Chronic oxidative stress compromises telomere integrity and accelerates the onset of senescence in human endothelial cells

Diseases of the vascular system have long been considered to be age-related in terms of their onset and progression (Lakatta and Levy, 2003). Ageing is associated with endothelial dysfunction (Celermajer et al., 1994; van der Loo et al., 2000; Weinsaft and Edelberg, 2001), impaired angiogenesis (Rivard et al., 1999), defective vascular repair (Weinsaft and Edelberg, 2001) and with an increasing prevalence of atherosclerosis (Strong and McGill, 1962; Weingand et al., 1986). The reasons for these associations are still unclear, but it is plausible that organismal ageing and vascular disease may share common cellular mechanisms. One phenomenon that has been increasingly implicated both in ageing and in the development of vascular pathologies is cellular replicative senescence. The replicative senescence hypothesis proposes that ageing in vivo is related to the limited replicative capacity that normal somatic cells show in vitro (Hayflick, 1965). Replicative senescence entails an irreversible arrest of cell proliferation along with selective changes in differentiated functions (Campisi, 1996). At the mechanistic level, replicative senescence has been linked to the erosion of telomeres (Campisi et al., 2001). These are high order structures formed by specific DNA sequences and a complex array of specialised proteins that cap and stabilise the physical ends of the chromosomes (Blackburn, 2001). In mammals, telomeric DNA consist of tandem hexanucleotide repeats of the sequence TTAGGG, which extend over several thousand base pairs (bp) in length and end in a G-rich single stranded 3′-overhang. Telomeres fold back on themselves to form a loop known as the T-loop that provides functional stability to the telomere, possibly by intercalating the single strand overhang into the double-stranded region of the telomere with the aid of several specialised telomere-related proteins. Telomeric DNA is synthesised by a specialised reverse transcriptase called telomerase (reviewed by Collins and Mitchell, 2002), an enzyme whose expression levels play an important role in determining the onset of replicative senescence (Bodnar et al., 1998). In the absence of telomerase or when this enzyme is expressed at very low levels, DNA synthesis during cell division results in the progressive shortening of telomeric DNA. This erosion eventually compromises telomere integrity, triggering senescence by activating DNA damage checkpoints that block further cell replication (Smogorzewska and de Lange, 2002).

The progressive oxidative damage of macromolecules resulting from the continual exposure of cellular components to oxidative stress is another process that has long been implicated in ageing and age-related pathologies (reviewed by Finkel and Holbrook, 2000). Interestingly, a substantial body of evidence, mostly from work in fibroblasts, indicates that oxidative stress can also induce or accelerate the development of replicative senescence (reviewed by Serrano and Blasco, 2001). This phenomenon has been collectively termed `stress-induced premature senescence' (Toussaint et al., 2000). In some studies this early onset of senescence has been attributed to accelerated telomere attrition (reviewed by von Zglinicki, 2002), most probably resulting from the generation of single strand breaks in the telomeric DNA (Petersen et al., 1998). Other reports claim that stress-induced premature senescence may not be related to telomere damage (Chen et al., 2001; Toussaint et al., 2002). It is not clear, however, whether these differences are the result of the cell type under study, the type and level of oxidative insult to which the cells were subjected, the degree of antioxidant protection that the cells could exert, and/or to some other experimental differences.

Human endothelial cells have been shown to be more susceptible to oxidative stress than fibroblasts, which results from substantial differences in the make-up of their antioxidant defence mechanisms (Michiels et al., 1990). In endothelial cells the glutathione redox-cycle plays a predominant role as an anti-oxidant defence against peroxides in general (Suttorp et al., 1986) and against homocysteine-induced oxidative stress in particular (Dayal et al., 2002). Therefore an imbalance in this system may have important implications for the onset of endothelial cell senescence in various pathophysiological contexts related to vascular disease, including hypercholesterolaemia, diabetes mellitus, hyperhomocysteinaemia and chronic inflammation. In this study we demonstrated that long-term exposure of human endothelial cells to mild oxidative stress caused by perturbation of the glutathione redox-cycle results in accelerated telomere erosion, loss of telomeric integrity and the premature onset of replicative senescence.

Human umbilical vein endothelial cells (HUVEC) and endothelial cell growth medium 2 (EGM-2) were purchased from Biowhittaker (Wokingham, UK). Tert-butyl hydroperoxide (t-BHP) and L-buthionine-[S,R]-sulphoximine (BSO) were from Sigma. Stock solutions of t-BHP and BSO were freshly made in water and diluted in medium immediately before addition to the cultures. All other reagents were from standard suppliers or as listed in the text.

First passage cryopreserved HUVEC were grown in EGM-2 and serially passaged until they reached senescence, as previously described (Kurz et al., 2000). The number of population doublings (PD) was calculated using the formula PD=(ln[number of cells harvested] – ln[number of cells seeded]) / ln2. To induce oxidative stress, from the second passage onwards parallel cultures were grown in EGM-2 lacking ascorbic acid and exposed at each passage to 0.1 μM t-BHP or 10 μM BSO. These compounds were added to the culture medium 48 hours after seeding and then every 2-3 days at the time of feeding. Pro-oxidant treatments were carried out on five independent cultures (three times with t-BHP, two times with BSO) from two batches of genetically distinct single donor HUVEC, and unless otherwise indicated results from one representative experiment are shown.

Viability was assessed by flow cytometry using dual staining with calcein-AM and ethidium homodimer-1 (Molecular Probes, OR, USA) as previously described (Levesque et al., 1995).

Early passage HUVEC were seeded into 96-well plates at a density of 10⁴ cells per well and grown in EGM-2 for 2 days. Then the cells were washed in Krebs-Ringer-Hepes (KRH) buffer (20 mM Hepes, 145 mM NaCl, 1.8 mM CaCl₂, 5.4 mM KCl, 1 mM MgCl₂, 0.8 mM Na₂PO₄, 5 mM glucose, pH 7.4) and incubated with 100 μM dichlorodihydrofluorescein diacetate (H₂DCF-DA; Molecular Probes) at 37°C for 30 minutes. After washing with KRH buffer, the cells were incubated at 37°C with 0.1 μM t-BHP or 10 μM BSO for 60 minutes in a Cytofluor 4000 plate reader. The fluorescence of 2′,7′-dichlorofluorescein (DCF), the oxidation product of H₂DCF, was detected under an excitation wavelength of 485±10 nm and an emission of 520±25 nm. Alternatively, after incubation with H₂DCF-DA, cultures were washed with ice-cold PBS, resuspended by trypsinization and analysed immediately using a FACScan flow cytometer (Becton Dickinson, Oxford, UK). Data were acquired and analysed with Cellquest software (Becton Dickinson). The DCF signal was measured on the FL1 detector and relative intracellular oxidant capacity was estimated using the median fluorescence intensity of the population.

Cell proliferation was determined using bromodeoxyuridine (BrdU) labelling. Cultures were incubated under normal or pro-oxidant conditions with 10 μg/ml BrdU and after various time intervals fixed in ethanol. Labelled cells were identified by immunocytochemistry using a mouse monoclonal antibody directed against BrdU (clone Bu20a), a biotinylated F(Ab′)₂ fragment of rabbit anti-mouse immunoglobulin as the secondary reagent and streptavidin-horseradish peroxidase with diaminobenzidine for visualisation (all from DAKO, UK). Labelled cells were viewed with a Zeiss Axiovert 25 CFL inverted microscope and the percentage of positive cells was determined by counting at least 500 cells in each culture dish.

Cytochemical staining for senescence-associated β-galactosidase (SA-β-gal) was performed as previously described (Kurz et al., 2000).

Genomic DNA was extracted at the end of each passage and digested overnight at 37°C with HinfI and RsaI. Southern blot determination of mean terminal restriction fragment (TRF) length was performed using a biotinylated telomeric 51-mer oligonucleotide probe (BD Pharmingen, UK) at a concentration of 6.5 ng/ml. Hybridisation was carried out at 65°C for 16 hours in hybridisation buffer (BD Pharmingen) followed by three 5-minute washes with 1× SSC/0.1% SDS at 42°C, and then three 15-minute washes with 0.1 × SSC/0.1% SDS at 42°C. Hybridisation signals were detected by chemiluminescence and quantified by computer-assisted scanning densitometry using the public domain NIH Image software (version 1.62, available at http://rsb.info.nih.gov/nih-image). The mean TRF lengths were calculated by integrating the signal intensity over the entire TRF distribution as a function of TRF length using the formula TRF length=Σ(OD_i) / Σ(OD_i/L_i), where OD_i and L_i are the signal intensity and TRF length respectively at position i on the autoradiogram.

HUVEC were treated overnight with 10 ng/ml colcemid followed by hypotonic swelling and fixation in 3:1 (v/v) methanol/acetic acid. Cell suspensions were then dropped onto slides, post-fixed with 70% acetic acid and dehydrated by passing through graded ethanol solutions following standard protocols. In situ hybridisation was performed using a fluorescein isothiocyanate (FITC)-conjugated (CCCTAA)₃ peptide nucleic acid probe (Perseptive Biosystems) as previously described (Lansdorp et al., 1996). Chromosomes were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and images of metaphase spreads were viewed by fluorescence microscopy. Images were captured using a cooled CCD camera (Photometrics) and Smartcapture software (Vysis). The fluorescence intensities of individual telomeres were calculated from the grey-scale digital images of the metaphase spreads after correction for background, using NIH Image software. Analysis of telomere length heterogeneity was performed by normalising the fluorescence intensity of each telomere spot to the median signal intensity of all telomeres in the metaphase spread.

Bal31 digestion was carried out in 600 mM NaCl, 20 mM Tris-HCl, 12 mM CaCl₂, 12 mM MgCl₂ and 1 mM EDTA (pH 8.0) in a final volume of 25 μl containing 10 μg of genomic DNA and 0.25 units of Bal31 nuclease (New England Biolabs) at 30°C for various lengths of time. The reaction was stopped by addition of 3 μl 0.2 M EGTA and incubation for 10 minutes at 65°C. The digests were then extracted with 25/24/1 (v/v) phenol/chloroform/isoamyl alcohol, precipitated in ethanol, digested with HinfI and RsaI, and processed for Southern blot analysis of TRF length as described above.

HUVEC were harvested by trypsinisation, counted using a haemocytometer, and lysed at 4°C in 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid buffer. Aliquots of the cleared lysate equivalent to 2×10⁴ cells were assayed for telomerase activity by a modified telomeric repeat amplification protocol (TRAPeze, Intergen, Oxford, UK), as previously described (Kurz et al., 2003). The activity of each sample was calculated from the ratio of the intensity of the telomeric repeat ladder (starting at 50 bp) to the 36 bp internal standard. Results, expressed in arbitrary units (AU), were normalized to the signal obtained from an extract of 500 HeLa cells routinely assayed in parallel as a positive control.

Protein oxidation was analysed by immunoblot detection of carbonyl groups following derivatisation with 2,4-dinitrophenylhydrazine and using a commercially available detection system (Oxyblot, Oncor). Briefly, at the indicated passages 7.5×10⁵ cells were extracted in 100 μl NP-40 lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, 1 mM Na₃VO₄, 1% NP-40, 5 mg/ml leupeptin, 1 mM PMSF) for 30 minutes on ice. Following centrifugation at 14,000 g for 30 minutes at 4°C, 1.5% β-mercaptoethanol was added to the supernatant, carbonyl group side chains were derivatised and proteins were separated by polyacrylamide gel electrophoresis. After transfer to PVDF membranes, oxidised proteins were identified by immunoblotting with antibodies against the dinitrophenyl moiety. Immunoreactive bands were detected using a horseradish peroxidase-labelled antibody and enhanced chemiluminescence. The relative intensity of each lane was quantified by scanning densitometry using NIH Image. Immunoblotting for hTERT was performed according to standard protocols using 50 μg of nuclear protein extract and a goat polyclonal antibody against the C terminus of hTERT (Santa Cruz Biotechnologies, CA, USA). Nuclear extracts were prepared as previously described (Andrews et al., 1991).

Association of TRF1 with PML bodies was detected by immunofluorescent staining essentially as previously described (Wu et al., 2000). In brief, cells were grown on coverslips coated with 0.2% gelatine and then arrested in the G2/M phase of the cell cycle by incubation with 500 ng/ml nocodazole (Calbiochem) for 16 hours. Unless otherwise indicated all further incubations were carried out at room temperature. After washing with PBS, cells were fixed and permeabilised for 30 minutes with 3.7% formaldehyde in PBS containing 0.1% Triton X-100, followed by a further 30-minute incubation with 0.05% saponin in PBS. After five washes with PBS, coverslips were pre-treated with serum-free protein blocking reagent (DAKO) for 30 minutes and then incubated with primary antibodies for 1 hour. TRF1 was detected using a 1:1000 dilution of a goat polyclonal antibody raised against a peptide mapping to the amino terminus of human TRF1 (Santa Cruz Biotechnology), followed by a FITC-conjugated chicken anti-goat IgG (Molecular Probes, dilution 1:2000). Detection of the PML nuclear body was carried out using a 1:1000 dilution of a mouse monoclonal antibody raised against a peptide mapping to residues 37-51 of human PML (Santa Cruz Biotechnology), followed by Texas Red-conjugated rabbit anti-mouse IgG (Molecular Probes, 1:2000). A corresponding primary control antibody was used in each case as a negative control. After antibody staining, coverslips were washed extensively in PBS/0.5% Nonidet P-40 and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) in Vectashield mounting medium (Vector Laboratories). Immunofluorescence microscopy was performed with a Zeiss Axioskop Microscope using a 63×/1.4 NA Plan Apochromat objective and fluorescence filter sets appropriate for fluorescein, Texas Red and DAPI. Fluorescence images were acquired with a Hammatsu C4742-95 digital camera using Improvision acquisition and analysis software (Openlab, version 3).

t-BHP, a substrate of glutathione peroxidase, and BSO, an inhibitor of γ-glutamyl cysteine synthase, are two compounds known to interfere with the glutathione-dependent antioxidant defences of the cell (Griffith and Meister, 1979; Jewell et al., 1986) (Fig. 1). In preliminary experiments these compounds were tested to find conditions that would cause chronic oxidative stress during long-term serial culture, without inducing cytotoxic effects. To this end, HUVEC were cultured for four passages in the presence of BSO or t-BHP at various doses. Viability assays using dual staining with calcein-AM and ethidium homodimer-1 demonstrated that the highest concentrations of these compounds that could be used without inducing cell death were 0.1 μM t-BHP and 10 μM BSO (data not shown). These concentrations were thus chosen for subsequent experiments. Fig. 2A shows that under these conditions the intracellular oxidative capacity increased by 2.9±0.7-fold with t-BHP and by 2.2±0.3-fold with BSO. Labelling with BrdU during the first two passages demonstrated that under these pro-oxidant conditions there was a small but noticeable lengthening of the cell-cycle. Nevertheless, consistent with an absence of cytostatic or cytotoxic effects, there was no effect on the overall fraction of cells passing through S-phase (Fig. 2B). Taken together these results indicated that t-BHP and BSO might be used at low concentrations to cause chronic oxidative stress. In addition, immunoblot detection of protein carbonyl groups revealed a significant increase of intracellular protein oxidation in treated cultures (Fig. 2C), demonstrating that this mild oxidative treatment was biochemically effective also at the macromolecular level. Furthermore, the level of protein oxidation appeared to increase with treatment duration, although this was not statistically significant.

To investigate whether long-term pro-oxidant treatment affected the response to further oxidative insults, intracellular oxidant capacity was measured in late passage cultures before addition of the oxidant (at baseline) and immediately after treatment. As shown in Fig. 2D, baseline intracellular oxidative capacities in control and t-BHP-treated cultures (passage 11-12) were comparable, but in BSO-treated cultures (passage 8-10) were increased by approximately fourfold. This difference suggests that BSO induces a more sustained pro-oxidant state, presumably by causing a persistent depletion of glutathione stores, whereas t-BHP, being rapidly metabolised, has a transient effect. However, in both cases exposure to an oxidative challenge with hydrogen peroxide resulted in an increase in intracellular oxidative status of a similar or larger magnitude to that seen in untreated cells (Fig. 2D). This demonstrated that the response to further oxidant stress was not being blunted by adaptive mechanisms, but in contrast, persisted despite long-term culture under pro-oxidant conditions.

Long-term cultures maintained under normal conditions ceased to replicate after about 70 days, reaching 46 cumulative population doublings (CPD). In comparison, parallel cultures treated with t-BHP or BSO, continued to grow for a similar length of time but replicated at a slower rate (Fig. 3A). Evaluation of these cultures for senescence characteristics revealed that under these pro-oxidant conditions there was a premature decrease in cell density at confluence (Fig. 3B), an accelerated accumulation of SA-β-gal positive cells (Fig. 3C) and an increase in cell size (data not shown). Thus, cultures with a replicative age of only about 25-30 CPD grown under oxidative stress resembled senescent cultures of about 45 CPD grown under normal conditions.

Telomere restriction fragment analysis revealed that in HUVEC grown under normal conditions telomere erosion occurred at a constant rate of ∼55 bp/PD and in a relatively uniform manner (Fig. 4A,B). In contrast, telomere shortening in cells cultured in the presence of t-BHP showed a biphasic behaviour. Initially, there was one population of telomeres, which, compared to control cultures, decreased in size at a much faster rate (∼110 bp/PD). However, beginning at about 15-18 CPD, a more complicated pattern became evident involving two distinct TRF populations. One population consisted of the conventionally sized fragments, while the other population was composed of longer, more uniformly sized species. Both of these TRF species continued to decrease in size at similar rates (∼110 bp/PD for conventional sized fragments and ∼100 bp/PD for long fragments). It was noteworthy that when these long TRFs first appeared, they were larger than the longest conventional size TRFs from younger cultures, suggesting that some form of active telomere elongation had occurred. A similar pattern of telomere length changes was also found in cultures treated with BSO (Fig. 4C). These findings were reproducibly present in all HUVEC series grown under pro-oxidant conditions (three times t-BHP, two times BSO).

In order to investigate in more detail the nature of these long TRF species, telomeres were examined on metaphase spreads of intermediate to late-passage cultures by FISH. Fig. 5A shows that telomeric signals from cells exposed to chronic oxidative stress were more heterogeneous than control cultures of similar replicative ages. Thus, while the majority of telomeric signals were clearly reduced, a minority appeared conspicuously brighter (Fig. 5A, arrows). The greater heterogeneity of telomere lengths within individual metaphases of cultures subjected to oxidant stress was confirmed by quantifying the relative fluorescence intensities of the individual telomere spots in ten representative examples each from control and t-BHP treated cells (Fig. 5B). Furthermore, histograms representing the cumulative frequency distributions of individual telomere fluorescence intensities demonstrated that a distinct subpopulation of long telomeres was present in oxidant-treated cultures, consistent with the results of the Southern blot analysis (Fig. 5C). These findings also suggested that the appearance of long telomeres occurred in a subset of chromosomes within the majority of the treated cells.

We further considered the possibility that these long TRF species reflected chromosomal end-to-end fusions or other gross chromosomal abnormalities induced by telomere damage. Although FISH could not detect such aberrations in treated cells, we could not rule out that this was only because cells with these abnormalities would not progress through metaphase. Therefore, digestion of genomic DNA from late passage cultures with Bal31 nuclease prior to TRF analysis was undertaken. As shown in Fig. 6, the susceptibility of telomeres to Bal31 digestion was identical in DNA from all culture conditions, effectively ruling out the presence of chromosomal end-to-end fusions.

The possibility that the long TRF species represented a subpopulation of cells in which telomeres had been elongated as a result of telomerase reactivation was then considered. Consistent with previous reports (Vasa et al., 2000) cultures grown under normal conditions showed a decrease in telomerase activity as a function of replicative age (Fig. 6A,B). Furthermore, in cultures grown under chronic oxidative stress telomerase activity was down regulated even faster, becoming virtually undetectable as early as 12 CPD (Fig. 6A,B). As shown in Fig. 6C, even the first oxidant treatment of HUVEC caused a very rapid (49% of control at 3 hours) and sustained (32% of control at 72 hours) decrease in telomerase activity. Furthermore, western blot analysis demonstrated that nuclear protein levels of hTERT, the catalytic subunit of telomerase, also decreased after the first oxidant treatment (Fig. 7D). Taken together, these results demonstrated that the decrease in telomerase activity caused by oxidative stress was not a consequence of premature senescence, but rather the result of a direct effect. Furthermore, these findings also indicate that appearance of long TRF species is not the outcome of telomerase up-regulation.

A pattern of increased telomere length heterogeneity within a single nucleus, similar to that found by us during long-term oxidant treatment has been demonstrated to be a characteristic of cell lines that use the `alternative lengthening of telomeres' (ALT) mechanism of telomere maintenance (reviewed by Henson et al., 2002). These immortalised cells lack substantial telomerase activity and use instead recombinatorial mechanisms to maintain telomere length. In an attempt to find evidence for the existence of ALT in HUVEC subjected to chronic oxidative stress, we searched for the presence of PML bodies associating with telomeres in the nuclei of these cells. These structures have become known as `ALT-associated PML bodies' (APBs), and have been repeatedly found to be present in ALT cell lines, where they colocalise with the telomere-associated proteins TRF1 and TRF2 (Henson et al., 2002). As shown in Fig. 8, PML bodies were readily visualised in HUVEC by immunofluorescence microscopy. However, neither TRF1 nor TRF2 could be detected by this method in either young control HUVEC or late passage HUVEC grown under pro-oxidant conditions (Fig. 8 and data not shown). In contrast, under the same experimental conditions, APBs were readily detected in the human ALT-positive osteosarcoma cell line U2OS (Fig. 8). In summary, these findings could neither support nor exclude a role for the ALT pathway in the generation of the long telomeres under oxidant culture conditions. Furthermore, they suggest that in HUVEC, the telomere-related proteins TRF1 and TRF2 are expressed at much lower levels than in immortalised cell lines.

The present study demonstrates that in normal endothelial cells mild chronic oxidative stress induced by interference with the glutathione-dependent anti-oxidant defences accelerates telomere erosion and the onset of replicative senescence. Furthermore, we show that under these pro-oxidant conditions, concomitantly with the faster telomere attrition, a distinct fraction of telomeres increase their length in the absence of telomerase activity. Our findings extend evidence from previous studies, the majority of which were performed in cultured fibroblasts (e.g. von Zglinicki et al., 1995; Petersen et al., 1998; Furumoto et al., 1998; Toussaint et al., 2000; Dumont et al., 2000; Xu et al., 2000; Chen et al., 2001; Tchirkov and Lansdorp, 2003) that have shown that both acute and chronic oxidative stress can induce a senescent phenotype. In a number of studies the premature onset of senescence has been further linked to accelerated telomere attrition (reviewed by von Zglinicki, 2002). However, the finding that oxidative stress leads to the emergence of abnormally long telomeres has not previously been reported.

In human endothelial cells the glutathione redox cycle has been shown to play a predominant role in the maintenance of redox homeostasis (Suttorp et al., 1986). Therefore, we chose to interfere with this system as a means to exert physiologically relevant oxidative stress. To this end, we used two distinct compounds, t-BHP and BSO, which are known to affect the metabolism of glutathione at different levels, namely utilisation (Jewell et al., 1986) and synthesis (Griffith and Meister, 1979), respectively. In preliminary experiments we set up culture conditions that would allow us to investigate the long-term effect of mild oxidative stress on telomere dynamics. To achieve this goal, it was essential to avoid acute growth arrest or cytotoxic cell death. In addition, since it has been shown that endothelial cells approaching senescence may undergo spontaneous apoptosis (Wagner et al., 2001; Zhang et al., 2002) it was necessary to define pro-oxidant culture conditions that would not enhance this propensity. By using significantly lower concentrations of t-BHP than those employed in other studies of senescence, we were able to induce a moderate, but effective increase in the intracellular oxidant capacity without causing cell death or cell cycle arrest (Fig. 2A-C). This allowed us to treat the cultures over several months through their whole replicative life span. We undertook a similar strategy with regards to the use of BSO, a substance that has not been employed previously in studies of telomere dynamics. Importantly, both compounds gave essentially identical results throughout the study, indicating that our findings were the consequence of effects related to the control of redox homeostasis.

The involvement of accelerated telomere erosion as a mechanism of stress-induced premature senescence has been a matter of debate (reviewed by von Zglinicki, 2002; Toussaint et al., 2002). In fibroblasts oxidant stress may induce or accelerate telomere shortening via generation of single strand breaks (Petersen et al., 1998). Conversely, it has been proposed that an apparent acceleration of telomere shortening may result from the compensatory expansion of a subpopulation of cells that escapes a stress-induced cell-cycle arrest (Dumont et al., 2001). In the present study, however, this possibility was ruled out by demonstrating in young cultures that cells did not drop from the cell cycle as a result of exposure to mild t-BHP or BSO treatment (Fig. 2B).

A major finding of this study was the discovery that a mild but chronic oxidative burden led to the emergence of distinct TRFs, which were longer and more uniform in size than the conventional TRF species (Fig. 4). Our initial assumption was that these long TRFs represented chromosomal end-to-end fusions, a phenomenon that has been documented in other models of telomere instability (Karlseder et al., 2002). This idea was further based on the fact that the long TRFs were approximately double the size of the shortest TRFs found in the same sample. However, both FISH and Bal31 digestion analysis could not confirm the presence of gross chromosomal abnormalities, making it a very unlikely source of the long TRF species (Fig. 5A, Fig. 6). We were also able to exclude the possibility that these long TRFs resulted from reactivation of telomerase activity (Fig. 7A,B). Indeed, in agreement with previous studies in HUVEC-treated with oxidised low density lipoprotein or H₂O₂ (Breitschopf et al., 2001), we found that oxidative stress generated by t-BHP or BSO in fact down-regulated telomerase in a rapid and persistent matter (Fig. 7C,D). It could be argued however, that up-regulation of telomerase was still occurring in a small subset of cells that could not be detected by the TRAP assay, and that this subpopulation might have been responsible for generating the long TRF species. However, evidence from the FISH analysis demonstrating increased telomere heterogeneity within the majority of the analysed cells did not support this possibility. One could further speculate that the reduction of telomerase activity may have contributed to the generation of these telomeric abnormalities. This hypothesis is consistent with more recent evidence suggesting that in human cells telomerase preserves telomere function and integrity independently of the maintenance of telomere length (Blasco, 2002; Masutomi et al., 2003).

A number of important differences between the present and other studies may explain why the emergence of long TRF species has not been found previously as a result of chronic oxidative stress. First, as stated above, a number of studies were performed with relatively high concentrations of oxidants, which most likely induce higher levels of macromolecular damage, preventing long-term cell culture. Second, the majority of studies were performed on fibroblasts, which are known to behave differently to endothelial cells with regard to their anti-oxidant defence mechanisms, their response to stress and their cell-cycle control (Michiels et al., 1990; Serrano and Blasco, 2001). Third, of the two studies performed on endothelial cells, one examined the effect of anti-oxidant rather than pro-oxidant treatment on telomere erosion (Furumoto et al., 1998), while the other, in which homocysteine was used to generate oxidative stress, did not examine the cultures beyond 15 population doublings (Xu et al., 2000).

It has been postulated that in the absence of telomerase activity telomeres can be elongated by recombinatorial mechanisms collectively known as `alternative lengthening of telomeres' (ALT). Up until now, ALT activity has been found only in telomerase-deficient mouse cells and in human immortalised cell lines and tumours (reviewed by Henson et al., 2002). The present study suggests that recombinatorial mechanisms related to the ALT pathway might also be activated in normal cells when these are subjected to chronic oxidative stress, thus explaining the emergence of the long TRFs. This hypothesis is based on the fact that we have excluded the possibility of chromosomal end-to-end fusions or telomerase reactivation, and that oxidative stress is known to activate the DNA damage repair machinery (Zhou and Elledge, 2000). Indeed, there is some evidence suggesting that the DNA damage repair machinery may be mechanistically involved in the ALT pathway (Henson et al., 2002). In this context it should be emphasized that telomeres, on account of their guanine-rich composition, are several fold more susceptible to oxidative stress-mediated damage than non-telomeric DNA (Retel et al., 1993; Henle et al., 1999) and thus represent a prime site for recombinatorial events. The pattern of increased TRF heterogeneity within the same cell (Fig. 5A-C), a characteristic that has been suggested to be typical of cell lines that use the ALT pathway (Henson et al., 2002), would seem to support this concept. However, at present this possibility should be considered with caution. Indeed, since we were unable to identify the telomere-associated proteins TRF1 or TRF2 by immunofluorescence, we could neither confirm nor exclude the presence of ALT-associated PML bodies in our cultures. Furthermore, other characteristics indicate that HUVEC, with long TRF species, differ from ALT cell lines: (1) the long TRF species in HUVEC continued to decrease in length at a similar rate to the conventionally sized telomere population, whereas in ALT cell lines, these remain stable, and (2) no immortalisation was observed in our cultures. The fact that we could not detect TRF1 and TRF2 in HUVEC, whereas these proteins, and indeed APBs, were readily found in the ALT cell line U20S (Fig. 8), suggests that the levels of expression of these telomere binding proteins is much lower in HUVEC than in immortalised cells.

In summary, the following sequence of events is consistent with the findings of this study. (1) Persistent mild oxidative stress in replicating HUVEC accelerates telomere attrition and the loss of telomerase activity. (2) These events culminate in the disruption of telomere integrity and the consequent triggering of premature senescence. (3) In parallel to, and possibly facilitated by these effects, oxidative DNA damage induces the activation of the DNA damage repair machinery, facilitating recombinatorial events at the telomere, which will lengthen certain telomeres. It remains to be established whether cells in which telomeres have been elongated in this way would have their senescence time point reset and thus continue to replicate. At present we have no evidence suggesting that this process could lead to the immortalisation of endothelial cells.

The present findings may have important implications for the understanding of the relationship between ageing and the development of vascular pathologies. Studies from other laboratories have shown that telomeres in the vasculature shorten with age (Chang and Harley, 1995; Okuda et al., 2000; Aviv et al., 2001) and that this erosion is more pronounced in atherosclerosis-prone areas (Chang and Harley, 1995). In addition, we and others have shown that senescent endothelial cells are present in vivo in pathophysiological contexts (Fenton et al., 2001; Vasile et al., 2001; Minamino et al., 2002) known to be associated with increased oxidative stress (Griendling and Alexander, 1997). Our findings that glutathione-dependent redox homeostasis plays a pivotal role in the preservation of telomere integrity in endothelial cells suggest that perturbation of this system may be an important mechanism for oxidative stress-induced premature senescence of the endothelium and is entirely consistent with a role for telomere dysfunction in the pathogenesis of vascular disease.

This work was supported by grants from the Swiss National Science Foundation (D.J.K.), the Swiss Heart Foundation (D.J.K.) and the British Heart Foundation (S.D., Y.H. and J.D.E.). The authors wish to thank Dr Margareth Fox from the Department of Biology at UCL for her contribution to the FISH analysis of chromosomes.