Telomere uncapping in progenitor cells with criticaltelomere shortening is coupled to S-phaseprogression in vivoSripriya Rajaraman*, Jinkuk Choi*†, Peggie Cheung*, Veronica Beaudry*†, Helen Moore‡, and Steven E. Artandi*†§

*Department of Medicine and †Cancer Biology Program, Stanford University School of Medicine, Stanford, CA 94305; and ‡Genentech, 1 DNA Way,South San Francisco, CA 94080-4990

Edited by Elizabeth Blackburn, University of California, San Francisco, CA, and approved September 14, 2007 (received for review July 11, 2007)

Telomeres protect chromosome ends and serve as a substrate fortelomerase, a reverse transcriptase that adds DNA repeats tothe telomere terminus. In the absence of telomerase, telomeresprogressively shorten, ultimately leading to telomere uncap-ping, a structural change at the telomere that activates DNAdamage responses and leads to ligation of chromosome ends.Telomere uncapping has been implicated in aging and cancer,yet the precise mechanism of uncapping and its relationship tocell cycle remain to be defined. Here, we show that telomeresuncap in an S-phase-dependent manner in gastrointestinal pro-genitors of TERT�/� mice. We develop an in vivo assay thatallows a quantitative kinetic assessment of telomere dysfunc-tion-induced apoptosis and its relationship to cell cycle. Byexploiting the mathematical relationship between rates of gen-eration and clearance of apoptotic cells, we show that 86.2 �8.8% of apoptotic gastrointestinal cells undergo programmedcell death either late in S-phase or in G2. Apoptosis is primarilytriggered via a signaling cascade from newly uncapped telo-meres to the tumor suppressor p53, rather than by chromosomefusion-bridge breakage, because mitotic blockade did not alterthe rate of newly generated apoptotic bodies. These datasupport a model in which rapidly dividing progenitor cells withina tissue with short telomeres are vulnerable to telomere uncap-ping during or shortly after telomere replication.

apoptosis � cell cycle � DNA damage � stem cells � telomerase

Telomeres are nucleoprotein structures that serve a criticalrole in maintaining chromosome stability and cell viability

(1). Telomere length is maintained by telomerase, the enzymethat adds telomere repeats. Chromosome ends in mammalscomprise double-stranded TTAGGG repeats that terminate ina single stranded 3� overhang. This overhang stably invades moreproximal telomere sequences to form a t-loop that is highlybound by the shelterin protein complex (2). In this protectedconformation, telomeres suppress DNA damage responses andprevent recombination at chromosome ends. Disruption oftelomere conformation through telomere shortening or othermeans results in telomere dysfunction, a process also termedtelomere uncapping (3). Telomere uncapping triggers eithercellular senescence or apoptosis; however, the precise changesthat occur at uncapped telomeres and the relationship betweentelomere uncapping and cell cycle are not well understood.

In human cells with insufficient telomerase, telomeres shortenwith each cell division because of the end-replication problem,the inability of DNA polymerase to fully replicate the laggingDNA strand, ultimately leading to replicative senescence as asubset of dysfunctional telomeres recruit DNA damage focicontaining phosphorylated histone-2AX and 53BP1 to un-capped chromosome ends (4, 5). The replicative senescenceresponse depends on p53 and Rb, because inactivation of thesepathways enables cells to bypass senescence and to continueproliferating (6). In this setting of deactivated checkpoints,dysfunctional telomeres are ligated, yielding fused chromosomes

that break during a subsequent anaphase when two independentcentromeres are pulled to opposite spindle poles (7). In contrastto the senescence response seen in primary human cells inculture, critically short telomeres in telomerase-deficient micecommonly induce apoptosis, including male germ cells, gastro-intestinal cells, and lymphocytes (8–11). Senescence in telom-erase knockout tissues has also been seen, particularly in regen-erating liver (12), in lymphomas (13), and in the context of a p53mutant that lacks the ability to induce apoptosis (14).

The relationship between cell cycle and telomere uncappinghas been most extensively studied in the setting of interferencewith telomere-binding protein function (2). Loss of function ofthe telomere-binding protein TRF2 disrupts the protective telo-mere protein complex, leading to rapid telomere uncapping,generation of damage foci at chromosome ends, processing ofthe telomere single-stranded overhang, and covalent ligation togenerate fused chromosomes (5, 15, 16). Impairing TRF2 func-tion can cause telomere uncapping even in quiescent cells in vivoand can lead to chromosomal fusion in both G1 and G2 phasesof cell cycle (16–19). In contrast, the relationship between cellcycle and telomere uncapping in cells whose chromosomesencounter the end-replication problem because of insufficientlevels of telomerase remains less well understood.

To understand the relationship between telomere uncappingand cell cycle, we developed a means for studying the kineticsand cell cycle timing of apoptosis triggered by telomere uncap-ping in gastrointestinal progenitors in mice lacking TERT, thecatalytic component of telomerase. By relating the rate ofgeneration and clearance of apoptotic progenitor cells, wedetermined that apoptosis is tightly coupled to passage throughS-phase, indicating that telomeres uncap in late S-phase or G2 ina stereotyped fashion. These results support a model in whichtelomere uncapping is coupled to telomere replication as a sub-set of critically short telomeres signal a DNA damage response.

ResultsApoptosis and p53 Activation in Proliferating Progenitor Cells inLate-Generation TERT-Deficient Mice. Epithelium of the small in-testine is continuously renewed through the division of stem cellsand progenitor cells located in the intestinal crypt, a niche thatlies at the base of each villus (20, 21). To address the questionof when in the cell cycle critically short telomeres uncap, we

Author contributions: S.R., H.M., and S.E.A. designed research; S.R., J.C., and V.B. performedresearch; P.C. contributed new reagents/analytic tools; S.R., H.M., and S.E.A. analyzed data;and S.R., H.M., and S.E.A. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviation: GI, gastrointestinal.

§To whom correspondence should be addressed. E-mail: sartandi@stanford.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0706485104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0706485104 PNAS � November 6, 2007 � vol. 104 � no. 45 � 17747–17752

GEN

ETIC

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020

studied telomere dysfunction-induced apoptosis in TERTknockout mice. TERT�/� mice were intercrossed to yield first-generation (G1) TERT�/� mice, which were viable and lackedtelomerase activity, but retained functional telomeres. G1TERT�/� mice were bred to yield G2 TERT�/� mice, and thisstrategy was continued until G5 TERT�/� mice were obtained.Cytogenetics and telomere FISH analysis of bone marrow cellsshowed that telomeres remained functional in TERT�/� miceand G1 TERT�/� mice. In contrast, metaphase spreads from G5TERT�/� mice showed a significant increase in chromosomeends lacking detectable telomere signal by FISH (signal-freeends) [supporting information (SI) Fig. 5]. To characterize theresponse to dysfunctional telomeres in TERT�/� mice, we as-sayed apoptosis by TUNEL in gastrointestinal crypts. Apoptoticcells were rare in age-matched TERT�/� and G1 TERT�/�

intestinal crypts, consistent with preservation of telomere func-tion in early-generation telomerase knockout mice (Fig. 1 a andb) (10, 14, 22). In contrast, critical telomere shortening achievedthrough successive generations of intercrossing TERT�/� miceresulted in a profound increase in apoptosis in the crypts of G5

TERT�/� mice (Fig. 1c). Apoptotic cells were seen only in theproliferating progenitor cell compartments, because TUNEL-positive cells were restricted to the population of cells that wereKi-67-positive (Fig. 1 g–i). Telomere shortening induced stabi-lization of p53 protein in crypt progenitor cells of G5 TERT�/�

mice (Fig. 1 d–f ), consistent with activation of a p53 DNAdamage response by telomere uncapping in late-generationTERT�/� cells and similar to results seen in TERC�/� mice (11,23). Consistent with a DNA damage response in a subset of cryptprogenitor cells, strong �-H2AX staining was seen only in G5TERT�/� mice (Fig. 1 j–l). These data indicate that telomereuncapping induces apoptosis in rapidly dividing epithelial pro-genitor cells in vivo, suggesting that continual cycling is impor-tant for telomere uncapping.

Determining the Percentage of Cells That Require S-Phase Progressionfor Telomere Uncapping-Induced Apoptosis. Although short telo-meres show an increased probability of uncapping, this uncap-ping could occur either stochastically in different phases of thecell cycle or in a specific cell cycle phase. For example, progres-sion through S-phase, the period during which telomeres must bereplicated and repackaged, could be important for physicaluncapping of the telomere end and activation of the apoptoticresponse. To determine the percentage of apoptotic cells thatpass through S-phase, G5 TERT�/� mice were pulsed with asingle BrdU injection, killed 4 hours later, and analyzed simul-taneously for BrdU incorporation and TUNEL by using fluo-rescent double immunostaining (Fig. 2 a–d). Interestingly, asignificant percentage of TUNEL-positive cells were also BrdU-positive (Fig. 2d, arrows). To determine whether overlappingTUNEL and BrdU signals originated within the same cell, weanalyzed samples by confocal microscopy. Confocal imagesshowed that the TUNEL signal was distributed in a ring-shapedpattern at the nuclear periphery, a classic pattern in apoptoticcells visualized by confocal imaging (24). Importantly, BrdUsignal was clearly seen within the interior of the TUNEL signalin a subset of apoptotic cells (Fig. 2 h and i and SI Movie 1). Thus,many cells undergoing apoptosis recently incorporated BrdUbefore initiation of the apoptotic program.

Of the TUNEL-positive cells, 36.7% showed strong BrdUstaining, indicating that they passed through S-phase and be-came apoptotic during this 4-hour period (Table 1). However,this is only a lower limit on the percentage of cells that becomeapoptotic after S-phase, because many of the BrdU-negativeapoptotic cells were generated before the labeling period and arecleared gradually from the tissue over time. We reasoned that wecould correct for the persistence of these slowly cleared apo-ptotic cells and accurately determine the percentage of newlygenerated apoptotic cells that passed through S-phase by con-structing a mathematical model.

In constructing the model, we assume that apoptosis in thetissue is in equilibrium, such that the number of apoptotic cellsgenerated during a specific time interval equals the number ofapoptotic cells cleared during that interval. This is a reasonableassumption because the intrinsic rate of telomere uncapping willnot change over this short period. Under equilibrium conditions,we can calculate the rate of apoptosis from the half-life forclearance of apoptotic fragments, a value that we can measureempirically (see below). We wish to determine what fraction ofthe apoptotic cells generated during a short time interval passedthrough S-phase. We therefore define the following two popu-lations of apoptotic cells:

A � A(t) � the total number of apoptotic cells per 100 cryptsat time t.

S � S(t) � the number of cells within the apoptotic populationA(t) that entered S-phase after time 0 and before time t.

γ γγ −−

TERT +/+ G5 TERT -/-G1 TERT -/-

TERT +/+ G1 TERT -/- G5 TERT -/-

TERT +/+ G1 TERT -/- G5 TERT -/-

TERT +/+ G1 TERT -/- G5 TERT -/-

a b c

d e f

g ih

kj l

TUNEL

p53

Ki67

γ−−H2AX

Fig. 1. Increased apoptosis, p53 expression, and �-H2AX foci in proliferatingcrypt progenitors in late-generation TERT�/� mice. (a–c) Representative �40images from TUNEL-stained GI crypts. Arrows indicate apoptotic cells. (d–f )p53 immunostaining shows increased expression (arrows) in G5 TERT�/�

crypts. (g–i) Ki67 staining reveals the proliferating progenitor cells in all threegenotypes. (j–l) �-H2AX foci formation is seen only in G5 TERT�/� crypts(arrows) and not in controls.

17748 � www.pnas.org�cgi�doi�10.1073�pnas.0706485104 Rajaraman et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020

S � A, because S represents a subset of the cells counted inA. We then create a model based on the rates of change for thesepopulations over time, dA/dt and dS/dt:

dAdt

� p � cA, [1]

dSdt

� kp � cS. [2]

Eq. 1 describes how A changes over time, with cells produced atrate p and lost at rate cA, where c is the clearance rate constantfor apoptotic cells. Eq. 2 describes the rate of change inpopulation S over time, with cells produced at rate kp and lostat rate cS. The constants p and c are the same as in Eq. 1. Theconstant k represents the fraction of apoptotic cells that weregenerated during, or shortly after, S phase. We wish to solve fork, and, by definition, 0 � k � 1. For example, if k � 1, then allepithelial progenitors initiate apoptosis after S-phase entry.Alternatively, if k � 0.5, then one-half of epithelial cells enterapoptosis after S-phase and the remaining half undergo apo-ptosis in other cell cycle phases. Therefore, determining thevalue of k will tell us the percentage of gastrointestinal progen-itor cells that undergo telomere dysfunction-induced apoptosisafter S-phase entry.

We use the differential equations in Eqs. 1 and 2 together withthe experimental data above to obtain an estimate of k. Eq. 2 isa separable differential equation, and its solution is:

S � S�t� � kpc�1 � e�ct� [3]

(see SI Text for details of this calculation). Our experimentaldata (Table 1) show that 36.7% of apoptotic cells were BrdU-positive at t � 4 h, which is a direct measure of S(t) and can berepresented as S(4) � 0.367 � A. Incorporating these empiricdata, we thus solve Eq. 3 for k, and substitute t � 4, and S � 0.367� A, to obtain

k �SA

1�1 � e�ct�

,

k �0.367 � A

A1

�1 � e�4c�,

k � 0.3671

�1 � e�4c�. [4]

Eq. 4 allows the calculation of k, providing that we know c, theclearance rate constant for apoptotic bodies; and the value of cis related to half-life t1/2 as c � (ln 2)/t1/2. The half-life ofapoptotic bodies in small intestinal crypts of wild-type micetreated with DNA-damaging agents has been empirically mea-sured and ranges from t1/2 � 3.5 h, for apoptosis caused byhydroxyurea treatment, to t1/2 � 11.8 h, for apoptosis caused byionizing radiation (25). Because this range is too large for anaccurate measurement of k in our study, we designed an exper-

TUNEL

BrDU

MERGE

Single BrdUinjection

(-/+ colcemid)

4 hrs BrdU

TUNEL

a

Newly generated apoptotic cellsthat passed through S phase during pulse

Total number of apoptotic cells in steady-state

Cells that entered S-phase during pulse

b

c

d

TUNEL

BrDU

MERGE

e

f

g

h i

(-) Colcemid (+) Colcemid

TUNELBrDU

TUNELBrDU

Fig. 2. Apoptotic crypt progenitor cells readily incorporate BrdU during ashort labeling period. (a) G5 TERT�/� mice were injected with BrdU and killedat 4 h, and small intestine crypts were analyzed by immunohistochemistry forapoptosis by TUNEL and for BrdU incorporation. (b) TUNEL. (c) BrdU. (d) Mergeshows double-positive apoptotic cells (arrows). BrdU incorporation in apo-ptotic cells was unchanged with mitotic blockade with colcemid. (e) TUNEL. ( f)BrdU. (g) Merge shows double-positive apoptotic cells (arrows). Representa-tive 40� images, n � 3 mice for each experiment. (h) High-resolution confocalimage of TUNEL and BrdU double-labeled G5 TERT�/� crypts showing com-plete overlap between apoptotic and BrDU-positive cells (white arrows). (i)Higher magnification of a small region seen in h.

Table 1. Quantitation of BrDU� and TUNEL� cells in GI crypts

(�) Colcemid (�) Colcemid

Genotype

No. ofTUNEL�

nuclei

No. ofTUNEL�

and BrDU�

Doublepositive, % Genotype

No. ofTUNEL�

nuclei

No. ofTUNEL�

and BrDU�

Doublepositive, %

G5 Tert�/� 131 51 38.9 G5 Tert�/� 177 76 42.9G5 Tert�/� 122 44 36.0 G5 Tert�/� 179 72 40.2G5 Tert�/� 136 48 35.2 G5 Tert�/� 178 85 47.7Total 389 143 36.7 1.95 Total 534 233 43.6 3.81

Rajaraman et al. PNAS � November 6, 2007 � vol. 104 � no. 45 � 17749

GEN

ETIC

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020

iment to measure the half-life of apoptotic fragments inTERT�/� crypts.

Measurement of t1/2 for Apoptotic Cells in Late-Generation TERT�/�

Crypts. To measure t1/2 for apoptotic cells in TERT�/� mice invivo, groups of three G5 TERT�/� mice were injected with BrdUand then killed for analysis at intervals of 4, 6, 8, 10, and 12 h.For each sample, the percentage of double BrdU� TUNEL�

cells per total number of TUNEL� cells was scored in gastro-intestinal crypts by immunohistochemistry (Fig. 3a). The per-centage of BrdU� TUNEL� cells per total number of TUNEL�

cells was 37.1%, 34.0%, 27.0%, 16.6%, and 6.7% at 4, 6, 8, 10,and 12 h, respectively (SI Table 2). From these data, wecalculated t1/2 using a standard one-phase exponential decaymodel, yielding a value of t1/2 � 5.00 0.68 h (c � 0.139 0.09;r2 � 0.896) (Fig. 3b). Interestingly, this value for t1/2 falls betweenthe previously published values for t1/2 in gastrointestinal crypts

and closer to the value for t1/2 in mice treated with hydroxy-urea (25).

Apoptosis Induced by Telomere Uncapping Is Coupled to S-Phase inProgenitor Cells. Having measured the clearance rate c, we cannow represent graphically how S(t), the fraction of apoptoticcells whose telomeres uncap during or shortly after S-phase,accumulate over time as a fraction of all apoptotic cells in thepopulation, A (Fig. 4a). Furthermore, with the measured valueof c � 0.139, we can use Eq. 4 to calculate k, which yields k �86.2 8.8%. These data indicate that 86.2% of gastrointestinalcrypt cells passed through S-phase shortly before the initiationof the apoptotic program. Importantly, this value for k markedlyexceeds the labeling index of 23%, the percentage of total cellsin small-intestine crypts that are labeled with short pulses ofBrdU or tritiated thymidine (26, 27). Thus, the high value for kdoes not simply reflect the fact that many progenitor cells areactively cycling. If apoptosis were initiated in a cell cycle-

6hr 8hr

12hr 10hr

a

b

Fig. 3. Measurement of half-life of apoptotic progenitor cells in crypts of G5TERT�/� mice. (a) Late-generation mice were injected with BrdU, followed bya chase period of increasing length. The abundance of BrdU� TUNEL� cells wasmeasured as a percentage of total TUNEL� cells after chase periods of 4, 6, 8,10, and 12 h. Apoptosis was measured by TUNEL (green) and BrdU incorpo-ration by immunofluorescence (red). Double-positive cells (yellow) diminishedover time as the newly generated apoptotic cells were cleared; n � 3 for eachtime point. (b) Data from a were used to plot the ratio of BrdU� TUNEL� cellsper total number of TUNEL� cells remaining at each time point. A one-phaseexponential decay model was fit to these data and was used to calculate theclearance rate constant (c � 0.139 0.09 hr�1) and the half-life of apoptoticcells in G5 TERT�/� crypts (t1/2 � 5.00 0.68 h).

Telomere uncapping

DamagePathway

Apoptosis(mouse GI progenitors)

Senescence in asubsequent G1 phase

(human fibroblasts)

End-to-endchromosome fusion

Chromosomerupture in mitosis

FusionPathway

Apoptosis

a

b

Fig. 4. Telomere uncapping is linked to S-phase in progenitor cells oflate-generation TERT�/� mice. (a) Graph depicting the dependence of telo-mere uncapping-induced apoptosis on S-phase. The number of cells thatbecome apoptotic during or shortly after S-phase as a percentage of the totalnumber of apoptotic cells, S(t)/A, is graphed versus time, based on Eq. 3 andby using our empirically derived values for k and c. Beginning at t � 0, themodel depicts the gradual accumulation of cells becoming apoptotic as aresult of telomere uncapping in S-phase. At t � 4 h, S(t) � 36.7% (point ongraph), representing our empiric measurement using BrdU incorporation. Thegraph asymptotically approaches 86.2%, our calculated value for k, whichrepresents the percentage of apoptotic cells that die in S-phase or G2 (dashedred line). (b) A model for telomere uncapping in the cell cycle. Telomeresuncap in late S-phase or in G2, resulting in either a DNA damage response thatsignals checkpoints (solid line) or a repair-mediated joining process leading tochromosomal fusion (dashed lines). The damage pathway results in apoptosisin crypt progenitor cells. In human fibroblasts, telomere uncapping may besimilar in its cell cycle timing but induces senescence in a subsequent G1 phase.

17750 � www.pnas.org�cgi�doi�10.1073�pnas.0706485104 Rajaraman et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020

independent manner, we would anticipate a value for k muchcloser to the labeling index of 23%.

Eq. 4 can also be used to predict t1/2 for any value of k. If weassume that all cells must pass through S-phase before under-going apoptosis (k � 1), we calculate, using our empiric data, thatt1/2 for clearance of apoptotic cells in TERT�/� mice equals6.06 0.236 h, a value very close to our measured t1/2 of 5.00 h(see SI Text for details of these calculations) (25). Takentogether, these results show that S-phase progression is a pre-requisite for telomere uncapping-induced apoptosis for the vastmajority, and perhaps all, gastrointestinal progenitor cells inlate-generation TERT�/� mice.

Apoptosis Induced by Uncapped Telomeres Does Not Require Mitosis.Telomere uncapping can induce cellular responses by directactivation of DNA damage response pathways. Alternatively,uncapped telomeres can fuse to yield dicentric chromosomeswhose breakage during mitosis may also activate DNA damageresponses to induce apoptosis. To understand whether gastro-intestinal progenitor cells initiate apoptosis before or aftermitosis, we performed our BrdU and TUNEL analysis on micetreated with mitotic inhibitor colcemid. We used a colcemidprotocol previously shown to block all gastrointestinal (GI)progenitor cells in vivo in early mitosis (28, 29). Consistent withthese data, we found that the prevalence of mitotic figures wasincreased 2-fold in mice treated with colcemid for 4 hours,compared with untreated controls (data not shown). Analysis ofG5 TERT�/� mice treated with colcemid and injected with BrdUrevealed that 43.6% of apoptotic cells were BrdU-positive (Table1 and Fig. 2 e–g), results that were in close agreement with thosederived from G5 TERT�/� mice that did not receive colcemid.These findings show that mitotic blockade had little effect onapoptosis, consistent with previous results in telomerase knock-out lymphocytes cultured in vitro (30). Using this value of 43.6%for the percentage of apoptotic cells that were BrdU-positive ina setting of mitotic inhibition, we can reformulate S as S(4) �0.436 � A and rederive Eq. 4 as above. Solving Eq. 4 with c �0.139, we find that, for mice treated with colcemid, k � 1.02.These data indicate that, in the presence of colcemid, 100% ofgastrointestinal progenitor cells undergo telomere uncapping-induced apoptosis shortly after S-phase but before mitosis. Weconclude that telomere uncapping in telomerase-deficient epi-thelial cells occurs in late S-phase or in G2, a period consistentwith the timing of telomere replication. The inability to repack-age these telomere ends in a protected conformation inducesapoptosis directly, without requiring progression through mitosisand is therefore independent of a fusion-bridge breakage mech-anism (Fig. 4b).

DiscussionTelomere Uncapping Is a Stereotyped Process Linked to Cell CycleProgression. The probability that a telomere can switch from aprotected, capped state to a deprotected, uncapped one in-creases as the telomeric repeats become very short, ultimatelyactivating a DNA damage response causing cell cycle arrest orapoptosis (3). These responses can contribute to certain aspectsof aging, can dramatically impair stem cell renewal, and candestabilize chromosomes, promoting tumorigenesis. To under-stand the relationship between uncapping and cell cycle, weanalyzed epithelial progenitor cells in TERT�/� mice in vivo,which, instead of senescence, respond to telomere uncappingthrough apoptosis. By labeling S-phase cells with a brief pulse ofBrdU, we tracked the fate of these cells that passed throughS-phase and found that many apoptotic cells (37%) had incor-porated BrdU, even during this short period. We reasoned thatthe true percentage of cells that initiate apoptosis during orshortly after S-phase was substantially higher than 37% becausemost of the BrdU-negative apoptotic cells were generated before

the BrdU pulse and are cleared from the tissue gradually overtime.

To correct for the persistence of apoptotic cells in the tissue,we developed a mathematical model based on the fact thatapoptosis in the small intestine of TERT�/� mice is in equilib-rium, indicating that the production rate of apoptotic cells mustequal the clearance rate. We empirically measured the half-lifeand clearance rate of apoptotic cells in TERT�/� mice, allowingus to finalize the model and determine the relationship betweenS-phase and apoptotic entry. This approach allowed us to definea population of cells that undergo apoptosis during S-phase andrevealed the kinetics with which these S-phase cells becomeapoptotic (Fig. 4a). Using our empiric data and this mathemat-ical model, we calculated the value of the variable k, representingthe fraction of apoptotic cells undergoing apoptosis during orshortly after S-phase to be 86 8.8%. Together, these data showthat apoptosis triggered by telomere uncapping is a stereotyped,cell cycle-dependent process, rather than a stochastic one re-sulting in uncapping in different cell cycle phases.

A Model for Synchronized Uncapping with Replication of CriticallyShort Telomeres. We propose that critically short telomeres areespecially vulnerable to uncapping during telomere replicationfor two reasons. First, replication of both the leading and laggingstrands requires that the three-dimensional structure of thetelomere is temporarily disrupted by the passage of replicationforks. Second, loss of telomere sequences because of the end-replication problem occurs precisely during this phase of telo-mere replication at the lagging strand. In addition, telomere endsare likely subject to exonucleolytic attack, which may furtherexacerbate the instability of critically short telomeres (31).Unlike chromosome ends with sufficient reserves of telomererepeats, a critically short telomere may not support the forma-tion of a stable conformation after replication. Uncapping oftelomeres in this manner would activate the DNA damagemachinery and a corresponding cellular response, consistentwith the stereotypic apoptotic response seen in mouse gastro-intestinal tract.

These data fit with a large body of evidence linking telomerereplication to late S phase or G2 in diverse species (32–34). Inaddition, components of Mre11-Rad50-Nbs1, a DNA repaircomplex, were shown to transiently associate with telomeresduring S phase in both yeast and human cells, suggesting a rolefor these proteins in telomere replication or repair (35, 36).Recently, telomeres were shown to elicit a transient DNAdamage response during the G2 phase of the cell cycle. Duringthis stage of replication, telomeres recruit �-H2AX, 53BP1, andphospho-ATM, but these damage complexes rapidly dissipatefrom telomeres, presumably as telomeres become properly pack-aged (37). Perhaps, these damage foci persist at an uncappedtelomere, initiating a full-f ledged DNA damage response. Theensuing apoptosis would occur in a synchronized fashion duringor shortly after S-phase, consistent with our findings.

Telomere uncapping has been extensively studied by interfer-ing with TRF2 function, which, in TRF2 conditional knockoutmice, causes rapid uncapping in quiescent hepatocytes (19). Inaddition, induction of apoptosis by dominant-negative TRF2 inproliferating human T cells did not require S-phase entry (38).Thus, telomere uncapping per se does not require S-phase whentelomere structure is perturbed by removal of critical telomere-binding proteins. The contrast between these results and ours intelomerase knockout mice likely reflects differences in thenature of the stimulus that causes uncapping. Our data suggestthat telomere uncapping because of insufficient telomeraseoccurs specifically in S-phase, as a subset of the shortest telo-meres can no longer be maintained in a protected conformation.

Rajaraman et al. PNAS � November 6, 2007 � vol. 104 � no. 45 � 17751

GEN

ETIC

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020

Implications for Telomere Uncapping at Replicative Senescence.These data suggest that the timing of telomere uncapping in cellsundergoing apoptosis in telomerase knockout mice and humanfibroblasts at replicative senescence in human fibroblasts may besimilar. Telomere uncapping in both cases may occur in anS-phase-dependent fashion, but the principal difference may bein the cellular response. Whereas mouse gastrointestinal pro-genitors possess a potent apoptotic response to telomere uncap-ping and other DNA damage signals, fibroblasts lack such anapoptotic response, instead undergoing cell cycle arrest whenconfronted with DNA damage. In the absence of a strongapoptotic response, telomere uncapping in late S-phase inhuman fibroblasts may go incompletely recognized until the nextG1 phase, where activation of p53 and Rb pathways enforce cellcycle arrest and senescence (Fig. 4b). Alternatively, telomereuncapping in S-phase could activate an S-phase checkpoint,adaptation to which allows eventual progression to G1, where theterminal senescence program is revealed (39, 40). Together,these findings provide insight into the mechanism of telomeredysfunction, linking the uncapping response to S-phase and totelomere replication.

Materials and MethodsMice. TERT knockout mice were generated by gene targeting inembryonic stem cells as described (S.R. and S.E.A., unpublishedwork). All mice were maintained on a mixed genetic background,50:50 C57BL/6:129S4. TERT�/� mice were intercrossed to yieldG1 TERT�/� mice, which, in turn, were intercrossed to create G2TERT�/� mice and so on, until G5 TERT�/� mice were derived.Cousin mating schemes were used to prevent the generation ofsubstrains (10). All mice were treated in accordance withAssociation for Assessment and Accreditation of LaboratoryAnimal Care-approved guidelines at Stanford University.

Kinetic Analyses Using BrdU Incorporation. BrdU (0.1 mg/g of bodyweight) was injected i.p. To assess rates of BrdU injection duringapoptosis, mice were killed 4 h after injection. Three mice wereanalyzed per genotype, and GI tracts were cleaned and fixedimmediately after sacrifice. To block mitosis, mice were injectedwith colcemid (1 �g/g of body weight) and BrdU simultaneously

and then analyzed at 4 h. For half-life measurements, G5TERT�/� mice in groups of three were injected with BrdU andchased for a total of 4, 6, 8, 10, and 12 h and then analyzed asabove.

Immunohistochemistry of GI Crypts. The entire small intestine wasopened longitudinally and rolled into a Swiss roll with theluminal side facing outward and then fixed in 10% formalin.TUNEL was performed on paraffin-embedded 5-�m tissuesections by using an Apoptag kit according to the manufacturer’sinstructions (Chemicon, Temecula, CA). TUNEL/BrdU doubleimmunostaining was performed sequentially on the same tissuesection. Each section was first processed for TUNEL by using afluorescein-conjugated secondary antibody (Chemicon), fol-lowed by incubation with a rat monoclonal primary antibodyagainst BrdU (Abcam, Cambridge, MA) and anti-rat cy-3 sec-ondary antibody. The following antibodies were used for immu-nohistochemistry: p53, CM5 rabbit polyclonal antibody (Novo-castra, Newcastle-upon-Tyne, U.K.); Ki-67, mouse anti-humanmonoclonal antibody (BD Pharmingen, Franklin Lakes, NJ);and �H2AX, mouse monoclonal antibody (Upstate Biotechnol-ogy, Lake Placid, NY).

Mathematical Calculations and Graphs. One-phase exponential de-cay was fit to data from SI Table 2 to generate the apoptotic cellhalf-life analysis in Fig. 3b. Eq. 3 was used together with ourempiric data to generate the model to describe the relationshipbetween S-phase and apoptosis in Fig. 4a. Both graphs weregenerated by using Mathematica 5.2 (Wolfram Research, Cham-paign, IL).

We acknowledge expert technical assistance from the Stanford Trans-genic Core Facility, Pauline Chu in the Stanford Comparative MedicineHistology Research Core Laboratory, Kitty Lee at the Stanford CellScience Imaging Facility, and helpful comments and discussions from L.Attardi, J. Sage, and members of the S.E.A. laboratory. S.R. wassupported by a Susan G. Komen Postdoctoral Fellowship, and J.C. wassupported by a graduate student fellowship from the Samsung Lee KunHee Scholarship Foundation. This work was supported by NationalInstitutes of Health (NIH)/National Cancer Institute Grant R01CA111691 and by a Pilot Project Award from the Stanford DigestiveDisease Center (NIH P30 DK56339).

1. Blackburn EH (2001) Cell 106:661–673.2. de Lange T (2005) Genes Dev 19:2100–2110.3. Blackburn EH (2000) Nature 408:53–56.4. d’Adda di Fagagna F, Reaper PM, Clay-Farrace L, Fiegler H, Carr P, Von

Zglinicki T, Saretzki G, Carter NP, Jackson SP (2003) Nature 426:194–198.5. Takai H, Smogorzewska A, de Lange T (2003) Curr Biol 13:1549–1556.6. Shay JW, Pereira-Smith OM, Wright WE (1991) Exp Cell Res 196:33–39.7. Counter CM, Avilion AA, LeFeuvre CE, Stewart NG, Greider CW, Harley CB,

Bacchetti S (1992) EMBO J 11:1921–1929.8. Hemann MT, Rudolph KL, Strong MA, DePinho RA, Chin L, Greider CW

(2001) Mol Biol Cell 12:2023–2030.9. Blasco MA, Lee HW, Hande MP, Samper E, Lansdorp PM, DePinho RA,

Greider CW (1997) Cell 91:25–34.10. Lee HW, Blasco MA, Gottlieb GJ, Horner JW, Greider CW, DePinho RA

(1998) Nature 392:569–574.11. Wong KK, Maser RS, Bachoo RM, Menon J, Carrasco DR, Gu Y, Alt FW,

DePinho RA (2003) Nature 421:643–648.12. Satyanarayana A, Wiemann SU, Buer J, Lauber J, Dittmar KE, Wustefeld T,

Blasco MA, Manns MP, Rudolph KL (2003) EMBO J 22:4003–4013.13. Feldser DM, Greider CW (2007) Cancer Cell 11:461–469.14. Cosme-Blanco W, Shen MF, Lazar AJ, Pathak S, Lozano G, Multani AS,

Chang S (2007) EMBO Rep 8:497–503.15. van Steensel B, Smogorzewska A, de Lange T (1998) Cell 92:401–413.16. Celli GB, de Lange T (2005) Nat Cell Biol 7:712–718.17. Smogorzewska A, Karlseder J, Holtgreve-Grez H, Jauch A, de Lange T (2002)

Curr Biol 12:1635–1644.18. Bailey SM, Cornforth MN, Kurimasa A, Chen DJ, Goodwin EH (2001) Science

293:2462–2465.

19. Lazzerini Denchi E, Celli G, de Lange T (2006) Genes Dev 20:2648–2653.20. Radtke F, Clevers H (2005) Science 307:1904–1909.21. Marshman E, Booth C, Potten CS (2002) BioEssays 24:91–98.22. Choudhury AR, Ju Z, Djojosubroto MW, Schienke A, Lechel A, Schaetzlein

S, Jiang H, Stepczynska A, Wang C, Buer J, et al. (2007) Nat Genet 39:99–105.23. Chin L, Artandi SE, Shen Q, Tam A, Lee SL, Gottlieb GJ, Greider CW,

DePinho RA (1999) Cell 97:527–538.24. Monaghan P, Watson PR, Cook H, Scott L, Wallis TS, Robertson D (2001) J

Microsc 203:223–226.25. Potten CS (1996) Br J Cancer 74:1743–1748.26. Kellett M, Potten CS, Rew DA (1992) Epithelial Cell Biol 1:147–155.27. Chwalinski S, Potten CS, Evans G (1988) Cell Tissue Kinet 21:317–329.28. Nome O (1975) Pathol Eur 10:221–231.29. Cheng H, Bjerknes M (1983) Anat Rec 207:427–434.30. Hao LY, Strong MA, Greider CW (2004) J Biol Chem 279:45148–45154.31. Zhu XD, Niedernhofer L, Kuster B, Mann M, Hoeijmakers JH, de Lange T

(2003) Mol Cell 12:1489–1498.32. Ferguson BM, Brewer BJ, Reynolds AE, Fangman WL (1991) Cell 65:507–515.33. Bianchi A, Shore D (2007) Cell 128:1051–1062.34. Wright WE, Tesmer VM, Liao ML, Shay JW (1999) Exp Cell Res 251:492–499.35. Zhu XD, Kuster B, Mann M, Petrini JH, de Lange T (2000) Nat Genet

25:347–352.36. Takata H, Tanaka Y, Matsuura A (2005) Mol Cell 17:573–583.37. Verdun RE, Crabbe L, Haggblom C, Karlseder J (2005) Mol Cell 20:551–561.38. Karlseder J, Broccoli D, Dai Y, Hardy S, de Lange T (1999) Science

283:1321–1325.39. Toczyski DP, Galgoczy DJ, Hartwell LH (1997) Cell 90:1097–1106.40. Galgoczy DJ, Toczyski DP (2001) Mol Cell Biol 21:1710–1718.

17752 � www.pnas.org�cgi�doi�10.1073�pnas.0706485104 Rajaraman et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

2, 2

020