INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The ends of linear chromosomes are protected by telomeres that consist of double-stranded repetitive sequences complexed with specific telomere-binding proteins. Each organism has a unique telomere length set-point. This set-point is the result of the balance between telomerase activity which extends the telomere (Greider and Blackburn, 1985) and activities that lead to shortening of the telomere such as DNA replication and nuclease activities (Reviewed in Greider, 1996). This equilibrium length maintenance results in a different number of telomere repeats on individual telomeres and heterogeneity in telomere length in a population of cells.

In the yeast Saccharomyces cerevisiae, telomeres consist of ~300 base pairs of TG1-3 repeats (Zakian, 1989). The catalytic components of S. cerevisiae telomerase are the telomerase reverse transcriptase Est2p (Lendvay et al., 1996; Lingner et al., 1997a, 1997b) and the RNA component TLC1 (Singer and Gottschling, 1994). In addition, Cdc13p binds the single-stranded G-rich overhang of the telomere and is required to maintain telomere function (Garvik et al., 1995; Lin and Zakian, 1996; Nugent et al., 1996; Hughes et al., 2000). Cdc13p recruits both telomerase via an interaction with Est1p (Evans and Lundblad, 1999; Qi and Zakian, 2000) and the telomere end-protection complex consisting of Stn1p and Ten1p to the telomere (Lin and Zakian, 1996; Nugent et al., 1996; Grandin et al., 1997, 2001; Pennock et al., 2001).

Telomere function can be perturbed in two different ways. Immediate telomere dysfunction can be induced by disrupting the interaction between the telomere and telomere binding proteins (Garvik et al., 1995; Kirk et al., 1997; Karlseder et al., 1999), whereas inhibition of telomerase activity with subsequent telomere shortening induces delayed telomere dysfunction (Lee et al., 1998a; Hemann et al., 2001a, 2001b).

Cells containing a temperature-sensitive allele of CDC13 grown at the nonpermissive temperature arrest at the G2/M stage of the cell cycle in a RAD9-dependent process (Garvik et al., 1995). In Tetrahymena thermophila, a mutant telomere sequence that prevents telomere binding proteins from interacting with the telomere, causes cells to accumulate in anaphase (Kirk et al., 1997). In transformed and primary human cell-lines, overexpression of a dominant negative allele of the telomere binding protein TRF2 leads to p53 activation and p53- and ATM-dependent arrest or apoptosis (Karlseder et al., 1999).

In S. cerevisiae, inactivation of telomerase results in progressive telomere shortening with accompanying loss of growth potential, chromosome instability, and cell death (Lundblad and Szostak, 1989; Hackett et al., 2001). When the telomeres reach a critical short length, a recombination process mediated by Rad52p and other components leads to telomere elongation that allows a few cells to regain wild-type growth potential (Lundblad and Blackburn, 1993). These survivor cells subsequently reestablish the culture (Lundblad and Blackburn, 1993). Survivors fall into two classes: type I cells that show amplification of the telomere-associated Y' elements and have very short TG1-3 repeat tracts and type II cells that have long variable tracts of TG1-3 repeats and only modest Y' element amplification (Lundblad and Blackburn, 1993; Teng and Zakian, 1999). Type I-mediated recombination depends on RAD51, RAD54, RAD57, CDC13 and the replicative polymerases, whereas type II survivors depend on RAD50, RAD59, MRE11, XRS2, TEL1, and MEC1 (Ritchie and Petes, 2000; Teng et al., 2000; Chen et al., 2001; Tsai et al., 2002). Survivor generation is prevented by either deleting RAD52 alone or deletion of any combination of genes that prevent both type I and type II survivor pathways (Le et al., 1999).

In mice that lack the RNA component of telomerase (Blasco et al., 1997), telomere dysfunction results in T-cell apoptosis, germ cell apoptosis, and testicular atrophy in late generation males (Lee et al., 1998a; Hemann et al., 2001a). Inactivation of p53 in the background of telomerase null mice decreases the degree of testicular atrophy, suggesting that p53 mediates the apoptotic response (Chin et al., 1999). The apoptosis induced in the telomerase null mice is due to the shortest telomere to reach a minimal critical length (Hemann et al., 2001b).

This short dysfunctional telomere may resemble a double-strand DNA break (DSB), which can activate a DNA damage response. DSBs induced by ionizing radiation or the HO endonuclease lead to a G2/M cell cycle arrest that allows time for repair of the damage (Weinert and Hartwell, 1988; Sandell and Zakian, 1993). In S. cerevisiae, the DNA damage response to a DSB is dependent on RAD9, RAD24, RAD17, MEC3, DDC1, DDC2, MEC1, RAD53, and CHK2 (Lydall and Weinert, 1995, 1997; Zhou and Elledge, 2000; Melo and Toczyski, 2002; Rouse and Jackson, 2002). The DSB sensing complex is proposed to consist of RAD24 in a complex with RFC2, RFC3, RFC4, and RFC5. This RFC-like complex might have a function similar to RFCs role of clamploader: recruiting the PCNA complex for initiation of DNA replication. This RFC-like complex recruits a PCNA-like complex consisting of Rad17p, Mec3p, and Ddc1p to the DSB (Shimomura et al., 1998; Kondo et al., 1999, 2001; Green et al., 2000; Venclovas and Thelen, 2000; Melo et al., 2001). This complex activates the signal transducing kinase Mec1p, which is a homolog of human ATR. Mec1p induces the phosphorylation of the effector kinases Rad53p, a homolog of human CHK2 (Allen et al., 1994; Sanchez et al., 1996; Sun et al., 1996) and Chk1p, the homolog of human CHK1(Sanchez et al., 1999).

Activation of Rad53p, which also depends on its interaction with Rad9p, leads to G2/M cell cycle arrest by preventing mitotic exit (Sun et al., 1996, 1998; Charles et al., 1998; Cheng et al., 1998; Emili, 1998; Sanchez et al., 1999; Soulier and Lowndes, 1999; Schwartz et al., 2002). Rad53p also mediates activation of the kinase Dun1p, which controls the transcriptional induction of RNR3, a component of ribonucleotide reductase (Elledge and Davis, 1990; Gardner et al., 1999).

The DSB-induced DNA damage response is also mediated by Tel1p, Mre11p, Rad50p, and Xrs2p. This pathway specifically responds to unprocessed DSBs in the absence of functional Mec1p (Usui et al., 2001). The Mre11p, Rad50p, and Xrs2p (MRX) complex interacts with the DSB and activates the central kinase Tel1p. Tel1p subsequently activates Rad53p and Chk1p to establish G2/M arrest and to induce a DNA damage repair (D'Amours and Jackson, 2001; Grenon et al., 2001; Usui et al., 2001). Interestingly Tel1p, the MRX complex, and MEC1 are also important in the recombination mediated process of type II survivor generation in telomerase mutant cells (Le et al., 1999; Ritchie and Petes, 2000; Teng et al., 2000; Chen et al., 2001; Tsai et al., 2002).

To investigate the primary response to telomere shortening in S. cerevisiae, we examined the rate of chromosome loss, cell cycle arrest, and DNA damage response in telomerase-deficient yeast as these cells were losing growth potential. Chromosome loss occurred, as previously described (Lundblad and Szostak, 1989), however the increase in chromosome loss was only significant after the growth potential of the population reached a minimum. This suggests that chromosome instability is not the primary determinant of the initial loss of growth potential. As the telomerase mutant cells lost growth potential, they accumulated in the G2/M phase of the cell cycle, with Rad53p phosphorylation and RNR3 mRNA upregulation. We propose that in S. cerevisiae in the absence of telomerase, short dysfunctional telomeres induce a DNA damage response, which subsequently leads to the loss of growth potential via a G2/M cell cycle arrest.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Yeast strains constructions and growth media S. cerevisiae strains were grown as previously described (Rose et al., 1990). All strains are derived from JHU500 (YPH983) MATa/MAT ura3-52/ura3-52 lys2-801/lys2-801 ade2-101/ade2-101 his3200/his3200 trp11/trp11 leu21/leu21 CFIII (CEN3.L.YPH983) URA3 SUP11 (kindly provided by Phil Hieter; Connelly and Hieter, 1996). The cells used in the chromosome loss assay carried the chromosome fragment. In all other experiments the chromosome fragment was not present. The strains used in this study are:rad52/RAD52: JHU504(MATa/MAT rad52::TRP1/RAD52)est2 rad52/EST2 RAD52: yAY139-(JHU504 est2::HIS3/EST2);tlc1/TLC1: JHU502(MATa/MAT tlc1:: LEU2/TLC1),tlc1 rad52/TLC1 RAD52: JHU508(JHU502 tlc1::LEU2/TLC1),tlc1 rad9/TLC1 RAD9: yAY212(JHU502 rad9::KANMX4/RAD9)tlc1 rad24/TLC1 RAD24: yAY103(JHU502 rad24::HIS3/ RAD24)tlc1 rad9 rad24/TLC1 RAD9 RAD24: yAY217(yAY212 rad24::HIS3/RAD24)tlc1 mad2/TLC1 MAD2: yAY120(JHU502 mad2::HIS3/MAD2)tlc1 mec1 sml1/TLC1 MEC1 SML1: yAY143(JHU502 mec1::HIS3/MEC1 sml1::TRP1/SML1) tlc1 mec1 sml1 tel1/TLC1 MEC1 SML1 TEL1: yAY222(yAY143 tel1:: HPH/TEL1). All gene disruptions were complete deletions of the open reading frames and were constructed by transforming the cells with either a plasmid-derived restriction fragment or a PCR-generated gene disruption fragment (Brachmann et al., 1998) using the lithium acetate transformation method (Schiestl and Gietz, 1989). Gene disruptions were confirmed by restriction digest and subsequent Southern blotting. JHU502 was constructed by deleting TLC1 in JHU500. JHU500 cells were transformed with XhoI-digested pBLUE61::LEU2 (Singer and Gottschling, 1994). JHU504 was constructed by transformation of BamHI fragment from JH185 into JHU500. JH185 is pBR322 containing the BamHI genomic fragment of RAD52 (JH182 was obtained from David Schild) where the BglII/BglII RAD52 fragment is replaced by a BamHI/BglII fragment containing TRP1 from PRS316. EST2 was deleted in JHU504 using a PCR product generated by AY10 (5'-AAGCATGGCAATGAATGA-CACAAGTGAAATAGAAAAGTGAGATTGTACTGAGAGTGCA-C-3') and AY11 (5'-CAGCATCATAAGCTGTCAGTATTTCATGTATTATTAGTACTGTGCGGTATTTCACACCG-3') on PRS403 (Brachmann et al., 1998). RAD9 was deleted in JHU502 using a PCR product generated by AY41F (5'-CGGCCTTGTTAGCGTTAGAT-3') and AY41R (5'-GACCTTACCAACGTTGTTGG-3') on genomic DNA from haploid rad9::KANMX4 obtained from the S. cerevisiae Genome Deletion Project (Winzeler et al., 1999). RAD24 was deleted in JHU502 using a PCR product generated byAY58F (5'-ATGGATAGTACGAATTTGAACAAACGGCCCTTATTACAAAGATTGTA-CTGAGAGTGCAC-3') and AY58R (5'-TTAGAGTATTTCCAGATCTGAATCTGAAAGGGACTCACTCTGT GC GG TA TT TCA-CACCG-3') on PRS403 (Brachmann et al., 1998). MAD2 was deleted in JHU502 using a PCR product generated by OLFS365 (5'-AGTAATTCCTTGGTCCACTG-3') and OLFS366 (5'-GGTCAAAATTTGTGAGGCAAG-3') on genomic DNA of YFS1104/YPH1238 (kindly provided by Forrest Spencer; Hyland et al., 1999). MEC1 was deleted in JHU502 sml1::TRP/SML1 using a PCR product generated by AY37F (5'-CAAATAGATGGAACGCACGCTCCAAAACTAGTCA-ACTAGAGATTGTACTGAGAGTGCAC-3') and AY37R (5'-AGCCAACCAATATACATCTTGCTTAGATTGTCTTCTGATCTGTGC-GGTATTTCACACCG-3') on PRS403 (Brachmann et al., 1998). SML1 was deleted in JHU502 using a PCR product generated by AY48F (5'-CTGCTCCTTTGTGATCTTACGGTCTCACTAACCTCTCTTA-GATTGTACTGAGAGTGCAC-3') and AY39R (5'-CAATGTTGGCGCTAGCGATATCTAGCTGTATCAAACGTACTGTGCGGTA-TTTCACACGC-3') on PRS404 (Brachmann et al., 1998). TEL1 was deleted in yAY143 using a PCR product generated by AY130F (5'- GATGAGCTAACAACAATTTTAAAAGAAGATCCGGAAAGG-ATACCAAGATTGTACTGAGAGTGCAC-3') and AY130R (5'-AAGCATCTGCATAGCAATTAATAAAAAGGTGACCATCCCA-GCATAGGCCACTAGTGGATCTG-3') on pAG32 (Goldstein and McCusker, 1999). All deletion strains were always maintained as heterozygous diploids. All haploid deletion mutants used were generated by sporulation of the heterozygous diploids, followed by tetrad dissection and genotypic analysis of all the spores.

Chromosome Loss Assay

The chromosome loss assay was carried out as described (Spencer et al., 1990). In this assay the loss of an artificial chromosome CFIII (CEN3.L.YPH983) URA3 SUP11 is quantitated. This ~150-kb chromosome fragment, containing most of the left arm of chromosome III, is marked by the URA3 gene and carries an ochre suppressor of the ade2-101 allele, SUP11. The loss of this and similar chromosome fragments in wild-type cells is 1.7 chromosome loss events/10,000 mitosis (Hegemann et al., 1988). In our sectoring assay only colonies that display half sectoring were counted as a chromosome loss event. In our assay, we kept the cultures in log phase while maintaining selection for the URA3 gene on the chromosome fragment. After every five population-doublings, the cells were spread to single colonies on nonselective YPD plates. The five population doublings were obtained by measuring the starting concentration of the culture by hemacytometer and monitoring the population until the culture had reached a concentration of (starting concentration ×2⁵). For every time-point, a total of 20,000 colonies were analyzed for the appearance of half white/half red sectored colonies. Plating efficiency was normalized to the wild-type plating efficiency: ([actual colonies of sample on the plate/calculated number of cells spread on plate]/[actual wild-type colonies on the plate/calculated number of cells spread on plate])*100%.

Liquid Growth Potential Assay

Haploid cells of the appropriate genotypes were picked from a fresh dissecting plate after 48 h at 30°C and grown in yeast extract-peptone-dextrose (YPD) at a starting concentration of 1 × 10⁴ cells/ml. Cells were kept in log phase for 17 h, and then the cell density was measured by counting cells in a hemacytometer (Bright-Line, Hausser Scientific, Horsham, PA). The culture was then diluted back to a density of 1 × 10⁴ cells/ml. This cycle was repeated for 8-11 d. At all time points during growth, cells were examined for possible contamination. Every day, the concentration of the culture after 17 h of growth was plotted on a curve. This curve represents the growth potential of the different mutants. Wild-type cells when started at 1 × 10⁴ cells/ml reach a maximum concentration of 5 × 10⁷ cells/ml after 17 h of growth. For all liquid growth potential experiments, all haploid cells compared within one experiment were derived from sporulation of one parent diploid strain.

Cell Viability Assay

To measure cell viability, the cells were grown for 4 h in YPD medium containing phloxin at a concentration of 20 µg/ml (Schupbach, 1971; Kohli et al., 1977). This dye enters metabolically dead cells and therefore can be used to quantitate percent of dead cells in a population. For every data-point, 2000-2500 cells were analyzed in a hemacytometer for the presence of red cells.

Individual Cell Growth Rate Assay on Plates

Cells from various time points during the growth potential assay were plated onto YPD plates and allowed to grow for 6 h. The growth rate during this time was measured by capturing an image directly after plating the cells and again at 6 h after growth at 30°C. Images were acquired at 400× magnification using a Zeiss Axioskop and images were captured with a CCD camera (Photometrix Sensys) and processed with IP-Lab Spectrum acquisition software (Scanalytics). The number of divisions was then calculated by a comparison of the two images. Only cells that started as a single cell, a doublet, or triplet were measured.

Division number was calculated as follows: For the single cells, after 6-h growth, a single cell was counted as 0 divisions, increase to 2-3 cells was counted as 1 division, increase to 4-6 cells was counted as 2 divisions, increase to 7-12 cells was counted as 3 divisions, and increase to 13 or more was measured as 4 divisions. For the doublets, after 6 h growth, 2-3 cells was counted as 0 divisions, increase to 4-6 cells was counted as 1 division, increase to 7-12 cells was counted as 2 divisions, increase to 13-24 cells was counted as 3 divisions, and increase to 25 or more was measured as 4 divisions. For the triplets, after 6 h growth, 3 or 4 cells was counted as 0 divisions, increase to 5-9 cells was counted as 1 division, increase to 10-18 cells was counted as 2 divisions, increase to 19-36 cells was counted as 3 divisions and increase to 37 or more was measured as 4 divisions. 60 microcolonies were analyzed for every time point.

Cell Cycle Stage Quantitation

Cell cycle stage quantitation was performed on DAPI-stained cells, using fluorescence microscopy. Cells from log phase cultures were fixed in 70% ethanol. After rehydration the cells were resuspended in 0.5 mg/ml DAPI in 10% glycerol and mounted on a slide. For quantitation purposes, the cells were divided into four stages: G1/S, S/G2, G2/M, and M/G1, based on morphology (Hartwell, 1974). Images were acquired at ×1000 magnification using a Zeiss Axioskop, and images were captured with a CCD camera (Photometrix sensys) and processed with IP-Lab Spectrum acquisition software (Scanalytics). Some cells were not quantifiable; those were put into a separate group, termed "others." Single unbudded cell with one nucleus, were designated as G1/S, a mother cell with a small bud and nucleus close to budneck was designated S/G2, a mother cell with equal sized daughter and nucleus at budneck was designated G2/M, and mother cell with smaller or equal sized daughter, each containing one nucleus, was designated M/G1 cells. For every time-point at least 250 cell-bodies were counted. For cell cycle quantitation experiments, all haploid cells compared within one experiment were derived from sporulation of one parent diploid strain. The data for each mutant is derived from the analysis of one mutant spore, which is representative for the multiple isolates of the same genotype analyzed. In these experiments only comparisons between haploid genotypes that were sporulated from a given multiple heterozygous diploid were made, because the starting telomere length may differ in the independent experiments. In addition, although the results for independent spores with the same genotype within an experiment were qualitatively very similar, there still was spore to spore variation in the timing of onset of loss of growth potential and appearance of survivors.

Total RNA blot analysis

Cells from log phase cultures from different days of the growth potential assay were frozen on dry ice and ethanol and stored at 80°C. Total yeast RNA was prepared by a hot acidic phenol extraction method (Collart and Oliviero, 1993). RNA concentration was measured using GeneSpecI spectrophotometer (Hitachi). Twenty-five micrograms of total RNA was fractionated in a formaldehyde agarose (1%) gel and subsequently transferred to a Hybond-N⁺ nylon membrane (NEN Life Science Products) in 10× SSC. The RNA was cross-linked to the filter by UV irradiation in a UV Stratalinker 2400 (Stratagene). Probes were made by random-primed labeling using Klenow enzyme (New England Biolabs) and [-³²P]dGTP and [-³²P]dATP (NEN Life Science Products) at 30°C for 4 h (Feinberg and Vogelstein, 1983). The actin-specific probe template spans the ACT1 coding region from nucleotide 339 to nucleotide 1421. It was amplified using AY43F-2 (5'-GATAACGGTTCTGGTATG-3') and AY43R (5'-GGTGAACGATAGATGGACCA-3') using yeast genomic DNA as a template. The RNR3 probe template spans the RNR3 coding region from nucleotide 51 to nucleotide 1532. It was amplified using AY42F (5'-ATTACCTCCCGTATCACCCG-3') and AY42R (5'-ACCCTGGACACCAAGAGCAA-3') using yeast genomic DNA as template. Before adding to the formamide hybridization solution, labeled probes were purified from unincorporated [-³²P]dNTP using NAP^TM5 columns (Pharmacia Biotech). Hybridization was performed at 42°C for 16 h. The filters were washed in 0.1 × SSC/0.1% SDS at 42°C three times for 20 min. Quantitation was performed using a BAS1500 phosphoimager (Fujifilm) and Image Quant (Molecular Dynamics) software.

Western Analysis

For each sample, 3 × 10⁸ cells from log phase cultures from different days of the growth potential assay were frozen on dry ice and ethanol and stored at 80°C. Protein extract was prepared using the glass bead lysis method. The cell pellet was thawed and resuspended in yeast lysis buffer (50 mM Tris, pH 7.5, 1% SDS, 5 mM EDTA, 0.001% -mercapto-ethanol) containing protease and phosphatase inhibitors. Glass beads were added to the meniscus and the cells were vortexed 4-5 times for 30 s pulses with 1 min on ice in between the pulses to keep the samples cold. Next the lysate was boiled for 3 min after which 50 µl of fresh YLB was added with subsequent vortexing. The sample was centrifuged at 4°C for 20 min. Then the lysate was transferred to a new tube and protein concentration was determined using the Bradford assay in a UV160U spectrophotometer (Shimadzu). Forty micrograms of total protein extract was fractionated on a 8% SDS-polyacrylamide (SDS-PAGE) gels, and transferred to PVDF Immobilin-P membranes (Millipore) using the TRANS-BLOT SD semi-dry transfer cell (Bio-Rad). The protein blot was incubated for 12 h at 4°C with goat polyclonal IgG anti-Rad53p (Santa Cruz) at 1:300 dilution in 5% milk TTBS (100 mM Tris, pH 7.5, 2.7 mM KCl, 137 mM NaCl, 0.1% Tween-20). After washing in TTBS, the blot was incubated with rabbit anti-goat HRP IgG (H+L; Jackson Immunoresearch) 1:10,000 dilution in 5% milk TTBS for 1 h at room temperature. After washing in TTBS, the signal was generated using ECL Western blotting detection reagents (Amersham Pharmacia Biotech). The signal was detected using X-OMAT AR film (Kodak) developed in M35A X-OMAT processor (Kodak).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Chromosome Loss and Cell Death Do Not Account for the Decrease in Growth Potential in Telomerase-deficient Cells

When telomerase mutants are grown in culture, telomere shortening is apparent in the first few divisions, however, cell growth potential is normal until about the 40th division, when a decline in growth potential is seen (Lundblad and Szostak, 1989; Singer and Gottschling, 1994). The term "loss of growth potential" is introduced here to describe the difference in growth of the telomerase mutant and the wild-type culture and is neutral with regard to the mechanistic basis of this poor growth. This decrease in growth potential, previously termed senescence, has been attributed to loss of telomere function followed by chromosome instability and subsequent cell death (Lundblad and Szostak, 1989).

To investigate the role of chromosome instability in the loss of growth potential in telomerase mutants in more detail, we carried out a quantitative chromosome loss assay (Hieter et al., 1985; Spencer et al., 1990). Telomerase was inactivated by deleting EST2 encoding the catalytic protein component of telomerase. The est2 strain contained a nonessential 150-kb artificial chromosome that carries most of the left arm of chromosome III and the SUP11 gene, which encodes a tRNA that suppresses the ochre mutation of ade2-101 (Spencer et al., 1990). Colonies with an ade2-101 mutation that carry the suppressor are white; however, loss of the chromosome with SUP11 results in red colonies. With this strain, chromosome loss can be quantitated by measuring the red sectoring in white colonies. Half-sectored colonies represent loss of the artificial chromosome at the first division of a single cell (Hieter et al., 1985). We grew the telomerase mutant cells in liquid culture while maintaining selection for the artificial chromosome. Every five population doublings, we plated the cells on nonselective plates and counted the number of half-sectored colonies in a total of 20,000 colonies (Figure 1B). At the same time the growth potential was measured by monitoring the plating efficiency of the cells (Figure 1A).

View larger version (48K): [in this window] [in a new window]   Figure 1.   Chromosome loss and plating efficiency of wild-type, est2 and rad52 cells. (A) The plating efficiency of all cells was measured and was normalized to the wild-type plating efficiency. For plating efficiency, wild-type is denoted as blue diamonds, est2 as red circles and rad52 as green triangles. (B) Chromosome loss was measured in the wild-type, est2, and rad52 cultures. Wild-type is denoted as blue filled bars, est2 as red horizontally striped bars, and rad52 as green hatched bars. The rad52 cells were only grown until generation 75 because of a loss of function mutation in the SUP11 gene after this point. No chromosome loss data was generated for wild-type cells at generation 35 and for rad52 cells at generation 45.

Hegemann et al. (1988) determined that the loss of the chromosome fragment in wild-type cells was 1.7 chromosome loss events/10,000 mitosis. We found a similar average chromosome loss of 1.9 loss events/10,000 mitosis in our wild-type cells. Chromosome loss in the est2 cells was comparable to that of wild-type cells for the first 50 generations. Chromosome loss of the est2 cells only increased to threefold higher than that of the wild-type cells by the time the est2 population displayed a decreased growth potential at generation 60. The chromosome loss was maximal when the growth potential was at a minimum at 75 generations, at which point the est2 cells displayed a 17-fold higher chromosome loss than the wild-type cells. tlc1 cells showed a similar increase in chromosome loss compared with est2 cells (our unpublished results). rad52 cells, which are known to have a high chromosome loss (Mortimer et al., 1981; Sandell and Zakian, 1993), had a five- to eightfold higher chromosome loss than wild-type cells throughout the experiment and did not show a progressive decrease in growth potential. The chromosome loss in the est2 culture was less than the loss in rad52 cultures for the first 70 generations, although the est2 cells lost growth potential and the rad52 cells did not. This suggests that that increased chromosome loss is not solely responsible for the decreased growth potential of est2 cells.

We next tested whether the loss of growth potential correlated with an increased rate of cell death. Cell viability was measured directly by assaying the percent of cells that stain with phloxin (Figure 2). Phloxin dye only enters cells that are metabolically dead (Schupbach, 1971; Kohli et al., 1977). All cells were grown in log phase culture for 10 d and the growth potential of tlc1 and wild-type cells was monitored each day (Figure 2A). Cells were kept in log-phase by starting the cultures at a concentration of 1 × 10⁴cells/ml. The cultures were then grown for 17 h during which time the wild-type cells reach a concentration of 5 × 10⁷cells/ml. Cells with reduced growth potential reached lower concentrations after 17 h of log-phase growth. The cell concentration was measured after growth for all cultures and the cells were rediluted to 1 × 10⁴ cells/ml and grown for another 17 h. We will refer to this assay as the growth potential assay. The increase in the average population doubling time in this assay correlated well with the decrease in colony forming units shown in Figure 1A . 

View larger version (22K): [in this window] [in a new window]   Figure 2.   Cell viability of wild-type and tlc1 cells in the growth potential assay. (A) Wild-type (black lines) and tlc1 (gray lines) cells were grown in log-phase for 11 d. Every day the growth potential of the cultures was measured. For each genotype four independent isolates were analyzed. (B) The percent of phloxin positive cells was plotted for wild-type (black line) and tlc1 cells (gray line) derived from the corresponding days of the growth potential assay in A.

To measure viability, each day an aliquot of the cells was tested for phloxin staining (Figure 2B). Four independent wild-type and four independent tlc1 cultures were tested. Wild-type cells grown for several days in liquid culture showed no decline in growth potential and no phloxin staining. In tlc1 cultures there was an increase in phloxin staining as the growth potential declined. At day 4 of the growth potential assay, 2% of the tlc1 cells stained with phloxin. On day 6 or 7, when tlc1 cells were at the lowest point of the growth potential assay, only between 5 and 12% of the tlc1 cells stained with phloxin. If cell death were the direct cause for the decreased growth potential, a much higher cell death rate would be expected. These results suggest that most of the cells that have stopped growing in a tlc1 culture are still metabolically viable.

Individual Telomerase Mutant Cells Show Decreased Growth Rate

Growth of cells in liquid culture only gives information on the mean growth potential of the population. To evaluate the growth rate of individual cells, we measured the growth rate of wild-type and tlc1 cells in a microscopic assay. After each day of growth in liquid culture (Figure 3A), single cells were plated on YPD plates and examined for their ability to form microcolonies (Figure 3B). The number of divisions that each individual cell underwent was monitored during a 6-h time span (Figure 3C). Most wild-type cells divided three times during this time. On the first day of growth, most of the tlc1 cells also divided three times. However starting at day 3, 80% of tlc1 cells divided only 0-2 times and by day 5; 90% of tlc1 cells did not divide at all or divided only once. Fifty percent of tlc1 cells were not dividing at day 5, which is significantly higher than the 12% phloxin-positive cells in the viability assay, suggesting that most nondividing cells were arrested rather than dead. The heterogeneity in the timing of the arrest of individual cells is likely due to the heterogeneity of the initial telomere length in the population.

View larger version (42K): [in this window] [in a new window]   Figure 3.   Individual telomerase mutant cells show decreased growth potential. (A) The growth potential assay was carried out for wild-type () and tlc1 () cultures. (B) Wild-type and tlc1 cells from day 5 of the growth potential assay were plated out and incubated at 30°C for 6 h. Pictures were taken at the beginning (0 h) and the end (6 h) of the experiment. (C) Quantitation of the growth potential of individual wild-type () and tlc1 () cells. The increase in cell number in the microcolonies in the 6-h time span was counted and converted to total number of divisions. All images were captured at 400× magnification.

Telomerase-deficient Cells Accumulate in the G2/M Phase of the Cell Cycle

To examine whether the loss in growth potential is caused by a cell cycle arrest, the distribution of cells at specific cell cycle stages was examined microscopically. Wild-type and tlc1 cells were grown in log phase for 9 d (Figure 4A). Cells were examined at days 1, 3, and 5 when the culture was losing growth potential. With increasing days of growth in liquid culture, the tlc1 cells accumulated as dumbbell-shaped cells with the nucleus in the bud neck. In addition the tlc1 cells increased in size as the growth potential of the culture declined. Both the G2/M accumulation and the cell size increase were dependent on the DNA damage checkpoint gene RAD24 (Figure 4B). Next, a quantitative assay was performed to measure the amount of G2/M arrest by counting the number of cells in four defined cell cycle stages. Each day the number of unbudded cells (G1/S), cells with small buds (S/G2), dumbbell-shaped cells with one nucleus (G2/M), and dumbbell-shaped cells with two nuclei (M/G1; Hartwell, 1974) were quantitated (Figure 5, A and B). The results from one isolate for each genotype are shown. However each genotype was analyzed multiple times and similar results were obtained in the independent experiments (see MATERIALS AND METHODS). The wild-type cultures displayed a distribution of G1/S (15-20%), S/G2 (55-65%), and M/G1 (10-25%), with only a small fraction (2-4%) of the cells in G2/M for all days (Figure 5B and unpublished data). The wild-type cells retained normal cell size throughout the experiment. In contrast, the tlc1 cells only had a cell cycle distribution and cell size similar to that of wild-type cells on the first day of growth. On day 4, the tlc1 culture contained 38% cells in G2/M and on days 5 and 6, these numbers were 52 and 58%, respectively (Figure 5B). Most of these G2/M-arrested cells were large-sized cells (Figure 4B and unpublished data). At day 10, after survivors were generated, the tlc1 cells reverted to wild-type cell size and showed a cell cycle distribution similar to wild-type cells. These results show that an increasing proportion of tlc1 cells arrest at G2/M as telomeres undergo progressive shortening. The increase in cell size could be due to continued metabolic growth of the arrested cells.

View larger version (71K): [in this window] [in a new window]   Figure 4.   Telomerase-deficient cells accumulate in the G2/M phase of the cell cycle. (A) A growth potential assay was carried out for wild-type () and tlc1 () and tlc1 rad24 (). (B) DAPI (red) stained wild-type, tlc1 and tlc1 rad24 cells from days 1, 3, and 5 of the growth potential assay. All images were captured at the same (×1000) magnification.

View larger version (33K): [in this window] [in a new window]   Figure 5.   Accumulation of G2/M-arrested cells in tlc1 cells depends on RAD24 and to a lesser extent on RAD9. (A) Growth potential assay for wild-type (gray diamonds), tlc1 (black circles), tlc1 rad9 cells (red X), tlc1 rad24 cells (blue squares) and tlc1 rad9 rad24 (green triangles) (B) Quantitation of cell cycle distribution of DAPI-stained cells. Wild-type (gray bars), tlc1 (black bars), tlc1 rad9 (horizontally red striped bars), tlc1 rad24 (hatched blue bars) and tlc1 rad9 rad24 (vertically striped green bars) cells were divided into different cell cycle stages as described in the materials and methods and as depicted at the bottom of the graphs. Wild-type cells are only shown for first and last day of experiment but showed similar profiles for days 4, 5, and 6. The data for each mutant is derived from the analysis of one mutant spore that is representative for the multiple isolates analyzed. For each mutant three independent isolates were analyzed; for the tlc1 rad9 rad24 mutant two independent isolates were analyzed.

Accumulation of G2/M-arrested Cells in Response to Short Telomeres Depends on RAD24 and MEC1

To examine if the G2/M arrest in tlc1 cells was mediated by a DNA damage checkpoint, the requirement of specific checkpoint genes for this arrest was tested. The cell cycle distribution of mutants deleted for TLC1 in combination with either rad9, rad24, rad9 rad24, mec1, tel1, or mad2 was investigated. As described above, tlc1 cells showed an increase in the number of G2/M-arrested cells. In contrast, the tlc1 rad24 double mutants showed a significant reduction in the number of arrested cells at all points in the growth potential assay (Figures 5B). Although there still remained a fraction of G2/M-arrested cells in the tlc1 rad24 mutants, these were predominantly normal-sized cells. A less pronounced reduction in G2/M-arrested cells was seen when the tlc1 mutant was compared with the tlc1 rad9 cells (Figures 5B). However, the tlc1 rad9 cells showed a similar decrease in large cells compared with the tlc1 rad24 mutants (our unpublished results). Deletion of both RAD9 and RAD24 in the tlc1 cells, again resulted in a reduction of G2/M-arrested cells, similar to that observed in the tlc1 rad24 double mutants (Figure 5B).

The role of MEC1 and TEL1 in the short telomere mediated cell cycle arrest was also investigated (Figures 6 and 7). Inactivation of MEC1 in a tlc1 mutant, abrogated the G2/M arrest (Figure 6, A and B). However, a deletion of TEL1 did not significantly change the cell cycle distribution in the tlc1 cells (Figure 7, A and C). These results imply that the cell cycle arrest is mediated, at least in part, by RAD9, RAD24, and MEC1.

View larger version (28K): [in this window] [in a new window]   Figure 6.   Accumulation of large G2/M-arrested cells in tlc1 cells depends on MEC1. (A) Growth potential assay for wild-type (), tlc1 () and tlc1 mec1 sml1 cells (). (B) Quantitation of cell cycle distribution of DAPI-stained cells. Wild-type (white bars), tlc1 (black bars) and tlc1 mec1 sml1 (dotted bars) cells. Cells were divided into different cell cycle stages as depicted at the bottom of the graphs. SML1 was also deleted in these cells to rescue the lethality of a MEC1 deletion. Deletion of SML1 in the background of tlc1 did not change the G2/M cell cycle arrest. Wild-type cells are only shown for first and last day of experiment but showed similar profiles for days 4, 6, and 7. The data shown for each mutant is derived from the analysis of one mutant spore. Another isolate that was analyzed for each mutant gave similar results.

View larger version (55K): [in this window] [in a new window]   Figure 7.   Accumulation of G2/M-arrested cells in tlc1 cells does not depend on TEL1 and MAD2. (A) Growth potential assay for wild-type (), tlc1 () and tlc1 tel1 cells (filled triangles). (B) Growth potential assay for wild-type (), tlc1 () and tlc1 mad2 cells (). (C) Quantitation of cell cycle distribution of DAPI-stained cells. Wild-type (white bars), tlc1 (black bars) and tlc1 tel1 (hatched bars) cells. Cells were divided into different cell cycle stages as depicted at the bottom of the graphs. (D) Quantitation of cell cycle distribution of DAPI-stained cells. Wild-type (white bars), tlc1 (black bars) and tlc1 mad2 (waved bars) cells. Cells were divided into different cell cycle stages as depicted at the bottom of the graphs. Wild-type cells are only shown for first and last day of both experiments but showed similar profiles for the other days that were tested. The data shown for each mutant is derived from the analysis of one mutant spore. Another isolate that was analyzed for each mutant gave similar results.

In Tetrahymena thermophila, a mutant telomere sequence disrupted telomere length maintenance and resulted in the accumulation of cells in anaphase, possibly because of a problem in chromosome segregation (Kirk et al., 1997). To investigate the role of the spindle checkpoint in the observed G2/M arrest, MAD2 was deleted in the tlc1 cells. Quantitation of the cell cycle stages revealed no difference between the tlc1 and tlc1 mad2 cells (Figure 7, B and D). Thus, the spindle checkpoint does not play a major role in the accumulation of tlc1 cells at G2/M.

Short Telomeres Induce an Increase in RNR3 mRNA and Rad53p Phosphorylation

To further examine whether short telomeres activate a DNA damage checkpoint, we examined the induction of RNR3 mRNA and Rad53p phosphorylation. The same samples analyzed for cell cycle distribution (Figure 4) were analyzed for RNR3 mRNA expression. There was an increase in the RNR3 mRNA beginning at day 3 in tlc1 cells (Figure 8A). As the number of G2/M-arrested cells increased (see Figure 4, A and B), the RNR3 mRNA expression level reached a maximum on day 5 (Figure 8A). This increase in RNR3 was not seen in the tlc1 rad24 double mutant (Figure 8B; see also Figure 4, A and B), indicating that the upregulation of the RNR3 mRNA is due to a RAD24-dependent DNA damage checkpoint response.

View larger version (69K): [in this window] [in a new window]   Figure 8.   RNR3 mRNA induction and Rad53p phosphorylation in tlc1 cells. (A) Northern blot analysis was performed for the induction of RNR3 mRNA in tlc1 and (B) tlc1 rad24 cells at different days of the growth potential assay (see Figure 4A). The ACT1 mRNA signal was used to normalize the RNR3 mRNA signal. As a positive control for RNR3 induction wild-type cells were treated with 0.04% MMS for 2 h. Quantitation of the signal is plotted below the blot. (C) Rad53p phosphorylation was assayed by Western blot. Rad53p is indicated by an arrow and phosphorylated species of Rad53p are indicated by the bracket. Cells from day 1 and day 5 of the growth potential assay (Figures 5A and 6A) were analyzed. As a positive control for Rad53p phosphorylation wild-type cells were treated with 0.04% MMS for 2 h with 1 h recovery.

We next tested whether Rad53p was phosphorylated in tlc1 cells. Rad53p phosphorylation can be measured by the appearance of slower migrating species of Rad53p on a Western blot (Allen et al., 1994; Sanchez et al., 1996; Sun et al., 1996). Rad53p is phosphorylated in response to treatment with methyl methanesulfonate (MMS), ionizing radiation, and HO-endonuclease-induced DSB formation (Allen et al., 1994; Sanchez et al., 1996; Sun et al., 1996; Pellicioli et al., 1999). When wild-type yeast cells were treated with MMS or bleomycin, a DSB-inducing agent (Grenon et al., 2001), the slower migrating species of Rad53p were observed (Figure 8C and our unpublished results). These slower migrating species disappeared upon calf intestinal phosphatase treatment, indicating that the shift in migration is due to phosphorylation (our unpublished results). After 1 day of growth in culture, Rad53p was unphosphorylated in tlc1 cells; however, at day 5, the phosphorylated form of Rad53p was seen. Inactivation of RAD9 in the tlc1 cells resulted in a complete loss of Rad53p phosphorylation, whereas inactivation of RAD24 in the tlc1 cells decreased the level of Rad53p phosphorylation but did not abolish it (Figure 8C). The tlc1 rad9 rad24 triple mutant displayed complete loss of Rad53p phosphorylation similar to the tlc1 rad9 cells. In addition a tlc1 mec1 sml1 triple mutant strongly reduced the Rad53p phosphorylation, whereas the tlc1 sml1 double did not. These experiments show that tlc1 cells arrest in the G2/M phase of the cell cycle due to a RAD24, MEC1-dependent DNA damage response.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Telomere Dysfunction Due to Short Telomeres Initiates a G2/M Checkpoint Arrest

Double-strand DNA breaks induce a cell cycle arrest as a mechanism to allow DNA repair before proceeding with cell division (Weinert and Hartwell, 1988). In mammalian cells, short telomeres or a loss of telomere structure can activate a checkpoint response that results in cell cycle arrest or apoptosis (Lee et al., 1998a; Karlseder et al., 1999; Hemann et al., 2001a). This checkpoint response requires p53 (Chin et al., 1999; Karlseder et al., 1999) and in some cases ATM (Karlseder et al., 1999). This suggests that, when telomere function is lost, the chromosome end is sensed by the cell as a double-strand DNA break and activates a checkpoint response (reviewed in Hemann et al., 2000; de Lange, 2002).

In yeast cells, experiments have shown that loss of telomere function leads to chromosome loss, chromosome rearrangements and chromosome end-to-end fusion (Lundblad and Szostak, 1989; Sandell and Zakian, 1993; Hackett et al., 2001). Cell cycle arrest occurs in yeast in response to DNA damage such as double-strand breaks (Weinert and Hartwell, 1988; Sandell and Zakian, 1993). In addition there is evidence that telomere dysfunction may induce a cell cycle checkpoint (Garvik et al., 1995; Lydall and Weinert, 1995; Ritchie et al., 1999; Johnson et al., 2001). To examine whether chromosome loss or cell cycle arrest is the major contributor to the decreased growth potential in the absence of telomerase, we used a quantitative assay to measure chromosome loss rates. After the initiation of telomere shortening by telomerase inactivation, chromosome loss occurred, but the amount of loss did not explain the decreased plating efficiency of the est2 population. In contrast, the fraction of G2/M-arrested cells did correlate with the loss of growth in a telomerase null culture.

Inactivating components of the checkpoint pathway, RAD9, RAD24, and MEC1 resulted in loss of the DNA damage response and decreased cell cycle arrest; thus, we conclude that loss of growth potential in a telomerase null culture is primarily due to a DNA damage-induced G2/M cell cycle arrest. This damage response could be a direct response to short telomeres or a response to breaks induced by chromosome fusions.

RAD24 and MEC1 are required for the cell cycle arrest response to short telomeres, whereas RAD9 plays a minor role in mediating the arrest. In contrast, deletion of RAD9 dramatically reduced Rad53p phosphorylation, whereas a deletion of RAD24 gave only partial reduction in phosphorylation. This suggests that telomere damage signaling might occur via kinases downstream of RAD24 and MEC1 and that the interaction between RAD9 and RAD53 is not essential for this process. It was previously shown that RAD9 and RAD24 have a similar contribution to the damage response in the temperature-sensitive mutant cdc13-1 at nonpermissive temperature (Lydall and Weinert, 1995). This might indicate that the damage response to a telomere of wild-type length that abruptly loses end protection via the loss of Cdc13p is subtly different from the damage response to short telomeres. RAD9 and RAD24 single mutants are known to act somewhat synergistically in their response to DNA damage in several experimental settings such as response to UV irradiation. A more severe loss of G1/S and G2/M arrest responses is often seen when both genes are mutated together (Eckardt-Schupp et al., 1987; Lydall and Weinert, 1995; de la Torre-Ruiz et al., 1998). However, our results indicate that telomere damage is mainly sensed via RAD24 because the tlc1 rad24  mutants and the tlc1 rad9 rad24 mutants show a very similar loss of the G2/M-arrested cells.

While this article was under review, Enomoto et al. (2002) published similar results showing a G2/M arrest in response to short telomeres. They showed that the G2/M arrest was dependent on RAD24, MEC3, DDC2, and MEC1, whereas RAD9, TEL1, and RAD53 did not play a role in this arrest, consistent with our results. One apparent difference with our results is that the deletion of RAD53 did not change the G2/M arrest in the Enomoto et al. experiments, whereas we find strong induction of Rad53p phosphorylation in response to short telomeres. These results can both be explained because the damage signal from Mec1p is transduced to two parallel downstream kinases Rad53p and Chk1p that have additive functions in mediating arrest (Gardner et al., 1999; Sanchez et al., 1999).

Multiple Dysfunctional Telomeres May Not Allow Adaptation

Checkpoint arrest induced by DNA-damage is thought to be a response that allows cells time to repair the damage before proceeding in the cell cycle (Weinert and Hartwell, 1988). One unrepaired double-stranded break created by induction of the site-specific HO-endonuclease in a disomic haploid strain is sufficient to cause G2/M arrest (Sandell and Zakian, 1993). However these cells are not permanently arrested. After 8-12 h, they resume progression through the cell cycle (Sandell and Zakian, 1993). This process was called adaptation because the cells resumed the cell cycle while the DNA damage signal was still present (Sandell and Zakian, 1993; Toczyski et al., 1997; Lee et al., 1998b). Adaptation is accompanied by disappearance of Rad53p phosphorylation (Pellicioli et al., 2001).

The transient arrest followed by adaptation would not be expected to block the growth of a telomerase-null culture. However, a more permanent arrest is found when there are multiple sites of DNA damage. Cells that have two DSBs induced simultaneously show permanent cell cycle arrest (Lee et al., 1998b). We found that, as telomeres shorten in tlc1 cultures, G2/M arrest and an induction of Rad53p phosphorylation occurred. It is not known how many dysfunctional telomeres the cell must sense before arresting; however, because the Rad53p phosphorylation remained high, multiple dysfunctional telomeres may be preventing adaptation. During the initial cell divisions, when most telomeres are still long, the tlc1 cells might adapt to the initial dysfunctional telomere, but with the subsequent burden of additional dysfunctional telomeres, the cells display a more prolonged arrest. This may be similar to the arrest and adaptation seen in cdc13-1 mutants. Growth of these cells at the nonpermissive temperature initially allows adaptation, but arrest occurs again at G2/M in the next cell-cycle (Toczyski et al., 1997).

This response to multiple short telomeres may be similar to the increased radiosensitivity of mTR/ mouse embryonic fibroblasts. When telomeres are long, in early generations, mTR/ cells show radiosensitivity similar to wild-type cells. However in later generations (mTR/ G4-G6) when significant telomere shortening has occurred, there is an increased response to ionizing radiation (Goytisolo et al., 2000; Wong et al., 2000). The increased response mimics the response to higher doses of irradiation, suggesting that the dysfunctional telomeres mimic additional double-stranded DNA breaks.

Repair of Dysfunctional Telomeres

Telomerase-null yeast cells may also respond to telomere dysfunction by attempting to repair the damage. This repair can occur through homologous recombination, which ultimately leads to survivors in S. cerevisiae (Lundblad and Blackburn, 1993) and may also occur through nonhomologous-end-joining (NHEJ). In both cases, the damage signals from the dysfunctional end would be temporarily eliminated.

In the yeast Kluyveromyces lactis, chromosomes with short telomeres undergo higher rates of subtelomeric recombination than those with long telomeres (McEachern and Iyer, 2001; Lundblad, 2002). This suggests that telomere-telomere recombination may be one mechanism to repair loss of telomere function at short telomeres, even before the generation of survivors. In the case of telomerase-null cells, this recombination will have to repair many chromosome ends simultaneously. When many telomeres become short, only those cells that have induced recombination at all telomeres can grow, and these cells likely represent survivors (Lundblad and Blackburn, 1993). Inactivation of the damage recognition pathway in tlc1 rad9 rad24 or tlc1 mec1 sml1 mutants did not affect the rate of survivor generation (Figures 5A and 6A). This suggests that cell cycle arrest is not the limiting step in generation of survivors or that survivors have inactivated the response to DNA-damage checkpoints.

End-to-end chromosome fusion will also eliminate the apparent double-strand break at dysfunctional telomeres and allow cells to overcome checkpoint arrest. This fusion pathway clearly occurs in cells with dysfunctional telomeres (Nakamura et al., 1998; van Steensel et al., 1998; Hackett et al., 2001; Hemann et al., 2001b). However, this mechanism leads to dicentric chromosomes and ultimately to further chromosome breaks and rearrangements (Hackett et al., 2001).

A Function for the DNA Damage Checkpoint in Both Telomere Lengthening and Cell Cycle Arrest

Genes involved in the DNA damage checkpoint are also involved in telomere length maintenance. In S. pombe, double mutants in the two ATM homologues rad3 and tel1 show telomere shortening similar to telomerase mutants and ultimately generate survivors with circular chromosomes (Naito et al., 1998). Similarly, in S. cerevisiae double mutants in both checkpoint genes MEC1 and TEL1 display a senescence phenotype analogous to that seen in a telomerase mutant (Ritchie et al., 1999). Telomerase activity is intact in these mutants; however, access of telomerase to the telomere is impaired (Chan et al., 2001). In wild-type cells telomerase is targeted to the shortest telomeres (Marcand et al., 1999; Hemann et al., 2001b; Hathcock et al., 2002). Because ATM homologues are clearly involved in allowing telomerase access to telomeres, these genes may be involved in the elongation of short telomeres. Thus, even in the absence of telomere dysfunction, the ATM DNA damage pathway may be involved in sensing short telomeres and allowing their elongation.

If the ATM checkpoint is normally activated to signal elongation of short telomeres, how do short dysfunctional telomeres send a signal through this same pathway that results in G2/M arrest? Perhaps a large number of short telomeres sends a strong enough signal that will activate an arrest, as described above for overcoming adaptation to arrest. Alternatively, the persistence of short telomeres after G2/M, when telomerase elongation may occur (Diede and Gottschling, 1999), may send a qualitatively different signal that results in arrest in the next cell cycle. It is also possible that there are specific proteins that are bound at dysfunctional telomeres and that the ATM pathway responds differently, depending on whether these proteins are present at telomeres or not. Understanding how dysfunctional telomeres signal cell cycle arrest will be an exciting challenge for the future.