Abstract
Highly regenerative tissues continuously produce terminally differentiated cells to replace those that are lost. How they orchestrate the complex transition from undifferentiated stem cells towards post-mitotic, molecularly distinct and often spatially segregated differentiated populations is not well understood. In the adult skin epidermis, the stem cell compartment contains molecularly heterogeneous subpopulations1–4 whose relationship to the complete trajectory of differentiation remains unknown. Here we show that differentiation, from commitment to exit from the stem cell layer, is a multi-day process wherein cells transit through a continuum of transcriptional changes with upregulation of differentiation genes preceding downregulation of typical stemness genes. Differentiation-committed cells remain capable of dividing to produce daughter cells fated to further differentiate, demonstrating that differentiation is uncoupled from cell cycle exit. These cell divisions are not required as part of an obligate transit-amplifying programme but help to buffer the differentiating cell pool during heightened demand. Thus, instead of distinct contributions from multiple progenitors, a continuous gradual differentiation process fuels homeostatic epidermal turnover.
Subject terms: Stem-cell differentiation, Multiphoton microscopy, Skin stem cells
Cockburn et al. report that epidermal differentiation is a multi-day process through which cells undergo a continuum of transcriptional alterations initiated independently of cell cycle exit.
Main
During regeneration, the onset of stem cell differentiation and cell cycle exit are highly temporally correlated. However, many tissues contain intermediate cell types (often termed transit-amplifying cells) that have begun to differentiate but retain some proliferative capacity5–7. How such intermediate cell states arise and whether the overlap between proliferation and differentiation contributes to cellular maturation or tissue maintenance remains largely unclear.
In the mammalian skin epidermis, proliferative basal layer cells differentiate and move upwards to replace barrier-forming cells that are continuously shed (Extended Data Fig. 1a). Lineage tracing of randomly labelled basal cells as well as label dilution studies support the existence of a single type of proliferating progenitor that generates dividing and differentiating cells with equal probability at the long-term, populational level8–11. In contrast, lineage tracing that reflects the more short term activity of distinct promoters such as Involucrin-CreERT2 (refs. 1,2), Tbx3-CreERT2 (ref. 3), Dlx-CreERT2 and Slc1a3-CreERT4 produces different distributions of clone numbers and sizes, indicating underlying differences in proliferative capacity or kinetics. Specifically, basal cells labelled by Involucrin-CreER have been suggested to represent a distinct and more differentiation-primed population of progenitors that contribute differentially to homeostasis versus wound repair1 (Extended Data Fig. 1b). How such progenitors might fit into the epidermal differentiation journey itself is not yet clear.
To understand this journey in more detail, in this Letter, we focused on the long-standing observation that a subset of basal cells express the well-established differentiation marker keratin 10 (K10) (refs. 12,13). As these cells have typically been considered a post-mitotic population in the process of exiting the basal layer, neither their real-time behaviours nor their relevance to any of the aforementioned models of epidermal homeostasis have been closely examined. We first performed whole-mount immunostaining and single-molecule RNA fluorescence in situ hybridization (smRNA-FISH) in ear and dorsal skin (Fig. 1a and Extended Data Figs. 1c,d and 2). In total, 39–44% of basal cells stained positive for K10 protein or messenger RNA (mRNA), with the majority of these K10+ cells making a small area of contact (footprint) with the underlying extracellular matrix (ECM) as would be expected from delaminating basal cells (Fig. 1a,b and Extended Data Fig. 2a,b). However, we also observed K10+ cells with a typical basal-cell morphology including a normal-sized footprint (Fig. 1a,b and Extended Data Fig. 2a). These results indicate that, in contrast to previous models14, the differentiation process of basal epidermal cells may begin before a detectable change in cell morphology.
To understand the real-time behaviours of K10+ basal cells, we performed intra-vital imaging in mouse ear skin10 using a reporter system where the Krt10 promoter drives H2BGFP fluorescence (K10rtTA; pTRE-H2BGFP)15 (Fig. 1a). This reporter labels 44% of basal cells, 80% of which co-express K10 protein (Fig. 1c) and/or Krt10 mRNA (Extended Data Fig. 2c). By revisiting the same epidermal regions every 24 h (Fig. 1d), we found that the vast majority (93%) of K10-reporter-positive basal cells delaminate over the following 5 days (Fig. 1e), indicating that this population has largely committed to differentiate. We then focused on delaminating cells and tracked their H2BGFP signal in the days leading up to this event (Fig. 1f and Extended Data Fig. 3). Notably, K10 reporter expression always preceded delamination (Extended Data Fig. 3b), typically beginning 3–4 days before basal layer exit (Extended Data Fig. 3a,c–e). The corresponding morphological changes occurred surprisingly slowly over approximately 36 h (Fig. 1g–i). In contrast to the rapid, actomyosin-based extrusion events that have been described in embryonic epidermis and other systems16–18, basal cell delamination lacked obvious signs of ring-like actin or myosin accumulation (Extended Data Fig. 3f,g). These temporal observations demonstrate that adult basal cell differentiation, from commitment to the completion of delamination, is a gradual multi-day process.
We next aimed to understand how K10 expression relates to the global transcriptional changes associated with basal cell differentiation and to other previously described differentiation-primed progenitor populations1,2. Our previous single-cell transcriptome-based reconstruction of the epidermal differentiation trajectory grouped cells according to their individual gene expression from basal (Krt14high) to mature (Krt10high) to terminally differentiated (Lorhigh) cell states19. To define the basal-suprabasal border on this trajectory, we generated single-cell transcriptomes of fluorescence-activated cell sorting (FACS)-isolated basal cells (Live/ITGA6+/SCA1+/CD34−; Extended Data Fig. 4a). We then merged these transcriptomes with two published datasets encompassing both basal and suprabasal cells (Extended Data Fig. 4b–d and Methods)19,20, which allowed us to assign the basal-suprabasal border according to the sorted basal (ITGA6+) cells (Fig. 2a and Methods). Notably, the previously defined intermediate Krt10dim/Ptgs1dim/Mt4+ cell group (Differentiated I) and part of the mature Krt10high/Ptgs1high group (Differentiated II) (ref. 19), are basal-layer cells (Fig. 2b,c). This is exemplified by ~40% of basal cells already expressing Krt10 transcripts (Extended Data Figs. 2a,b and 5a,b), as well as other basal and differentiation transcripts in the same cells (Fig. 2d,e and Extended Data Fig. 5e–i), consistent with recent studies in oral21 and skin epithelia (Extended Data Fig. 5c)22.
To examine early transcriptional changes signifying the onset of differentiation more closely, we grouped the cells along the trajectory into ten differentiation bins (Fig. 2f, Extended Data Fig. 5d–f and Methods), revealing Krt10 as the first upregulated differentiation marker (bin 3), followed by Krtdap and Krt1 (bin 4), Mt4 and Sbsn (bin 5) and Ivl and Lor (bin 6) (Fig. 2f,g and Extended Data Fig. 5e–i). Consistently, immunostaining revealed a small subset of K10+ basal cells with very low levels of K1, but not vice versa (Extended Data Fig. 5j). Basal marker-gene expression starts to decrease in bin 3–4 (Itga6 and Ly6a), with the decrease of Krt14 being most pronounced (Fig. 2f,g and Extended Data Fig. 5e,f). Thus, the earliest molecular changes associated with differentiation start in bin 3, marked by an increase of Krt10 expression in cells displaying otherwise typical characteristics of stem cells (Krt14high, large ECM footprint) (Extended Data Fig. 5g–i). In addition, we observed gene-expression patterns associated with known as well as less explored transcription factors (TFs; Id1, Hes1, Trp63, Ovol1 and Csnk2b), all leading to a gradual onset of differentiation-associated gene expression (bins 2–4; Extended Data Fig. 6).
Our results indicate that, instead of discrete intermediate cell states, basal cells differentiate through a series of progressive transcriptional changes, raising questions about how previously proposed differentiation-primed progenitors (that is, Involucrin-CreER traced cells1,2) fit within this continuum. Thus, we acquired our ITGA6-sorted dataset from Involucrin-CreER traced (committed; Tom+) and non-traced cells (non-committed; Tom−) (Extended Data Fig. 4a). The majority of Tom+ basal cells indeed mapped with the Krt10+/Mt4+ cells and the majority of Tom− cells mapped with basal Krt14+ cells (Fig. 2h,i and Extended Data Fig. 5h,i), suggesting that the Involucrin-traced cell population largely represents committed cells but not necessarily a discrete progenitor cell state.
We next sought to investigate whether upregulation of differentiation genes precedes the downregulation of stem cell signature genes or vice versa. As splicing typically occurs co-transcriptionally 5–15 min after crossing an exon–intron junction23,24 we analysed spliced and unspliced Krt10 and Krt14 mRNA fragments during the defined differentiation process (Fig. 2j). Focusing on differentiation pseudotime stages that had both high Krt10 and Krt14 expression (bins 4–6), we observed that Krt14 is still actively transcribed (unspliced mRNAs) in cells with high expression of mature (spliced) Krt10 mRNA (Fig. 2j,k). Moreover, active transcription (unspliced) of both Krt14 and Krt10 mRNA can occur within the same cells (Fig. 2l). Additionally, Krt14 transcription can remain active even in differentiating, Krtdap-expressing cells (Fig. 2j–l). In sum, this fine-tuned differentiation trajectory appoints Krt10 expression at the molecular onset of a continuum of transcriptional changes associated with epidermal differentiation.
Further, analysis of the cell cycle stages revealed that approximately 15% of Krt10-expressing cells are actively proliferating (Extended Data Fig. 7a–c). Krt10+ cells account for ~24% of the S/G2/M phase population (Fig. 3a,b), while K10 protein is detectable in 9–19% of these cells (Fig. 3c,d and Extended Data Fig. 7d–g). Through detailed timelapse imaging, we observed that K10+ cells undergo mitosis parallel to the basement membrane and produce daughter cells that are fully integrated within the basal layer and retain reporter expression (Fig. 3e and Supplementary Video 1). Moreover, 20% of all divisions were performed by reporter-positive cells (Extended Data Fig. 7h). Conversely, 22% of cells that induced K10 reporter expression underwent division as a subsequent behaviour (Extended Data Fig. 7i). In the 5 days following their birth, most daughter cells (78%) from K10+ divisions had delaminated, completing the differentiation trajectory begun by their mother cell (Fig. 3f,g). More rarely (12%), daughter cells underwent an additional round of division. In cases when daughters from these subsequent divisions could be further tracked, we often witnessed them also delaminating (Extended Data Fig. 7j). These results contrast with K10− divisions, where 33% of daughters delaminated and 53% divided in the same time frame (Fig. 3g). Overall, of the 266 daughter cells whose subsequent behaviour after birth could be resolved within 5 days, 51% delaminated and 49% divided (Extended Data Fig. 7k), indicating that the population tracked in our short-term revisits is representative of the homeostatic basal population.
We next focused on basal K10-reporter-positive divisions to understand how they contribute to asymmetric or symmetric fate outcomes (Extended Data Fig. 1b, right)1,8. Among divisions where the behaviour of both daughter cells could be resolved in subsequent days of imaging, K10+ mother cells produced 76% of all divisions with two symmetrically differentiating daughters, 18% of all asymmetric divisions and no divisions with symmetrically dividing daughters (Fig. 3h,i). This preponderance of symmetric, differentiation-fated divisions contrasts with the largely asymmetric, self-renewing mode of division that has been proposed to characterize differentiation-primed progenitors in other models1,2. Together these results indicate that, although K10 reporter expression signifies commitment to eventually delaminate, differentiating cells remain capable of dividing to generate short-term lineages of basal daughter cells fated to delaminate.
We next sought to understand whether differentiating cells, just like their undifferentiated neighbours, divide owing to basal layer density changes when nearby cells delaminate25 (Fig. 4a). To test this, we quantified the behaviours taking place within 10 μm (a one-cell distance) preceding K10-reporter-positive division events. If these divisions occur in response to a neighbouring delamination, they should be preceded by the net loss of one nearby cell. If K10+ divisions occur cell autonomously, or in response to other cues, this imbalance will not be observed (Fig. 4a). Notably, a clear net loss of one neighbour preceded K10+ divisions, just like those of K10− cells, indicating that differentiating basal cells indeed divide as a response to delamination in their local neighbourhood (Fig. 4b and Extended Data Fig. 8a).
To directly test the role of K10+ divisions during epidermal homeostasis, we used the Krt10 promoter to induce the cell cycle inhibitor Cdkn1b (or p27) (ref. 26) in differentiating cells and monitored these cells with the K10 reporter (K10rtTA; pTRE-Cdkn1b; pTRE-H2BGFP) (Extended Data Fig. 8b). CDKN1B induction caused a rapid and penetrant block of proliferation in K10+ cells (Extended Data Fig. 8c) but did not alter epidermal thickness or expression of terminal markers (Extended Data Fig. 8d,e). The number of delaminating cells was comparable in tissue with and without CDKN1B induction, and all cells acquired K10 reporter expression before delamination regardless of genotype (Extended Data Fig. 8f,g), demonstrating that cell division is not required for the later maturation nor delamination of differentiating cells. Most interestingly, expression of CDKN1B did not affect the number of K10-reporter-positive basal cells (Fig. 4c,d), indicating that the size of the differentiating cell pool is maintained even when these cells are entirely unable to divide. Notably, the number of new K10-reporter-positive cells that emerged over a 24 h period was significantly higher in Cdkn1b mice (Extended Data Fig. 8h,i), and the number of basal cells expressing Krt10dim mRNA within the K10-reporter-negative population was increased (Extended Data Fig. 8j,k), suggesting a more rapid onset of differentiation-associated transcriptional changes in these mice. Thus, in the absence of amplifying K10 divisions, an increased number of basal cells initiate differentiation to maintain the size of the differentiating cell pool.
To next understand how the density of the basal layer was maintained even after 4 days of CDKN1B induction (Extended Data Fig. 8l), we directly tracked the number of cell divisions taking place in Cdkn1b versus control tissue over a 24 h period. We found that K10− basal cells in the Cdkn1b epidermis increased their proliferation rate to equal the number of divisions performed by both K10+ and K10− cells in control tissue (Fig. 4e,f). These results indicate that the homeostatic need for proliferation is normally satisfied by the contribution of both differentiating and undifferentiated cells. Thus, proliferation of differentiating cells in the adult epidermis occurs as a consequence of the need to replace delaminating basal neighbours, and not as part of an obligate transit-amplifying programme that fuels proper numbers of differentiating cells.
Finally, we evaluated the behaviours of differentiating basal cells in response to disruption of the epidermal barrier using tape stripping (Extended Data Fig. 9a). This perturbation caused a reduction of basal cell density (Extended Data Fig. 9b) due to a 40% depletion of K10+ basal cells through a rapid wave of delaminations within 24 h (Fig. 4g,h and Extended Data Fig. 9c). Notably, the differentiating basal pool in wild-type epidermis was quickly replenished to homeostatic levels within 3 days (Fig. 4g,h) through a combination of increased K10+ cell divisions and increased de novo K10 expression (Extended Data Fig. 9d–f). We observed a more severe initial reduction of the basal cell density and basal K10+ population when we performed the same perturbation in Cdkn1b epidermis (Fig. 4g,h and Extended Data Fig. 9b), indicating that K10+ cell divisions help to initially buffer the basal layer from excessive K10 cell loss. Consequently, Cdkn1b tissue was delayed in its ability to recover this population compared with wild-type epidermis (Fig. 4h and Extended Data Fig. 9b), and more undifferentiated basal cells initiated K10 expression in Cdkn1b tissue than in wild-type controls (Extended Data Fig. 9d,e). Thus, proliferation of K10+ basal cells helps replenish the differentiating cell pool during recovery after barrier disruption.
In summary, this study demonstrates the continuous, progressive and multi-day nature of the transcriptional changes associated with basal cell differentiation, reveals the cell-to-cell variability in the timing and behaviours associated with this process, and indicates that the onset of K10 expression both precedes downregulation of stem cell genes and marks cells committed to leaving the basal layer. We also show that the initiation of differentiation is both temporally and functionally uncoupled from cell cycle exit, with a subset of K10+ cells undergoing proliferation in response to local density changes. When this proliferative capacity is blocked, differentiating cells remain capable of maturing and stratifying normally, and the nearby stem cells compensate through increased rates of proliferation and differentiation. Even upon acute barrier disruption, the differentiating population recovers with only a mild delay when K10+ divisions are blocked.
Despite representing a robust marker of early basal cell differentiation, K10-deficient epidermis stratifies normally and forms a well-developed barrier27,28. Previous work has indicated highly complex transcriptional regulation of K10 expression via factors such as the TFAP2 and CEBP TF families and the Notch pathway29–34. Using SCENIC35 to analyse potential regulons, we identified several TFs that could play an important role in initiating differentiation and/or inhibiting stemness-related genes (Extended Data Fig. 6 and Methods), including Cebpb, Bhlhe40, Hes1, Id1, Ovol1 and Trp6319,31,36–38, and the interesting additional candidate Csnk2b, which plays a role in astrocyte differentiation39. Future studies that combine in vivo imaging of K10 dynamics with more precise spatial and temporal manipulation will help to unravel the upstream mechanisms driving epidermal stem cell differentiation.
Our results demonstrate that all delaminating cells transition through a K10-positive state in which they retain a limited capacity to divide, but proliferation is not necessary for their further maturation or stratification. This argues against the presence of a long-lived and functionally distinct committed epidermal progenitor population. At the same time, we reveal direct evidence of a molecular heterogeneity among the proliferating basal population. This helps to clarify the distinct results obtained by different lineage tracing approaches1–4, and explains the strong temporal correlation in lifetimes between sister cells that both go on to delaminate10. Taken together, our results reveal that a single continuous differentiation process shaped by feedback from the local environment fuels epidermal turnover.
Methods
Mice and experimental treatment
mTmG40, K14-CreER41, tetO-Cdkn1b26, Ivl-CreERT2 (ref. 42) and R26-tdTomato43 mice were obtained from the Jackson Laboratory. K10-rtTA15 mice were obtained from T. Lechler (Duke University), pTRE-H2BGFP44 mice were obtained from E. Fuchs (Rockefeller University), Lifeact-GFP45 mice were obtained from R. Weigert (National Institute of Dental and Craniofacial Research, National Institutes of Health (NIH)), GFP-NMMIIB46 mice were obtained from R. Adelstein (National Heart, Lung, and Blood Institute, NIH) and R26p-Fucci2 (ref. 47) mice were obtained from S. Aizawa (RIKEN). K14-H2BmCherry48 mice were generated in the laboratory and described previously. To visualize clonally labelled basal cells as they delaminate, K14CreER; mTmG mice were given a single dose of tamoxifen (20 μg/g in corn oil) 3 days before imaging. IvlCreERT2; R26-Tomato mice were treated with tamoxifen once at 8 weeks (0.2 mg g−1 in corn oil intra-peritoneally; Sigma Aldrich, catalogue number T5648), 2 days before cell isolation. To visualize K10 expression, K10-rtTA; pTRE-H2BGFP; mTmG mice were given doxycycline (2 mg ml−1) in drinking water with 2% sucrose continuously, starting 3 days before imaging. To block proliferation in the K10-positive population, littermates were genotyped to identify mutant (K10-rtTA; pTRE-H2BGFP; mTmG; tetO-Cdkn1b) and control (K10-rtTA; pTRE-H2BGFP; mTmG) mice. Animals were given doxycycline (2 mg ml−1) in drinking water with 2% sucrose continuously for the times specified. To disrupt the epidermal barrier via tape stripping, standard coloured lab tape (Fisher Scientific) was applied to the surface of the ear skin and removed in ten sequential repetitions. All mice used in this study were between 6 and 10 weeks old and were maintained either on a CD1 background (intra-vital imaging) or C57BL/6J background (single-cell RNA sequencing, scRNA-seq). Mice from experimental and control groups were randomly selected from either sex for live imaging experiments. Data collection and analysis were not performed blind to the conditions of the experiments. All procedures involving animal subjects were performed under the approval of the Institutional Animal Care and Use Committee of the Yale School of Medicine or the Linköping Animal Ethics Committee in accordance with Swedish legislation.
In vivo imaging
All imaging was performed in distal regions of the ear skin during prolonged telogen, with hair removed using depilatory cream (Nair) ≥3 days before the start of each experiment. Mice were anaesthetized with vapourized isofluorane delivered by a nose cone. Image stacks were acquired with a LaVision TriM Scope II (LaVision Biotec) laser scanning microscope equipped with both a Chameleon Vision II and Discovery two-photon lasers (Coherent). For collection of serial optical sections, the laser beam was focused through a 40× water immersion lens (Nikon; numerical aperture 1.15) and scanned with a field of view of 0.3 mm × 0.3 mm at 600 Hz. z-Stacks were acquired with 0.5–1 µm steps to image a total depth of ~40 µm of tissue, covering the entire thickness of the epidermis. Visualization of ECM was achieved via second harmonic signal using the blue channel at 940 nm imaging wavelength. To follow the same epidermal cells over multiple days, inherent landmarks of the skin together with a micro-tattoo were used to navigate back to the same epidermal regions every 12 or 24 h. For time-lapse imaging, serial optical sections were obtained in a range of 5–8 min intervals for a total duration of 1–3 h.
Image analysis
For quantifications in Figs. 1, 3, and 4c–h and Extended Data Figs. 1–5 and 7–9, raw image stacks were imported into Fiji49,50 for further analysis. Basal footprint and suprabasal spreading were quantified by manually outlining cell boundaries directly above the ECM signal and at the widest section in the top half of each cell, respectively. K10 reporter levels were quantified by measuring H2BGFP signal intensity at the widest plane of each cell nucleus. All values shown in this paper represent absolute fluorescence (with the exception of Fig. 4b; see below), and the same cut-off for positivity was applied to all cells within each experiment (1,500 for Figs. 1 and 3; 2,000 for Fig. 4c–h; for details of threshold for Fig. 4b, see below). K10 protein levels in whole-mount images were quantified by measuring average cytoplasmic signal intensity at the midpoint of each cell. Cell behaviours were tracked by visually comparing epidermal regions at subsequent timepoints, and cells were scored as suprabasal at the first timepoint when they made no observable contact with the underlying ECM signal. Prism software (GraphPad) was used to graph data and perform statistical analysis.
To track dynamics in the basal layer for Fig. 4b, we adapted the procedure from ref. 25. To height correct the 3D images, we first Gaussian blurred the signal from the ECM (mouse 1) or the epidermal cell nuclei (K14H2BmCherry) (mouse 2) spatially in the xy plane (width 6 μm) to create a 3D mask representing the region covering the whole epidermis. We then defined the height of the interface between the epidermis and the dermis from the 3D mask and subtracted this height from the original 3D data to level the basal layer position. From the height-corrected 3D images, we took three consecutive z-positions containing the nuclei of all the basal layer cells and averaged the intensity over the three slices to obtain 2D images in each channel. We calculated the local minima or maxima of the epithelial cell membranes (mem-tdTomato) or epithelial cell nuclei (K14H2BmCherry) to represent the cell positions in mouse 1 or mouse 2, respectively, and automatically corrected the shifts between time frames by minimizing the square distance between the nearest cell positions across the frames. The intensity of K10 reporter in the nucleus was calculated by taking the mean of the H2BGFP signal within a circle of 3 μm radius around the cell position (mouse 1) or by taking the sum within the segmented cell regions (mouse 2). After dividing the K10 signal levels by the mean within dividing cells for each region, we manually set a threshold to define cell divisions with high K10, as shown in Supplementary Fig. 8A (13 out of 49 cell divisions in mouse 1, 102 out of 374 cell divisions in mouse 2).
To assess neighbour fate imbalance before cell division, we took the sum over the net imbalance of the fates in the frames leading up to each division event. The net imbalance was calculated by taking the timepoint right after the division as zero, and going back in time while adding one (differentiation) or subtracting one (division) whenever there was a fate event within in the 10 μm neighbourhood of the division event of interest. We then averaged this net imbalance track over all the division events in mice 1 and 2. The error bar represents the fluctuation of net imbalance over each division event.
Single-molecule FISH image quantification was performed with the Cell Counter plugin in Fiji, and results were analysed with custom Python scripts. All quantifications were performed on n = 3 biological replicates.
Immunostaining
To isolate epidermis for whole-mount staining, ear and dorsal skin were incubated in 5 mg ml−1 dispase II solution (Sigma, 4942078001) at 37 °C for 10 min or 4 °C overnight, respectively, and the epidermis was separated from dermis using forceps. Epidermal tissue was fixed in 4% paraformaldehyde in PBS for 45 min at room temperature, washed in PBS, permeabilized and blocked for >1 h (0.2% Triton-X, 5% normal donkey serum and 1% BSA in PBS) and then incubated in primary antibodies overnight at 4 °C and secondary antibodies for 3 h at room temperature. For histological analysis of terminal differentiation markers, 10% formalin-fixed, paraffin-embedded (FFPE) skin was cut in 5 μm sections. Primary antibodies used were as follows: rabbit anti-K10 (1:1,000; Biolegend Poly19054), guinea pig anti-K10 (1:400; Progen GP-K10), rabbit anti-pH3 (1:1,000; Millipore 06-570), chicken anti-GFP (1:1,000; Invitrogen A10262), rabbit anti-involucrin (1:750; Biolegend Poly19244) and rabbit anti-loricrin (1:1,000; Biolegend Poly19051). All secondary antibodies used for immunofluorescence were raised in a donkey host and were conjugated to AlexaFluor 488, 568 or 647 (1:400; Thermo Fisher A78950, A21206, A10042, A31573, A21202, A10037 or A31507). When used, AlexaFluor 647 Phalloidin (Thermo Fisher) was incubated at the same time as secondary antibodies. 5-Ethynyl-2′-deoxyuridine (EdU) was administered via intra-peritoneal injection (50 μg g−1 in PBS) 2 h before collecting tissue, and EdU labelling was performing using the Click-iT AlexaFluor 568 kit (Thermo Fisher) according to the manufacturer’s instructions. Fixed whole-mount tissue was mounted on a slide with Vectashield anti-fade mounting medium (Vector Laboratories) with a #1.5 coverslip. For brightfield immunohistochemistry, biotinylated species-specific secondary antibodies followed by detection using the ABC kit (Vector Labs) and DAB kin (Vector Labs) were used according to the manufacturer’s instructions.
FISH
smRNA-FISH was performed using the RNAscope Multiplex Fluorescent Detection Kit v2 (323100, Advanced Cell Diagnostics) according to the manufacturer’s instructions using tyramide signal amplification (TSA) with Cy3, Cy5 and/or fluorescein (NEL760001KT, Perkin Elmer) on FFPE sections of dorsal skin. FFPE sections were hybridized with combinations of the following mRNA probes (all from Advanced Cell Diagnostics): Krt10 (457901 and 457901-C2), Krt14 (422521-C3), Krt5 (547901), Mt4 (447121-C3) and Krtdap (500671), together with cell membrane counterstaining using WGA (1:50, 29028-1 and 29059-1, Biotium). RFP (1:100, 600-401-379, Rockland) and GFP (1:200, ab13970, Abcam or 1:100 Cell Signaling 2965) immunohistochemistry was performed together with smRNA-FISH stainings according to manufacturer’s instructions with secondary antibodies that were raised in donkey or goat host and conjugated to AlexaFluor 405, 488, 546 or 647 (1:500; Thermo Fisher A10040, A11008, A21245 or A31556). Stainings were performed on skin samples isolated from the same mice that were used for ITGA6-sorted cell sequencing, wild-type 8-week-old mice, K10rtTA; pTRE-H2BGFP mice, or Cdkn1b mice and their respective controls. Images were acquired on a Nikon A1R spinning disk confocal as tiled images (10–15% overlap) and stitched by NIS Elements. Subsequently, all images were processed in the same way (maximum intensity projection, brightness adjustment and pseudocolouring) using Fiji36,37. In the case of Supplementary Fig. 8j, CDKN1B ear sample Krt10 staining bled through into the Krt14 channel, which was corrected for by subtracting Krt10 staining intensities from Krt14 intensities.
scRNA-seq and library preparation
For the scRNA-seq, epidermal cells were isolated from the back skin of 8-week-old Ivl-traced mice as described previously19. Cells were stained for 1 h with CD49f (Itga6)-AlexaFluor 647 (1:50; BD Biosciences; catalogue number 551129), Sca1(Ly6a)-PeCy7 (1:50; BD Biosciences; catalogue number 558162) and Cd34-FITC (1:50; BD Pharmingen catalogue number 553733) and Sytox blue (1:1,000; Life Technologies catalogue number S34857) was added just before (2 min before) FACS sorting on FACSAria III machine with BD FACSDiva 8.0.1 software (BD Biosciences). Tomato-traced and non-traced live cells gated for ITGA6+/SCA1+/CD34− were collected in 400 μl Defined Keratinocyte serum-free medium (Thermo Fisher, catalogue number 10744019) with 1% DNaseI (Stem Cell Technologies, catalogue number 07900), loaded into Fluidigm C1 chips (Fluidigm) for library preparation and sequenced as described previously19.
Data analysis
Pre-processing of the sequencing results into count matrices was performed as in ref. 19, and all subsequent data analysis was performed using Scanpy51. From the previously published datasets19,20 only cells that were categorized as ‘interfollicular epidermis cells’ (IFE) were included. Each dataset (ITGA6 sorted, Joost 2016 and Joost 2020 main dataset) was separately normalized using size factors and logarithmitized (ln(X + 1)), before filtering out non-expressed genes. Further, each dataset was regressed for the effects of total counts per cell, percentage of External RNA Controls Consortium (ERCC) spike-in counts for the ITGA6-sorted and Joost 2016 datasets (Fluidigm C1 based), and cell cycle (as scored by ‘score_genes_cell_cycle’ method). To combine all data, we first merged the C1-based datasets using the shared genes from the top 4,000 highly variable genes from each individual dataset. For integrating the Joost 2020 dataset, we used genes present in the C1-merged data. To compensate for the lower per-cell read counts in the Joost 2020 dataset (10x Chromium based), Scanpy implementation of MAGIC imputation52 was used with the following parameters specified: k-nearest neighbour (kNN) = 10; t = 2. This merged and imputed dataset revealed a small outlying group of cells (n = 9 cells) and a population of infundibulum cells (n = 147 cells), which were removed from subsequent analysis (Jupyter notebook). Cell cluster classification was performed on the imputed data using the previously defined differentiation trajectory from Joost 2016 as a reference. A kNN classifier (Scikit-learn53) was used to assign the cluster identity of cells from the ITGA6-sorted and Joost 2020 datasets according to the reference dataset with the following parameters specified: kNN = 20; weight = ‘distance’. For gene expression analyses, raw count matrices were log-normalized and downsampled to 2,000 counts per cell, compensating for the higher read counts in Fluidigm C1-based datasets, and merged. Pseudotime analysis was performed using diffusion pseudotime implementation in Scanpy51,54 and by ordering cells on the basis of their position along the pseudotime. The cut-off for basal and suprabasal populations was defined by the 95th percentile of ITGA6-sorted cells on the ordered differentiation pseudotime. The cut-off for cells to be classified as Krt10 positive was defined by the mean expression of Krt10 in all cells (1.84 log-normalized counts). For further analysis of gene expression over pseudotime, cells were grouped into six and four equally sized bins for basal and suprabasal populations, respectively. Fitted gene expression trends along the pseudotime were compared with the average expression levels in bin 1 (considered to be the most basal) and plotted as the change (log normalized) compared with this baseline expression. Spliced and unspliced mRNA analysis was performed on the annotated dataset from Joost 2020 (mapped with Velocyto package55). To avoid any bias due to sequencing methods used and by merging the datasets, cell cycle analysis was performed on the largest dataset (before regression), covering males and females of different ages20. Cell cycle phase was assigned with ‘score_genes_cell_cycle’ using a cut-off of 0.05 for positive classification. For mapping the Aragona et al. 2020 dataset onto our differentiation timeline, only those cells defined in Aragona et al. as ‘basal IFE’, ‘cycling IFE’ and ‘suprabasal IFE’ were included (log-normalized and subset to include only genes that are expressed in all datasets). Finally, each dataset was independently scaled to unit variance and zero mean before mapping the Aragona et al. dataset to our combined uniform manifold approximation and projection (UMAP) using ingest implementation in Scanpy.
SCENIC analysis
SCENIC analysis was performed using the pySCENIC (v0.11.2) Python package together with Scanpy (v1.7.2) and loompy (v3.0.6) for data processing and visualization. Overall, the analysis was performed as described in the full SCENIC analysis example and the extended analysis examples with default parameters. Network inference was performed with GRNBoost2 algorithm using the downsampled dataset from our analysis and the murine TF list (mm_mgi_tfs.txt) supplied by SCENIC. Regulon and motif prediction was done within 10 kb of the transcription start site, using the following databases (from cisTarget): mm10__refseq-r80__10kb_up_and_down_tss.mc9nr.feather and motifs-v9-nr.mgi-m0.001-o0.0.tbl. Unfortunately, we did not find enriched motifs with a clear link to differentiation (for example, targeting Krt10) during the downstream analysis.
To identify differentiation associated TFs and their potential target genes, we performed the following analysis. (1) Find genes that are correlated with Krt10 expression (Pearson r > 0.3) during the earliest differentiation steps (bins 2, 3 and 4). (2) Identify TFs that target any of these differentiation-associated genes (based on the adjacency list from GRNBoost2 analysis), and include genes that are targeted by the same TF. (3) Give each cell a score based on the expression levels of the target genes for each of the TF (using scanpy.tl.score_genes() function). Note that this step in SCENIC analysis includes genes that are both up- and downregulated during the differentiation process (that is, TF expression levels can be used to predict the expression of the gene), so the resulting target gene list also included several basal-layer-associated genes (that is, Krt14, Krt5, Col17a1, Mt2 and so on). To show only the relationships between TF and differentiation, we first identified differentially expressed genes (using scanpy.tl.rank_genes_groups()) in bin 1 (adjusted P value < 0.05, log fold change >0.5) and excluded them from gene scoring.
Statistics and reproducibility
Statistical parameters including the exact value of n for each experiment and statistical significance are reported in figure legends. Significance was determined using an unpaired Student’s t-test, pairwise Tukey’s honest significant difference (HSD) when multiple comparisons were made or analysis of variance (ANOVA) for the time-series analysis (see below). Asterisks denote statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001). Statistical calculations were performed using the Prism software package (GraphPad), Scanpy for differential gene expression analysis or statsmodels (v0.12.2) library in Python for time-series analysis. No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications10,19,20,25. All figures represent data from at least three independent experiments from at least n = 3 mice, except Figs. 1b,c and 3h and Extended Figs. 1e, 7g and 8k, which represent data from one independent experiment with n = 3 mice.
Comparison of the effects of Cdkn1b mouse model on the number of K10+ basal cells (Fig. 4h) and basal cell density (Supplementary Fig. 9b) after tape stripping was performed as follows. First, cubic linear models were constructed (statsmodels.regression.linear_model.OLS) to explain the observed proportion of cells depending either only on the day after tape stripping (H0) or on a combination of the day and genotype (H1):
where Y is proportion of cells, X is day after tape stripping and G is genotype.
Subsequently, ANOVA for linear models (statsmodels.stats.anova.anova_lm) was used to compare if the model with the addition of genotype (H1) is better at explaining the observed data than the model using only days (H0), using the default F-test.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41556-022-01021-8.
Supplementary information
Acknowledgements
We thank all members of the Greco and Kasper labs for critical feedback on the manuscript. We thank T. Lechler for K10-rtTA mice, E. Fuchs for pTRE-H2BGFP mice, R. Weigert for Lifeact-GFP mice, R. Adelstein for GFP-NMMIIB mice and S. Aizawa for R26p-Fucci2 mice. This work is supported by an HHMI Scholar award and NIH grants number R01AR063663, R01AR072668 and DP1AG066590 (V.G.), grants from the Swedish Foundation for Strategic Research (FFL12-0133), Swedish Cancer Society (CAN 2018/793; 21 1821 Pj), Swedish Research Council (2018-02963), LEO Foundation (GOLD award 2017), Gösta Miltons Donationsfond, Karolinska Institutet (2-2111/2019) and StratRegen (SFO at KI) (M.K.), and JSPS KAKENHI grant numbers JP18H04760, JP18K13515, JP19H05275 and JP19H05795 and the Human Frontier Science Program (K.K.). K.C. was supported by the Canadian Institutes of Health Research and a New York Stem Cell Foundation Druckenmiller Fellowship. K.A. was supported by PhD (KID) funding from Karolinska Institutet.
Extended data
Source data
Author contributions
K.C., K.A., K.K., M.K. and V.G. designed experiments and analysed data. K.C. performed two-photon imaging, whole-mount staining, mouse genetics and image analysis. K.A. performed Ivl-CreERT2 mouse genetics, smRNA-FISH and image analysis, and scRNA-seq experiments. K.A. and M.K. performed single-cell transcriptome analysis and interpretation. D.G.G. and D.P.M. performed data analysis. K.K. performed data analysis and statistical modelling. S.G. assisted with whole mounts and mouse genetics. K.R.M. performed time-lapse imaging. K.C., K.A., M.K. and V.G. wrote the manuscript with input from all authors.
Peer review
Peer review information
Nature Cell Biology thanks Salvador Benitah, Philip Jones and Sabine Werner for their contribution to the peer review of this work.
Funding
Open access funding provided by Karolinska Institute.
Data availability
Sequencing data that support the findings of this study have been deposited in the Gene Expression Omnibus (GEO) under accession code GSE152044. Previously published scRNA-seq data that were re-analysed here are available under accession codes GSE129218, GSE67602 and GSE146637. Annotated and analysed sequencing data have been deposited in Zenodo: 10.5281/zenodo.6998285. Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding authors on reasonable request.
Code availability
The complete computational analysis workflow for the scRNAseq is available in the form of Jupyter notebooks at https://github.com/kasperlab. The Python scripts for the image analysis will be available on request.
Competing interests
The authors declare no competing financial interests.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Katie Cockburn, Karl Annusver.
Contributor Information
Maria Kasper, Email: maria.kasper@ki.se.
Valentina Greco, Email: valentina.greco@yale.edu.
Extended data
is available for this paper at 10.1038/s41556-022-01021-8.
Supplementary information
The online version contains supplementary material available at 10.1038/s41556-022-01021-8.
References
- 1.Mascré G, et al. Distinct contribution of stem and progenitor cells to epidermal maintenance. Nature. 2012;489:257–262. doi: 10.1038/nature11393. [DOI] [PubMed] [Google Scholar]
- 2.Sánchez-Danés A, et al. Defining the clonal dynamics leading to mouse skin tumour initiation SCs but not CPs initiate BCC formation. Nature. 2016;536:298–303. doi: 10.1038/nature19069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ichijo R, et al. Vasculature-driven stem cell population coordinates tissue scaling in dynamic organs. Sci. Adv. 2021;7:eabd2575. doi: 10.1126/sciadv.abd2575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sada A, et al. Defining the cellular lineage hierarchy in the interfollicular epidermis of adult skin. Nat. Cell Biol. 2016;18:619–631. doi: 10.1038/ncb3359. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Barker N. Adult intestinal stem cells: critical drivers of epithelial homeostasis and regeneration. Nat. Rev. Mol. Cell Biol. 2013;15:19–33. doi: 10.1038/nrm3721. [DOI] [PubMed] [Google Scholar]
- 6.Zhang B, Hsu YC. Emerging roles of transit-amplifying cells in tissue regeneration and cancer. Wiley Interdiscip. Rev. Dev. Biol. 2017;6:1–14. doi: 10.1002/wdev.282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lavker RM, Sun T-T. Epithelial stem cells: the eye provides a vision. Eye. 2003;17:937–942. doi: 10.1038/sj.eye.6700575. [DOI] [PubMed] [Google Scholar]
- 8.Clayton E, et al. A single type of progenitor cell maintains normal epidermis. Nature. 2007;446:185–189. doi: 10.1038/nature05574. [DOI] [PubMed] [Google Scholar]
- 9.Doupé DP, Klein AM, Simons BD, Jones PH. The ordered architecture of murine ear epidermis is maintained by progenitor cells with random fate. Dev. Cell. 2010;18:317–323. doi: 10.1016/j.devcel.2009.12.016. [DOI] [PubMed] [Google Scholar]
- 10.Rompolas P, et al. Spatiotemporal coordination of stem cell commitment during epidermal homeostasis. Science. 2016;352:1471–1474. doi: 10.1126/science.aaf7012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Piedrafita G, et al. A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial maintenance in mice. Nat. Commun. 2020;11:1–15. doi: 10.1038/s41467-020-15258-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Braun KM, et al. Manipulation of stem cell proliferation and lineage commitment: visualisation of label-retaining cells in wholemounts of mouse epidermis. Development. 2003;130:5241–5255. doi: 10.1242/dev.00703. [DOI] [PubMed] [Google Scholar]
- 13.Schweizer J, Kinjo M, Furstenberger G, Winter H. Sequential expression of mRNA-encoded keratin sets in neonatal mouse epidermis: basal cells with properties of terminally differentiating cells. Cell. 1984;37:159–170. doi: 10.1016/0092-8674(84)90311-8. [DOI] [PubMed] [Google Scholar]
- 14.Connelly JT, et al. Actin and serum response factor transduce physical cues from the microenvironment to regulate epidermal stem cell fate decisions. Nat. Cell Biol. 2010;12:711–718. doi: 10.1038/ncb2074. [DOI] [PubMed] [Google Scholar]
- 15.Muroyama A, Lechler T. A transgenic toolkit for visualizing and perturbing microtubules reveals unexpected functions in the epidermis. eLife. 2017;6:e29834. doi: 10.7554/eLife.29834. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Eisenhoffer GT, et al. Crowding induces live cell extrusion to maintain homeostatic cell numbers in epithelia. Nature. 2012;484:546–549. doi: 10.1038/nature10999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Marinari E, et al. Live-cell delamination counterbalances epithelial growth to limit tissue overcrowding. Nature. 2012;484:542–545. doi: 10.1038/nature10984. [DOI] [PubMed] [Google Scholar]
- 18.Miroshnikova YA, et al. Adhesion forces and cortical tension couple cell proliferation and differentiation to drive epidermal stratification. Nat. Cell Biol. 2018;20:69–80. doi: 10.1038/s41556-017-0005-z. [DOI] [PubMed] [Google Scholar]
- 19.Joost, S. et al. Single-cell transcriptomics reveals that differentiation and spatial signatures shape epidermal and hair follicle heterogeneity. Cell Syst. 1–27 (2016). [DOI] [PMC free article] [PubMed]
- 20.Joost S, et al. The molecular anatomy of mouse skin during hair growth and rest. Cell Stem Cell. 2020;26:441–457.e7. doi: 10.1016/j.stem.2020.01.012. [DOI] [PubMed] [Google Scholar]
- 21.Jones KB, et al. Quantitative clonal analysis and single-cell transcriptomics reveal division kinetics, hierarchy, and fate of oral epithelial progenitor cells. Cell Stem Cell. 2019;24:183–192.e8. doi: 10.1016/j.stem.2018.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Aragona M, et al. Mechanisms of stretch-mediated skin expansion at single-cell resolution. Nature. 2020;584:268–273. doi: 10.1038/s41586-020-2555-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Drexler HL, Choquet K, Churchman LS. Splicing kinetics and coordination revealed by direct nascent RNA sequencing through nanopores. Mol. Cell. 2020;77:985–998.e8. doi: 10.1016/j.molcel.2019.11.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Singh J, Padgett RA. Rates of in situ transcription and splicing in large human genes. Nat. Struct. Mol. Biol. 2009;16:1128–1133. doi: 10.1038/nsmb.1666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Mesa KR, et al. Homeostatic epidermal stem cell self-renewal is driven by local differentiation. Cell Stem Cell. 2018;23:677–686.e4. doi: 10.1016/j.stem.2018.09.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Pruitt SC, Freeland A, Rusiniak ME, Kunnev D, Cady GK. Cdkn1b overexpression in adult mice alters the balance between genome and tissue ageing. Nat. Commun. 2013;4:1–12. doi: 10.1038/ncomms3626. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Reichelt J, Büssow H, Grund C, Magin TM. Formation of a normal epidermis supported by increased stability of keratins 5 and 14 in keratin 10 null mice. Mol. Biol. Cell. 2001;12:1557–1568. doi: 10.1091/mbc.12.6.1557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wallace L, Roberts-Thompson L, Reichelt J. Deletion of K1/K10 does not impair epidermal stratification but affects desmosomal structure and nuclear integrity. J. Cell Sci. 2012;125:1750–1758. doi: 10.1242/jcs.097139. [DOI] [PubMed] [Google Scholar]
- 29.Maytin EV, et al. Keratin 10 gene expression during differentiation of mouse epidermis requires transcription factors C/EBP and AP-2. Dev. Biol. 1999;216:164–181. doi: 10.1006/dbio.1999.9460. [DOI] [PubMed] [Google Scholar]
- 30.Zhu S, et al. C/EBPbeta modulates the early events of keratinocyte differentiation involving growth arrest and keratin 1 and keratin 10 expression. Mol. Cell. Biol. 1999;19:7181–7190. doi: 10.1128/MCB.19.10.7181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Blanpain C, Lowry WE, Pasolli HA, Fuchs E. Canonical notch signaling functions as a commitment switch in the epidermal lineage. Genes Dev. 2006;20:3022–3035. doi: 10.1101/gad.1477606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Wang X, Pasolli HA, Williams T, Fuchs E. AP-2 factors act in concert with Notch to orchestrate terminal differentiation in skin epidermis. J. Cell Biol. 2008;183:37–48. doi: 10.1083/jcb.200804030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Lopez RG, et al. C/EBPα and β couple interfollicular keratinocyte proliferation arrest to commitment and terminal differentiation. Nat. Cell Biol. 2009;11:1181–1190. doi: 10.1038/ncb1960. [DOI] [PubMed] [Google Scholar]
- 34.Kim DS, et al. The dynamic, combinatorial cis-regulatory lexicon of epidermal differentiation. Nat. Genet. 2021;53:1564. doi: 10.1038/s41588-021-00947-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Van de Sande B, et al. A scalable SCENIC workflow for single-cell gene regulatory network analysis. Nat. Protoc. 2020;15:2247–2276. doi: 10.1038/s41596-020-0336-2. [DOI] [PubMed] [Google Scholar]
- 36.Genander, M. et al. BMP signaling and its pSMAD1/5 target genes differentially regulate hair follicle stem cell lineages. Stem Cell 1–16 (2014). [DOI] [PMC free article] [PubMed]
- 37.Truong AB, Kretz M, Ridky TW, Kimmel R, Khavari PA. p63 regulates proliferation and differentiation of developmentally mature keratinocytes. Genes Dev. 2006;20:3185–3197. doi: 10.1101/gad.1463206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Lee B, et al. Transcriptional mechanisms link epithelial plasticity to adhesion and differentiation of epidermal progenitor cells. Dev. Cell. 2014;29:47–58. doi: 10.1016/j.devcel.2014.03.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Yang CP, et al. Comprehensive integrative analyses identify GLT8D1 and CSNK2B as schizophrenia risk genes. Nat. Commun. 2018;9:1–16. doi: 10.1038/s41467-018-07401-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Muzumdar MD, Tasic B, Miyamichi K, Li L, Luo L. A global double-fluorescent Cre reporter mouse. Genesis. 2007;45:593–605. doi: 10.1002/dvg.20335. [DOI] [PubMed] [Google Scholar]
- 41.Vasioukhin V, Degenstein L, Wise B, Fuchs E. The magical touch: genome targeting in epidermal stem cells induced by tamoxifen application to mouse skin. Proc. Natl Acad. Sci. USA. 1999;96:8551–8556. doi: 10.1073/pnas.96.15.8551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Lapouge G, et al. Identifying the cellular origin of squamous skin tumors. Proc. Natl Acad. Sci. USA. 2011;108:7431–7436. doi: 10.1073/pnas.1012720108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Madisen L, et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 2010;13:133–140. doi: 10.1038/nn.2467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Tumbar T. Defining the epithelial stem cell niche in skin. Science. 2004;303:359–363. doi: 10.1126/science.1092436. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Riedl J, et al. Lifeact mice for studying F-actin dynamics. Nat. Methods. 2010;7:168–169. doi: 10.1038/nmeth0310-168. [DOI] [PubMed] [Google Scholar]
- 46.Bao J, Ma X, Liu C, Adelstein RS. Replacement of nonmuscle myosin II-B with II-A rescues brain but not cardiac defects in mice. J. Biol. Chem. 2007;282:22102–22111. doi: 10.1074/jbc.M702731200. [DOI] [PubMed] [Google Scholar]
- 47.Abe T, et al. Visualization of cell cycle in mouse embryos with Fucci2 reporter directed by Rosa26 promoter. Development. 2012;140:237–246. doi: 10.1242/dev.084111. [DOI] [PubMed] [Google Scholar]
- 48.Mesa KR, et al. Niche-induced cell death and epithelial phagocytosis regulate hair follicle stem cell pool. Nature. 2015;522:94–97. doi: 10.1038/nature14306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Rueden CT, et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinform. 2017;18:529. doi: 10.1186/s12859-017-1934-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Schindelin J, et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods. 2012;9:676–682. doi: 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 2018;19:15. doi: 10.1186/s13059-017-1382-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.van Dijk D, et al. Recovering gene interactions from single-cell data using data diffusion. Cell. 2018;174:716–729.e27. doi: 10.1016/j.cell.2018.05.061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Pedregosa, F. et al. scikit-learn: machine learning in Python. scikit-learnhttp://scikit-learn.sourceforge.net (2011).
- 54.Haghverdi L, Büttner M, Wolf FA, Buettner F, Theis FJ. Diffusion pseudotime robustly reconstructs lineage branching. Nat. Methods. 2016;13:845–848. doi: 10.1038/nmeth.3971. [DOI] [PubMed] [Google Scholar]
- 55.La Manno G, et al. RNA velocity of single cells. Nature. 2018;560:494–498. doi: 10.1038/s41586-018-0414-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Sequencing data that support the findings of this study have been deposited in the Gene Expression Omnibus (GEO) under accession code GSE152044. Previously published scRNA-seq data that were re-analysed here are available under accession codes GSE129218, GSE67602 and GSE146637. Annotated and analysed sequencing data have been deposited in Zenodo: 10.5281/zenodo.6998285. Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding authors on reasonable request.
The complete computational analysis workflow for the scRNAseq is available in the form of Jupyter notebooks at https://github.com/kasperlab. The Python scripts for the image analysis will be available on request.