A non-natural nucleotide uses a specific pocket to selectively inhibit telomerase activity

Telomerase, a unique reverse transcriptase that specifically extends the ends of linear chromosomes, is up-regulated in the vast majority of cancer cells. Here, we show that an indole nucleotide analog, 5-methylcarboxyl-indolyl-2′-deoxyriboside 5′-triphosphate (5-MeCITP), functions as an inhibitor of telomerase activity. The crystal structure of 5-MeCITP bound to the Tribolium castaneum telomerase reverse transcriptase reveals an atypical interaction, in which the nucleobase is flipped in the active site. In this orientation, the methoxy group of 5-MeCITP extends out of the canonical active site to interact with a telomerase-specific hydrophobic pocket formed by motifs 1 and 2 in the fingers domain and T-motif in the RNA-binding domain of the telomerase reverse transcriptase. In vitro data show that 5-MeCITP inhibits telomerase with a similar potency as the clinically administered nucleoside analog reverse transcriptase inhibitor azidothymidine (AZT). In addition, cell-based studies show that treatment with the cell-permeable nucleoside counterpart of 5-MeCITP leads to telomere shortening in telomerase-positive cancer cells, while resulting in significantly lower cytotoxic effects in telomerase-negative cell lines when compared with AZT treatment.


Introduction
Linear chromosomes end in telomeres, which are repetitive G-rich sequences (TTAGGG in mammals) that guard against illicit DNA repair and end-to-end fusions [1,2]. Because replicative DNA polymerases operate directionally, they are unable to fully synthesize the end of the lagging strand of DNA [3]. Telomeres function to absorb the shortening that occurs at each cell division and, thus, protect against the loss of genetic information. The progressive shortening of telomeres restricts the number of times a cell will divide until a state of replicative a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Screening of indolyl-2 0 -deoxynucleotide analogs for telomeric incorporation
Telomerase functions as a unique RT that uses a specialized, noncoding RNA as a template for telomere DNA synthesis [9,18,28,29]. This distinctive trait, along with additional differences in sequence, structure, and overall function compared with conventional replicative and repair polymerases [30,31], could provide key features of telomerase that can be exploited therapeutically. Pursuing this potential, we selected a set of indolyl-2 0 -deoxynucleotide analogs that exhibit different chemical properties and screened them for their ability to inhibit telomerase activity (Fig 1A and S1 Table). These analogs were designed to mimic the core structure of deoxyadenosine triphosphate (dATP), but with the introduction of various functional groups at the 5-or 6-position of the indole ring to systematically modulate biophysical features such as hydrophobicity, size variation, shape, and π-electron density that might contribute to selective interactions with telomerase. Modification of the nucleobase at these positions will additionally disturb proper Watson-Crick base pairing with the RNA template. Finally, the individual analogs investigated are already known to exhibit differential effects on diverse DNA polymerases that include HIV-1 RT, bacteriophage T4 DNA polymerase, and Escherichia coli Klenow fragment [32][33][34][35][36].
Alterations that these artificial nucleotide analogs had on telomerase activity were characterized using a modified direct in vitro telomerase activity assay [37]. Our first set of experiments were designed to determine whether any of the analogs serve as effective substrates for telomerase-mediated extension. As such, the indolyl analogs were used in place of the structurally similar native dATP nucleotide, along with deoxythymidine triphosphate (dTTP) and deoxyguanosine triphosphate (dGTP). After a 30-minute incubation period, the ability of telomerase to incorporate the nucleotide analogs into a telomere DNA product was evaluated by analyzing the extended products and comparing with those generated with dATP as the nucleotide substrate (S1 Fig). Extension of DNA products was negligible under these conditions, as product lengths were comparable to those measured in the complete absence of dATP and nucleotide analog (S1 Fig). These data suggest that telomerase cannot incorporate any of the analogs into telomere DNA products at the concentration tested or that selective incorporation of the indolyl analogs by telomerase causes chain termination to produce very short products.
To probe the mode of binding of 5-MeCITP to telomerase, another experiment was performed using the direct telomerase assay in the presence of 400 μM 5-MeCITP and with increasing concentrations of dATP. Fitting of the data revealed a nearly complete recovery of telomerase activity at elevated concentrations of dATP (S4 Fig). The K 1/2 for dATP increased by approximately four times when the experiments were performed in the presence of 5-MeCITP (S4 Fig). Cumulatively, these results suggest that 5-MeCITP inhibits telomerase activity by competing with dATP binding.

The noncanonical binding of 5-MeCITP disturbs telomerase RNA template base positioning
To identify the precise interactions that govern 5-MeCITP binding to telomerase, we crystalized the T. castaneum TERT-MeCITP complex along with a nucleotide hairpin that functionally mimics the telomerase RNA template-telomere DNA substrate [18]. The structure of the complex refined to 2.8 Å identifies density corresponding to an unpolymerized 5-MeCITP residing in the active site of the enzyme (Fig 3 and S2 Table). The triphosphate group of 5-MeCITP is coordinated by two metal ions in a manner that is structurally similar to that of a nonhydrolyzable nucleotide analog bound to the active site of other polymerases [38][39][40]. The modified nucleobase of 5-MeCITP adopts a noncanonical configuration in which it is rotated 180˚as compared with the expected orientation of a native dNTP forming Watson-Crick base pairs with the RNA template (S5 Fig). This configuration is very likely governed by an inability to properly base pair with RNA combined with selective interactions formed between the methylcarboxyl moiety of 5-MeCITP and TERT. Specifically, the methylcarboxyl moiety occupies a hydrophobic patch formed between motifs 1, 2 of the fingers domain and T-motif in the TRBD. The positioning of the methylcarboxyl group of 5-MeCITP is coordinated by interactions with Ile187 and Ile196 of motifs 1 and 2 and with Leu141 of the T-motif of T. castaneum TERT (Fig 3A-3C). The unique positioning of the modified nucleotide coincides with a displacement of the nucleobase of the template away from its canonical position (S5 Fig). It is worth noting that the RNA base displacement is likely to be subtle in the context of the fulllength RNA, where additional protein-RNA contacts within telomerase stabilize this interaction [41]. Nonetheless, an overlay of the available TERT structures of T. castaneum [18], Homo sapiens [42], and Tetrahymena thermophila [43] highlights a structural arrangement that is relatively conserved among species (S6 Fig). Similarly, the interaction with the methylcarboxyl moiety of the artificial nucleotide in T. castaneum TERT is coordinated by two isoleucine residues located in motifs 1 and 2 that are conserved in the human enzyme (Fig 3D). The methylcarboxyl group of 5-MeCITP is coordinated by a leucine in T. castaneum and a phenylalanine in H. sapiens, residing in T-motif of TERT. While either amino acid would be expected to make similar hydrophobic contributions to the pocket that is formed, the presence of a bulkier phenylalanine in human TERT might provide a tighter pocket for 5-MeCITP to bind to or require a structural rearrangement to accommodate the methylcarboxyl moiety. In any event, an analogous hydrophobic pocket is very likely formed in human telomerase to trap 5-MeCITP in a similarly inactive state.
The co-crystal structure shows that 5-MeCITP does not form direct interactions with the RNA template, suggesting that inhibition is not specific to incorporation of a specific dNTP. Rather, 5-MeCITP likely prevents access of all dNTPs to the active site of telomerase. To test this hypothesis, we performed the direct telomerase incorporation assay with 5-MeCITP and human telomerase along with dATP, dTTP, or dGTP at limiting concentrations in the reaction. As anticipated, 5-MeCITP displayed similar inhibitory properties under all three conditions (S7 Fig). These results further suggest that the inhibition of telomerase by 5-MeCITP is not specific against a particular dNTP substrate, but that it competes with all dNTPs for access to the telomerase active site.

Nucleoside 5-MeCIdR leads to telomere shortening in a telomerasedependent manner and is less toxic than AZT
We next characterized the growth properties of different cell lines that were treated with 5-MeCIdR, the cell-permeable nucleoside counterpart of 5-MeCITP. These studies used a panel of genetically distinct cell types that included A549 (lung cancer), HCT-116 (colon cancer), and MIA PaCa-2 (pancreatic cancer) cell lines that are telomerase positive, and U2OS (osteosarcoma) cells, which implement alternative lengthening of telomeres (ALT) instead of telomerase for maintaining telomere length. Finally, WI-38 (WI-38 VA13 subline) cells were used as human lung fibroblast cells that do not exhibit ALT or telomerase activity. In general, treatment with 5-MeCIdR (10-500 μM) for up to 3 days did not produce any adverse effects on cell viability for any of the five cell lines investigated (Fig 4A and 4B
https://doi.org/10.1371/journal.pbio.3000204.g002 We next monitored the effects of long-term exposure of cells to 5-MeCIdR (100 μM) on telomere length maintenance. Telomerase-negative U2OS cells treated with 5-MeCIdR maintained a constant telomere length distribution even after 180 population doublings (Fig 4C). In contrast, telomere length became progressively shorter in telomerase-positive A549, HeLa, and HCT116 cells treated with 5-MeCIdR and compared with DMSO control (Fig 4D and S9  Fig). Consistent with these data, long-term administration of 5-MeCIdR resulted in a significant increase in senescence-associated β-galactosidase activity in telomerase-positive cancer cells as compared with DMSO-treated control cells (S10 Fig). In summary, these data indicate that 5-MeCIdR treatment selectively leads to telomere shortening and initiates senescence in telomerase-positive cancer cells.

Discussion
As most tumors rely on telomerase to drive cellular immortality, the development of compounds designed to selectively impede telomerase activity remains a promising therapeutic strategy. In this study, we tested the ability of a set of indolyl nucleotide analogs that display diverse physicochemical properties to impede telomerase function. These results, and especially the identification of 5-MeCITP as an inhibitor of telomerase, provide exciting potential for further drug development.
While used with other antiviral agents in the treatment of HIV-1, the administration of AZT is associated with enhanced cytotoxicity in healthy cells, which has limited its clinical potential [44][45][46][47][48]. Here, we report that 5-MeCITP inhibits telomerase activity with a potency similar to that of AZT-TP. Compared with AZT, however, 5-MeCIdR administration was more tolerated by cells and produced minimal changes in cell viability. Together, these data indicate that 5-MeCIdR treatment is generally more tolerated by cells than AZT, potentially due to higher selectivity that limits off-target effects on metabolic pathways and non-telomerase polymerases.
Our results also provide essential and fundamental knowledge regarding the selectivity of small-molecule interactions with the telomerase catalytic site. For example, the structure of 5-MeCITP bound to T. castaneum TERT reveals that the carboxyl group helps to position the methyl moiety at a distance away from the ribose ring so that the methyl can fit into a specific hydrophobic pocket of TERT, while the phosphate groups and ribose ring occupy a native-like position in the active site of the enzyme. This finding could explain why the active site of telomerase is more selective for 5-MeCITP over the other nucleotide analogs we investigated, a notion that is supported by docking experiments. In this set of experiments, none of the other analogs were predicted to bind in a manner such that the modifications within the nucleobase could access the hydrophobic pocket formed by the T-motif and motifs 1 and 2 while concomitantly occupying the nucleotide binding pocket of TERT (S11 Fig). As such, it is very likely that the carboxyl group of 5-MeCITP merely serves as an extension to place the aliphatic methyl moiety in the vicinity of the hydrophobic pocket of TERT so that it may form favorable interactions there.
As the specific, physicochemical properties of 5-MeCITP facilitate its interactions with telomerase, the elements that form the unique binding pocket of telomerase might similarly explain the selective nature of 5-MeCITP binding to telomerase over other polymerases. Interestingly, the hydrophobic binding pocket of telomerase is formed by regions that are specific to telomerase (S12 Fig). The T-motif is a conserved structure that is unique to TERT proteins. It is postulated to interact with the template boundary element (TBE) [41] located just upstream of the RNA template to help mediate positioning of the RNA template to guide nucleotide selection and facilitate telomerase processivity [49][50][51]. Motifs 1 and 2 form an insertion within the fingers subdomain and are conserved among the RT family, including TERTs [7,29]. Interestingly, 5-MeCITP also inhibits HIV-1 RT activity [32]. Therefore, it is predicted that motifs 1 and 2 provide the more favorable interactions to promote 5-MeCITP binding to HIV-1 RT or TERT active sites and that contributions from the TERT-specific Tmotif play a less prominent role in the selective binding of this analog. In contrast to HIV-1 RT and TERT enzymes, other DNA polymerases lack the analogous environment formed by motifs 1 and 2 in the active site and coincidentally lack the putative hydrophobic pocket that facilitates interactions of 5-MeCITP with telomerase (S12 Fig). As a result, 5-MeCITP does not inhibit DNA polymerases such as T4 DNA polymerase and E. coli Klenow fragment. Instead, these polymerases can effectively incorporate 5-MeCITP, at least at positions that are opposite abasic sites in the template DNA [32,36].
The X-ray crystal structure of T. castaneum TERT bound with an RNA template mimic and a complementary telomeric single stranded DNA (ssDNA) has a 3 0 deoxyguanosine (dG) nucleotide occupying the catalytic site [18]. That particular structure represents a post-polymerization state in which the terminal dG is covalently bound to the synthesized ssDNA and its lone backbone phosphate group is coordinated by a single Mg 2+ ion. In contrast, the structure presented here reveals that 5-MeCITP is not incorporated into the DNA strand and instead binds to the T. castaneum TERT active site with the triphosphate coordinated by a pair of Mg 2+ ions (Fig 3). The α-phosphate of 5-MeCITP is situated at a distance that is approximately 5 Å from the nucleophilic 3 0 -hydroxyl group of the terminal dATP of the ssDNA that is being extended and is therefore too far for nucleophilic attack and incorporation into the 3 0 end of the DNA. Additionally, the inability of 5-MeCITP to properly base pair with the RNA template most likely prevents the nucleotide analog from adopting a position that would facilitate polymerization into the ssDNA chain. Without base pairing, the nucleobase of 5-MeCITP forms alternative interactions in the TERT active site. Specifically, our data reveal that it flips so that the methylcarboxyl group can selectively form hydrophobic interactions with the unique pocket formed by the T-motif and motifs 1 and 2 of TERT. This unorthodox positioning in the TERT active site further perturbs the complementary RNA base away from the active site (Fig 5). In summary, the inability of the modified nucleotide to form proper Watson-Crick base pairs with RNA combined with the favorable and unique interactions formed with TERT are important contributors to 5-MeCITP's unique mechanism of action. The structurefunction studies reported here therefore provide an opportunity to develop additional analogs to exploit these properties and to develop more selective and potent telomerase inhibitors.
where Norm. R f refers to the normalization of the radioactivity factor and is obtained by dividing the raw intensity values by the number of Gs added to the extended primer. LC of the lane refers to loading control for each individual lane, while Reference LC refers to a common loading control used for normalization between lanes. Dose response curves were generated by plotting the G incorporation value against the concentration of the drug being added.

Direct telomerase incorporation assay to determine K 1/2
The K 1/2 determination for each dNTP was conducted under the standard telomerase extension assay conditions described in the direct telomerase incorporation assay in the Methods section, but with increasing concentrations of that particular dNTP that ranged from 0-100 μM. Normalized G incorporation was plotted against that dNTP concentration. Normalized G incorporation was calculated by following Eq 1, and each reaction was then normalized as indicated in each figure. Normalized G incorporation was plotted against deoxynucleotide (dATP or dTTP) concentration to generate a dose-response curve that was fit by a hyperbolic equation using Kintek Explorer 7.6 software [53] and KaleidaGraph.version 3.5. The hyperbolic equation used to fit the data is described in Eq 2: where a is the amplitude, c is the offset, and [dNTP] is the deoxynucleotide triphosphate concentration that is changing (dATP or dTTP).

Direct telomerase incorporation assay to determine K' 1/2 of AZT and 5-MeCITP
For assays involving 5-MeCITP, concentrations of dTTP were maintained at 100 μM and dATP at a concentration near its determined K 1/2 value of 1 μM, along with increasing concentrations of 5-MeCITP that ranged from 0 to 2 mM. The observed K' 1/2 of AZT-TP was determined in a similar manner but with dTTP used at a limiting concentration near its K 1/2 value of 2 μM and dATP at 100 μM. For these reactions, increasing concentrations of AZT-TP (0-2 mM) were added to each reaction at the specified dose. Normalized G incorporation was calculated following Eq 1, and each reaction was then normalized to the reaction with no inhibitor or as indicated in each figure. Normalized G incorporation was plotted against inhibitor concentration to generate a dose-response curve that was fit by a hyperbolic equation using Kintek Explorer 7.6 software [53] and KaleidaGraph.version 3.5. The hyperbolic equation used to fit the data is described in Eq 3: where a is the amplitude, c is the offset, and [i] is the inhibitor concentration. This fitting generated an observed affinity constant rate value (K' 1/2 ) that describes the inhibitor concentration needed to observe half of the normalized G incorporation.

Cell culture
Human

Telomere restriction fragment analysis
Telomere restriction fragment (TRF) analysis was performed using a commercial kit using standard protocols (TeloTAGGG Telomere Length Assay, Catalog no. 12209136001, Roche Diagnostics Corporation, Indianapolis, IN). DNA was extracted from A459, HCT116, MIA-Pa-Ca-2, Wi-38, and U2OS cells after 6 months' treatment with 100 μM of 5-MeCIdR/ DMSO or DMSO alone. A total of 2 μg DNA was digested overnight with Rsa I and Hinf I at 37˚C and electrophoresed through 0.8% agarose gels in 1 × TAE at 50 V for 4 hours. Gels were denatured and neutralized prior to capillary transfer overnight. Telomeric DNA was transferred onto a Hybond-N + membrane (GE Healthcare, Chicago, IL) using 20× SSC buffer. The transferred DNA was fixed by UV cross-linking. The cross-linked membrane was then hybridized with a 32 P-labeled synthetic telomere probe with the sequence (GGGTTA) 4 overnight at 42˚C. After hybridization, the membrane was washed with buffer 1 (2× SSC, 0.1% SDS) at room temperature for 15 minutes and then washed twice with buffer 2 (0.5× SSC, 0.1% SDS) at 55˚C for 15 minutes. The membranes were exposed overnight to phosphorimager plates, which were imaged on a Typhoon FLA 9500 biomolecular imager (GE Healthcare, Chicago, IL). Image quantification was performed using ImageQuant software to measure the intensity value of each telomere smear. The weighted centers of mass of density plot profiles were generated for each sample. Numeric values generated from the histograms were compared against values of a known molecular weight standard, to obtain the average telomere length.

Colony forming assay
Cells were treated with vehicle, AZT, or 5-MeCIdn for 3 days and then harvested. For each treatment group, 500 viable cells (HCT116, A549, U2OS, or WI-38) were seeded in three 35-mm plates and evenly dispersed. The cells were then grown in drug-free media for 11 days. The colonies were fixed (75% methanol and 25% acetic acid) for 15 minutes at room temperature and stained (0.05% crystal violet and 95.05% methanol) for 30 minutes at room temperature.

Senescence-associated β-galactosidase staining
Cultured cells were plated into six-well dishes with a cover glass on it one week before staining. Cover glasses were collected and cells on it were fixed and stained as per the manufacturer's protocol (Senescence β-Galactosidase Staining Kit, Cell Signaling Technology no. 9860, Danvers, MA). Stained cells were imaged with a Leica SCN400 slide scanner. Ten random images were collected and cells counted with ImageJ 1.52e software. Staining was conducted 5 months after continual treatment with 100 μM 5-MeCIdR/DMSO or DMSO alone.

Protein crystallization and data collection
The T. castaneum TERT protein was purified as described by Mitchel and colleagues [18]. The protein was dialyzed in 10 mM Tris-HCl, 100 mM KCl, 1 mM TCEP, pH 7.5 prior to crystallization trials. The ternary complex was prepared by adding to the 10 mg/mL dialyzed protein, 1.2 molar excess nucleic acid consisting of the putative RNA template and the complementary telomeric DNA (rCrUrGrArCrCrUrGrArCTTCGGTCAGGTCA-Integrated DNA Technologies), 1 mM 5-MeCITP, and 2.5 mM MgCl 2 . Crystals of the monoclinic space group P2 1 were grown by the vapor diffusion sitting drop method. Drops were prepared by mixing one volume of the ternary complex with one volume of reservoir solution containing 0.1 M HEPES (pH 7.5), 8% PEG 8K, and 0.1 M KCl. The crystals were transferred into a cryosolution containing 0.1 M HEPES (pH 7.5), 8% PEG 8K, 30% PEG400, 0.1 M KCl, and 1 mM TCEP and flash frozen in liquid nitrogen. Data were collected from three crystals on a Rigaku MicroMax-007 HF rotating anode X-ray generator (wavelength 1.54178 Å) with VariMax optics and using a Saturn 944 HG CCD detector. The crystals were kept frozen with an Oxford Cryosystems Cobra system at 100 K during the data collection. The data were processed and scaled with XDS [54] (S2 Table). The coordinates and structure factors of the T. castaneum TERT-5-MeCITP complex have been deposited in the Protein Data Bank (PDB ID: 6E53).

Structure determination and refinement
Phases were calculated by molecular replacement (MR) using PHASER [55] as implemented in PHENIX [56] using the T. castaneum TERT-hybrid structure (PDB ID: 3KYL) as a search model. The maps revealed clear fo-fc density for 5-MeCITP at 3.0 sigma contour level bound at the active site of the enzyme. Model building was carried out in COOT [57] and the model was refined using REFMAC5 [58] (S2 Table). The structure was refined to good stereochemistry with 92.1% and 7.9% of the residues in the "most favorable" and "additional allowed" of the Ramachandran plot, respectively.

Statistical analysis
All statistical analyses were carried out using Excel software. Unless otherwise noted, experiments were conducted in triplicate. Comparisons between groups were performed using twotailed Student t test. Statistical significance was considered if p-values were <0.05. Individual p-values are indicated in figure legends. All results are expressed as means ± SD.