Human hair genealogies and stem cell latency
© Kim et al; licensee BioMed Central Ltd. 2006
Received: 16 October 2005
Accepted: 03 February 2006
Published: 03 February 2006
Stem cells divide to reproduce themselves and produce differentiated progeny. A fundamental problem in human biology has been the inability to measure how often stem cells divide. Although it is impossible to observe every division directly, one method for counting divisions is to count replication errors; the greater the number of divisions, the greater the numbers of errors. Stem cells with more divisions should produce progeny with more replication errors.
To test this approach, epigenetic errors (methylation) in CpG-rich molecular clocks were measured from human hairs. Hairs exhibit growth and replacement cycles and "new" hairs physically reappear even on "old" heads. Errors may accumulate in long-lived stem cells, or in their differentiated progeny that are eventually shed.
Average hair errors increased until two years of age, and then were constant despite decades of replacement, consistent with new hairs arising from infrequently dividing bulge stem cells. Errors were significantly more frequent in longer hairs, consistent with long-lived but eventually shed mitotic follicle cells.
Constant average hair methylation regardless of age contrasts with the age-related methylation observed in human intestine, suggesting that error accumulation and therefore stem cell latency differs among tissues. Epigenetic molecular clocks imply similar mitotic ages for hairs on young and old human heads, consistent with a restart with each new hair, and with genealogies surreptitiously written within somatic cell genomes.
Stem cells are few in number, but it is possible to infer how often they divide by measuring the genomes of their more abundant differentiated descendants. The mitotic age of a differentiated cell can be defined as the total numbers of divisions since conception, which includes the three genealogical phases. Development and differentiation are highly regulated, and numbers of divisions during these phases are similar regardless of age. In contrast, stem cells may divide throughout life. Therefore, any increase in average mitotic age primarily depends on stem cell division.
To apply this logic, we infer mitotic age by measuring the errors that occur during genome replication. The idea is that replication errors surreptitiously record cell divisions, the number of errors being proportional to the number of divisions. Such molecular clock approaches are commonly employed to trace species ancestry , but conceivably somatic cell genealogies are also recorded within genomes. Sequences are initially similar in all cells but will become increasingly polymorphic owing to rare and random replication errors. In theory, ancestry may be inferred from polymorphisms sampled from any population, including somatic cells. Sequences rarely mutate, but epigenetic errors also exhibit somatic inheritance and methylation measurably accumulates at certain CpG-rich regions during normal human aging [2, 3]. The 5' to 3' order of CpG methylation provides binary information strings (i.e. "0" for unmethylated and "1" for methylated) or tags analogous to sequences, which appear to function as somatic molecular clocks . In general, percent methylation is equivalent to mitotic age because our tags, like most CpG islands , are initially unmethylated.
Human intestinal studies revealed an age-related increase in tag methylation, suggesting that crypt stem cells divide throughout life [3, 5]. Although cell division appears to be required for de novo in vitro CpG island methylation , some in vivo methylation could also occur in non-dividing cells. Although it is difficult to test directly whether tag methylation primarily represents replication errors, such errors probably occur in all cells and should record genealogies in many tissues. Human hair can provide a further test of such an epigenetic molecular clock because its biology suggests much greater stem cell latency.
Hair is periodically shed because follicles exhibit cycles of growth (anagen), degeneration (catagen) and rest (telogen). Differentiated cells including the matrix and inner root (IRS) and outer root sheath (ORS) keratinocytes are shed when the lower follicle physically disappears during catagen [7, 8]. A new hair cycle is thought to be initiated after telogen from stem cells located in the bulge. According to a bulge activation hypothesis , stem cells briefly divide at the start of anagen to produce differentiated progeny that reconstruct and maintain a new lower follicle. Most divisions occur in differentiated cells that die at the end of anagen whereas stem cells seldom divide and survive each cycle.
This biology suggests hair genealogies with relatively few stem cell divisions, since most divisions occur in differentiated cells during anagen, and greater mitotic ages in longer hairs. Variations among hairs on a single individual will primarily reflect differences in anagen divisions because lower follicle mitotic activity is high, cycles are asynchronous, and anagen lasts a variable number of years [8, 10]. Age-related methylation should be minimal to undetectable because of relative bulge stem cell quiescence. However, hair biology has mainly been studied in mice, and human anagen cycles are longer (years versus weeks) and usually recur for decades. Relatively little is known about human hair stem cell dynamics because experimental manipulations are impractical.
Average hair methylation does not increase beyond infancy
Seven scalp hairs were randomly plucked from the vertex of 28 individuals of different ages. The plucked hairs were primarily in anagen and contained differentiated IRS and ORS cells, though some matrix cells may be present [10, 12]. Methylation patterns were sampled from each hair using bisulfite sequencing of two CpG rich regions or tags (Figure 1). One tag called CSX contains 8 CpG sites in the 3' untranslated region of the CSX gene and has been used to study the intestines . The other tag is in the promoter region of the SOX10 gene and contains 20 CpG sites.
Average CSX hair methylation varies during the hair cycle
Short, long, pigmented and white hairs
CSX % Methylation
SOX10 % Methylation
Average methylation is not significantly different between pigmented and white hairs
Methylation was compared between white and pigmented hairs from three individuals (Figure 3). There were no significant differences, suggesting similar cell kinetics (Table 1). SOX10 is important for melanin production  and the extensive SOX10 promoter methylation (>80%) in pigmented and white hair implies lack of expression. To test whether SOX10 is methylated in melanin producing cells, we examined three cell lines. SOX10 was unmethylated (<2%) in two pigmented melanoma cell lines (M624 and M888) and methylated (95%) in a breast cancer cell line (MCF7). CSX was extensively methylated (>90%) in all three tumor cell lines, consistent with multiple divisions before and after transformation.
Validation and reproducibility
Reproducibility was tested by resampling eight more bacterial clones from the same PCR/cloning experiment, or by repeating the PCR on the same hair and sampling eight more clones (Figure 4B). Although methylation patterns were not identical after resampling, average hair percent methylation values were similar (correlation coefficients of 0.86 for resampling the same PCR/cloning and 0.93 for repeat PCR of the same hair). Therefore, a sample of eight alleles appears to approximate the complex methylation patterns likely to be present in a human hair follicle.
Human cell fate studies are hampered by the inability to mark and follow individual cells experimentally. Here we employ a molecular clock strategy to reconstruct human hair genealogies from random epigenetic replication errors. Contemporary sequences are routinely used to reconstruct species ancestry, with times since a common ancestor proportional to numbers of sequence differences . By analogy, all cells are related through a single, continuously expanding somatic cell tree, and stem cells are common ancestors (Figure 1). For example, the zygote is the stem cell or common ancestor for all somatic cells, and a hair stem cell would produce differentiated follicle cells. Depending on genealogy, there may be multiple common ancestors or stem cells per follicle. Hair provides a well-defined empirical test of an epigenetic molecular clock because its genealogy would be consistent with only a limited number of potential methylation patterns.
The seemingly "random" methylation tag patterns observed in hair (Figure 1) are consistent with stochastic replication errors. Some of the scatter of average hair values is probably secondary to the small number of alleles sampled per hair, but control experiments illustrate reproducibility (Figure 4B). The ability to sample reproducibly a human hair follicle containing thousands of cells with just eight tags may reflect that follicle cells are closely related to each other. The lack of an age-related increase in average hair methylation is consistent with the bulge activation hypothesis , in which stem cells primarily divide at the start of anagen to produce progeny that reconstruct and maintain a new lower follicle. Replication errors in differentiated cells cannot accumulate because they are discarded when the lower follicle degenerates after several years. A new cycle begins after several weeks, and new hairs in young or old individuals would have similar mitotic ages if bulge stem cells are normally latent and only divide a few times every cycle.
Hair neogenesis or development precedes the stem cell phase. Human follicle neogenesis starts late in the first trimester [14, 15]. Stem cells appear to originate from rapidly dividing hair peg germ cells that grow fine lanugo hairs, which fall out around birth to be replaced by adult-type terminal scalp hairs. Consistent with the idea that bulge stem cells are descendents of mitotic germ cells, methylation rapidly increased from low levels before birth. Average tag methylation attained maximum or adult average values about two years after birth (Figure 2), suggesting bulge stem cell latency is not fully established until after infancy. SOX10 is a faster clock than CSX, and the lack of unmethylated SOX10 tags in older individuals suggests that most bulge stem cells had previously divided before becoming latent after infancy. High mitotic activity followed by latency would allow for the label-retaining hallmark of bulge stem cells.
To characterize the differentiated phase, hair length was employed as a surrogate for anagen cycle length. Consistent with new hairs emerging from descendants of infrequently dividing bulge stem cells, average CSX methylation was significantly lower for short versus long hairs and average diversity was also significantly lower in short hairs, suggesting a relative population bottleneck. However, there was no consistent relationship because some long hairs had very low methylation levels and some short hairs had very high ones. This variation may reflect the inability definitively to select hairs of different lengths, which can be confounded by hair breakage (short hairs may be broken long hairs), cycle length heterogeneity between hairs, and the variable cellular compositions of plucked hairs . Variation in methylation with respect to hair length may also reflect anagen bulge cell migration. Transplanted bulge progeny in whiskers migrate down the ORS throughout anagen , which could counteract a steady increase in average anagen hair mitotic age if these immigrants have lower mitotic ages. Anagen bulge migration may not necessarily require new stem cell divisions because these cells may originate from the suprabasal bulge, a region with the early progeny of basal bulge stem cells . In addition, cells from the previous anagen cycle (secondary hair germ) may survive through telogen and help reconstitute the new follicle .
White hairs were slightly more methylated than pigmented hairs from the same heads, possibly suggesting more lifetime divisions in white hairs. However, this difference was not statistically significant and further studies are necessary to determine whether mitotic ages differ between pigmented and white hairs. SOX10 is important for melanin production  and the SOX10 promoter methylation (>80%) in both pigmented and white hair, but not in melanoma cell lines, implies lack of expression . The melanocyte lineage appears to be infrequent in plucked hairs because only one of 234 SOX10 tags sampled after two years of age was unmethylated like the melanoma cell lines. Therefore, consistent with separate stem cell origins for melanocytes and other follicle cells , epithelial lineages appear distinct from the melanocyte lineage.
Niche structure may help explain different mitotic ages (Figure 5). Intestinal mitotic compartments, located at crypt bases, are preserved throughout life. Although most mitotic intestinal cells are considered to be transit amplifying rather than stem cells , dividing cells may persist as long as their mitotic compartment or niche persists. Without a bottleneck mechanism to eliminate mitotic cells, they may accumulate errors throughout life and effectively function as stem cells or common ancestors because they produce the majority of cells within a crypt. By contrast, the primary hair mitotic compartment (lower follicle) is physically destroyed every few years, which eliminates most dividing cells and their accumulated errors. Quiescent bulge stem cells can function as common ancestors because they briefly divide and replenish the lower follicle at critical periods after such bottlenecks. Configurations physically differ between stem cell niches , which could influence the genealogies of their tissues.
Age-related changes in methylation or mitotic ages appear to depend on how often stem cells divide. Although the stochastic nature of replication errors may hinder the interpretation of molecular clocks, and many questions remain , such approaches provide information that is difficult to obtain by other more precise methods. Further studies are needed to define human somatic molecular clocks better, and more information can be extracted with more data and formal quantitative modeling. Here we infer human stem cell genealogies from present day methylation patterns consistent with biology reconstructed by more direct methods, primarily in mice. Epigenetic molecular clocks are experimentally much more efficient than sequence somatic clocks because mutation frequencies in tumor cell lines are about one per million bases  whereas multiple epigenetic errors are readily found after sequencing only a few hundred bases (Figure 1). Replication errors are synonymous with division, and somatic molecular clocks that count these errors potentially facilitate the systematic reconstruction of many human tissue genealogies, allowing a fundamental distinction between old and new cells within the same individual.
Randomly chosen vertex scalp hair follicles were obtained by plucking from 28 individuals of different ages. This research was approved by the Health Sciences Institutional Review Board at the University of Southern California. Most of these hairs were in anagen because telogen hairs (obtained by pulling rather than plucking ) generally did not contain sufficient DNA for amplification (data not shown). For CSX, seven hairs each with eight tags per hair were sampled from each head. For SOX10, four to five hairs, each with six tags per hair, were sampled. Long hairs were greater than 40 cm and short hairs were shorter than 4 cm on the same head. DNA was extracted by placing the follicle in 10 μl of 10 mM Tris-HCl, 2 mM EDTA (pH 8.0) with 20 mg/ml Proteinase K for two hours at 56°C. The solution was boiled for 5 min and DNA was bisulfite converted using an agarose bead method . The final bead volume was about 25 μl. Approximately 10% (2.5 μl) of the bead was amplified with 42 PCR cycles using primers specific for CpG rich regions or tags in the CSX and SOX10 genes (Figure 1). Each follicle was amplified in duplicate and PCR products were combined prior to cloning (TOPO TA Cloning kit, Invitrogen, Carlsbad, CA). Tags with evidence of incomplete bisulfite conversion (Cs at non-CpG sites) were discarded from the analysis. Average hair methylation was defined as the average number of methylated sites among tags from each follicle. Follicle diversity of a hair was defined as the number of unique tags found among the sampled tags.
Methylation of the first five CSX CpG sites was also measured with pyrosequencing . The biotinylated primer was 5' -ACACCAAACTACAAAATCACTCATTA-3' and the pyrosequencing primer was 5' -TAGGAATTTTTTTTGTTTTA-3'. A suitable pyrosequencing primer could not be located for the last 3 CpG sites. The assay was performed on a PSQ HS 96 Pyrosequencing System with conditions recommended by the manufacturer (Pyrosequencing, Inc., Westborough, MA).
This work was supported by a grant from the National Institutes of Health (DK61140). ST is a Royal Society-Wolfson Research Merit Award holder, and is supported in part by a grant from CRUK. We thank the many individuals who donated their hair.
- Bromham L, Penny D: The modern molecular clock. Nat Rev Genet. 2003, 4: 216-224. 10.1038/nrg1020.View ArticlePubMedGoogle Scholar
- Issa JP: CpG-island methylation in aging and cancer. Curr Top Microbiol Immunol. 2000, 249: 101-118.PubMedGoogle Scholar
- Yatabe Y, Tavaré S, Shibata D: Investigating stem cells in human colon by using methylation patterns. Proc Natl Acad Sci USA. 2001, 98: 10839-10844. 10.1073/pnas.191225998.PubMed CentralView ArticlePubMedGoogle Scholar
- Bird A: DNA methylation patterns and epigenetic memory. Genes Dev. 2002, 16: 6-21. 10.1101/gad.947102.View ArticlePubMedGoogle Scholar
- Kim JY, Siegmund KD, Tavaré S, Shibata D: Age-related human small intestine methylation: evidence for stem cell niches. BMC Med. 2005, 3: 10-10.1186/1741-7015-3-10.PubMed CentralView ArticlePubMedGoogle Scholar
- Velicescu M, Weisenberger DJ, Gonzales FA, Tsai YC, Nguyen CT, Jones PA: Cell division is required for de novo methylation of CpG islands in bladder cancer cells. Cancer Res. 2002, 62: 2378-2384.PubMedGoogle Scholar
- Fuchs E, Tumbar T, Guasch G: Socializing with the neighbors: stem cells and their niche. Cell. 2004, 116: 769-778. 10.1016/S0092-8674(04)00255-7.View ArticlePubMedGoogle Scholar
- Paus R, Foitzik K: In search of the "hair cycle clock": a guided tour. Differentiation. 2004, 72: 489-511. 10.1111/j.1432-0436.2004.07209004.x.View ArticlePubMedGoogle Scholar
- Sun T-T, Cotsarelis G, Lavker RM: Hair follicular stem cells: the bulge-activation hypothesis. J Invest Dermatol. 1991, 96: 77S-78S. 10.1111/1523-1747.ep12471959.View ArticlePubMedGoogle Scholar
- Sperling LC: Hair anatomy for the clinician. J Am Acad Dermatol. 1991, 25: 1-17.View ArticlePubMedGoogle Scholar
- Colella S, Shen L, Baggerly KA, Issa JP, Krahe R: Sensitive and quantitative universal pyrosequencing methylation analysis of CpG sites. Biotechniques. 2003, 35: 146-150.PubMedGoogle Scholar
- Moll I: Proliferative potential of different keratinocytes of plucked human hair follicles. J Invest Dermatol. 1995, 105: 14-21. 10.1111/1523-1747.ep12312406.View ArticlePubMedGoogle Scholar
- Mollaaghababa R, Pavan WJ: The importance of having your SOX on: role of SOX10 in the development of neural crest-derived melanocytes and glia. Oncogene. 2003, 22: 3024-3034. 10.1038/sj.onc.1206442.View ArticlePubMedGoogle Scholar
- Holbrook KA, Minami SI: Hair follicle embryogenesis in the human. Characterization of events in vivo and in vitro. Ann N Y Acad Sci. 1991, 642: 167-196.View ArticlePubMedGoogle Scholar
- Akiyama M, Smith LT, Shimizu H: Changing patterns of localization of putative stem cells in developing human hair follicles. J Invest Dermatol. 2000, 114: 321-327. 10.1046/j.1523-1747.2000.00857.x.View ArticlePubMedGoogle Scholar
- Oshima H, Rochat A, Kedzia C, Kobayashi K, Barrandon Y: Morphogenesis and renewal of hair follicles from adult multipotent stem cells. Cell. 2001, 104: 233-245. 10.1016/S0092-8674(01)00208-2.View ArticlePubMedGoogle Scholar
- Blanpain C, Lowry WE, Geoghegan A, Polak L, Fuchs E: Self-renewal, multipotency, and the existence of two cell populations within an epithelial stem cell niche. Cell. 2004, 118: 635-648. 10.1016/j.cell.2004.08.012.View ArticlePubMedGoogle Scholar
- Panteleyev AA, Jahoda CA, Christiano AM: Hair follicle predetermination. J Cell Sci. 2001, 114: 3419-3431.PubMedGoogle Scholar
- Egger G, Liang G, Aparicio A, Jones PA: Epigenetics in human disease and prospects for epigenetic therapy. Nature. 2004, 429: 457-463. 10.1038/nature02625.View ArticlePubMedGoogle Scholar
- Nishimura EK, Jordan SA, Oshima H, Yoshida H, Osawa M, Moriyama M, Jackson IJ, Barrandon Y, Miyachi Y, Nishikawa S: Dominant role Dominant role of the niche in melanocyte stem-cell fate determination. Nature. 2002, 416: 854-860. 10.1038/416854a.View ArticlePubMedGoogle Scholar
- Kim KM, Calabrese P, Tavaré S, Shibata D: Enhanced Stem Cell Survival in Familial Adenomatous Polyposis. Am J Pathol. 2004, 164: 1369-1377.PubMed CentralView ArticlePubMedGoogle Scholar
- Potten CS, Loeffler M: Stem cells: attributes, cycles, spirals, pitfalls and uncertainties. Lessons for and from the crypt. Development. 1990, 110: 1001-1020.PubMedGoogle Scholar
- Komuro I, Izumo S: A murine homeobox-containing gene specifically expressed in the developing heart. Proc Natl Acad Sci USA. 1993, 90: 8145-8149.PubMed CentralView ArticlePubMedGoogle Scholar
- Kim JY, Tavaré S, Shibata D: Counting human somatic cell replications: Methylation mirrors endometrial stem cell divisions. Proc Natl Acad Sci USA. 2005, 102: 17739-17744. 10.1073/pnas.0503976102.PubMed CentralView ArticlePubMedGoogle Scholar
- Ro S, Rannala B: Methylation patterns and mathematical models reveal dynamics of stem cell turnover in the human colon. Proc Natl Acad Sci USA. 2001, 98: 10519-10521. 10.1073/pnas.201405498.PubMed CentralView ArticlePubMedGoogle Scholar
- Wang TL, Rago C, Silliman N, Ptak J, Markowitz S, Willson JK, Parmigiani G, Kinzler KW, Vogelstein B, Velculescu VE: Prevalence of somatic alterations in the colorectal cancer cell genome. Proc Natl Acad Sci USA. 2002, 99: 3076-3080. 10.1073/pnas.261714699.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.