Somatic stem cell heterogeneity: diversity in the blood, skin and intestinal stem cell compartments

Abstract

Somatic stem cells replenish many tissues throughout life to repair damage and to maintain tissue homeostasis. Stem cell function is frequently described as following a hierarchical model in which a single master cell undergoes self-renewal and differentiation into multiple cell types and is responsible for most regenerative activity. However, recent data from studies on blood, skin and intestinal epithelium all point to the concomitant action of multiple types of stem cells with distinct everyday roles. Under stress conditions such as acute injury, the surprising developmental flexibility of these stem cells enables them to adapt to diverse roles and to acquire different regeneration capabilities. This paradigm shift raises many new questions about the developmental origins, inter-relationships and molecular regulation of these multiple stem cell types.

In the late 1800s, descriptive pathology portrayed the cellular structure of many organs in exquisite detail. In the absence of modern experimental tools such as fluorescence imaging, tissue transplantation and animal models, developmental relationships among cell types were inferred from extensive observation of the tissues of interest and documented by detailed hand drawings. Observations of the bone marrow led to heated debate about whether the distinct lymphoid and myeloid components of the blood were continuously generated from a common cell or from distinct progenitor cells (a view championed by Paul Ehrlich). The term ‘stem cell’ (Stammzelle) first appeared in the literature around 1900, when it was used by Artur Pappenheim and others to promote the common progenitor concept¹ (reviewed in REF. 2). For the past century, this concept has been the foundation of our understanding of tissue regeneration.

In the mid-1950s, bone marrow transplantation in mice, combined with tracking of the cellular progeny of transplanted tissue on the basis of the presence of common chromosomal translocations, strongly supported the hypothesis that cells of both lymphoid and myeloid lineages originate from the same cell³. Corroboration of this idea came from studies in which transplanted cells were marked by retroviral transduction; integration sites that were shared by multiple blood lineages were considered to be indicative of a common originating cell^4,5. Thus, for more than 50 years, the research field has accepted the view that the haemato poietic system is maintained by a single type of stem cell that regenerates all of the blood lineages during adulthood.

Concepts of tissue regeneration that were developed from the study of haematopoiesis have provided a framework for understanding the mechanisms that underlie the maintenance of other tissues. As transplantation was not easily practicable in other tissues such as the skin and the intestine, the stem cell paradigm in these tissues gained support through experiments in which cell fates were followed using tritiated thymidine labelling⁶. Data from these experiments reinforced a model in which a single tissue-specific stem cell continuously regenerates some or all of the lineages in a given tissue.

Somatic stem cells have thus come to be defined as adult-derived cells that have two hallmark capabilities: the ability to undergo differentiation and generate multiple lineages over long periods of time; and the ability to simultaneously self-renew (that is, to regenerate themselves). Traditionally, the stem cell pool within a single tissue was thought to be uniform: all stem cells in the pool were presumed to have equivalent potential for differentiation and self-renewal.

However, recent data that were generated using new technologies in various systems (BOX 1) have indicated that, in many somatic tissues, the stem cell system is surprisingly heterogeneous, comprising different types of stem cells. Here, we review the evidence in support of such heterogeneity and present instructive examples from haematopoietic, skin and intestinal epithelium that argue in favour of a profound revision of traditional views. Stem cell systems in other tissues, including stomach, mammary gland and prostate tissues, which are not discussed in this Review, may also be worth reconsidering in light of this evidence^7–9. Similarly, malignancies may represent a special case of stem cell heterogeneity that fits into this broad conceptual framework (BOX 2).

Box 1. Strategies used to investigate heterogeneity.

Different strategies can be used to examine stem cell function and heterogeneity in different tissue systems. For example, in the case of bone marrow haematopoietic stem cells, the stem cells are often extracted from their niche and examined using functional assays such as transplantation. By contrast, transplantation is difficult to perform in the intestine; however, lineage tracing has been a powerful technique that has been used to identify and characterize the progeny of cells that are located in specific anatomical locations. Mouse hair follicle stem cells can be expanded in vitro and transplanted, and can generate hair follicles and sebaceous glands. Not all experimental strategies for studying stem cells are applicable to all tissues, and they could be better exploited in some cases, as detailed in the table.

Table 1.

Experimental strategy	Achieved in blood	Achieved in skin	Achieved in the intestine
Expansion of multipotent stem cells in vitro	No	Yes	Yes
Generation of tissue and/or stem cells from pluripotent stem cells	No	Yes	Yes, but needs work
Lineage tracing	No*	Yes	Yes
Transplantation	Yes	Yes	Yes, but inefficient
Single-cell analysis	Yes, but needs work	No	No
Barcoding of libraries of cells	Yes	No	No

^*Lineage relationships have been established largely through transplantation assays rather than by labelling of cells under homeostatic conditions.

Box 2. Cancer stem cell heterogeneity.

The concept of ‘cancer stem cells’ emerged in the 1990s to explain there emergence of cancer many years after apparent eradication, perhaps through a long-lived stem-like cell. Considerable debate about the existence of these cells and their importance in cancers of different tissues has continued for the past 15 years. As for normal somatic stem cells, our understanding of the identity and behaviour of cancer stem cells has evolved. It is now believed that different models explain there-emergence of cancer in different tissues and even within a given type of cancer: in some cases, a traditional hierarchical stem cell model may be accurate; in others, many (and sometimes all) of the cells in the tissue can function like stem cells.

Recently, ultra-deep genome sequencing of malignancies has begun to shed light on the identity and characteristics of cancer stem cells in some tissues. In acute myeloid leukaemia (AML), a small number of mutations arise in bona fide stem cells. These mutations, in genes such as DNMT3A and genes encoding members of the cohesin family, provide a growth advantage in these stem cells, which are otherwise fairly normal. Haematopoietic stem cells (HSCs) with such mutations represent a pre-leukaemic state, in which their properties are subtly altered such that further mutations have a large proliferative effect, quickly initiating leukaemia. In these cases, it is very likely that the mutations occur in an HSC^123–125. Multiple secondary or tertiary mutations can occur, generating a diversity of cell clones that coexist, compete and show distinct growth dynamics following chemotherapy¹²⁶. This clonal heterogeneity has enormous implications for how to ablate these kinds of malignancies, as using drugs that only target branches of the original malignant clone would almost certainly lead to a relapse. Thus, it will be crucial to develop new drugs that are designed to kill cells carrying the initiating mutation. Although this paradigm is well understood for at least some types of adult AML, other malignancies may be initiated by a progenitor¹²⁷. Furthermore, the extent to which all cells in a tumour, or a subset of stem‑like cells, can initiate the growth of a secondary tumour probably varies among malignancies of different types^128,129.

Evolving concepts in haematopoiesis

The haematopoietic system is one of the most dynamic systems in the body; billions of blood cells are generated every day to continuously replace the dozen or so different peripheral blood cell types that are expended (FIG. 1). Since the first bone marrow transplantation experiments in the 1950s, substantial experimental effort has been made to identify reconstituting haematopoietic stem cells (HSCs). This culminated in the establishment of several robust strategies for their purification in the 1990s, which have facilitated their study.

Figure 1. Stem cell models for the haematopoietic system.

a | The traditional hierarchical view of haematopoiesis is that there is one type of stem cell that has the capacity to give rise to lineage-restricted progenitors that differentiate into all the cell types of the blood with equivalent propensity. b | In the consortium model, a pool of stem cells with slightly different properties regenerates the system continuously through progenitors that are increasingly restricted in their potential. c | In a new speculative model, stem cells are rare reserve cells that occasionally generate lineage-restricted progenitors. These stem cells have different lineage biases and give rise to specific progenitors. Existing data suggest that the most primitive stem cells are primed towards the megakaryocyte lineage²². These stem cells give rise to progenitors that are largely restricted to specific fate choices, and these progenitors are the main drivers of haematopoiesis, generating massive numbers of differentiated cells over a long period of time. During extreme stress (such as major injury or transplantation), the progenitors may revert (dashed arrows pointing left) to a stem-like state while retaining some of their lineage preferences. This model is consistent with the reported existence of megakaryocyte-biased stem cells (cells at the top of the progenitor hierarchy) and lymphoid-primed multipotent progenitors (one-step-down stem-like cells that lack megakaryocyte differentiation potential) as well as with the increasing differentiation bias observed with age. Indeed, it has been shown that the progenitors lose their developmental flexibility during ageing. Models that are hybrids of the three that are outlined in this figure can also be envisioned.

Variations in HSC behaviour

HSCs are widely viewed as being a uniform population of cells with an equivalent capacity to generate diverse progeny. Nevertheless, data have for some time suggested that there is considerable variation among individual stem cells. For example, single purified HSCs showed large fluctuations in their contribution to myeloid and lymphoid lineages when engrafted in recipient mice, suggesting that there is inherent variability in self-renewal and multilineage differentiation despite the cell population that was used for transplantation being highly purified¹⁰. Similarly, transplants of clonally derived HSCs from in vitro cultures displayed marked variation in repopulation kinetics among the transplanted cells as well as differences in their myeloid-to-lymphoid cell output ratio¹¹. Importantly, serial transplantation of bone marrow derived from the clones showed that daughter and granddaughter HSCs recapitulated the behaviour of their parent clone. These studies all indicated that the self-renewal and differentiation capacity of individual HSCs was both varied and intrinsically predetermined. Subsequent work led to the hypothesis that there are two classes of HSC: myeloid-biased HSCs (which preferentially give rise to myeloid progeny), and lymphoid-biased HSCs (which produce proportionately more lymphoid than myeloid cells)¹².

Although intriguing, this hypothesis only took root following a landmark study of a large cohort of mice that had been transplanted with single purified HSCs. The differentiation of individual HSCs was followed for many months, in primary and secondary transplant recipients, offering an unprecedented view of the diversity of the adult mouse HSC pool¹³. Two main classes of cell with multilineage HSC-like activity, which were designated α-cells and β-cells, were considered to be bona fide HSCs with the ability to sustain all blood production in the long term. α-cells displayed reduced capacity for generating lymphoid lineages (similar to myeloid-biased HSCs), and β-cells showed diminished production of myeloid progeny (that is, they were lymphoid biased). These behaviours were remarkably durable. After transplantation into secondary recipients, these differentiation-biased behaviours were recapitulated, and the HSCs that were regenerated were primarily of the same type as the original transplanted cell, supporting the idea that cell behaviours are largely predetermined in nature (although some interconvertibility was also reported). Although it remains unclear when this heterogeneity arises, it is already apparent in HSCs that have been purified from mouse fetal tissue¹⁴, suggesting that HSC subtypes are established during embryogenesis or shortly after. A meta-analysis of marking studies in the human bone marrow indicated that human HSCs also have heterogeneous regenerative properties¹⁵.

These studies led to the view that HSCs exist as two distinct stem cell types, but in fact they show a range of intermediate behaviours (FIG. 1). Thus, purification of individual HSC types has been challenging, which has impeded both their strict classification and their study at the molecular level. Nevertheless, some strategies have emerged to partially separate HSCs with discrete properties. For example, the fluorescent DNA-binding dye Hoechst 33342, which is pumped out of HSCs, can be used to distinguish a range of HSCs that are otherwise similar with regard to cell surface markers. When tested by single-cell transplantation, HSCs with the greatest dye efflux showed enhanced production of myeloid progeny, whereas HSCs with the least dye efflux favoured lymphoid cell generation. Moreover, lymphoid-biased HSCs displayed lower overall reconstitution and produced progeny for a shorter period of time than the myeloid-biased HSCs¹⁶.

Other markers that have been used to partially distinguish between HSC subtypes include Cd150 (REFS 16–19) and other signalling lymphocytic activation molecule (SLAM) family surface markers¹³⁰, integrin α2 (REF. 20), Cd41 (also known as Itga2b)²¹ and von Willebrand factor homologue (Vwf)²². In addition, HSC classes have been reported to differ in their sensitivity to transforming growth factor-β (TGFβ) signalling^16,23 and in their response to gamma irradiation²⁴. Importantly, the more myeloid-biased HSCs show the greatest level of quiescence, whereas the more lymphoid-biased HSCs divide more actively^16,17,130. This difference in proliferative behaviour alone could account for many measurable differences in the HSCs, such as the more limited time frame during which they can contribute after transplantation (reviewed in more detail in REF. 25). Together, these and other studies have contributed to changing our view of stem cells from being a single cell type to a consortium of cell types (FIG. 1), all of which contribute to blood regeneration.

Intrinsic and extrinsic regulation

An important question that has arisen from these findings is whether intrinsic and/or extrinsic cues regulate and confer these different HSC behaviours. Differences in growth factor responsiveness are indicative of intrinsic regulation^16,23. However, technologies that facilitate analyses at the single-cell level by microarrays²⁶, real-time PCR^27,28, RNA sequencing^29,30 and epigenetic analysis^30,31 should ultimately yield new insights into the molecular mechanisms that underlie HSC heterogeneity. For example, DNA methyltransferases are differentially expressed among HSC types in mice³².

A particular challenge for understanding the molecular regulation of HSCs results from the fact that HSC types are probably a continuum of states, which are characterized by subtle differences such as small changes in gene expression (on the order of twofold–threefold, as reported in REF. 16), the functional significance of which may be difficult to discern.

Currently, little is known about extrinsic regulation of HSC types. However, we speculate that different HSC types may be associated with distinct niches that influence the differentiation and self-renewal capacity of these cells. In recent years, several cell types have been identified as potential niche constituents, including osteoblasts³³, endothelial cells^34,35, Schwann cells³⁶ and megakaryocytes³⁷ (reviewed in REF. 38). Most of these are candidates for physical association with different HSC subtypes, which may direct different differentiation patterns.

How do HSCs normally behave?

Transplantation experiments have shown that HSCs have a range of properties, which raises the question of how they naturally behave, in the absence of such perturbations. Despite advances in technologies, such as transplantation of single HSCs and barcoding, that enable the investigation of cellular behaviours with unprecedented detail, HSCs are still largely defined by their functional output in assays that broadly mimic extreme circumstances, such as recovery from injury. The stem cell activities that are observed in these assays are unlikely to reflect in vivo behaviours under normal homeostasis, and they may not occur even during typical injury recovery. For example, single-cell transplantation is considered to be the ultimate test to define the differentiation competence of a cell. However, a single cell is unlikely to ever need to regenerate an entire tissue. Thus, this extreme test may push a cell to behave in ways it would not typically behave.

To determine normal cellular dynamics and differentiation capacity, it is necessary to use strategies that mark and track cells with minimal perturbation of endogenous tissues. In vivo transposon tagging was the first approach of this kind and has now been used to track blood cell production in mice without cell transplantation³⁹. When transposition was stimulated, and transposon insertion sites were tracked in multiple blood lineages over time, tags that were common to multiple lineages were rarely observed. Instead, the majority of neutrophils (which are the blood cells with the highest turnover) were continuously generated from very long‑lived lineage‑restricted progenitors³⁹. These progenitor cells produced neutrophils for more than 1 year, a time frame in which only HSCs were thought to operate. These data could indicate the existence of stem-like cells that give rise exclusively to neutrophils or the existence of neutrophil-restricted progenitors with extraordinary longevity. These recent observations are not incompatible with the traditional HSC concept, but they do call into question the circumstances in which true multipotent cells or lineage restricted progenitors are used. Much more work remains to be done to understand these cellular dynamics.

A new model of haematopoiesis

Together, these recent studies challenge the traditional hierarchical view of HSC differentiation, in which a relatively uniform pool of HSCs regenerates all peripheral blood cells with equal propensity over a long period of time (FIG. 1a). However, even in traditional transplantation assays, HSCs have been shown to be predestined to produce one lineage in preference to others, with some HSCs potently contributing to single lineages, such as neutrophils and megakaryocytes, over long periods of time^22,39,40. These data suggested a consortium model of stem cells with slightly varied differentiation propensities (FIG. 1b). In this model, all of the HSCs could be used for blood production, generating all cell types in the differentiated progeny, with variations in their output (that is, bias towards myeloid or lymphoid differentiation). This view could be reconciled with the most recent data and lead to a new model (FIG. 1c), which would bring together many disparate observations.

One possibility is that stem-like cells are established early during development (perhaps different stem cells are associated with different niches). These stem-like cells mainly generate long-lived progenitors that then almost exclusively produce one or a few lineages of cells. Some of these dedicated progenitors may retain a degree of developmental flexibility, such that they can revert to a more primitive state under conditions of duress (for example, transplantation), in which they behave more like stem cells. Dedicated progenitors that revert to oligopotency (that is, stem-like progenitors) may differ in their differentiation capacity, which may correspond to behaviours that have been interpreted as HSC differentiation bias. The lineage relationship between these stem-like progenitors may be one of stepwise loss of differentiation potential (FIG. 1c), which would be consistent with the existence of cells such as the lymphoid-primed multipotent progenitor⁴¹ that lack only megakaryocyte–erythroid potential.

New experimental approaches are required for rigorous testing of this model of haematopoiesis. Nonetheless, the recent data on HSC potential should prompt a re-evaluation of the definitions of stem cells.

Skin epithelial stem cells

Similar to bone marrow cells, a large number of skin cells are continuously regenerated throughout life. The regenerative potentials of epidermal progenitors have been studied for decades and form a foundation for both experimental and therapeutic stem cell biology. The skin is an easily accessible tissue, which has enabled the three-dimensional landscape of its cell types and their interactions to be mapped in great detail.

The skin epithelium comprises the interfollicular epidermis (IFE) and its appendages, which include hair follicles, sebaceous glands and sweat glands⁴². During early development, a single layer of ectodermal cells gives rise to the entire IFE and its appendages. After morphogenesis, the hair follicle goes through cyclical phases of regression, rest and growth⁴³. During the resting phase of the hair cycle, the permanent part of the hair follicle is divided into several compartments, which are anatomically and biochemically distinct (FIG. 2). These include the infundibulum, which is contiguous with the IFE in the uppermost portion of the hair follicle; the isthmus, which is located directly below the infundibulum and above the hair follicle bulge; and the hair germ, which is situated directly below the bulge and on top of the dermal papilla, which is a cluster of specialized mesenchymal cells that are required for hair follicle regeneration^44,45.

Figure 2. Heterogeneity of skin epithelial stem cells.

The resting-phase hair follice(HF) contains several compartments (indicated by different colours), which are defined by cells that express distinct molecular markers: stem cell antigen 1 (Sca1; also known as Ly6a), leucine-rich repeats and immunoglobulin-like domains 1 (Lrig1), placenta-expressed transcript 1 (Plet1), leucine-rich repeat-containing G protein-coupled receptor 6 (Lgr6), Blimp1, Sox9, transcription factor 3 (Tcf3), Tcf4, LIM homeobox 2 (Lhx2), keratin 15 (K15), K19, CD34, nuclear factor of activated T cells, cytoplasmic calcineurin-dependent 1 (Nfatc1) and Gli1. The compartments above the HF bulge include the infundibulum, which is contiguous with the interfollicular epidermis (IFE) in the uppermost portion of the HF; the isthmus, which is located directly below the infundibulum; and the junctional zone, which is part of the upper isthmus and lies next to the sebaceous gland (SG). These compartments contain their own stem cells that maintain their homeostasis. The bulge itself contains HF stem cells, whereas the hair germ (HG), which is located directly below the bulge, comprises progenitor cells. The HG is marked by high levels of cadherin 3 (Cdh3) expression and is situated on top of the dermal papilla, which is a cluster of specialized mesenchymal cells that are required for HF regeneration. HFs comprise a heterogeneous population of stem cells, which express different markers as indicated. As listed in the table, lineage tracing using inducible Cre recombinases under the control of specific promoters, which are expressed in distinct cell populations, shows the restricted lineage potential of the heterogeneous populations during normal homeostasis as well as an expanded potential in response to injury. ND, not done.

Multiple epidermal populations with stem cell activity in vitro?

The bulge region, which was first identified as the hair follicle stem cell reservoir, contains cells that have high proliferative capacity in vitro^46,47 and are slow cycling in vivo, as demonstrated by their ability to retain nucleotide analogues or labelled histone in pulse-chase experiments (so‑called label‑retaining cells)^47–50. In mice, such bulge cells are uniquely marked by CD34 (REF. 51) and nuclear factor of activated T cells, cytoplasmic calcineurin-dependent 1 (NFATC1) expression⁵², and bulge cells and the hair germ express keratin 15 (K15; also known as KRT15)⁵³, K19 (also known as KRT19)⁵⁴, transcription factor 3 (TCF3)⁵⁵, TCF4 (REF. 56), LIM homeobox 2 (LHX2)⁵⁷, SOX9 (REF. 58) and leucinerich repeat containing G protein-coupled receptor 5 (LGR5)⁵⁹. When isolated bulge cells are combined with neonatal dermal cells in transplantation assays in immunodeficient mice, they can reconstitute all skin epithelial lineages, including the IFE, hair follicle and sebaceous gland^59–62. Thus, for some time, bulge cells were thought to be the true multipotent stem cells at the apex of the epidermal hierarchy.

Subsequently, cells in the isthmus, which are marked by LGR6 (REF. 63), and cells in the junctional zone in the upper isthmus⁶⁴, which are marked by leucine‑rich repeats and immunoglobulin‑like domains 1 (LRIG1) and placenta-expressed transcript 1 (PLET1)^65,66, were also found to have high proliferative capacities in vitro. These cells were able to contribute to all three skin epithelial lineages in transplantation assays. These data indicated that various populations of cells with distinct markers can reconstitute all skin epithelial lineages. The field then began to question whether the stem cell potential of these heterogeneous populations was indicative of endogenous stem cell function.

Normal epidermal stem cell function

Meanwhile, evidence was accumulating that the IFE and hair follicle are maintained by separate stem cell populations under physiological conditions. Although retroviral fate mapping had previously suggested the existence of discrete stem cell compartments⁶⁷, lineage tracing in which cells of the embryonic hair follicle buds were marked showed for the first time that the entire hair follicle, including the sebaceous gland, but not the IFE, was derived from these embryonic progenitor cells. This new evidence indicated that the IFE and its appendages are derived from separate pools of stem cells^58,68. In addition, a cell ablation experiment that used the K15 promoter to drive the expression of an inducible ‘suicide gene’ to ablate adult bulge cells led to the complete loss of hair follicles but did not affect the IFE, indicating that a separate pool of stem cells maintains IFE homeostasis⁶⁹. Finally, lineage tracing in adult tissue showed that the basal layer of the IFE contains its own stem cells, which maintain the stratified epidermis^70,71.

Although embryonic progenitors in hair follicle buds contribute to all cells in the hair follicle and the sebaceous gland^58,68, lineage-tracing experiments have shown that under physiological conditions discrete stem cell populations in adult tissues maintain the sebaceous gland as well as distinct components of the hair follicle (FIG. 2).

Various groups have shown, through lineage tracing using mutant mice expressing the Cre recombinase fusion proteins CrePR (which is fused to a truncated progesterone receptor) and CreER (which is fused to a truncated oestrogen receptor), that bulge and hair germ cells predominantly give rise to the lower part of the hair follicle. Specifically, these studies used bulge and hair germ-specific inducible K15–CrePR transgenic mice⁶¹, and K19–CreER^54,72, Lgr5–CreER^59,72, Gli1–CreER⁷³ and Tcf3–CreER⁷⁴ knock-in mice (see the activity of these promoter s in FIG. 2).

Lineage tracing has also shown that the infundibulum and the sebaceous gland are maintained by Lrig1⁺ cells, which are located in the junctional zone and the basal layer of the sebaceous gland⁷². Blimp1⁺ cells⁷⁵, which are located adjacent to the sebaceous glands and are distinct from Lrig1⁺ cells⁷², can also give rise to sebaceous gland cells⁷⁵; however, this finding has recently been disputed⁷⁶. Previously, Lgr6⁺ cells were reported to be located in the adult central isthmus and to give rise to the sebaceous gland and IFE⁶³. However, recent studies have shown that Lgr6⁺ cells are not restricted to the isthmus but are also present in the sebaceous gland and IFE^72,77.

Although lineage tracing is a very powerful tool for identifying stem cell populations in the skin, it relies on cell-type- and temporal-specific induction of Cre recombinases and therefore produces slightly different results depending on the timing of the labelling^74,78. Collectively however, the results that have been obtained using this approach strongly support the concept of stem cell heterogeneity in the hair follicle, in which several distinct pools of stem cells regenerate one or more parts of the hair follicle.

Not only are there heterogeneous populations of stem cells within the hair follicle, there is heterogeneity within the bulge itself. Lgr5⁺ cells, which are proliferative cells located in the lower part of the bulge and the hair germ, are distinct from the quiescent label-retaining cells and only partially overlap with Cd34⁺ and K15⁺ bulge cells⁵⁹. In addition, two distinct subcompartments of the bulge express Gli1 (REF. 73). A recent study, using lineage tracing of single cells in combination with live imaging, revealed differences between the fates of stem cells located in the upper bulge and those located in the lower bulge, further supporting the finding that bulge cells are heterogeneous. Cells in the upper bulge tend to remain in the bulge and do not contribute to hair follicle regeneration, whereas cells in the lower bulge regenerate the outer root sheath of the hair follicle⁷⁹. Cells of the hair germ, which express many of the same molecular markers as bulge cells, have their own molecular signature⁸⁰ and mainly contribute to differentiated lineages of the hair follicle, but they can also contribute to the regeneration of the outer root sheath⁷⁹.

Although they are heterogeneous in their molecular markers, proliferative propensity and cell fate, these discrete populations in the hair follicle can interconvert and diverge into different lineages to replace damaged cells. Elegant experiments that used laser ablation and in vivo imaging have shown that, after bulge ablation, neighbouring cells from the hair germ or upper hair follicle region repopulate the ablated area and acquire bulge cell identity, and are capable of regenerating the lower part of the hair follicle. Similarly, laser ablation of hair germ cells induces bulge cells to reconstitute the hair germ and give rise to new hair growth⁷⁹. Following wounding by removal of a full‑thickness piece of skin (that is, including the dermis as well as the epidermis), the majority of the re-epithelialized portion of the IFE is derived from the neighbouring basal cells of the IFE⁷¹, although cells from discrete compartments of the hair follicle have also been found to contribute to the regeneration and repair of the IFE^{58,63,69,72–74,81}.

The fate-determining role of the microenvironment

Stem cell niches, which support and regulate stem cell function, contain numerous cell types, although the precise cellular constitution is unique to each niche. In the skin, the dermal papilla is required for hair follicle stem cell activation⁴⁴. In addition, adipocytes, nerves and the arrector pili muscle have all been shown to affect either the characteristics or the behaviour of the hair follicle stem cells^73,77,82,83, and thus they are all candidate components of the hair follicle stem cell niche. Niche cells can be descendants of stem cells: for example, K6⁺ inner bulge cells (which are descendants of the bulge cells and function as a niche for bulge cells) secrete factors that promote hair follicle stem cell quiescence⁷⁸.

As discussed above, lineage-tracing experiments indicate that, under physiological conditions, hetero-geneous populations of stem cells are restricted in their lineage differentiation potential. However, transplantation experiments, in which cells are removed from their natural environments, clearly show that these cells have the capacity to give rise to all skin epithelial lineages. These observations suggest that the intrinsic fate of these cells is not irreversibly predetermined and that the microenvironment in which the heterogeneous populations reside restricts their lineage choice.

Different microenvironments probably influence the behaviour of stem cells by conferring on them differential proliferative properties and dictating the expression of their molecular markers. Recent studies in which denervation abolished the specific expression pattern of the stem cell markers Gli1 (REF. 73) and Lgr6 (REF. 77) in the hair follicle underscored the importance of this interaction between the microenvironment and the stem cell population. The role of the microenvironment in influencing stem cell fate has also been shown for melanocyte stem cells, which are another stem cell population in the hair follicle^36,84,85.

Further work will be required to identify the signals from the microenvironment that specify the identity of heterogeneous stem cell populations and that regulate fate determination. Understanding the mechanisms by which the microenvironment regulates stem cell function is challenging because of the complexity of the stem cell niches. However, advances in single-cell technologies now allow expression profiling of the single cells that constitute the different microenvironments. The remaining challenge is to functionally show how the factors produced by the different microenvironments dictate the identity and behaviour of a specific stem cell population.

Intestinal stem cells

The intestinal lining is also highly regenerative and has long been thought to be maintained by stem-like cells⁶. The lining comprises a single layer of columnar epithelium that is renewed every 3–5 days⁸⁶. This rapid replacement is supported by progenitor cells that are organized into proliferative units termed crypts of Lieberkühn, which are pockets of cells that are embedded in the wall of the intestine (FIG. 3). Intestinal stem cells (ISCs) located near the bases of these crypts continuously proliferate to support the constant turnover of differentiated cells from the surface.

Figure 3. Overview of the intestinal stem cell system.

a | Intestinal progenitor cells are organized in crypts of Lieberkühn. Intestinal stem cells (ISCs) reside near the crypt base and produce daughter cells (transient-amplifying (TA) cells), which proliferate in the mid-crypt and terminally differentiate near the crypt opening to produce the diverse range of intestinal epithelial cell types. Crypt base columnar (CBC) stem cells are interdigitated with Paneth cells at the crypt base. The differentiated cells that line the villus include absorptive enterocytes, goblet cells and enteroendocrine cells. Quiescent ISCs (qISCs; also known as +4 stem cells) are thought to reside just above the zone that contains Paneth and CBC cells. b | CBC cells compete for niche support and space at the base of the crypt, which confers a competitive advantage on the lower (tier 1) CBC cells. Tier 2 CBC cells differentiate as they leave the niche, using contact-dependent Notch signalling to determine their cell fate. Contact with ligand-expressing secretory cells (goblet, enteroendocrine or Paneth cells) activates Notch and directs cells into the absorptive lineage, which gives rise to enterocytes. Contact that is limited to enterocytes or their progenitors results in expression of the transcription factor atonal homologue 1 (ATOH1) and in secretory lineage commitment. Secretory progenitor cells subsequently differentiate into goblet, Paneth or enteroendocrine cells. Extreme injury that causes depletion of the CBC cell pool can induce reversion of committed progenitor cells to a stem cell state (indicated by the dashed arrow). qISCs are thought to give rise to CBC cells following injury. It remains unclear whether a dedicated pool of qISCs exists or whether these cells are transient progenitor cells under homeostatic conditions.

Two claims for intestinal stem cell identity

Historically, two types of ISC were identified. The first, known as crypt base columnar (CBC) cells, was defined using cytological lineage tracing⁶. The second, termed +4 cells, was defined on the basis of DNA label retention, proliferation and radiation injury response^87,88. More recently, CBC and +4 cells have been defined using molecular markers and transgenic lineage-tracing techniques. CBC cells are the proliferative engines that drive cellular production in the crypts: they divide daily, with frequent turnover⁸⁹. +4 cells have been redefined as reserve or quiescent ISCs (qISCs): they divide infrequently under homeostatic conditions but can be induced to produce new CBC cells in response to injury or other stimuli⁸⁹.

Proliferation of CBC cells is highly dependent on canonical WNT–β-catenin signalling. Analysis of β-catenin target genes led to the identification of the first molecular marker that was found to be specific to CBC cells, LGR5 (REF. 90), and hence to the discovery of additional CBC cell markers. These markers include achaete-scute homologue 2 (ASCL2), olfactomedin 4 (OLFM4), SPARC-related modular calcium-binding protein 2 (SMOC2) and tumour necrosis factor receptor superfamily member 19 (TNFRSF19)^91–94. qISCs were identified by their localization just above the Paneth cell zone at the base of the crypt and by their ability to regenerate the entire crypt in response to injury (typically radiation)⁹⁵. qISCs were marked by Polycomb complex protein BMI1, doublecortin-like kinase 1 (DCLK1), HOP homeobox (HOPX), LRIG1 and telomerase reverse transcriptase (TERT)^96–101.

Lineage-tracing experiments suggest that CBC cells and qISCs interconvert; in response to radiation injury or genetic ablation of Lgr5⁺ CBC cells, qISCs are thought to enter the cell cycle and produce new CBC cells (FIG. 3b), which are indispensable for recovery from radiation injury^102–104. Conversely, CBC cells can give rise to qISCs through unknown mechanisms⁹⁸. However, the existence of a dedicated qISC pool is cur‑ rently the subject of active debate and investigation, as discussed below.

Interestingly, some data suggest that intercellular signalling pathways can differentially regulate CBC cells and qISCs. In contrast to CBC cells, Bmi1⁺ qISCs were shown to be resistant to β-catenin-driven proliferative signals¹⁰³. In addition to WNT–β-catenin signalling, other signalling pathways that may differentially contribute to ISC activity include Notch and bone morpho genetic protein (BMP). BMPs counteract mitogenic signals in the crypt, probably by antagonizing canonical WNT signalling¹⁰⁵. Markers of BMP activity were found in label-retaining cells but have not been further evaluated in recently identified ISC populations; thus, it remains unclear whether BMP signalling controls CBC cell or qISC activity¹⁰⁵. Notch signalling is essential for both promoting self‑renewal and determining the cell fate of differentiating progenitor cells (reviewed in REF. 106). Activation of Notch signalling promotes the absorptive enterocyte fate, whereas cells that do not activate the Notch pathway express atonal homologue 1 (Atoh1) and commit to the secretory cell fate. Emerging data suggest that secretory progenitors may be able to contribute to the stem cell pool in response to injury^105,106. In this way, Notch activity could regulate the composition of both the active (CBC cell) and reserve (qISC) pools.

Reverse or reserve — how do crypts regenerate?

Two recent landmark studies suggested that non-stem cells in the crypts have remarkable developmental plasticity when subjected to stress^107,108. In one study, cells with high Delta-like 1 (Dll1) expression were shown to be secretory cell precursors, which were not part of the stem cell pool under homeostatic conditions but contributed to the stem cell pool during crypt regeneration following radiation injury. In another study, crypt cells retaining YFP-labelled histone H2B were shown to be secretory progenitor cells. The progeny of these label-retaining cells were marked using a dimerizable Cre enzyme fused to H2B. This approach enabled the authors to show that H2B-labelled cells did not normally contribute to the active stem cell compartment, but after various genotoxic injuries these label-retaining cells contributed to the stem cell pool and to all cell lineages that emerge from those crypts.

These studies suggest that, following severe injury, long-lived secretory progenitor cells can revert to function as stem cells to help to regenerate the intestinal epithelium. Other studies have cast doubt on the validity of qISC markers such as Bmi1 and Dclk1, reporting that they are either broadly expressed throughout the crypt or expressed in a subset of differentiated cells^92,109. Furthermore, qISC markers do not identify a homogeneous cell population, and the extent of the overlap of these different cell types has yet to be fully described. Despite the strong evidence that Lgr5⁻ cells can maintain the intestinal epithelium¹⁰², the existence of a dedicated qISC pool remains controversial. These recent results raise the question of whether cells previously reported to have stem cell activity are committed progenitors or differentiated cells under homeostatic conditions.

Role of the niche in regulating intestinal stem cell activity

The ISC niche, which consists of supporting epithelial cells, subepithelial stromal cells and the crypt luminal milieu, regulates ISC activity in several ways. Epithelial Paneth cells are interdigitated among the CBC stem cells at the crypt base and provide several important survival and growth factors, including secreted WNT ligands and transmembrane Notch ligands¹¹⁰. Stromal cells provide additional support to ISCs and can compensate for the loss of Paneth cells; this highlights the robustness and adaptability of the ISC niche^111,112. Bacterial products such as muramyl dipeptide, a component of the cell wall, can directly stimulate survival of ISCs and resistance to cytotoxic injury¹¹³. Moreover, quantitative analyses of clonal expansion, together with recent in situ imaging of cells that express GFP from the Lgr5 promoter, support the idea that the 14–16 CBC cells in each crypt compete with one another for niche support^114,115.

An important breakthrough was the development of a three-dimensional epithelial culture system that allowed continuous growth of Lgr5⁺ CBC cells while they produced the normal constituency of intestinal epithelial cells¹¹⁶. This culture system provides all essential non-epithelial niche signals, including epidermal growth factor, WNT–R-spondin, Noggin and basement membrane components. This experimental platform, known as Sato organoids or enteroids, have been used to identify putative ISCs that are capable of self-renewal and multilineage differentiation, to test the function of putative niche signals and to determine the effects of experimental manipulations on stem cell function^{98,111,117–119}. In addition, organoids have been used to show that Lgr5⁺ CBC cells are programmed according to their regional identity along the cephalocaudal axis¹²⁰. These regional identities of ISCs are maintained when the cells are transplanted into an ectopic site in the intestine (for example, transplantation of the small intestine into the colon)¹²¹. These data highlight another level of heterogeneity among ISCs.

Together, these data support a model in which CBC cells have equal intrinsic potential for self-renewal. As they divide, their daughter cells compete for niche support: cells that obtain a minimum threshold of niche signals remain stem cells, whereas other daughter cells are forced out of the lower crypt and begin to differentiate. Although all CBC cells may have equivalent potential, CBC cells that reside closer to the crypt base (known as tier 1 CBC cells) may have a self-renewal advantage compared with those that reside farther up the crypt (tier 2 CBC cells), because the daughters of tier 1 CBC cells can better compete for niche space, resulting in two effective levels of CBC self-renewal potential¹¹⁴ (FIG. 3b). A recent study reported that there are 5–7 functionally active stem cells per crypt, which may coincide with tier 1 CBC cells¹²². As these nascent progenitor cells exit the ISC niche, the composition of their immediate neighbourhood constrains their fate: contact with Notch-ligand-expressing secretory cells specifies these cells as absorptive enterocyte precursors, whereas absence of contact with a secretory neighbour enables those cells to express Atoh1 and commit to the secretory cell fate. This model does not exclude a role for dedicated qISCs, but more studies are needed to clarify the identity of these cells.

An evolving somatic stem cell perspective

It is striking that more than 100 years after the term ‘stem cell’ was coined, we are again debating the interrelationships among different cell types in various tissues. Research in the 1990s focused on identifying a ‘master’ stem cell population for each somatic tissue. As these master stem cells with different markers or behaviour were discovered, there was much heated debate about their identity and differentiation potential. Although many details remain to be resolved, the emerging view of stem cells in each of the three compartments — blood, skin and the intestine — is remarkably similar, leading to two broad conclusions. First, in each tissue, there is more than one population of cells, which express distinct markers, that can behave as stem cells. Thus, there is a new appreciation of stem cell diversity, even within a single stem cell compartment. Second, it seems that each of these stem cell types is more adaptable than had previously been thought: they have a ‘default’ role under normal conditions, but following perturbation, such as stimulation by injury, they can fulfil distinct functions when required.

A key remaining question is whether there is one ‘über’ stem cell in each of these tissues. In the intestine, the cell at the top of the hierarchy that produces all of the lineages during normal homeostasis is an Lgr5⁺ CBC cell. There is little information about normal homeostasis in the bone marrow, but transplantation experiments suggest that some cells have greater differentiation potential and longevity than others, and these cells may be the ‘über’ stem cells²². In the skin, multiple stem cell populations contribute to distinct compartments of the epidermis during normal homeostasis, indicating that it may lack a master stem cell.

Why are these definitions and cellular relationships important? As we move forward in our attempts to use stem cells for regenerative medicine, it may be very important to be able to distinguish the different regenerative requirements in different circumstances and to identify the best progenitor for a given purpose. The current approach to stem cell transplantation therapy is to provide highly purified stem cells of a specific type. With this new perspective, we can envision scenarios in which multiple types of progenitor are transplanted to support tissue regeneration with the longest duration and with the optimal array of cell types being produced.

Glossary

Quiescence

The state of being inactive. This usually implies limited mitotic activity and is sometimes referred to as dormancy

Niches

The local environments in which stem cells reside. Niches are thought to regulate stem cell activity through various types of interaction

CrePR

A Cre recombinase fused with a truncated progesterone receptor that translocates into the nucleus when the receptor binds to the progesterone antagonist RU486

CreER

A Cre recombinase fused with an oestrogen receptor that translocates into the nucleus when the receptor binds to the oestrogen antagonist tamoxifen. When it translocates to the nucleus, the Cre recombinase is activated and removes the sequences preceding the reporter gene, allowing expression of the reporter

Quiescent label-retaining cells

Cells that are labelled with a pulse of 5-bromo-2-deoxyuridin e (BrdU) and that retain the BrdU label when followed for a certain amount of time