Human marrow stromal cells (MSCs) can be isolated from bone marrow and differentiate into multiple tissues in vitro and in vivo. These properties make them promising tools in cell and gene therapy. The lack of a specific MSC marker and the low frequency of MSCs in bone marrow necessitate their isolation by in vitro expansion prior to clinical use. This may severely reduce MSC proliferative capacity to the point that the residual proliferative potential is insufficient to maintain long-term tissue regeneration upon reinfusion. In this study we determined the effect of in vitro expansion on the replicative capacity of MSCs by correlating their rate of telomere loss during in vitro expansion with their behavior in vivo. We report that even protocols that involve minimal expansion induce a rapid aging of MSCs, with losses equivalent to about half their total replicative lifespan.
Marrow stromal cells (MSCs), or mesenchymal stem cells, are a promising therapeutic tool. They can be readily isolated by plastic adherence, differentiated into tissues such as bone and cartilage, and genetically modified by viral vectors [1–4]. Clinically, MSCs may be used to enhance hematopoietic stem cell (HSC) engraftment post-transplantation, to correct inherited disorders of bone and cartilage, or as vehicles for gene therapy such as in osteogenesis imperfecta .
The use of MSCs to reconstitute bone marrow stroma or to sustain healthy osteogenesis relies on the long-lasting engraftment of MSCs with a residual replicative capacity to produce differentiated progeny, substitute damaged tissue, and sustain tissue turnover for life. To date, reinfusion of MSCs has resulted in poor engraftment and limited cell survival [6–8]. Previous studies have shown that human MSCs cultured in vitro display a tendency to lose their proliferative potential, homing capability, and in vivo bone-forming efficiency over time [9–15]. Whether even a limited expansion in vitro is sufficient to age the cells to a point where their successful clinical use is compromised has not yet been addressed.
Telomeres are specialized structures present at the ends of eukaryotic chromosomes. They have been associated with the molecular machinery critical for cell replicative lifespan, and their shortening is known to play an important part in the cell molecular aging process [16, 17]. In human cells, such as fibroblasts, telomeres shorten progressively during successive cell divisions, down to a threshold length at which they undergo replicative senescence [18, 19]. Consequently, mean telomere restriction fragment (mTRF) lengths can be used to extrapolate the replicative history and remaining replicative capacity of a cell population.
In the study reported here, we investigated the mTRFs of human MSCs during in vitro expansion and correlated this with mTRF changes with donor age. Our data show that, with present protocols, in vitro culture for just 7–10 population doublings (PDs) gives a reduction in mTRF length that, depending on the age of the MSC donor, is equivalent to the loss of more than half their lifespan by the time they are reinfused.
Materials and Methods
MSCs were derived from bone marrow aspirates from healthy donors obtained after informed consent (in accordance with the ethical standards of the local ethical committee and with the Helsinki Declarations of 1995). MSCs were isolated and cultured as previously described . Briefly, mononuclear cells (MNCs) were seeded at a concentration of 2 × 106 MNC/cm2 in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen, Paisley, U.K., http://www.invitrogen.com) with 10% fetal calf serum (FCS) (StemCell Technologies, London, http://www.stemcell.com). When cultures reached confluence, the MSCs were detached using 0.05% trypsin 5-mM EDTA (Invitrogen), counted using a hemocytometer; then one third of the cells were replated, and the remainder were used for mTRF analysis and differentiation assays. Analysis of karyotype was carried out as previously described .
The number of progenitor MSCs was determined by the colony-forming-unit fibroblast (CFU-F) assay, as previously described . The number of CFU-Fs at the start of the culture was used to determine the number of PDs that cells have undergone to reach primary confluence. Thereafter, the number of PDs was calculated by dividing the log N by log 2, in which N equals the total number of cells divided by the total number of cells in the previous passage.
For differentiation assays, cells were plated at 6 × 102/cm2 in MSC medium with either osteogenic supplements or adipogenic supplements, as previously described [22, 23]. The mineralized matrix was stained by the Von Kossa technique  and alkaline phosphatase expression by the alkaline phosphatase leukocyte cytochemical staining kit (Sigma Chemical Corp., Poole, Dorset, U.K., http://www.sigmaaldrich.com). Adipocytes were detected using the oil red O stain .
The mTRF length was determined as previously described . Briefly, cells were lysed in buffer containing 1% SDS, 10 mM tris-HCl, 100 mM NaCl, 1 mM EDTA, and 200 mg/ml proteinase K (Roche Diagnostics Ltd., Sussex, U.K., http://www.roche-diagnostics.com) at 37°C overnight. Genomic DNA was then extracted and digested with HinfI and RsaI (Roche Diagnostics). The determination of mTRF length was carried out by in-gel hybridization with a phosphorus-32-labeled 5′-(CCCTAA)3 telomeric probe and using ImageQuant software (Amersham, Buckinghamshire, U.K., http://www.amersham.co.uk). To limit variability, each DNA sample was run on two separate gels, and the average of the mTRF length of the two measurements was used. To determine variability inherent to the technique, the same DNA sample obtained from Jurkat or K562 cell lines was run in all lanes across two separate gels and was determined to be 500 bp.
All values are expressed as means ± standard error of the mean. Differences between groups were tested using Student's t-test. P values < .05 were considered significant. To determine mTRF length at 16 PDs, the best-fit linear regression equation was used.
In Vitro Expansion of MSCs
MSCs were derived from 10 donors of pediatric age (hMSC0–18) and five donors aged 59–75 years (hMSC59–75). After primary passage, fluorescence-activated cell sorter (FACS) analysis showed no expression of either CD34 or CD45 antigens in all MSC cultures, and all stained positive with SH2 antibody . MSC frequency in bone marrow samples, as determined by the CFU-F assay, showed a significant decrease in hMSC59–75 compared with hMSC0–18 (3.2 ± 1.7 versus 29.0 ± 4.7 per 106 mononuclear cells, p < .001).
The in vitro growth kinetics of MSCs were investigated from the primary passage until cells in culture ceased to replicate for at least 3 weeks. At this point, the cultures were considered senescent and were terminated. Most of hMSC0–18 had a faster expansion rate than hMSC59–75 (Table 1; Fig. 1A, 1B) while exhibiting the well-documented spindle-shape cell morphology (Fig. 2A) and low levels of alkaline phosphatase expression (Fig. 2B). After primary confluence, their rate of growth progressively declined. At growth arrest, hMSC0–18 lines had undergone on average 30.6 ± 2.2 PDs after 197.4 ± 25.3 days of culture.
Table Table 1.. Growth properties of hMSC0–18, hMSC0–18E, and hMSC59–75
Abbreviations: CFU-F, colony-forming unit–fibroblasts; hMSC, human marrow stromal cell; MNC, mononuclear cell; PD, population doubling.
hMSC0–18E (n= 2)
Donor age (range)
41 and 46 months
3.2 ± 1.7
29.0 ± 4.7
11.1, no data
MNC recovery following density gradient separation × 106
33.6 ± 18.0
107.3 ± 31.5
No. of days to reach primary confluence19.4 ± 4.4
10.4 ± 1.0
No. of PDs to reach primary confluence12.0 ± 1.1
12.8 ± 0.4
No. of cells at primary confluence × 104 cells/cm2
1.7 ± 0.9
6.2 ± 0.1
No. of days for 1 PD until primary confluence
1.7 ± 0.4
0.9 ± 0.1
No. of days for 1 PD between primary and secondary confluence
6.8 ± 1.7
3.4 ± 0.8
No. of CFU-F per 106 at 20–22 PDs
4.6 ± 2.7 (n = 3)
70.7 ± 41.6 (n = 3)
No. of days to growth arrest
67.7 ± 14.7
174 ± 25.2
No. of PDs to growth arrest
16.5 ± 1.0
28 ± 1.7
No. of passages to growth arrest
2.4 ± 0.5
10.3 ± 1.0
Interestingly, two of the hMSC0–18 cultures (hMSC0–18E) proliferated for over 40 PDs (Fig. 1A). They followed a similar pattern of growth kinetics to hMSC0–18 in the first 25 PDs. Thereafter, the rate of growth was maintained at an average steady rate of 1 PD every 14.8 ± 0.6 days for over 40 PDs, with no growth arrest observed. They maintained the spindle-shape morphology even at later time points (Fig. 2D) and exhibited low levels of alkaline phosphatase expression (Fig. 2E). This was in contrast to all other hMSC0–18 lines where, even at earlier PDs, a progressive change in cell morphology was observed, with cells becoming wider and flatter and showing increased alkaline phosphatase expression and Von Kossa–stained mineralized deposits (Fig. 2G). hMSC0–18E retained good osteogenic capacity (Fig. 2F) but reduced adipogenic capacity (data not shown).
hMSC59–75 showed severely reduced proliferative capacity (Table 1; Fig. 1B) with a slower growth rate than that in hMSC0–18. No cells with the spindle-shape morphology were observed; only cells exhibiting a larger and flatter morphology were present. They formed a significantly thinner monolayer at first confluence (1.7 ± 0.9 × 104 cells/cm2 versus 6.2 ± 0.1 × 104 cells/cm2, p < .001) and exhibited increased alkaline phosphatase expression and Von Kossa–stained mineralized matrix formation (Fig. 2H). hMSC59–75 cultures showed reduced alkaline phosphatase upregulation and calcium deposition with osteogenic differentiation (Fig. 2I) when compared with hMSC0–18 cultures at similar PD (Fig. 2C).
Telomere Kinetics during In Vitro Expansion and In Vivo
Length of mTRF was measured on genomic DNA taken from hMSC0–18 cells and hMSC59–75 at each passage when cell numbers were sufficient. A significant decrease was found in hMSC0–18 between the mTRF lengths at primary passage and at senescence by an average total of 1.0 kb ± 0.2 (range 0.4–2.5 kb, p = .002, Fig. 3A). The kinetic study in hMSC59–75 was limited, but no significant difference in the rate of mTRF loss was observed (on average, 88 bp/PD ± 10 for hMSC0–18 and 78 bp/PD ± 34 for hMSC59–75, p = .82). A strong correlation was found between total mTRF loss and total number of PDs occurring between the two measurements (R = .93 Pearson coefficient; Fig. 3B).
Interestingly, the two cultures hMSC0–18E that maintained the spindle-shape morphology and showed longer proliferative capacity also displayed mTRF shortening in a similar range to other hMSC0–18 over the first 25 PDs, but thereafter losses were below detection (Fig. 3A). No telomerase activity was detected in those cultures at 35 PDs (data not shown), and karyotypic analysis carried out at the same time did not show any abnormality (Fig. 3C).
To ascertain whether in vitro aging of MSCs was a natural property of MSC and not due to suboptimal culture conditions, we compared the mTRF of hMSC0–18 (n = 10) with that of hMSC59–75 (n = 5) after an equal number of divisions in culture. All MSC cultures were expanded in vitro, and mTRF was calculated at 16 PDs. We found that mTRF in hMSC0–18 was significantly longer than in hMSC59–75 (11.5 PDs ± 0.2 versus 10.3 PDs ± 0.1, p = .021; Fig. 3D). As the average total mTRF loss was 1.2 kb and the mean age difference between the two groups was 63 years, an average loss of 17 bp per year could be estimated. To determine whether there was an mTRF threshold at which MSCs stop proliferating regardless of their starting mTRF, mTRF length from cells at growth arrest was determined. We found no significant difference in mTRF length at growth arrest between hMSC0–18 and hMSC59–75 (10.4 ± 0.1 kb, n = 10 versus 10.4 ± 0.1, n = 5; p = .795; Fig. 3E). Interestingly, the mTRF length of hMSC59–75 at primary passage was found to be close to the mTRF threshold at which cells stopped proliferating.
Historically, human MSCs have been defined as stem cells  on the basis of their ability to differentiate into multiple cell types and on their extensive proliferative potential. In contrast, more recent studies have shown that while subpopulations of hMSCs retain multipotential capacity through a number of passages, most hMSC cultures tend to progressively lose this ability [12–15]. They commit to the osteoblast lineage or undergo senescence after 20 to 40 PDs [11, 29, 30]. Two morphologically distinct cell types are associated with the progressive change of multilineage to unilineage potential: type I cells that are spindle shaped, grow rapidly, and are associated with greatest potential to expand in culture; type II cells that are flat, broad, and proliferate slowly . The type II cells increase in number with time in culture and show signs of commitment to the osteoblast lineage.
First, our data suggest that the aging process, which encompasses loss of proliferative and differentiation capacity with increased commitment to the osteogenic lineage, is seen not only over time in culture but also with increasing donor age. We observed a slower growth rate and lower number of total PDs in hMSC59–75 cultures than in hMSC0–18. As donors age, we found a reduction in the number of type I hMSCs and a predominance of type II cells even at primary passage. This is consistent with the decrease in the number of CFU-F with donor age shown here and as previously observed .
Second, telomere length analysis suggests a clear correlation between proliferative capacity of hMSCs and telomere length, both in culture and with donor age. As found in our data, most hMSC lines lose telomeres at each cell division until they reach a threshold around 10 kb, where cells stop dividing and assume a senescence-type phenotype. This threshold, higher than in other cell types, is similar for all lines examined, regardless of their starting telomere length in culture, as we found no statistically significant difference in telomere length between hMSC0–18 and hMSC59–79 at the end of their lifespan. The high threshold may suggest that other mechanisms such as accumulated damage may contribute toward telomeres to signal the cell to exit the cycle .
A significant difference in telomere length was seen between hMSC0–18 and hMSC59–75 when compared after 16 PDs of expansion in culture, suggesting that telomere erosion is also a property of MSCs in vivo. The absence of a specific marker for MSCs and their low frequency in bone marrow necessitated minimal in vitro expansion for MSC isolation prior to mTRF determination. As we have shown that an equal number of PDs in vitro causes equivalent telomere erosion in both hMSC0–18 and hMSC59–75, it is fair to deduce that any difference in telomere length between the two groups reflects telomere losses occurred in vivo prior to any in vitro manipulation. To ascertain that cells in the cultures have undergone a similar number of doublings, we used the number of CFU-F in each culture at the start to estimate the number of PDs. Moreover, we cultured the cells for a short time to avoid the development of any dominant clonal component.
An encouraging finding is that two MSC cultures (hMSC0–18E), after an initial telomere shortening, were able to maintain telomere length for over 40 PDs. This correlated with their ability to undergo more divisions, the preservation of a higher number of type I cells in culture, and the retention of in vitro efficient bone formation. The presence of a normal karyotype excludes mechanisms such as alternative lengthening of telomere as being responsible for the absence of telomere erosion . No difference has been found in the bone marrow cellularity, number of progenitor cells, or age of the donor that could explain the extended proliferative capacity. The maintenance of telomere length may indicate a selection of more primitive MSCs within the cultures, as shown for MSC cultures maintained in fibroblast growth factor–2 . This population may be so small that such cells may not always be present in all samples due to the small volumes of bone marrow used. Recently, a very rare primitive MSC subpopulation—multipotent adult progenitor cells (MAPCs)—has been characterized . Similarly to the hMSC0–18E cultures described here, MAPC cultures exhibit no telomere shortening after 35 PDs. This may be due to expression of low levels of telomerase, the RNA-dependent DNA polymerase responsible in most cases for telomere length maintenance . Indeed, hMSC0–18E behaved similarly to human telomerase reverse transcriptase (hTERT), the catalytic subunit of telomerase, when it was ectopically expressed in MSC [36, 37]. hTERT expression abolished the onset of senescence and maintained unlimited proliferative capacity and in vivo bone-forming activity. We were unable to detect any telomerase activity by telomeric repeat amplification protocol (TRAP) assay, in agreement with previously published data [13, 38]. However this may be due to a methodological limitation. hTERT has recently been detected in fibroblasts, previously believed to lack telomerase expression, only by immunoprecipitation and after synchronization of the cells in S-phase and not by TRAP assay . Its abrogation accelerated in vitro senescence.
In this study we used telomere length as a marker of aging, to quantify the remaining replicative capacity following in vitro expansion. Cells with great telomere shortening are expected to have little remaining proliferative capacity. This study highlights the severe aging of MSCs by expansion using present protocols. From our data expansion of about 10 PDs (minimal expansion of MSCs used for transplantation in osteogenesis imperfecta ), leads to an average loss of 1 kb of telomere sequence. Taking into account that the total loss can be at most about 2.5–3.5 kb when cells are derived from young donors, this loss is significant. As we have shown that cells derived from an adult donor have already undergone some substantial telomere erosion in vivo (17 bp/year), the expansion may lead to reinfusion of cells that have short telomeres and are severely compromised in their remaining long-term proliferative, differentiative, and homing capacity at a time when proliferation should be at its best to be able to initiate regenerative processes. The presence of lines where expansion occurs with little or no telomere erosion is encouraging and requires further investigation into the mechanisms of telomere maintenance and on culture conditions where this can reliably be obtained.
We thank Steve Truman for technical help in the cytogenetic analysis and Trevor Carr for technical advice. We are grateful to Dominique Bonnet for critical review of the manuscript and to Richard Swindell for advice on the statistical analysis. M.A.B. and I.B. are supported by the Jeans for Genes appeal (Mucopolysaccharidosis Society, U.K.). L.J.F. is supported by Cancer Research, U.K.