Author contributions: T.G.F., J.B., L.G.-G., G.E.: Collection and/or assembly of data; S.A.C.M.: Conception and design, collection and/or assembly of data manuscript writing; A.H.: Collection and/or assembly of data, manuscript writing; S.I.: Collection and/or assembly of data, three-dimensional modeling; N.M.W.D.: Data analysis and interpretation; P.J.T.: three-dimensional modeling and software design for construction of crypt maps; H.M.K., S.B., L.M., M.E.: Provision of study material or patients; D.M.T., R.W.T., L.C.G., P.F.C., C.P.D.: Data analysis and interpretation; N.A.W., M.R.A.: Conception and design, data analysis and interpretation, manuscript writing. T.G.F., S.A.C.M., and J.B. contributed to this work as joint first authors.
First published online in STEM CELLSExpress March 19, 2009
Methods for lineage tracing of stem cell progeny in human tissues are currently not available. We describe a technique for detecting the expansion of a single cell's progeny that contain clonal mitochondrial DNA (mtDNA) mutations affecting the expression of mtDNA-encoded cytochrome c oxidase (COX). Because such mutations take up to 40 years to become phenotypically apparent, we believe these clonal patches originate in stem cells. Dual-color enzyme histochemistry was used to identify COX-deficient cells, and mutations were confirmed by microdissection of single cells with polymerase chain reaction sequencing of the entire mtDNA genome. These techniques have been applied to human intestine, liver, pancreas, and skin. Our results suggest that the stem cell niche is located at the base of colonic crypts and above the Paneth cell region in the small intestine, in accord with dynamic cell kinetic studies in animals. In the pancreas, exocrine tissue progenitors appeared to be located in or close to interlobular ducts, and, in the liver, we propose that stem cells are located in the periportal region. In the skin, the origin of a basal cell carcinoma appeared to be from the outer root sheath of the hair follicle. We propose that this is a general method for detecting clonal cell populations from which the location of the niche can be inferred, also affording the generation of cell fate maps, all in human tissues. In addition, the technique allows analysis of the origin of human tumors from specific tissue sites. STEM CELLS 2009;27:1410–1420
It is widely believed that adult tissue-specific stem cells reside within a specialized microenvironment, known as the niche, and stem cell behavior is regulated and maintained there . Many studies have attempted to identify stem cells and the location of the niche in a variety of epithelial tissues, including the gut and skin, and a number of controversies exist. It is rare indeed to be able to identify stem cells within tissues using histological methods. Stem cells are supposed to have inherent properties, such as DNA label retention, but specific molecular markers of stem cells have not been found in many tissues. In Drosophila, however, it has been possible to tag individual stem cells using genetic methods and to demonstrate their ability to self-renew and give rise to dependent cell lineages ; seven different types of stem cell have now been identified . In mammalian epithelial tissues progress has been much slower, although, for example, it is generally believed that germ-line stem cells lie within the basal cell layer of the seminiferous tubules , hair follicle stem cells reside within the bulge of hair follicles , and intestinal stem cells are located at or near the base of intestinal crypts . Markers for stem cells have long been sought in mammalian tissues, but even in comparatively well-defined instances such as the hematopoietic system, uncertainties still exist .
It has been proposed that the gold standard of stem cell identification involves marking putative stem cells to locate the neighboring niche and then performing lineage tracing to demonstrate that the proposed “stem cell” has multipotentiality . The application of Cre recombinase to map the fate of stem cells in mice has increased our knowledge of muscle satellite cells , spermatogonial stem cells , epidermal stem cells , and intestinal stem cells . In particular, lineages in the hair follicle have also been mapped, apparently demonstrating that cells within the bulge give rise to all of the epithelial cells within the hair follicle . Some of these approaches have made it possible to image mammalian stem cells within their niches and allow both single-cell resolution and lineage marking for the identification of a mammalian stem cell niche in vivo . The study by Barker et al. ., using the Wnt target gene, lgr5, showed restriction to epithelial cells at the base of the crypts: fate mapping of these cells with a Cre knock-in allele of lgr5 showed self-renewal with lgr5-positive cells giving rise to all intestinal epithelial lineages. Similarly, a 12.4-kilobase villin promoter/enhancer fragment driving several transgenes (enhanced green fluorescent protein, β-galactosidase, and Cre recombinase) allowed in vivo lineage tracing in the mouse antral gastric gland .
A niche consists of a specialized local tissue microenvironment capable of housing and maintaining one or more stem cells  and perturbing a precisely identified stem cell or its surroundings allows the existence, size, and regulatory properties of a corresponding niche to be revealed. In human epithelial tissues, however, progress in identifying stem cell niches and the stem cells themselves and in lineage tracing from these stem cells has been practically impossible. Importantly, because tumors are thought to arise from a single mutated cell and arguably that cell is a stem cell, methods are badly needed to identify stem cell niches and allow lineage tracing from stem cell progeny in human tissues. Studying stem cell dynamics and clonal populations in human tissue presents a number of unique challenges because of the inherent experimental limitations involved, and to date there has been no reliable means to do this. In this article we describe techniques for lineage tracing in human epithelial tissues, taking as our exemplars the intestine, liver, pancreas, and skin. We believe that the method has general applicability and is limited only by the availability of fresh-frozen human tissues. The major aim of this communication is to bring this general method for identifying clonally derived cell populations, from which fate maps can be traced, to the attention of the stem cell community.
MATERIALS AND METHODS
Patients and Tissues
Ethical approval was sought and obtained as per the requirements of the United Kingdom Human Tissue Act (2006), REC reference numbers 06/Q0603/1 and 07/Q1604/17. Colonic tissue was obtained from patients who had a colectomy for adenocarcinoma, with normal mucosa as far as possible (∼15 cm) from the tumor margin being analyzed. Small intestinal tissue was obtained from patients undergoing pancreaticoduodenectomy for pancreatic malignancy. Fresh-frozen blocks of human liver, of normal histological appearance, were obtained from hepatic resections for metastatic colorectal carcinoma. Cytochrome c oxidase (COX)-deficient patches were identified, COX-deficient and COX-positive cells were microdissected, and the mitochondrial DNA (mtDNA) was sequenced. Histologically normal pancreata, distant from tumors, were obtained from patients undergoing pancreaticoduodenectomy for periampullary carcinoma, snap-frozen, and analyzed as described for liver. Excision specimens from six basal cell carcinomas were snap-frozen and analyzed as described above. All COX-deficient cells (both normal and from tumors) were functional and expressed the epithelial marker pan-cytokeratin (Fig. 7).
Histological Detection of COX-Deficient Cells
Two-color enzyme histochemistry can be used to simultaneously detect the mtDNA-encoded cytochrome c oxidase (complex IV of the respiratory chain) and nuclear DNA-encoded succinate dehydrogenase (SDH), a component of complex II of the respiratory chain. Cells lacking COX activity appear blue (such cells have been previously shown to harbor random mutations, which lead to the deficiency in COX); nonmutated cells appear brown. Frozen sections were cut at a thickness of 8 μm. Sequential COX and SDH histochemistry was used to highlight any deficiencies in COX. Briefly, air-dried sections were incubated in cytochrome c oxidase medium containing 100 mmol/l cytochrome c, 20 mg/ml catalase, and 4 mmol/l diaminobenzidine tetrahydrochloride in 0.2 mol/l phosphate buffer, pH 7.0 (all from Sigma-Aldrich, Poole, U.K., http://www.sigmaaldrich.com) for a maximum of 50 minutes at 37°C. Sections were then washed in phosphate-buffered saline (PBS), pH 7.4, for 3-5 minutes and incubated in SDH medium (130 mmol/l sodium succinate, 200 mmol/l phenazine methosulfate, 1 mmol/l sodium azide, and 1.5 mmol/l nitroblue tetrazolium in 0.2 mol/l phosphate buffer, pH 7.0) for a maximum of 45 minutes at 37°C or until a strong blue stain had developed. Sections again were washed in PBS for 3-5 minutes and dehydrated in a graded ethanol series (70, 90, 100, and 100%), cleared in Histoclear (Lamb Laboratory Supplies, Eastbourne, U.K., http://www.ralamb.co.uk/), and mounted with Permount (Fisher Scientific, FairLawn, NJ, http://www.fisherscientific.com).
The absence of COX subunits can also be demonstrated using immunohistochemical methods. Formalin-fixed paraffin sections (4 μm) were cut and allowed to air dry overnight. All sections were dewaxed in xylene and rehydrated through decreasing alcohol concentrations and then blocked for endogenous peroxidase in methanol/H2O2. Sections then were microwaved for 10 minutes in boiling sodium citrate buffer (pH 6.0) and allowed to slowly cool under running distilled water. Sections first were blocked with a serum-free protein block (DAKO,) for 10 minutes followed by incubation with streptavidin for 15 minutes and then biotin, also for 15 minutes at room temperature (Vector Laboratories, Peterborough, U.K., http://www.vectorlabs.com). The primary antibody, mouse anti-human cytochrome c oxidase subunit 1 (1:250; Molecular Probes Invitrogen, Paisley, U.K., http://probes.invitrogen.com), diluted in PBS with 5% fetal calf serum, was applied for 35 minutes at room temperature in a humid chamber. Sections were then washed for 3-5 minutes in PBS followed by a 30-minute incubation with appropriate secondary antibodies conjugated to biotin. After the final wash, the sections were incubated with streptavidin peroxidase for another 30 minutes, washed, and developed in a solution containing 4 mmol/l diaminobenzidine and 0.2% hydrogen peroxide. Sections were dehydrated through alcohol, cleared with xylene, and mounted with DePeX. An identical protocol was used for all other antibodies: anti-pan-cytokeratin (1:100; DAKO, Ely, U.K.), anti-Ki-67 (1:100, DAKO), and anti-α-fetoprotein (1:25; Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com).
Cells from COX-positive and -deficient areas were cut by laser microdissection. Cell digestion and DNA extraction were performed by overnight incubation in a DNA extraction kit (PicoPure, Arcturus; Molecular Devices, Sunnyvale, CA, http://www.moleculardevices.com) at 65°C. The extracted DNA was used to sequence the entire mitochondrial genome from microdissected areas. A two-round amplification method was followed, whereby the first round consisted of amplifying nine fragments spanning the entire genome, and the second round consisted of 36 M13-tailed primer pairs to amplify overlapping segments of the first-round products. Sequencing was performed using the BigDye terminator cycle sequencing method on an ABI Prism Genetic Analyzer and compared with the revised Cambridge reference sequence using sequence alignment software of the European Molecular Biology Open Software Suite (EMBOSS, http://www.ebi.ac.uk/emboss/align/).
Tracking Cell Clones in the Intestine: Generation of Crypt Maps
By combining the two-color enzyme histochemical staining with image analysis and computer reconstruction, we have devised a method of mapping the origin and spread of stem cell progeny and mutated cell populations within the intestinal crypt. Multiple serial sections in the transverse plain are taken through a frozen sample. Digital images are taken of each serial section and are used to create what we call a “crypt map” using image analytical algorithms and in-house software developed by one of the authors (P.J.T.). A crypt map is a representation of the whole three-dimensional (3D) tubular crypt unfurled and laid flat (like a map) with color enhancement postprocessing designed to increase the contrast between blue-staining and non-blue-staining cells. Briefly, serial section images of the crypt were digitally aligned to form a 3D reconstruction series of the whole crypt. The center point and perimeter of the cross-sectional image of the crypt in such a series was then marked out by hand using freely available interactive image analysis software (BiaQIm, version 2.5 alpha, http://www.bialith.com). For each cross-sectional image a line was then subtended from the center point to the perimeter at a fixed angle (known as the “cut-angle”). The cut-angle, although arbitrary, was the same for all cross-sectional images in a given crypt series. For each image in the series, the approximately oval cross-sectional profile of the crypt is then digitally “cut” and then warped into a straight profile. For each straight profile generated, the basal-most pixels (i.e., the pixels starting at the basement membrane aspect and continuing toward the center for a distance of approximately 25% of the epithelial thickness) were then averaged into a single pixel strip. This step acquired most of the cytoplasmic color information before the cytoplasm gets “diluted” by mucin as one gets close to the luminal aspect of the crypt epithelium. These single pixel strips were then arranged in order of the 3D series, from crypt base at the bottom to gut luminal surface at the top, forming the initial crypt map, which underwent color discrimination analysis to isolate the blue hues and finally was false-colored to give the final crypt map.
The result of the above is a crypt map in which the intensity value of each pixel is a log-transformed representation of the blue component in the original RGB image multiplied by its saturation. Thus, a pixel that is white in the original image (e.g., mucin or other unstained areas) will have lots of blue in it but will be very unsaturated and so will be dark in this image. A pixel that is brownish, reddish, or greenish in the original image will have a high saturation value but will not have a hue in the blue range and so will appear dark in this enhanced image. Only those pixels that have both a hue in the blue range and are also saturated will appear bright in this postprocessed image. Thus, these final crypt maps display a representation of visual blue staining in the crypt.
By this method the mutated (blue) clonal cell lineage is differentiated from the nonmutated (brown) cells and any nonstaining areas (the latter appear black on the final crypt maps). The crypt map thus shows the path taken by the mutated blue clone that can be fully visualized from crypt base to crypt luminal surface.
Mitochondrial DNA Mutations as a Marker of Clonal Expansion
It has recently been shown that human stem cells and their progeny contain nonpathogenic mutations in their mitochondrial DNA, including mutations in the cytochrome c oxidase (COX) gene (a component of complex IV of the respiratory chain), that are relatively common [12–14]. Each mitochondrion contains several hundred molecules of its own circular genome, and in most cells several thousand of these organelles exist. The mitochondrial genome is prone to mutation because of a lack of protective histones and poor DNA repair mechanisms. Mutations can expand stochastically within a cell, and over time cells will become either homoplasmic (all the mitochondria in the cell are mutated) or heteroplasmic (the cell contains a mixture of mutated and wild-type mitochondria). This stochastic expansion is a lengthy process, often taking many years , and for a mutated cellular phenotype to be observed, homoplasmy or a high degree of heteroplasmy must be present. Thus, probably only stem cells or possibly committed long-lived progenitors (Fig. 3B) are the only cells that have a sufficient life-span to accumulate these mitochondrial mutations to a level that results in a detectable biochemical deficiency.
Two-color enzyme histochemistry can be used to simultaneously detect activity of the mtDNA-encoded COX and nuclear DNA-encoded SDH, a component of complex II of the respiratory chain (Fig. 1). Cells lacking in COX activity appear blue; COX-active cells appear brown. Laser microdissection followed by polymerase chain reaction (PCR) coupled with mtDNA sequencing can determine the presence of clonal mitochondrial mutations, and this has been shown to be a reliable marker for observing the clonal expansion of mutated stem cells in the intestinal crypt and stomach [12–14].
Tracking Cell Clones in Intestinal Crypts
Figure 1 demonstrates that the expansion of stem cell clones in a crypt is not always straightforward from the base to the surface, hence the need to accurately map such clones. After histochemical staining for COX activity in the human intestine, three different types of crypts are observed: (a) wild-type brown crypts, (b) wholly mutated blue crypts, and (c) crypts containing populations of both brown and blue cells (mixed or partially mutated crypts) [12, 13]. These mixed crypts provide us with a unique opportunity to identify and map stem cells and the spread of their progeny within individual human intestinal crypts (Fig. 2). Figure 3 shows colonic and small intestinal crypt maps that include a variety of different patterns of spread of stem cell progeny within crypts. In the colonic crypt it can clearly be seen that the first mutated cells appear toward the base of the crypt: Figure 3A and 3C shows that the mutated clones arise just at or very close to the crypt base (which represents the likely site of the stem cell niche) and expand laterally to a varying degree as they migrate upward through the crypt to reach the very top. More complex patterns are also seen: Figure 3B shows a clone that appears to originate approximately halfway up the crypt. Of course, this could represent a mutated stem cell clone that has been lost from the niche by clonal succession with wild-type stem cells or it could represent a mutation occurring in a committed progenitor cell: by marking random intestinal epithelial cells by somatic mutation of the Dlb-1 locus in mice, Bjerknes and Cheng  were able to demonstrate short-lived (days) progenitors yielding one or two cell types and long-lived (months) mucus cell progenitors, columnar cell progenitors, and multipotential stem cells capable of giving rise to all epithelial cell types. Consequently, it is possible that the clone illustrated in Figure 3B results from a mutation in a committed mucus or columnar cell progenitor, a proposal that could be tested by lineage analysis of the cells in this clone. Thus, this method has the potential of identification of these elusive cells and also eventually of completing a lineage map of the crypt. Note that the pattern of clone expansion is not linear, as has been suggested previously  but that clones expand laterally as they migrate up the crypt, implying lateral expansion of the clone as stem cell progeny move through the putative committed progenitor region.
In the small intestine, however, the mutated clone appears to begin above the level of the Paneth cells (Fig. 3D), with no mutated cells being seen within the Paneth cell zone itself, suggesting that Paneth cells are long-lived and that, in the human, the niche is indeed sited immediately above the Paneth cell zone. Figure 3E shows that these COX-deficient cells can be tracked onto the villus surface. Thus, these maps allow the visualization of the spread of (presumed) stem cell progeny and also the demonstration of cell migration from committed progenitor cells, not possible in human epithelial tissues before. Combining these mapping methods with lineage analysis will allow considerably further insight into the biology of the system.
Slow Turnover Tissues: Liver and Pancreas
Immunohistochemical and double histochemical staining of normal human liver identified patches of COX-deficient hepatocytes located in close proximity to or in direct contact with the portal tract region (Fig. 4A–4C). The patches varied in size but had clearly defined boundaries, seemingly always extending to the perivenous regions, although not always recognized in two-dimensional sections (Fig. 4A). Interestingly, a segment of nearby biliary epithelium was also COX-deficient; if these cells and the adjacent patch of hepatocytes shared a similar mutation, then this result would lend weight to the belief that biliary cells are bipotential progenitors. Unfortunately, mtDNA sequencing was not possible on this fixed material. mtDNA sequencing of a COX-deficient patch observed in a frozen section (highlighted by blue SDH staining) revealed that an identical mutation was present in each cell of the patch (Fig. 4B–4E), thus demonstrating the clonal expansion of a population of cells within morphologically normal liver tissue. Because of the long time required for mtDNA mutations to become established in the population, it is likely that these negative patches represent the progeny of a long-lived cell type such as a stem cell or committed progenitor cell, because it has been estimated that liver has a turnover time on the order of 200 days. Clonal COX-negative patches showed no extra Ki-67 positivity and had normal expression patterns of albumin and CYP1A2 (data not shown), indicating that the mutated cells retain the synthetic and metabolic function of normal hepatocytes within the patch. The tumor marker α-fetoprotein was also negative, suggesting that these patches are not preneoplastic (data not shown).
Figure 5a shows a clonal population of human exocrine acinar cells that appears to be derived from the pancreatic duct or from an area adjacent to it. The cells in the duct have the same 4098A>G mutation as those in the neighboring patch of exocrine tissue, strongly indicating that they are of the same clonal origin. Patches of COX-negative cells could often be found located adjacent to interlobular ducts (Fig. 5B), and close inspection of the negative patch of acini strongly indicates that the acinar cells, as expected, have a separate cellular ancestry distinct from that of stellate cells, which here are COX-positive (Fig. 5C). Individual acini appeared to be clonally derived, because acini composed of both COX-positive and -deficient cells were never seen (Fig. 5C), but lobules were always a mixture of COX-positive and -negative acini, possibly indicating a polyclonal origin of lobules.
Origins of a Skin Tumor
Maintenance of the epidermal layer is achieved through a highly complex and orchestrated process, possibly involving distinct stem cell compartments in the hair follicle, interfollicular epidermis (IFE), and sebaceous glands . The common tumors that arise in the skin are the basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs), which show differences in differentiation pattern, arguably reflecting the stem cell pool from which they originate. Because of their resemblance to IFE, SCCs are thought by some to originate from the IFE stem cell pool. However, it has also been argued that they arise from the pilosebaceous follicle. BCCs are the most common skin tumors in humans; they can be invasive but do not metastasize.
Figure 6 shows COX deficiency throughout the outermost layer of the outer root sheath (ORS), but the layers beneath, including the companion layer, are COX positive, suggesting that the companion layer is not derived from the same stem cell population as the ORS. Sequence analysis of mtDNA from microdissected areas in the COX-deficient ORS showed a 13,740T>C mutation, whereas the inner root sheath (IRS) and the companion layer were wild type. These results could be interpreted as showing that the ORS is derived from bulge stem cells, whereas the IRS lineages are derived from a separate stem cell pool at the base of the hair follicle. Alternatively, this whole follicle could be in the process of monoclonal conversion from stem cells in the bulge area, with, in this follicle, only the ORS mutated. Further studies on many more follicles would confirm this: a common mutation within the ORS, companion layer, and IRS would mean a common origin from a multipotential stem cell, and serial section reconstruction of early mutated follicles should allow definition of the follicle niche within the bulge area, if indeed that is where this is housed.
The BCC showed a homogeneous distribution of COX deficiency (Fig. 7), which included the entire lesion and interestingly was also observed to include the tumor stroma. The pattern of COX deficiency, with its identical mutation profile across various tumor locations, strongly suggests that BCCs are typically monoclonal. COX deficiency extended throughout this BCC and into the histologically normal ORS of the hair follicle (Fig. 6): mutation analysis of COX-deficient areas from the tumor and ORS demonstrated the presence of a clonal 13,740 T>C mutation.
In this article we showed how the demonstration of clonal mtDNA mutations can be used to trace cell lineages in human epithelial tissues. Specifically we showed how this can be applied to the human colonic and small intestinal crypt, liver and pancreas, and skin. The monoclonal origins of the COX-deficient areas are based on the minimal odds of finding two random identical mutations occurring at the same site in two separate cells being calculated as [16,600:1 (genome size) × 3 (possible transition mutations)] × [16,600:1 × 3] = 2.48 × 109:1. Moreover, although the method only allows a snapshot in time, the technique does demonstrate the two major aspects of stemness, multipotentiality and self-renewal. In both the stomach  and small intestine , we have demonstrated all appropriate cell lineages within a clonally derived COX-deficient proliferative unit (gastric gland and crypt, respectively).
The organization and position of different cell types within the intestinal crypt have been studied extensively in the mouse. Intestinal stem cells are not identifiable morphologically, and, therefore, a variety of methods have been used to specifically demonstrate their location. In animals, current evidence suggests that the cells of the intestinal epithelium are arranged hierarchically, with the stem cell niche positioned toward the base of the crypt; as cells leave the neighborhood of the niche they become progressively more differentiated as they migrate toward the luminal surface. The specific location of the niche is believed to differ between the small intestine and colon, with the putative stem cell compartment at the base in the colonic crypt, but at cell positions 4-5 in the small intestine, above the basal Paneth cells . More recent studies in mice have questioned the exact location of the niche in the small intestine; Barker et al.  showed that the lgr5-positive cells, which are able to clonally replace the crypt, are found at the base of the crypt in the small intestine, mixed with the Paneth cells, but a similar approach in which a tamoxifen-activated Cre recombinase was knocked into the Bmi1 locus again produced long-lived clones populated by all four cell lineages . However, Bmi1 was expressed just above the Paneth cells at around cell position +4. Thus, in the mouse small intestine, the exact location of the niche is in dispute. In the present study we suggest that the niche is located at the base of the human colonic crypt and above the Paneth cell zone in the small intestine. Moreover, we describe a crypt mapping method through which patterns of expansion of stem cell clones can be analyzed, giving the prospect of the generation of a detailed fate map and also of identification of the stem cell, in addition to the analysis of the dynamics of niche succession in this important stem cell system.
The liver has a remarkable ability to activate its usually proliferatively quiescent parenchymal hepatocytes as a compensatory response to lost tissue mass. The ability of hepatocytes to clonally expand and repopulate damaged liver has been impressively demonstrated using hepatocyte transplant models in the mouse, with cells dividing at least 69 times without loss of function [22, 23]. Thus, hepatocytes can be considered to encompass the role of a unipotent stem cell. The liver can also stimulate a facultative stem cell compartment residing in the intrahepatic biliary tree (canal of Hering) when replication of hepatocytes is compromised. This results in ductules of biliary-type epithelia that emanate from the portal areas before hepatocyte differentiation. Thus, the most popular site identified as a possible niche is the periportal region . In the rat, Zajicek et al. . advanced the concept of the liver as a slowly renewing cell population of hepatocytes formed adjacent to the portal tract (the proposed niche), streaming en masse toward the terminal hepatic vein. We have highlighted the presence of defined populations of clonal cells in the normal human liver apparently originating from the periportal area. We therefore propose that the liver stem cell niche is periportally located, housing stem cells capable of giving rise to very large populations of hepatocytes. This suggestion is consistent with a recent report of DNA label-retaining cells in mouse liver being found within or close to the biliary ducts  and the observation of rare periportally located hepatocytes expressing embryonic stem cell-associated transcription factors in the human liver . It is also possible that the biliary epithelium can give rise to hepatocytes in the normal liver, as in the damaged liver, and proof of this concept lies in the demonstration of the same clonal mutation in the biliary epithelium and associated mutated hepatocytes. We have indeed observed a loss of COX expression in a segment of an interlobular bile duct adjacent to a COX-deficient patch of hepatocytes (Fig. 4A). Whether the two populations are from a common ancestor awaits mtDNA sequencing from a fresh sample exhibiting similar COX deficiency in the two cell types.
Murine studies have concluded that pancreatic islets lack stem cells. For example, a transient labeling of all insulin-expressing cells by a reporter gene failed to find evidence for the emergence of further insulin-producing cells from undifferentiated cells, even after severe pancreatic injury . Sequential administration of two different DNA labels showed that double-labeled cells were very rare, even after massive pancreatic damage, again leading to the conclusion that there was no pancreatic cell hierarchy . Similarly Brennand et al. , studying fluorescent label dilution in β-cells, found equal dilution across the β-cell population with time, interpreted in terms of no stem cell population for β-cells. Furthermore, we currently have little information about the lineage relationships between ductal and acinar cells in the human pancreas or about the possible location of any stem cell niche. From our initial observations presented here, we suggest that a stem cell niche exists in the human pancreas, either in the duct or close to it, which provides both ductal and acinar cell lineages. In view of the recent finding that Ngn3-expressing cells in mouse pancreatic ducts could differentiate into insulin-expressing cells after pancreatic injury , we propose that detailed observations of human pancreatic ducts and nearby islets for clonal mtDNA mutations could solve the long-standing question of whether β-cells can be derived from stem cells housed in the niche we have proposed for ductal and acinar cells. It has been suggested that neither liver nor pancreas possesses stem cells, but our finding of clonal populations in both argues for their existence.
In the skin, cells of the IFE and sebaceous glands follow a single differentiation pathway; however, hair follicles comprise eight different cell lineages: medullary, cortical, and cuticle cells comprise the hair shaft that is surrounded by the three cell lineages of the IRS. Encircling the IRS and separated by the companion layer is the outermost layer, the ORS. Controversy exists about the nature of stem cell populations in the hair follicle. Some studies indicate the presence of multiple stem cells in mouse cutaneous epithelium, some with restricted lineages in the absence of major injury . However, it is unclear in the human whether there is a common stem cell that gives rise to both the ORS and IRS. The origin of the companion layer and its attribution to either the IRS or the ORS is also subject to debate: analysis of keratin expression patterns ascribed it to the IRS as its outermost layer .
A widely accepted location of hair stem cells is in the hair bulge, a specialized region in the ORS . According to this unitarian hypothesis, bulge stem cells form the ORS and, in addition, move along the ORS to the matrix of the hair bulb, where they function as progenitor cells to replenish the IRS and hair shaft at the end of each hair growth cycle . However, there may be multiple stem cell populations within the follicle itself , including actively cycling lgr5-positive cells , and lineage tracing studies have indicated that even the matrix contains restricted self-renewing stem cells for each inner structure . These conclusions have all been drawn from studies on mouse epidermis. Our finding here that the ORS and an adjacent BCC share the same mutation strongly indicates that the tumor has arisen from the ORS. It has often been suggested that BCCs may originate in the stem cells of the hair follicle bulge, but this theory has never been confirmed experimentally: here we show a BCC that exhibited a distinct pattern of COX deficiency, displaying a continuity extending from a histologically normal ORS of a hair follicle to the entire tumor. The common origin of this monoclonal phenomenon was confirmed by mutation analysis of various COX-deficient areas. These findings provide evidence for a common origin of BCCs and the ORS in the stem cells of the hair bulge. Thus, this technique offers a unique opportunity of analyzing the tissues or even the cell of origin of human tumors.
We argue that the method described in this article has considerable potential for cell lineage tracing and for the subsequent identification of human stem cells and their niches, especially, but not exclusively, in epithelial tissues, in which organized architecture is the rule. We also see applications for studying the origins of human tumors. Limitations of the method include the fact that such mutations take some time to be established and thus observations so far have been confined to individuals older than age 40. There is also the limitation that snap-frozen tissue is necessary for definitive analysis by double-enzyme histochemistry and mtDNA mutation analysis, although presumptive observations can be made using immunohistochemistry. Thus, it would be sensible for the scientific community to pool such important tissue resources to advance our knowledge of human stem cells. Moreover, if robust stem cell markers can be determined for the various tissues, then we would expect that there should be overlap with the earliest COX-deficient cells in our lineage maps.
This work was supported by Barts and The London Charity (T.G.F), the Wellcome Trust (S.I), Cancer Research UK (J.B. and A.H), CORE (formerly Digestive Disorders Foundation) (S.A.C.M), and Fundacion Mutua Madrileña (L.G.-G). Liver tissue was provided by Ajit T. Abraham of Barts and The London HPB Centre.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.