Systematic Analysis of Compounds Specifically Targeting Telomeres and Telomerase for Clinical Implications in Cancer Therapy

Aneuploidy or an abnormal chromosome number is a common feature of many cancers and is often accompanied by an elevated rate of chromosome instability (CIN; ref. 1). Gain or loss of entire chromosomes leads to changes in gene copy number and expression levels. Recent findings point to that experimentally induced CIN may be a barrier to tumor growth that can be exploited therapeutically (2–5). However, drugs that increase CIN beyond the therapeutic threshold are currently limited. Developing novel strategies to screen the compounds specifically modulating CIN and to exploit the fitness cost associated with excessive aneuploidy is important for the successful treatment of cancer.

Telomeres, protective caps at the ends of human chromosomes, are potential targets for increasing CIN. Telomeres shorten with each successive cell division in normal cells, whereas in tumors they are continuously elongated by telomerase reverse transcriptase (hTERT). Telomerase is overexpressed in 80%–95% of cancers and is present at very low levels in normal cells. Because telomerase plays a key role in cancer cell proliferation, it may serve as a useful target for anticancer therapeutics. Inhibition of telomerase in telomerase-positive tumors could potentially lead to a decrease of telomere length, resulting in CIN, cell senescence, and apoptosis.

Several strategies have been developed for telomerase inhibition based on targeting either hTERT or its RNA subunit with small-molecule inhibitors or antisense oligonucleotides (6). However, interest in their use for therapeutic approaches is diminished by the observation that significant effects on tumor growth were obtained only after long-term drug administration required for telomeres to reach a critical length. Such long-term telomerase inhibition has the potential to select for cells expressing the alternative lengthening of telomeres (ALT) pathway in which chromosomal ends are maintained by a recombination-based mechanism (7).

The telomeric structure itself offers a potential target for telomere-binding compounds in short-term treatments (8). Thus, G4 ligands, a class of molecules that are able to interact with G-quadruplex (abbreviated G4) structures formed by the G-rich overhang of telomeric DNA have recently received considerable attention (9–11). G4-stabilizing molecules such as Telomestatin, Braco-19, 360A (PDC) and RHPS4 can reduce cancer growth by rending telomeres dysfunctional. In contrast to hTERT inhibitors, G4 ligands represent a good example of a multimodal class of drugs, able even after a short treatment to simultaneously affect multiple targets participating in several distinct pathways (including the complex mechanism of gene expression). This simplifies the treatment modalities, improving the selectivity against cancer cells and avoiding the selection of ALT cells. Until now, there has been no reliable method to quantify the effect of different G4 ligands on chromosome instability.

In our previous work, we developed a quantitative assay to measure CIN in response to cell treatment by different compounds (12). This assay was based on the use of a circular human artificial chromosome (HAC) constructed in our laboratory (13). The HAC has previously been used for the efficient and regulated expression of genes of interest (14) and kinetochore studies (15–17). It contains centromeric repeats that form a functional centromere/kinetochore allowing its stable inheritance as a nonessential chromosome, albeit with a loss rate roughly 10× that of the native chromosomes (15). To develop a quantitative (“loss of signal”) assay for chromosome mis-segregation, we used a modified HAC carrying a constitutively expressed EGFP transgene (12). Cells that inherit the HAC display green fluorescence, while cells lacking the HAC do not. This allows the measurement of HAC loss rate by flow cytometry, providing a quick and efficient way to screen hundreds of drugs and identify those affecting chromosome mis-segregation. The assay was successfully used to rank different anticancer drugs according to their effects on chromosome transmission (18). More recently, this HAC-based assay was adapted for high-throughput screening of chemical libraries using a fluorescence microplate reader to identify compounds that elevate chromosome mis-segregation and drive lethal aneuploidy (19). In the modified assay, cells carry the EGFP transgene integrated in the genome and the HAC carries a constitutively expressed shRNA against EGFP. Cells display green fluorescence only after loss of the HAC (“gain of signal” assay). Using both HAC-based systems we have identified new and potentially less toxic agents that selectively elevate CIN in cancer cells.

In this work, we further modified our HAC-based screening protocol to identify compounds specifically targeting telomeres or telomerase. This protocol is based on the use of two isogenic cell lines expressing the EGFP transgene: one carrying a circular HAC lacking telomeres and the other carrying a linear HAC with telomeres. We hypothesized that compounds specifically inhibiting telomerase or other telomere functions would induce loss of the linear HAC but not the circular HAC. Our screen included known telomerase inhibitors as well as a set of known and newly developed G4 ligands. Among this last group, Pt-ttpy and Cu-ttpy (20, 21) induced the highest rate of loss of the linear HAC.

Identification of new compounds that greatly increase chromosome mis-segregation rates as a result of telomere dysfunction may expedite the development of new therapeutic strategies for cancer.

The human fibrosarcoma (HT1080; ATCC CCL-121), human colon carcinoma [HCT116; (ATCC CCL-247)] and human osteosarcoma (U2OS; ATCC HTB-96) cell lines were obtained from the ATCC and were authenticated both morphologically and by short tandem repeat analysis. All cell lines were tested regularly to confirm lack of Mycoplasma infection with Mycoplasma Detection Kit PlasmoTest from InvivoGen.

The human fibrosarcoma HT1080 cells (telomerase positive) harboring either alphoid^tetO-HAC-EGFP or 21ΔqHAC-EGFP were cultured in DMEM (Invitrogen) supplemented with 10% (v/v) tet system–approved FBS (Clontech Laboratories, Inc) at 37°C in 5% CO₂. Human colon carcinoma HCT116 cells (telomerase positive) were cultured in McCoy 5A medium supplemented with 10% FBS at 37°C and 5% CO₂. Human osteosarcoma U2OS cells (telomerase negative) were cultured in DMEM supplemented with 10% FBS at 37°C and 5% CO₂.

For chromosome instability experiments, HT1080 cells were grown in blasticidin-containing medium to prevent HAC loss prior to treatment with the drugs being tested (both linear and circular HACs contain the BS marker). After drug treatment, the cells were cultured in a nonselective medium to allow HAC loss, that is, under conditions when the cells that have lost a HAC are able to grow. For mitotic abnormality experiments, the cells were not exposed to blasticidin because the experiments were carried out in HT1080, HCT116, and U2OS cell lines not containing any HAC.

Analysis of EGFP expression was performed on a FACSCalibur instrument (BD Biosciences) using CellQuest acquisition software and analyzed statistically with FlowJo software. The cells were harvested by trypsin treatment. Intensities of fluorescence were determined by flow cytometry. A minimum of 4 × 10⁴ cells was analyzed for each cell sample.

Twenty-three different compounds were used in our experiments (Supplementary Table S1). Our experiment protocol was as follows. HT1080 cells containing a EGFP-HAC were maintained on blasticidin selection to select for the presence of the HAC. Approximately, 1 × 10⁵ cells were cultured either in the presence or absence of blasticidin selection in parallel with a third culture that was exposed to the agent under examination to test its effect on EGFP-HAC segregation. The compound concentration applied for measuring chromosome instability was adjusted to the LC₅₀ level for each compound (determined using a proliferation assay described below). Concentrations of compounds and lengths of treatment are presented in Supplementary Table S2. After treatment, the compound was removed by performing three consecutive medium washes and the cells were subsequently grown without blasticidin selection for 1–14 days. At the end of the experiment, cells were collected and analyzed by flow cytometry to detect the proportion of cells that retain EGFP fluorescence. This served as a measure of EGFP-HAC stability following compound treatment. For each compound, the experiments on measuring EGFP-HAC loss were carried out in triplicate. The results were reproducible and the std were small (for example, Pt-ttpy: SD ± 0.9%; MST-312: SD ± 1.1%).

To study mitotic abnormalities induced by Pt-ttpy and MST-312 in the telomerase-positive HT1080 and HCT116 and telomerase-negative U2OS cell lines, we used much lower concentrations of the drugs as at the LC₅₀ only very small numbers of mitotic cells were seen. For these experiments, the cells were treated by noncytotoxic drugs concentrations as a dose based on published data and proven experimentally for our cell lines (see Supplementary Table S2).

In our study, we determined the normal rate of spontaneous HAC mis-segregation (R_Normal) in the host cell line HT1080 using the formula, |${P_{normal}} = {P_0}{( {\frac{{2 - {R_{{\rm{Normal}}}}}}{2}} )^{{n_1}}}$|⁠; where P₀ is the percentage of EGFP(+) cells at the start of the experiment as determined by FACS. These cells were cultured under HAC selection conditions using blaticidin. P_Normal is the percentage EGFP(+) cells after culturing without HAC selection (no blasticidin) for a duration of t₁. In this study, t₁ was 14 days. n₁ is the number of cell doublings that occurs during culturing without blasticidin selection. The doubling time of HT1080 under normal growth conditions is approximately 18 hours. The number of cell divisions (n) is calculated by (t/host cell doubling time).

Once (R_Normal) was obtained, the rate of HAC loss induced by drug treatment (R_Drug) is determined using the formula, |${P_{{\rm{Treated}}}} = {P_0}{( {\frac{{2 - {R_{{\rm{Drug}}}}}}{2}} )^{{n_2}}}{( {\frac{{2 - {R_{{\rm{Normal}}}}}}{2}} )^{{n_3}}}$|⁠. As before, P₀ represents the percentage of EGFP(+) cells at the start of the experiment, cultured under HAC selection condition. P_Treated is the percentage of EGFP(+) cells at the end of the drug treatment experiment with a duration of (t₂ + t₃), where t₂ is the duration of drug treatment and t₃ is the duration of culturing after the drug is removed. (t₂ + t₃) was 14 days in this study. n₂ is the number of cell doublings that occurs during drug treatment, while n₃ is the number of cell doublings that occurs during the culturing without selection after drug treatment.

In this study, the duration of most drug treatments was less than the duration of a single cell cycle of HT1080 (t₂ < 18 hours). We made the assumption that any significant increase in HAC loss occurs only during the first mitotic division after wash-out of the drug (n₂ = 1). Thus, n₃ = (14 days/18 hours - 1). The algorithm we used is valid between the ranges R = 0 to 1. R values large than 1 indicate that the assumptions made in this model are incorrect. The assumption of synchronous growth in the model means that the estimated mis-segregation rate is lower than real values. As the spontaneous rate of HAC mis-segregation (R_Normal) was found to be low, this algorithm is relatively insensitive to the number of cell divisions that occurs post drug treatment (12).

For each compound LC₅₀ was determined using a MTS tetrazolium cell viability assay according to the manufacturer's instructions (CellTiter 96 AQueous Assay reagent; Promega). Briefly, the CellTiter 96 AQueous One Solution Reagent was added to each well and incubated at 37°C for 3 hours. Cell proliferation was determined by measuring the absorbance at 490 nm using a microtiter plate reader (Molecular Devices). The LC₅₀ was obtained from the MTS viability curves using GraphPad Prism 5. Experiments were carried out in triplicate.

The presence of the circular HAC in an autonomous form was confirmed by FISH analysis as described previously (22). HT1080 cells containing the HAC were grown in DMEM to 70%–80% confluence. Metaphase cells were obtained by adding colcemid (Gibco) to a final concentration of 0.05 μg/mL and incubating overnight. Media was aspirated, and the plate washed with 1× PBS. Cells were detached from the plate by 0.25% Trypsin, washed off with DMEM, pelleted, and resuspended in 10 mL of 50 mmol/L KCl hypotonic solution for 30 minutes at 37°C. Cells were fixed by three washes of fixative solution (75% acetic acid, 25% methanol). Between washes, cells were pelleted by centrifugation at 900 rpm for 4 minutes. Metaphase cells were evenly spread on a microscope slide and the fixative solution evaporated over boiling water. Dry slides were rehydrated with 1× PBS for 15 minutes and fixed in 4% formaldehyde-1× PBS for 2 minutes, followed by three 5-minute 1× PBS washes and ethanol series dehydration. PNA (peptide nucleic acid) labeled probes used were telomere (CCCTAA)₃-Cy3; PerSeptive Biosystems, Inc.) and tetO-alphoid array (FITC-OO-ACCACTCCCTATCAG; Panagene, South Korea). Ten nanomoles of each PNA probe was mixed with hybridization buffer and applied to the slide, followed by denaturation at 80°C for 3 minutes. Slides were hybridized for 2 hours at room temperature in the dark. Slides were washed twice in 70% formamide, 10 mmol/L Tris pH 7.2, 0.1% BSA, followed by three washes with 1× TBS, 0.08% Tween-20. Slides were dehydrated gradually with a series of 70%, 90%, and 100% ethanol washes and mounted (Vectorshield with DAPI). Images were captured using a Zeiss Microscope (Axiophot) equipped with a cooled charge-coupled device (CCD) camera (Cool SNAP HQ, Photometric) and analyzed by IP lab software (Signal Analytics). The PNA-DNA hybrid probes demonstrated a high hybridization efficiency and staining.

Slides with metaphases were rehydrated with 1× PBS for 15 minutes at room temperature, fixed with 4% paraformaldehyde, prepared on 1× PBS, for 2 minutes and washed three times with 1× PBS for 5 minutes. Then slides were gradually dehydrated at room temperature for 5 minutes each: 70% EtOH, 90% EtOH, 100% EtOH and then dried. Hybridization mix (10 mmol/L Tris-HCl, pH7.4; 70% formamide; 5% dextran sulfate; 10 ng PNA-TRITC-telomere (Panagen); 10 ng Biotin-labeled-EGFF sequence in the linear HAC was applied onto each dehydrated slide in 20 μL of volume and covered with a coverslip. Slides were next denaturated at 80°C on heating table for 3 minutes in the dark and then incubated 2–6 hours at room temperature in the dark. Cover glasses were then removed and slides washed two times in washing Solution I (70% formamide/10 mmol/L Tris-HCl pH 7.4/0.1% BSA) for 15 minutes, three times in washing solution II (20 mmol/L Tris-HCl pH 7.4; 136 mmol/L NaCl; 0.08% Tween) for 5 minutes, and briefly rinsed once in PBS. For biotin detection, 60 μL of Avidin-FITC solution (Sigma) was applied, covered with a 24 × 60 mm coverslip, and incubated for 40 minutes at 37°C in a moist chamber in the dark. To amplify the signal, slides were washed in 4× SSC/0.1% Tween-20 three times for 2 minutes at 45°C, after which, 60 μL of anti-avidin solution (Sigma, 1:200 dilution in 4× SSC/0.1 Tween-20) was applied, covered with 24 × 60 mm coverslip and incubated for 40 minutes at 37°C in a moist chamber in the dark. Next, a second round of detection was performed in the same manner. Then slides were gradually dehydrated at room temperature for 5 minutes each: 70% EtOH, 90% EtOH, 100% EtOH, and dried. Slides were mounted in Vectashield mounting media, containing DAPI (Vector Laboratories). Images were captured and analyzed using a DeltaVision microscopy imaging system and software in the CRC, LRBGE Fluorescence Imaging Facility (NIH, Bethesda, MD).

For Figs. 1 and 8, treated cells were fixed in 4% paraformaldehyde in PBS for 15 minutes at room temperature. Cells were rinsed two times quickly with PBS followed by incubation for 15 minutes with a last PBS wash at room temperature. two-hundred milliliters of 5% BSA in PBS-TT (PBS containing 0.5% Tween-20, 0.1% Triton X-100) were added to the washed cells and incubated for 30 minutes in a humid chamber. Cells were rinsed once in PBS-T (PBS, containing 0.1% Tween-20) for 5 minutes. Two-hundred microliters of γH2AX antibody (Abcam, catalog no. # 05-636, mouse origin, dilution 1:500) and TRF2 antibody (Santa Cruz Biotechnology, catalog no. # sc-9143, rabbit origin, dilution 1:200) in 1% BSA in PBS-TT were added for 2 hours at room temperature in the humid chamber. The samples were washed three times for 5 minutes in PBS-T. Two-hundred microliters of secondary antibodies (goat-anti-mouse Alexa 488, dilution 1:500, Life Technology A11029; goat-anti-rabbit Alexa 555, dilution 1:500, Life Technology A21428) were applied at room temperature in the humid chamber for 1 hour. The samples were washed three times in PBS-T for 5 minutes. The samples were counterstained with DAPI and mounted with mounting media (ProLong Diamond Antifade Mountant with DAPI, Life Technology, P36962). Samples were analyzed using DeltaVision Microscopic System at the CRC, LRBGE Fluorescence Imaging Facility (NIH, Bethesda, MD). For each compound, at least 120 nuclei were analyzed.

For Figs. 4–7 and Supplementary Figs. S5–S14, after drug treatment, the cells were rinsed in PBS and then fixed in 4% paraformaldehyde in PBS. Cells were then permeabilized and blocked before incubation with primary antibodies. Primary antibodies used in this study include: mouse monoclonal anti α-tubulin (SIGMA), rabbit polyclonal anti-Survivin (Novus Biologicals), rabbit polyclonal anti-Aurora B (Abcam), rabbit polyclonal anti-INCENP (Cell Signaling Technology), mouse mAb anti-Borealin (MBL), rabbit polyclonal anti-phosphorylated Histone3 Ser10 (Millipore), mouse monoclonal anti-LAP2 (BD Biosciences), mouse monoclonal anti-phosphorylated γH2A.X Ser139 (Abcam). Fluorophore-conjugated secondary antibodies were purchased from Jackson Immunoresearch. Samples were counterstained with DAPI and mounted in Vectashield. Microscope images were acquired on a DeltaVision Core system (Applied Precision) using an inverted Olympus IX-71 stand, with an Olympus UPlanSApo × 100 oil immersion objective [numerical aperture (NA) 1.4] and a LED light source. The camera (Photometrics Cool Snap HQ), shutter, and stage were controlled by SoftWorx 5.5.0 (Applied Precision). Z-series were collected with a spacing of 0.2 μm, and image stacks were subsequently deconvolved using SoftWorx. Projected Z-sections were exported as TIFF files into Adobe Photoshop.

Aliquots of approximately 2.4 × 10⁶ HT1080 cells carrying a circular alphoid^tetO-HAC or a linear 21ΔqHAC both carrying EGFP were incubated in 2 mL of nonselective medium containing 100 ng/mL of Trichostatin A (TSA; Wako) for 24 hours in the presence of doxycycline. Then, cell samples were collected and analyzed by FACS.

For Western blot analysis, HT1080 lysates were subjected to electrophoresis on SDS-PAGE followed by transfer to Hybond membrane (Amersham Biosciences) according to the manufacturer's instructions. Concentrations for blot analysis of anti-hTERT as well as secondary antibodies varied according to the manufacturer's recommendation (Santa Cruz Biotechnology, Jackson ImmunoResearch). The blot was developed with an ECL system (Amersham Biosciences).

To identify compounds specifically targeting telomeres or telomerase, we developed an assay based on the use of two different human artificial chromosomes (HACs; Fig. 1). Both HACs are of approximately the same size (5 Mb; refs. 23, 24) but one of them is circular while the other is linear and has telomeres (Fig. 1A). Both HACs have functional kinetochores and are maintained as nonessential extra chromosomes, replicating and segregating like normal chromosomes in human cells. The rate of spontaneous loss (no treatment) was approximately the same for circular and linear HACs (loss rate = 0.0024 and 0.0054, correspondingly). Both HACs carry the same enhanced GFP (EGFP) transgene flanked by HS4 insulators that is stably expressed from the HACs. Cells that inherit the HAC display green fluorescence, while cells that lack it do not (Fig. 1B). After drug treatment, there are two expected outcomes: either there is no change in the percentage of EGFP-expressing cells (no effect on HAC stability) or, if the compound induces chromosome segregation errors, there is an increase in the percentage of EGFP-negative cells (Fig. 1C). Control untreated cells containing either of these HACs show uniform green fluorescence. The actual percentage of cells carrying EGFP-HAC can be measured by FACS as described previously (18).

The circular HAC used in our experiments is the alphoid^tetO-HAC (15), which was engineered from a synthetic alphoid DNA array and was previously used to screen for drugs inducing CIN (18). The linear 21ΔqHAC was engineered by truncation of human chromosome 21 using telomere containing plasmids (25). These HACs possess several useful and similar features: (i) they have a well-defined architecture (24, 26); (ii) they are present episomally, independently of the host chromosomes (13, 14, 23, 27); (iii) they are mitotically stable both in human cell lines and mice, but slightly less stable than natural human chromosomes, so that the system is sensitized (15, 25, 28–32); (iv) EGFP expression from the HACs is stable (33, 34). No epigenetic silencing of EGFP was observed after 24 months of continuous culturing of HAC-containing human HT1080 cells under blasticidin selection (33, 34). Importantly, for these experiments, HT1080 cells containing either the linear or circular HAC express human telomerase (hTERT; Supplementary Fig. S1). The experimental design comparing segregation fidelity of the circular and linear HACs reported here can be used not only to identify new compounds that specifically target telomeres or telomerase but also to rank those compounds according to their effect on the segregation of the linear HAC.

Initially, we chose several groups of known antitelomerase compounds. They included antisense oligonucleotides that target the RNA template of human telomerase (hTERT), a set of small-molecule telomerase inhibitors that block the catalytic activity of the enzyme and compounds that block the access of telomerase to telomeres, among them G-quadruplex (G4) stabilizers. These compounds can bind to and stabilize the secondary DNA structures formed by genomic G-rich sequences such as telomeres. This blocks telomerase access to telomeres, uncapping them, and mimicking ssDNA overhang exposure. Thus, these agents are good candidates to increase chromosome instability. In this study, we used a panel of G-quadruplex stabilizers belonging to various chemical classes: the bisquinolinium series comprised of Phen-DC3, Phen-DC6, PDC and their functional derivatives Phen-DC3-C4Bn, PDC-C4Bn, Phen-DC3 Bisalk1; the styryl derivatives series PhenDV, PhenDV-An, and Bisphenyl-Vpy; the metal complexes series CuBisQ, Pt-ttpy and Cu-ttpy; and a series of benchmark compounds of diverse structures MMQ3, TrisQ, TMPyP4, BRACO19, PIPER, and Pyridostatin (Supplementary Table S1). A negative control CN1 that has no G4-binding ability was also included in the analysis. The compounds effecting telomerase were GRN163L, BIBR1532, MST-312, and 6-Thio-dG (Supplementary Table S1). Chemical structures of most of the compounds have been published previously (see references in Supplementary Table S1). The structures and synthesis of four G4-stabilizers developed recently in our group, Phen-DC3 Bisalk1, Phen-DC3C4Bn, PhenDMA, and Bisphenyl-VPy, are shown in Supplementary Fig. S2A–S2D. The structures of previously developed components, Pt-ttpy and Cu-ttpy, are shown in Supplementary Fig. S2E and S2F.

We previously demonstrated in other screens that the highest rate of drug-induced HAC loss occurs at the compound's LC₅₀ (12). Higher concentrations of drugs killed more cells but did not increase the rate of HAC loss (12, 18). In contrast, treatment at lower concentrations induced either no or lower frequency HAC loss. Because the LC₅₀ provides a useful parameter to normalize the results from different drugs/compounds we determined LC₅₀ values for all compounds, except the specific telomerase inhibitor GRN163L, that is, conditions under which HT1080 cell viability was around 50%. For GRN163L, we could not determine the LC₅₀ because even the highest concentration of this drug did not lead to detectable cell death. Thus, for GRN163L we used concentrations recommended in the literature (10, 35). The complete list of compounds used in this work is presented in Supplementary Table S1. The results of LC₅₀ determinations are shown in Supplementary Table S2.

Drugs that induce CIN via mechanisms that do not involve telomeres/telomerase should affect HAC stability independently of the presence or absence of telomeres. In our previous work using a circular HAC to rank a set of anticancer drugs according to their effect on chromosome instability (18), significant effects were observed after treatment with Taxol (microtubule-stabilizing agent), LMP400, and camptothecin [specific inhibitors of topoisomerase 1 (TOP1)], olaparib (PARP inhibitor), cisplatin and gemcitabine (DNA damage compounds). In the current control experiments, the rate of loss in response to drug treatment was indistinguishable between linear and circular HACs (Supplementary Fig. S3A). Notably, no increase in green cells was observed when the histone deacetylase inhibitor TSA (trichostatin A) was added to the culture (Supplementary Fig. S3B), indicating that these inhibitors do not induce transgene silencing but rather induces a high rate of linear HAC loss.

We next investigated the effect on the stability of circular versus linear HAC of one of the most specific telomerase inhibitors described to date, imetelstat or GRN163L. This compound consists of a 13-mer oligonucleotide N3’-P5’ thio-phosphoramidate covalently attached to a C16 (palmitoyl) lipid moiety. It directly binds to the RNA component of telomerase (hTERT) with very high affinity in the active site of the telomerase enzyme (10, 36). GRN163L has been shown to inhibit telomerase in a wide range of human solid tumor cells, including lung, breast, prostate liver, brain, and also in hematologic malignancies, including multiple myeloma and lymphoma (10, 35).

Cells carrying either a linear or a circular HAC, were treated with increasing concentrations of GRN163L. Treatments were carried out for 3 days, as shorter treatment times do not result in a significant shortage of the telomeres and chromosome instability (10, 35). As expected, treatment with increasing concentrations of GRN163L resulted in a specific concentration–dependent destabilization of the linear HAC (Fig. 2A). Importantly, the mitotic stability of the circular HAC was not affected under the same conditions. This proof-of-principle experiment indicated that our dual-HAC system allows to identify new compounds specifically targeting telomeres or telomerase.

Four known telomerase and telomere inhibitors, 6-Thio-dG (36), BIBR1532 (37), GRN163L (10, 35), and MST-312 (38), were included in the analysis. The LC₅₀ of these compounds in HT1080 cells is shown in Supplementary Table S2. HT1080 cells carrying either a linear or a circular EGFP-HAC were treated with each compound overnight. Figure 2B summarizes the results on HAC stability. Treatment with BIBR1532 increased the rate of loss of the linear HAC approximately 15 times compared with control untreated cells. The two other inhibitors tested, 6-Thio-dG and MST-312, also resulted in a significant increase of the rate of loss of the linear HAC, with the highest effect observed after treatment with MST-312 (approximately a 40-fold increase). Interestingly, this inhibitor also affected the stability of the circular HAC (approximately a 15-fold increase). Overnight treatment by GRN163L at 50 μmol/L only slightly and equally affected stability of linear and circular HACs despite the fact that much lower concentrations of the same compound (1 μmol/L and 10 μmol/L) but longer treatment (3 days) specifically and significantly increased the rate of loss of only the linear HAC (Fig. 2A).

To summarize, the dual-HAC assay allowed us to rank the strongest and most specific inhibitors of telomerase and telomeres according to their effects on chromosome stability. Two inhibitors in particular, 6-Thio-dG and MST-312, caused a significant increase of the rate of loss of the linear HAC after one day of treatment. 6-Thio-dG appeared to be the best among the compounds analyzed, as it exhibited both the highest and most specific effect on linear HAC stability (Fig. 2B).

Next, we investigated how structurally diverse G-quadruplex ligands affect chromosome stability. HT1080 cells carrying the circular or the linear EGFP-HAC were treated overnight with 19 different G4 ligands: Phen-DC3, Phen-DC6, TrisQ, Phen-DC3C4Bn, Phen-DV-Py, Phen-DV-An, CN 1, Phen-DC3 Bisalk1, PDC, PDC C4Bn, CuBisQ, Pt-ttpy, Cu-ttpy, Bisphenyl-Vpy, MMQ3, TMPyP4, BRACO19, PIPER, and Pyridostatin (LC₅₀ values of these compounds are shown in Supplementary Table S2). Figure 2C illustrates the effect of these compounds on the rate of HAC loss. As seen, no significant increase in loss of the circular HAC was detected with most of the analyzed compounds. Eight compounds, PDC C4Bn, TrisQ, Phen-DC3, BRACO19, Pyridostatin, Phen-DC6, Cu-ttpy, and Pt-ttpy, increased the rate of linear HAC loss. After treatment by Cu-ttpy and Pt-ttpy this rate was almost 70- and 90-fold higher compared with that observed in control untreated cells (Fig. 2C).

In summary, our dual-HAC assay allowed us to rank 19 G-quadruplex ligands according to their effects on chromosome stability. The two compounds, Cu-ttpy and Pt-ttpy, exhibited the highest effects on stability of the linear but not the circular HAC. This makes them potentially interesting candidates for future development as therapeutic agents, and therefore we decided to investigate further their mechanism of action in cancer cells.

Pt-ttpy and Cu-ttpy are metallo-organic complexes of a similar structure containing a tolylterpyridine moiety (ttpy) coordinated with either copper (Cu²⁺) or platinum (Pt²⁺) cation (Supplementary Fig. S2E and S2F; refs. 20, 21). To explore the mechanism by which they disturb chromosome transmission, we have performed the first detailed analysis of the effect of Pt-ttpy in mitosis. In parallel, we analyzed the effect of the telomerase inhibitor MST-312, a compound that also induces a high rate of instability in our assay for the linear HAC (Fig. 2B). In these experiments, the cells were treated by noncytotoxic drugs concentrations as a dose based on published data and proven experimentally for our cell lines (see Supplementary Table S2). Preliminary studies of MST-312 have revealed some mitotic abnormalities, but still many questions remained unanswered in this regard (38–40). To rule out cell line–specific phenotypes, all experiments were carried out in two different telomerase-expressing cell lines, HT1080 (Supplementary Fig. S1) and HCT116 (41).

Pt-ttpy- and MST-312–treated HT1080 and HCT116 cells were stained with antibodies against phosphorylated HistoneH3 (Ser10), a mitosis-specific modification, and against α-tubulin to visualize the mitotic spindles (Figs. 3A–D and 4A–H; Supplementary Figs. S4A–S4F and S5A–S5I). MST-312-treatment of HT1080 cells caused a significant decrease of the mitotic index (the fraction of cells in a population undergoing mitosis; Fig. 3A), associated with a significant increase in the proportion of mitotic cells in prometaphase and a decrease of the proportion of cells in cytokinesis (Fig. 3B). Further quantification revealed a significant increase in mitotic abnormalities in MST-312–treated cells (Fig. 3C). These included defects in chromosome alignment (Fig. 4C and D) and chromatin bridges in late mitosis (Fig. 4G). Quantification of chromatin bridges is of special interest because they can form when telomeres of sister chromatids fuse and fail to completely segregate into the respective daughter cells (42). In contrast, Pt-ttpy treatment of HT1080 cells caused no significant decrease in the mitotic index (Fig. 3A) or total percentage of mitotic abnormalities (Fig. 3C). Notably, Pt-ttpy also induced chromatin bridges (Fig. 4H).

HTC116-treated cells showed similar phenotypes to HT1080-treated cells (Supplementary Figs. S4A–S4F and S5A–S5I) but in addition we observed an increased frequency of spindle pole defects (Supplementary Fig. S5E and S5F). It thus appears that the spindle phenotypes are specific to the HCT116 cell line. We have observed similar effects previously in this cell line when studying the effect of other drugs treatments (Carmena, unpublished results). In contrast, the MST-312–induced chromosome misalignment phenotype was observed in both cell lines and is consistent with the increase in the frequency of prometaphases.

The defects in chromosome alignment shown by the MST-312-treated cells are reminiscent of the phenotypes caused by defects in chromosomal passenger complex (CPC) function. Aurora B kinase (the enzymatically active component of the CPC) destabilizes incorrect kinetochore–microtubule attachments and, therefore, promotes correct and timely chromosome biorientation (43, 44).

Levels and localization of the CPC components were analyzed in HT1080 (Fig. 5) and HCT116 cells (Supplementary Fig. S6). Levels of Survivin were reduced in MST-312–treated cells and its localization was abnormal in early mitosis (Fig. 5A–D; Supplementary Fig. S6A–S6C). Similar results were obtained when studying the distribution of the other CPC components, Borealin (Fig. 5A–D), INCENP, and Aurora B kinase (Fig. 5E–H; Supplementary Fig. S6D and S6E). To assess whether this had an effect on Aurora B activity, we stained MST-312–treated cells with an antibody against the activated form of the kinase (anti-Aurora B phospho-T232; Supplementary Fig. S7A–S7H). This revealed that the level of the active Aurora B was reduced in early mitosis. Thus, MST-312 affects, either directly or indirectly, the function of Aurora B kinase in both HT1080 and HCT116 cells. This is likely to be the cause of the defects in chromosome alignment and consequent delay in prometaphase observed in drug-treated cells.

Pt-ttpy treatment did not result in chromosome alignment defects as described above (Fig. 4; Supplementary Fig. S5). As known, MST-312 inhibits telomerase activity (45) while Pt-ttpy binds to G-quadruplexes, hindering telomerase access to telomeres (but not affecting telomerase activity; refs. 20, 21). Thus, one possible explanation is that telomerase activity is somehow required for Aurora B/CPC function, so this would not be altered after treatment with Pt-ttpy, as telomerase remains active. Another possible explanation is that inhibition of Aurora B kinase is an off-target effect of MST-312.

To test this possibility, we used U2OS, an ALT cell line that does not express telomerase (46). MST-312–treated U2OS cells showed the chromosome misalignment phenotype [Supplementary Figs. S8A–S8F (DMSO), S8A’–S8F’ (MST-312), and S9A and S9B], suggesting that indeed Aurora B kinase inhibition is likely an off-target, telomerase-independent effect of MST-312. This effect possibly contributed to the elevated rate of chromosome instability detected in our original screen and could also explain why this compound has also an effect on segregation of the circular HAC (Fig. 2A).

Our analysis of mitosis in drug-treated cells revealed the presence of chromatin bridges in late mitosis and cytokinesis (see Fig. 4 and Supplementary Fig. S5). To carry out a detailed quantification of the frequency of chromosome bridges in late mitosis in HT1080 and HCT116 cells treated with Pt-ttpy and MST-312, we used an antibody against Survivin and LAP2, a nuclear envelope protein that decorates chromatin bridges. This works because during late anaphase and telophase LAP2 and other nuclear envelope components decorate the decondensing chromatin while it is still in the process of moving polewards. With LAP2 antibody, bridges are readily visible, even when they are too fine to be visualized by DAPI staining (Fig. 6A–C; Supplementary Fig. S10A–S10E). These experiments revealed that our previous analysis had underestimated the number of chromatin bridges in late mitosis (anaphase to cytokinesis). As described above, we observed a statistically significant increase in the percentage of late mitotic cells showing chromatin bridges in cells treated with either Pt-ttpy or MST-312 (Fig. 3D; Supplementary Fig. S4F). A similar analysis in telomerase-negative U2OS cells showed no significant increase in chromatin bridges (Supplementary Fig. S9B). We, therefore, concluded that this phenotype is likely to be a consequence of the disruption of telomerase activity and/or telomere function.

We also analyzed the formation of UFBs following Pt-ttpy and MST-312 treatment. HCT116-treated cells (Supplementary Fig. S11A–S11C) were stained with antibodies against Blooms Syndrome protein (BLM) and against phosphorylated histone H2AX (γH2AX). UFBs were increased in anaphase/telophase after either treatment (Supplementary Fig. S11D). The similar results were obtained in HT1080-treated cells (Supplementary Fig. S12A–S12D). Importantly, no significant increase in the frequency of UFBs was observed in the telomerase-negative U2OS cells (Supplementary Fig. S13A–S13D).

To begin to investigate the mechanism(s) by which drug treatment resulted in chromosome loss, we stained double-stranded breaks (DSB) in interphase drug-treated cells with an antibody against phosphorylated γH2AX. The number of DSBs was increased after treatment with either Pt-ttpy or MST-312 (Supplementary Fig. S14A–S14F). To determine whether DSBs are associated with telomeric sequences, we examined colocalization of γH2AX foci with the telomeric protein TRF2 (telomeric repeat binding factor 2; Fig. 7A). Because the nucleoside analogue 6-Thio-dG is known to be incorporated into de novo–synthesized telomeres and cause DNA breaks (36, 47) it was used as a control in this experiment. In addition, we included in this analysis Cu-ttpy, a compound with a structure similar to Pt-ttpy. Immunostaining with antibodies against TRF2 and γH2AX was carried out every 24 hours for 3 days (see Materials and Methods). A statistically significant colocalization of γH2AX foci and TRF2 protein was observed at day 3 for 6-Thio-dG (4.5%), Pt-ttpy (6.2%), and Cu-ttpy (9%) (Fig. 7B). Note that telomeres represent only approximately 1/6,000 of the total human genome. Therefore, if the γH2AX foci were distributed evenly along the chromosomes, at telomeres we would expect to observe only 0.016% colocalization. As shown above, Pt-ttpy also induces additional DSBs outside of telomeres. This is consistent with the description of nontelomeric targets for G4 stabilizers in the promoter sequences of some oncogenes and other G-rich genomic regions (21, 48, 49). G-quadruplex–forming genomic sequences, similar to telomeres, represent natural replication fork barriers.

To summarize all of the above experiments, chromosome loss after Pt-ttpy treatment may be due to the induction of DSBs that are predominantly localized at telomeres. Such telomere damage could explain the increased formation of chromosome bridges and UFBs in the drug-treated cells.

As telomerase is constitutively expressed in many human cancers, telomerase-targeting therapy has been considered to be a potentially promising approach for cancer treatment (6, 50–52). However, a limited number of chemical compounds that target telomerase or telomeres have been identified and only some of these are in clinical trials. Moreover, protocols that allow quantitative comparison of the efficiency and specificity of these compounds were lacking.

To address this point, we developed a novel assay allowing comparison of libraries of compounds for their ability to induce telomere dysfunction leading to chromosome loss. Even transient telomere dysfunction can induce chromosomal instability (CIN) in human cells (53). Therefore, the activity of each compound can be evaluated on the basis of its effect on CIN. To quantify this effect, we used two isogenic cell lines with linear and circular HACs carrying the EGFP color marker. Specific destabilization of the linear HAC (containing telomeres) in response to drug treatment is consistent with specific targeting of telomeres or telomerase. Conversely, destabilization of both linear and circular HACs suggests that a compound has off-target effects in addition to its effect on telomerase. We initially verified our dual-HAC system using GRN163L, a well-characterized and highly specific inhibitor of telomerase. We found that this experimental design can be used to rank known and newly developed compounds targeting telomerase or telomeres.

In this study, the dual-HAC assay was applied to analyze a set of known and newly developed compounds targeting telomerase or telomeres, including G4 ligands. For each drug, the rate of HAC loss was quantified and within each analyzed group, the compounds were ranked according to their HAC-destabilizing potency. Analysis of four well-known telomerase inhibitors revealed the nucleoside analogue 6-Thio-dG (36, 47) as the highest and the most specific drug with respect to its effect on the linear HAC stability in this assay. It was quite unexpected that this drug, which may be incorporated into DNA strands during replication, induced a more specific effect on linear HAC stability than other analyzed telomerase inhibitors such as MST-312. Furthermore, our studies of MST-312 action in ALT cells that lack active telomerase suggested that this drug has other targets that are important for chromosome segregation, possibly including the CPC.

Ranking of 19 G-quadruplex (G4) ligands revealed that two recently developed compounds, Cu-ttpy and Pt-ttpy, exhibited the highest rate of linear HAC mis-segregation, that is, a 70- and 90-fold increase compared with the controls, respectively. Our results suggest that these two compounds or their derivatives may be useful for a new telomere-addressed anticancer approach, by exacerbating the CIN phenotype in cancer cells. One possible mechanism that might explain chromosome loss after Pt-ttpy and Cu-ttpy treatment is an increased number of double-stranded breaks (DSB) at or near telomeres. Such telomere damage may lead to the formation of chromosome bridges and UFBs that we observed in anaphase/telophase and could ultimately result in chromosome mis-segregation. Alternatively, treatment with these agents might interfere with telomere replication, leading to UFBs formation and subsequent DNA damage. It is worth noting that we cannot exclude additional mechanisms of action for these G4 ligands such as transcriptional inactivation of genes important for chromosome segregation or induction of chromosomal DSBs as previously was described for other G4 stabilizers (21, 48, 49). Indeed, other authors have reported that nonspecific induction of DNA damage enhances telomeric dysfunction induced by G-quadruplex–stabilizing agents (54).

Among other possible applications of the dual-HAC system described here is analysis of the genetic control of the alternative lengthening of telomeres (ALT) pathway in telomerase-negative cells. The ALT pathway is specific for cancer cells (7). Until now the development of ALT-targeted therapy has been challenging as the identification of proteins specific to ALT has proven elusive: all enzymes thus far shown to play any role in ALT are also critical for normal cellular functions. Transfer of circular and linear HACs into telomerase-negative cells might be used to identify genes or compounds that disturb the ALT pathway (e.g., by siRNA gene depletion).

To summarize, this study describes a novel approach that may be applied to the quantitative analysis of known and novel compounds used in cancer therapy to determine their ability and potency to specifically target telomeres or telomerase. The identification of compounds that selectively inactivate telomere replication and interfere with cell proliferation could lay the foundation for new treatment strategies for cancer.

No potential conflicts of interest were disclosed.

Conception and design: H.-S. Lee, M. Carmena, M. Oshimura, M.-P. Teulade-Fichou, Y. Pommier, V. Larionov

Development of methodology: Y. Pommier, N. Kouprina

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Carmena, M. Liskovykh, E. Peat, J.-H. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-S. Lee, M. Carmena, M. Liskovykh, J.-H. Kim, H. Masumoto, Y. Pommier, W.C. Earnshaw, N. Kouprina

Writing, review, and/or revision of the manuscript: H.-S. Lee, M. Carmena, J.-H. Kim, M. Oshimura, M.-P. Teulade-Fichou, Y. Pommier, W.C. Earnshaw, V. Larionov, N. Kouprina

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.-S. Lee, E. Peat

Study supervision: M. Carmena, W.C. Earnshaw, N. Kouprina

Other (provided several batches of chemical compounds used in the screening assays and newly synthesized batches of best candidates): M.-P. Teulade-Fichou

The authors are grateful to Dr. Kazuki from Institute of Regenerative Medicine and Biofunction, Tottori University for providing information about 21ΔqHAC. The authors are also grateful to Dr. Shay from University of Texas, Southwestern Medical Center for providing us GRN163L and 6-Thio-dG and fruitful discussion. The authors also would like to thank the CRC, LRBGE Fluorescence Imaging Facility (NIH) and personally Drs. Karpova and Dr. McNally for instructions, consultations, and help with the usage of a DeltaVision microscopy imaging system. This work was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (to V. Larionov, N. Kouprina, and Y. Pommier; grant number BC 006150), a Wellcome Trust Principal Research Fellowship (to W.C. Earnshaw; grant number 073915), the Grand-in-Aid for Scientific Research from Ministry of Education, Culture, Sports, Science and Technology of Japan (to H. Masumoto; grant numbers 16H04747 and 16H01414) and the Kazusa DNA Research Institute Foundation (to H. Masumoto).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.