Telomere length and proliferation potential of hematopoietic stem cells

The blood-forming or hematopoietic system is and has been extensively studied for a variety of reasons. It represents the prototype of a self-renewing biological system in that large numbers of blood cells have to be produced daily in order to compensate for the loss of relatively short-lived mature blood cells. The relative ease at which blood and bone marrow samples can be obtained and single cell suspensions prepared has greatly facilitated in vitro and in vivo experimentation with hematopoietic cells. Monoclonal antibodies specific for a variety of cell surface antigens expressed on hematopoietic cells have been produced and techniques to separate cells using combinations of such reagents have been developed. A large number of molecules with activity on various hematopoietic cells are now available as are various in vitro assays to measure functional properties of hematopoietic cells.

The hematopoietic system has been subdivided into a hierarchy of three distinct populations (Till et al., 1964; Metcalf, 1984). In this model the most mature cells are morphologically identifiable as belonging to a particular lineage and have very limited proliferative potential. The cells in this most mature compartment are derived from committed progenitor cells with a higher but still finite proliferative potential. Committed progenitor cells in turn are produced by a population of multipotential hematopoietic stem cells with selfrenewal potential, i.e. the capacity to give rise to more cells with indistinguishable properties and developmental potential. Self-renewal of stem cells is believed to be essential for maintenance of hematopoiesis over time.

Studies with cultured hematopoietic cells have shown that the formation of hematopoietic colonies as well as the proliferation and survival of various hematopoietic cells typically requires the presence of ‘hematopoietic growth factors’, many of which have been cloned over the last decade (Metcalf, 1989, 1993). Many investigators have interpreted these observations as evidence that the behavior of hematopoietic cells is ultimately controlled by extracellular signals. This notion, together with the possibility of purifying various hematopoietic cells with unprecedented precision has led to frantic research efforts over the last five years to achieve clinical useful manipulation of blood-cell production in vitro. Two sets of observations cast doubt about the feasibility of some of these perceived applications in the short term. First, there are a number of observations indicating that extracellular factors do not appear to control lineage commitment and self-renewal but instead seem to permit exhibition of a predetermined cellular proliferative and differentiation potential (Ogawa, 1989; Lansdorp et al., 1993; Mayani et al., 1993). These findings suggest that the decisions that ultimately determine the fate of hematopoietic stem cells are not dictated by the environment but instead are controlled by currently ill-defined, intrinsic genetic mechanisms with a developmental component. A second hurdle to meaningful in vitro stem cell ‘expansion’ is the observation that hematopoietic cells, including purified hematopoietic stem cells, appear to loose telomeric DNA with each cell division and with age (Vaziri et al., 1994). This Commentary is focused on the role of telomeres and telomerase in normal and abnormal hematopoietic cells. As telomerase activity has so far not been found in any somatic tissues (Counter et al., 1994), the findings in the hematopoietic system appear to be applicable to other self-renewing somatic tissues as well. More extensive and general reviews on telomeres and telomerase have been published elsewhere (Blackburn, 1991, 1992; Harley, 1991).

Without extremely reliable mechanisms to duplicate and segregate complete copies of the genome into daughter cells, life in any form could not exist. It seems certain that given this importance, many factors and pathways involved in securing the fidelity of gene duplication and chromosome segregation remain to be uncovered. Recently, the role of telomeres in some of these processes has received increased attention. Telomeres are the physical ends of eukaryotic chromosomes and contain both DNA and protein (Blackburn, 1991). Various functional properties of telomeres were recently reviewed (Blackburn, 1994). The DNA component of telomeres consist of (TTAGGG)_n in all vertebrates including humans. Telomeres distinguish intact from broken chromosomes (Sandell and Zakian, 1993). This distinction is important because the ends of broken chromosomes (but not intact telomeres) indirectly induce cell cycle arrest and chromosome repair before cells can proceed to cell division(s). Telomeres furthermore appear to stabilize chromosome ends, in that telomeres (but not broken chromosomes) are protected from nuclear degradation and end-to-end fusion. Telomere proteins are likely to be involved in the positioning of telomeres and chromosomes within the nucleus and are important for the spatial structure of telomeres and chromosome segregation (Giraldo and Rhodes, 1994; Chikashige et al., 1994). Finally, telomeres appear to play an important role in the replication of the very end of chromosomal DNA.

Two characteristics of the DNA polymerases involved in duplication of DNA prior to cell division are: (1) the requirement of label RNA primers to initiate DNA synthesis; and (2) synthesis of new DNA in the 5′r3′ direction only. That these characteristics pose a problem for the complete replication of the 3′ end of linear chromosomes was realized by Olovnikov (1971, 1973) and Watson (1972), who described it as the ‘end replication problem’ (Fig. 1). The effect of the end replication problem on DNA duplication at chromosome ends is illustrated in Fig. 2. From this figure, it can be understood that the 3′ ends of all linear chromosomes are expected to shorten progressively with each round of DNA duplication (Levy et al., 1992). It now appears that such chromosome shortening can indeed be observed in all human somatic cells studied to date, including primitive hematopoietic cells from adult bone marrow (Vaziri et al., 1994).

A number of studies have documented that the length of telomeric (TTAGGG)_n repeats in human cells decreases with in vitro and in vivo cell divisions (Harley et al., 1990; Hastie et al., 1990; Lindsey et al., 1991; Counter et al., 1992; Allsopp et al., 1992; Vaziri et al., 1993). In these studies, telomeric length showed considerable variation between individuals and between different tissues. Studies with cultured fibroblasts have indicated that telomere length is a better predictor of replicative capacity than the actual age of the fibroblast donor (Allsopp et al., 1992). On the basis of these studies, it appears that loss of telomeric DNA (and resulting cell cycle exit signals) may adequately explain the observation by Hayflick and Moorhead that normal human fibroblast become senescent after a fixed number of doublings in vitro (Harley, 1991). This possibility has been expanded by Harley into the ‘telomere hypothesis of cellular aging’ (Harley, 1991; Harley et al., 1992). Several recent observations are in agreement with predictions of this theory and it appears that the possible implications of telomere biology could be wide-spread indeed. Not only could measurements of telomere length be used to estimate the proliferative potential of cells or, by repeating measurements over time, the in vivo turn-over rate of cells and tissues, but such measurements should also be valuable in studies of aging and associated disorders. More information on the role of telomeres in cellular aging may furthermore lead to meaningful predictions of organ failure with implications in diverse areas such as gerontology and preventive medicine.

A major problem in the exploration in some of these possibilities is that telomere length measurements currently require DNA from at least 10⁵ cells and that such assays are time consuming. Typically, DNA (extracted from 1×10⁶ to 2×10⁶ cells) is first digested with restriction enzymes that cleave internal but not telomeric sequences. The resulting fragments are separated by gel electrophoresis, immobilized and hybridized with ³²P-(CCCTAA)₃. Specifically bound telomere probe results in a smear on autoradiographs that ranges between 10 and 15 kb for ‘young’ and 5 and 10 kb for ‘old’ cells. Results of these experiments are typically expressed as the mean telomere restriction fragment (TRF) size (in kilobases) after quantitation using densitometry. The actual size of telomeric repeats is shorter because the restriction enzymes used cut 2-5 kilobases upstream from the start of the telomeric repeats themselves (Harley, 1991). The presumed heterogeneity in TRF length between cells within a tissue and between individual chromosomes make alternative measurements of telomere length highly desirable. Ideally, such measurements should be done on individual cells. Alternatives that are being explored are in situ hybridization and immunological methods for quantitation of proteins that bind specifically to telomeric repeats (Palladino et al., 1993). Potential problems with both approaches may be the presence of sequences with homology to telomeric repeat sequences at subtelomeric or intra-chromosomal sites. The presence of the latter is particularly a problem in inbred mice, and has prevented the use of this animal model in studies of telomeres and cellular aging (de Lange, 1994).

Most, if not all, somatic human cells lose 50-100 bp of telomeric DNA with each successive round of cell division (Harley et al., 1990; Allsopp et al., 1992; Vaziri et al., 1993, 1994). Because such cells ‘start’ their proliferative life at an unidentified point in fetal life with 5-10 kb of (T₂AG₃)_n repeats, the replicative capacity of early fetal cells is expected to be of the order of 50-200 doublings. Even 50 doublings represents a tremendous proliferative potential, which can, in theory, yield up to 10¹⁵ cells or approximately 1000 kg of cells. Differentiation and cell death will almost certainly limit the actual size of clones to a fraction of this estimate. The actual proliferative potential of cells may furthermore be less, as the exact length of telomeric repeats required for full telomere function is not known. In addition, the length of telomeric DNA repeat sequences will vary to some extent between individual chromosomes (Moyzis et al., 1988), presumably limiting the proliferative potential of cells to the chromosome with the shortest length of such repeats.

As discussed above, the replicative potential of somatic cells is expected to decrease as a function of their proliferative history. How does critical shortening of telomeres signal cell cycle exit and cellular senescence? Some answers to this important question have emerged from elegant studies in yeast (Sandell and Zakian, 1993). It was found that inducible elimination of a telomere on a single chromosome resulted in arrest of the cell cycle at the G₂ checkpoint. Interestingly, many of the yeast cells ultimately recovered from the arrest without repairing the damaged chromosome. Such damaged chromosomes were much more likely to be lost upon division than chromosomes with intact telomeres (Sandell and Zakian, 1993). From this study, it appears that loss of telomeric DNA induces the same cellular machinery (cell cycle arrest, DNA repair) as broken chromosomes and that intact telomeres have important functions, by allowing cell cycle progression and avoiding chromosome loss.

If somatic cells are indeed limited by the length of their telomeric DNA to a finite replicative capacity, this could be an important natural barrier to prevent unlimited proliferation by malignant clones. It is currently not known whether this mechanism indeed limits the growth of certain tumors. Furthermore, given a proliferative potential of 50-100 doublings, it can easily be envisioned how some tumors could kill their hosts without exhausting available telomere DNA at the time of the initial malignant transformation. Nevertheless, it now appears that malignant human tumors may use the same mechanism by which cells of the germ line presumably remain immortal: expression of the enzyme telomerase (Counter et al., 1992, 1994; de Lange, 1994). Telomerase synthesizes telomeric repeats using an RNA template complementary to the G-rich repeats on the 3′ telomeric chromosome end (Greider and Blackburn, 1985; Blackburn, 1992). The human gene(s) encoding for this remarkable RNA-dependent DNA polymerase (a reverse transcriptase) have not yet been cloned. However, biochemical assays, using substrates of artificial telomeric repeats, have provided clear evidence for the existence of telomerase activity (Counter et al., 1992, 1994). Interestingly, this enzyme activity has been found so far only in immortal human cell lines (Counter et al., 1992) and certain human tumors (Counter et al., 1994) but not in primary somatic cells from normal individuals (Counter et al., 1992, 1994). These observations suggest that telomerase could be a tumorspecific enzyme and that non-toxic inhibitors of this enzyme may be useful chemotherapeutic agents in the management of patients with certain malignant tumors (Counter et al., 1994).

When DNAs from fetal liver cells, umbilical cord blood cells and adult bone marrow cells were compared, a striking and highly significant (P≤0.0001) difference in the mean length of terminal restriction fragments (containing the (T₂AG₃)_n telomeric repeats) between the fetal/neonatal and adult tissues was observed (Fig. 3). The observed loss of telomeric DNA in these hematopoietic tissues appears not to be restricted to more mature cells as purified CD34⁺CD38⁻ primitive progenitor cells from adult bone marrow were also found to have shorter telomeres than fetal liver cells (Vaziri et al., 1994). In these studies, there also appeared to be an age-related loss of telomeric DNA. The measured rate of telomere loss in lymphocytes was calculated to reflect 0.4 stem cell doubling/year (Vaziri et al., 1993), whereas the corresponding value was 0.25 stem cell doubling/year for (limited) adult bone marrow data (Vaziri et al., 1994). Considerable variation in the mean telomere length between individuals will require extension of these studies to a larger number of samples in order to obtain more accurate estimates of in vivo cell turn-over. Interestingly, leukemic blast cells from a pediatric patient were found to have significantly shorter telomeres than normal cells from the same patient (Adamson et al., 1992). These findings are in support of proliferation-dependent loss of telomeric DNA in cells of hematopoietic origin. The most straightforward explanation for these combined observations is that hematopoietic stem cells do not express telomerase, have a very low turn-over rate in adults and decrease their proliferative potential with each division and, as a consequence, with age. However, the possibility that minor populations of CD34⁺CD38⁻ stem cells either stop dividing at a fetal cell stage or, alternatively, express telomerase activity and by either mechanism maintain ‘fetal length’ telomeres throughout life has not been formally excluded. Techniques that could be used to measure telomere length and telomerase activity in individual cells would be extremely useful in the further study of these possibilities.

Several observations with murine as well as human cells are in agreement with there being a finite number of divisions of early hematopoietic cells and some investigators have speculated about the nature of such restrictions. In discussing differences between embryonic and adult stem cells, Metcalf and Moore (1971) calculated that embryonic (murine) stem cells can undergo 20-80 more doublings than equivalent cells from adult marrow (Metcalf and Moore, 1971). As one of the possible explanations for the calculated differences, these authors proposed that stem cells may be programmed to undergo a fixed number of divisions. Reincke et al. (1982) observed limitations in the proliferative capacity of normal and irradiated stem cells in long-term bone marrow cultures. The invariable occurrence of cellular senescence after a fixed number of doublings was interpreted as indicating a general biological limit to the division capacity of cells. Other reports documenting that embryonic and fetal hematopoietic cells have a higher in vivo (Rosendaal et al., 1979; Albright and Makinodan, 1976) and in vitro (Nakahata and Ogawa, 1982; Hows et al., 1992; Lansdorp et al., 1993) proliferative potential than adult hematopoietic cells may also be in support of there being absolute limits in replicative capacity. Similarly, numerous observations indicating limitations in the proliferative potential of stem cells using serial transplantation experiments (Metcalf and Moore, 1971; Moore, 1992) could be interpreted as providing support for possible telomere-related restrictions in the proliferative potential of hematopoietic cells. A detailed review of these studies is thought to be outside the scope of this Commentary as the fate of primitive hematopoietic cells in vitro and in vivo is undoubtedly controlled by a variety of intrinsic and extrinsic factors, each of which may have contributed in various degrees to these observations.

If all hematopoietic stem cells show age- and proliferationdependent shortening of telomeric DNA, several notions about the ‘self-renewal’ of stem cells and their use in transplantation and gene therapy probably need to be revised. Culture of purified candidate stem cells with the goal of expanding their number would have to take ultimate telomere-related restrictions in their replicative potential into account. From this perspective, the use of fetal liver (Lansdorp et al., 1993) and cord blood cells (Broxmeyer et al., 1992) appears to have significant advantages over cells derived from adult peripheral blood or bone marrow. The ±4 kb of extra telomeric repeats that cord blood and fetal liver cells have in comparison to cells from adult bone marrow (Fig. 3) represent an estimated extra proliferation potential of 20-40 cell doublings (assuming loss of 100-200 bp of telomeric DNA per cell division). Even a minimal estimated difference of 20 doublings could produce, in theory, a 10⁶-fold difference in cell numbers generated from cord blood versus adult progenitor cells. Clearly, this number compares favourable with the 7- to 30-fold fewer CD34⁺ progenitor cells that can be directly harvested from cord blood compared to adult bone marrow (Broxmeyer et al., 1992). In order to take advantage of this theoretical superiority of cord blood cells, it will be important to understand how their extensive proliferation potential can be selectively exploited in order to increase the number of transplantable stem cells. At this point, it seems unlikely that growth factors alone will be sufficient to achieve such net expansion of cord blood stem cells. For this purpose, studies towards determining the genetic mechanisms that ultimately determine the fate of very primitive hematopoietic cells may be more rewarding. The selective expression of certain transcription factors in the most primitive normal hematopoietic cells such as homeobox genes A2, B3 and B4 (Sauvageau et al., 1994) may provide a starting point for such studies. If self-renewal decisions of stem cells could be manipulated in a meaningful way, such measures would most likely need to be combined with techniques that allow induced extension or maintenance of telomere length in order to be clinically useful with cells from adult tissues. Clearly such sophisticated cellular engineering techniques require a few breakthroughs and will not be available in the immediate future.

In view of the likelihood that telomere length ultimately determines the replicative capacity of hematopoietic cells, the word proliferation potential should probably be used with caution. It now seems reasonable to distinguish cell biological or functional from telomeric or genetic restrictions in the number of divisions a cell can potentially undergo. Thus, hematopoietic cells may differ in replicative capacity as determined by the length of their telomeres and yet have similar functional proliferative potentials.

It is possible that the functional proliferative potential of primitive hematopoietic cells is somehow linked to the length of telomeres in such cells. This notion is certainly in agreement with the higher proliferative capacity of fetal and cord blood cells as compared to cells with a similar phenotype purified from adult bone marrow (Lansdorp et al., 1993). However, a more likely explanation for this correlation is that the probability of self-renewal (reflected in the replating potential of cells) changes with different stages of development, similar perhaps to the well-known developmental control of hemoglobin gene expression. Without conditions that allow full display of actual genetic replicative capacity, correlation between telomere length and functional proliferative potential are likely to be coincidental. It could be argued that all correlations between telomere length and replicative history/proliferative potential of cells that have been observed to date are just that: correlations. Although numerous observations discussed in this Commentary are in agreement with the functional involvement of telomere length in determining the number of divisions a cell can undergo, direct proof of this linkage in normal cells has not yet been provided. For this reason, expression of endogenous or exogenous functional telomerase activity in normal somatic cells and demonstration that this results in extension of their proliferative potential are eagerly awaited.

This work was supported by NIH grant AI29524, and by grants from the Medical Research Council and the National Cancer Institute of Canada. Dr H. Mayani and C. Harley are thanked for comments on the manuscript, which was typed by Colleen MacKinnon.