Department of Pathology, Norris Cancer Center, University of Southern California School of Medicine, Los Angeles, California, USA

Correspondence to: D Shibata, 1200 N. State St., Unit I, Room 2428, Los Angeles, CA 90033, USA; E-mail: dshibata@hsc.usc.edu

Little it known about human stem cells although they are likely to be the earliest progenitors of carcinomas. Just as methylation can substitute for mutations to inactivate tumor suppressor genes, methylation can also substitute for mutations in a phylogenetic analysis. This review explains why stem cell dynamics may be important to tumor progression and how methylation patterns found in a normal human colon can be used to reconstruct the behavior of crypt stem cells. Histories are recorded in sequences and strategies used to reconstruct phylogenies from sequences likely apply to methylation patterns because both exhibit somatic inheritance. Such a quantitative analysis of colon methylation patterns infers stem cells live in niches containing multiple 'stem' cells. Although niche stem cell numbers remain constant, clonal succession is inherent to niches because periodically progeny from a single stem cell become dominant. These niche succession cycles may potentially accumulate multiple alterations because they resemble superficially the clonal succession of tumor progression except that they occur invisibly in the absence of selection or phenotypic change. Alterations without immediate selective value may hitchhike passively in the stem cells that become dominant during niche succession cycles. The inherent ability of a niche to fix alterations (Muller's ratchet) is another potential mechanism besides instability and selection to sequentially accumulate multiple alterations. Many alterations found in colorectal tumors may reflect such occult clonal progression in normal colon.

Oncogene (2002) 21, 5441-5449. doi:10.1038/sj.onc.1205604

colorectal cancer; progression; stem cells; niche; crypt

This review will focus on using methylation patterns to reconstruct stem cell fates or phylogenies in normal human colon. It will describe why such phylogenies may be useful for understanding tumor progression, how to trace phylogenies with methylation patterns, and implications of certain stem cell dynamics for cancer. To appeal to a general audience, the quantitative analysis required to reconstruct phylogenies will be referenced but only superficially outlined.

Why normal stem cell dynamics may be important to tumor progression

Multiple alterations are thought to accumulate in a stepwise during tumor progression (Nowell, 1976). A well-known progression model is the adenoma-cancer sequence in which alterations are associated with phenotypic progression to colorectal cancer (Fearon and Vogelstein, 1990). One dilemma of tumor models is defining or recognizing a 'start' to progression (Foulds, 1954; Nowell, 1976). A visible start may be defined by so called 'gatekeeper' mutations (Kinzler and Vogelstein, 1996) or the initial alterations that allow detectable clonal expansion. However, it has become increasingly evident that many alterations found in cancers may also be found in phenotypically normal cells. For example, the intestines of mice and men with germline mutations in genes thought to have critical roles in cancers may be tumor prone but are otherwise morphologically normal (Table 1). Presumably the same mutations acquired somatically would also sometimes not alter the phenotype of normal cells.

If alterations always changed phenotype, then genetic and phenotypic progression would start together. Genetic progression is defined here as the accumulation of alterations that ultimately confer survival or growth advantages to a cancer cell. Observations that mutations critical to cancer phenotypes (Hanahan and Weinberg, 2000) may also be found in normal cells (Table 1) suggest that genetic progression may sometimes start before phenotypic progression. Somatic alterations that accumulate before gatekeeper mutations would be difficult to directly detect because it would be nearly impossible to identify or isolate the rare phenotypically normal cells containing the alterations. A different sort of experimental approach may be necessary to characterize how occult alterations accumulate in normal colon. What do cancer progenitors do before they become capable of tumorigenesis?

Normal colon crypts

Normal human colon consists of millions of crypts (Figure 1). Crypts contain about 2000 cells (Cheng et al., 1984; Potten and Loeffler, 1990; Potten et al., 1992; Booth and Potten, 2000) and are maintained by stem cells located near or at the bottom of each crypt. The exact number and phenotypic characteristics of crypt stem cells are unknown, but murine studies suggest that crypts are maintained by multiple stem cells (Potten and Loeffler, 1990; Booth and Potten, 2000). Differentiated cells produced by stem cells migrate towards the surface of the colon and subsequently die. Epithelial cell turnover is rapid and essentially all cells except stem cells are replaced within a week. Alterations acquired in non-stem cells will be quickly lost and therefore multiple alterations likely accumulate only in stem cells (Cairns, 1975).

Stem cells typically divide asymmetrically to yield another stem cell and a cell destined to differentiate. Stem cells may be immortal or defined extrinsically by a niche (Figure 2). Niches are specialized regions that instruct cells within them to function as 'stem' cells (Spradling et al., 2001). Immortal stem cells always divide asymmetrically and therefore their lineages never become extinct. In contrast, 'stem' cells in niches may sometimes expand by producing two daughters that remain within the niche, or become extinct by producing two daughters that leave the niche and differentiate.

Distinguishing between immortal stem cells and stem cells maintained by niches is difficult because asymmetrical division usually also occurs in niches. However, rare stem cells loss or expansion can be detected with certain experimental scenarios (Figure 3). Crypts can be made heterogeneous for histologically detectable markers through chimeric mice or by mutagenesis (Potten and Loeffler, 1990; Booth and Potten, 2000). Such heterogeneous crypts should persist with immortal stem cells but will become homogeneous (for one or the other marker state) with a niche because of its inherent random stem cell extinction and expansion. Murine studies are consistent with crypt stem cells maintained by niches (Potten and Loeffler, 1990; Williams et al., 1992; Booth and Potten, 2000). Niches have also been observed in recent studies of Drosophila (Spradling et al., 2001).

By analogy to mice, human colon crypts are also likely maintained by stem cell niches. Studies of crypt heterogeneity in human colons after therapeutic radiation are consistent with stem cell niches (Campbell et al., 1996). However, one potential problem of interpreting these studies is that the manipulations such as mutagenesis required for the induction of the crypt heterogeneity needed to observe stem cells, may alter the normal behavior of stem cells (Potten and Loeffler, 1990). For example, stem cells that are normally immortal may expand to repair the damage induced by mutagenesis. In this case, the observed niche behavior reflects abnormal tissue damage rather than normal crypt maintenance. Moreover, the experimental manipulations used in murine stem cell studies are impractical for human studies.

Observing stem cells in normal human colon through methylation patterns

Detection of a stem cell niche requires an ability to distinguish between cells. Crypt cells are morphological identical and therefore visual detection of stem cell expansion or extinction is impossible. A seminal observation indicating that morphologically identical stem cells may be distinguished from each other was that methylation of certain CpG islands increases with age in normal human colon (Ahuja et al., 1998; Issa, 2000). If this age-related methylation occurs independently in different cells, then methylation patterns may differ between cells in a single individual.

Intact individual colon crypts can be isolated from fresh human colons (Cheng et al., 1984). DNA isolated from single crypts can be bisulfite treated, and their methylation patterns read after PCR and sequencing of individual cloned PCR products (Yatabe et al., 2001). Examples of methylation patterns sampled from single crypts of a single colon are illustrated in Figure 4. Methylation patterns are different between crypts and between cells within the same crypt, indicating that methylation patterns can be used to distinguish between cells. The data is complex and intuitively useless because patterns appear to randomly arise. However, just like random sequence errors record evolutionary histories, random methylation may record somatic cellular histories. The quantitative approaches used to trace phylogenies from sequences may apply to methylation because both exhibit somatic inheritance.

A quantitative analysis of age-related methylation is consistent with crypts maintained by niches containing multiple stem cells (Yatabe et al., 2001). Ignoring the mathematics (for more details see sidebar), a simple summary is that methylation patterns should differ between crypts with immortal stem cells or crypt niches. However, within a crypt, patterns should be very different with immortal stem cells but more similar for a stem cell niche. Methylation patterns as in Figure 4 are more consistent with crypts maintained by stem cell niches. Random stem cell loss with replacement means that periodically all niche lineages except one become extinct. This niche clonal succession counteracts further methylation changes to ensure that crypt cells are always relatively closely related.

Implications of stem cell niches to tumor progression

The detection of stem cell niches from crypt methylation patterns in normal human colon is not surprising considering prior mouse and human studies. Methylation studies (Yatabe et al., 2001), however, indicate that niche succession cycles occur continuously in normal human colon and are not experimental artifacts because no prior manipulations are required to read the cellular histories recorded in methylation. Although visually nothing changes in normal colon, clonal succession is inherent to a niche. It appears that approximately every 8.2 years (95% confidence intervals of 2.7 to 19 years) all stem cell lineages within a crypt niche except for one become extinct (Yatabe et al., 2001). By this model, each crypt in an 82 year-old colon has undergone on average 10 succession cycles.

Clonal succession is usually associated with tumor progression. However, a stem cell niche is also synonymous with clonal succession except it occurs invisibly without a change in phenotype and is not driven by mutation. This natural niche rhythm provides another mechanism other than selection to accumulate alterations through sequential clonal successions because the genotype of the dominant stem cell becomes the genotype of the crypt (Figure 5). Instead of driving clonal succession, early alterations, including those critical to tumorigenesis that do not yet appear to confer immediate growth advantages in normal cells (Table 1), may passively hitchhike to clonal dominance along with these niche succession cycles.

Crypts could become increasing 'fit' if differences between stem cells lead to increasing capable successors. However, the accumulation of alterations in small populations has been examined for decades and a somewhat intuitively opposite conclusion is that despite selection, fitness usually decreases over time without sex. Most mutations are deleterious and one benefit of sex may be to reduce mutation burden. A niche with its small and finite asexual population size would be prone to Muller's ratchet (Muller, 1964). Muller proposed that an asexual population inevitably accumulates deleterious mutations because of a ratchet-like irreversible loss of individuals with fewer mutations (Muller, 1964; Felsenstein, 1974; Smith and Nee, 1990; Chao, 1997). Muller's ratchet is exacerbated in smaller populations (Chao, 1997) because selection effects (positive or negative) on fixation are decreased relative to drift (defined as random sampling in a finite population). Niches decrease the onerous requirement that each new mutation associated with progression confers a selective growth advantage, and potentially fix even deleterious alterations that only later contribute to a tumor phenotype. The physical partition of stem cells into small niches may overall decrease the risk of cancer (Cairns, 1975) because of the fitness decline predicted by Muller's ratchet, although asexual populations sometimes become more fit (Chao, 1990). Muller's ratchet has been evoked to explain gene loss on Y-chromosomes (Charlesworth, 1978), which superficially resembles losses of heterozygosity commonly found in cancers.

Similar to tumor progression, multiple alterations in normal crypts could accumulate in a stepwise process during sequential niche succession cycles. Differences between progression associated with tumors or normal crypt niches are outlined in Table 2. A niche progression sequence (Figure 5) illustrates clonal succession to a visible tumor and incorporates many tenets of multistep progression (Nowell, 1976). Alterations accumulate in a series of clonal successions except selection, phenotypic changes, and abnormal clonal expansions are not required. Instead of sequential alterations that progressively increase the fitness and frequency of their clones, alterations may accumulate regardless of immediate selective value and only the final combination is critical.

The efficiency of progression may be greatly altered whether alterations occur in cells that allow hitchhiking compared to cells that prohibit hitchhiking. Mutations rarely occur in normal cells and the probability that a single cell will accumulate multiple mutations is extremely low (Loeb, 1991). However, like the clonal succession of tumor progression, which increases the probability of further hits because more cells are at risk (Tomlinson et al., 1996), alteration followed by niche succession may also expedite further progression because multiple stem cells are present in a niche. Estimates of human niche stem cell numbers range from four to 512 (Yatabe et al., 2001). Therefore, progeny of a cell with an alteration may have up to a 500-fold higher chance of acquiring another alteration after attaining clonal dominance in a niche compared to cells that remain single.

The gladiator-like life or death dynamics inherent to a niche are illustrated for APC loss (Figure 6). Crypts in a normal individual are wild type or APC+/+. By either mutation or methylation, one APC allele is inactivated in one stem cell. The fate of this APC+/- stem cell must be decided by the end of the next succession cycle because only one lineage survives the niche arena. It will either become fixed or extinct. Most of the time this APC+/- stem cell will be lost because most crypt lineages become extinct. However, sometimes the niche will be completely populated by offspring from this APC+/- cell, essentially a conversion into a tumor-prone familial adenomatous polyposis type crypt.

A teleological temptation is to attribute succession to a selective advantage, but because niche cycles occur in all crypts, Muller's ratchet suggests that change or drift may be more important in the choice of a victor. Alteration alone is insufficient for niche progression because of the high probability of extinction. Alteration linked to succession is irreversible progression, equivalent to a click of Muller's ratchet. The vast majority of crypts will not collect specific alterations because replication fidelity is high. However, a few crypts out of millions of colon crypts may collect multiple, potentially tumorigenic, hitchhiker alterations.

Clonal succession due to a niche succession cycles may occur in many normal tissues other than the colon because most mammalian self-renewing tissues appear to be maintained by stem cell niches (Watt and Hogan, 2000). Multiple alterations are found in most epithelial cancers including those from the breast, lung, prostate, and bladder. Many mutations found in these cancers are also compatible with normal phenotypes in mouse models (Table 1 and Pearson (2002)), therefore also allowing for the possibility that such somatic mutations may invisibly hitchhike to clonal dominance during niche succession cycles. The niches in different tissues likely differ with respect to stem cell numbers, division rates, and succession rates, and these differences in niche dynamics may ultimately help explain why cancer frequencies and epidemiology differs between tissues.

Hitchhikers or hijackers?

The inherent ability of a niche to collect and fix alterations through drift regardless of immediate selective value adds another mechanism besides instability and selection for progression through sequential clonal successions. This niche potential is inferred and invisible because 'normal' cells are morphologically indistinguishable. The conversion from an APC+/+ to an APC+/- crypt would be invisible because APC+/+ and APC+/- phenotypes are visually identical. If many cancer alterations accumulate during niche progression (Figure 5), one prediction and requirement is that many alterations found in cancers will sometimes also be found in phenotypically normal cells. The redundancy revealed by mouse models (Table 1 and most genes altered in viable mouse transgenics or knockouts (Pearson, 2002)) suggests that although tumor-prone, crypts with many of the same mutations found in tumors may remain visibly normal.

The monotonous appearance of normal crypts delegates the accumulation of alterations that confer visible growth advantages to tumors. However, the compatibility of many alterations with key roles in tumorigenesis with normal phenotypes indicates that many such alterations may invisibly hitchhike in normal crypts and only later in the right combinations become hijackers. Instead of alterations immediately hijacking normal cells, normal cells may control the early fates of these alterations. This dual hitchhiker/hijacker potential of loci with critical roles in tumorigenesis potentially allows the start of genetic progression to precede visible changes in phenotype.

Whether or not and to what extent alterations hitchhike in normal crypts is uncertain. Conceivably all but the final hit may accumulate with niche succession because correction of a single tumor defect often suppresses tumorigenicity (Baker et al., 1990; Tanaka et al., 1991; Goyette et al., 1992). Also, multiple engineered hits are required to confer a tumorigenic phenotype to normal human cells (Hahn et al., 1999). Approaches that directly attempt to isolate and detect such altered crypts are unlikely to be successful (see sidebar). The succession of niches in normal human colon inferred from a quantitative analysis of methylation patterns illustrates a type of approach needed to illuminate otherwise occult behaviors that occur before gatekeeper mutations.

Sidebar: methylation patterns as molecular clocks

Just as methylation can replace mutations as a mechanism to silence genes in cancer cells (Jones and Laird, 1999), methylation can also potentially replace mutations in molecular clocks. Molecular clocks are sequences that change through time and these alterations essentially record cell histories. Methylation can substitute for mutation as clock alterations because both are somatically inherited. Unlike the experimental manipulations needed to 'see' stem cells in model systems, human methylation tags inherently record stem cell fates in anyone. More importantly for studies of somatic cell evolution, epigenetic fidelity is likely to be less than genetic fidelity. Therefore, sufficient numbers of methylation changes may accumulate in adults to infer past cellular behavior.

Genomic stability in mitotic tissues depends on faithful replication but no process is error-free. The error rate for DNA replication is extremely low and estimated to be ~10^-9 per base per division (Loeb, 1991). Methylation changes rapidly during development (Reik et al., 2001), but few studies have measured in vivo adult methylation fidelity. Somatic epigenetic fidelity is likely less than DNA replication fidelity because methylation changes are more easily detected in vitro (Wigler et al., 1981; Shmookler Reis and Goldstein, 1982). In one study, in vitro re-methylation patterns were consistent with a site autonomous mechanism (methylation random and independent between CpG sites) after demethylation with 5-azacytidine (Pfeifer et al., 1990).

One approach to enhance estimates of small error rates is to compare genomes replicated independently over long intervals. The adult colon can be thought of as an old multi-well tissue culture plate because there are millions of crypts, and individual (~2000 cell) crypts can be isolated from fresh colons (Cheng et al., 1984; Potten and Loeffler, 1990; Potten et al., 1992; Booth and Potten, 2000). Around birth each crypt is seeded with essentially identical cells and then 'cultured' for decades. Most CpG islands are unmethylated (Cross and Bird, 1995) but some CpG islands unmethylated at birth demonstrate substantial age-related methylation in normal human colons (Ahuja et al., 1998; Issa, 2000). These somatic changes can be further characterized by comparing methylation patterns between crypts from a single adult.

Bisulfite genomic sequencing (Clark et al., 1994) of cloned PCR products can identify which of a number of possible methylation patterns are present at each CpG 'locus'. For example, there are (2)⁸ or 256 different possible patterns when a sequence with eight CpG sites is examined. Although methylation status is usually summarized as proportions of methylated CpG sites, molecular clocks require that the 5' to 3' order of CpG sites be preserved. The 'sequence' rather the composition of a locus is informative. Per cent methylation is as useful to a phylogenetic analysis as is the ratio of bases (A,T,G,C) in a sequence. Methylation represented as 5' to 3' binary strings or tags (0 if unmethylated and 1 if methylated) facilitates pattern comparisons or molecular bookkeeping. For example the difference between 0000 and 1111 is the same as the difference between 0101 and 1010 because four changes are needed to convert one tag to the other.

Tags with age-related methylation are illustrated in Figure 4. Many different tag patterns are observed in different crypts from the same adult colon. Tags from the same colon may differ between crypts and within crypts, but differences between crypts are generally greater than differences within crypts (Yatabe et al., 2001). Methylation errors in these tags should not be subject to selection because their genes are not expressed in colon. Methylation differences are likely due to random errors because presumably normal and phenotypically identical cells within a single individual are examined. Random errors allow for different methylation patterns in functionally identical normal adult cells.

Methylation exhibits somatic inheritance and a wealth of information can be inferred by mining this binary data with quantitative techniques. Although these quantitative approaches are not described here, they depend on the ability to reconstruct histories recorded in sequences. However, instead of examining sequences, cellular somatic evolution is reconstructed from epigenetic changes. A quantitative analysis describes the dynamics of age-related methylation for certain CpG islands (Yatabe et al., 2001). Errors (methylation and demethylation) appear to occur independently between CpG sites with a rate estimated at ~2´10^-5 changes per CpG site per division. This error rate is over 10 000 times higher than for DNA replication. By this model methylation at each CpG site is faithfully replicated except sometimes (once every ~50 000 divisions) the methylation status is changed. One prediction of this model is that methylation at different CpG islands or even between alleles of CpG islands may differ within the same clone. As predicted, the methylation status of one allele does not correlate with another allele, or another CpG island within individual crypts (Yatabe et al., 2001). Random errors would inherently increase methylation with aging because CpG islands are initially unmethylated.

Methylation bookkeeping can account for how tag values change with time and also audit inheritance (who gave what to whom). Multiple different tags within a single crypt (Figure 4) imply multiple stem cells because most non-stem cells die within a week (Potten and Loeffler, 1990)). With such short lifetimes, new tags are seldom generated and are soon lost in non-stem cells. Therefore numbers of unique methylation tags primarily reflect stem cell numbers. The number and properties of human crypt stem cells are largely uncharacterized because they cannot be physically identified (Potten and Loeffler, 1990). Stem cell definitions differ but essentially stem cells are equivalent to common ancestors. A phylogenetic analysis of somatically inherited methylation patterns may reveal otherwise invisible stem cell behaviors.

Stem cells may be immortal or may be defined by niches (Figure 2). Immortal stem cells intrinsically reproduce themselves through asymmetrical division. In contrast, a cell within a niche is extrinsically instructed to function as a stem cell because of its specialized environment (Spradling et al., 2001). In both cases stem cell numbers are constant, but identities of niche stem cells change with time because some 'stem' cells exit the niche through symmetrical division. Unlike the tenure of immortal stem cells, ancestors of a single cell periodically and randomly dominate a niche. The variability in numbers of unique tags per crypt is more consistent with multiple stem cells in niches rather than multiple immortal stem cells (Yatabe et al., 2001). A stem cell niche allows adult crypt tags to remain similar because periodic succession 'bottlenecks' ensure stem cells are recently related, whereas tags in immortal stem cells may become more different because they were last related at birth.

The utility of methylation patterns as telemetry for tissue reconstruction is still preliminary and uncertain (Ro and Rannala, 2001). The same strategies used to reconstruct phylogenies from sequences likely apply to methylation patterns because both exhibit somatic inheritance. Based on relatively simple assumptions, measurements and quantitative analysis of a few small stretches of CpG strings are consistent with long-lived crypts maintained by stem cell niches (Yatabe et al., 2001). Many issues remain because crypt methylation patterns are not completely random. For example, not every CpG island demonstrates age-related changes (Ahuja et al., 1998; Issa, 2000) suggesting that some sites are protected (Turker and Bestor, 1997).

Why alterations are difficult to detect in normal colon

One prediction of niche progression is that alterations accumulate in normal colon. If alterations are present in normal colon, why can't such alterations be easily detected? The primary reason is that there are no visual clues to 'select' phenotypically normal crypts with alterations. For example, even if a thousand crypts contained heterozygous somatic APC alterations, the frequency of such altered crypts would be less than 0.1% because there are millions of colon crypts. Blind sampling and the relatively poor sensitivity of screening techniques would hinder the direct detection of crypts with specific alterations.

Crypts may divide by fission (Park et al., 1995) and small clusters of aberrant crypt foci are found in normal colons (Bird, 1995). Therefore, it is possible that a single crypt with alterations could expand to cover relatively large colon patches, which would facilitate detection. However, such a field would still be visually undetectable if its crypts remained morphologically normal. Methylation tags can reveal relationships between similar appearing crypts. Tags from different crypts isolated from 1.0 cm² colon patches do not appear closely related (Yatabe et al., 2001). Tags from different crypts are no more related whether isolated from the same 1.0 cm² colon patch or from another patch more than 15 cm away (unpublished data). Tags from adjacent crypts are not closely related (unpublished data). In contrast to niche stem cell turnover, it appears that crypts themselves are stable entities. Most normal colon crypts appear to be distantly related to their neighbors and may survive the lifetime of an individual. Therefore, each crypt may accumulate alterations independent of its neighbors. Detecting single crypts with an alteration among millions of other crypts lacking the alteration would require sensitivities higher than most commonly used techniques.

Acknowledgements

Supported by Grant DK61140 from the National Institute of Diabetes and Digestive and Kidney Disease and a postdoctoral fellowship from the Korea Science & Engineering Foundation.

Ahuja N, Li Q, Mohan AL, Baylin SB, Issa JP. (1998). Cancer Res., 58: 5489-5494. MEDLINE

Baker SJ, Markowitz S, Fearon ER, Willson JK, Vogelstein B. (1990). Science, 249: 912-915. MEDLINE

Bird RP. (1995). Cancer Lett., 93: 55-71. Article MEDLINE

Booth C, Potten CS. (2000). J. Clin. Invest., 105: 1493-1499. MEDLINE

Cairns J. (1975). Nature, 255: 197-200. MEDLINE

Campbell F, Williams GT, Appleton MA, Dixon MF, Harris M, Williams ED. (1996). Gut, 39: 569-573. MEDLINE

Chao L. (1990). Nature, 348: 454-455. MEDLINE

Chao L. (1997). Gene, 205: 301-308. Article MEDLINE

Charlesworth B. (1978). Proc. Natl. Acad. Sci. USA, 75: 5618-5622. MEDLINE

Cheng H, Bjerknes M, Amar J. (1984). Gastroenterology, 86: 78-85. MEDLINE

Clark SJ, Harrison J, Paul CL, Frommer M. (1994). Nucl. Acids Res., 22: 2990-2997.

Clarke AR, Cummings MC, Harrison DJ. (1995). Oncogene, 11: 1913-1920. MEDLINE

Cross SH, Bird AP. (1995). Curr. Opin. Genet Dev., 5: 309-314. MEDLINE

Donehower LA, Harvey M, Slagle BL, McArthur MJ, Montgomery JrCA, Butel JS, Bradley A. (1992). Nature, 356: 215-221. MEDLINE

Fearon ER, Vogelstein B. (1990). Cell, 61: 759-767. MEDLINE

Felsenstein J. (1974). Genetics, 78: 737-756. MEDLINE

Foulds L. (1954). Cancer Res., 14: 327-339.

Goyette MC, Cho K, Fasching CL, Levy DB, Kinzler KW, Paraskeva C, Vogelstein B, Stanbridge EJ. (1992). Mol. Cell. Biol., 12: 1387-1395. MEDLINE

Groden J, Thliveris A, Samowitz W, Carlson M, Gelbert L, Albertsen H, Joslyn G, Stevens J, Spirio L, Robertson M, Sargeant L, Krapcho K, Wolff E, Burt R, Hughes JP, Warrington J, McPherson J, Wasmuth J, Le Paslier D, Abderrahim H, Cohen D, Leppert M, White R. (1991). Cell, 66: 589-600. MEDLINE

Hahn WC, Counter CM, Lundberg AS, Beijersbergen RL, Brooks MW, Weinberg RA. (1999). Nature, 400: 464-468. Article MEDLINE

Hanahan D, Weinberg RA. (2000). Cell, 100: 57-70. MEDLINE

Issa JP. (2000). Curr. Top. Microbiol. Immunol., 249: 101-118. MEDLINE

Jones PA, Laird PW. (1999). Nat. Genet., 21: 163-167. Article MEDLINE

Kinzler KW, Vogelstein B. (1996). Cell, 87: 159-170. MEDLINE

Loeb LA. (1991). Cancer Res., 51: 3075-3079. MEDLINE

Malkin D, Li FP, Strong LC, Fraumeni JrJF, Nelson CE, Kim DH, Kassel J, Gryka MA, Bischoff FZ, Tainsky MA, Friend SH. (1990). Science, 250: 1233-1238. MEDLINE

Muller HJ. (1964). Mutat. Res., 1: 2-9.

Nowell PC. (1976). Science, 194: 23-28. MEDLINE

Oshima M, Oshima H, Taketo MM. (1996). Dev. Biol., 179: 297-302. Article MEDLINE

Park HS, Goodlad RA, Wright NA. (1995). Am. J. Pathol., 147: 1416-1427. MEDLINE

Parsons R, Li GM, Longley M, Modrich P, Liu B, Berk T, Hamilton SR, Kinzler KW, Vogelstein B. (1995). Science, 268: 738-740. MEDLINE

Pearson H. (2002). Nature, 415: 8-9. Article MEDLINE

Pfeifer GP, Steigerwald SD, Hansen RS, Gartler SM, Riggs AD. (1990). Proc. Natl. Acad. Sci. USA, 87: 8252-8256. MEDLINE

Potten CS, Loeffler M. (1990). Development, 110: 1001-1020. MEDLINE

Potten CS, Kellett M, Roberts SA, Rew DA, Wilson GD. (1992). Gut, 33: 71-78. MEDLINE

Reik W, Dean W, Walter J. (2001). Science, 293: 1089-1093. MEDLINE

Ro S, Rannala B. (2001). Proc. Natl. Acad. Sci. USA, 98: 10519-10521. MEDLINE

Shmookler Reis RJ, Goldstein S. (1982). Proc. Natl. Acad. Sci. USA, 79: 3949-3953. MEDLINE

Shoemaker AR, Haigis KM, Baker SM, Dudley S, Liskay RM, Dove WF. (2000). Oncogene, 19: 2774-2779. MEDLINE

Smith JM, Nee S. (1990). Nature, 348: 391-392.

Spradling A, Drummond-Barbosa D, Kai T. (2001). Nature, 414: 98-104. Article MEDLINE

Tanaka K, Oshimura M, Kikuchi R, Seki M, Hayashi T, Miyaki M. (1991). Nature, 349: 340-342. MEDLINE

Tomlinson IP, Novelli MR, Bodmer WF. (1996). Proc. Natl. Acad. Sci. USA, 93: 14800-14803. Article MEDLINE

Turker MS, Bestor TH. (1997). Mutat. Res., 386: 119-130. MEDLINE

Watt FM, Hogan BL. (2000). Science, 287: 1427-1430. Article MEDLINE

Wigler M, Levy D, Perucho M. (1981). Cell, 24: 33-40. MEDLINE

Williams ED, Lowes AP, Williams D, Williams GT. (1992). Am. J. Pathol., 141: 773-776. MEDLINE

Yatabe Y, Tavare S, Shibata D. (2001). Proc. Natl. Acad. Sci. USA, 98: 10839-10844. MEDLINE

Figure 1 A colon crypt. There are millions of crypts per human colon and each contains about 2000 cells. The crypt is maintained by small numbers of stem cells (black circles) located at or near the crypt base. Progeny from the stem cells differentiate and migrate towards the lumen of the colon. The differentiated cells die at the lumen. Epithelial cell turnover is rapid such that most cells other than stem cells are replaced within a week

Figure 2 Differences between immortal stem cells and a stem cell niche. Immortal stem cells always replace themselves with asymmetrical divisions and therefore their lineages never become extinct. In contrast, in a niche, stem cells numbers are constant but both asymmetrical and symmetrical divisions are possible. With symmetrical division, sometimes one stem cell lineage becomes extinct because both daughters leave the niche. To maintain a constant niche stem cell number, this extinction is balanced by stem cell expansion in which both daughters remain as stem cells. This random stem cell loss with replacement will eventually lead to the extinction of all lineages except one, or clonal succession

Figure 3 Niches revealed by visible crypt conversion. Murine crypts can be made heterogeneous for a histologically detectable marker (black or stripped stem cells) by mutagenesis. Heterogeneous appearing crypts subsequently disappear over several months with crypts becoming homogeneous for one or the other marker. This loss of heterozygous appearing crypts is evidence for the stem cell succession of niches. Although serial observations and mutagenesis of humans is impractical, the same process can be inferred by examining crypt methylation patterns

Figure 4 Examples of 24 methylation tags sampled from two different crypts from the same 87 year-old colon. Methylated sites are represented by '1' and unmethylated sites by '0'. Tags represent sequences of individual cloned PCR products derived from bisulfite treated crypt DNA. Each tag represents a single cell because the locus (BGN) is on the X-chromosome and the colon is male. Tags are different between crypts and similar within crypts. A 'consensus' tag is 10100111 for crypt 1 and 00001000 for crypt 2. There is intra-crypt diversity with nine different unique tags in crypt 1 and three different unique tags in crypt 2. These tags are inter-related and ultimately trace back to a single common ancestor. Quantitative analysis of intra- and inter- crypt tags may infer this history

Figure 5 A crypt niche progression sequence. This diagram separates visual progression from a potentially invisible genetic progression that starts before tumorigenesis. Visually tumors are rare until adulthood. However, alterations may invisibly accumulate in histologically normal cells as long as they do not confer phenotypic changes. From birth, stem cell alterations (short arrows) may be passively fixed due to the natural rhythm of niche succession cycles. Most alterations will be lost, but sometimes a single crypt may sequentially collect multiple hitchhiker alterations. Rarely combinations of hitchhiker alterations collectively confer a tumor phenotype, leading to the visible start of phenotypic progression. The numbers of alterations that eventually help confer a tumorigenic phenotype but first hitchhike in normal crypts are unknown and may differ between tumors. Perhaps some alterations are common in tumors because they have greater propensities to hitchhike. Although a stem cell niche of four cells is illustrated, niches may contain hundreds of cells

Figure 6 The niche arena. Like Figure 5, a stem cell niche with four cells is illustrated. An APC alteration results in a crypt with both APC+/- and APC+/+ cells. This situation is unstable because only one lineage survives the niche. The APC alteration will either become fixed or extinct. The APC alteration may confer dominance, but drift can cause its extinction regardless of a selective advantage. Conversely, by Muller's ratchet, drift may lead to fixation even if the APC alteration has a deleterious or neutral effect. A click of Muller's ratchet is both APC alteration and its fixation. Once fixed the APC+/- crypt cannot return back to APC+/+ because restorative changes are rare

Table 1 Genotypes associated with normal crypt morphologies in mice and men

Table 2 Comparisons between an adenoma-cancer sequence and a niche progression sequence