An Ovol2‐Zeb1 transcriptional circuit regulates epithelial directional migration and proliferation

Abstract

Directional migration is inherently important for epithelial tissue regeneration and repair, but how it is precisely controlled and coordinated with cell proliferation is unclear. Here, we report that Ovol2, a transcriptional repressor that inhibits epithelial‐to‐mesenchymal transition (EMT), plays a crucial role in adult skin epithelial regeneration and repair. Ovol2‐deficient mice show compromised wound healing characterized by aberrant epidermal cell migration and proliferation, as well as delayed anagen progression characterized by defects in hair follicle matrix cell proliferation and subsequent differentiation. Epidermal keratinocytes and bulge hair follicle stem cells (Bu‐HFSCs) lacking Ovol2 fail to expand in culture and display molecular alterations consistent with enhanced EMT and reduced proliferation. Live imaging of wound explants and Bu‐HFSCs reveals increased migration speed but reduced directionality, and post‐mitotic cell cycle arrest. Remarkably, simultaneous deletion of Zeb1 encoding an EMT‐promoting factor restores directional migration to Ovol2‐deficient Bu‐HFSCs. Taken together, our findings highlight the important function of an Ovol2‐Zeb1 EMT‐regulatory circuit in controlling the directional migration of epithelial stem and progenitor cells to facilitate adult skin epithelial regeneration and repair.

Introduction

Directional migration of epithelial cells is an integral component of tissue development, regeneration, and repair 1, 2, 3. While the cellular and molecular machineries responsible for cell movements are relatively well delineated, little is known about the transcriptional mechanisms that regulate gene expression in epithelial stem and progenitor cells to ensure their directional and collective migration for efficient regeneration and repair. Moreover, how migration of epithelial stem and progenitor cells is coordinated with their proliferative activity during tissue expansion is unclear.

Mouse skin serves as a leading model to study the molecular and cellular mechanisms that control epithelial stem cell function and tissue regeneration 4. During homeostasis, skin epidermis is maintained by epithelial stem cells in the innermost basal layer, which can either self‐renew, or follow an upward path to differentiate into spinous and granular cells of the suprabasal layers culminating in the formation of a protective outer permeability barrier 4. Upon skin injury and integral to the healing process, epidermal cells at the wound margin become activated to re‐epithelialize the wound 5, 6. Central to re‐epithelialization are the collective migration of epidermal basal/spinous cells immediately adjacent the wound and the proliferation of basal cells further out on the wound periphery 7. Migrating keratinocytes exhibit molecular, morphological, cytoskeletal, and adhesive changes that bear resemblance to those occurring during epithelial‐to‐mesenchymal transition (EMT) 8, leading to the prevailing notion that wound re‐epithelialization is a partial EMT process 9. However, to date, the precise role of EMT‐regulatory proteins in epidermal cell migration during wound healing remains to be elucidated.

Adult hair follicles (HFs) undergo cycles of regression (catagen), resting (telogen), and growth (anagen) 10. Regeneration of a new HF is fueled by stem cells that reside in the bulge (Bu‐HFSCs) and their early progenies in the secondary hair germ (HG) 4, 11. During telogen‐to‐anagen transition, HG cells and Bu‐HFSCs become sequentially activated, unleashing a coordinated program of active proliferation and differentiation to generate multiple cell types constituting the new HF 12. Tantalizing clues implicate the pro‐migratory nature of Bu‐HFSCs and their progenies in the HG and outer root sheath. Live cell imaging and lineage tracing have detected Bu‐HFSC migration within and out of the niche 13, 14. During wound healing, Bu‐HFSCs cells also travel toward the skin surface to transiently participate in repair of the interfollicular epidermis (IFE) 15, 16. HG cells also undergo dynamic movements to support the downward extension of the new HF, and contain a keratin 79‐positive subset that becomes specified during early anagen and migrates upwards to initiate hair canal formation 13, 17. Furthermore, outer root sheath cells of the lower HF are capable of rapid and long‐range migration during anagen progression 13, 18. It has been reported that Bu‐HFSC‐specific transcription factor (TF) Sox9 activates other Bu‐HFSC‐TFs when ectopically expressed in epidermal keratinocytes 19 and confers enhanced invasive migration 20. However, at present very little is known about the transcriptional mechanisms that regulate Bu‐HFSC migration and the importance of such control mechanisms in HF regeneration.

Ovol2, a member of the Ovo family of zinc finger TFs, has recently been identified as a critical EMT‐inhibitory factor in developing skin and mammary epithelia 21, 22. These studies prompted us to ask whether Ovol2 also regulates epidermal repair and HF regeneration in adult skin, and if yes whether this role is mechanistically linked to the control of cell migration. We show that epithelia‐specific loss of Ovol2 results in aberrant wound re‐epithelialization, delayed anagen progression, and compromised HF regeneration. We provide ex vivo and in vivo evidence that Ovol2‐deficient epidermal cells and Bu‐HFSCs migrate faster than their control counterparts, but with reduced directionality and proliferation. The migratory defects are near‐completely rescued by simultaneous deletion of Zeb1, which encodes an EMT‐inducing TF 23. Together, our findings highlight an important role for EMT‐regulatory factors in fine‐tuning the migration and proliferation of skin epithelial stem cells to facilitate optimal tissue regeneration and repair.

Results

Ovol2‐deficient newborn keratinocytes show compromised colony formation, and altered expression of EMT and cell cycle genes

To probe a functional requirement for Ovol2 in skin, we first examined the in vitro behavior of newborn primary keratinocytes (NBPKs) derived from skin epithelia‐specific Ovol2 knockout (Ovol2 SSKO: K14‐Cre; Ovol2 ^f/−) mice. Despite lack of a remarkable embryonic epidermal phenotype in vivo 21, compared with control counterparts, Ovol2‐deficient NBPKs grew slower at a high density (Fig 1A) and formed fewer colonies at a clonal density (Fig 1B). Closer examination of individual colonies over time revealed that Ovol2 SSKO cells were capable of initial attachment and divisions, but failed to continuously expand and/or to maintain proper intercellular adhesion (Figs 1B and C, and EV1A and B).

Figure 1. Growth and gene expression defects in Ovol2‐deficient NBPKs.

1. Growth curve of cells cultured at high densities (n = 3 pairs).
2. Results of clonal assays at 14 days after plating. Representative images are shown on the left, and quantification of multiple assays on the right (n = 3 pairs).
3. Distribution of colony size (i.e., number of cells per colony) at 3 days after plating (n = 3 pairs).
4. Heat map of genes differentially expressed (P < 0.05) in control and Ovol2 SSKO NBPKs. Plotted values are log₁₀(FPKM).
5. GO analysis of the top up‐ or downregulated gene sets.
6. GSEA analysis of RNA‐seq data. NES, normalized enrichment score. FDR, false discovery rate.
7. RT–qPCR of the indicated genes normalized to Gapdh (n = 3 pairs).

Data information: For statistical analysis in (A), (B), (C), and (G), we used an unpaired two‐tailed Student's t‐test. Error bars in (A), (B), (C), and (G) represent mean ± SEM.

Figure EV1. Supplemental data on NBPK analysis.

1. Tracking individual colonies starting at 3 days after plating.
2. Fold increase in cell number per colony.
3. Linear regression analysis of RNA‐Seq replicate samples.
4. GSEA analysis of control and Ovol2 SSKO NBPKs with the indicated gene sets.
5. List of enriched/de‐enriched GO terms identified in GSEA.

Data information: Scale bar, 100 μm in (A). For statistical analysis in (B), we used an unpaired two‐tailed Student's t‐test. Error bars in (B) represent mean ± SEM.

To seek molecular insights, RNA sequencing (RNA‐seq) was performed to examine global gene expression differences between control and Ovol2 SSKO NBPKs. This analysis revealed 748 significantly upregulated and 740 significantly downregulated genes (greater than 2‐fold difference, P < 0.05) in Ovol2‐deficient cells (Figs 1D and EV1C; for a complete list of the differentially expressed genes, see Dataset EV1). Gene Ontology (GO) analysis of the differentially expressed genes revealed EMT genes to be the most highly enriched, and cell cycle genes to be the most reduced, in Ovol2 SSKO NBPKs (Fig 1E). Gene Set Enrichment Analysis (GSEA) confirmed enrichment or de‐enrichment of these GO‐identified gene signatures (Figs 1F and EV1D and E). Furthermore, the upregulated expression of select EMT genes, Zeb1 and Vim, as well as the downregulated expression of select cell cycle genes, Ki67 and Cdk1, was validated by RT–qPCR (Fig 1G). Together, these data uncover reduced proliferative expansion and increased EMT tendency as two major defects of Ovol2‐deficient epidermal cells in culture.

Ovol2 is expressed in the proliferative compartments of adult skin epithelia during homeostasis and repair

Toward examining the in vivo function of Ovol2, we determined its expression in adult skin. Indirect immunofluorescence revealed the presence of nuclear Ovol2 protein in IFE basal and some suprabasal cells (Fig 2A). In HFs, nuclear Ovol2 was detected at telogen in cells within the bulge and HG, and at early anagen in the presumptive matrix with the highest expression in cells directly abutting dermal papilla (DP; Fig 2B and C). During excisional wound healing, nuclear Ovol2 protein was abundant in epidermal cells of the highly proliferative zone just outside the wound margin, but completely absent in the migrating front that is known to be devoid of active proliferation 7 (Fig 2D–I). Mixed nuclear (predominantly in suprabasal cells) and cytoplasmic (predominantly in basal cells) signals were detected in the intermediate regions (Fig 2F–H), which is curious and may implicate the possibility of previously undocumented nuclear‐cytoplasmic shuttling of Ovol2, a notion that is outside the scope of this work. Consistent with regional protein distribution, Ovol2 mRNA expression was higher in the wound proliferative zone than the leading edge 7 (Fig 2J). Overall, these data show that Ovol2 protein is present in stem and progenitor cells of the IFE and HF, with expression seemingly correlating with a more proliferative but less migratory cellular state.

Figure 2. Ovol2 expression in normal (A–C) and wounded (D–I) adult skin.

• A–I
  Ovol2 protein expression revealed by indirect immunofluorescence. Enlarged images of the boxed areas in (E) are shown as (F–I) to indicate Ovol2 protein distribution in the intermediate regions (F–H) between the proliferative zone (D) and the migrating front (I). White dashed line in (E) indicates the wound margin. DAPI stains the nuclei.
• J
  RT–qPCR analysis of unwounded skin (control) and microdissected wound regions (n = 3 mice).

Data information: Scale bar, 50 μm in (A–D and F–I); 100 μm in (E). For statistical analysis in (J), we used an unpaired two‐tailed Student's t‐test. Error bars in (J) represent mean ± SEM.

Ovol2 is required for efficient wound repair and promotes directional migration of wound epidermal cells

To investigate the role of Ovol2 in epidermal wound repair, we turned to an excisional wound splinting model 24 to minimize myofibroblast‐driven contraction and to enable measuring wound closure as a direct consequence of re‐epithelialization (Fig EV2A). Compared to littermate controls, Ovol2 SSKO mice showed delayed wound closure during a 7‐day period. By post‐wounding day (PWD) 7, Ovol2 SSKO and control wounds were 30 and 15%, respectively, of the original wound area (Figs 3A and EV2B). Interestingly, at the histological level on PWD 3, the migrating front in Ovol2 SSKO wounds had a significantly longer contour and showed apparently abnormal cellular adhesions compared to control counterparts (Fig 3B and C). When epidermal cells were genetically marked (mediated by K14‐Cre) for β‐galactosidase (β‐gal) or GFP reporter expression to better visualize the neo‐epidermis, it became obvious that the leading edges in Ovol2 SSKO mice had multiple spikes, rough edges, branches, and grooves (Figs 3C and EV2C). Moreover, some Ovol2‐deficient epidermal cells appeared to have disseminated away from the main body of the leading edges (Fig 3C).

Figure EV2. Supplemental data on wound healing analysis.

1. Diagram of wound splint model and measurement strategy.
2. H/E analysis at 7 days after wounding. Control wounds had completed re‐epithelialization and the wound bed had become flush with the epidermis, whereas Ovol2 SSKO mice had the distinctive “U” shape morphology suggesting a delay in the overall wound healing process.
3. Confocal Z‐stack of 30‐μm sections from control (Ovol2 ^f/+;K14‐Cre;ROSA ^mTmG) and Ovol2 SSKO (Ovol2 ^f/−;K14‐Cre;ROSA ^mTmG) mice at 3 days after wounding.
4. Comparison of the distribution of turning angles between steps for cells in control and Ovol2 SSKO wound explants.

Data information: Scale bar, 500 μm in (B), 100 μm in (C).

Figure 3. In vivo and ex vivo evidence for skin wound healing defects in Ovol2 SSKO mice.

• A
  Percent of original wound area over 7 days after wounding (n = 3 pairs; *P < 0.05; **P < 0.005).
• B
  Length of the neo‐epidermis (outer contour of the migrating front up to the wound margin) at PWD 3 (n = 3 pairs).
• C
  Morphology of the leading edges in control (Ovol2 ^f/+ ;K14‐Cre;ROSA26R) and Ovol2 SSKO (Ovol2 ^f/− ;K14‐Cre;ROSA26R) wounds. Yellow arrow points to disseminating cells.
• D, E
  Proliferation analysis by Ki67 immunostaining. Percent Ki67⁺ cells (E) were calculated as the number of Ki67⁺ cells within the neo‐epidermis over the total number of DAPI‐stained nuclei (n = 3 pairs).
• F
  Representative wound explants stained with crystal violet.
• G
  Quantification of the outgrowth distance of multiple explants from (F) (n = 3 pairs).
• H
  Movement tracks of individual cells in control and Ovol2 SSKO explants over 18 h of live imaging.
• I–L
  Quantitative analysis of total distance traveled (I), directionality (J), full step length (K), and straightness distribution (L) of multiple cells in control and SSKO explants (n = 3 pairs).

Data information: Scale bar, 100 μm in (C) and (D), 1,000 μm in (F). For statistical analysis in (A), (B), (E), (G), (I), and (J), we used an unpaired two‐tailed Student's t‐test. For statistical analysis in (K) and (L), we used Kolmogorov–Smirnov test with P = 0.005 and P = 0.0001, respectively. Error bars in (A), (B), (E), (G), (I), and (J) represent mean ± SEM.

Next, we compared cell proliferation in control and Ovol2 SSKO wounds. Consistent with previous reports 7, 25, epidermal cells distal to the wound in control mice at PWD 3 exhibited a higher proliferative activity than epidermal cells proximal to the wound (Fig 3D). Compared to control mice, significantly reduced numbers of Ki67‐positive cells were observed in both wound‐distal and wound‐proximal regions of the Ovol2 SSKO mice (Fig 3D and E).

Wound explant outgrowth is an established method to measure ex vivo wound epidermal cell migration 26. Analysis of explant cultures of control and Ovol2 SSKO wounds revealed decreased overall outgrowth in the latter (Fig 3F and G). Live cell imaging analysis shows that even though individual cells from Ovol2‐deficient explants migrated greater distances within a given time period (18 h), their directionality was reduced as compared to the controls (Fig 3H–J, Movies EV1 and EV2). When the full step length and straightness distributions of the migrating cells were plotted, it was clear that Ovol2‐deficient keratinocytes took larger and more variable steps (Fig 3K), but with less straight paths (Fig 3L). Examination of the distribution of turning angles between steps corroborated the loss of persistence of Ovol2‐deficient cells (much lower density near 0 than for control cells; Fig EV2D). Decreased proliferation and aberrant migration of epidermal cells thus underlie the wound healing defect in Ovol2 SSKO mice.

Loss of Ovol2 compromises HF regeneration and Bu‐HFSC expansion

Data in the preceding sections demonstrate a functional involvement of Ovol2 in regulating IFE cell proliferation, adhesion, and migration under culturing and wound healing conditions. Previously, we reported that Ovol2 overexpression results in hair loss and HF defects characterized by precocious telogen‐to‐anagen progression 21, prompting us to examine a physiological role of Ovol2 in HF regeneration during adult hair cycle. Indeed, a mild but reproducible delay in the progression of the first postnatal anagen was observed in Ovol2 SSKO mice (Figs 4A–C and EV3A). Although Ki67‐positive HG cells were observed in Ovol2 SSKO mice at P22, the presumptive matrix structures were less developed in the new mutant HFs than control counterparts (Fig 4D). Moreover, anagen HFs were generally less elongated in Ovol2 SSKO skin, their bulbs were smaller, and their differentiated lineages were poorly represented compared to control counterparts (Figs 4A–C and EV3B and C). Even in residual mutant HFs with lengths comparable to control HFs, hair keratin expression was reduced as revealed by AE13 antibody staining (Fig 4E). It appears as if the outer root sheath cells are migrating downwards to extend the HF, but the matrix cells are not efficiently dividing and migrating upwards to produce the differentiated progenies. Despite these defects, HFs in Ovol2 SSKO mice were able to reach second telogen by P49 just as in the WT (Fig 4A) and were able to produce new hair shafts (Fig EV3D). There was no hair loss phenotype, and the epidermal thickness was not affected throughout these stages examined (Fig EV3E). Taken together, our data demonstrate a functional requirement for Ovol2 in the timely downgrowth of, and subsequent differentiation within, postnatal HFs.

Figure 4. In vivo and ex vivo evidence for Ovol2 loss‐induced HF defects.

• A
  H/E analysis of control and Ovol2 SSKO skin at the indicated ages.
• B, C
  Fold difference in HF length (B) and bulb width (C) between control and Ovol2 SSKO mice at the ages of P22‐P26 (n = 5 pairs).
• D
  Ki67 immunostaining in control and Ovol2 SSKO skin. K15 stains the bulge/HG cells.
• E
  AE13 immunostaining. Arrows and “*” indicate specific and background signals, respectively.
• F
  RNA expression of the indicated genes normalized to Gapdh in freshly sorted P23‐Bu‐HFSCs (n = 3 pairs).
• G
  Clonal analysis of control and Ovol2 SSKO Bu‐HFSCs.
• H, I
  Number (H) and average size (I) of Bu‐HFSC colonies from (G) (n = 3 pairs).
• J
  Representative images of patch assay results using control and Ovol2 SSKO epidermal cells along with dermal‐only control.
• K
  Quantification of number of HFs (n = 3 pairs; each biological replicate is an average of three injections).
• M
  K14 immunostaining of skin from the injection site at 3 days after injection. Arrow points to epidermal spheres.
• N
  Quantification of epidermal sphere diameter (three independent experiments using three pairs of control and Ovol2 SSKO mice) at 3 days after injection.

Data information: Scale bar, 50 μm in (A), (D), (E), and (L), 250 μm in (J). For statistical analysis in (B), (C), (H), (I), (K), and (M), we used an unpaired two‐tailed Student's t‐test. For statistical analysis in (F), we used a paired two‐tailed Student's t‐test. Error bars in (B), (C), (F), (H), (I), (K), and (M) represent mean ± SEM.

Figure EV3. Supplemental data on HF analysis.

• A
  Whole‐mount back skin images of same‐sex littermates at different stages of the hair cycle. The particular underside region that is being examined in each row was indicated on the mouse diagrams on the left.
• B, C
  Fold difference in HF length (B) and bulb width (C) between control and Ovol2 SSKO mice at the indicated ages (n = 1 pair at P22, 3 pairs at P25, and 1 pair at P26).
• D
  H/E analysis of P49 Ovol2 SSKO skin indicating new hair shaft formation. Arrows indicate old and new hair shafts from the same follicle.
• E
  Epidermal thickness of skin of the indicated genotypes (n = 6 pairs: 2 at P21, 2 at P25, and 2 at P49).
• F, G
  Percentage of CD49f^HiCD34⁺ Bu‐HFSC at P23 (F) and P49 (G).

Data information: Scale bar, 100 μm in (D). For statistical analysis in (B), (C), (E), (F), and (G), we used an unpaired two‐tailed Student's t‐test. Error bars in (B), (C), (E), (F), and (G) represent mean ± SEM.

Next, we asked if Bu‐HFSCs are affected by Ovol2 loss. The percentage of CD49f^HiCD34⁺ Bu‐HFSCs of total epithelial cells (CD49f⁺) was similar between control and Ovol2 SSKO mice at P23, but was significantly reduced at P49 (Fig EV3F and G) 27. However, even at P23, gene expression differences were already evident, as freshly sorted SSKO Bu‐HFSCs exhibited upregulated (more than 10‐fold) expression of Zeb1 but downregulated expression of Ki67 and Cdk1 compared to controls (Fig 4F). Subsequent functional characterizations focused on Bu‐HFSCs isolated at P49, when control and SSKO HFs were both in telogen. When plated in culture at a clonal density, control Bu‐HFSCs were able to produce colonies that continuously expanded over a period of 2 weeks. In sharp contrast, Ovol2‐deficient Bu‐HFSCs, while able to form small colonies initially, failed to sustain clonal growth leading to dramatically reduced number and size of the colonies at the end of culture (Figs 4G–I and EV4A). Furthermore, the few mutant colonies that reached large sizes exhibited abnormal morphology with enlarged intercellular gaps, increased levels of Zeb1 and Vim mRNAs but reduced level of Cdk1 mRNA, and a higher frequency of nuclear Zeb1‐positive cells (Fig EV4B–E). Collectively, these data show that Ovol2 is required for maintaining the epithelial state and active expansion of Bu‐HFSCs.

Figure EV4. Supplemental data on Bu‐HFSC analysis.

1. Tracking individual Bu‐HFSC colonies from day 7 to day 10 in culture. Yellow dashed lines indicate colony outline at the indicated times, whereas red dashed lines indicate colony outline at day 7.
2. High magnification images of single Bu‐HFSC colonies at 14 days after plating.
3. Expression analysis of the indicated genes in cultured Bu‐HFSCs (n = 3 pairs).
4. Zeb1 immunostaining in control and Ovol2 SSKO Bu‐HFSCs.
5. Quantification of percent Zeb1⁺ cells per total cells in each colony (n = 2 pairs).
6. Cell cycle analysis of P49 Bu‐HFSCs freshly sorted from control and Ovol2 SSKO littermates.

Data information: Scale bar, 100 μm in (A), (B), and (D). For statistical analysis in (C), we used an unpaired two‐tailed Student's t‐test. Error bars in (C) represent mean ± SEM.

To investigate the in vivo consequence of defective clonal adhesion and expansion of Ovol2‐deficient cells, we turned to “patch” assay, a well‐established transplantation model to examine the hair/HF regenerative capacity of isolated skin epithelial cells 28. Control or Ovol2 SSKO epidermal cells in single‐cell suspension were combined with wild‐type newborn dermal cells and subcutaneously injected into the backs of Nu/J mice. Two weeks later, significantly fewer hairs/HFs were observed in patches derived from Ovol2 SSKO epidermal cells than the littermate control cells (Fig 4J and K). Interestingly, at earlier time points when injected epidermal cells first aggregated and began to proliferate 28, the K14⁺ spheres formed by Ovol2 SSKO cells were significantly smaller than those formed by control cells (Fig 4L and M). These data demonstrate that in the absence of Ovol2, adult epidermal cells have a reduced capacity to regenerate HFs, which appears to stem from early defects in epithelial sphere formation.

Live cell imaging reveals an intimate link between aberrant migration and defective cell cycle progression of Ovol2‐deficient Bu‐HFSCs

To probe the potential mechanism of Ovol2 deficiency‐induced Bu‐HFSC clonal expansion defects, we utilized live cell imaging to track individual colonies over time. Freshly sorted control Bu‐HFSCs generated colonies after 7 days in culture and cells within these colonies exhibited a remarkably rapid doubling rate leading to a > 2‐fold increase in cell number over an 18‐h period, whereas the doubling rate of Ovol2 SSKO cells was significantly lower (Fig 5A, Movies EV3 and EV4). More than 80% of the dividing cells in control colonies rounded up slightly and transiently when undergoing mitosis—a phenomenon previously identified as mitotic cell rounding 29, but they quickly reattached in situ after generating two daughter cells (Fig 5B and C; type‐1 division). Less than 20% of the dividing cells, particularly those that reside in the periphery of the colonies, rose up significantly above the plane of the neighboring cells (Fig 5B and C; type‐2 division). In stark contrast, the relative numbers of rounded‐up cells and type‐2 divisions were significantly higher for Ovol2‐deficient cells (Figs 5B and C, and EV4A). Furthermore, these dividing mutant cells often migrated away from their original positions, failed to reattach, and the two daughter cell nuclei remained connected after an extensive period of time (Fig 5C).

Figure 5. Ovol2‐deficient Bu‐HFSCs display aberrant cell division behavior and arrest in G2/M‐>G1 transition.

• A
  Fold change in cell number per colony in an 18‐h period (n = 3 pairs; each biological replicate is an average of three different colonies).
• B, C
  Live imaging reveals an increased frequency of type‐II divisions in SSKO culture, as quantified in (B) (n = 3 pairs; each biological replicate is an average of three different colonies); shown in (C) are representative images from a single experiment. Red and blue arrows in (C) mark two individual cells and their division products during time‐lapse.
• D
  Cell cycle analysis of cultured Bu‐HFSCs from control and Ovol2 SSKO mice (n = 3 pairs).

Data information: Scale bar, 100 μm in (C). For statistical analysis in (A), (B), and (D), we used an unpaired two‐tailed Student's t‐test. Error bars in (A), (B), and (D) represent mean ± SEM.

The prolonged connection of daughter cell nuclei in Ovol2‐deficient colonies led us to wonder whether without Ovol2 cell cycle is arrested post‐mitosis. Indeed, flow cytometry‐mediated cell cycle analysis of actively growing Bu‐HFSCs revealed a significantly increased number of G2/M cells accompanied by a decreased number of G1 cells in Ovol2 SSKO culture, whereas the percent of S‐phase cells was unaffected (Fig 5D). As expected 12, flow cytometric analysis of freshly isolated Bu‐HFSCs from telogen back skin revealed few cycling (G2/M/S phase) cells, and no difference was detected between control and Ovol2‐deficient cells (Fig EV4F). These data suggest that during active expansion, Ovol2‐deficient Bu‐HFSCs cannot efficiently transition from G2/M to G1 phase of the cell cycle.

Ovol2 promotes directional migration of Bu‐HFSCs through suppressing Zeb1 expression

The ability to track Bu‐HFSC colonies over time enabled us to monitor the migratory behavior of single cells as colonies are expanding. While control Bu‐HFSCs migrated outwards with relatively straight trajectories, Ovol2 SSKO cells followed less straight/persistent paths (Fig 6A, Movies EV3 and EV4). On average, mutant cells migrated greater overall distances with greater velocities than control cells, but the directionality of migration was severely compromised (Fig 6B). Moreover, Ovol2‐deficient cells took larger and more variable steps (Fig 6C), moved in less straight patterns (Fig 6D), and exhibited a decreased frequency of turning angles near 0 indicative of loss of persistence (Fig EV5A). In keeping with this aberrant migratory behavior, phalloidin staining revealed a higher prevalence of stress fiber type of actin network in Ovol2 SSKO colonies than in controls, which displayed predominantly cortical type of actin organization (Fig 6E and F). Collectively, these data indicate that Ovol2 is required for directional migration of Bu‐HFSCs ex vivo.

Figure 6. Compromised directional migration of Ovol2‐deficient Bu‐HFSCs is rescued by Zeb1 deletion.

• A
  Representative movement tracks of individual cells over an 18‐h period.
• B–I
  (B, G) Quantitative analysis of migration distance, velocity, and directionality of Bu‐HFSCs with the indicated genotypes (n = 3 pairs; each biological replicate is an average of three different colonies with eight cells tracked per colony). Ovol2 KO, Ovol2‐deficient; DKO, Ovol2‐ and Zeb1‐deficient. (C, H) Step‐length comparison among the indicated genotypes. Kolmogorov–Smirnov test: CTL vs. SSKO (C) or Ovol2 KO (H), P < 10⁻¹⁰; CTL vs. DKO (H), P = 0.06. (D, I) Straightness distribution comparison among the indicated genotypes. Kolmogorov–Smirnov test: Control vs. SSKO (D) or Ovol2 KO (I), P < 10⁻¹⁰, P < 10⁻⁵; Control vs. DKO (I), P = 0.45. (E) Representative plots that show distribution of fluorescence intensity of actin staining in control and Ovol2 SSKO Bu‐HFSCs. E1 (Edge‐1) and E2 (Edge‐2) signals correspond to cortical actin near the cell border, whereas center signals correspond to stress fiber actin. (F) Quantification of the ratio between E1 + 2 and center signals (E) in control and Ovol2 SSKO mice (n = 2 pairs).

Data information: For statistical analysis in (B) and (G), we used an unpaired two‐tailed Student's t‐test. Error bars in (B), (F), and (G) represent mean ± SEM.

Figure EV5. Supplemental data on Bu‐HFSC migration.

• A
  Comparison of the distribution of turning angles between steps for control and Ovol2 SSKO Bu‐HFSCs.
• B
  Diagram of strategy for adenoviral infection of cultured Bu‐HFSCs.
• C, D
  GFP expression in infected Bu‐HFSCs analyzed by epifluorescence (C) or flow cytometry (D; 2A‐CRE‐infected).
• E
  RT–qPCR analysis confirming decreased expression of Ovol2 and/or Zeb1 at 2 weeks after 2A‐CRE infection of Bu‐HFSCs (n = 2 pairs).
• F
  Fold increase in cell number in control, Ovol2 KO, and DKO Bu‐HFSC cultures as measured by live cell imaging over an 18‐h period (n = 2 pairs).
• G
  Comparison of the distribution of turning angles between steps for control, Ovol2 KO, and DKO Bu‐HFSCs.

Data information: Scale bar, 100 μm in (C). Error bars in (E) and (F) represent mean ± SEM.

The upregulation of Zeb1 expression in the absence of Ovol2 and a known role for EMT in conferring cell motility 23 led us to ask whether loss of Zeb1 might normalize the aberrant migratory behavior of Ovol2‐deficient Bu‐HFSCs. To do this, we generated mice containing floxed alleles of both Ovol2 and Zeb1 and acutely deleted the two genes, singly or in combination, in Bu‐HFSCs using Cre‐expressing adenoviruses (Ade‐Cre; Fig EV5B–E). Distinct from Ovol2 SSKO Bu‐HFSCs where Ovol2 deficiency is chronic, acute Ovol2 deletion did not result in any detectable difference in Bu‐HFSC expansion (Fig EV5F). Live cell imaging revealed decreased directionality/persistence in the migration of Bu‐HFSCs with acute Ovol2 deletion similar to that of Ovol2 SSKO cells, and that this defect was near‐completely rescued by the simultaneous deletion of Zeb1 (Figs 6G–I and EV5G). These data provide strong evidence for a key role of the Ovol2‐Zeb1 regulatory axis in controlling the migratory behavior of adult Bu‐HFSCs.

Discussion

While our previous work describes the importance of Ovol1 and Ovol2 TFs in epidermal morphogenesis 21, the current study highlights Ovol2 as a crucial player in skin epithelial repair and regeneration in adult mice. Most notably, our findings underscore a key role for the Ovol2‐Zeb1 EMT‐regulatory circuit in modulating the migratory behavior of skin epithelial cells, specifically by restricting migration speed and conferring optimal directionality.

To date, involvement of a classical EMT TF in mammalian cutaneous wound healing has only been shown for Slug (Snai2), the loss of which leads to compromised migration of epidermal cells at the wound leading edge 9, 30. Very little was known about the molecular mechanisms that keep migrating epidermal cells in check so that they are able to maintain or resume epithelial traits. We now find that Ovol2, likely by virtue of its ability to inhibit the expression of EMT‐inducing Zeb1, enables wound epidermal cells to migrate slower but with improved directionality and persistence. Thus, counteracting molecular pathways that normally promote and restrict EMT are both active during epidermal wound healing. Conceivably, EMT‐promoting TFs such as Slug and possibly Zeb1 can temporarily and mildly relax epithelial rigidity to initiate and facilitate the migration of epidermal cells around the wound, whereas EMT‐inhibitory TFs such as Ovol2 restrict this form of epithelial plasticity to ensure that migration is collective and directional toward the wound center for efficient re‐epithelialization to occur.

The molecular regulation of epithelial cell migration during HF regeneration remains an even less charted area. Our data reveal for the first time that EMT‐regulatory TFs play critical roles in controlling the migratory behavior of Bu‐HFSCs and that HF regeneration is inefficient when such mechanisms go awry. Compared to adult IFE stem cells during normal tissue homeostasis, Bu‐HFSCs and their early progenies have to travel much further as they embark on the journey to differentiate into specialized cell types during HF regeneration, much like wound margin epidermal cells during re‐epithelialization. It is tempting to speculate that epithelial plasticity, specifically partial, reversible EMT‐like changes in cell adhesion and cytoskeleton, is an integral part of Bu‐HFSC/progeny migration that drives the formation of a new HF. Indeed, as the HF extends downwards, the nuclei of HG cells are more separated from each other than those in the bulge, and they realign as the HG cells reach the epithelial‐DP interface 13, implicating dynamic but reversible changes in cell–cell associations. Additional work such as intravital imaging of the Ovol2 mutant skin is needed to determine whether and how such dynamic cellular events are affected in vivo.

Interestingly, coupled to aberrant cell migration, Ovol2‐deficient IFE cells, HF matrix cells, and Bu‐HFSCs also display significantly reduced proliferation potential. In particular, the aberrant migration and inefficient reattachment of post‐mitotic Ovol2‐deficient Bu‐HFSCs along with an inability to transition into the next cell cycle (G2/M‐>G1) provide correlative evidence for a potential mechanistic link between aberrant migration/adhesion and cell proliferation. However, the lack of a proliferation defect in Bu‐HFSCs following acute Ovol2 deletion precluded us from using our established assay to ask whether simultaneous deletion of Zeb1 is able to normalize the cell cycle arrest. Along a similar line, we note that while previous studies identified distinct proliferative and migratory epidermal zones in the healing wound, a recent study found a third epidermal region where migration and proliferation co‐exist and where tissue expansion peaks 7. Intriguingly, here migration regulates the directionality of the cell division plane, providing yet another case of coordinated control of epidermal cell migration and proliferation. Molecular and signaling pathways are known to regulate both migration and proliferation during wound healing 8. This said, the distinct mode of regulation by Ovol2 vs. known examples such as in the case of BMP signaling 31 is worth noting. It is also important to note that Bu‐HFSCs and epidermal cells are two distinct cell types with unique identities and functions; thus, future work is needed to elucidate the potential differences and similarities in mechanism of Ovol2 function in these two cell types in in vivo settings.

In Ovol2 SSKO mice, K14‐Cre‐mediated Ovol2 deletion occurs during embryogenesis 21. It is possible that some Ovol2 deficiency‐induced cellular aberrancies take longer time to develop than allowed by the Ade‐Cre acute deletion system. A context‐dependent notion receives further support from our finding that the migratory and proliferative defects of Ovol2‐deficient IFE cells and Bu‐HFSCs appeared much more pronounced when cells were experiencing non‐physiological and/or stressful conditions, such as in culture, upon transplantation, or during wound repair. Perhaps under such conditions, microenvironmental cues such as elevated growth factor concentration and immune cell infiltration cause greater epithelial plasticity, thereby creating a higher demand for molecular mechanisms that maintain epithelial traits such as cell–cell and cell–matrix associations during migration.

Lack of keratinocyte migration is observed in chronic wounds such as those in diabetic patients and is apparently associated with epidermal hyperproliferation 6. Our animal model studies raise the intriguing possibility that aberrantly increased cell migration at the cost of directionality can be accompanied by epidermal hypoproliferation and together they may underlie some non‐healing wounds in human patients. As the current treatment strategies emphasize the promotion of epidermal migration 6, caution should be exercised to create an optimal wound healing tissue microenvironment that enables migration but in a persistent and directional manner.

Materials and Methods

Mice

K14‐Cre transgenic mice, floxed (f) and null (−) alleles of Ovol2, as well as floxed (f) allele of Zeb1 have been previously described 32, 33, 34. ROSA ^mTmG, ROSA26R, and Nu/J mice are from the Jackson Laboratory (Stock #s 007576, 003474, 002019, respectively). In all mutant analyses, same‐sex control littermates were used for comparison. All experiments have been approved and abide by regulatory guidelines of the International Animal Care and Use Committee (IACUC) of the University of California, Irvine.

Morphology and immunostaining

For histological analysis, mouse back skin was shaved, removed, fixed in 4% paraformaldehyde (MP; 150146) in 1× PBS, embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H/E). HF stage identification was based on 35. For whole‐mount analysis, mice were shaved, their skin collected and fat carefully removed, and pigmented HFs were visualized from the dermal side. For indirect immunofluorescence, mouse back skin was freshly frozen in OCT (Fisher; 4585), sectioned at 5–8 μm, and staining was performed using the appropriate antibodies and DAPI (Thermo Fisher; D1306: 1:1,000). The following primary antibodies were used for immunofluorescence: Ovol2 36; (rabbit, 1:100), K14 (chicken, 1:1,000; rabbit, 1:1,000; gift of Julie Segre, National Institutes of Health, Bethesda), K15 (Covance, PCK‐153P‐100, 1:1,000), Ki67 (Cell Signaling, D3B5, 1:1,000), and AE13 (Abcam, ab16113, 1:200). For β‐galactosidase histochemistry, mouse back skin wounds were fixed with 0.5% glutaraldehyde (Sigma; G6257), stained overnight at 37°C with X‐gal (Denville; CX3000‐3), and counterstained with Nuclear Fast Red (VWR; 1B1369). For epidermal sphere analysis in patch assays, injection sites were isolated and fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned. Antigen retrieval was performed by incubating slides in 0.01 M citrate buffer (pH 6.0) in microwave at full power for 3–5 min. For ROSA ^mTmG analysis, mouse back skin wounds were freshly frozen in OCT, sectioned, and then stained with DAPI. For actin staining of cultured Bu‐HFSCs, cells were fixed on plate in 4% paraformaldehyde and subsequently stained for K14 and actin (Phalloidin, Invitrogen A12379, 1:250). Thin section images were taken with an inverted fluorescence microscope (Eclipse E600; Nikon) using the Plan‐Fluor 10X DIC L 0.30, Plan‐Apo‐chromat 20× N.A. 0.75, or Plan‐Fluor 40× N.A. 0.75 objectives (Nikon) and a camera (RT Slider; Diagnostic Instruments) equipped with SPOT 4.0.9 software (Diagnostic Instruments). Thick section images were taken at room temperature with the Zeiss LSM700 confocal microscope using EC Plan‐Neofluar 10×/0.30 objective. Analysis of Phalloidin staining intensity was done using the Plot Profile tool in Fiji.

Wound healing

Mice were anesthetized using isoflurane (Primal Healthcare; NDC‐66794‐017‐25), backs shaved, and then 4‐mm punch (Integra; 33–34) was used to generate a full‐thickness wound on each side of the mouse. For re‐epithelialization measurements, splints were fastened surrounding the punch. To generate splints, a 5‐mm punch biopsy (Integra; 33–35) was used to generate a hole in 1.5‐cm‐diameter circular piece of silicone (Life; P18178) and then stitched over the wound. Wound diameter was measured every 24 h. For expression analysis, 6‐mm punch biopsies (Integra; 33–36) were made, and then, a 12‐mm‐diameter portion was collected 4 days after wounding. An 8‐mm‐diameter portion was removed from the center and used as leading edge sample, while the remaining was used as the proliferative zone sample. For ex vivo explant assay, mouse back skin was shaved, removed, fat was scrapped away; remaining hair was removed with brief treatment with Nair and then thoroughly rinsed with 1× PBS. A 4‐mm punch biopsy was collected as wound explants, which were then placed onto a well‐containing 6 μl Matrigel (BD; 354230) in a 6‐well tissue culture dish that was pre‐coated with 10 μM fibronectin (EMD Millipore; FC010) for 1 h.

Isolation, culture, and/or infection of NBPKs and Bu‐HFSCs

NBPKs were isolated as reported 37 with minor modifications. Briefly, epidermal and dermal separation was achieved with an overnight dispase treatment (Stem Cell Technologies; 07913). Keratinocytes were then isolated from separated epidermis using 0.25% trypsin (Sigma; T4799). Dermal cells were isolated from separated dermis using a 0.25% collagenase (Sigma; C9091) digestion at 37°C for 2 h. For growth curve and clonal assays, freshly isolated NBPKs were cultured on tissue culture plastics without fibroblast feeders in low‐Ca²⁺ keratinocyte E‐media. For gene expression analysis, NBPKs were cultured on mitotically inactivated J2‐3T3 fibroblasts in low‐Ca²⁺ keratinocyte E‐media. Detailed instructions for J2‐3T3 culture and mitotic inactivation can be found in 38. For growth curve analysis, 10,000 NBPKs were plated per well onto 12‐well plates in triplicate. Every 24 h, cells were harvested using 0.1% trypsin and then counted. For clonal growth analysis, NBPKs were plated at a concentration of 1,000 cells/cm² onto 6‐cm gridded plates. Two weeks later, a 0.5% crystal violet (Sigma; HT90132) solution made in 1:1 water:methanol was added to the plates to fix and stain the cells. Plates were then rinsed with DI water and imaged.

Isolation and subsequent culture of Bu‐HFSCs were as reported 38 with minor modifications. Briefly, back skin was cut into ~8 equal pieces and spread dermis side down in 0.25% trypsin and incubated at 37°C for 60 min. Subsequently, epidermis was gently scraped off and then mixed with low‐Ca²⁺ keratinocyte E‐media, filtered with 70‐ and 40‐μm filters, spun down, and resuspended in low‐Ca²⁺ keratinocyte E‐media. Adult keratinocytes were stained for CD49f‐PE (BD; 555736), CD34‐Alexa700 (BD, 560518; clone RAM34), and 7AAD (BD; 559925), followed by sorting using a BD FACSAria™ Fusion Sorter. Adenoviral infection of Bu‐HFSCs was based on 39, using Ade‐Cre viruses at a multiplicity of infection of 50 (Vector Biolabs; 1772).

Cell cycle analysis

Cell cycle analysis was performed on Bu‐HFSCs after 2 weeks of culture. Cells (Bu‐HFSCs and feeders) were removed from plate using 0.1% trypsin and then fixed with cold 70% EtOH for 30 min at 4°C. Cells were washed twice in 1× PBS and then stained with CD49f‐FITC (BD; 555735) for 30 min at room temperature in the dark. After two washes in 1× PBS, cells were stained with FxCycle™ PI/RNase Staining Solution (Life; F10797) for 15 min and then immediately analyzed using ACEA NovoCyte™ Flow Cytometer.

Patch assay

Patch assay was performed according to published procedures 28, 40. Briefly, total adult epidermal cells (500,000) from 7‐ to 8‐week‐old mice were combined with dermal cells (1,000,000) freshly isolated from newborn C57BL/6J mice. The mixture was subcutaneously injected into 7‐week‐old Nu/J mice, which were examined 2 weeks later.

Live cell imaging

Wound explants and cells were cultured on either 6‐cm or 6‐well plates and incubated within a microscope chamber at 37°C with 5% CO₂. Imaging was performed using the Keyence BZ‐X700 microscope for 18‐h periods with images taken every 15 min. Images were then exported and analyzed in FIJI using the Manual Tracking plugin. Each initial cell position was set to coordinates of (0,0). Step lengths were calculated for each step for each cell track, and turning angles were calculated based on pairs of cell track segments, i.e., ((x_t−1, x_t), (x_t, x_t+1)). The straightness was defined as D_w/L_w where D_w is the total displacement for a given number of steps (window size), and L_w is the total path length in this window. A window size of 70 steps was used.

RNA isolation, quantitative RT–PCR, and RNA‐Seq

For RNA isolation, Trizol (Life; 15596018) was used as per manufacturer's instructions. One microgram of RNA was used to generate cDNA (Applied Biosystems; 4368814) as per manufacturer's instructions. qPCR was performed using a Bio‐Rad CFX96 Real‐Time System and SsoAdvanced Universal SYBR^® Green Supermix (Bio‐Rad; 172‐5271). Primers sequences used for expression analysis are as follows: Ovol2 F: AGCTTCACGACGCCCAAGGC; Ovol2 R: GCCGCAGAAGGTGCACAGGT; Zeb1 F: ACCGCCGTCATTTATCCTGAG; Zeb1 R: CATCTGGTGTTCCGTTTTCATCA; Vim F: GGAGATGCTCCAGAGAGAGG; Vim R: ATTCCACTTTCCGTTCAAGG; Ki67 F: CATCAGCCCATGATTTTGCAAC; Ki67 R: CTGCGAAGAGAGCATCCATC; Cdk1 F: TTCCACGGCGACTCAGAGAT; Cdk1 R: AGCAAATCCAAGCCGTTCTC; Gapdh F: CCTGCCAAGTATGATGAC; Gapdh R: GGAGTTGCTGTTGAAGTC. For RNA‐Seq, optimal‐quality RNAs (RNA integrity numbers > 9) were used for cDNA library preparation. Full‐length cDNA library amplification was performed as previously described 41, 42. Briefly, 1 ng of total RNA was reverse‐transcribed and the resulting cDNA was preamplified for 17 cycles. Tagmentation of cDNA was carried out using the Nextera DNA Sample Preparation Kit (Illumina; FC‐121‐1031). The Tn6 tagmentation reaction was carried out at 55°C for 5 min and purified using a PCR Purification Kit (Qiagen; 28104). Adapter‐ligated fragments were amplified using limited cycle enrichment PCR with Nextera barcodes for seven continuous cycles. The resulting libraries were purified using AMPure XP beads (Beckman Coulter; A63880) and were multiplexed and sequenced as paired end on a HiSeq 2500 Illumina sequencing platform. Sequencing reads were mapped to the mm9 mouse genome using BowTie2 43, and splice junctions between exons were mapped using Tophat2 44. Analysis of differential gene expression was accomplished using CuffDiff 45. Differentially expressed genes were used for GO analysis. Gene FPKM values were used for downstream analysis that included GSEA 46.

Data availability

Gene expression data are provided in the Excel files and have been deposited to GEO (GSE118915). https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118915

Author contributions

XD conceived the study, and DH and XD designed the project and experiments; DH, PS, ALM, XM, and YZ performed the experiments; MVP assisted in patch assay; ALM and QN assisted in cell tracking analysis; MPS, SB, GB, and TB provided animals with a Zeb1 flox allele; DH and XD interpreted the data; DH, ALM, and XD wrote the manuscript; DH, ALM, MVP, QN, TB, and XD edited the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Dataset EV1

Review Process File

EMBO Reports (2019) 20: e46273