CDK1 differentially regulates G-overhang generation at leading- and lagging-strand telomeres in telomerase-negative cells in G₂ phase

Abstract

Human telomeres contain single-stranded 3' G-overhangs that function in telomere end protection and telomerase action. Previously we have demonstrated that multiple steps involving C-strand end resection, telomerase elongation and C-strand fill-in contribute to G-overhang generation in telomerase-positive cancer cells. However, how G-overhangs are generated in telomerase-negative human somatic cells is unknown. Here, we report that C-strand fill-in is present at lagging-strand telomeres in telomerase-negative human cells but not at leading-strand telomeres, suggesting that C-strand fill-in is independent of telomerase extension of G-strand. We further show that while cyclin-dependent kinase 1 (CDK1) positively regulates C-strand fill-in, CDK1 unlikely regulates G-overhang generation at leading-strand telomeres. In addition, DNA polymerase α (Polα) association with telomeres is not altered upon CDK1 inhibition, suggesting that CDK1 does not control the loading of Polα to telomeres during fill-in. In summary, our results reveal that G-overhang generation at leading- and lagging-strand telomeres are regulated by distinct mechanisms in human cells.

Introduction

Telomeres protect genome stability and integrity by preventing chromosome ends from inappropriate degradation, fusions and recombination.¹ Human telomere consists of many kilobases of (TTAGGG)_n tandem repeats and terminates with a single-stranded 3′-overhang at the G-rich strand termed G-overhang. The importance of G-overhang is multifaceted. First, G-overhang serves as the binding site for telomerase.^2,3 Lack of G-overhang prevents telomerase binding to telomeres. Second, G-overhang is essential for the formation of the unique telomere nucleoprotein structure termed t-loop, which “caps” the chromosome end and protects the chromosome end from degradation and inappropriate DNA repair activities.⁴ Failure of maintaining the t-loop structure leads to chromosome end-to-end fusions.⁵ In addition, the mechanism for G-overhang generation contributes to telomere shortening. It has been shown that the length of G-overhang is influenced by a number telomere binding factors,^6-13 suggesting that telomere binding factors regulate G-overhang generation. Despite the importance of G-overhang in telomere maintenance, the mechanism underlying G-overhang generation is largely unclear.

During DNA replication, the TTAGGG (G-rich) strand serves as the template for discontinuous lagging-strand synthesis, while the CCCTAA (C-rich) strand serves as the template for continuous leading-strand synthesis. At least four molecular events are thought to contribute to G-overhang generation. First, the removal of the final RNA primer and/or the failure to position the final RNA primer at the very end of the chromosome immediately after replication creates an intermediate overhang at the lagging-strand telomere. Second, the 5′ end of C-rich strand of the initially blunt-ended leading-strand daughter telomere is resected by nucleases/helicases such as mammalian Apollo, ExoI¹⁴ as well as undefined nucleases, generating a G-overhang that is essential for telomerase binding. It is currently unknown whether end resection also takes place at the lagging-strand telomere, although a recent mammalian study suggests so.¹⁴ Next, telomerase extends G-rich strands of both leading- and lagging-strand telomeres, further lengthening G-overhang.¹⁵ Finally, at least in telomerase-positive cells, a distinct DNA synthesis step termed C-strand fill-in synthesizes more C-rich repeats in the late S/G₂ phase after the synthesis of genome DNA completes, shortening G-overhang length.^7,15,16

The C-strand fill-in step is of significant importance to telomerase-expressing cells, because it replenishes C-strand length after telomerase elongates G-strand.¹⁶ C-strand fill-in is also present at yeast and mouse telomeres.^17-20 Interestingly, C-strand fill-in in yeast cells is independent of telomerase expression,^18,21-23 suggesting that telomerase extension of telomeres is not a prerequisite for fill-in. Another possible significance of C-strand fill-in is that, since the final Okazaki fragment may not be placed at the very end of lagging-strand telomere,²⁴ it will leave the daughter strand significantly shorter and telomere will undergo rapid shortening without C-strand fill-in.

G-overhang length at human telomeres undergoes cell cycle-regulated dynamics, with G-overhang length transiently increasing during S phase and then reducing during late S/G₂ phase in telomerase-positive cells.⁷ Inhibition of the activity of CDK1 results in the retention of long G-overhangs at global telomeres in the G₂/M phase, suggesting that CDK1 may regulate G-overhang generation.⁷ CDKs are a series of kinases controlling the precise progression of the cell cycle. CDK consists of a catalytic kinase subunit and a positive regulatory subunit (cyclin).²⁵ It is generally thought that CDK4/6 with D cyclins control G₁ phase progression; CDK2/cyclins E and A control S phase progression; and CDK1/cyclins A and B trigger G₂/M progression.²⁶ In human cells, CDK1 regulates the resolution of sister telomeres through phosphorylating TRF1.²⁷ In budding yeast, which contains only one CDK, cdc28, Cdc28 is required for the generation of G-overhangs by controlling end resection^28-30 as well as by regulating the recruitment of telomerase to telomeres.^31,32 The role of CDK1 in regulating telomere maintenance in human cells is largely elusive.

In the G₂ phase of the cell cycle, a variety of proteins, including the ones involved in general DNA replication, recombination and repair, become associated with telomere DNA,^33,34 suggesting that telomeres are under active reconfiguration, and that critical events take place at telomeres in the G₂ phase. To better understand the mechanism regulating this reconfiguration, we performed a comprehensive analysis of G-overhangs at leading- and lagging-strand telomeres in G₂ phase. We report here that C-strand fill-in is present at lagging-strand telomeres in telomerase-negative human cells, suggesting that C-strand fill-in occurs independent of telomerase. In contrast, C-strand fill-in is absent at leading-strand telomeres in both telomerase-negative and -positive cells. Chromatin immunoprecipitation (ChIP) assay reveals that CDK1 has no effect on the association of DNA Polα to telomeres at G₂ phase, indicating that CDK1 may regulate Polα activity during C-strand fill-in rather than Polα recruitment to telomeres. In addition, we also found that while CDK1 positively regulates C-strand fill-in, it unlikely controls G-overhang generation at leading telomeres. Our results provide insights into the molecular mechanism underlying the regulation of G-overhang generation at human telomeres.

Results

Global G-overhangs are lengthened during S phase and shortened in G₂ phase in telomerase-negative human cells

In a previous study we analyzed G-overhang dynamics in telomerase-negative human somatic cell line IMR90, and found that G-overhangs were lengthened in S phase and then shortened in G₂ phase.⁷ This was a surprising finding, because C-strand fill-in was viewed as a means for replenishing C-strand length after telomerase extension of G-strand. To validate the G₂ phase-specific G-overhang shortening in the absence of telomerase, we tested a different telomerase-negative human somatic cell line BJ/E6/E7³⁵ and analyzed G-overhang dynamics during the cell cycle in BJ/E6/E7. Cells was synchronized at G₁/S boundary using serum starvation/aphidicolin arrest, released into drug-free media and collected at indicated time during the cell cycle (Fig. 1). FACS analysis shows that nearly 80% cells were synchronized (Fig. 1A). Genome DNA was isolated, and the overhang protection assay³⁵ was used to measure G-overhang length. The length of G-overhangs in BJ/E6/E7 increased during S phase, reached the longest in late S/G₂ phase (6 h to 8 h) and then shortened in G₂ phase (Fig. 1B and C), largely similar to the G-overhang dynamics in IMR90.⁷ Thus, our results establish that C-strand fill-in occurs independent of telomerase, arguing against the concept that C-strand fill-in is exclusively for replenishing C-strand length after telomerase extends G-strand.

Figure 1. G-overhang length dynamics of telomerase-negative BJ/E6/E7 cells during the cell cycle. (A) FACS analysis of DNA content in synchronized BJ/E6/E7 cells (PD39). Cells were collected from 0 h to 12 h after release from G₁/S boundary. (B) G-overhang length measurement of genomic DNA using the overhang protection assay. (C) Quantitation of mean overhang lengths. Two-tailed t-test was used to calculate statistical significance. Results were from two independent experiments. Error bars: s.e.m.

C-strand fill-in occurs at lagging-strand telomeres in telomerase-negative cells

To determine whether G-overhangs are equally shortened in late S/G₂ phase at leading- and lagging-strand telomeres in telomerase-negative cells, we measured G-overhangs dynamics of separated leading and lagging telomeres from late S to the completion of G₂/M phase. BJ/E6/E7 cells were synchronized at G₁/S boundary and then released into S phase. BrdU was added into media following release, and cells were collected from late S phase to the next G₁ phase (Fig. 2B). Leading- and lagging-strand telomeres were separated and purified by CsCl ultracentrifugation (Fig. 2A; Fig. S1).^7,36 Non-denaturing in-gel hybridization was used to determine the G-overhang abundance. We found that in BJ/E6/E7, G-overhangs of lagging-strand telomeres shortened during G₂ phase (Fig. 2C and D, DMSO-treated samples), suggesting that C-strand fill-in takes place at lagging telomeres in telomerase-negative cells.

Figure 2. G-overhang dynamics at lagging telomeres in BJ/E6/E7 with or without CDK1 inhibition. (A) Scheme of separation of leading and lagging daughter telomeres.³⁶ (B) FACS analysis of DNA content in synchronized BJ/E6/E7 cells (PD37). DMSO and 8 μM PA was added at 6 hr after release. Cells were collected at 7 h, 9.5 h, 12 h and 14.5 h (next G₁) after release. (C) G-overhang abundance at lagging-strand telomeres measured by non-denaturing in-gel hybridization assay. (D) Quantitation of G-overhang abundance. Two-tailed t-test was used to calculate statistical significance. Results were from at least three independent experiments. Error bars: s.e.m.

CDK1 positively regulates C-strand fill-in at lagging-strand telomeres

Our previous results indicate that CDK1 plays a role in G-overhang generation.⁷ Because only one CDK1 inhibitor, purvalanol A (PA), was used in previous study, we therefore first determined the specificity of CDK1 inhibition using other highly selective CDK1 inhibitors CGP74514A (CGP) and RO3306,^37-40 together with PA. HeLa was synchronized at G₁/S boundary using double-thymidine block and then released into S phase. Since CDK1 activity is low in S phase, increases in late S/G₂ and peaks at M phase, whereas CDK2 activity peaks in late G₁/S phase and mainly controls G₁ to S phase transition,⁴¹ CDK1 inhibitors were added to media in late S/G₂ phase (6.5 hr after release) to avoid possible nonspecific inhibition of other CDKs during S or G₁ phase. In addition, the concentrations of inhibitors were carefully determined, so that minimal amounts of inhibitors were used to achieve CDK1 inhibition (Fig. S2). Under our experimental conditions, 8 μM PA, 6 μM CGP and 9 μM RO3306 were sufficient to inhibit CDK1 activity (Fig. S2). Cells were then collected at different time points, and the telomere overhang protection assay was used to determine the mean length of G-overhangs (Fig. S3). In DMSO-treated control cells, G-overhang length shortened from late S/G₂ to the next G₁ phase, whereas treatments with PA, CGP and RO3306 attenuated such G-overhang shortening (Fig. S3C). We also treated HeLa with roscovitine, a commonly used CDK1/2 inhibitor,⁴² at late S phase (6 h after release), and found that roscovitine also blocked G-overhang shortening in G₂ (Fig. S4). We previously showed that the attenuation of G-overhang shortening was not due to cell cycle arrest.⁷ Collectively, our results demonstrate that CDK1 regulates C-strand fill-in in G₂ phase.

To determine whether CDK1 regulates G-overhang generation at both daughter telomeres, synchronized BJ/E6/E7 cells were released into BrdU-containing media, and PA was added into the media at late S phase to inhibit CDK1 activity. Cells were collected from late S phase to the next G₁ phase (Fig. 2B). Addition of PA abolished G-overhang shortening at lagging-strand telomeres in BJ/E6/E7 (Fig. 2C and D), suggesting that CDK1 positively regulates the fill-in step at lagging telomeres.

C-strand fill-in is absent at leading-strand telomeres, and CDK1 unlikely regulates G-overhang generation at leading telomeres

When G-overhang dynamics of leading telomeres was measured, we found that G-overhang length of leading telomeres did not decrease from S to G₂/M phase in BJ/E6/E7 cells (Fig. 3A and B, DMSO-treated samples). Similarly, leading G-overhangs were not shortened in HeLa cells (Fig. 3C and D, DMSO-treated samples). Collectively, our results suggest that C-strand fill-in is absent at leading telomeres.

Figure 3. G-overhang dynamics at leading telomeres from BJ/E6/E7 and HeLa with or without CDK1 inhibitor. (A) G-overhang abundance at leading daughter telomeres from BJ/E6/E7 measured by non-denaturing in-gel hybridization assay. (B) Quantitation of G-overhang abundance at leading telomeres from BJ/E6/E7. (C) G-overhang abundance at leading-strand telomeres from HeLa measured by non-denaturing in-gel hybridization assay. (D) Quantitation of G-overhang abundance at leading telomeres from HeLa. Two-tailed t-test was used to calculate statistical significance. Results were from at least three independent experiments. Error bars: s.e.m. (E and F) show that leading overhangs are much shorter than lagging overhangs in BJ/E6/E7 without CDK1 inhibition (E) and with CDK1 inhibition (F).

We noticed that G-overhang abundance increased at leading telomeres from late S to G₂ phase in BJ/E6/E7, and this increase was statistically significant (Fig. 3B, DMSO-treated samples). Since telomerase activity is absent in BJ/E6/E7 cells, this increase suggested that in addition to the initial end resection in S phase, an additional end resection step takes place at C-strand of the leading telomere in late S/G₂ phase. This resection might be due to the continuation of the initial resection at a portion of leading telomeres or a distinct second resection independent of the initial end resection. The nature of C-strand end resection, both the initial and the second ones, remains to be elucidated.

We then analyzed whether CDK1 regulated G-overhang generation at leading-strand telomeres in G₂ phase. Treatment with PA did not significantly alter G-overhang length at leading telomeres in either BJ/E6/E7 or HeLa (Fig. 3B and D), indicating that CDK1 does not play a regulatory role in G-overhang generation at leading telomeres.

The recruitment of Polα to telomeres is not regulated by CDK1

Polα activity is required for C-strand fill-in.⁷ We then determined whether CDK1 regulates the recruitment of Polα to telomeres during the late S to G₂ phase using chromatin immunoprecipitation (ChIP) assay (Fig. 4). HeLa cells were synchronized, released into S phase and collected at indicated time (Fig. 4A). PA was added at middle to late S phase (5.5 h after release). ChIP was performed with the anti-Polα antibody, and the precipitated DNA was denatured, loaded on slot blot and hybridized to telomere probe. Consistent with previous report,³⁴ Polα re-associates with telomeres at G₂ phase in DMSO control cells, likely to complete C-strand fill-in at this time. Interestingly, CDK1 inhibition with PA showed no significant change in Polα telomeric association (Fig. 4C), indicating that CDK1 did not regulate the recruitment of Polα to telomeres. It is possible that CDK1 may control C-strand fill-in by regulating the RNA priming and/or synthesis activity of Polα.

Figure 4. The association of Polα to telomeres was not altered upon CDK1 inhibition. (A) FACS analysis of DNA content in synchronized HeLa used for ChIP. (B) Representative slot-blot results from ChIP of Polα. (C) Quantitation of ChIP of Polα. Two-tailed t-test was used to calculate statistical significance. Results were from at least three independent experiments. Error bars: s.e.m.

Discussion

We used both telomerase-negative and -positive cells to determine the molecular mechanism regulating G-overhang generation at telomeres replicated by leading- and lagging-strand syntheses. Our results show that in telomerase-negative cells, G-overhang shortening in G₂ phase exists at global telomeres (Fig. 1C). Further study reveals that such G₂-specific G-overhang shortening is caused by C-strand fill-in at lagging telomeres, and CDK1 positively regulates the fill-in step (Fig. 2D). In addition, our results demonstrate that C-strand fill-in is absent at leading-strand telomeres, and CDK1 plays a minimal role in regulating G-overhang generation at leading telomeres (Fig. 3B and D). Our results firmly establish that G-overhang generation at leading- and lagging-strand telomeres are regulated by distinct mechanisms.

Two pieces of evidence support our conclusion that C-strand fill-in is independent of telomerase expression in human cells. First, G-overhang shortens in G₂ phase in telomerase-negative cells (Fig. 2D).⁷ Second, if C-strand fill-in is exclusively used for replenishing C-strand after telomerase extends G-strand, a fill-in step should be observed at leading telomeres in telomerase-positive HeLa cells. However, C-strand fill-in is lacking at leading telomeres in HeLa (Fig. 3D), arguing against the notion that fill-in is dependent on telomerase extension of G-strand. So why does C-strand fill-in exist in telomerase-negative cells? We consider that the fill-in step is most likely necessary due to two reasons. First, the final RNA primer may not be positioned at the very end of a telomere, as suggested by a recent study.²⁴ In addition, during the processing of the final Okazaki fragment, a significant portion of C-strand DNA can be removed from the 5′ end of lagging-strand telomere.^14,43 If no mechanism fills in the gap, the lagging daughter strand would be rapidly shortened. C-strand fill-in, perhaps with a mechanism distinct from conventional lagging-strand synthesis,¹⁵ can attenuate such rapid telomere shortening in each replication. At the leading-strand telomere, such a problem does not exist, and end resection produces a much shorter G-overhang (less than 1/2 of the length of lagging overhang) (Fig. 3E and F).³⁶ This short G-overhang may not trigger a signal for C-strand fill-in. Given the importance of C-strand fill-in in telomere shortening, understanding the mechanism regulating C-strand fill-in will aid in developing novel therapies for preventing accelerated aging caused by rapid telomere shortening.

One important step controlling the generation of G-overhang as well as the rate of telomere shortening is end resection of C-strand. It will be important to understand the mechanisms underlying the regulation of end resection, because end resection needs to be tightly regulated to avoid excessive degradation of C-strand. In yeast, end resection is limited by the telomere protection protein complex Cdc13/Stn1/Ten1.^44-46 In mammalian cells, Pot1b appears to play this role.^14,47,48 However, depletion of human Pot1 leads to G-overhang shortening,^6,8 contradicting the expectation from protecting C-strand from degradation. This paradox highlights the complexity of end resection and telomere protection in human cells. Since the blunt-ended leading daughter telomeres resemble DSBs, it will be interesting to know whether enzymes involved in resecting double-strand breaks are the key players in resecting telomere ends.

Methods

Cell culture and synchronization

All cells were cultured at 37°C under 5% CO₂ in DMEM supplemented with 10% fetal bovine serum (FBS) or cosmic calf serum (Hyclone). Double thymidine block was used to synchronize HeLa cells as described previously.⁷ For synchronization of BJ/E6/E7, cells were serum starved for 48 h, split into DMEM media containing 20% FBS with 1 μg/mL aphidicolin for 24 h and then released into DMEM-20% FBS. Cells were then collected at different time points for analyzing DNA contents using a Beckman Coulter EPICS^® XL^™ flow cytometer. To inhibit CDK1 activity, the CDK1 inhibitors were dissolved in DMSO and added at indicated times and concentrations. Cells treated with DMSO were used as the control.

Separation of leading and lagging daughter telomeres

Leading and lagging daughter telomeres were isolated as described⁷ with minor modifications. Briefly, BJ/E6/E7 or HeLa cells were synchronized at G₁/S boundary and released into media containing 100 µM BrdU. Genomic DNA isolated from BrdU-labeled cells was digested with three restriction enzymes HaeIII, RsaI and AluI, mixed with CsCl solution and subjected to ultracentrifugation at 44,000 rpm for 13 h. After fractions were collected from bottom (high density) to top (low density), CsCl density at each fraction was measured by refractometer, and the amount of telomere DNA in each fraction was determined by hybridization to C-rich probe using slot-blot. DNA analyzed in the upper panel in Figure S1 was crosslinked to membrane with UV treatment, while DNA samples in the lower panel in Figure S1 were crosslinked to the membrane by baking at 95°C for 2 h. It has been shown that UV light, but not baking, is damaging to leading-strand telomeres due to the high content of BrdU and can lead to degradation of leading telomeres.⁴⁹ Thus, the amount of leading-strand telomeres appeared to be lower than that of lagging telomere in UV-treated samples, whereas the amounts of leading and lagging telomeres were similar in samples treated with baking (Fig. S1). Fractions containing leading or lagging telomeres were combined; the DNA was desalted, concentrated and used in G-overhang analysis. All procedures were performed in dark.

Telomeric G-overhang length measurement

The mean length of the telomeric G-overhang was measured by overhang protection assay as described previously.³⁵

Non-denaturing in-gel hybridization assay

Non-denaturing in gel hybridization was performed as described previously.⁷ Briefly, genomic DNA or isolated leading and lagging telomeres was treated with or without ExoI at 37°C for 3 h, followed with proteinase K at 55°C for 2 h and was resolved on 0.8% 0.5 × TBE agarose gel. The gel was then dried at r.t., hybridized to telomeric C-rich probe at 37°C for overnight, washed with washing buffer (0.1XSSC, 0.1% SDS) two times at 30°C and exposed to PhosphoImager (GE Healthcare). In the native gel, the signal from the ExoI-treated samples (ExoI +) represented the background and the signal from untreated samples (ExoI-) represented the G-overhang signal. The gel was then denatured in denaturing buffer (0.5 M NaOH, 1.5 M NaCl), rinsed with H₂O, neutralized with neutralization buffer (0.5 M TRIS-HCl pH7.5, 1.5 M NaCl), rehybridized to telomeric C-rich probe at 42°C for overnight, washed and exposed to PhosphoImager. The signal from the denatured gel represented the total amount of telomere DNA. The signal from native gel normalized by signal from denatured gel represented the G-overhang abundance.

ChIP assay

About 10 million cells were cross-linked with 1% formaldehyde in PBS for 10 min. Glycine (0.2 mM) was added to stop crosslinking. Cells were then collected with scraper, resuspended in lysis buffer [1% SDS, 10mM EDTA pH8.0, 50 mM TRIS-HCl pH8.0, 1 mM PMSF, 1 × protease inhibitor (Roche)], sonicated (10 sec pulse with 1 min intervals on ice, eight times) and centrifuged at 4°C for 10 min at 20,000 g. Supernatant was then diluted with dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM TRIS-HCl pH8.0, 150 mM NaCl, 1 mM PMSF, 1 × protease inhibitor) and incubated with 30 μL 50% protein G beads (Roche) at 4°C for 1 h to preclear the lysate. Precleared lysate was then incubated with antibody at 4°C for overnight. Protein G beads were then added to the mixture and incubated for 1 h at 4°C. After pelletting at 2,000 g for 2 min, beads were washed sequentially with buffer A (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM TRIS-HCl pH8.0, 150 mM NaCl, 1 mM PMSF, 1 × protease inhibitor), buffer B (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM TRIS-HCl pH8.0, 500 mM NaCl), buffer C (250 mM LiCl, 1% NP-40, 1% Na-Deoxycholate, 1 mM EDTA, 10 mM TRIS-HCl pH8), buffer D (1 mM EDTA, 10 mM TRIS-HCl pH8.0). For each wash, 1 ml of buffer was incubated with beads for 5 min at 4°C. Beads were then washed again with buffer D for 5 min and eluted with 300 μL buffer (1% SDS, 100 mM NaHCO₃) at 55°C for 15 min. Elutes were reverse crosslinked in 200 mM NaCl at 65°C for 6 h to overnight, treated with 20 μg of DNase-free RNase A at 37°C for 30 min, followed by 20 μg protease K treatment at 45°C for 1 h. DNA was then precipitated by ethanol and used for slot-blot to hybridize to telomeric probe.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Author Contributions

X.D. and C.H. performed experiments and analyzed data. X.D. and W.C. wrote the paper.

Acknowledgment

This work was supported by National Institute of Health R15GM099008 to W.C.