Stem Cells of the Skin and Their Lineages
Adult stem cells reside in most tissues of our body. Although more restricted than embryonic stem cells in the types of tissues they make, these stem cells survive long-term, and regenerate both themselves and their resident tissues during normal wear and tear (homeostasis) and following injury. Tissue stem cells typically reside in specialized microenvironments (niches) and are used sparingly, but can be mobilized quickly and expand to repair wounds when the tissue is damaged. During normal homeostasis, tissue stem cells often exist in two distinct states: a quiescent state in which they are not making tissue and a primed state in which they are either actively making tissue or more readily activated when needed. Once activated, tissue stem cells produce shorter-lived progenitors (sometimes called ‘transit-amplifying cells’) that divide rapidly several times, but then progress to terminally differentiate to form their respective tissue.
My laboratory uses mammalian skin as a model system to study adult stem cells. At the surface of our body, skin epidermis is readily accessible, and its stem cells can be cultured and passaged long-term without losing their stemness. Shortly after gastrulation, embryonic skin begins as a single layer of multipotent epithelial cells. These cells will give rise to the epidermis, and its appendages, which include hair follicles, sebaceous glands, sweat glands and even mammary glands. In the adult, each of these epithelial tissues has its own niche(s) of stem cells, presenting a rich and fascinating diversity for exploring the molecular mechanisms underlying the “fountain of youth” properties of stem cells and their remarkable ability to make and repair tissues.
The mouse hair follicle offers a particularly attractive model for dissecting the mechanisms involved. Like most mammals, mice have a thick hair coat. Each hair follicle has a resident niche (the bulge) of stem cells, responsible for regenerating the follicle below the bulge and producing the hair and its channel. Across the hair coat, follicles undergo synchronized bouts of activity. Thus, the stem cells can rest in quiescence for weeks, even months, adjacent to a specialized mesenchymal cell stimuli, the dermal papilla. During this time, epithelial-mesenchymal crosstalk at the base of the bulge stem cell niche (“the hair germ”) accumulates stimulating factors, and when the threshold is reached, these primed stem cells in the hair germ launch a new hair cycle (Tumbar et al., 2004). We now know that this initial activation step arises from a combination of WNT signaling and inhibition of BMP signaling (Greco et al., 2009).
During skin embryogenesis, WNT signaling is also essential to form an epithelial bud, which in certain respects, resembles the hair germ, but at its earliest stage, could be a hair bud, tooth bud, mammary bud or sweat gland bud. Recently, we learned that BMP signaling plays a critical role in bud fate decision. Specifically, we discovered that in contrast to making a hair follicle, which like the adult hair germ, requires blocking BMP signaling within the bud, a sweat gland requires the opposite, namely high BMP signaling in the bud (Lu et al., 2016). In mice and most mammals, eccrine sweat glands only form on the paws, whereas hair follicles occur elsewhere on body skin. By contrast, humans evolved to have both sweat glands and hair follicles on body skin, enabling them to better control body temperature and withstand temperature extremes. Probing mechanism, we discovered that in humans, body skin begins in a BMP-inhibitory state, causing buds to become hair follicles. Then, a spike of BMP occurs, causing the last bud waves to become glands. Moreover downstream, BMP-inhibition promotes expression of sonic hedgehog (SHH), while BMPs antagonize SHH production, and this antagonism is essential for proper fate specification. Our findings not only solve a fascinating evolutionary conundrum but also have important implications for future therapeutics of patients suffering from sweat gland loss in genetic disorders and/or bad burns.
Interestingly, once activated by WNT signaling and BMP inhibition, the primed stem cells in the adult hair germ generate short-lived progenitors which start to make SHH (sonic hedgehog). By conditionally inactivating the genes encoding either SHH or its receptor, we discovered that SHH signals in two ways: first, it instructs the dermal papilla to make more BMP inhibitors and induce expression of proactivating FGF7; second, it stimulates the otherwise quiescent bulge to restock the stem cell pool and to make the shaft (outer root sheath or ORS) that pushes the SHH-expressing short lived progenitors away from the bulge niche (Hsu et al., 2014). Distanced from the SHH signal, the stem cells return to quiescence until the next hair cycle. By contrast, the short-lived progenitors continue to interact with the dermal papilla and receive the necessary signals to continue to proliferate and differentiate, producing the hair and its channel, or inner root sheath (IRS). As short-lived progenitors exhaust their proliferative capacity, the follicle degenerates, drawing the dermal papilla upward again to the base of the stem cell niche.
Another remarkable feature of bulge stem cells is that they normally only function in hair regeneration, but they possess the potential to also make epidermis and sebaceous glands. We first learned of this plasticity of bulge stem cells when we fluorescently marked and purified the stem cells from their native niche, cultured them and then engrafted the progeny from individual stem cell-derived clones back onto the skin of a hairless mouse. The graft produced hundreds of fluorescent hair follicles, sebaceous glands and epidermis (Blanpain et al., 2004). This experiment not only illustrated the enormous tissue-regenerating potential of these stem cells, but also showed that these stem cells possess multipotency. It also tells us that the niche microenvironment imposes restraints upon what these stem cells do.
We observed similar stem cell plasticity when we purified and tested the myoepithelial stem cells from sweat glands (Lu et al., 2012; Blanpain and Fuchs, 2014). Similar to myoepithelial stem cells of mammary glands, these stem cells normally act unipotently and only replenish dying myoepithelial cells of the gland. However, when purified by fluorescence activated cell sorting (FACS) and transplanted directly into a mammary fat pad, the stem cells can regenerate the complete bi-layered gland, and the new luminal cells secrete sweat. Moreover, when engrafted to the skin, these stem cells can make epidermis. An area of interest in my lab is to understand the environmental cues that dictate the fascinating plasticity of epithelial stem cells, and to elucidate the chromatin remodeling that leads to the changes in gene expression necessary to generate different tissues from a common progenitor.
To understand how a stem cell chooses its differentiation pathway, we have taken several approaches. An ongoing approach of the lab is to express different fluorescent proteins under the control of various skin promoters, active at different stages in stem cells and their lineages. Through FACS, we've purified cells at different time points along the lineages and generated a battery of lineage-specific profiles, enabling us to define at an mRNA (RNA-seq) and chromatin (ChIP-seq) level how stem cells change as they transition from quiescence to activation to lineage determination. Our global objective is to exploit this information to understand how stem cells receive signals, change their program of gene expression and select a lineage. We also want to understand the functional significance of these changes. The beauty of the hair follicle as a model is that it is currently the only system where sufficient quantities of stem cells can be isolated directly from their native niche in order to carry out whole-genome wide analyses in vivo. This eliminates the caveats arising from culturing cells, namely induction of a stress response and large-scale epigenetic changes in gene expression.
For the hair follicle, >150 mRNAs are selectively upregulated in the bulge stem cells relative to their short-lived progeny (Tumbar et al., 2004; Blanpain et al., 2004; Keyes et al., 2013). A number of these changes are in transcription factors and epigenetic regulators. We’ve conducted in vivo chromatin immunoprecipitation and high throughput sequencing (ChIP-seq) on chromatin from hair follicle stem cells (HFSCs) and their short-lived progeny (Lien et al., 2011; 2014; Adam et al., 2015). Bioinformatics reveals which genes bind these transcription factors, and how this changes as the stem cells progress to form transiently dividing cells that then terminally differentiate along one of the 7 distinct concentric cell layers that constitute the hair and its channel. By conducting high throughput RNA sequencing (RNA-seq) on HFSCs lacking each of these genes, we’ve learned which target genes depend upon binding these transcription factors. Finally, by engineering inducible-conditional knockouts to selectively remove these transcription factors in the stem cells, we’ve learned the physiological relevance of these factors.
Based upon these analyses, TCF3/TCF4, LHX2 and SOX9 are all essential for maintaining the HFSCs in their native niche (Nguyen et al., 2006; 2009; Rhee et al., 2006; Nowak et al., 2008). In addition, LHX2 represses sebaceous gland differentiation: following its loss, the stem cell niche soon becomes a sebaceous gland (Folgueras et al., 2013). SOX9 represses epidermal differentiation: following its loss, the niche becomes an epidermal cyst (Kadaja et al., 2014). TCF3 and TCF4 repress HF differentiation: following their loss, quiescent HFSCs precociously activate a new hair cycle (Lien et al., 2014). TCF3 and TCF4 can partner with β-catenin, which is stabilized and becomes nuclear upon WNT signaling: if β-catenin is silenced in the quiescent HFSCs, they never reenter a new hair cycle. In their native niche, quiescent HFSCs express a transcriptional repressor TLE4 which binds to TCF3 and TCF4: our findings are consistent with the view that WNT signaling functions by relieving TCF3/4/TLE4-mediated repression (Lien et al., 2014). However, the precise mechanisms underlying the antagonistic phenotypic behavior between TCF3/4 and β-catenin still remain to be elucidated.
NFATc1 and FOXC1 are required for maintaining HFSC quiescence, and in its absence, HFs cycle precociously (Horsley et al., 2008; Lay et al., 2016). Both of these factors are downstream of BMP signaling, offering a potential explanation as to why BMP signaling must be lowered to activate hair cycling. A major feature of aging skin is elevated BMPs in the skin and elevated NFATc1 in HFSCs. We can stimulate old follicles by lowering BMPs or inhibiting NFATc1 (Keyes et al., 2013), and we can prevent the age-related delays between hair cycles by blocking either NFATc1 or FOXC1 (Lay et al., 2016). While a tantalizing possible route to possible therapies for treating hair thinning in aging humans, there still is much to learn, as mice that continuously generate hair deplete their follicle stem cells over time, suggesting that our stem cells may not have an endless capacity for self-renewal. At best, there has to be a balance!
NFiB is a transcription factor which is specific to the HFSCs, but functions by repressing the expression of genes that are essential for the differentiation of the melanocyte stem cells, which reside within the same stem cell niche (Chang et al., 2013). These two stem cell populations must be activated at the same time so that differentiating melanocytes can transfer pigment to the differentiating hair cells to provide the natural coloring to our hair. Loss of NFiB uncouples this crosstalk and leads to the precocious activation of a key NFiB target gene that encodes a secreted melanocyte differentiation factor (Chang et al., 2013).
There are a number of additional transcription factors and epigenetic regulators which are enhanced in the complex milieu of HFSC chromatin, and there is still much to be learned. Of the epigenetic regulators, we’ve thus far examined only the role of polycomb chromatin repressor complexes and their targeted methylation of H3K27, which we’ve shown function critically in controlling the fate switch from a stem cell to a committed, transit-amplifying state (Ezhkova et al., 2009; 2011; Lien et al., 2011). Our recent in vivo ChIP-seq studies with mutually exclusive H3K27ac has unearthed an additional component of this fascinating switch: a small cohort (~5%) of the total genes expressed by HFSCs are regulated by large open chromatin domains, called ‘super-enhancers’. These enhancers are typified by the presence of ‘epi-centers’, which are strong peaks of H3K27ac associated with clusters of densely packed binding motifs for the key stemness genes, noted above (Adam et al., 2015). In a paradigm first set for cultured embryonic stem cells, the genes encoding the key stemness transcription factors of in vivo tissue stem cells are themselves regulated by super-enhancers. Interestingly, the super-enhancer landscape of HFSCs differs dramatically from that of its short-lived progenitors, and also from HFSCs placed in culture, an environment that mimics a wound state. Moreover, a GFP transgene driven by a HFSC epi-center recapitulates this plasticity: In vivo, GFP is only activated in the bulge; in vitro, GFP is silenced, but regained upon grafting the cultured HFSCs and re-establishing HFs in vivo (Adam et al., 2015).
This information is offering us new avenues to tailor-make drivers to target genes to a particular cell type and state. Finally, the short-list of super-enhancer regulated genes is rich in stemness genes and human skin disease genes, offering a plethora of new, unexplored genes on this list for future study. As we continue to systematically work our way through the functional significance and mechanism of action of epigenetic and transcriptional controls on stem cells, we are taking advantage of our ability to efficiently and selectively knockdown genes in skin progenitors in a few days using lentiviral-mediated shRNA delivery (Beronja et al., 2010). When coupled with recent CRISPR/CAS knockout technology, we can now deliver guide RNAs to conditionally ablate genes at a rate that would have taken us years using conventional mouse genetics.
Our ultimate goal is to understand how external signals from the surrounding niche microenvironment impact chromatin dynamics to achieve tissue production in normal homeostasis and how this changes in wound-healing. Recently, we addressed the mechanisms underlying a decades-old observation that aging individuals heal their wounds slowly, rendering them at risk of infections. By transcriptome profiling of epidermal stem cells at the wound edge of young and old mice, we learned that during aging, molecular crosstalk breaks down between the skin stem cells and the resident immune cells (Keyes et al., 2016). Thus far, we’ve only had a glimpse of the tip of the iceberg of immune-epithelial stem cell interactions that take place in homeostasis, wound-healing, aging and malignant progression. This field, and that of other niche: stem cell interactions offer a rich ground for future explorations.
Equally important will be the expression of specific genes that enables epithelial progenitors to remodel their cytoskeleton and adhesive contacts and either form a stratified epidermis or an epithelial bud that can then develop into a hair follicle (Perez-Moreno et al., 2003; Blanpain and Fuchs, 2009; Hsu et al., 2014). While our model is the skin, the problem is a general one of how a single epithelial stem cell gives rise to a spatially organized, functional tissue. It is also integrally linked to understanding the basis of cancer progression.