PROTECTION BY THE CONTROL OF LINEAGE

In each region of a rapidly multiplying tissue there are a few long-lived “stem” cells, which periodically divide asymmetrically, with one of the daughter cells replacing its mother and with the other generating a clone that replaces the existing rapidly dividing population. The rate of accumulation of mutations in each tissue therefore reflects the properties of the tissue's stem cells. One obvious way to lower the rate of accumulation of errors of replication would therefore be to slow the rate of the division of stem cells. For example, consider a tissue in which the cycle time of the rapidly multiplying cells is one Nth of the animal's lifetime. If this were true for every cell in the tissue, any cell still present at the end of the animal's life would necessarily have had N divisions in its ancestry (ignoring the relatively small number of divisions initially needed to create the tissue). If, however, this population of cells is periodically completely replaced by a clone arising from an underlying slowly dividing stem cell and if that stem cell produces a replacing clone S times in the animal's life, then the greatest number of divisions that the fast-dividing cells in these clones can have is N/S, the number of divisions of the stem cells will be S, and at the end of the animal's life the maximum number of divisions in any cell's ancestry will be (S + N/S) (note that S and N reflect the frequency of division of the two classes of cell, not the number of such cells). The minimum value of (S + N/S) is and is reached when . For example, in the small intestine of a mouse the cycle time of the 10¹⁰ fast-dividing cells is ∼1/2000th of the usual maximum life span of a mouse (3 years), and therefore N = 2000. If the cells in each region of this epithelium were being replenished by the divisions of a single stem cell, that stem cell ideally should have divided (i.e., 45 times) during a mouse's life span, or roughly once every 3–4 weeks. Instead of the 10¹⁰ cells in this old mouse's small intestine having 2000 divisions in their ancestry, none could have >90 divisions in their ancestry () and so there would be slightly less than one-twentieth as many mutant cells. And if there were a hierarchy of stem cells in a tissue, each with a slower rate of division than its stem cell daughters, the maximum number of divisions could be further reduced.

Some systems of proliferating cells are known to be periodically replenished from separate groups of cells that divide rarely if at all. For example, human hair follicles are totally replaced roughly every 3 years by a clone arising from a cell exported from a stationary population of cells in what is called the “bulge” region of the hair follicle (Cotsarelis et al. 1990). Similarly, the apical stem cells in the growing shoots of many plants are probably periodically replenished by a cell coming from the nearby “quiescent center” (Barlow 1976). But, in the best-studied epithelia, such as mouse epidermis and gut, the rate of division of the stem cells at the base of the epithelium is about one-half the rate of the fast-growing cells and there is no evidence for the existence of rarely dividing stem cells, let alone precisely ordered hierarchies of stem cells. Furthermore, it seems unlikely that precise hierarchies exist, because mutant frequencies have been observed to rise roughly linearly with age both in human lymphocytes and in mouse small intestine (Winton et al. 1988; Finette et al. 1994). Contrary to what one might have expected, therefore, the control of mutation rates in epithelia may not depend on a drastic reduction in the length of lineages achieved by a great reduction in the rate of division of stem cells. Yet the accumulation of mutations in epithelia is somehow slowed because most cancers arise only in old age, and this protection must be occurring in stem cells because they are the only epithelial cells that survive from one year to the next.

PROTECTION BY THE SEGREGATION OF MUTATIONS

Changes in DNA sequence arise from errors during replication or repair. But a change in sequence is not fixed irreversibly until it is present in both strands—i.e., has been copied at the next round of replication. It seemed possible, therefore, that stem cells might avoid accumulating mutations if, at mitosis, they always kept, for each chromosome, the chromatid with the older template strand (Cairns 1975). Bacteria were known to keep together the template strands of separate replicons (Cuzin and Jacob 1965) and this form of control has been observed in mouse embryonic cells in culture (Lark et al. 1966) and in the growing root tips of plants (Lark 1967) and perhaps in mouse tongue (Potten et al. 1978). Recently, studies of stem cells in the mouse small intestine (Potten et al. 2002), breast (Smith 2005), brain (Karpowicz et al. 2005), and muscle (Shinin et al. 2006) have shown that these stem cells do indeed keep the same parental DNA strands through successive divisions and, in one case, that their non-stem-cell daughters do not (Shinin et al. 2006). This arrangement ensures that any errors arising in stem cells during gene duplication avoid being permanently fixed because they are passed on, at the asymmetric division of each stem cell, to the differentiating daughter cell and will therefore soon be discarded.

The interactions of gene products underlying these properties of stem cells are not understood, but it seems likely that they involve the action of P53. Thus, when p53 in non-stem cells grown in vitro is continuously overexpressed, the cells behave like stem cells in that they keep their old template strands together (Merok et al. 2002). Furthermore, under these conditions, one of each pair of daughter cells is unable to divide again, suggesting that there is a strand-dependent P53-dependent checkpoint that forces one of the daughters to behave like a stem cell.

The preservation of “immortal strands” entails certain restraints on the types of DNA repair available to stem cells. Recombinational repair, for example, involves the interchange of new and old strands, and so if stem cells are to keep the same template strands through many divisions, they must avoid sister-chromatid exchange. It was therefore not surprising to learn that embryonic stem cells are deficient in certain kinds of repair (Roth and Samson 2002) and apparently cannot carry out mitotic recombination (Cervantes et al. 2002), which may be why they are extremely sensitive to certain kinds of DNA damage (Potten 1977).

The only DNA lesions (potential mutations) that persist in epithelia (i.e., accumulate over a lifetime) may be those that are fixed in a cell that later is promoted into being a stem cell. When a stem cell is killed (e.g., by radiation), it is replaced by the dedifferentiation of a differentiating daughter cell (Hendry et al. 1992). This may be why the frequency of mutation in epithelial stem cells during continuous mutagenesis increases as the square of time (Shaver-Walker et al. 1995); for, to put a mutation in a stem cell, first the existing stem cell must be killed and second its replacement must have become mutant in both strands (Cairns 2002).

What we see here, therefore, is a system that slows the accumulation of mutations. When stem cells are not being killed, they multiply slowly and may therefore have a long period in which to check their genome for errors, and whenever they are killed, they are replaced by transiently repair-proficient daughter cells. Another factor that saves epithelia from the selection of fitter mutants is their stem cells' dependence on signals coming from neighboring cells in the stem cell niche (Watt and Hogan 2000). For example, in human skin, mutant stem cells can produce aggressive clones of descendants that out-compete the daughters of neighboring nonmutant stem cells, but these clones usually cannot take over neighboring niches if there are stem cells in those niches (Zhang et al. 2001); similarly, mouse blood contains circulating stem cells that can repopulate marrow but only if the existing marrow stem cells have been killed (Micklem et al. 1975). Indeed, in the absence of continual cell damage, it seems that such expanded mutant clones tend to do less well than their normal neighbors. For example, skin tumors produced in rabbit ears by coal tar usually disappear when the skin is not being continually treated but duly reappear at the same place (presumably from the same mutant stem cells) when treatment is resumed (Mackenzie and Rous 1941). Those classical experiments in carcinogenesis showed how the natural selection of fitter variants is inhibited in the rapidly multiplying tissues of long-lived animals. There is, however, clear epidemiological and experimental evidence for yet another final, rate-limiting barrier to the development of many human cancers that apparently does not require either cell death or mutagenesis (Peto 2001).

THE FINAL, POSSIBLY EPIGENETIC, STEP IN CARCINOGENESIS

Once again we start with what is known about the timing of events in the creation of the commonest human cancers. The annual death rates from breast cancer and cervical cancer rise steeply up to middle age but remain almost constant for the next 10–20 years. As pointed out many years ago (Prehn 1977), it is as if creation of breast and cervical cancers (and certain experimental cancers) requires a final step that is of a different kind from the preceding steps. Prehn (1997) has suggested that the critical event is when a stem cell has to undergo symmetrical rather than its normally asymmetric mitosis. A slightly different interpretation now seems more likely. Several studies have shown that in ex-smokers the annual death rate from lung cancer remains for the next 20 years close to the level it had reached when the smoker quit smoking. This suggests that the last step is a rare event affecting existing mutant cells that, astonishingly, is not accelerated by smoking.

The main cause of lung cancer in smokers is presumably the continuing cell death and mutagenesis that occurs in lung epithelium exposed to the toxins and mutagens in burning tobacco. The main cause of breast cancer is less obvious; mammary epithelium is known to undergo periodic expansion and contraction during the period of accumulating risk (Masters et al. 1977), and it therefore seemed possible that the rising risk of both breast and cervical cancer in young women could be due to the accumulation of mutations associated with the repeated generation and loss of stem cells (Cairns 1975), helped in the case of cervical cancer by HPV infection, which inactivates the p53- and Rb1-dependent mechanism for recognizing DNA damage. However, the example of lung cancer in ex-smokers adds a totally unexpected element.

At any moment, the frequency of any class of cancer presumably reflects the number of cells that have undergone all but the final step in carcinogenesis, multiplied by the rate of whatever happens to be the final step. If this final step required DNA damage, the death rate from lung cancer would quickly drop when the mutagenic stimulus (smoking) was removed. In fact, as already mentioned, at that point the death rate shows no decline and for the next 10–20 years actually stays roughly at the level it had reached in the year before the smoker stopped smoking (Halpern et al. 1993). This suggests that the final step, which starts the growth of three common human cancers (cancer of the lung, breast, and cervix), is an event of a different kind from the steps that lead to the accumulation of mutant cells because, unlike the earlier steps, it does not have to be stimulated by the toxicity and mutagenicity of tobacco or by the ebb and flow of mammary and cervical epithelium, but occurs at a constant rate. It must, however, be a rare event because the incidence of lung cancer in ex-smokers, even in those who smoked for >40 years, is <1% a year.

The final step in carcinogenesis initiates the growth of a presumably fully mutant stem cell, creating a perpetually expanding cancerous clone in which more mutations will inevitably arise. We have to ask what kind of rare event can alter the fate of stem cells and of their immediate descendants. Normally, each time a stem cell divides, its two daughters have different fates: one has to replace its parent and stay within the confines of the stem cell niche, while the other is programmed to differentiate and multiply and eventually all its descendants die and are discarded. The final step could be the rare failure of this dichotomy, when a fully mutant cancerous stem cell produces a daughter cell that is freed from some nongenetic imperative to differentiate and die (an event analogous to Prehn's “symmetrical division”). The most important feature and perhaps the evolutionary origin of the preservation of immortal strands may have been that it provided the basis for a strand-based epigenetic mechanism that distinguished daughter cells from their mothers and, later in evolution, could not only determine the phenotypes of the cells in the different tissues of multicellular organisms but also distinguish the stem cells in those tissues from their differentiating daughters.