Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial maintenance in mice

A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial... ARTICLE https://doi.org/10.1038/s41467-020-15258-0 OPEN A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial maintenance in mice 1,2 3 1 1 1 Gabriel Piedrafita , Vasiliki Kostiou , Agnieszka Wabik , Bartomeu Colom , David Fernandez-Antoran , ✉ ✉ 1 1 3 1,3 Albert Herms , Kasumi Murai , Benjamin A. Hall & Philip H. Jones In adult skin epidermis and the epithelium lining the esophagus cells are constantly shed from the tissue surface and replaced by cell division. Tracking genetically labelled cells in trans- genic mice has given insight into cell behavior, but conflicting models appear consistent with the results. Here, we use an additional transgenic assay to follow cell division in mouse esophagus and the epidermis at multiple body sites. We find that proliferating cells divide at a similar rate, and place bounds on the distribution cell cycle times. By including these results in a common analytic approach, we show that data from eight lineage tracing experiments is consistent with tissue maintenance by a single population of proliferating cells. The outcome of a given cell division is unpredictable but, on average, the likelihood of producing pro- liferating and differentiating cells is equal, ensuring cellular homeostasis. These findings are key to understanding squamous epithelial homeostasis and carcinogenesis. 1 2 Wellcome Sanger Institute, Hinxton CB10 1SA, UK. Spanish National Cancer Research Centre (CNIO), C/Melchor Fernández Almagro 3, Madrid 29029, Spain. MRC Cancer Unit, University of Cambridge, Hutchison-MRC Research Centre, Box 197, Cambridge Biomedical Campus, Cambridge CB2 0XZ, UK. email: bh418@mrc-cu.cam.ac.uk; pj3@sanger.ac.uk NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 1 1234567890():,; ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 he squamous epithelia that cover the external surface of the and sweat ducts, which form distinct proliferative compartments 5,8–10 body and line the mouth and esophagus consist of layers of independent of the epidermis (Fig. 1a) . The structure of the Tkeratinocytes. In the mouse epidermis and esophagus cell epidermis also varies with body site. In typical mouse epidermis, division is confined to the deepest, basal cell layer (Fig. 1a). On such as that on the back (dorsum), hair follicles are frequent but commitment to terminal differentiation, proliferating cells exit there are no sweat glands . In contrast, in the mouse paw epi- the cell cycle and migrate to the suprabasal cell layers, before dermis hair and, particularly, sweat ducts are common in the being ultimately shed from the tissue surface. Cellular home- anterior, acrosyringial region around the foot pads, while the 10–12 ostasis requires that cells are generated by proliferation at the posterior, plantar epidermis is devoid of appendages . The ear same rate at which they are shed. Further, to maintain a constant epidermis is different again; it has uniform columns of differ- number of proliferating cells, on average each cell division must entiating cells, not present elsewhere . Finally, the mouse tail has generate one daughter that will go on to divide and one that will the most unusual structure, being a scale forming epidermis like differentiate after first exiting the cell cycle. However, the nature that of chicken legs and Crocodillia rather than typical mam- 1–5 14–16 of the dividing cell population has been subject to controversy . malian skin . This structural diversity has motivated a range Resolving proliferating cell behavior is key for understanding not of studies to define the properties of proliferating cells at each site. only normal tissue maintenance but also processes such as wound Genetic lineage tracing in transgenic mice has emerged as a healing and the accumulation of somatic mutations in normal powerful technique for tracking the behavior of cells within tis- 6,7 17 tissues during aging and carcinogenesis . sues (Fig. 1b) . This is performed in mice expressing two Whilst murine epidermis and esophageal epithelium share the transgenic constructs (Fig. 2a,b). The first is a genetic switch, same basic organization, there are significant differences between using a bacterial recombinase enzyme Cre, expressed either from the tissues. The esophageal epithelium is uniform, with no a transgenic promoter or targeted to a specific gene . A variety of appendages, while the epidermis is punctuated by hair follicles Cre expressing mouse strains have been used for studies of Cell loss Differentiation Skin Oesophagus Proliferation SG HF Lineage tracing Induction Time chase H2BGFP pulse-chase +Dox –Dox Time chase SP model 2xSC model SC-CP model Slow-cycling Rapidly-dividing Progenitor Differentiating stem cell stem cell Fig. 1 Quantitative approaches to cell behavior in murine epithelium. a Structure of the stratified squamous epithelia from the interfollicular epidermis (skin) and esophagus of adult mice. Proliferating keratinocytes are located in the basal layer. Upon differentiation they migrate through suprabasal layers until they are ultimately lost by shedding. A balance should be established between cell division and cell loss to guarantee tissue homeostasis. HF, hair follicle; SG, sebaceous gland. b Rationale of genetic lineage tracing. Low-dose induction in transgenic mice allows recombination and conditional labeling of punctuated keratinocyte progenitors in the basal layer. These cells and their progeny remain labeled and can be tracked to study clonal dynamics over time. c Rationale of Histone 2B-GFP (H2BGFP) dilution experiments (a top-down view of the basal-layer plane is sketched). Transgenic-mouse keratinocytes express H2BGFP protein while on doxycycline (Dox) treatment. After Dox withdrawal cycling cells dilute their H2BGFP content with every division, allowing to study cell-proliferation rate. d Different stochastic cell-proliferation models invoked to explain epithelial self-renewal. Branches reflect different possible fates for a given proliferating cell upon division. SP: single-progenitor model. 2xSC: two stem-cell model, involving two independent types of proliferating cells dividing at different rates. SC-CP: stem cell-committed progenitor model, involving slow-cycling stem cells underpinning a second population of quickly-dividing progenitors. 2 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE Lineage tracing Endogenous X-promoter Tissue Reference ERT locus X cre Axin2 Paw 11 Cre +Tamoxifen (knockin) Lgr6 Back 33 Krt15 Esophagus 31 Reporter R26 R26 STOP Reporter Lrig1 Esophagus (this work) Y-promoter Tissue Reference Cyp1a1 (Ah) Tail 3 Transgene ERT Cyp1a1 (Ah) Ear 32 // // Y cre (random Cre Cyp1a1 (Ah) Esophagus 5 +Tamoxifen insertion) Cyp1a1 (Ah) Back 22 (+β-NF) Reporter R26 Ivl Tail 4,21 R26 STOP Reporter Krt14 Tail 4,21 H2BGFP pulse-chase X-promoter Tissue Reference tTA Krt5 Tail 4 w/o Dox tTA Tet-Off x Krt5 Back 20 tetO H2BGFP H2B-GFP tetO Y-promoter Tissue Reference rtTA Dox d +Dox Gt(ROSA)26Sor Esophagus 5 Tet-On x rtTA Gt(ROSA)26Sor Back 22 tetO H2BGFP Gt(ROSA)26Sor All (this work) H2B-GFP tetO Fig. 2 Transgenic-mouse models used for lineage tracing and cell-proliferation studies. a, b Transgenic mice for lineage tracing are designed with two genetic constructs. The first codes for a bacterial Cre recombinase- mutant estrogen receptor fusion protein (CreERT), which can be targeted to a specific endogenous locus (a) or be under control of a transgenic promoter, randomly inserted in the genome (b). The second construct codes for a conditional fluorescent protein reporter, typically targeted to the ubiquitously expressed Rosa26 locus. Treatment with tamoxifen induces Cre protein internalization to the nucleus, allowing expression of the reporter following Cre-mediated excision of a loxP-flanked STOP cassette. Specific details of constructs used in previous literature for lineage tracing in squamous epithelia are listed on the right. Note expression of Cre from the transgenic Cyp1a1 arylhydrocarbon receptor, Ah) promoter requires additional treatment with a Ah inducer, β-napthoflavone (β-NF). c, d Transgenic mice for H2BGFP-dilution experiments are designed with a first construct, typically targeted to a constitutive promoter, coding either for a tetracycline-controlled transactivator (tTA; Tet-Off system) (c) or a reverse tetracycline-controlled transactivator (rtTA; Tet-On system) (d). A second construct codes for a Histone 2B-green fluorescent protein fusion (H2BGFP) controlled by a tetracycline-response promoter element (tetO; sometimes referred to as pTRE). Treatment with tetracycline or its derivative doxycycline (Dox) preempts tTA protein from binding to tetO elements in Tet-Off systems, causing repression of pTRE-controlled H2BGFP expression, whilst it is required for binding of rtTA to tetO elements in Tet-On systems, hence having an opposite effect. Dox is administered for induction and withdrawn during the H2BGFP-dilution chase in Tet-On mice, while in Tet-Off animals its application gets required for the duration of the experiment. esophageal epithelium and epidermis (Fig. 2a,b). Cre is fused to a The stable H2B-GFP protein is diluted by cell division, so if the mutant hormone receptor so it is only active following treatment tissue contains cell populations dividing at different rates, the with a drug, giving control over when recombination is induced. more slowly dividing cells will retain higher levels of protein . Using low doses of inducing drug allows the labeling of scattered Measurements of the rate of loss of fluorescence have been used 4,5,20 single cells. The second construct is a reporter, such as a fluor- to estimate the rate of cell division . escent protein, typically targeted to the ubiquitously expressed Gt Lineage tracing has ruled out older deterministic models of a (ROSA)26Sor (Rosa26) locus. The reporter is only expressed fol- proliferative hierarchy of asymmetrically dividing stem cells lowing the excision of a “stop” cassette by Cre and expression generating ‘transit amplifying’ cells that undergo a fixed number persists in the progeny of the labeled cell. If the cells are labeled at of divisions prior to differentiation . These models predict that a low frequency, single-cell-derived clones of reporter expressing clone sizes will rise and then remain stable. In multiple lineage cells result. If a representative sample of proliferating cells is tracing experiments, however, mean clone size has been found to labeled and their progeny tracked over a time course, statistical increase progressively with time. However, several mutually analysis of the evolving clone-size distributions may be used to incompatible models in which proliferating cells have stochastic infer cell behavior . fate have been proposed that do appear consistent with the data Alongside lineage tracing, a complementary transgenic assay in one or more experiments (Fig. 1d; Supplementary Methods). may be used to detect cells cycling at different rates and infer the The simplest stochastic model, the single-progenitor (SP) average rate of cell division (Fig. 1c). This uses a transgenic, drug hypothesis, proposes that all dividing keratinocytes are func- regulated synthetic promoter to control expression of a protein tionally equivalent and generate dividing and differentiating 3,5 comprising Histone 2B fused to green fluorescent protein (H2B- daughters with equal probability . An alternative stem cell- GFP) (Fig. 2c, d). The H2B-GFP is initially expressed at high committed progenitor (SC-CP) paradigm, applied to the epi- levels in keratinocytes. Its transcription is then shut off and levels dermis proposes a hierarchy of rare, slowly cycling stem cells of H2B-GFP protein measured by microscopy or flow cytometry. which generate stem and progenitor daughters. The progenitors NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 are biased toward differentiation so continual stem cell pro- populations (Figs. 3d, 4; Supplementary Data 3; Supplementary 4,21 liferation is required . A third model argues that two inde- Methods). pendent populations of stem cells (2xSC) dividing at different To further challenge the hypothesis that there is a single rates exist in the epidermis . These models all give comparably proliferating cell population, we examined whether this model good fits to the results from individual experiments. However, can recapitulate the observed H2BGFP intensity distributions at each has been proposed on the basis of distinct data sets analyzed each time point. For a given average division rate, we performed by different inference and fitting procedures, with limited testing simulations of H2BGFP-dilution kinetics under a wide range of of alternative hypotheses. possible underlying (Gamma) distributions for individual cell- Motivated by the disparity of the proposed models of cell cycle times (Figs. 3e, 4b, d, f, Supplementary Methods). We find dynamics we set out to determine if a single model was consistent that the form of the H2BGFP histograms over time can indeed be across multiple data sets in both esophagus and different epi- fully described by a single population of cells, dividing within a dermal regions. We use cell-cycle properties from the H2B-GFP relatively narrow range of cell-cycle times, further supporting the dilution data to fit lineage tracing results by maximum likelihood SP model (Figs. 3c, f, g, 4; Supplementary Data 4). parameter inference. We find that the data are consistent with a Altogether, these observations strongly argue against scenarios simple SP model of homeostasis. We also show that the fates of of heterogeneous proliferating cell populations, such as the SC- pairs of sister cells are anti-correlated, and that the basal layer CP or 2xSC models, at all sites other than in the tail where contains a substantial proportion of cells which will differentiate marked variation between animals precluded reliable inference on rather than going on to divide. cell-proliferation rates (Fig. 5c). We conclude that in the basal layer of the epidermis at multiple body sites and in the esophagus proliferating cells divide at a unique average rate with highly Results homogenous cell-cycle periods, consistent with the SP model Cell-cycle times in epidermis and esophagus. Analysis of cell (Table 1). proliferation in epithelia offers a simple way to test the predic- tions of the disparate models of epithelial homeostasis by iden- tifying the level of heterogeneity in the division rate of basal-layer A common analytical approach to resolve cell behavior. The cells. The SP model predicts a single-cell population dividing at ability of lineage tracing to track the behavior of cohorts of the same average rate while the alternative hypotheses argue for proliferating cells and their progeny over time courses extending discrete populations dividing at different average rates. We to many rounds of cell division offers the potential to validate therefore investigated the dilution of H2B-GFP in the epidermis models of homeostasis. Having established the homogeneity in M2rtTA and esophagus of R26 /TetO-H2BGFP mice (Fig. 3a). The the division rate of basal-layer cells, we then set out to determine animals were treated with doxycycline (Dox) for 4 weeks to whether clonal dynamics across different lineage tracing data sets induce H2BGFP expression. Dox was then withdrawn and were consistent with the SP paradigm. H2BGFP protein levels in individual basal keratinocytes tracked Multiple lineage tracing studies have been published but these by direct, in situ measurement of GFP fluorescence from confocal used distinct approaches to infer models of cell behavior and did images of epithelial wholemounts at multiple time points. We not apply the additional constraint imposed by measuring the 3,4,11 examined esophagus and epidermis from plantar area of the cell-cycle time distribution . Computational simulations hindpaw, ear, and tail (Figs. 3c, 4, and 5; Supplementary showed that the SP, SC-CP and 2xSC models all predict very similar development of clonal features over time, which rendered Movies 1–3). Optical sections through the deepest, basal cell layer were taken over at least 5 fields of view per tissue/animal and them hardly distinguishable from lineage tracing data alone H2BGFP fluorescence quantified for all non-mitotic nuclei fol- (Supplementary Methods). However, as our cell-proliferation lowing image segmentation based on DAPI staining (Fig. 3b; analyses do not support the SC-CP and 2xSC paradigms we Methods). Non-epithelial cells in the form of CD45 leukocytes, focused on testing the SP model. which retain high levels of H2BGFP, were excluded from the By incorporating the measurement of the average division rate, analysis, but served as internal reference for label retention (e.g., we could reduce the uncertainty in the parameter estimation, a Fig. 3c, insert; Figs. 4b, d, f and 5b, Supplementary Data 1). In problem that has been generally overlooked in these stochastic addition, for the analysis below we included a recently published models (Supplementary Fig. 1A-B). For example, a relatively high data set from dorsal epidermis performed using an identical division rate and modest proportion of symmetric division protocol (Fig. 4f; Supplementary Movie 4). outcomes predict a similar clone-size distribution to those with We first examined images for the presence of label-retaining a slower turnover rate but higher level of symmetric divisions. In cells (LRCs) (Supplementary Data 1). We found no keratinocyte turn, whilst long-term model predictions on clone-size distribu- LRC in the basal cell layer of the esophagus or any epidermal site tions remained largely unaffected by the assumptions on the cell- other than the interscale region of the tail (Fig. 5a). Rare cycle time distribution, introducing realistic estimates for the keratinocyte LRCs (4/1923, i.e., 0.2% of basal-layer keratinocytes) distribution of individual cell-cycle lengths affected short-term were observed in interscale epidermis, in a single animal, 18 days clone-size predictions, impacting on the inferred parameter values after DOX withdrawal (Fig. 5b). Their scarcity however suggests (Supplementary Fig. 2). This is due to the probability of a chain of that they are unlikely to make a substantial contribution to tissue consecutive division events deviating from the average rate, for maintenance. example a run of several consecutive divisions shorter than Next, we performed a quantitative analysis of the time series of average cell-cycle times (Supplementary Fig. 2a). This results in a the individual-cell H2BGFP intensity histograms (Supplementary broadening of the clone-size distribution at early time points after Data 2). If there were multiple subpopulations of cells proliferat- labeling. At later times, where many rounds of division have ing at different rates the distribution of H2BGFP intensities occurred in each clone, these random cycle time variations regress would progressively diverge, becoming wider over time. We toward the mean cycle time of the population (Supplementary found no evidence of such behavior in the esophagus and at Fig. 2a). This disputes most analyses that use a Markovian multiple sites in the epidermis (Fig. 3c; Supplementary Fig. 4A, B, implementation which makes the biologically implausible assump- D, F). Specifically, several statistical tests of the modality of the tion that cell cycle times are distributed exponentially (i.e., the distribution were applied, showing no evidence for multiple likeliest time for a cell to divide is immediately after the division 4 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a M2rtTA b Rosa26 TetO-HGFP DAPI CD45 H2B-GFP R26 rtTA tetO H2B-GFP ≥ 5 fields of view per tissue/mouse +Dox Basal Collection * +Dox layer –4 weeks 0 days 7 days 12 days 18 days ROIs ROIs Exclude Exclude H2B-GFP mitotic nuclei (*) CD45+ (�) quantification Unimodality test c d 0 days 7 days 12 days 18 days Animal no. 123 4 Oesophagus n.s. n.s. n n.s. .s. Paw n.s. n.s. n n.s. .s. n n.s. .s. Ear n.s. n.s. n n.s. .s. n n.s. .s. Back n.s. n.s. n n.s. .s. n n.s. .s. n n.s. .s. 100 ± 2.1% 19.0 ± 0.9% 3.4 ± 0.2% 0.8 ± 0.1% 0.8 Tail scale § § 0.01 0.01 n n.s. .s. 0.6 0.4 Tail interscale n.s. n.s. n n.s. .s. 0.2 Unimodal Multimodal –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 (p > 0.05) (p < 0.05) (Fig. S3C) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 EXP t fit GAM t fit cc cc H : EXP t 0 cc Time Time t end H : GAM t t 1 cc Time Time t end f g –2 –4 –6 –8 EXP t cc –10 GAM t cc 02468 10 –12 Cell-cycle period, t (days) cc 0 4 8 12 16 20 Time (days) that generated it). We therefore developed a robust quantitative Clonal dynamics in esophageal epithelium. In order to explore approach where cell-cycle attributes estimated from H2B-GFP in vivo clonal dynamics, we began by studying a lineage tracing experiments were embodied in (non-Markovian) model simula- data set from mouse esophageal epithelium (Fig. 7a). In this tions, and a subsequent maximum likelihood estimation (MLE) experiment we used a strain (Lrig1-cre) in which a tamoxifen- method was applied across the available data sets for each body regulated form of Cre recombinase and enhanced green fluor- site to challenge whether each of them was consistent with the SP escent protein (EGFP) are targeted to one allele of the Lrig1 8,23–25 paradigm (Fig. 6; Supplementary Methods). locus . We found LRIG1 protein was ubiquitously expressed NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 5 Rel. frequency log (H2BGFP intensity) Simulations Non-Markovian Markovian Prob. division Prob. division Frequency ^ ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 M2rtTA Fig. 3 Analysis of cell proliferation in epidermis and esophageal epithelium. a Protocol: R26 /TetO-H2BGFP mice were treated with doxycycline (Dox) to induce H2BGFP expression (green). Following Dox withdrawal, H2BGFP transcription ceases and protein levels dilute with cell division. b H2BGFP fluorescence was quantified in non-mitotic basal cell nuclei in optical sections of the basal layer of wholemounts. Scale bar, 20 μm. c Representative confocal z stacks of the esophageal basal layer showing H2BGFP (green) and pan-leukocyte marker CD45 (red). Images are representative of a total of 15 fields of view from 3 individual biologically independent mice at 0, 7, and 12 days and 10 fields of view from two individual mice at 18 days. Infrequent label- retaining cells (LRCs) (arrowhead) are positive for CD45 (insert; blue: DAPI). Scale bars, 20 μm. Histograms show keratinocyte H2BGFP intensity for each time point (in green; bottom panels) with mean ± s.e.m. values from each field of view. Best fits for the SP model with exponential- (gray) or gamma- distributed cell-cycle periods (orange lines) are shown. Raw values for H2B-GFP intensity are given in Supplementary Data 2. d Outcome of Silverman’s unimodality test applied to individual-cell H2BGFP distributions at 18 days in esophagus and epidermis (analyses are separated per animal; this testis effectively two tailed, no multiple-testing corrections are made on p values, exact p values are given in Supplementary Data 3) . A single tissue in a single animal was found to be bimodal, in this case due to variability between fields of view where cells differed by a single round of division (see Fig. 5c). e SP- model simulations using different distributions of cell-cycle times t (EXP: exponential; GAM: Gamma) with the same average division rate (blue vertical cc line), the Gamma-shaped distributions predict a more homogeneous dilution. f Time course in the H2BGFP intensity distributions from esophageal epithelium (normalized to average keratinocyte intensity at time 0). Green boxplots: experimental data from n = 3 biologically independent mice at 0, 7, and 12 days and 2 at 18 days. Centre line of box is median value, box indicates 25th and 75th centiles and whiskers indicate minimum and maximum values. The computed average division rate ⟨λ⟩ = 2.9/week; solid black line. Gray region: range of H2BGFP intensities predicted from models assuming exponentially distributed cell-cycle times (interquartile range, inner dashed black lines). Light orange region: range of H2GFP intensities inferred with a gamma cell-cycle time distribution (interquartile range in dark orange, delimited by inner dashed orange lines). g Most-likely (gamma) shapes for the distribution of the cell-cycle period of esophageal keratinocytes, estimated from fits to the H2BGFP-dilution data. A conservative solution (in dark orange) is used for further inference. Vertical blue line: mean cell-cycle period. in the basal layer of esophageal epithelium in wild type mice hallmarks of neutral competition, in which clonal dynamics result (Supplementary Fig. 3A). Consistent with this finding, in Lrig1- from stochastic cell fates, with an average cell division generating cre animals, EGFP, reporting Lrig1 transcription was detected in one proliferating and one differentiating daughter cell, a scenario 3,29,30 94 ± 0.3% (s.e.m.) of basal cells (Supplementary Fig. 3B). These consistent with the SP model (Supplementary Methods) . observations indicate Lrig1 is widely expressed in the proliferative The measurement of the average cell division rate (λ) and compartment of esophageal epithelium and is suitable for lineage inference of the cell-cycle time distribution constrain the fitting of tracing of proliferating esophageal keratinocytes (Supplementary lineage tracing data, providing a stringent test of the candidate SP Methods). model (Supplementary Figs. 1A–C, 2E). Within this paradigm To track the fate of basal cells, Lrig1-cre mice were crossed with unknown parameters are the probability of a progenitor cell flConfetti/wt the Rosa26 (Confetti) reporter strain which labels cells division generating two dividing (PP), or two differentiating (DD) with one of four possible fluorescent proteins (green, GFP, cyan, daughters (r), and the stratification rate (Γ), which in homeostasis CFP, yellow, YFP, or red RFP) after recombination (Supplemen- sets the fraction of progenitor cells in the basal layer (ρ) (Fig. 6). 26,27 tary Fig. 3c, d) . In some Cre inducible mouse lines, reporter Our technique for identifying the most appropriate cell-cycle expressing clones can appear without induction with Tamoxifen. distribution coupled with an MLE grid search estimated However, no fluorescent protein expression was found in adult parameter values that gave an excellent fit with the clone-size uninduced Lrig1-cre/Confetti mice (Supplementary Data 5) . distributions at both early and late time points for the Lrig1/ Next, cohorts of Lrig1-cre/Confetti animals were treated with a confetti data set (Fig. 7d–e; Table 1). The model predictions were low dose of Tamoxifen that resulted in labeling of only 1 in 300 ± within the 95% confidence interval of the measured proportion of 106 (mean ± s.e.m.) basal cells at 10 days post induction. Clones clones of a given size at each time point in 27/28 cases. To containing one or more basal cells were imaged in esophageal quantify the quality of the fit, we calculated both the determina- epithelial wholemounts from at least three mice at multiple time tion coefficient between the model prediction and measured clone points over 6 months following induction (Fig. 7b; Supplemen- sizes, averaged across all time points, R , and the standard error tary Fig. 4a; Supplementary Data 5). Only CFP, YFP, and RFP of the fit, S , a measure of the standard deviation between the expressing clones were counted because of Lrig1-driven GFP model estimates and the experimental data, averaged over all time expression in all basal cells. points. For the fit of the SP model to the Lrig1/Confetti data set, 2 2 The pooled Confetti clone data set displayed several important R = 0.93, S = 4.3. Values of R and S for experimental data at T T features, which were recapitulated by clones labeled with each each time point are given in Supplementary Data 4. individual reporter. No statistically significant differences were Next, we applied the same approach to an independent, observed between CFP, YFP, and RFP clone-size distributions at published lineage tracing data set from esophageal epithelium each time point (see Methods). The density (clones/area) of where clones were labeled with YFP by Cre expressed from an labeled clones decreased progressively, consistent with clone loss inducible Cyp1a1 (Ah) promoter in AhYFP mice . Parameter through differentiation, while the number of basal and suprabasal values very similar to those from the Lrig1/confetti experiment cells in the remaining clones rose (Fig. 7c; Supplementary Fig. 4B, gave predictions from the SP model within the 95% CI for all 49 C). The proportion of labeled basal cells remained constant points in the experimental data set (Fig. 7d, e; Supplementary during the experiment, indicating the labeled population was self- Fig. 4e; Table 1; Supplementary Data 4; Supplementary Methods). maintaining over a 6-month period, consistent with labeled cells Quantifying the quality of fit, we found R = 0.98, S = 2.8. We T T being a representative sample of all proliferating cells in the noted that including the cell-cycle time constraints resulted in an homeostatic tissue (Fig. 7c). At late time points, the clone-size improved agreement with early time point clone sizes compared distribution scaled with time. This means that if, for example, with the original publication (R = 0.97, S = 3.3), where cell- T T time doubles, not only the average clone-size shape and breadth cycle time distributions were assumed exponential (see Supple- of the clone-size distribution also double. More formally, the mentary Data 4 for detailed goodness-of-fit statistics) . probability of seeing clones larger than x times the average clone As a further validation, we tested the predictions of the SP model PR1 mT/mG size became time-invariant, following a simple exponential f(x) = against a third, more limited data set from Krt15-cre R26 −x 29 e (Supplementary Fig. 4D) . Collectively, these features are mice in which a red-to-green fluorescent reporter was used with 6 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications ^ NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a H : one population H : two populations 0 1 (SP model) (SC-CP or 2xSC) Rapidly-dividing (progenitors) Time Slow-cycling (stem cells) –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 SC-CP model SC-CP model 2xSC model (Sada et al. 2016) (Mascré et al. 2012) (Sánchez-Danés et al. 2016) –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 b c 0 days 7 days 12 days 18 days ac. pl. 02468 10 Cell-cycle period, t (days) cc 100 ± 4.1% 34.4 ± 3.4% 19.9 ± 2.0% 2.2 ± 0.4% 0.8 EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 d e 0 days 7 days 12 days 18 days 02468 10 12 Cell-cycle period, t (days) cc 100 ± 3.6% 38.8 ± 3.1% 26.5 ± 3.8% 8.4 ± 0.9% 0.8 EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 f g 0 days 5 days 14 days 02468 10 12 Cell-cycle period, t (days) cc 0 days 5 days 11 days 14 days 21 days 0.8 100 ± 4.6% 56.9 ± 5.2% 37.1 ± 3.6% 19.9 ± 1.7% 9.3 ± 0.7% EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 2 inducible Cre expressed from a Krt15 promoter (Supplementary to three independent lineage tracing data sets using different Methods) . Although the SP paradigm was criticized by these combinations of transgenic Cre and reporter alleles strongly authors, it yielded an adequate fit(R = 0.91, S = 2.6) with their supports the conclusion that the esophageal epithelium is T T own data over the experimental time course (Supplementary Fig. 4f; maintained by a single-progenitor population and argues for the Supplementary Data 4). The consistent agreement of the SP model reliability of our parameter estimates. NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 7 Frequency Rel. frequency Rel. frequency Rel. frequency Frequency Frequency Frequency ^ ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 Fig. 4 H2BGFP-dilution and cell-cycle inference in skin epidermis at different sites. a Theoretical distributions of individual-cell H2BGFP intensities expected after 3 weeks dilution under the SP, SC-CP and 2xSC scenarios assuming gamma-distributed cell-cycle periods. Simulations considered an average division rate for stem cells 4× slower than for progenitors in SC-CP and 2xSC (top panels). Predictions of each model using published parameters are shown below. All theoretical SC-CP or 2xSC scenarios represented in this figure were significant by 6 different unimodality tests (see Supplementary M2rtTA Data 3). b, d, f Representative confocal z stacks of hindpaw (plantar), ear and dorsal epidermal basal layer, respectively, from R26 /TetO-H2BGFP mice, showing H2BGFP (green) and immunostaining for pan-leukocyte marker CD45 (red). Analysis of hindpaw epidermis was confined to the posterior, plantar region (pl.), excluding the acrosyringia (ac.), cartoon. Label-retaining cells (LRCs) are CD45+ leukocytes (arrowheads). Scale bars, 20 μm. Images shown are representative of a total of 18 fields of view from 3 mice at 0 and 12 days, 14 fields of view from 3 mice at 7 and 18 days in (b), 18 fields of view from 3 mice at 0, 7 and 12 days and 12 fields of view from 2 mice at 18 days in (d), and 20 fields of view from 4 mice at 0 and 11 days and 17 fields of view from 4 mice in (f). Lower panels: Individual-keratinocyte H2BGFP intensity levels (in green) with mean ± s.e.m. values from different fields of view. Best fits for the SP model with exponential- (gray) or gamma-distributed cell-cycle periods (orange lines) are shown. See Supplementary Data 2 for raw intensity values and summary statistics. c, e, g Best estimates for the (gamma) distribution of the keratinocyte cell-cycle times in hindpaw, ear, and dorsal epidermis, respectively, estimated from fits to H2BGFP-dilution data. Conservative solutions (in dark orange) are used for further inference. Vertical blue lines: average cell-cycle period per site: ⟨λ⟩ = 2.0, 1.5, 1.2/week for paw (c), ear (e), and dorsum (g), respectively. 4,21 Clonal dynamics in skin epidermis. We next investigated clonal scale regions (Fig. 5a) . These claims were primarily supported dynamics in the epidermis through available lineage tracing data by the observation of LRCs in the interscale region in H2BGFP- tTA sets from the typical interfollicular epidermis of the mouse dilution experiments in Krt5 /pTRE-H2BGFP mice. However, hindpaw (plantar), ear and back (dorsum). Applying the MLE our quantitative reanalysis of this data set showed the SP-model approach constrained by the cell-cycle time analysis at each body fits the reported H2BGFP intensity histograms over time as well 2 2 site yielded slightly improved fits of the SP model to data from as the SC-CP model (SP R = 0.89, S = 0.04 vs. SC-CP R = T T T ERT Rainbow 2 Axin2-cre R26 animals in paw epidermis (R = 0.98, 0.89, S = 0.04) (Supplementary Fig. 6A; Supplementary Data 4) T T S = 2.8) and also ear epidermis (R = 0.97, S = 3.5) and dorsal T T T . Even though we cannot discard the possibility of a interfollicular epidermis (R = 0.94, S = 5.0) from AhYFP mice T T subpopulation of slow-cycling stem cells in the tail, such cells compared with those fits reported in the original publications would seem to be rare in interscale epidermis (Supplementary 2 2 2 (R = 0.93, S = 5.16; R = 0.97, S = 3.52; R = 0.92, S = T T T T T T Data 1). We noted that a large proportion of the rare LRCs were 6.01; respectively) (Fig. 8; Supplementary Methods, Supplemen- identified as CD45 expressing leukocytes in our data set (Fig. 5b; 11,22,32 tary Data 4) . Despite differences in average keratinocyte Supplementary Data 1). Further analysis argued that there was no division rates across territories (λ ≈ 2.0, 1.5, 1.2/week for plantar conflict between the reported tail lineage tracing data and the SP hindpaw, ear and dorsum, respectively), all analyzed regions model (Supplementary Fig. 6B-H; Supplementary Methods; share comparable intermediate proportions of progenitor basal Supplementary Data 4). cells ρ (~55%, the rest corresponding to differentiating basal cells) and a predominance of asymmetric cell divisions (i.e., low inferred values for the probability of symmetric division, r < 0.25) Discussion (Table 1; Supplementary Data 4). Overall, we find that combining cell-cycle distribution analysis Particularly relevant are the implications for the mode of with lineage tracing argues mouse esophageal epithelium and keratinocyte renewal in back skin, as a previous work claims that epidermis are maintained by a single population of progenitor two stem cell populations dividing at different rates coexist at cells, with the sole possible exception of the interscale compart- this site (2xSC model) . This argument was supported by a ment of tail skin. The quality of the fit of the SP model to the data tTA quantitative analysis of H2BGFP-dilution patterns in Krt5 / is equivalent to or exceeds that of more complex models, ren- pTRE-H2BGFP mice, a system that differs from that we use dering the need to invoke additional cell populations redundant. above in that mice are treated with Dox to suppress H2B-GFP The nine lineage tracing data sets analyzed include a variety of expression instead of using it as activator (Supplementary Cre and reporter strain combinations, and are all consistent with Fig. 5A) . However, in that publication, which rejected the SP the SP model. In addition, live imaging studies of the epidermis model, exponential distributions for cell division/cell stratifica- are consistent with a single proliferating cell population main- tion rates were assumed. Here we have shown this to be taining the tissue . inappropriate for the short time scale of the experiment Quantitative analysis of cell proliferation in the different tissue (Supplementary Methods). Our computational reanalysis, con- types identifies further constraints that must be considered by strained by cell-cycle time distributions demonstrated the SP researchers exploring the appropriateness of alternative models. tTA model gave as good a fit to the Krt5 /pTRE-H2BGFP-dilution The original SC-CP and 2xSC models invoked 12% and 30% of data as the more complex 2xSC hypothesis (SP R = 0.85, S = T T basal-layer keratinocytes constitute slow-cycling stem cells, 2 2 4,20 0.06 vs. 2xSC R = 0.87, S = 0.05 for basal layer; SP R = 0.78, T T T respectively . Histone dilution experiments have allowed us to S = 0.07 vs. 2xSC R = 0.79, S = 0.06 for spinous layer) T T T make strong statements about the nature of any proposed second (Supplementary Fig. 5B; Supplementary Methods; Supplemen- population. For each body site, with the exception of tail inter- tary Data 4). Indeed, the inferred parameter values from the scale epidermis, no keratinocyte label-retaining cells were detec- AhYFP mouse back skin epidermis proved robust, providing ted in over 2000 cells imaged at each location. It follows that any good fits to another lineage tracing data set from the same body slow-cycling stem-cell population must have substantially fewer ERT flConfetti 2 site in Lgr6-eGFPcre Rosa26 mice (R = 0.96, S = T T than one slow-cycling cell per thousand basal keratinocytes to be 2.39) (Supplementary Fig. 5C; Supplementary Methods; Supple- compatible with observations reported here, making it unlikely mentary Data 4) . that such slow-cycling cells will make a detectable contribution to Finally, we turned to revisit clonal dynamics in the mouse tail tissue homeostasis. The hypothesis that two subpopulations exist, epidermis. Previous studies of tail have argued that the but that they both divide at a similar rate, is hard to sustain in the hierarchical SC-CP paradigm applies to proliferating cells in the face of the close agreement of the simpler SP model across all the interscale areas while the SP paradigm describes behavior in the analyzed data sets. 8 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications ^ NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a M2rtTA b Rosa26 0 days 7 days 12 days 18 days TetO-HGFP Scale Interscale Hair follicle openings Unimodality test –2 –4 –2 –6 –8 –4 1 2 34567 –10 Field –2 –4 –6 –8 –10 048 12 16 20 Time (days) Fig. 5 H2BGFP-dilution analysis in tail epidermis. a Structure of the mouse tail epidermis. 3D reconstruction of confocal z stacks (top panel) and orthogonal xyz view (mid panel). Apical blister-shaped regions (scale) alternate with deeper regions (interscale) (boundaries delineated by dotted lines), being arranged in lines separated by hair follicles (asterisks). Green: H2BGFP expression; white: KRT14 immunostaining (as a marker of basal layer); red: CD45 immunostaining; blue: DAPI. Scale bars, 200 μm. Bottom panel: Cartoon illustrating the tail skin structure. b Representative confocal z stacks of scale M2rtTA and interscale regions of tail epidermis during H2BGFP chase experiments in R26 /TetO-H2BGFP mice, showing H2BGFP (green), immunostaining for KRT14 (white) and immunostaining for pan-leukocyte marker CD45 (red). Images from the same time point correspond to different mice to illustrate inter- animal variation. Label-retaining cells (LRCs) are highlighted with arrowheads (CD45+ cells) or full arrows (CD45− cells) (details in inserts; blue: DAPI). Scale bars, 20 μm. c Time course of basal-layer keratinocyte H2BGFP intensity distributions from scale and interscale regions of tail. Experimental data shown as boxplots per individual biologically independent mouse (intensities normalized to average keratinocyte intensity at time 0, raw H2B-GFP intensity values are given in Supplementary Data 2). n = 3 animals at each time point except 18 days where n = 2 mice. Centre line of box is median value, box indicates 25th and 75th centiles and whiskers indicate minimum and maximum values. Solid black lines: average H2BGFP-dilution rates (within 95% CI limits—shaded gray areas—where the value of λ = 1.2/week reported by Mascre et al. falls). Insert: Detail of H2BGFP intensity distributions separated per field of view for the single tissue found to be bimodal in the overall, per-animal modality test (see Fig. 3d). Analyses per field of view all resulted non- significant (unimodality). NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 9 n.s. n.s. n.s. n.s. n.s. Overall n.s. n.s. Interscale Scale ** * * * log (H2BGFP int.) log (H2BGFP int.) 2 2 * * * * * * Interscale Scale ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 Table 1 Parameter values inferred for progenitor cell behavior in different murine epithelial regions as derived from quantitative lineage tracing. Tissue Experimental model Reference min Division rate, Symmetric % of progenitor Stratification rate, tcc (days) λ (/week) division prob., r cells, ρ Γ (/week) ERT Esophagus Lrig1-eGFP-Cre / 0.5 2.9 (2.7; 3.0) 0.10 (0.07; 0.15) 65 (50; 96) 5.4 (2.9; 69.6) flConfetti R26- ERT flEYFP Ah-Cre /R26 5 0.5 2.9 (2.7; 3.0) 0.06 (0.04; 0.10) 56 (50; 89) 3.7 (2.9; 23.5) ERT2 Paw Epidermis Axin2-Cre /R26- 11 1 2.0 (1.7; 2.3) 0.14 (0.12; 0.17) 53 (49; 58) 2.3 (1.9; 2.8) Rainbow ERT flEYFP Ear Epidermis Ah-Cre /R26 32 1 1.5 (1.2; 1.7) 0.04 (0.03; 0.06) 54 (47; 72) 1.8 (1.3; 3.9) ERT flEYFP Back Epidermis Ah-Cre /R26 22 2 1.2 (1.1; 1.3) 0.04 (0.03; 0.07) 61 (55; 76) 1.9 (1.5; 3.8) Parameter values indicated correspond with the maximum likelihood estimate (MLE), values in parentheses are 95% confidence bounds (see Supplementary Methods for details). H2BGFP dilution Lineage tracing Average cell cycle time, Clone size distributions at cell cycle time distribution multiple time points Inferred t cc Time Compute likeliest values of: l 1 – 2r {r, r, G } to best fit clone size distributions Data 3–4 5–8 Model 9–16 17–32 33–64 40 >64 1 10 100 Time (weeks) Compare model with experimental data Fig. 6 Method for single-progenitor model testing and parameter inference. Method to single-progenitor model testing and infer model parameters. Orange boxes indicate experiments and resulting data, gray box computational model and parameter estimation. Italics indicate parameters in the SP model. The multimodality testing of H2B-GFP data showed that there is a single population dividing at the same average rate in epidermis and esophagus, consistent with the SP model (Fig. 3d). To test the SP model, the average cell-cycle time (λ) and cell-cycle time distribution were inferred from H2B-GFP experiments. These values are used in computational analysis to estimate the values of the other parameters in the SP model, the proportion of progenitor cells in the basal layer ρ, the proportion of symmetric cell division outcomes r, and the stratification rate of differentiating cells leaving the basal cell layer (Γ). Multiple sets of values for the unknown parameters were tested. For each set of unknown parameter values 100,000 progenitor-derived clones were simulated (lines) and inferred clone-size distributions compared with experimental ones (points) obtained from lineage tracing. The likeliest sets of parameter values were obtained by maximum likelihood estimation for each linage tracing data set. The quality of the fit was assessed by determining whether the simulated values lie within the 95% confidence interval of the experimental clone-size measurements at each time point. The improved resolution of parameter estimates identifies cells cycle comparatively slower in the epidermis, on average differences in cell division rates across the epidermis and the between 3.5 and 6 days depending on body site . However, our esophagus (Table 1). Proliferating cells divide rapidly in the study suggests individual cell-cycle periods are tightly con- esophageal epithelium, on average every ~2.4 days (similar to trolled, showing little variation around average division rate, keratinocyte turnover rate in oral mucosa), while progenitor per territory. 10 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications Computation Experiments Prob. division Clone frequency (%) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE Lrig1-eGFPcreERT/wt R26Confetti/wt ERT Lrig1 EGFP x R26 STOP Confetti cre Label Collection 0 10 days 30 days 84 days 180 days 10 days 30 days 84 days 180 days 15 0.3 0.2 5 0.1 0 0 0 0 50 100 150 200 0 50 100 150 200 10 days 30 days 84 days 180 days Time (days) Time (days) Time (days) d e EXP t GAM t cc cc 1.0 80 3–4 0.8 8 5–8 9–16 0.6 17–32 33–64 2 40 0.4 >64 0.2 0.5 N = 1222 0 0 1.0 100 0.8 80 3–4 5–8 9–16 0.6 17–32 33–64 0.4 2 40 >64 0.2 0.5 N = 1925 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob., r Time (weeks) The proportion of progenitor cells in the basal layer and the our analysis. Instead, the consistent values of r < 0.25 indicate the probability of symmetric cell division outcomes (r) are similar fate of sister cells is preferentially anti-correlated. This phenom- across body sites (Table 1). The insight that a substantial pro- enon can be associated to local coordination of neighboring cell portion of cells in the basal layer will proceed to differentiate stratification and division events . Our results argue that antic- rather than divide will be important for the interpretation for the orrelation of sister cell fates applies generally in the epidermis and growing body of single-cell RNA sequencing data in these tis- esophagus, pointing to common mechanisms of keratinocyte cell 33,35 sues . In addition, the low values of r we identify give insight fate regulation. into the basis of cell fate determination (Fig. 9). In principle, if The single-progenitor model captures the average behavior of every basal cell divides or differentiates with equal probability, as progenitor cells during homeostasis. However, epithelia are fre- proposed by Leblond, r will be 0.25, as expected from any pair of quently subject to wounding. To repair the tissue requires a uncorrelated basal cells . However, this scenario is excluded by temporary imbalance in cell fate, with the progenitors close to the NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 11 ERT EYFP ERT Confetti Average basal clone size Ah-cre /R26 Lrig1-cre /R26 Proportion of P-cells, ρ Proportion of P-cells, ρ Number of clones/mm Stratif. rate, Γ (/week) Stratif. rate, Γ (/week) Clone frequency (%) Clone frequency (%) % labelled basal cells ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ERT/wt flConfetti/wt Fig. 7 Quantitative lineage tracing in esophageal epithelium. a Protocol: clonal labeling was induced in Lrig1-eGFPcre R26 mice and samples analyzed at different times from 10 to 180 days post induction, as single labeled cells develop into clones. See Supplementary Data 5 for source data for panels (c) and (e). b Rendered confocal z stacks of the esophageal basal layer showing typical RFP clones (red) at the times indicated. Blue is DAPI. Scale bars, 10 μm. Images are representative of 104 RFP clones (10 days), 75 RFP clones (30 days), 106 RFP clones (84 days), and 274 RFP clones (180 days). c Quantitative characteristics of the labeled clone population over time: average basal-layer clone size (i.e., mean number of basal cells/ surviving clone) (left panel), average density of labeled clones in the basal layer (middle panel), average fraction of labeled basal cells at the indicated time points (right panel). Observed values are shown in individual biologically independent mice (blue circles, n mice = 3 at 10 and 30 days, 6 at 84 days, and 4 at 180 days) with error bars (black) indicating mean ± s.e.m. of all mice at each time point. A total of 300 or more clones was quantified at each time point. Orange lines: SP-model fit (shaded area corresponds with 95% plausible intervals). Orange line and shading in last panel show mean and s.e.m. across all ERT time points (from n= 16 mice), which is consistent with homeostatic behavior. d SP-model parameter inference on Lrig1- and Ah-Cre driven lineage tracing data sets from esophagus . Parameter estimates are affected by the underlying modeling assumptions on the cell-cycle period, whether default exponential cell-cycle time distributions were considered (solutions in gray) or realistic gamma distributions implemented, as inferred from the cell- proliferation analysis (solutions in orange). Regions within the dashed gray lines fall consistent with the predicted ρ/r ratios from the linear scaling of the average clone size. The total number of clones counted in each data set is displayed in the corresponding graph and previous parameter estimates given in 5 ERT 5 ref. shown as black error bars. e Experimental Lrig1- and Ah-Cre -derived basal-layer clone sizes from ref. (dots with error bars indicating the standard error of a proportion) fit well with the SP model, with gamma-distributed cell-cycle times (lines; prediction from maximum likelihood estimation). Dim dashed lines: fits from ref. . Frequencies for each clone size (basal cell number) are shown in different colors. n = 3 biologically independent mice at each time point except for the Lrig1/confetti where n = 6 mice at 12 weeks and 4 mice at 26 weeks. wound producing an excess of progenitor over differentiating Methods Animals. All experiments were conducted according to the UK Home Office daughters on average. This occurs as part of a coordinated set of Project Licenses 70/7543, P14FED054 or PF4639B40. Male and female adult mice responses that includes cell migration and altered cell differ- aged 3–18 months were used for in vivo experiments. Animals were housed in 5,37,38 entiation . Once the epithelial defect is resolved, the pro- individually ventilated cages and fed on standard chow and maintained in SOPF genitors revert to homeostatic balance. In esophageal epithelium health status. ERT/wt flConfetti/wt Doubly transgenic, Lrig1-eGFPcre R26 mice on a C57/Bl6N and the plantar epidermis, wound repair is achieved by pro- background were generated for lineage tracing studies in esophageal epithelium, by 5,11 genitors alone . In the epidermis at other sites, cells migrating ERT2 8 flConfetti crossing Lrig1-eGFP-ires-cre mice onto a Rosa26 multicolor reporter from other proliferative compartments, the hair follicles and line . Transcription of the Cre recombinase-mutant estrogen receptor fusion 9,10,39,40 ERT sweat ducts, may also contribute to wound healing . The protein (Cre ) is under the control of an endogenous allele of Lrig1. Following ERT induction with tamoxifen, Cre protein internalizes into the nuclei and excises a ability to transiently increase the likelihood of progenitors gen- LoxP-flanked “STOP” cassette resulting in the expression of one of the four erating proliferating progeny provides a rapid and robust M2rtTA Confetti fluorescent reporters (YFP, RFP, CFP, or GFP). R26 /TetO-H2BGFP response to injury. The down side of this adjustable progenitor mice, doubly transgenic for a reverse tetracycline-controlled transactivator (rtTA- fate is that it may be subverted by mutations acquired during M2) targeted to the Rosa26 locus and a HIST1H2BJ/EGFP fusion protein tissue aging, leading to mutant clonal expansions that may (H2BGFP) expressed from a tetracycline promoter element, were used for label- 5,22 22,41–43 retaining experiments . H2BGFP expression is induced by treatment with undergo malignant transformation . doxycycline (Dox) and dilution of H2BGFP protein content can be chased upon How might these findings in mice relate to homeostasis human Dox withdrawal. All animals were induced at 8–12 weeks age. Cohorts of at least epidermis? Human skin differs from that of mice with many two or three animals per time point were culled and esophagus and/or skin more epidermal cell layers and undulates in thickness at most epidermis collected for analysis. body sites creating folds called rete ridges and dermal papillae . Nevertheless, a population of cells with balanced stochastic cell Wholemount preparation and immunostaining. Esophageal epithelium whole- fate generating equal proportions of proliferating and differ- mounts for lineage tracing were prepared as follows: The esophagus was cut longitudinally and the middle two-thirds of the tract was incubated for 3 h in 5 mM entiating cells has been identified in a live imaging study of EDTA in PBS at 37 °C. The epithelium was then peeled away from the underlying human keratinocytes in primary culture . In vivo lineage tracing submucosa, stretched and fixed for 30 min in 4% paraformaldehyde in PBS. in humans is not feasible. However, human epidermis has been Samples were stored in PBS at 4 °C until subsequent analysis. Skin pieces of ~0.5 grafted onto immune compromised mice and injected with len- cm were cut and incubated for 1 h in 5 mM EDTA in PBS at 37 °C. Skin epidermis was then peeled away using fine forceps and processed as described above for the tiviral vectors carrying fluorescent protein reporters. When the esophageal epithelium. resulting clones were imaged 6 months later they were found to For staining, wholemount samples were incubated in Permeabilization Buffer vary widely in size and shape and arise from any point in the (PB) (0.5% BSA, 0.25% Fish Skin Gelatin (FSG), 0.5% Triton X-100/PBS) for 15 basal layer, both in rete ridges and dermal papillae . These min at room temperature (RT), then blocked in 10% goat or donkey serum/PB (according to the secondary antibody used) for 1 h at RT and incubated overnight findings are consistent with the single-progenitor paradigm, but with primary antibody at 4 °C. Primary antibodies used were Lrig1 antibody (R&D cannot provide quantitative challenge to the model available Systems, Cat. AF3688), ITGA6 antibody (clone GoH3, Biolegend, Cat. B204094), in mice. Alexa Fluor® 647 anti-CD45 (clone 30-F11, Biolegend, Cat. 103124), Keratin 14 The lineage tracing approaches considered above have been antibody (clone Poly19053, Biolegend, Cat. 905301). Samples were subsequently 34,36 enriched by live imaging studies of mouse epidermis . Whilst washed four times for 30 min in 0.2% Tween-20/PBS and incubated with an appropriate secondary antibody for 3 h at RT. Secondary antibodies used were lineage tracing resolves the average behavior of a population of Goat or Donkey Alexa Fluor 488/546/555/647 (Molecular Probes). A washing step proliferating cells over many cell generations, live imaging allows with 0.2% Tween-20/PBS was repeated and samples were incubated for 30 min the fate of individual cells to be resolved. Insights gained from live with DAPI (Sigma-Aldrich) and finally mounted in DAKO Vectashield Mounting imaging include showing that cell fate is stochastic, the prob- Medium with DAPI (Vector Labs). ability of generating progenitor and differentiated daughters is M2rtTA equal and that the fate of cells is not coordinated across cell Dilution of Histone 2B-GFP protein content. R26 /TetO-H2BGFP animals generations, all of which are key features of the SP model . were treated with doxycycline (Dox, 2 mg/ml in drinking water sweetened with 10% blackcurrant & apple) for 4 weeks. Dox was then withdrawn and animals We conclude that the single-progenitor model is consistent with culled at different time points to track H2BGFP florescence dilution. Epithelial a large body of lineage tracing and cell-cycle data collated from wholemounts from esophageal epithelium and skin epidermis were imaged on a multiple studies and identifies the behavior of proliferating cells Leica TCS SP8 confocal microscope using ×20 or ×40 objectives at 1024 × 1024 that underpins epidermal and esophageal epithelial homeostasis. resolution, line average 4 and 400 Hz scan speed. Individual-cell H2BGFP 12 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE ERT2 Axin2 cre x R26 STOP Rainbow (Lim et al. 2013) 0 –4 –8 log-Like. ratio statistic 1.0 100 3–4 0.8 80 5–8 9–16 0.6 60 17–32 2 33–64 >64 0.4 40 0.2 0.5 20 N = 9643 0 0 0 0.1 0.2 0.3 0.4 0.5 Symmetric division prob, r Time (weeks) ERT Cyp1A1 R26 STOP YFP (Doupé et al. 2010) cre x 0 –4 –8 log-Like. ratio statistic 1.0 8 3–4 0.8 4 5–8 9–16 0.6 2 60 17–32 >32 0.4 1 0.5 0.2 N = 1233 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob, r Time (weeks) ERT Cyp1A1 R26 STOP YFP (Murai et al. 2018) cre x 0 –4 –8 log-Like. ratio statistic 1.0 100 3–4 0.8 80 5–8 9–16 0.6 60 17–32 >32 0.4 40 0.5 0.2 N = 1682 0 0 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob, r Time (weeks) Fig. 8 The single-progenitor model fits clone dynamics in different regions of skin epidermis. a–c Left panels: SP-model parameter inference on lineage 11 ERT Rainbow 32 22 tracing data sets from paw epidermis using Axin2-cre R26 animals (a), and ear and dorsal interfollicular epidermis in AhYFP mice (b and c, respectively). Parameter estimates are obtained by MLE based on SP-model simulations constrained by the cell-cycle period distribution inferred from each corresponding cell-proliferation analysis. Regions within the dashed gray lines fall consistent with the predicted ρ/r ratios from the linear scaling of the average clone size. The total number of clones counted in each data set is displayed in the corresponding graph. Black bars are parameter estimates given in the original publications shown, centre is the mid range and bars indicate the maximum and minimum plausible parameter values in simulations in each 11 32 22 paper (a, b, c). Right panels: Experimental Axin2- (a; from ref. ) and Ah-(b and c; from refs. and ) derived basal-layer clone sizes (dots indicate mean ± standard error of proportion) give an excellent fit with the SP model with gamma-distributed cell-cycle times (lines; prediction from MLE). Dim dashed lines: fits obtained with parameter estimates given in the original publications. Frequencies for each clone size (basal cell number) are shown in different colors. See Supplementary Data 4 for goodness-of-fit statistics. intensities were determined by image segmentation/nuclear identification, using a of 1 in 301 ± 106 (mean ± s.e.m.) basal cells by 10 days post induction (allowing semi-automated object-recognition macro (based on the DAPI channel) built in individual clone tracking without merging). Between three and six mice were culled ImageJ, and the process completed by manual curation. Per-cell intensity values per time point. Confocal images of immunostained wholemounts were acquired on given are averaged over all nuclear pixels. All H2BGFP samples were stained for a Leica TCS SP8 confocal microscope (×10, ×20, and ×40 objectives; typical settings CD45 and positive cells excluded from the analysis. for z-stacks acquisition: optimal pinhole, line average 4, bi-directional scan with 400–600 Hz speed, resolution of 1024 × 1024 pixels). The number of nucleated basal and suprabasal cells per labeled clone was Lineage tracing. Low-frequency expression of the Confetti reporters in the Lrig1- counted under live acquisition mode. GFP-labeled clones were not scored due to ERT2 flConfetti/wt eGFP-ires-cre R26 mouse esophagus was achieved by inducing 10- the difficulty of distinguishing them from the constitutive basal GFP expression driven by the Lrig1 cassette. CFP, RFP, and YFP clones were pooled together for week-old animals with intraperitoneal injection of a single dose of 1 mg tamoxifen (100 μl of 10 mg/ml) on two consecutive days . This resulted in a labeling efficiency further analysis (histograms (distributions) of basal-layer clone sizes and average NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 13 Proportion of P-cells, ρ Proportion of P-cells, ρ Proportion of P-cells, ρ –1 –1 Stratif. rate, Γ (week ) –1 Stratif. rate, Γ (week ) Stratif. rate, Γ (week ) Clone frequency (%) Clone frequency (%) Clone frequency (%) ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 l (1 + d) 0.5 (1 – d) 0.5 Progenitor cell Differentiating cell Suprabasal cell #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 0.5 1.5 cd Sibling #1: Sibling #2: <0.25 >0.25 >0.25 <0.25 Local fate coordination Fig. 9 Cell fate coordination underpins single-progenitor dynamics in epidermis and esophagus. a Epidermis and esophageal epithelium are maintained by a single population of progenitor cells. Left panel: Progenitor cells (in green) share the basal layer with post-mitotic keratinocytes in early stages of differentiation (pale red), which are transiently retained in the basal compartment before stratification. Right panel: Simplified representation of the single- progenitor model focusing on individual basal cell fates. Epithelial cell dynamics are dominated by stochastic but skewed fates through spatial coordination between neighboring or sibling cells. Individual basal cells undertake dichotomic decisions: they have the potential to divide but can alternatively differentiate exiting the basal layer. Both probabilities are balanced (50–50%) across the entire proliferating cell population, but can be decompensated or skewed for individual cells depending on their local niche, as reflected with the parameter δ (which works as a context-dependent modulatory factor). b Stochastic progenitor fates explain a scenario of neutral clonal competition dynamics where clones develop into heterogeneous sizes, constrained by cell-cycle time control and fate coordination effects. Displayed is a representative set of epithelial clone dynamics simulated using the parameters inferred for murine esophageal epithelium homeostasis. c, d Our results demonstrate that the outcome of sibling keratinocyte cells is commonly biased toward an excess of asymmetric fates where one decides to divide while the other differentiates, in agreement with a single-progenitor model with low values of r (r < 0.25). number of basal cells). A total of 300, 315, 302, and 305 labeled clones from 3, 3, 6, Reporting summary. Further information on research design is available in and 4 mice at 10, 30, 84, and 180 days post induction, respectively, were quantified. the Nature Research Reporting Summary linked to this article. Regarding the time courses in the number of clones per unit area and the proportion of labeled basal cells, only RFP clones were considered given the low, Data availability variable induction of the other florescent reporters and their overall small The authors declare that the experimental data supporting the findings of this study are contribution (including these numbers did not alter the conclusions). available within the paper and its supplementary information files. ERT flEYFP 5 Lineage-tracing data from Ah-cre R26 derived clones in esophagus , 32 22 ear , and dorsal epidermis were obtained from experimental colleagues (data ERT Rainbow available upon request). Data on induced Axin2-cre R26 clones in Code availability 11 ERT flConfetti 33 hindpaw and Lgr6-eGFPcre R26 in back epidermis were kindly Code used in computational modeling is available in Github: https://github.com/gp10/ provided by the authors. Data from lineage tracing in scale and interscale tail Piedrafita_etal_SI_code/ epidermis were accessed through the online publication material, while authors 4 PR1 mT/mG were unable to provide original data from ref. . Data on Krt15-cre R26 mouse esophagus were retrieved by digitalizing Fig. 2e and Figure S3B from the Received: 29 July 2019; Accepted: 26 February 2020; tTA original publication. A similar procedure was used to extract Krt5 /pTRE- H2BGFP-dilution data from back skin (Fig. 3 from ref. ) and tail epidermis (Fig. 3k from ref. ). Mathematical modeling and statistical inference. Model dynamics were simu- lated using Markovian (Gillespie algorithm) and non-Markovian exact stochastic References Monte Carlo methods implemented in Matlab. A maximum likelihood estimation 1. Marques-Pereira, J. P. & Leblond, C. P. Mitosis and differentiation in the (MLE) approach was followed for parameter inference, except when stated stratified squamous epithelium of the rat esophagus. Am. J. Anat. 117,73–87 otherwise, and best-fit parameters obtained with 95% confidence intervals based on (1965). the likelihood-ratio test (alpha = 0.05). The coefficient of determination (R ) and 2. Allen, T. D. & Potten, C. S. Fine-structural identification and organization of the standard error of the fit(S) were calculated for the evaluation of goodness of fit the epidermal proliferative unit. J. Cell Sci. 15, 291–319 (1974). (GoF). In those cases involving data at various time points, GoF values averaged 3. Clayton, E. et al. A single type of progenitor cell maintains normal epidermis. across the different time points (R and S ) are displayed. A comprehensive list of T T detailed R and S values can be found in Supplementary Data 4. For details of Nature 446, 185–189 (2007). theoretical modeling and computational methods used to infer cell behavior and 4. Mascre, G. et al. Distinct contribution of stem and progenitor cells to clonal dynamics, see Supplementary Methods section below. epidermal maintenance. Nature 489, 257–262 (2012). 14 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications Time (weeks) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE 5. Doupe, D. P. et al. A single progenitor population switches behavior to 39. Levy, V., Lindon, C., Harfe, B. D. & Morgan, B. A. Distinct stem cell maintain and repair esophageal epithelium. Science 337, 1091–1093 (2012). populations regenerate the follicle and interfollicular epidermis. Dev. Cell 9, 6. Martincorena, I. et al. Somatic mutant clones colonize the human esophagus 855–861 (2005). with age. Science 362, 911–917 (2018). 40. Levy, V., Lindon, C., Zheng, Y., Harfe, B. D. & Morgan, B. A. Epidermal stem 7. Martincorena, I. et al. Tumor evolution. High burden and pervasive positive cells arise from the hair follicle after wounding. Faseb J. 21, 1358–1366 (2007). selection of somatic mutations in normal human skin. Science 348, 880–886 41. Fernandez-Antoran, D. et al. Outcompeting p53 mutant cells in the normal (2015). esophagus by redox manipulation. Cell Stem Cell 25, 329–341 (2019). 8. Page, M. E., Lombard, P., Ng, F., Gottgens, B. & Jensen, K. B. The epidermis 42. Alcolea, M. P. et al. Differentiation imbalance in single oesophageal progenitor comprises autonomous compartments maintained by distinct stem cell cells causes clonal immortalization and field change. Nat. Cell Biol. 16, populations. Cell Stem Cell 13, 471–482 (2013). 615–622 (2014). 9. Ito, M. et al. Stem cells in the hair follicle bulge contribute to wound repair but 43. Alcolea, M. P. & Jones, P. H. Cell competition: winning out by losing notch. not to homeostasis of the epidermis. Nat. Med. 11, 1351–1354 (2005). Cell Cycle 14,9–17 (2015). 10. Lu, C. P. et al. Identification of stem cell populations in sweat glands and ducts 44. Lavker, R. M. & Sun, T. T. Heterogeneity in epidermal basal keratinocytes: reveals roles in homeostasis and wound repair. Cell 150, 136–150 (2012). morphological and functional correlations. Science 215, 1239–1241 (1982). 11. Lim, X. et al. Interfollicular epidermal stem cells self-renew via autocrine Wnt 45. Ghazizadeh, S. & Taichman, L. B. Organization of stem cells and their progeny signaling. Science 342, 1226–1230 (2013). in human epidermis. J. Invest Dermatol 124, 367–372 (2005). 12. Kamberov, Y. G. et al. A genetic basis of variation in eccrine sweat gland and 46. Silverman, B. W. Using kernel density estimates to investigate multimodality. hair follicle density. Proc. Natl Acad. Sci. USA 112, 9932–9937 (2015). J. R. Stat. Soc.: Ser. B (Methodol.) 43,97–99 (1981). 13. Mackenzie, J. C. Ordered structure of the stratum corneum of mammalian skin. Nature 222, 881–882 (1969). Acknowledgements 14. Spearman, R. I. & Hardy, J. A. Ultrastructure of the contrasting types of We are grateful to Xinhong Lim and Roeland Nusse; Kyogo Kawaguchi, Allon M. Klein keratinization seen in the tail epidermis of the laboratory mouse Mus and Valentina Greco; Anja Füllgrabe and Maria Kasper; and David Shalloway for kindly musculus. Arch. Dermatol Res. 258,33–40 (1977). sharing raw lineage-tracing data and computational algorithms from previous publica- 15. Gomez, C. et al. The interfollicular epidermis of adult mouse tail comprises tions, and for their valuable comments. We thank María P. Alcolea for l feedback on two distinct cell lineages that are differentially regulated by Wnt, Edaradd, and H2BGFP experiments and Kim Jensen for the Lrig1-creERT mouse strain, Esther Lrig1. Stem Cell Rep. 1,19–27 (2013). Choolun and staff at the MRC ARES and Sanger RSF facilities for excellent technical 16. Braun, K. M. et al. Manipulation of stem cell proliferation and lineage support and Hall and Jones’ group members and Fernando Pozo, Fátima Al-Shahrour, commitment: visualisation of label-retaining cells in wholemounts of mouse and Francisco X Real at CNIO for critical comments. This work was supported by grants epidermis. Development 130, 5241–5255 (2003). from the Wellcome Trust to the Wellcome Sanger Institute, 098051 and 296194, Cancer 17. Alcolea, M. P. & Jones, P. H. Lineage analysis of epidermal stem cells. Cold Research UK Programme Grants to P.H.J. (C609/A17257, C609/A27326) and the Royal Spring Harb. Perspect. Med 4, a015206 (2014). Society (UF130039 to B.A.H.). G.P. is supported by a Talento program fellowship from 18. Alcolea, M. P. & Jones, P. H. Tracking cells in their native habitat: lineage Comunidad de Madrid. tracing in epithelial neoplasia. Nat. Rev. Cancer 13, 161–171 (2013). 19. Tumbar, T. et al. Defining the epithelial stem cell niche in skin. Science 303, 359–363 (2004). Author contributions 20. Sada, A. et al. Defining the cellular lineage hierarchy in the interfollicular G.P. performed the computational simulations and carried out the analysis of experi- epidermis of adult skin. Nat. Cell Biol. 18, 619–631 (2016). mental results. A.W. carried out the linage tracing experiments. B.C. and D.F.A. per- 21. Sanchez-Danes, A. et al. Defining the clonal dynamics leading to mouse skin formed H2BGFP-dilution experiments. A.H. and K.M. contributed with image tumour initiation. Nature 536, 298–303 (2016). acquisition and image analysis. V.K. performed preliminary, conceptual work and helped 22. Murai, K. et al. Epidermal tissue adapts to restrain progenitors carrying clonal implementing the Non-Markovian simulation algorithm. B.A.H. and P.H.J. supervised p53 mutations. Cell Stem Cell 23, 687–699 (2018). and designed the work. G.P., B.A.H., and P.H.J. wrote the paper. All authors reviewed 23. Nakamura, T. et al. LRIG1 inhibits STAT3-dependent inflammation to and edited the final version. maintain corneal homeostasis. J. Clin. Invest. 124, 385–397 (2014). 24. Lu, L. et al. LRIG1 regulates cadherin-dependent contact inhibition directing epithelial homeostasis and pre-invasive squamous cell carcinoma Competing interests development. J. Pathol. 229, 608–620 (2013). The authors declare no competing interests. 25. Wong, V. W. et al. Lrig1 controls intestinal stem-cell homeostasis by negative regulation of ErbB signalling. Nat. Cell Biol. 14, 401–408 (2012). Additional information 26. Snippert, H. J. et al. Intestinal crypt homeostasis results from neutral competition Supplementary information is available for this paper at https://doi.org/10.1038/s41467- between symmetrically dividing Lgr5 stem cells. Cell 143, 134–144 (2010). 020-15258-0. 27. Frede, J., Greulich, P., Nagy, T., Simons, B. D. & Jones, P. H. A single dividing cell population with imbalanced fate drives oesophageal tumour growth. Nat. Correspondence and requests for materials should be addressed to B.A.H. or P.H.J. Cell Biol. 18, 967–978 (2016). 28. Jensen, K. B. et al. Lrig1 expression defines a distinct multipotent stem cell Peer review information Nature Communications thanks Qing Nie and the other, population in mammalian epidermis. Cell Stem Cell 4, 427–439 (2009). anonymous, reviewer(s) for their contribution to the peer review of this work. Peer 29. Klein, A. M. & Simons, B. D. Universal patterns of stem cell fate in cycling reviewer reports are available. adult tissues. Development 138, 3103–3111 (2011). 30. Klein, A. M., Doupe, D. P., Jones, P. H. & Simons, B. D. Kinetics of cell Reprints and permission information is available at http://www.nature.com/reprints division in epidermal maintenance. Phys. Rev. E Stat. Nonlin Soft Matter Phys. 76, 021910 (2007). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 31. Giroux, V. et al. Long-lived keratin 15+ esophageal progenitor cells contribute published maps and institutional affiliations. to homeostasis and regeneration. J. Clin. Invest. 127, 2378–2391 (2017). 32. Doupe, D. P., Klein, A. M., Simons, B. D. & Jones, P. H. The ordered architecture of murine ear epidermis is maintained by progenitor cells with random fate. Dev. Cell 18, 317–323 (2010). Open Access This article is licensed under a Creative Commons 33. Fullgrabe, A. et al. Dynamics of Lgr6(+) progenitor cells in the hair follicle, Attribution 4.0 International License, which permits use, sharing, sebaceous gland, and interfollicular epidermis. Stem Cell Rep. 5, 843–855 (2015). adaptation, distribution and reproduction in any medium or format, as long as you give 34. Rompolas, P. et al. Spatiotemporal coordination of stem cell commitment appropriate credit to the original author(s) and the source, provide a link to the Creative during epidermal homeostasis. Science 352, 1471–1474 (2016). Commons license, and indicate if changes were made. The images or other third party 35. Jones, K. B. et al. Quantitative clonal analysis and single-cell transcriptomics material in this article are included in the article’s Creative Commons license, unless reveal division kinetics, hierarchy, and fate of oral epithelial progenitor cells. indicated otherwise in a credit line to the material. If material is not included in the Cell Stem Cell 24, 183–192.e188 (2019). article’s Creative Commons license and your intended use is not permitted by statutory 36. Mesa, K. R. et al. Homeostatic epidermal stem cell self-renewal is driven by regulation or exceeds the permitted use, you will need to obtain permission directly from local differentiation. Cell Stem Cell 23, 677–686.e674 (2018). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 37. Roshan, A. et al. Human keratinocytes have two interconvertible modes of licenses/by/4.0/. proliferation. Nat. Cell Biol. 18, 145–156 (2016). 38. Park, S. et al. Tissue-scale coordination of cellular behaviour promotes © The Author(s) 2020 epidermal wound repair in live mice. Nat. Cell Biol. 19, 155–163 (2017). NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 15 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Communications Springer Journals

A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial maintenance in mice

Loading next page...
 
/lp/springer-journals/a-single-progenitor-model-as-the-unifying-paradigm-of-epidermal-and-1RaZ3A2VWh

References (62)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2020
eISSN
2041-1723
DOI
10.1038/s41467-020-15258-0
Publisher site
See Article on Publisher Site

Abstract

ARTICLE https://doi.org/10.1038/s41467-020-15258-0 OPEN A single-progenitor model as the unifying paradigm of epidermal and esophageal epithelial maintenance in mice 1,2 3 1 1 1 Gabriel Piedrafita , Vasiliki Kostiou , Agnieszka Wabik , Bartomeu Colom , David Fernandez-Antoran , ✉ ✉ 1 1 3 1,3 Albert Herms , Kasumi Murai , Benjamin A. Hall & Philip H. Jones In adult skin epidermis and the epithelium lining the esophagus cells are constantly shed from the tissue surface and replaced by cell division. Tracking genetically labelled cells in trans- genic mice has given insight into cell behavior, but conflicting models appear consistent with the results. Here, we use an additional transgenic assay to follow cell division in mouse esophagus and the epidermis at multiple body sites. We find that proliferating cells divide at a similar rate, and place bounds on the distribution cell cycle times. By including these results in a common analytic approach, we show that data from eight lineage tracing experiments is consistent with tissue maintenance by a single population of proliferating cells. The outcome of a given cell division is unpredictable but, on average, the likelihood of producing pro- liferating and differentiating cells is equal, ensuring cellular homeostasis. These findings are key to understanding squamous epithelial homeostasis and carcinogenesis. 1 2 Wellcome Sanger Institute, Hinxton CB10 1SA, UK. Spanish National Cancer Research Centre (CNIO), C/Melchor Fernández Almagro 3, Madrid 29029, Spain. MRC Cancer Unit, University of Cambridge, Hutchison-MRC Research Centre, Box 197, Cambridge Biomedical Campus, Cambridge CB2 0XZ, UK. email: bh418@mrc-cu.cam.ac.uk; pj3@sanger.ac.uk NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 1 1234567890():,; ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 he squamous epithelia that cover the external surface of the and sweat ducts, which form distinct proliferative compartments 5,8–10 body and line the mouth and esophagus consist of layers of independent of the epidermis (Fig. 1a) . The structure of the Tkeratinocytes. In the mouse epidermis and esophagus cell epidermis also varies with body site. In typical mouse epidermis, division is confined to the deepest, basal cell layer (Fig. 1a). On such as that on the back (dorsum), hair follicles are frequent but commitment to terminal differentiation, proliferating cells exit there are no sweat glands . In contrast, in the mouse paw epi- the cell cycle and migrate to the suprabasal cell layers, before dermis hair and, particularly, sweat ducts are common in the being ultimately shed from the tissue surface. Cellular home- anterior, acrosyringial region around the foot pads, while the 10–12 ostasis requires that cells are generated by proliferation at the posterior, plantar epidermis is devoid of appendages . The ear same rate at which they are shed. Further, to maintain a constant epidermis is different again; it has uniform columns of differ- number of proliferating cells, on average each cell division must entiating cells, not present elsewhere . Finally, the mouse tail has generate one daughter that will go on to divide and one that will the most unusual structure, being a scale forming epidermis like differentiate after first exiting the cell cycle. However, the nature that of chicken legs and Crocodillia rather than typical mam- 1–5 14–16 of the dividing cell population has been subject to controversy . malian skin . This structural diversity has motivated a range Resolving proliferating cell behavior is key for understanding not of studies to define the properties of proliferating cells at each site. only normal tissue maintenance but also processes such as wound Genetic lineage tracing in transgenic mice has emerged as a healing and the accumulation of somatic mutations in normal powerful technique for tracking the behavior of cells within tis- 6,7 17 tissues during aging and carcinogenesis . sues (Fig. 1b) . This is performed in mice expressing two Whilst murine epidermis and esophageal epithelium share the transgenic constructs (Fig. 2a,b). The first is a genetic switch, same basic organization, there are significant differences between using a bacterial recombinase enzyme Cre, expressed either from the tissues. The esophageal epithelium is uniform, with no a transgenic promoter or targeted to a specific gene . A variety of appendages, while the epidermis is punctuated by hair follicles Cre expressing mouse strains have been used for studies of Cell loss Differentiation Skin Oesophagus Proliferation SG HF Lineage tracing Induction Time chase H2BGFP pulse-chase +Dox –Dox Time chase SP model 2xSC model SC-CP model Slow-cycling Rapidly-dividing Progenitor Differentiating stem cell stem cell Fig. 1 Quantitative approaches to cell behavior in murine epithelium. a Structure of the stratified squamous epithelia from the interfollicular epidermis (skin) and esophagus of adult mice. Proliferating keratinocytes are located in the basal layer. Upon differentiation they migrate through suprabasal layers until they are ultimately lost by shedding. A balance should be established between cell division and cell loss to guarantee tissue homeostasis. HF, hair follicle; SG, sebaceous gland. b Rationale of genetic lineage tracing. Low-dose induction in transgenic mice allows recombination and conditional labeling of punctuated keratinocyte progenitors in the basal layer. These cells and their progeny remain labeled and can be tracked to study clonal dynamics over time. c Rationale of Histone 2B-GFP (H2BGFP) dilution experiments (a top-down view of the basal-layer plane is sketched). Transgenic-mouse keratinocytes express H2BGFP protein while on doxycycline (Dox) treatment. After Dox withdrawal cycling cells dilute their H2BGFP content with every division, allowing to study cell-proliferation rate. d Different stochastic cell-proliferation models invoked to explain epithelial self-renewal. Branches reflect different possible fates for a given proliferating cell upon division. SP: single-progenitor model. 2xSC: two stem-cell model, involving two independent types of proliferating cells dividing at different rates. SC-CP: stem cell-committed progenitor model, involving slow-cycling stem cells underpinning a second population of quickly-dividing progenitors. 2 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE Lineage tracing Endogenous X-promoter Tissue Reference ERT locus X cre Axin2 Paw 11 Cre +Tamoxifen (knockin) Lgr6 Back 33 Krt15 Esophagus 31 Reporter R26 R26 STOP Reporter Lrig1 Esophagus (this work) Y-promoter Tissue Reference Cyp1a1 (Ah) Tail 3 Transgene ERT Cyp1a1 (Ah) Ear 32 // // Y cre (random Cre Cyp1a1 (Ah) Esophagus 5 +Tamoxifen insertion) Cyp1a1 (Ah) Back 22 (+β-NF) Reporter R26 Ivl Tail 4,21 R26 STOP Reporter Krt14 Tail 4,21 H2BGFP pulse-chase X-promoter Tissue Reference tTA Krt5 Tail 4 w/o Dox tTA Tet-Off x Krt5 Back 20 tetO H2BGFP H2B-GFP tetO Y-promoter Tissue Reference rtTA Dox d +Dox Gt(ROSA)26Sor Esophagus 5 Tet-On x rtTA Gt(ROSA)26Sor Back 22 tetO H2BGFP Gt(ROSA)26Sor All (this work) H2B-GFP tetO Fig. 2 Transgenic-mouse models used for lineage tracing and cell-proliferation studies. a, b Transgenic mice for lineage tracing are designed with two genetic constructs. The first codes for a bacterial Cre recombinase- mutant estrogen receptor fusion protein (CreERT), which can be targeted to a specific endogenous locus (a) or be under control of a transgenic promoter, randomly inserted in the genome (b). The second construct codes for a conditional fluorescent protein reporter, typically targeted to the ubiquitously expressed Rosa26 locus. Treatment with tamoxifen induces Cre protein internalization to the nucleus, allowing expression of the reporter following Cre-mediated excision of a loxP-flanked STOP cassette. Specific details of constructs used in previous literature for lineage tracing in squamous epithelia are listed on the right. Note expression of Cre from the transgenic Cyp1a1 arylhydrocarbon receptor, Ah) promoter requires additional treatment with a Ah inducer, β-napthoflavone (β-NF). c, d Transgenic mice for H2BGFP-dilution experiments are designed with a first construct, typically targeted to a constitutive promoter, coding either for a tetracycline-controlled transactivator (tTA; Tet-Off system) (c) or a reverse tetracycline-controlled transactivator (rtTA; Tet-On system) (d). A second construct codes for a Histone 2B-green fluorescent protein fusion (H2BGFP) controlled by a tetracycline-response promoter element (tetO; sometimes referred to as pTRE). Treatment with tetracycline or its derivative doxycycline (Dox) preempts tTA protein from binding to tetO elements in Tet-Off systems, causing repression of pTRE-controlled H2BGFP expression, whilst it is required for binding of rtTA to tetO elements in Tet-On systems, hence having an opposite effect. Dox is administered for induction and withdrawn during the H2BGFP-dilution chase in Tet-On mice, while in Tet-Off animals its application gets required for the duration of the experiment. esophageal epithelium and epidermis (Fig. 2a,b). Cre is fused to a The stable H2B-GFP protein is diluted by cell division, so if the mutant hormone receptor so it is only active following treatment tissue contains cell populations dividing at different rates, the with a drug, giving control over when recombination is induced. more slowly dividing cells will retain higher levels of protein . Using low doses of inducing drug allows the labeling of scattered Measurements of the rate of loss of fluorescence have been used 4,5,20 single cells. The second construct is a reporter, such as a fluor- to estimate the rate of cell division . escent protein, typically targeted to the ubiquitously expressed Gt Lineage tracing has ruled out older deterministic models of a (ROSA)26Sor (Rosa26) locus. The reporter is only expressed fol- proliferative hierarchy of asymmetrically dividing stem cells lowing the excision of a “stop” cassette by Cre and expression generating ‘transit amplifying’ cells that undergo a fixed number persists in the progeny of the labeled cell. If the cells are labeled at of divisions prior to differentiation . These models predict that a low frequency, single-cell-derived clones of reporter expressing clone sizes will rise and then remain stable. In multiple lineage cells result. If a representative sample of proliferating cells is tracing experiments, however, mean clone size has been found to labeled and their progeny tracked over a time course, statistical increase progressively with time. However, several mutually analysis of the evolving clone-size distributions may be used to incompatible models in which proliferating cells have stochastic infer cell behavior . fate have been proposed that do appear consistent with the data Alongside lineage tracing, a complementary transgenic assay in one or more experiments (Fig. 1d; Supplementary Methods). may be used to detect cells cycling at different rates and infer the The simplest stochastic model, the single-progenitor (SP) average rate of cell division (Fig. 1c). This uses a transgenic, drug hypothesis, proposes that all dividing keratinocytes are func- regulated synthetic promoter to control expression of a protein tionally equivalent and generate dividing and differentiating 3,5 comprising Histone 2B fused to green fluorescent protein (H2B- daughters with equal probability . An alternative stem cell- GFP) (Fig. 2c, d). The H2B-GFP is initially expressed at high committed progenitor (SC-CP) paradigm, applied to the epi- levels in keratinocytes. Its transcription is then shut off and levels dermis proposes a hierarchy of rare, slowly cycling stem cells of H2B-GFP protein measured by microscopy or flow cytometry. which generate stem and progenitor daughters. The progenitors NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 are biased toward differentiation so continual stem cell pro- populations (Figs. 3d, 4; Supplementary Data 3; Supplementary 4,21 liferation is required . A third model argues that two inde- Methods). pendent populations of stem cells (2xSC) dividing at different To further challenge the hypothesis that there is a single rates exist in the epidermis . These models all give comparably proliferating cell population, we examined whether this model good fits to the results from individual experiments. However, can recapitulate the observed H2BGFP intensity distributions at each has been proposed on the basis of distinct data sets analyzed each time point. For a given average division rate, we performed by different inference and fitting procedures, with limited testing simulations of H2BGFP-dilution kinetics under a wide range of of alternative hypotheses. possible underlying (Gamma) distributions for individual cell- Motivated by the disparity of the proposed models of cell cycle times (Figs. 3e, 4b, d, f, Supplementary Methods). We find dynamics we set out to determine if a single model was consistent that the form of the H2BGFP histograms over time can indeed be across multiple data sets in both esophagus and different epi- fully described by a single population of cells, dividing within a dermal regions. We use cell-cycle properties from the H2B-GFP relatively narrow range of cell-cycle times, further supporting the dilution data to fit lineage tracing results by maximum likelihood SP model (Figs. 3c, f, g, 4; Supplementary Data 4). parameter inference. We find that the data are consistent with a Altogether, these observations strongly argue against scenarios simple SP model of homeostasis. We also show that the fates of of heterogeneous proliferating cell populations, such as the SC- pairs of sister cells are anti-correlated, and that the basal layer CP or 2xSC models, at all sites other than in the tail where contains a substantial proportion of cells which will differentiate marked variation between animals precluded reliable inference on rather than going on to divide. cell-proliferation rates (Fig. 5c). We conclude that in the basal layer of the epidermis at multiple body sites and in the esophagus proliferating cells divide at a unique average rate with highly Results homogenous cell-cycle periods, consistent with the SP model Cell-cycle times in epidermis and esophagus. Analysis of cell (Table 1). proliferation in epithelia offers a simple way to test the predic- tions of the disparate models of epithelial homeostasis by iden- tifying the level of heterogeneity in the division rate of basal-layer A common analytical approach to resolve cell behavior. The cells. The SP model predicts a single-cell population dividing at ability of lineage tracing to track the behavior of cohorts of the same average rate while the alternative hypotheses argue for proliferating cells and their progeny over time courses extending discrete populations dividing at different average rates. We to many rounds of cell division offers the potential to validate therefore investigated the dilution of H2B-GFP in the epidermis models of homeostasis. Having established the homogeneity in M2rtTA and esophagus of R26 /TetO-H2BGFP mice (Fig. 3a). The the division rate of basal-layer cells, we then set out to determine animals were treated with doxycycline (Dox) for 4 weeks to whether clonal dynamics across different lineage tracing data sets induce H2BGFP expression. Dox was then withdrawn and were consistent with the SP paradigm. H2BGFP protein levels in individual basal keratinocytes tracked Multiple lineage tracing studies have been published but these by direct, in situ measurement of GFP fluorescence from confocal used distinct approaches to infer models of cell behavior and did images of epithelial wholemounts at multiple time points. We not apply the additional constraint imposed by measuring the 3,4,11 examined esophagus and epidermis from plantar area of the cell-cycle time distribution . Computational simulations hindpaw, ear, and tail (Figs. 3c, 4, and 5; Supplementary showed that the SP, SC-CP and 2xSC models all predict very similar development of clonal features over time, which rendered Movies 1–3). Optical sections through the deepest, basal cell layer were taken over at least 5 fields of view per tissue/animal and them hardly distinguishable from lineage tracing data alone H2BGFP fluorescence quantified for all non-mitotic nuclei fol- (Supplementary Methods). However, as our cell-proliferation lowing image segmentation based on DAPI staining (Fig. 3b; analyses do not support the SC-CP and 2xSC paradigms we Methods). Non-epithelial cells in the form of CD45 leukocytes, focused on testing the SP model. which retain high levels of H2BGFP, were excluded from the By incorporating the measurement of the average division rate, analysis, but served as internal reference for label retention (e.g., we could reduce the uncertainty in the parameter estimation, a Fig. 3c, insert; Figs. 4b, d, f and 5b, Supplementary Data 1). In problem that has been generally overlooked in these stochastic addition, for the analysis below we included a recently published models (Supplementary Fig. 1A-B). For example, a relatively high data set from dorsal epidermis performed using an identical division rate and modest proportion of symmetric division protocol (Fig. 4f; Supplementary Movie 4). outcomes predict a similar clone-size distribution to those with We first examined images for the presence of label-retaining a slower turnover rate but higher level of symmetric divisions. In cells (LRCs) (Supplementary Data 1). We found no keratinocyte turn, whilst long-term model predictions on clone-size distribu- LRC in the basal cell layer of the esophagus or any epidermal site tions remained largely unaffected by the assumptions on the cell- other than the interscale region of the tail (Fig. 5a). Rare cycle time distribution, introducing realistic estimates for the keratinocyte LRCs (4/1923, i.e., 0.2% of basal-layer keratinocytes) distribution of individual cell-cycle lengths affected short-term were observed in interscale epidermis, in a single animal, 18 days clone-size predictions, impacting on the inferred parameter values after DOX withdrawal (Fig. 5b). Their scarcity however suggests (Supplementary Fig. 2). This is due to the probability of a chain of that they are unlikely to make a substantial contribution to tissue consecutive division events deviating from the average rate, for maintenance. example a run of several consecutive divisions shorter than Next, we performed a quantitative analysis of the time series of average cell-cycle times (Supplementary Fig. 2a). This results in a the individual-cell H2BGFP intensity histograms (Supplementary broadening of the clone-size distribution at early time points after Data 2). If there were multiple subpopulations of cells proliferat- labeling. At later times, where many rounds of division have ing at different rates the distribution of H2BGFP intensities occurred in each clone, these random cycle time variations regress would progressively diverge, becoming wider over time. We toward the mean cycle time of the population (Supplementary found no evidence of such behavior in the esophagus and at Fig. 2a). This disputes most analyses that use a Markovian multiple sites in the epidermis (Fig. 3c; Supplementary Fig. 4A, B, implementation which makes the biologically implausible assump- D, F). Specifically, several statistical tests of the modality of the tion that cell cycle times are distributed exponentially (i.e., the distribution were applied, showing no evidence for multiple likeliest time for a cell to divide is immediately after the division 4 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a M2rtTA b Rosa26 TetO-HGFP DAPI CD45 H2B-GFP R26 rtTA tetO H2B-GFP ≥ 5 fields of view per tissue/mouse +Dox Basal Collection * +Dox layer –4 weeks 0 days 7 days 12 days 18 days ROIs ROIs Exclude Exclude H2B-GFP mitotic nuclei (*) CD45+ (�) quantification Unimodality test c d 0 days 7 days 12 days 18 days Animal no. 123 4 Oesophagus n.s. n.s. n n.s. .s. Paw n.s. n.s. n n.s. .s. n n.s. .s. Ear n.s. n.s. n n.s. .s. n n.s. .s. Back n.s. n.s. n n.s. .s. n n.s. .s. n n.s. .s. 100 ± 2.1% 19.0 ± 0.9% 3.4 ± 0.2% 0.8 ± 0.1% 0.8 Tail scale § § 0.01 0.01 n n.s. .s. 0.6 0.4 Tail interscale n.s. n.s. n n.s. .s. 0.2 Unimodal Multimodal –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 (p > 0.05) (p < 0.05) (Fig. S3C) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 EXP t fit GAM t fit cc cc H : EXP t 0 cc Time Time t end H : GAM t t 1 cc Time Time t end f g –2 –4 –6 –8 EXP t cc –10 GAM t cc 02468 10 –12 Cell-cycle period, t (days) cc 0 4 8 12 16 20 Time (days) that generated it). We therefore developed a robust quantitative Clonal dynamics in esophageal epithelium. In order to explore approach where cell-cycle attributes estimated from H2B-GFP in vivo clonal dynamics, we began by studying a lineage tracing experiments were embodied in (non-Markovian) model simula- data set from mouse esophageal epithelium (Fig. 7a). In this tions, and a subsequent maximum likelihood estimation (MLE) experiment we used a strain (Lrig1-cre) in which a tamoxifen- method was applied across the available data sets for each body regulated form of Cre recombinase and enhanced green fluor- site to challenge whether each of them was consistent with the SP escent protein (EGFP) are targeted to one allele of the Lrig1 8,23–25 paradigm (Fig. 6; Supplementary Methods). locus . We found LRIG1 protein was ubiquitously expressed NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 5 Rel. frequency log (H2BGFP intensity) Simulations Non-Markovian Markovian Prob. division Prob. division Frequency ^ ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 M2rtTA Fig. 3 Analysis of cell proliferation in epidermis and esophageal epithelium. a Protocol: R26 /TetO-H2BGFP mice were treated with doxycycline (Dox) to induce H2BGFP expression (green). Following Dox withdrawal, H2BGFP transcription ceases and protein levels dilute with cell division. b H2BGFP fluorescence was quantified in non-mitotic basal cell nuclei in optical sections of the basal layer of wholemounts. Scale bar, 20 μm. c Representative confocal z stacks of the esophageal basal layer showing H2BGFP (green) and pan-leukocyte marker CD45 (red). Images are representative of a total of 15 fields of view from 3 individual biologically independent mice at 0, 7, and 12 days and 10 fields of view from two individual mice at 18 days. Infrequent label- retaining cells (LRCs) (arrowhead) are positive for CD45 (insert; blue: DAPI). Scale bars, 20 μm. Histograms show keratinocyte H2BGFP intensity for each time point (in green; bottom panels) with mean ± s.e.m. values from each field of view. Best fits for the SP model with exponential- (gray) or gamma- distributed cell-cycle periods (orange lines) are shown. Raw values for H2B-GFP intensity are given in Supplementary Data 2. d Outcome of Silverman’s unimodality test applied to individual-cell H2BGFP distributions at 18 days in esophagus and epidermis (analyses are separated per animal; this testis effectively two tailed, no multiple-testing corrections are made on p values, exact p values are given in Supplementary Data 3) . A single tissue in a single animal was found to be bimodal, in this case due to variability between fields of view where cells differed by a single round of division (see Fig. 5c). e SP- model simulations using different distributions of cell-cycle times t (EXP: exponential; GAM: Gamma) with the same average division rate (blue vertical cc line), the Gamma-shaped distributions predict a more homogeneous dilution. f Time course in the H2BGFP intensity distributions from esophageal epithelium (normalized to average keratinocyte intensity at time 0). Green boxplots: experimental data from n = 3 biologically independent mice at 0, 7, and 12 days and 2 at 18 days. Centre line of box is median value, box indicates 25th and 75th centiles and whiskers indicate minimum and maximum values. The computed average division rate ⟨λ⟩ = 2.9/week; solid black line. Gray region: range of H2BGFP intensities predicted from models assuming exponentially distributed cell-cycle times (interquartile range, inner dashed black lines). Light orange region: range of H2GFP intensities inferred with a gamma cell-cycle time distribution (interquartile range in dark orange, delimited by inner dashed orange lines). g Most-likely (gamma) shapes for the distribution of the cell-cycle period of esophageal keratinocytes, estimated from fits to the H2BGFP-dilution data. A conservative solution (in dark orange) is used for further inference. Vertical blue line: mean cell-cycle period. in the basal layer of esophageal epithelium in wild type mice hallmarks of neutral competition, in which clonal dynamics result (Supplementary Fig. 3A). Consistent with this finding, in Lrig1- from stochastic cell fates, with an average cell division generating cre animals, EGFP, reporting Lrig1 transcription was detected in one proliferating and one differentiating daughter cell, a scenario 3,29,30 94 ± 0.3% (s.e.m.) of basal cells (Supplementary Fig. 3B). These consistent with the SP model (Supplementary Methods) . observations indicate Lrig1 is widely expressed in the proliferative The measurement of the average cell division rate (λ) and compartment of esophageal epithelium and is suitable for lineage inference of the cell-cycle time distribution constrain the fitting of tracing of proliferating esophageal keratinocytes (Supplementary lineage tracing data, providing a stringent test of the candidate SP Methods). model (Supplementary Figs. 1A–C, 2E). Within this paradigm To track the fate of basal cells, Lrig1-cre mice were crossed with unknown parameters are the probability of a progenitor cell flConfetti/wt the Rosa26 (Confetti) reporter strain which labels cells division generating two dividing (PP), or two differentiating (DD) with one of four possible fluorescent proteins (green, GFP, cyan, daughters (r), and the stratification rate (Γ), which in homeostasis CFP, yellow, YFP, or red RFP) after recombination (Supplemen- sets the fraction of progenitor cells in the basal layer (ρ) (Fig. 6). 26,27 tary Fig. 3c, d) . In some Cre inducible mouse lines, reporter Our technique for identifying the most appropriate cell-cycle expressing clones can appear without induction with Tamoxifen. distribution coupled with an MLE grid search estimated However, no fluorescent protein expression was found in adult parameter values that gave an excellent fit with the clone-size uninduced Lrig1-cre/Confetti mice (Supplementary Data 5) . distributions at both early and late time points for the Lrig1/ Next, cohorts of Lrig1-cre/Confetti animals were treated with a confetti data set (Fig. 7d–e; Table 1). The model predictions were low dose of Tamoxifen that resulted in labeling of only 1 in 300 ± within the 95% confidence interval of the measured proportion of 106 (mean ± s.e.m.) basal cells at 10 days post induction. Clones clones of a given size at each time point in 27/28 cases. To containing one or more basal cells were imaged in esophageal quantify the quality of the fit, we calculated both the determina- epithelial wholemounts from at least three mice at multiple time tion coefficient between the model prediction and measured clone points over 6 months following induction (Fig. 7b; Supplemen- sizes, averaged across all time points, R , and the standard error tary Fig. 4a; Supplementary Data 5). Only CFP, YFP, and RFP of the fit, S , a measure of the standard deviation between the expressing clones were counted because of Lrig1-driven GFP model estimates and the experimental data, averaged over all time expression in all basal cells. points. For the fit of the SP model to the Lrig1/Confetti data set, 2 2 The pooled Confetti clone data set displayed several important R = 0.93, S = 4.3. Values of R and S for experimental data at T T features, which were recapitulated by clones labeled with each each time point are given in Supplementary Data 4. individual reporter. No statistically significant differences were Next, we applied the same approach to an independent, observed between CFP, YFP, and RFP clone-size distributions at published lineage tracing data set from esophageal epithelium each time point (see Methods). The density (clones/area) of where clones were labeled with YFP by Cre expressed from an labeled clones decreased progressively, consistent with clone loss inducible Cyp1a1 (Ah) promoter in AhYFP mice . Parameter through differentiation, while the number of basal and suprabasal values very similar to those from the Lrig1/confetti experiment cells in the remaining clones rose (Fig. 7c; Supplementary Fig. 4B, gave predictions from the SP model within the 95% CI for all 49 C). The proportion of labeled basal cells remained constant points in the experimental data set (Fig. 7d, e; Supplementary during the experiment, indicating the labeled population was self- Fig. 4e; Table 1; Supplementary Data 4; Supplementary Methods). maintaining over a 6-month period, consistent with labeled cells Quantifying the quality of fit, we found R = 0.98, S = 2.8. We T T being a representative sample of all proliferating cells in the noted that including the cell-cycle time constraints resulted in an homeostatic tissue (Fig. 7c). At late time points, the clone-size improved agreement with early time point clone sizes compared distribution scaled with time. This means that if, for example, with the original publication (R = 0.97, S = 3.3), where cell- T T time doubles, not only the average clone-size shape and breadth cycle time distributions were assumed exponential (see Supple- of the clone-size distribution also double. More formally, the mentary Data 4 for detailed goodness-of-fit statistics) . probability of seeing clones larger than x times the average clone As a further validation, we tested the predictions of the SP model PR1 mT/mG size became time-invariant, following a simple exponential f(x) = against a third, more limited data set from Krt15-cre R26 −x 29 e (Supplementary Fig. 4D) . Collectively, these features are mice in which a red-to-green fluorescent reporter was used with 6 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications ^ NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a H : one population H : two populations 0 1 (SP model) (SC-CP or 2xSC) Rapidly-dividing (progenitors) Time Slow-cycling (stem cells) –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 SC-CP model SC-CP model 2xSC model (Sada et al. 2016) (Mascré et al. 2012) (Sánchez-Danés et al. 2016) –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 b c 0 days 7 days 12 days 18 days ac. pl. 02468 10 Cell-cycle period, t (days) cc 100 ± 4.1% 34.4 ± 3.4% 19.9 ± 2.0% 2.2 ± 0.4% 0.8 EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 d e 0 days 7 days 12 days 18 days 02468 10 12 Cell-cycle period, t (days) cc 100 ± 3.6% 38.8 ± 3.1% 26.5 ± 3.8% 8.4 ± 0.9% 0.8 EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 f g 0 days 5 days 14 days 02468 10 12 Cell-cycle period, t (days) cc 0 days 5 days 11 days 14 days 21 days 0.8 100 ± 4.6% 56.9 ± 5.2% 37.1 ± 3.6% 19.9 ± 1.7% 9.3 ± 0.7% EXP t fit 0.6 cc 0.4 GAM t fit cc 0.2 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) log (H2BGFP int.) 2 2 2 2 2 inducible Cre expressed from a Krt15 promoter (Supplementary to three independent lineage tracing data sets using different Methods) . Although the SP paradigm was criticized by these combinations of transgenic Cre and reporter alleles strongly authors, it yielded an adequate fit(R = 0.91, S = 2.6) with their supports the conclusion that the esophageal epithelium is T T own data over the experimental time course (Supplementary Fig. 4f; maintained by a single-progenitor population and argues for the Supplementary Data 4). The consistent agreement of the SP model reliability of our parameter estimates. NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 7 Frequency Rel. frequency Rel. frequency Rel. frequency Frequency Frequency Frequency ^ ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 Fig. 4 H2BGFP-dilution and cell-cycle inference in skin epidermis at different sites. a Theoretical distributions of individual-cell H2BGFP intensities expected after 3 weeks dilution under the SP, SC-CP and 2xSC scenarios assuming gamma-distributed cell-cycle periods. Simulations considered an average division rate for stem cells 4× slower than for progenitors in SC-CP and 2xSC (top panels). Predictions of each model using published parameters are shown below. All theoretical SC-CP or 2xSC scenarios represented in this figure were significant by 6 different unimodality tests (see Supplementary M2rtTA Data 3). b, d, f Representative confocal z stacks of hindpaw (plantar), ear and dorsal epidermal basal layer, respectively, from R26 /TetO-H2BGFP mice, showing H2BGFP (green) and immunostaining for pan-leukocyte marker CD45 (red). Analysis of hindpaw epidermis was confined to the posterior, plantar region (pl.), excluding the acrosyringia (ac.), cartoon. Label-retaining cells (LRCs) are CD45+ leukocytes (arrowheads). Scale bars, 20 μm. Images shown are representative of a total of 18 fields of view from 3 mice at 0 and 12 days, 14 fields of view from 3 mice at 7 and 18 days in (b), 18 fields of view from 3 mice at 0, 7 and 12 days and 12 fields of view from 2 mice at 18 days in (d), and 20 fields of view from 4 mice at 0 and 11 days and 17 fields of view from 4 mice in (f). Lower panels: Individual-keratinocyte H2BGFP intensity levels (in green) with mean ± s.e.m. values from different fields of view. Best fits for the SP model with exponential- (gray) or gamma-distributed cell-cycle periods (orange lines) are shown. See Supplementary Data 2 for raw intensity values and summary statistics. c, e, g Best estimates for the (gamma) distribution of the keratinocyte cell-cycle times in hindpaw, ear, and dorsal epidermis, respectively, estimated from fits to H2BGFP-dilution data. Conservative solutions (in dark orange) are used for further inference. Vertical blue lines: average cell-cycle period per site: ⟨λ⟩ = 2.0, 1.5, 1.2/week for paw (c), ear (e), and dorsum (g), respectively. 4,21 Clonal dynamics in skin epidermis. We next investigated clonal scale regions (Fig. 5a) . These claims were primarily supported dynamics in the epidermis through available lineage tracing data by the observation of LRCs in the interscale region in H2BGFP- tTA sets from the typical interfollicular epidermis of the mouse dilution experiments in Krt5 /pTRE-H2BGFP mice. However, hindpaw (plantar), ear and back (dorsum). Applying the MLE our quantitative reanalysis of this data set showed the SP-model approach constrained by the cell-cycle time analysis at each body fits the reported H2BGFP intensity histograms over time as well 2 2 site yielded slightly improved fits of the SP model to data from as the SC-CP model (SP R = 0.89, S = 0.04 vs. SC-CP R = T T T ERT Rainbow 2 Axin2-cre R26 animals in paw epidermis (R = 0.98, 0.89, S = 0.04) (Supplementary Fig. 6A; Supplementary Data 4) T T S = 2.8) and also ear epidermis (R = 0.97, S = 3.5) and dorsal T T T . Even though we cannot discard the possibility of a interfollicular epidermis (R = 0.94, S = 5.0) from AhYFP mice T T subpopulation of slow-cycling stem cells in the tail, such cells compared with those fits reported in the original publications would seem to be rare in interscale epidermis (Supplementary 2 2 2 (R = 0.93, S = 5.16; R = 0.97, S = 3.52; R = 0.92, S = T T T T T T Data 1). We noted that a large proportion of the rare LRCs were 6.01; respectively) (Fig. 8; Supplementary Methods, Supplemen- identified as CD45 expressing leukocytes in our data set (Fig. 5b; 11,22,32 tary Data 4) . Despite differences in average keratinocyte Supplementary Data 1). Further analysis argued that there was no division rates across territories (λ ≈ 2.0, 1.5, 1.2/week for plantar conflict between the reported tail lineage tracing data and the SP hindpaw, ear and dorsum, respectively), all analyzed regions model (Supplementary Fig. 6B-H; Supplementary Methods; share comparable intermediate proportions of progenitor basal Supplementary Data 4). cells ρ (~55%, the rest corresponding to differentiating basal cells) and a predominance of asymmetric cell divisions (i.e., low inferred values for the probability of symmetric division, r < 0.25) Discussion (Table 1; Supplementary Data 4). Overall, we find that combining cell-cycle distribution analysis Particularly relevant are the implications for the mode of with lineage tracing argues mouse esophageal epithelium and keratinocyte renewal in back skin, as a previous work claims that epidermis are maintained by a single population of progenitor two stem cell populations dividing at different rates coexist at cells, with the sole possible exception of the interscale compart- this site (2xSC model) . This argument was supported by a ment of tail skin. The quality of the fit of the SP model to the data tTA quantitative analysis of H2BGFP-dilution patterns in Krt5 / is equivalent to or exceeds that of more complex models, ren- pTRE-H2BGFP mice, a system that differs from that we use dering the need to invoke additional cell populations redundant. above in that mice are treated with Dox to suppress H2B-GFP The nine lineage tracing data sets analyzed include a variety of expression instead of using it as activator (Supplementary Cre and reporter strain combinations, and are all consistent with Fig. 5A) . However, in that publication, which rejected the SP the SP model. In addition, live imaging studies of the epidermis model, exponential distributions for cell division/cell stratifica- are consistent with a single proliferating cell population main- tion rates were assumed. Here we have shown this to be taining the tissue . inappropriate for the short time scale of the experiment Quantitative analysis of cell proliferation in the different tissue (Supplementary Methods). Our computational reanalysis, con- types identifies further constraints that must be considered by strained by cell-cycle time distributions demonstrated the SP researchers exploring the appropriateness of alternative models. tTA model gave as good a fit to the Krt5 /pTRE-H2BGFP-dilution The original SC-CP and 2xSC models invoked 12% and 30% of data as the more complex 2xSC hypothesis (SP R = 0.85, S = T T basal-layer keratinocytes constitute slow-cycling stem cells, 2 2 4,20 0.06 vs. 2xSC R = 0.87, S = 0.05 for basal layer; SP R = 0.78, T T T respectively . Histone dilution experiments have allowed us to S = 0.07 vs. 2xSC R = 0.79, S = 0.06 for spinous layer) T T T make strong statements about the nature of any proposed second (Supplementary Fig. 5B; Supplementary Methods; Supplemen- population. For each body site, with the exception of tail inter- tary Data 4). Indeed, the inferred parameter values from the scale epidermis, no keratinocyte label-retaining cells were detec- AhYFP mouse back skin epidermis proved robust, providing ted in over 2000 cells imaged at each location. It follows that any good fits to another lineage tracing data set from the same body slow-cycling stem-cell population must have substantially fewer ERT flConfetti 2 site in Lgr6-eGFPcre Rosa26 mice (R = 0.96, S = T T than one slow-cycling cell per thousand basal keratinocytes to be 2.39) (Supplementary Fig. 5C; Supplementary Methods; Supple- compatible with observations reported here, making it unlikely mentary Data 4) . that such slow-cycling cells will make a detectable contribution to Finally, we turned to revisit clonal dynamics in the mouse tail tissue homeostasis. The hypothesis that two subpopulations exist, epidermis. Previous studies of tail have argued that the but that they both divide at a similar rate, is hard to sustain in the hierarchical SC-CP paradigm applies to proliferating cells in the face of the close agreement of the simpler SP model across all the interscale areas while the SP paradigm describes behavior in the analyzed data sets. 8 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications ^ NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE a M2rtTA b Rosa26 0 days 7 days 12 days 18 days TetO-HGFP Scale Interscale Hair follicle openings Unimodality test –2 –4 –2 –6 –8 –4 1 2 34567 –10 Field –2 –4 –6 –8 –10 048 12 16 20 Time (days) Fig. 5 H2BGFP-dilution analysis in tail epidermis. a Structure of the mouse tail epidermis. 3D reconstruction of confocal z stacks (top panel) and orthogonal xyz view (mid panel). Apical blister-shaped regions (scale) alternate with deeper regions (interscale) (boundaries delineated by dotted lines), being arranged in lines separated by hair follicles (asterisks). Green: H2BGFP expression; white: KRT14 immunostaining (as a marker of basal layer); red: CD45 immunostaining; blue: DAPI. Scale bars, 200 μm. Bottom panel: Cartoon illustrating the tail skin structure. b Representative confocal z stacks of scale M2rtTA and interscale regions of tail epidermis during H2BGFP chase experiments in R26 /TetO-H2BGFP mice, showing H2BGFP (green), immunostaining for KRT14 (white) and immunostaining for pan-leukocyte marker CD45 (red). Images from the same time point correspond to different mice to illustrate inter- animal variation. Label-retaining cells (LRCs) are highlighted with arrowheads (CD45+ cells) or full arrows (CD45− cells) (details in inserts; blue: DAPI). Scale bars, 20 μm. c Time course of basal-layer keratinocyte H2BGFP intensity distributions from scale and interscale regions of tail. Experimental data shown as boxplots per individual biologically independent mouse (intensities normalized to average keratinocyte intensity at time 0, raw H2B-GFP intensity values are given in Supplementary Data 2). n = 3 animals at each time point except 18 days where n = 2 mice. Centre line of box is median value, box indicates 25th and 75th centiles and whiskers indicate minimum and maximum values. Solid black lines: average H2BGFP-dilution rates (within 95% CI limits—shaded gray areas—where the value of λ = 1.2/week reported by Mascre et al. falls). Insert: Detail of H2BGFP intensity distributions separated per field of view for the single tissue found to be bimodal in the overall, per-animal modality test (see Fig. 3d). Analyses per field of view all resulted non- significant (unimodality). NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 9 n.s. n.s. n.s. n.s. n.s. Overall n.s. n.s. Interscale Scale ** * * * log (H2BGFP int.) log (H2BGFP int.) 2 2 * * * * * * Interscale Scale ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 Table 1 Parameter values inferred for progenitor cell behavior in different murine epithelial regions as derived from quantitative lineage tracing. Tissue Experimental model Reference min Division rate, Symmetric % of progenitor Stratification rate, tcc (days) λ (/week) division prob., r cells, ρ Γ (/week) ERT Esophagus Lrig1-eGFP-Cre / 0.5 2.9 (2.7; 3.0) 0.10 (0.07; 0.15) 65 (50; 96) 5.4 (2.9; 69.6) flConfetti R26- ERT flEYFP Ah-Cre /R26 5 0.5 2.9 (2.7; 3.0) 0.06 (0.04; 0.10) 56 (50; 89) 3.7 (2.9; 23.5) ERT2 Paw Epidermis Axin2-Cre /R26- 11 1 2.0 (1.7; 2.3) 0.14 (0.12; 0.17) 53 (49; 58) 2.3 (1.9; 2.8) Rainbow ERT flEYFP Ear Epidermis Ah-Cre /R26 32 1 1.5 (1.2; 1.7) 0.04 (0.03; 0.06) 54 (47; 72) 1.8 (1.3; 3.9) ERT flEYFP Back Epidermis Ah-Cre /R26 22 2 1.2 (1.1; 1.3) 0.04 (0.03; 0.07) 61 (55; 76) 1.9 (1.5; 3.8) Parameter values indicated correspond with the maximum likelihood estimate (MLE), values in parentheses are 95% confidence bounds (see Supplementary Methods for details). H2BGFP dilution Lineage tracing Average cell cycle time, Clone size distributions at cell cycle time distribution multiple time points Inferred t cc Time Compute likeliest values of: l 1 – 2r {r, r, G } to best fit clone size distributions Data 3–4 5–8 Model 9–16 17–32 33–64 40 >64 1 10 100 Time (weeks) Compare model with experimental data Fig. 6 Method for single-progenitor model testing and parameter inference. Method to single-progenitor model testing and infer model parameters. Orange boxes indicate experiments and resulting data, gray box computational model and parameter estimation. Italics indicate parameters in the SP model. The multimodality testing of H2B-GFP data showed that there is a single population dividing at the same average rate in epidermis and esophagus, consistent with the SP model (Fig. 3d). To test the SP model, the average cell-cycle time (λ) and cell-cycle time distribution were inferred from H2B-GFP experiments. These values are used in computational analysis to estimate the values of the other parameters in the SP model, the proportion of progenitor cells in the basal layer ρ, the proportion of symmetric cell division outcomes r, and the stratification rate of differentiating cells leaving the basal cell layer (Γ). Multiple sets of values for the unknown parameters were tested. For each set of unknown parameter values 100,000 progenitor-derived clones were simulated (lines) and inferred clone-size distributions compared with experimental ones (points) obtained from lineage tracing. The likeliest sets of parameter values were obtained by maximum likelihood estimation for each linage tracing data set. The quality of the fit was assessed by determining whether the simulated values lie within the 95% confidence interval of the experimental clone-size measurements at each time point. The improved resolution of parameter estimates identifies cells cycle comparatively slower in the epidermis, on average differences in cell division rates across the epidermis and the between 3.5 and 6 days depending on body site . However, our esophagus (Table 1). Proliferating cells divide rapidly in the study suggests individual cell-cycle periods are tightly con- esophageal epithelium, on average every ~2.4 days (similar to trolled, showing little variation around average division rate, keratinocyte turnover rate in oral mucosa), while progenitor per territory. 10 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications Computation Experiments Prob. division Clone frequency (%) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE Lrig1-eGFPcreERT/wt R26Confetti/wt ERT Lrig1 EGFP x R26 STOP Confetti cre Label Collection 0 10 days 30 days 84 days 180 days 10 days 30 days 84 days 180 days 15 0.3 0.2 5 0.1 0 0 0 0 50 100 150 200 0 50 100 150 200 10 days 30 days 84 days 180 days Time (days) Time (days) Time (days) d e EXP t GAM t cc cc 1.0 80 3–4 0.8 8 5–8 9–16 0.6 17–32 33–64 2 40 0.4 >64 0.2 0.5 N = 1222 0 0 1.0 100 0.8 80 3–4 5–8 9–16 0.6 17–32 33–64 0.4 2 40 >64 0.2 0.5 N = 1925 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob., r Time (weeks) The proportion of progenitor cells in the basal layer and the our analysis. Instead, the consistent values of r < 0.25 indicate the probability of symmetric cell division outcomes (r) are similar fate of sister cells is preferentially anti-correlated. This phenom- across body sites (Table 1). The insight that a substantial pro- enon can be associated to local coordination of neighboring cell portion of cells in the basal layer will proceed to differentiate stratification and division events . Our results argue that antic- rather than divide will be important for the interpretation for the orrelation of sister cell fates applies generally in the epidermis and growing body of single-cell RNA sequencing data in these tis- esophagus, pointing to common mechanisms of keratinocyte cell 33,35 sues . In addition, the low values of r we identify give insight fate regulation. into the basis of cell fate determination (Fig. 9). In principle, if The single-progenitor model captures the average behavior of every basal cell divides or differentiates with equal probability, as progenitor cells during homeostasis. However, epithelia are fre- proposed by Leblond, r will be 0.25, as expected from any pair of quently subject to wounding. To repair the tissue requires a uncorrelated basal cells . However, this scenario is excluded by temporary imbalance in cell fate, with the progenitors close to the NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 11 ERT EYFP ERT Confetti Average basal clone size Ah-cre /R26 Lrig1-cre /R26 Proportion of P-cells, ρ Proportion of P-cells, ρ Number of clones/mm Stratif. rate, Γ (/week) Stratif. rate, Γ (/week) Clone frequency (%) Clone frequency (%) % labelled basal cells ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ERT/wt flConfetti/wt Fig. 7 Quantitative lineage tracing in esophageal epithelium. a Protocol: clonal labeling was induced in Lrig1-eGFPcre R26 mice and samples analyzed at different times from 10 to 180 days post induction, as single labeled cells develop into clones. See Supplementary Data 5 for source data for panels (c) and (e). b Rendered confocal z stacks of the esophageal basal layer showing typical RFP clones (red) at the times indicated. Blue is DAPI. Scale bars, 10 μm. Images are representative of 104 RFP clones (10 days), 75 RFP clones (30 days), 106 RFP clones (84 days), and 274 RFP clones (180 days). c Quantitative characteristics of the labeled clone population over time: average basal-layer clone size (i.e., mean number of basal cells/ surviving clone) (left panel), average density of labeled clones in the basal layer (middle panel), average fraction of labeled basal cells at the indicated time points (right panel). Observed values are shown in individual biologically independent mice (blue circles, n mice = 3 at 10 and 30 days, 6 at 84 days, and 4 at 180 days) with error bars (black) indicating mean ± s.e.m. of all mice at each time point. A total of 300 or more clones was quantified at each time point. Orange lines: SP-model fit (shaded area corresponds with 95% plausible intervals). Orange line and shading in last panel show mean and s.e.m. across all ERT time points (from n= 16 mice), which is consistent with homeostatic behavior. d SP-model parameter inference on Lrig1- and Ah-Cre driven lineage tracing data sets from esophagus . Parameter estimates are affected by the underlying modeling assumptions on the cell-cycle period, whether default exponential cell-cycle time distributions were considered (solutions in gray) or realistic gamma distributions implemented, as inferred from the cell- proliferation analysis (solutions in orange). Regions within the dashed gray lines fall consistent with the predicted ρ/r ratios from the linear scaling of the average clone size. The total number of clones counted in each data set is displayed in the corresponding graph and previous parameter estimates given in 5 ERT 5 ref. shown as black error bars. e Experimental Lrig1- and Ah-Cre -derived basal-layer clone sizes from ref. (dots with error bars indicating the standard error of a proportion) fit well with the SP model, with gamma-distributed cell-cycle times (lines; prediction from maximum likelihood estimation). Dim dashed lines: fits from ref. . Frequencies for each clone size (basal cell number) are shown in different colors. n = 3 biologically independent mice at each time point except for the Lrig1/confetti where n = 6 mice at 12 weeks and 4 mice at 26 weeks. wound producing an excess of progenitor over differentiating Methods Animals. All experiments were conducted according to the UK Home Office daughters on average. This occurs as part of a coordinated set of Project Licenses 70/7543, P14FED054 or PF4639B40. Male and female adult mice responses that includes cell migration and altered cell differ- aged 3–18 months were used for in vivo experiments. Animals were housed in 5,37,38 entiation . Once the epithelial defect is resolved, the pro- individually ventilated cages and fed on standard chow and maintained in SOPF genitors revert to homeostatic balance. In esophageal epithelium health status. ERT/wt flConfetti/wt Doubly transgenic, Lrig1-eGFPcre R26 mice on a C57/Bl6N and the plantar epidermis, wound repair is achieved by pro- background were generated for lineage tracing studies in esophageal epithelium, by 5,11 genitors alone . In the epidermis at other sites, cells migrating ERT2 8 flConfetti crossing Lrig1-eGFP-ires-cre mice onto a Rosa26 multicolor reporter from other proliferative compartments, the hair follicles and line . Transcription of the Cre recombinase-mutant estrogen receptor fusion 9,10,39,40 ERT sweat ducts, may also contribute to wound healing . The protein (Cre ) is under the control of an endogenous allele of Lrig1. Following ERT induction with tamoxifen, Cre protein internalizes into the nuclei and excises a ability to transiently increase the likelihood of progenitors gen- LoxP-flanked “STOP” cassette resulting in the expression of one of the four erating proliferating progeny provides a rapid and robust M2rtTA Confetti fluorescent reporters (YFP, RFP, CFP, or GFP). R26 /TetO-H2BGFP response to injury. The down side of this adjustable progenitor mice, doubly transgenic for a reverse tetracycline-controlled transactivator (rtTA- fate is that it may be subverted by mutations acquired during M2) targeted to the Rosa26 locus and a HIST1H2BJ/EGFP fusion protein tissue aging, leading to mutant clonal expansions that may (H2BGFP) expressed from a tetracycline promoter element, were used for label- 5,22 22,41–43 retaining experiments . H2BGFP expression is induced by treatment with undergo malignant transformation . doxycycline (Dox) and dilution of H2BGFP protein content can be chased upon How might these findings in mice relate to homeostasis human Dox withdrawal. All animals were induced at 8–12 weeks age. Cohorts of at least epidermis? Human skin differs from that of mice with many two or three animals per time point were culled and esophagus and/or skin more epidermal cell layers and undulates in thickness at most epidermis collected for analysis. body sites creating folds called rete ridges and dermal papillae . Nevertheless, a population of cells with balanced stochastic cell Wholemount preparation and immunostaining. Esophageal epithelium whole- fate generating equal proportions of proliferating and differ- mounts for lineage tracing were prepared as follows: The esophagus was cut longitudinally and the middle two-thirds of the tract was incubated for 3 h in 5 mM entiating cells has been identified in a live imaging study of EDTA in PBS at 37 °C. The epithelium was then peeled away from the underlying human keratinocytes in primary culture . In vivo lineage tracing submucosa, stretched and fixed for 30 min in 4% paraformaldehyde in PBS. in humans is not feasible. However, human epidermis has been Samples were stored in PBS at 4 °C until subsequent analysis. Skin pieces of ~0.5 grafted onto immune compromised mice and injected with len- cm were cut and incubated for 1 h in 5 mM EDTA in PBS at 37 °C. Skin epidermis was then peeled away using fine forceps and processed as described above for the tiviral vectors carrying fluorescent protein reporters. When the esophageal epithelium. resulting clones were imaged 6 months later they were found to For staining, wholemount samples were incubated in Permeabilization Buffer vary widely in size and shape and arise from any point in the (PB) (0.5% BSA, 0.25% Fish Skin Gelatin (FSG), 0.5% Triton X-100/PBS) for 15 basal layer, both in rete ridges and dermal papillae . These min at room temperature (RT), then blocked in 10% goat or donkey serum/PB (according to the secondary antibody used) for 1 h at RT and incubated overnight findings are consistent with the single-progenitor paradigm, but with primary antibody at 4 °C. Primary antibodies used were Lrig1 antibody (R&D cannot provide quantitative challenge to the model available Systems, Cat. AF3688), ITGA6 antibody (clone GoH3, Biolegend, Cat. B204094), in mice. Alexa Fluor® 647 anti-CD45 (clone 30-F11, Biolegend, Cat. 103124), Keratin 14 The lineage tracing approaches considered above have been antibody (clone Poly19053, Biolegend, Cat. 905301). Samples were subsequently 34,36 enriched by live imaging studies of mouse epidermis . Whilst washed four times for 30 min in 0.2% Tween-20/PBS and incubated with an appropriate secondary antibody for 3 h at RT. Secondary antibodies used were lineage tracing resolves the average behavior of a population of Goat or Donkey Alexa Fluor 488/546/555/647 (Molecular Probes). A washing step proliferating cells over many cell generations, live imaging allows with 0.2% Tween-20/PBS was repeated and samples were incubated for 30 min the fate of individual cells to be resolved. Insights gained from live with DAPI (Sigma-Aldrich) and finally mounted in DAKO Vectashield Mounting imaging include showing that cell fate is stochastic, the prob- Medium with DAPI (Vector Labs). ability of generating progenitor and differentiated daughters is M2rtTA equal and that the fate of cells is not coordinated across cell Dilution of Histone 2B-GFP protein content. R26 /TetO-H2BGFP animals generations, all of which are key features of the SP model . were treated with doxycycline (Dox, 2 mg/ml in drinking water sweetened with 10% blackcurrant & apple) for 4 weeks. Dox was then withdrawn and animals We conclude that the single-progenitor model is consistent with culled at different time points to track H2BGFP florescence dilution. Epithelial a large body of lineage tracing and cell-cycle data collated from wholemounts from esophageal epithelium and skin epidermis were imaged on a multiple studies and identifies the behavior of proliferating cells Leica TCS SP8 confocal microscope using ×20 or ×40 objectives at 1024 × 1024 that underpins epidermal and esophageal epithelial homeostasis. resolution, line average 4 and 400 Hz scan speed. Individual-cell H2BGFP 12 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE ERT2 Axin2 cre x R26 STOP Rainbow (Lim et al. 2013) 0 –4 –8 log-Like. ratio statistic 1.0 100 3–4 0.8 80 5–8 9–16 0.6 60 17–32 2 33–64 >64 0.4 40 0.2 0.5 20 N = 9643 0 0 0 0.1 0.2 0.3 0.4 0.5 Symmetric division prob, r Time (weeks) ERT Cyp1A1 R26 STOP YFP (Doupé et al. 2010) cre x 0 –4 –8 log-Like. ratio statistic 1.0 8 3–4 0.8 4 5–8 9–16 0.6 2 60 17–32 >32 0.4 1 0.5 0.2 N = 1233 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob, r Time (weeks) ERT Cyp1A1 R26 STOP YFP (Murai et al. 2018) cre x 0 –4 –8 log-Like. ratio statistic 1.0 100 3–4 0.8 80 5–8 9–16 0.6 60 17–32 >32 0.4 40 0.5 0.2 N = 1682 0 0 0 0.1 0.2 0.3 0.4 0.5 1 10 100 Symmetric division prob, r Time (weeks) Fig. 8 The single-progenitor model fits clone dynamics in different regions of skin epidermis. a–c Left panels: SP-model parameter inference on lineage 11 ERT Rainbow 32 22 tracing data sets from paw epidermis using Axin2-cre R26 animals (a), and ear and dorsal interfollicular epidermis in AhYFP mice (b and c, respectively). Parameter estimates are obtained by MLE based on SP-model simulations constrained by the cell-cycle period distribution inferred from each corresponding cell-proliferation analysis. Regions within the dashed gray lines fall consistent with the predicted ρ/r ratios from the linear scaling of the average clone size. The total number of clones counted in each data set is displayed in the corresponding graph. Black bars are parameter estimates given in the original publications shown, centre is the mid range and bars indicate the maximum and minimum plausible parameter values in simulations in each 11 32 22 paper (a, b, c). Right panels: Experimental Axin2- (a; from ref. ) and Ah-(b and c; from refs. and ) derived basal-layer clone sizes (dots indicate mean ± standard error of proportion) give an excellent fit with the SP model with gamma-distributed cell-cycle times (lines; prediction from MLE). Dim dashed lines: fits obtained with parameter estimates given in the original publications. Frequencies for each clone size (basal cell number) are shown in different colors. See Supplementary Data 4 for goodness-of-fit statistics. intensities were determined by image segmentation/nuclear identification, using a of 1 in 301 ± 106 (mean ± s.e.m.) basal cells by 10 days post induction (allowing semi-automated object-recognition macro (based on the DAPI channel) built in individual clone tracking without merging). Between three and six mice were culled ImageJ, and the process completed by manual curation. Per-cell intensity values per time point. Confocal images of immunostained wholemounts were acquired on given are averaged over all nuclear pixels. All H2BGFP samples were stained for a Leica TCS SP8 confocal microscope (×10, ×20, and ×40 objectives; typical settings CD45 and positive cells excluded from the analysis. for z-stacks acquisition: optimal pinhole, line average 4, bi-directional scan with 400–600 Hz speed, resolution of 1024 × 1024 pixels). The number of nucleated basal and suprabasal cells per labeled clone was Lineage tracing. Low-frequency expression of the Confetti reporters in the Lrig1- counted under live acquisition mode. GFP-labeled clones were not scored due to ERT2 flConfetti/wt eGFP-ires-cre R26 mouse esophagus was achieved by inducing 10- the difficulty of distinguishing them from the constitutive basal GFP expression driven by the Lrig1 cassette. CFP, RFP, and YFP clones were pooled together for week-old animals with intraperitoneal injection of a single dose of 1 mg tamoxifen (100 μl of 10 mg/ml) on two consecutive days . This resulted in a labeling efficiency further analysis (histograms (distributions) of basal-layer clone sizes and average NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 13 Proportion of P-cells, ρ Proportion of P-cells, ρ Proportion of P-cells, ρ –1 –1 Stratif. rate, Γ (week ) –1 Stratif. rate, Γ (week ) Stratif. rate, Γ (week ) Clone frequency (%) Clone frequency (%) Clone frequency (%) ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 l (1 + d) 0.5 (1 – d) 0.5 Progenitor cell Differentiating cell Suprabasal cell #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 0.5 1.5 cd Sibling #1: Sibling #2: <0.25 >0.25 >0.25 <0.25 Local fate coordination Fig. 9 Cell fate coordination underpins single-progenitor dynamics in epidermis and esophagus. a Epidermis and esophageal epithelium are maintained by a single population of progenitor cells. Left panel: Progenitor cells (in green) share the basal layer with post-mitotic keratinocytes in early stages of differentiation (pale red), which are transiently retained in the basal compartment before stratification. Right panel: Simplified representation of the single- progenitor model focusing on individual basal cell fates. Epithelial cell dynamics are dominated by stochastic but skewed fates through spatial coordination between neighboring or sibling cells. Individual basal cells undertake dichotomic decisions: they have the potential to divide but can alternatively differentiate exiting the basal layer. Both probabilities are balanced (50–50%) across the entire proliferating cell population, but can be decompensated or skewed for individual cells depending on their local niche, as reflected with the parameter δ (which works as a context-dependent modulatory factor). b Stochastic progenitor fates explain a scenario of neutral clonal competition dynamics where clones develop into heterogeneous sizes, constrained by cell-cycle time control and fate coordination effects. Displayed is a representative set of epithelial clone dynamics simulated using the parameters inferred for murine esophageal epithelium homeostasis. c, d Our results demonstrate that the outcome of sibling keratinocyte cells is commonly biased toward an excess of asymmetric fates where one decides to divide while the other differentiates, in agreement with a single-progenitor model with low values of r (r < 0.25). number of basal cells). A total of 300, 315, 302, and 305 labeled clones from 3, 3, 6, Reporting summary. Further information on research design is available in and 4 mice at 10, 30, 84, and 180 days post induction, respectively, were quantified. the Nature Research Reporting Summary linked to this article. Regarding the time courses in the number of clones per unit area and the proportion of labeled basal cells, only RFP clones were considered given the low, Data availability variable induction of the other florescent reporters and their overall small The authors declare that the experimental data supporting the findings of this study are contribution (including these numbers did not alter the conclusions). available within the paper and its supplementary information files. ERT flEYFP 5 Lineage-tracing data from Ah-cre R26 derived clones in esophagus , 32 22 ear , and dorsal epidermis were obtained from experimental colleagues (data ERT Rainbow available upon request). Data on induced Axin2-cre R26 clones in Code availability 11 ERT flConfetti 33 hindpaw and Lgr6-eGFPcre R26 in back epidermis were kindly Code used in computational modeling is available in Github: https://github.com/gp10/ provided by the authors. Data from lineage tracing in scale and interscale tail Piedrafita_etal_SI_code/ epidermis were accessed through the online publication material, while authors 4 PR1 mT/mG were unable to provide original data from ref. . Data on Krt15-cre R26 mouse esophagus were retrieved by digitalizing Fig. 2e and Figure S3B from the Received: 29 July 2019; Accepted: 26 February 2020; tTA original publication. A similar procedure was used to extract Krt5 /pTRE- H2BGFP-dilution data from back skin (Fig. 3 from ref. ) and tail epidermis (Fig. 3k from ref. ). Mathematical modeling and statistical inference. Model dynamics were simu- lated using Markovian (Gillespie algorithm) and non-Markovian exact stochastic References Monte Carlo methods implemented in Matlab. A maximum likelihood estimation 1. Marques-Pereira, J. P. & Leblond, C. P. Mitosis and differentiation in the (MLE) approach was followed for parameter inference, except when stated stratified squamous epithelium of the rat esophagus. Am. J. Anat. 117,73–87 otherwise, and best-fit parameters obtained with 95% confidence intervals based on (1965). the likelihood-ratio test (alpha = 0.05). The coefficient of determination (R ) and 2. Allen, T. D. & Potten, C. S. Fine-structural identification and organization of the standard error of the fit(S) were calculated for the evaluation of goodness of fit the epidermal proliferative unit. J. Cell Sci. 15, 291–319 (1974). (GoF). In those cases involving data at various time points, GoF values averaged 3. Clayton, E. et al. A single type of progenitor cell maintains normal epidermis. across the different time points (R and S ) are displayed. A comprehensive list of T T detailed R and S values can be found in Supplementary Data 4. For details of Nature 446, 185–189 (2007). theoretical modeling and computational methods used to infer cell behavior and 4. Mascre, G. et al. Distinct contribution of stem and progenitor cells to clonal dynamics, see Supplementary Methods section below. epidermal maintenance. Nature 489, 257–262 (2012). 14 NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications Time (weeks) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15258-0 ARTICLE 5. Doupe, D. P. et al. A single progenitor population switches behavior to 39. Levy, V., Lindon, C., Harfe, B. D. & Morgan, B. A. Distinct stem cell maintain and repair esophageal epithelium. Science 337, 1091–1093 (2012). populations regenerate the follicle and interfollicular epidermis. Dev. Cell 9, 6. Martincorena, I. et al. Somatic mutant clones colonize the human esophagus 855–861 (2005). with age. Science 362, 911–917 (2018). 40. Levy, V., Lindon, C., Zheng, Y., Harfe, B. D. & Morgan, B. A. Epidermal stem 7. Martincorena, I. et al. Tumor evolution. High burden and pervasive positive cells arise from the hair follicle after wounding. Faseb J. 21, 1358–1366 (2007). selection of somatic mutations in normal human skin. Science 348, 880–886 41. Fernandez-Antoran, D. et al. Outcompeting p53 mutant cells in the normal (2015). esophagus by redox manipulation. Cell Stem Cell 25, 329–341 (2019). 8. Page, M. E., Lombard, P., Ng, F., Gottgens, B. & Jensen, K. B. The epidermis 42. Alcolea, M. P. et al. Differentiation imbalance in single oesophageal progenitor comprises autonomous compartments maintained by distinct stem cell cells causes clonal immortalization and field change. Nat. Cell Biol. 16, populations. Cell Stem Cell 13, 471–482 (2013). 615–622 (2014). 9. Ito, M. et al. Stem cells in the hair follicle bulge contribute to wound repair but 43. Alcolea, M. P. & Jones, P. H. Cell competition: winning out by losing notch. not to homeostasis of the epidermis. Nat. Med. 11, 1351–1354 (2005). Cell Cycle 14,9–17 (2015). 10. Lu, C. P. et al. Identification of stem cell populations in sweat glands and ducts 44. Lavker, R. M. & Sun, T. T. Heterogeneity in epidermal basal keratinocytes: reveals roles in homeostasis and wound repair. Cell 150, 136–150 (2012). morphological and functional correlations. Science 215, 1239–1241 (1982). 11. Lim, X. et al. Interfollicular epidermal stem cells self-renew via autocrine Wnt 45. Ghazizadeh, S. & Taichman, L. B. Organization of stem cells and their progeny signaling. Science 342, 1226–1230 (2013). in human epidermis. J. Invest Dermatol 124, 367–372 (2005). 12. Kamberov, Y. G. et al. A genetic basis of variation in eccrine sweat gland and 46. Silverman, B. W. Using kernel density estimates to investigate multimodality. hair follicle density. Proc. Natl Acad. Sci. USA 112, 9932–9937 (2015). J. R. Stat. Soc.: Ser. B (Methodol.) 43,97–99 (1981). 13. Mackenzie, J. C. Ordered structure of the stratum corneum of mammalian skin. Nature 222, 881–882 (1969). Acknowledgements 14. Spearman, R. I. & Hardy, J. A. Ultrastructure of the contrasting types of We are grateful to Xinhong Lim and Roeland Nusse; Kyogo Kawaguchi, Allon M. Klein keratinization seen in the tail epidermis of the laboratory mouse Mus and Valentina Greco; Anja Füllgrabe and Maria Kasper; and David Shalloway for kindly musculus. Arch. Dermatol Res. 258,33–40 (1977). sharing raw lineage-tracing data and computational algorithms from previous publica- 15. Gomez, C. et al. The interfollicular epidermis of adult mouse tail comprises tions, and for their valuable comments. We thank María P. Alcolea for l feedback on two distinct cell lineages that are differentially regulated by Wnt, Edaradd, and H2BGFP experiments and Kim Jensen for the Lrig1-creERT mouse strain, Esther Lrig1. Stem Cell Rep. 1,19–27 (2013). Choolun and staff at the MRC ARES and Sanger RSF facilities for excellent technical 16. Braun, K. M. et al. Manipulation of stem cell proliferation and lineage support and Hall and Jones’ group members and Fernando Pozo, Fátima Al-Shahrour, commitment: visualisation of label-retaining cells in wholemounts of mouse and Francisco X Real at CNIO for critical comments. This work was supported by grants epidermis. Development 130, 5241–5255 (2003). from the Wellcome Trust to the Wellcome Sanger Institute, 098051 and 296194, Cancer 17. Alcolea, M. P. & Jones, P. H. Lineage analysis of epidermal stem cells. Cold Research UK Programme Grants to P.H.J. (C609/A17257, C609/A27326) and the Royal Spring Harb. Perspect. Med 4, a015206 (2014). Society (UF130039 to B.A.H.). G.P. is supported by a Talento program fellowship from 18. Alcolea, M. P. & Jones, P. H. Tracking cells in their native habitat: lineage Comunidad de Madrid. tracing in epithelial neoplasia. Nat. Rev. Cancer 13, 161–171 (2013). 19. Tumbar, T. et al. Defining the epithelial stem cell niche in skin. Science 303, 359–363 (2004). Author contributions 20. Sada, A. et al. Defining the cellular lineage hierarchy in the interfollicular G.P. performed the computational simulations and carried out the analysis of experi- epidermis of adult skin. Nat. Cell Biol. 18, 619–631 (2016). mental results. A.W. carried out the linage tracing experiments. B.C. and D.F.A. per- 21. Sanchez-Danes, A. et al. Defining the clonal dynamics leading to mouse skin formed H2BGFP-dilution experiments. A.H. and K.M. contributed with image tumour initiation. Nature 536, 298–303 (2016). acquisition and image analysis. V.K. performed preliminary, conceptual work and helped 22. Murai, K. et al. Epidermal tissue adapts to restrain progenitors carrying clonal implementing the Non-Markovian simulation algorithm. B.A.H. and P.H.J. supervised p53 mutations. Cell Stem Cell 23, 687–699 (2018). and designed the work. G.P., B.A.H., and P.H.J. wrote the paper. All authors reviewed 23. Nakamura, T. et al. LRIG1 inhibits STAT3-dependent inflammation to and edited the final version. maintain corneal homeostasis. J. Clin. Invest. 124, 385–397 (2014). 24. Lu, L. et al. LRIG1 regulates cadherin-dependent contact inhibition directing epithelial homeostasis and pre-invasive squamous cell carcinoma Competing interests development. J. Pathol. 229, 608–620 (2013). The authors declare no competing interests. 25. Wong, V. W. et al. Lrig1 controls intestinal stem-cell homeostasis by negative regulation of ErbB signalling. Nat. Cell Biol. 14, 401–408 (2012). Additional information 26. Snippert, H. J. et al. Intestinal crypt homeostasis results from neutral competition Supplementary information is available for this paper at https://doi.org/10.1038/s41467- between symmetrically dividing Lgr5 stem cells. Cell 143, 134–144 (2010). 020-15258-0. 27. Frede, J., Greulich, P., Nagy, T., Simons, B. D. & Jones, P. H. A single dividing cell population with imbalanced fate drives oesophageal tumour growth. Nat. Correspondence and requests for materials should be addressed to B.A.H. or P.H.J. Cell Biol. 18, 967–978 (2016). 28. Jensen, K. B. et al. Lrig1 expression defines a distinct multipotent stem cell Peer review information Nature Communications thanks Qing Nie and the other, population in mammalian epidermis. Cell Stem Cell 4, 427–439 (2009). anonymous, reviewer(s) for their contribution to the peer review of this work. Peer 29. Klein, A. M. & Simons, B. D. Universal patterns of stem cell fate in cycling reviewer reports are available. adult tissues. Development 138, 3103–3111 (2011). 30. Klein, A. M., Doupe, D. P., Jones, P. H. & Simons, B. D. Kinetics of cell Reprints and permission information is available at http://www.nature.com/reprints division in epidermal maintenance. Phys. Rev. E Stat. Nonlin Soft Matter Phys. 76, 021910 (2007). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 31. Giroux, V. et al. Long-lived keratin 15+ esophageal progenitor cells contribute published maps and institutional affiliations. to homeostasis and regeneration. J. Clin. Invest. 127, 2378–2391 (2017). 32. Doupe, D. P., Klein, A. M., Simons, B. D. & Jones, P. H. The ordered architecture of murine ear epidermis is maintained by progenitor cells with random fate. Dev. Cell 18, 317–323 (2010). Open Access This article is licensed under a Creative Commons 33. Fullgrabe, A. et al. Dynamics of Lgr6(+) progenitor cells in the hair follicle, Attribution 4.0 International License, which permits use, sharing, sebaceous gland, and interfollicular epidermis. Stem Cell Rep. 5, 843–855 (2015). adaptation, distribution and reproduction in any medium or format, as long as you give 34. Rompolas, P. et al. Spatiotemporal coordination of stem cell commitment appropriate credit to the original author(s) and the source, provide a link to the Creative during epidermal homeostasis. Science 352, 1471–1474 (2016). Commons license, and indicate if changes were made. The images or other third party 35. Jones, K. B. et al. Quantitative clonal analysis and single-cell transcriptomics material in this article are included in the article’s Creative Commons license, unless reveal division kinetics, hierarchy, and fate of oral epithelial progenitor cells. indicated otherwise in a credit line to the material. If material is not included in the Cell Stem Cell 24, 183–192.e188 (2019). article’s Creative Commons license and your intended use is not permitted by statutory 36. Mesa, K. R. et al. Homeostatic epidermal stem cell self-renewal is driven by regulation or exceeds the permitted use, you will need to obtain permission directly from local differentiation. Cell Stem Cell 23, 677–686.e674 (2018). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 37. Roshan, A. et al. Human keratinocytes have two interconvertible modes of licenses/by/4.0/. proliferation. Nat. Cell Biol. 18, 145–156 (2016). 38. Park, S. et al. Tissue-scale coordination of cellular behaviour promotes © The Author(s) 2020 epidermal wound repair in live mice. Nat. Cell Biol. 19, 155–163 (2017). NATURE COMMUNICATIONS | (2020) 11:1429 | https://doi.org/10.1038/s41467-020-15258-0 | www.nature.com/naturecommunications 15

Journal

Nature CommunicationsSpringer Journals

Published: Mar 18, 2020

There are no references for this article.