle pattern formation and spacing in mouse (Mou et al., 2006). However, the key regulatory effect of BMPs in suppressing Edar expression that was observed in mouse we failed to detect in cultured human skin, suggesting that this interaction is not relevant to human hair patterning, perhaps because humans do not have a faithful analogue to the mouse primary hair follicle type. Thus, the two simple model networks proposed to explain hair follicle patterning based on mouse models (Mou et al., 2006; Sick et al., 2006) do not appear to apply in human skin, suggesting that human hair follicle spacing either relies on decisive regulatory interactions not yet observed or upon a complex interweaving of a set of regulatory pathways with multiple regulatory connections between them (Glover et al., 2017). The longer time taken to produce a pattern in human may permit the engagement of more signalling pathways and interactions in this process. In conclusion, while there are some key differences to be recognised, the development of mouse primary hair follicles serves as a good model for early human hair follicle development. However, an understanding of the evolution of human‐specific hair characteristics and human ’hairlessness’ needs to consider events that occur subsequent to initial hair follicle patterning and morphogenesis. AUTHOR CONTRIBUTIONS ZRS, JDG and DJH contributed to the study concept/design and acquisition of data; ZRS, JDG, CB, BS and DJH contributed to data analysis/interpretation; ZRS and DJH drafted the manuscript; ZRS, JDG, CB, BS and DJH critically revised and approved the manuscript. Supporting information Supplementary Figure 1: Follicular units in human fetal skin. (a) Four series of near‐adjacent transverse H&E‐stained sections from Week 19 and Week 17 scalp skin showing examples of follicle units with flanking secondary follicles joining to the epidermis separately from the primary follicle. Asterisks indicate main follicle hair canals; arrowheads indicate accessory follicles. Block arrows indicate the images in series. The secondary follicles appear to have begun development significantly later than the primaries. (b) Longitudinal H&E‐stained series of sections from week 19 neck, front of trunk (ventrum) and back of trunk (dorsum) skin showing examples of follicle units where the secondary flanking follicles attach to the epidermis independently of the primary follicle. Arrowheads indicate follicles in a unit. Scale bars: 100 μm. Supplementary Figure 2: Arrector pili muscles in fetal human skin. (a) Longitudinal H&E‐stained sections of 19 week scalp skin showing likely arrector pili muscles (APMs), indicated by arrowheads. (b) Immunofluorescent detection of smooth muscle actin (SMA), which highlights the APM, in longitudinal (top) and transverse (bottom) Week 19 scalp skin sections. The APM attaches to a region deeper than the sebaceous gland, here termed the bulge. In transverse sections of hair follicles the APMs can be seen uniformly to one side of each hair follicle unit. Open structures circumscribed by SMA detection are likely blood vessels, with expression in the vascular smooth muscle. (c) Immunofluorescent detection of SMA in longitudinal (top) and transverse (bottom) 17 week scalp skin showing APMs. The state of development is similar to that observed at 19 weeks. (d) Immunofluorescent detection of SMA in Week 16 scalp skin showing the lack of APMs at this age. Scattered SMA‐positive cells are present in the dermis, some adjacent to hair follicles, but no coherent APM is observed. Arrowheads indicate APMs. SG = sebaceous glands, B = bulge. Scale bars: 100 μm. Supplementary Figure 3: Keratin 14 is expressed in ventral skin prior to its expression dorsally. Immunofluorescent detection of Keratin 14 in 11 week human torso skin. No expression was detected in the epidermis of the back skin but was detected in the front skin at this stage of development. Arrowheads indicate Keratin 14 positive epithelial cells. Scale bar = 25 μm. Supplementary Figure 4: LEF1 detection in chicken feather tracts and the developing human facial skin. Immunofluorescent detection of LEF1, counterstained with DAPI, in transverse sections of chicken embryos and saggital sections of human facial skin at the embryonic days or weeks indicated. LEF1 is detected in the chicken dorsal (spinal) and wing (humeral) skin tracts at (a–c) embryonic day 6 (E6), (d–f) E7 and (g) E8, and in human foetal facial skin at (h–j) 10 weeks EGA, and (k–n) 12 weeks EGA. (a–g) Arrows and arrowheads indicate epidermal and mesenchymal LEF1 expression, denoting the presumptive feather tracts in these regions of the chicken embryos. Areas of epidermal LEF1 expression are indicated by arrows and strong mesenchymal LEF1 detection by arrowheads. In chicken embryonic skin strong LEF1 expression is present in the epidermis at the site of presumptive feather tracts with strong mesenchymal LEF1 in the upper dermis in the dorsal and wing feather tracts. Expression in the spaces between feather tracts is absent from the epidermis. (e, f) Enlarged image of the gap between the dorsal and wing tracts. No epidermal LEF1 is seen in the gap, and only weak mesenchymal LEF1 is detected. Mesenchymal LEF1 is present in the tracts. (h‐j) In human facial skin at Week 10, epidermal LEF1 is present across the face, particularly in the presumptive eyebrow, upper lip and lip vermilion. Low mesenchymal LEF1 is observed generally, with stronger expression in the dermis underneath the areas of highest epidermal expression. Strong epidermal and mesenchymal LEF1 is present in the eyebrow, with limited epidermal and no mesenchymal LEF1 continuing in the forehead. (k‐n) At 12 weeks, LEF1 expression in the epidermis is detected across the face with strong mesenchymal expression detected in the upper lip, and chin surrounding developing hair follicles. Dermal LEF1 appears weaker than that at 10 weeks but extends deeper into the dermis. Epidermal LEF1 is strong in the eyebrow and all hair follicles, but is slightly weaker in the forehead. Strong mesenchymal LEF1 expression is present in the upper lip, reducing to moderate in the cheek. DT = Dorsal tract, WT = Wing tract, FF = Feather follicle bud, SC = Spinal cord, NC = Notochord, BC = Brain cavity, OC = Oral cavity, EL = Eyelid, LLV = Lower lip vermilion, ULV = Upper lip vermilion, HF = Hair follicle. Scale bars: a, d, g, h, and k = 500 μm; b, c, e, and f = 100 μm; i, j, l, m, and n= 250 μm. Supplementary Figure 5: Proportion of SOX2 + ve dermal condensate cells in developing human hair follicles. The percentage of the dermal condensate that is SOX2 + ve by area is plotted by hair follicle developmental stage. All anatomical sites had SOX2 + ve cells in the dermal condensate but in no hair follicle examined did all of the dermal condensate cells express detectable SOX2. A total of 258 hair follicles were measured from 15 individual specimens. Dots indicate the mean ± SEM at each anatomical site and hair follicle stage. Supplementary Figure 6: SOX2 expression is not detected at the earliest stages of human hair follicle morphogenesis. Wholemount immunofluorescent detection of SOX2 in dorsal digit skin at 14 weeks. Various developmental stages of hair follicle are present on the skin above phalanx 1 (P1), in a cluster between the knuckle joints. The hair follicle primordia have a radial organisation, with the most advanced and deepest primordia at the centre of the cluster, giving way to progressively less developed primordia towards its periphery. From this arrangement, we infer that hair follicle formation in this region is likely to occur in a spreading wave. SOX2 expression is observed in the dermal condensates of the more mature follicle primordia, but morphologically distinct mesenchymal condensations at the periphery of the cluster, and above P2, lack SOX2 immunoreactivity, demonstrating the late activation of its expression in human hair follicle development. Boxes (i) and (ii) are expanded in the lower panels. Arrowheads indicate dermal condensations of incipient follicles that are detectable by cell arrangement (DAPI stain) but that lack detectable SOX2 expression. Scale bars as indicated.| Supplementary Figure 7: Gene expression cluster identification in snRNASeq and scRNASeq data, and DKK4 transcript detection. (a) Unbiased clustering of snRNAseq from human 14 week back skin. All clusters are labelled based on differential gene expression and previously reported markers of different cell types. Populations for comparison were defined as follows: BKII = Placode, BKI = IFE, FbV = Dermal condensate, FbIV = Dermis. (b) Feature plot for KRT10 on human back skin identifying the suprabasal keratinocyte population. (c) Feature plot for PDGFRA on human back skin identifying fibroblast populations. (d) Dotplot of placode and epithelial cell markers from human back skin. (e) Dotplot of dermal condensate and fibroblast markers in human back skin. (f) Unbiased clustering of scRNAseq from published E14.5 mouse data (Qu et al., 2022). All clusters are labelled based on differential gene expression and known markers of cell types. Populations for comparison were defined as follows: BKIII = Placode, BKI + BKII = IFE, FbIV = Dermal condensate, FbI = Dermis. (g) Feature plot for Krt10 on mouse back skin identifying the suprabasal keratinocyte population. (h) Feature plot for Pdgfra on mouse back skin identifying fibroblast populations. (i) Dotplot of placode and epithelial cell markers from mouse back skin. (j) Dotplot of dermal condensate and fibroblast markers in mouse back skin. (k) Diagram of DKK4 gene with primers used for PCR. Length of the gDNA fragment is shown with a solid green line, length of the corresponding RNA/cDNA fragment with a dashed green line. (l) PCR detecting DKK4 on different ages and areas of human foetal skin. DKK4 cDNA was amplified in all skin samples, both treated with recombinant WNT protein, and untreated. Only DKK4 DNA was identified in the gDNA and no RT (no reverse transcriptase) reactions, as expected. Supplementary Figure 8: Expression of cell signalling factors during development of volar sweat glands. (a) H&E‐stained sections showing sweat gland development on the volar skin from 16 to 19 weeks EGA. Sweat glands bud off from the primary fingerprint/dermatoglyph ridges and grow down as thin epithelial cords with a wider bulb at the leading edge and no overt dermal component. (b–f) Immunofluorescent detection of (b) Keratins 8, 14 and 10, (c) LEF1, (d) phospho‐SMAD1/5, (e) β‐catenin and (f) SOX2 in developing sweat glands. (g–k) RNA in situ hybridisation to detect expression of (g) BMP7, (h) EDA and EDAR, (i) DKK1 and DKK2, (j) EDAR and FGF20 and (k) BMP2 and SHH in sweat glands of the volar skin. Arrowheads indicate primary fingerprint ridges, arrows indicate sweat gland ducts, and chevrons indicate sweat gland bulbs. Scale bars a = 50 μm; b–k = 25 μm. Supplementary Table 1: Shared Placode‐Enriched Genes. U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for mouse and human. Supplementary Table 2: Human‐Specific Placode‐Enriched Genes. This table was generated through a two‐step process. First, U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for human, <0.5 for mouse was calculated using data from Glover et al. (2023) and Qu et al. (2022). Second, any genes found to be placode‐enriched in mouse by either Sennett et al. (2015) or Sulic et al. (2023) were not included in this table. Supplementary Table 3: Mouse‐Specific Placode‐Enriched Genes. U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for mouse, <0.5 for human. Supplementary Table 4: Shared Dermal Condensate‐Enriched Genes. U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for mouse and human. Supplementary Table 5: Human‐Specific Dermal Condensate‐Enriched Genes. This table was generated by a two‐step process. First U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for human, <0.5 for mouse using data from Glover et al. (2023) and Qu et al. (2022). Second, any genes found to be dermal condensate‐enriched in mouse by Sennett et al. (2015) were not included in this table. Supplementary Table 6: Mouse‐Specific Dermal Condensate‐Enriched Genes. U score = avg_log2FC × pct.1/pct.2. U score > 1.5 for mouse, <0.5 for human. Supplementary Table 7: Shared IFE‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for mouse and human. Supplementary Table 8: Human‐Specific IFE‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for human, >−0.5 for mouse. Supplementary Table 9: Mouse‐Specific IFE‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for mouse, >−0.5 for human. Supplementary Table 10: Shared Dermis‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for mouse and human. Supplementary Table 11: Human‐Specific Dermis‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for human, >−0.5 for mouse. Supplementary Table 12: Mouse‐Specific Dermis‐Enriched Genes. U score = avg_log2FC × pct.2/pct.1. U score < −1.5 for mouse, >−0.5 for human. Supplementary Table 13: Placode and IFE‐enriched genes identified in either the mouse (Qu et al., 2022) or human where no orthologue was found in the other species. Supplementary Table 14: Dermal condensate and dermis‐enriched genes identified in either the mouse (Qu et al., 2022) or human where no orthologue was found in the other species.