An innovative in vitro system unveils IGF1R signaling regulating Merkel cell generation

Summary

Merkel cells (MCs) are specialized mechanoreceptors crucial for tactile sensation, yet their developmental investigation remains challenging, particularly in humans, due to the lack of validated in vitro culture system. Here, we establish novel approaches, including short-term ex vivo vibrissae explants, innovative mouse skin organoids (mSKOs), and human pluripotent stem cell-derived skin organoids (hSKOs), to monitor MC development. We demonstrate that Polycomb repressive complex inhibitors (PRCis) efficiently promote MC generation in these culture systems. Through single-cell and spatial transcriptomics analysis, together with pharmacological screening, we identify IGF1R as a potential regulator of MC formation, which likely exerts its effects through the AKT pathway. Furthermore, we validate the role of FGFR2 signaling in MC generation. These systems constitute a versatile platform that harnesses complementary strengths to not only advance MC biology and skin development but also enable stem cell research, supporting organoid-based disease modeling, therapeutic compound screening, and regenerative medicine.

Graphical abstract

Highlights

• •Advanced in vitro system enables MC generation within 1 week
• •PRC inhibition enhances MC yields in mouse and human skin organoids
• •IGF1R signaling participates in MC development revealed by multi-omics
• •Ex vivo model confirms the role of IGF1R in MC generation

Beyond Merkel cells, this work presents a broadly useful stem cell toolkit: mouse/human skin organoids plus vibrissae explants that reveal and tune lineage specification. PRC inhibition accelerates sensory lineage differentiation, while multi-omics nominates IGF1R and confirms FGFR2 as manipulable pathways in MC development. The platform supports scalable culture, mechanism-guided screening, disease modeling to advance regenerative medicine, and stem cell-based therapies.

Introduction

Merkel cells (MCs), specialized mechanoreceptors residing in the basal layer of the epidermis in vertebrates’ skin, are essential to the sense of touch. They are particularly abundant in tactile-sensitive regions such as the fingertips and whisker follicles, where they convert mechanical stimuli into electrical signals and relay precise sensory information to the nervous system via neurotransmitter release (Yamada et al., 2024). The function of MC is essential for discerning subtle textural differences and enabling fine motor skills and object manipulation. Consequently, their dysfunction results in altered tactile sensitivity and has been implicated in abnormal sensations such as alloknesis (itch in response to light touch) and allodynia (pain in response to normally non-painful stimuli) (Feng and Hu, 2019; Feng et al., 2018). Beyond their well-established mechanosensory role, emerging evidence implicates MC function in neuroendocrine signaling and potential involvement in processes like nerve regeneration (Yuan et al., 2022). Therefore, a comprehensive understanding of MC biology is vital for both normal skin physiology and pathological conditions, including somatosensory disorders and skin diseases.

The molecular orchestration of MC development involves a complex interplay of transcription factors, stem cell dynamics, and signaling pathways. MCs originate from the K14⁺ basal keratinocyte layer and are further specified from progenitor populations marked by K17 and SOX9 in the developing hair follicle (Doucet et al., 2013; Nguyen et al., 2018; Van Keymeulen et al., 2009). The specification of MC during embryogenesis requires critical transcription factors such as ATOH1, SOX2, ISL1, and POU4F3 (Perdigoto et al., 2014; Yu et al., 2021), among which ATOH1 is the essential fate-decision factor interacting with other mediators. SOX2 cooperatively regulates ATOH1 with ISL1, supporting MC differentiation and maturation, while POU4F3 enhances ATOH1 activity through feedforward epigenetic mechanisms. In addition, epigenetic regulation by Polycomb group proteins (Oss-Ronen and Cohen, 2021), particularly the PRC1 (Polycomb repressive complex 1) and PRC2 (Polycomb repressive complex 2), functioning as an epigenetic regulator by catalyzing trimethylation of lysine 27 on histone H3 (H3K27me3), modulates gene expression programs essential for MC specification (Cohen et al., 2018). In particular, Polycomb subunits Ezh1 and Ezh2 regulate the MC differentiation program in skin stem cells via suppression of Sox2 (Bardot et al., 2013). Moreover, multiple signaling pathways orchestrate MC development (Xiao et al., 2014). Wnt signaling initiates a cascade that ultimately induces Shh expression in developing hair follicles, crucial for touch dome MC specification (Perdigoto et al., 2016; Xiao et al., 2016), while Shh signaling maintains MC stem cell renewal through Gli1-mediated transcription (Xiao et al., 2015). FGFR2-mediated signaling is essential for MC formation in both hairy and glabrous skin in mouse (Nguyen et al., 2019). Notably, the Notch pathway functions as a negative regulator preventing ectopic MC formation (Logan et al., 2018). Nevertheless, the crosstalk among these signaling pathways during development and the involvement of other regulatory mechanisms in MC specification remain to be fully elucidated.

The study of MC development presents significant challenges due to the intricate regulatory processes and methodological constraints. Current understanding of MC development primarily derives from in vivo studies. The research timeline is relatively extended, and certain developmentally crucial genes cannot be directly knocked out for in vivo verification and analysis. Particularly, the development of MC in humans is barely studied due to the lack of experimental model. To address these limitations, we have developed innovative in vitro tissue and organoid culture systems for MCs. This system utilizes advanced approaches, including mouse skin organoids (mSKOs) through a novel organoid assembly strategy and human skin organoids derived from pluripotent stem cells (human pluripotent stem cells [hPSCs]), substantially reducing experimental timelines for evaluating candidate signals and small-molecule compounds.

While in vivo studies remain indispensable for capturing systemic complexity, their limitations in cost, throughput, and mechanistic resolution are well-recognized. The in vitro culture system established here bridges critical gaps in in vivo research, allowing for the validation of signal pathways and factors identified from in vivo studies, discovery of novel regulators, and elucidation of their roles in MC development and maintenance. Additionally, the in vitro culture system serves as an effective platform for screening small-molecule compounds that enhance MC generation, potentially advancing therapeutic approaches for MC-associated disorders, including MC carcinoma, while contributing to developments in dermatology, regenerative medicine, and sensory neurobiology.

Results

Inhibition of PRCs promotes Merkel cell generation in mouse ex vivo tissue culture

To establish an ex vivo model for MC research, determining the initiation point of MC development is essential. We examined MC presence at different developmental time points (embryonic day [E] 13.5, E14.5, E15.5, and E16.5) in the dorsal skin and vibrissae tissues with its key identifier SOX2 and cytokeratin 8 (K8) (antibody information can be found in Tables S1 and S2 of the supplemental information). The results of whole-mount immunofluorescence (IF) staining of SOX2/K8 for embryonic tissues revealed MC emergence in the dorsal skin at E15.5, while early differentiated MCs were present in the vibrissae at E14.5 (Figure 1A), which was consistent with previous studies (Perdigoto et al., 2014). Given the accessibility of vibrissa tissue and its stability during culture, we selected vibrissae tissue at the critical developmental stage of E14.5 for subsequent experiments. We treated the vibrissae with small molecules targeting signaling pathways involved in MC development, which had been identified through previous in vivo studies (Rui et al., 2022) (Figure 1B). Genetic knockout of certain functional subunits of PRC1 and PRC2 in mice results in an increase in MCs (Cohen et al., 2018; Perdigoto et al., 2016). To further investigate whether this phenomenon persists in the ex vivo culture system for MC, we added small-molecule inhibitors targeting PRC1 and PRC2 to vibrissae tissue cultures for 6 days. Whole-mount IF analysis demonstrated that treatment with a combination of PRC1/2 inhibitors ([PRCis] EED226 + PRT4165 + GSK126) significantly increased SOX2⁺/K8⁺ MCs approximately 2-fold compared to the vehicle group treated with an equal volume of DMSO (Figures 1C and 1D, p < 0.01). Furthermore, UTX (ubiquitously transcribed tetratricopeptide repeat, X chromosome), a histone demethylase that antagonizes PRC2 activity (Kruidenier et al., 2012), was inhibited using a specific inhibitor GSK-J4 in the explant culture system. As expected, the results showed that UTX inhibitor (GSK-J4) attenuated SOX2⁺/K8⁺ MC generation (Figure S1A, p < 0.01). However, inhibitor targeting NOTCH (IMR-1) or agonist targeting SHH signaling (SAG) showed no effect on MC numbers in vibrissae explant culture (Figures S1B and S1C, p > 0.05). These findings demonstrate that PRC inhibition sufficiently promotes MC generation in developing mouse vibrissae tissues in the in vitro explant culture system.

Figure 1.

Pharmacological effects of small molecules on MCs in ex vivo vibrissa explant culture system

(A) Whole-mount IF staining for SOX2 (magenta) and K8 (yellow) was performed on dorsal skin or vibrissae of mice at different developmental stages (E13.5, E14.5, E15.5, and E16.5) and the quantifications of MC number were shown on the right, with mean ± SEM; each data point represents a touch dome (TD) or a hair follicle (HF). (For TD, n = 5 [E13.5], n = 8 [E14.5], n = 13 [E15.5], n = 4 [E16.5]; for HF, n = 6 [E13.5], n = 6 [E14.5], n = 6 [E15.5], n = 3 [E16.5]).

(B) Schematic illustration of the experimental procedure for vibrissa explant culture.

(C) Following 6 days treatment with vehicle (DMSO) or PRCi (10 μM EED +10 μM PRT4165 + 5 μM GSK126) on vibrissae of mice at E14.5, whole-mount IF staining for SOX2 (magenta) and K8 (yellow) was performed.

(D) The number of MCs in the vehicle or PRCi-treated vibrissae was quantified and visualized through boxplots, in which the box represents the 25th and 75th percentiles, the middle line indicates the median, and the whiskers represent the minimum and maximum values; each data point represents an independent experiment (n = 4).

Statistical significance was assessed using the t test. ^∗∗p < 0.01. Scale bars, 50 μm (see also Figure S1).

Inhibition of PRCs promotes MC generation in mSKOs

While vibrissae ex vivo cultures preserve native follicular architecture, they do not recapitulate the cutaneous niche in which MCs predominantly reside. We established a three-dimensional mSKO model to model dermal-epidermal organization of skin, enabling MC specification studies that better align with the most MC distribution context. Following a recent study, the organoid was formed by assembling epithelial and mesenchymal cells isolated from E18.5 mouse skin (Figure S2A). During subsequent culture for 9 days, the organoid developed appendage hair follicles (Figure S2B) and expressed markers (Figure S2C) of hair follicle stem cells and epithelial cells (K17, K15, and LEF1) (Fan et al., 2022), resulting in a mouse hair follicle organoid (mHFO) model. These mHFOs were treated with small molecules starting from 2 days post the assembly (D2). After 7 days of treatment (D9), E18.5-derived mHFOs showed few or no SOX2⁺/K8⁺ MCs and remained unresponsive to PRCi, Notchi (IMR-1), or combined signaling perturbations (Mp-1: PRCi + NOTCHi + SHHa [SAG]) (Figure S2D). We then tried to generate an mSKO model that incorporated cell sources containing MC precursor cells at earlier developmental stage, by co-culturing E14.5 dorsal skin cells with E18.5 mesenchymal cells (Figure 2A). Small molecules were added to mSKOs on the second day of culture and maintained for 7 days, followed by whole-mount IF staining analysis. Results showed that PRCi treatment induced a 27-fold increase in SOX2⁺/K8⁺ MCs compared to vehicle group (Figures 2B and 2C, p < 0.001). However, neither sequential treatment with Notchi followed by PRCi nor combined treatment with signaling perturbations by Mp-1 enhanced MC formation in this mSKO model (Figure S3, p > 0.05). These results indicate that inhibition of PRCs could promote MC generation in mSKOs but relies on early MC precursors from E14.5 mice skin.

Figure 2.

Pharmacological effects of small molecules on MCs in mouse embryonic skin-derived organoids

(A) Schematic diagram illustrating the generation of mSKOs. Skin cells and mesenchymal cells were isolated from the dorsal skin of E14.5 and the dermal skin of E18.5 mice, respectively. These cells were then co-assembled on D0 of culture. Vehicle or PRCi treatment was initiated on D2.

(B) Whole-mount IF staining for SOX2 (magenta) and K8 (yellow) was performed on mSKOs at D9 of culture (7 days after the initiation of vehicle or PRCi treatment).

(C) The number of MCs in the vehicle or PRCi-treated groups of mSKOs was quantified and visualized through boxplots, each data point represents an mSKO sample, n = 27 mSKOs per group from 5 independent experiments.

Statistical significance was assessed using the non-parametric t test, with ^∗∗∗p < 0.001. Scale bar, 50 μm. (see also Figure S3).

hSKOs recapitulate MC development

Given that the regulatory mechanism for human MC development remains unclear due to limitations of available models, we sought to establish a human skin organoid (hPSC-derived skin organoid [hSKO]) platform to enable mechanistic studies of human MC development from hPSC (H1 and iPSC). In our system (Figure 3A), the key checkpoints of hSKO formation were consistent with previous reports (Lee et al., 2020) with developed follicle-like buds at day 56 and hair follicle observation at day 87 (Figure 3B). While the time when differentiation initiated was recorded as day 0, the hSKO expressed non-neuroectodermal (NNE) markers (E-cad/TFAP2A) and neural crest cell (NCC) markers (P75 and SOX10) (Lee et al., 2020; 2022) at day 36. At this time of differentiation, the hSKO also expressed epithelial markers (K10, K17, and K15) (Jung et al., 2022), although the epidermal stem cell marker K14 remained undetectable during this developmental phase (Figure 3C). By day 62 (Figure 3D), the hSKOs demonstrated distinct epidermal and dermal structures with the staining of mesenchymal marker PDGFRα (Lee et al., 2022), containing K14⁺ epidermal stem cells, along with epithelial markers K10 and K8. Furthermore, SOX2⁺/K8⁺ MCs were positioned above dermal papilla-like structures, resembling their arrangement in vivo. Whole-mount staining of K20, another unique marker for MC in human skin, revealed the first appearance of K20⁺ MCs at day 42, with rapid proliferation afterward in hSKOs (Figure 3E). To assess whether hSKO is a comparable system for studying human MC development, we initially examined the responsiveness of hSKO to PRCi treatment. Upon adding PRCi to the hSKO culture system at 42 days post-hSKO differentiation initiation (D42), frozen sections and whole-mount staining of the MC marker K20 revealed a significant 4- to 13-fold increase in K20⁺ MCs following 2, 3, and 4 weeks of PRCi treatment (Figures 4A–4F). Notably, there were no obvious morphological or marker abnormalities for MCs; we also checked MCs’ unique marker ATOH1 and K20, as well as the functional synaptic protein SNAP25 and SYN-2. PRCi-treated group and vehicle group exhibited normal expression of these markers (Figures 4G–4I). These findings validate that hSKOs are an effective platform for studying human MC development and indicate that PRC inhibitors can also promote MC generation in hSKOs.

Figure 3.

Development of MCs in hSKOs

(A) Schematic illustration of hSKOs.

(B) Representative phase images of hSKOs at various time points during differentiation (the initiation time of differentiation was recorded as D3, D6, D9, D12, D33, D46, D56, and D87). At D87, the dotted line highlights the fully developed hair follicle.

(C) Frozen-section IF staining of hSKOs at D36 post-induction revealed the expression of the NNE markers, including E-cad (E-cadherin, yellow) and TFAP2A (magenta); the NCC markers, including P75NTR (P75, magenta) and SOX10 (magenta); and the epithelial markers, including K10 (yellow), K17 (magenta), K8 (yellow), and K15 (gray).

(D) Frozen-section IF staining of hSKOs at D62 post-induction demonstrated the expression of the dermal marker PDGFRa (yellow), the epidermal stem cell marker K14 (magenta), the epithelial markers K10 (gray) and K15 (gray), and the MC markers SOX2 (gray) and K8 (magenta); arrows indicate SOX2⁺/K8⁺ MC.

(E) Whole-mount IF staining of hSKOs at D42 post-induction showed the expression of MC marker K20 (magenta); arrows indicate MC. Analysis of the MC number at different time points was shown on the right. (D35, n = 4 fields from 3 hSKOs; D42, n = 5 fields from 3 hSKOs; D49, n = 8 fields from 4 hSKOs, D56, n = 10 fields from 4 hSKOs, field area = 0.3364 mm²).

Scale bars: 500 μm in (B) and 50 μm in (C, D, and E).

Figure 4.

PRCi treatment increases the number of MCs in hSKOs

(A–F) hSKOs were treated with PRCi for 2 weeks (A), 3 weeks (C), or 4 weeks (E) (time 0 was the PRCi treatment initiation time), followed by IF staining of whole-mount (A and C) or frozen section (E) for K20⁺ MC (magenta). The quantification of MC number per field was shown on the right as box blots; each data point represents an hSKO sample; n = 9 hSKOs per group from 3 independent experiments (B), n = 9 hSKOs for vehicle-treated group and n = 11 hSKOs for PRCi-treated group from 3 independent experiments (D), and n = 11 hSKOs per group from 3 independent experiments (F) (4 fields averaged per hSKO, field area = 0.3364 mm²).

(G–I) Whole-mount IF staining reveals ATOH1⁺ (magenta) and K20⁺/SNAP25⁺/SYN-2⁺ (cyan) MC in vehicle or PRCi-treated groups of hSKOs.

Statistical significance was assessed using the t test, with ^∗∗∗p < 0.001. Scale bars: 100 μm in (A and C) and 50 μm in (E, G, H, and I).

Single-cell and spatial transcriptomics reveal IGF-IGF1R as a paracrine signaling participating in MC development

To explore potential novel regulators of MC development, we utilized an integrative approach combining single-cell and spatial transcriptomics analysis. We first examined previously published single-cell RNA sequencing (scRNA-seq) data from mouse dorsal skin samples collected at E14.5, E16.5, and postnatal day 0 (P0) (Lin et al., 2020) (Figure 5A). The analysis identified a distinct cluster of MCs with 81 cells (Figure 5A), which were characterized by the expression of the MC markers (K8, Sox2, and Atoh1) (Figure 5B). Other identified skin cell types included basal interfollicular epidermis (IFE), differentiated IFE, fibroblast, hair follicle, melanocyte, adipocyte, and other cell types (Figures 5A and 5B), with fibroblasts representing the predominant cell type. Cell-cell communication analysis using the iTALK revealed that fibroblasts expressed insulin-like growth factors IGF1 and IGF2, which potentially activate the IGF1R receptor in MCs (Figures 5C and 5D).

Figure 5.

IGF2-IGF1R signaling pathways regulate MC development

(A) Uniform manifold approximation and projection visualization of cells from integrated published scRNA-seq data (GSE154579) collected from E14.5, E16.5, and P0 skin of mice. The MC cluster is indicated in red (n = 81 cells) with dashed box enlarged. IFE, interfollicular epidermis.

(B) Dot plot shows the expression of selected marker genes of different types of cells.

(C) Circle plot shows the significant ligand-receptor interactions identified by cell-cell communication analysis by iTALK. Line thickness and arrowheads are scaled to represent interaction strength between ligands and receptors. IGF1/2-IGF1R interaction was highlighted in red.

(D) Violin plots show the normalized expression levels of Igf1/2 and Igf1r in the indicated cells in scRNA-seq data.

(E and F) (E) Spatial transcriptomics analysis of P0 skin by Stereo-seq. Image of single-stranded DNA staining section is shown. The blue box highlights the touch dome region. Spatial distribution of representative cell clusters in the region is shown on the right. The dashed lines indicate the edges of hair follicle. IGF1/2-IGF1R-mediated adjacent fibroblast (FB)-MC communications are indicated by arrows. (F) Spatial expression of indicated genes in the highlighted region of (E) (see also Figure S5).

To confirm whether this finding reflects local cell communication, we conducted high-resolution spatial transcriptomics analysis on P0 dorsal skin using Stereo sequencing (Stereo-seq). Using molecular features of distinct cell types derived from the scRNA-seq, we identified IFE, hair follicle cells, fibroblasts, smooth muscle cells, and adipocytes in the dataset. Despite their low abundance, MCs were detected in the basal layer of IFE and the upper region of hair follicles, surrounded by fibroblasts (Figures 5E, S4A, S4B, S5A, and S5B). The FGF20-FGFR2 pathway, previously established as crucial for MC development, showed FGFR2 expression in MCs and FGF20 expression in hair follicles, consistent with earlier studies (Figures S4C and S4D) (Nguyen et al., 2018). In line with the scRNA-seq findings, MCs expressed IGF1R, whereas adjacent fibroblasts expressed IGF1 or IGF2 (Figures 5F, S5C, and S5D). These findings suggest that the IGF1/2-IGF1R axis may facilitate paracrine signaling from fibroblasts in the niche, potentially influencing early MC development.

IGF1R signaling is necessary but not sufficient for MC development

Given the time-consuming nature of in vivo studies on IGF1R signaling in MC development, we conducted a preliminary investigation using our ex vivo system to determine whether IGF1R signaling affects MC generation in a significantly shorter time frame. Concurrently, we examined the FGF20-FGFR2 pathway, which has been reported to regulate MC development in vivo (Nguyen et al., 2018). Our results in mouse vibrissae explants demonstrated that IGF1R inhibition (IGF1Ri) reduced formation of SOX2⁺/K8⁺ MCs significantly (Figures 6A and 6G, p < 0.001), an effect comparable to FGFR2 inhibition (FGFR2i) (Figures 6B–6H, p < 0.001). Notably, in hSKOs, the number of K20⁺ MC was decreased by more than 80% after 4 weeks treatment of either IGF1Ri or FGFR2i (Figures 6C–6F, 6I, and 6J, p < 0.001). Meanwhile, we observed that FGFR2i treatment led to atrophy of the treated organoid; however, no significant difference in nuclear size and no nuclear morphological abnormalities (such as pyknosis/condensation or swelling) were observed. Furthermore, we analyzed the IGF1R expression patterns across different developmental stages in mouse dorsal skin. Bioinformatic analysis revealed that Igf1r is expressed in MCs as well as in various other skin cells, such as endothelial cells and melanocytes at E14.5 and E16.5 (Figure S6A). Immunohistochemistry analysis indicated slightly increased IGF1R expression in the dermal-epidermal junction area from E14.5 to P0 (Figure S6B), although there was no obvious specific staining for MCs. Notably, at P0, in the dorsal skin, IGF1R was more abundantly expressed at the dermal-epidermal junction where MCs reside.

Figure 6.

FGF20-FGFR2 and IGF2-IGF1R are involved in MC development

(A and G) Whole-mount IF staining for SOX2 (magenta) and K8 (yellow) was performed on E14.5 mouse vibrissae following 6 days’ treatment with 10 μM IGF1R inhibitor (IGF1Ri, BMS-754807). The quantification of MC number per vibrissa in (G) was shown as box blots (each data point represents a vibrissa explant, n = 9 vibrissae per group from 1 independent experiment).

(B and H) Whole-mount IF staining for SOX2 (magenta) and K8 (yellow) was performed on E14.5 mouse vibrissae following 6 days’ treatment with 5 nM FGFR2 inhibitor (FGFR2i, infigratinib). The quantification of MC number per vibrissa was shown in (H) (each data point represents a vibrissa explant, n = 8 vibrissa per group from 2 independent experiments).

(C–F and I–J) Whole-mount IF staining for K20⁺ MC (magenta) and DAPI (blue) was performed on hSKO following 4 weeks’ treatment with vehicle, 10 μM IGF1Ri, or 10 μM FGFR2i. The quantification of MC number per field was shown in (I and J) as box blots (each data point represents a hSKO sample, n = 7 hSKOs per group from 2 independent experiments) (4 fields averaged per hSKO, field area = 0.3364 mm²).

(K–M) Western blot (K) and statistical analysis were performed to assess the protein expression of total AKT (L) and phospho-AKT (pAKT-Ser473) (M) in vibrissae following 6 days’ treatment with vehicle or IGF1Ri. Data are presented as mean ± SEM; each data point represents an independent experiment, (n = 5). (N) The quantification of MC number per vibrissa after 6 days’ treatment with DMSO (vehicle), 10 μM IGF1Ri, 10 μM IGF1Ri and 10 μM AKT agonist SC79 (each data point represents a vibrissa explant, n = 11 vibrissa per group from 3 independent experiment).

Statistical significance was assessed using the t test (G, H, I, L, and M) or non-parametric t test (J and N), with ns, p > 0.05; ^∗p < 0.5; ^∗∗∗p < 0.001. Scale bars: 100 μm in (A and B) and 50 μm in (I–L).

Previous studies suggested that IGF1R influences cell growth and proliferation via the AKT pathway (Jiang et al., 2022). We investigated the downstream pathways by which IGF1R exerts its effects to further show the involvement of IGF1R signaling in MC development. The protein levels of IGF1R downstream signals, total AKT and ERK, as well as phosphorylated AKT and ERK (pAKT and pERK), were checked by western blot following 6 days’ treatment of IGF1Ri in E14.5 vibrissae. Our results showed that IGF1Ri significantly decreased pAKT/AKT ratio without altering total AKT (Figure 6K–6M), while the ERK and pERK showed no significant change (Figure S7, p > 0.05). To further investigate whether AKT influences MC development, the vibrissa treated with IGF1Ri were cultured with the AKT agonist SC79 for 6 days. The results showed that SC79 partially rescued the IGF1Ri-induced reduction of MCs in the explants (Figure 6N). These results suggested that IGF1R probably exerts its effect on MC development via the AKT pathway. However, exogenous IGF2 or FGF20 supplementation did not increase MC numbers in mouse vibrissae explants or in hSKOs (Figure S8, p > 0.05). Taken together, these results demonstrate that IGF1R is indispensable for MC development; however, exogenous IGF2 supplementation alone is not sufficient to promote MC generation.

Discussion

In this study, we established multiple in vitro culture systems, including explant cultures, mHFOs, mSKOs, and hSKOs, for modeling MC development. These systems enabled the validation of known signaling pathways and the discovery of novel regulators, overcoming limitations of traditional in vivo approaches. Taking advantage of these innovative culture systems, we identified PRCi (a combination of EED226, PRT4165, and GSK126) as a potent cocktail inducer of MC generation in vitro and indicated the involvement of the IGF1R signaling axis in MC development.

In vivo studies have established that PRC subunits Ezh1 and Ezh2 suppress the MC differentiation program in skin stem cells by repressing Sox2 (Bardot et al., 2013). Sox2 is a transcription factor that cooperates with Isl1 to activate Atoh1, a master regulator essential for MC differentiation and maturation (Perdigoto et al., 2014). Prior in vivo mouse studies have further validated the regulatory roles of PRC complexes in MC development. Multiple genetic ablation experiments demonstrated that disrupting PRC function increases skin MC numbers. These include simultaneous knockout of PRC2’s EZH1/2 (Bardot et al., 2013), knockout of PRC2’s EED (Perdigoto et al., 2016), ablation of PRC1’s Ring1a/Ring1b (enzymatically inactive) or Pcgf2/4, and knockout of PRC2’s Suz12 (Cohen et al., 2018). Collectively, these findings laid a foundation for our in vitro validation. However, targeted MC genetic manipulation via CRISPR/small hairpin RNA (relying on viral delivery) faces technical barriers stemming from MCs’ inherent traits. MCs account for only 0.21% of epithelial cells in P0 dorsal skin (per our results), limiting the infection probability. As terminally differentiated cells, MCs have weak proliferation; they show lower viral integration efficiency than proliferating precursors. To bypass these barriers, we used small-molecule inhibitors targeting PRC subunits (aligned with in vivo genetic targets) and validated their effects in E14.5 vibrissa explants. Notably, combined inhibition of PRCs (PRCi) induced the most robust MC increase in explants and was further validated in mSKO and hSKO systems, while the effect of a UTX inhibitor that decreased MC number confirmed the specificity of PRC-mediated MC induction, consistent with in vivo findings in mice (Bardot et al., 2013; Cohen et al., 2018; Perdigoto et al., 2016). Building on this foundation, we treated ex vivo explants and in vitro mSKO/hSKO models with PRCi and observed significant promotion of MC generation consistent with in vivo studies.

While PRCi demonstrated robust effects, interventions targeting other pathways—such as Notch signaling—showed limited effect on MC generation in our system; although Notch is a known inhibitor of MC fate in vivo, preventing ectopic MC differentiation (Logan et al., 2018), its minimal impact in vitro likely reflects the absence of the complex in vivo interactions that define its context-dependent function, and the inability of certain exogenous factors to modulate MC generation may be attributed to signaling redundancy or saturating levels of endogenous signals within our system, which could obscure the effects of additional manipulation.

Our study reveals the involvement of both IGF and FGF signaling in MC development. Blocking the receptors IGF1R and FGFR2 significantly impaired MC formation, underscoring the essential role of ligand-receptor interactions in MC specification. However, exogenous addition of excess IGF2 or FGF20 produced no apparent effect on MC generation. This lack of response is consistent with the fact that FGF20 is endogenously secreted by dermal papillae and epithelial cells in vivo, and its local concentration in the MC niche is already saturated (Song et al., 2025); therefore, exogenous supplementation in vitro may not be able to further activate FGFR2. This interpretation aligns with literature indicating that receptor antagonism often reveals phenotypic effects, as redundant signaling pathways may compensate for single-factor loss (Perdigoto et al., 2016). Regarding signal sources, FGF20 signaling likely originates from adjacent hair follicle epithelium, whereas IGF1/2 may derive from fibroblasts adjacent to MCs, as suggested by single-cell sequencing and spatial transcriptomics data. Consistent with this, our findings support the notion that IGF1/2 act as paracrine signals, with their receptor IGF1R expressed on MC progenitors. Furthermore, our western blot results suggest that IGF1R likely exerts its effects through the AKT signaling pathway. Administration of the AKT agonist SC79 partially rescued the IGF1Ri-induced reduction of MCs in explants, further validating that IGF1R mediates its effects through AKT. In hSKOs, inhibition of FGFR2 led to atrophy, suggesting that FGFR2 may also be involved in other developmental processes.

The development of robust in vitro systems for studying MCs has presented significant challenges due to their scarcity and difficulties in long-term culture. Previous in vivo studies indicate MCs can survive long-term under intact conditions (Wright et al., 2017), whereas existing in vitro systems face substantial limitations. Isolated murine MCs typically survive only 7–8 days and require supplemental growth factors such as NT3 and NGF (Shimohira-Yamasaki et al., 2006). An organotypic co-culture method combining whisker follicles with trigeminal ganglion explants offered observational insights but lacked experimental flexibility (Tsutsumi et al., 2023). Our system introduces essential optimizations, including reduced serum concentration, suspension culture, and a shortened 6-day culture period, which substantially improves efficiency. Furthermore, our mSKO system represents a novel approach. MCs occur sparsely in the skin (Yuan et al., 2022); thus, attempts for mHFOs constructed from E18.5 skin tissue using previous method (Kageyama et al., 2022) failed to yield observable MCs, even with PRCi supplementation. Based on the observation that MCs emerge around E14.5, we hypothesized that MC precursors were enriched in E14.5 skin. Since isolated E14.5 skin cells cannot form mSKOs independently without a suitable niche, we innovatively combined them with E18.5 dermal mesenchymal cells as supporting elements, successfully establishing a system capable of generating abundant MCs when treated with PRCi.

Current in vivo MC research has been conducted primarily in mouse models, with recent work expanding to zebrafish systems for developmental and regeneration study (Brown et al., 2023; Craig et al., 2025) models. Due to the scarcity of culture systems and the significant challenges associated with long-term MC maintenance, human MC studies are particularly limited, which has impeded progress in tactile regeneration. A major obstacle in cultivation lies in the disparity between in vitro and in vivo environments. To better approximate physiological conditions, complex organoid systems have been developed for modeling and differentiation. As a rare functional cell type, MCs were first identified in hPSC-derived human skin organoids (hSKOs) by the Koehler team (Lee et al., 2020). Building upon this organoid differentiation strategy, we have implemented partial optimizations and now consistently generate hSKOs containing MCs. Through temporal tracking, we observed that MCs emerge in hSKOs within a critical window around 6 weeks (day 42) of differentiation, and PRCi intervention at this stage demonstrated significant enhancement of MC populations. This advancement both establishes an earlier MC detection time point than previously documented and determines the optimal intervention window for maximizing MC generation through PRCi treatment. Consequently, this system enables detailed investigation of human MC development and regeneration mechanisms.

The simplified culture conditions, however, may not fully recapitulate the complex three-dimensional architecture and cellular diversity present in the in vivo environment. Future investigations could incorporate co-culture with sensory neurons or other niche cell types identified through spatial transcriptomics to better approximate the in vivo niche. Furthermore, optimization of the system to support long-term MC maintenance would facilitate studies of MC survival and homeostasis. This could include supplementation with trophic factors such as NT3, NGF and SHH, which are crucial for MC survival after birth (Xiao et al., 2015). A more robust culture system that builds upon our system would enable comprehensive characterization, including genetic, protein, and electrophysiological profiling. Moreover, with the rapid advancement of gene editing technologies, genetic manipulation based on hSKOs in future studies will significantly accelerate research on the fate regulation of stem cell differentiation into MCs or other cutaneous cell types. For instance, constructing viral vectors driven by MC-specific promoters (e.g., Atoh1) would be useful to manipulate genes in MCs specifically. Additionally, it would also be helpful to generate hESC cell lines that carry inducible CRISPR elements to activate or repress specific genes at targeted differentiation stages in order to investigate their roles in MC development in a stage-specific manner.

In conclusion, this in vitro culture system represents a pivotal advance by providing a versatile platform for MC research. Integrating spatial transcriptomics with hPSC-derived organoids enhances translational potential, yielding mechanistic insights into human MC biology. Prospectively, the deliberate integration of in vitro and in vivo approaches will harness complementary strengths to accelerate progress in stem cell-based regenerative medicine, disease modeling, and therapeutic development.

Materials and methods

Experimental animal

C57BL/6J (average body weight 25 ± 1.2 g) mice were obtained from the GemPharmatech Company. All animal experiments were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals and approved by ZJU-Laboratory Animal Welfare and Ethics Review Committee (approval no. ZJU20220357). To obtain embryos at defined developmental stages, male and female mice were co-housed overnight and separated the following morning. The day of separation was designated as E0.5.

Explant culture

Pregnant C57BL/6J mice at E14.5 were anesthetized and euthanized humanely, and then the vibrissae pads and dorsal skin were harvested. The experimental groups were treated with a PRC inhibitor cocktail (PRCi), the FGFR2i infigratinib, the IGF1R inhibitor (BMS-754807), or with the AKT agonist SC79 (HY-18749) (all from MCE). The PRCi was a mixture of EED226 (HY-101117), PRT4165 (HY-19817), and GSK126 (HY-13470). An equal volume of DMSO (Solarbio, D8371) was used as the vehicle control. The medium was refreshed every 2 days. Stocks of EED226 (50 mM), PRT4165 (50 mM), GSK126 (20 mM), infigratinib (10 mM), BMS-754807 (10 mM), and SC79 (10 mM) were all dissolved in DMSO. IGF2 (Human IGF-II Recombinant Protein, Peprotech, 100-12) and FGF20 (Human FGF-20 Recombinant Protein, Peprotech, 100-41) were prepared according to the manufacturer’s instructions, with stock concentrations of 100 and 10 μg/mL. Complete medium for explant culture contains DMEM/F12 (GIBCO, 11330032), 20% fetal bovine serum (FBS, ExCell Bio, FSP500), and 1% PS (penicillin-streptomycin, BI, C3420).

Generation of mHFO and mSKO

The mHFOs were generated as previously described (Kageyama et al., 2022) with slight modification. The dorsal skin of E18.5 mice was harvested and incubated in dispase II (4 mg/mL Gibco, 17105041) at 37°C for 60 min to separate the epithelial and mesenchymal layers. The epithelial layer was further treated with 0.25% trypsin-EDTA (Genom, GNM25200-1) for 10 min at 37°C. The dermal layer was then digested in collagenase A (2 mg/mL, Roche, 10103586001) for 60 min at 37°C. Cell suspensions were filtered through a 40-μm cell strainer (SAINING, 5020000) and centrifuged at 500 × g for 5 min. The epithelial and mesenchymal cells were suspended separately in mHFO medium (Advanced DMEM/F-12 [Gibco, 12634010] containing 1% GlutaMAX [Gibco, 35050061] and 1% PS). After counting, epithelial (3,000 cells/well) and mesenchymal (12,000 cells/well) cells were mixed in mHFO medium containing 2% Matrigel (Corning, 354230) and seeded into the ultralow-attachment, round-bottomed 96-well plates (Corning, 7007). Plates were chilled at 4°C for 60 min and centrifuged at 300 × g for 3 min to help cells aggregate and then transferred to a 37°C, 5% CO₂ incubator. Cells were maintained in mHFO medium, and the medium in each well was refreshed by replacing 0.1 mL every 2 days.

For mSKO generation, dorsal skin of E14.5 embryos was harvested and digested in 2 mg/mL collagenase A for 60 min at 37°C. A single-cell suspension was prepared using the protocol described above and then combined with mesenchymal cells isolated from E18.5 embryos at a 1:1 ratio (20,000 total cells/well), and cultured under identical conditions with mHFO.

Generation of the hSKO

The hSKOs were generated from hPSCs following established protocols (Lee et al., 2020; 2022) with slight modification. H1 cells at P13–P56 were dissociated using Accutase (Gibco, A1110501) to obtain single-cell suspensions in E8 medium (Gibco, A2858501) supplemented with 10 μM Y-27632 (Stemgent, 04-0012) called E8-10Y; 4,000 cells in 100 μL E8-10Y medium per well were seeded in ultralow-attachment 96-well plates. Plates were centrifuged at 200 × g for 3 min to promote aggregate formation (designated as day −2) and then incubated at 37°C with 5% CO₂. On day −1, 100 μL E8 medium was added per well. Differentiation was initiated on D0 with E6 medium (Gibco, A1516501) containing 2% Matrigel (Corning, 354230), 10 ng/mL BMP-4 (R&D Systems, 314-BP), 10 μM SB431542 (Stemgent, 04-0010-05), and 4 ng/mL bFGF (PeproTech, 100-18B). On D3, 25 μL of E6 medium supplemented with 1 μM LDN (BMP inhibitor, Stemgent, 04-0074-02) and 250 ng/mL bFGF was added. Medium was supplemented with 100 μL E6 on D6, with half-medium changes on D8 and D10. On D12, aggregates were transferred to 24-well low-attachment plates containing 500 μL organoid maturation medium (OMM) containing 1% Matrigel and cultured at 65 rpm on an orbital shaker (Miulab, GSP-20) at 37°C with 5% CO₂. OMM consisted of 1:1 Advanced DMEM/F12: Neurobasal media (Gibco, 21103049) supplemented with 1 × GlutaMAX, 0.5 × B-27 (minus vitamin A, Gibco, 12587010), 0.5 × N2 (Gibco,17502048), 0.1 mM 2-mercaptoethanol (Gibco, 21985023), and 1% PS. Half-medium changes were performed every 3 days, with gradual volume increases as organoids matured.

Intercellular ligand-receptor prediction

Publicly available single-cell datasets from E14.5, E16.5, and P0 were acquired from the Gene Expression Omnibus (GEO) (GEO: GSE154579; https://www.ncbi.nlm.nih.gov/geo/). Cell-cell communication networks were computationally inferred by analyzing receptor-ligand interactions between MCs and other identified cell subsets using iTALK (v.0.1.0) (Wang et al., 2019).

Spatial Stereo-seq data processing

Paired-end sequencing was performed on the MGI DNBSEQ-Tx platform, generating FASTQ files containing spatially encoded reads. Read 1 encapsulated both the Coordinate ID ([CID] spatial barcode) and Molecular Identifier (unique molecular tag), while Read 2 carried cDNA sequences. Reads were mapped to the mouse reference genome (GRCm38/mm10) using STAR aligner with default parameters (Dobin et al., 2013). HandleBam filtered low-confidence alignments (MAPQ [Mapping Quality] <10) to retain uniquely mapped reads. Finally, exon-derived reads with valid CID annotations were aggregated to generate a gene-by-DNB matrix, where rows represented genes and columns corresponded to individual DNBs. Transcript-captured bin 50 × 50 DNBs were merged as on bin50, which was treated as the fundamental analysis. The data analysis followed the official tutorial (https://stereopy.readthedocs.io/en/latest/index.html). Visualization was performed using the Seurat package in R.

Statistical analyses

p values and statistical tests performed are provided in the figure legends or supplemental information. Most of the data are displayed as the boxplot, where the box represents the interquartile range (25th and 75th percentiles), the line within the box indicates the median, and the whiskers represent the minimum and maximum values, with some data presented as mean ± SEM indicated in figure legends. All data were assessed for normality and homogeneity of variance using SPSS (v.27). For datasets that satisfied the assumptions of normality, an independent sample t test was performed utilizing GraphPad Prism (v6.01) to compare two groups. When the assumption of normality was not met, non-parametric tests were employed for the comparison of two or more groups. One-way ANOVA followed by Tukey’s HSD (Honestly Significant Difference) post hoc test was applied to compare multiple groups when the data satisfied both normality and homogeneity of variance assumptions.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact, Ying Xiao (xiaoying.srr@zju.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The accession number for the Stereo-seq data reported in this paper is GSA: CRA023727. The code supporting the findings of this study is available from Zenodo at https://doi.org/10.5281/zenodo.17933885.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by fundings from National Key Research and Development Program of China (2021YFA1101100) and National Natural Science Foundation of China (32070816, 32300674, and 82003333). We would like to thank Oasisbiofarm Company for providing the ATOH1 antibody, Xiaoli Hong and Chao Bi from the core facilities at Zhejiang University School of Medicine for their assistance in confocal imaging, and the BioRender (biorender.com) for providing the tools to create scientific illustrations in this publication.

Author contributions

H.Y., C.R., and Y.Z. contributed equally to this project; H.Y. designed the study, performed organoids-related experiments, prepared the manuscript, and acquired funding; C.R. performed explant culture experiments and helped with manuscript preparation; Y.Z. collected omics data, performed bioinformatics analysis, and helped with manuscript preparation; J.L. supervised hSKO differentiation; Y.H. performed mouse breeding and mating; X.W. and T.W. participated in discussion and edited the manuscript; Z.Z. performed statistical data analysis; C.W. supervised data analysis and edited the manuscript; Y.X. designed and supervised the project, finalized the manuscript, and acquired funding.

Declaration of interests

The authors declare no competing interests.

Published: December 26, 2025