Cross‐species difference in telomeric function of tankyrase 1

Expression vectors.  GST fusion for mouse TAB182 C‐terminal fragment (GST‐mTAB182C) was constructed by cloning the PCR fragment above into pGEX‐5X‐2 (GE Healthcare, Buckinghamshire, UK). GST‐hTAB182C and GST‐hTRF1^{( 12 )} and GST‐mTRF1 and GST‐mTRF2^{( 16 )} fusions were constructed as described previously. FN‐tankyrase 1 constructs were made as described previously.^{( 13 )} For Myc‐NLS‐mouse tankyrase 1 (MN‐tankyrase 1), the full‐length cDNA of mouse tankyrase 1 was cloned into pcDNA3 (Invitrogen) with an N‐terminal Myc epitope tag and an NLS.

Subcellular fractionation and western blot analysis.  Subcellular fractions were obtained by using a ProteoExtract Subcellular Proteome Extraction kit (Calbiochem, San Diego, CA, USA). Each fraction derived from the same number of the cells was subjected to SDS‐PAGE. Western blot analysis was carried out as previously described^{( 12 )} with the following primary antibodies: rabbit anti‐tankyrase (H‐350, 2 µg/mL, Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse anti‐tankyrase (19A449, 1 µg/mL, Imgenex, Carlsbad, CA, USA), mouse anti‐GM130 (#35, 1 µg/mL, BD Transduction Laboratories, Lexington, KY, USA), rabbit anti‐calpain I (H65, 1 µg/mL, Santa Cruz Biotechnology), rabbit anti‐H2AX (NB100‐383, 0.2 µg/mL, Novus Biologicals, Littleton, CO, USA), or rabbit anti‐vimentin (H‐84, 1 µg/mL, Santa Cruz Biotechnology).

Northern blot analysis.  The mouse multiple tissue northern blot was obtained from Clontech. Mouse tankyrase 1 cDNA corresponding to amino acid 1–189 (HPS domain) was obtained by PCR and cloned into a pCR2.1 vector (Invitrogen). This fragment was purified and labeled using [α‐³²P]dCTP (3000 Ci/mmol) and a Megaprime labeling system (GE Healthcare). The blot was hybridized with the probe in 0.5 M sodium phosphate, pH 7.2, 1 mM EDTA, 7% SDS, 1% BSA at 65°C overnight and washed in 40 mM sodium phosphate, pH 7.2, 1 mM EDTA, 1% SDS at 65°C for 15 min, three times. Signals were detected on XAR‐5 X‐ray films (Kodak, Rochester, NY, USA) with a Lightning Plus intensifying screen at –80°C. Quantity and quality of RNA were monitored using a β‐actin probe (Clontech).

In vitro pull‐down assay.  GST‐fusion proteins were produced in the bacterium BL21‐CodonPlus‐RP strain (Stratagene, La Jolla, CA, USA) and purified with Glutathione‐Sepharose 4B (GE Healthcare). FN‐tankyrase 1 and MN‐tankyrase 1 proteins were prepared using the T_NT T7 transcription‐translation system (Promega, Madison, WI, USA). For GST pull‐down assays, 8 µL of FN‐ or MN‐tankyrase 1 was incubated for 1 h at 4°C with the GST fusion protein‐bound Glutathione‐Sepharose (containing about 1 µg of fusion protein) in TNE buffer [10 mM Tris‐HCl (pH 7.8), 1% Nonidet P‐40 (NP‐40), 150 mM NaCl, and 1 mM EDTA]. The beads were extensively washed with TNE buffer and subjected to SDS‐PAGE. Signals were detected by western blot analysis with anti‐tankyrase (H‐350).

In vitro PARP assay.  In vitro PARP assay was carried out as described,^{( 13 )} with a slight modification. One 15‐cm dish culture of HTC75 cells constitutively overexpressing FN‐tankyrase 1^{( 13 )} was resuspended on ice for 30 min in 1 mL of ice‐cold Buffer C/NP‐40 consisting of 20 mM HEPES, pH 7.9, 0.42 M KCl, 25% glycerol, 0.1 mM EDTA, 5 mM MgCl₂, 1 mM DTT, 1/40 volume of the protease inhibitor cocktail (Sigma, P8340, St. Louis, MO, USA), and 0.2% NP‐40. After centrifugation at 9000g for 10 min at 4°C, the supernatant was collected as whole cell extract. For four PARP reactions, 0.45 mL of the cell extract was supplemented with 30 µL (50% slurry) of EZview Red ANTI‐FLAG M2 Affinity Gel (Sigma) and was slowly rotated for 1 h at 4°C to form the FN‐tankyrase 1/anti‐FLAG immunocomplex. The beads were washed with ice‐cold Buffer C/NP‐40 and then washed twice in 1 mL of ice‐cold PARP buffer without DTT. Then, the immunocomplex was resuspended in 0.8 mL of ice‐cold PARP buffer containing 0.2 mM DTT, and the suspension was divided into four microtubes. Finally, the beads were resuspended in 14 µL of PARP buffer containing 0.2 mM DTT and 0.5 µg of GST‐TRF1 or GST protein on ice. The PARP reaction was started by adding 2 µL of the [³²P]‐NAD solution to the microtube. To prepare the [³²P]‐NAD solution, 1.65 µL of 1 mM NAD (non‐radioactive) was added to 25 µL of [³²P]‐NAD (29.6 TBq/mmol, PerkinElmer, Wellesley, MA, USA). The resulting solution was diluted 1:4 with distilled water and used as the [³²P]‐NAD solution. The PARP reaction was carried out at 25°C for 45–60 min and subjected to conventional SDS‐PAGE. The gel was dried and subjected to autoradiography at –80°C with an intensifying screen.

DNA transfection and fluorescence microscopy.  Cells were transfected with 15 µg of vector DNA using a standard electroporation method. After incubation for 18–21 h at 37°C, cells were fixed with 2% paraformaldehyde/PBS and permeabilized with 0.5% NP‐40/PBS. For the Triton‐X100 extraction, cells were processed and fixed as described previously.^{( 17 )} For immunofluorescence staining, the fixed cells were blocked in PBS containing 1% BSA and incubated with the following antibodies: mouse anti‐FLAG (M2, 2 µg/mL, Sigma), mouse anti‐Myc (9E10, 2 µg/mL, Santa Cruz Biotechnology), rabbit anti‐Myc (2 µg/mL, Santa Cruz Biotechnology), rabbit anti‐human TRF1 (5747, affinity purified, 1 µg/mL),^{( 13 )} rabbit anti‐mouse TRF1 (affinity purified, 3.5 µg/mL),^{( 16 )} or mouse anti‐γH2AX (JBW301, 2 µg/mL, BD Transduction Laboratories). The primary antibodies were detected with FITC‐ or Texas Red‐conjugated donkey anti‐rabbit or sheep anti‐mouse Ig (1:25, GE Healthcare). DNA was stained with 0.2 µg/mL 4,6‐diamino‐2‐phenylindole (DAPI). For iFISH assay, the paraformaldehyde‐fixed cells were rinsed in water, 70% ethanol, and 95% ethanol and air‐dried. Then the cells were incubated with 300 ng/mL Cy3‐PNA (CCCTAA)₃ probe in PNA dilution buffer consisting of 70% formamide, 10 mM Tris‐HCl, pH 7.5, 0.05% blocking reagent (Roche, cat. #1096176, Basel, Switzerland) at 83°C for 4 min and at room temperature for 2 h. Then the cells were washed with PNA wash solution consisting of 70% formamide, 10 mM Tris‐HCl, pH 7.5, and 0.1% BSA for 15 min twice. After further washing with 0.1% Tween 20/PBS for 5 min three times, the cells proceeded to immunofluorescence staining. Images were acquired using an Olympus IX‐71 microscope with a DP70 digital camera and a Lumina Vision software (Mitani Corporation, Tokyo, Japan).

Co‐immunoprecipitation assay.  HeLa I.2.11 cells were transfected with 15 µg of vector DNA. After incubation for 23 h at 37°C, the cell lysate was prepared and then subjected to coimmunoprecipitation assay, as described previously.^{( 13 )}

Mouse TRF1 does not bind tankyrase 1.  In humans, the nuclear fraction of tankyrase 1 binds and poly(ADP‐ribosyl)ates TRF1. This post‐translational modification of TRF1 is a prerequisite for tankyrase 1 to execute its function at telomeres.^{( 7} , ⁹ , ^{19 )} A highly conserved amino acid sequence and its presence in the nucleus (Fig. 1) suggest that mouse tankyrase 1 has a telomeric function, as human paralog does. Meanwhile, because mouse TRF1 does not have a canonical tankyrase 1 binding motif, RXX(P/A)DG,^{( 14 )} it is also possible that tankyrase 1 does not recognize TRF1 in mice. To examine these possibilities, we carried out an in vitro pull‐down assay by using GST‐fused full‐length TRF1 and in vitro‐translated full‐length tankyrase 1. As shown in Fig. 2a, human TRF1 bound both human and mouse tankyrase 1. However, mouse TRF1 did not bind human or mouse tankyrase 1. The interaction between human/mouse tankyrase 1 and human TRF1 was specific because GST alone did not pull down tankyrase 1. Also, mouse TRF2, as a negative control, did not form a complex with tankyrase 1.

So far, whether RXX(P/A)DG recognition by tankyrases is conserved among species has not been investigated. To examine the interaction between tankyrase 1 and another binding partner, TAB182 (tankyrase‐binding protein of 182 kDa),^{( 12 )} in mice, we cloned a 1.9‐kb partial cDNA fragment of mouse TAB182. The amino acid sequence deduced from the partial cDNA fragment exhibited 74% identity and 79% homology with that of human TAB182 (data not shown). Mouse TAB182 retained the RXXPDG motif (RPQPDG) as did the human paralog (Fig. 2b). As shown in Fig. 2c, we confirmed that both human and mouse TAB182 associated with human/mouse tankyrase 1 in vitro. Since the RXX(P/A)DG motives in other known tankyrase‐binding proteins seem to be conserved among several species,^{( 14 )} TRF1 could be a species‐restricted partner of tankyrases.

In humans, to poly(ADP‐ribosyl)ate TRF1, tankyrase 1 needs to recognize the RXXADG motif (RGCADG) of TRF1 via the ANK subdomain closest to the C‐terminal, called ARC V.^{( 13 )} Fig. 2d shows poly(ADP‐ribosyl)ation of human TRF1 by tankyrase 1 in vitro (lane 5). 3‐Aminobenzamide (3AB), a PARP inhibitor, diminished the amount of radioactive TRF1 (lane 6), confirming that the incorporated radioactivity to the protein was derived from genuine PARP activity of tankyrase 1. Consistent with there being no physical interaction between the two proteins, mouse TRF1 was not poly(ADP‐ribosyl)ated by tankyrase 1 at all (lane 7).

Mouse TRF1 resists tankyrase 1‐mediated dissociation from telomeres.  Telomeric function of tankyrase 1 can be monitored by the transient overexpression system in intact human cells.^{( 7} , ^{20 )} Thus, in a cell that overexpresses exogenous tankyrase 1 in the nucleus, TRF1 dissociates from telomeres in the tankyrase 1 PARP activity‐dependent manner.^{( 7} , ^{9 )} This phenomenon can be visualized as the disappearance of the telomeric spots of TRF1 by means of indirect immunofluorescence microscopy.^{( 7} , ⁹ , ^{13 )} In analogy with human tankyrase 1, mouse tankyrase 1 efficiently removed TRF1 from telomeres in human HeLa I.2.11 cells (Fig. 3a). This is consistent with the result that mouse tankyrase 1 recognized human TRF1 in vitro (Fig. 2). In this TRF1 release assay, tankyrase constructs that were tagged with an N‐terminal NLS have been conventionally used.^{( 7} , ⁸ , ⁹ , ¹³ , ^{19 )} Because more than half of the endogenous tankyrase 1 reside outside the nucleus (Fig. 1), NLS tagging to the exogenous gene allows us to enhance the function of tankyrase 1 exclusively in the nucleus. By using an iFISH analysis, we confirmed that the dot‐like signals of TRF1 coincided with telomere DNA in HeLa I.2.11 cells (data not shown).

In contrast with the results in HeLa I.2.11 cells, mouse tankyrase 1 failed to dissociate TRF1 in mouse NIH3T3 cells (Fig. 3b). The levels of exogenous protein expression were comparable between the two cell lines (data not shown). Similarly, mouse TRF1 still resided on telomeres in human tankyrase 1‐overexpressing NIH3T3 cells (Fig. 3c). These observations are consistent with a lack of in vitro association between mouse TRF1 and tankyrase 1 (Fig. 2).

In the presence of PARP inhibitors that are effective against tankyrase 1, TRF1 poly(ADP‐ribosyl)ation and its release from telomeres are inhibited both in vitro and in intact cells. Because such PARP inhibitors do not interfere with TRF1 binding to tankyrase 1, telomeric colocalization of these two proteins can be observed in PARP inhibitor‐treated HeLa I.2.11 cells.^{( 19 )} In other words, the emergence of telomeric tankyrase 1 dots in PARP inhibitor‐treated cells is proof of the TRF1–tankyrase 1 interaction in intact cells. As shown in Fig. 3d), overexpression of mouse tankyrase 1 in the presence of 3AB colocalized the exogenous tankyrase 1 with TRF1 in HeLa I.2.11 cells. Telomeric localization of tankyrase 1 was also confirmed by costaining with anti‐TRF2, another telomere binding protein, and was not detected in 3‐aminobenzoic acid‐treated cells (data not shown). 3‐aminobenzoic acid is a non‐PARP inhibitory analog that we used as a negative control. In 3AB‐treated NIH3T3 cells, overexpressed tankyrase 1 did not reside with TRF1 (Fig. 3e), indicating that these two proteins had no interaction even in a whole cell. In fact, we failed to detect an intracellular complex formation between mouse TRF1 and tankyrase 1 by a coimmunoprecipitation assay (Fig. 2e). Taken together, these data indicate a lack of functional interaction between tankyrase 1 and TRF1 in mice and suggest that tankyrase 1 does not regulate telomerase access to mouse telomeres.

Non‐telomeric nuclear foci of tankyrase 1.  Although the 3AB‐treated NIH3T3 cells exhibited no colocalization of tankyrase 1 with TRF1, we found that they showed significant numbers of nuclear tankyrase 1 foci without colocalization with TRF1 (Fig. 3e). These foci were resistant to extraction with Triton X‐100 detergent prior to paraformaldehyde fixation and therefore seemed to bind DNA and/or nuclear matrices (Fig. 3f). iFISH analysis that costained the exogenous tankyrase 1 and telomere DNA confirmed that these tankyrase 1 foci were, indeed, non‐telomeric (Fig. 4a,e). In 3AB‐treated human HeLa I.2.11 cells, up to 70% of the exogenous tankyrase 1 foci colocalized with telomeres, whereas, of note, the rest of the foci were non‐telomeric (Fig. 4c,f). Another PARP inhibitor, 4‐amino‐1,8‐naphtalimide (4ANI), which is effective against tankyrase 1,^{( 19 )} gave similar results to those with 3AB (Fig. 4b,d).

Among the five ARCs that work as independent TRF1 recognition units in the tankyrase 1 ANK domain, the C‐terminal ARC V is essential, whereas ARCs I–IV are auxiliary to telomere elongation.^{( 13 )} While ectopic expression of the wild‐type FN‐tankyrase 1 gave its nuclear foci in 80–90% of 3AB‐treated HeLa I.2.11 cells, its P906L point mutant (Fig. 5a, upper panel; FN‐tank‐ARC Vmut), which lacked TRF1 binding at ARC V,^{( 13 )} formed nuclear foci in only 41% of the cells (data not shown). Furthermore, deletion of ARCs III–V (FN‐tank‐ARC I + II) almost completely abolished the tankyrase 1 foci formation in 3AB‐treated HeLa I.2.11 and NIH3T3 cells (Fig. 5a, lower panel and data not shown). These observations indicate that the foci formation requires an array of multiple ARCs as a platform for protein–protein interactions.

Meanwhile, PARP inhibitors may increase damaged DNA by inhibiting PARP‐1, which is the most abundant enzyme among the PARP family and is involved in DNA repair.^{( 21 )} Under our experimental conditions, however, PARP inhibitors did not induce γH2AX foci formation, an indication of DNA double‐strand breaks. Furthermore, 3AB‐treated tankyrase 1 foci did not colocalize with γH2AX foci, which happened spontaneously in a very minor fraction of the cells (data not shown).

Tankyrase 1 does not accumulate on rat telomeres.  According to the NCBI nucleotide database, rat TRF1 (NM_001012464) has an RSNAGG motif, which does not correspond to the tankyrase 1 recognition consensus, RXX(P/A)DG. Thus an important aspartic acid at the fifth position, which is required for tankyrase 1 recognition,^{( 14 )} is not conserved. In the nucleus of the rat REF cell, we overexpressed mouse tankyrase 1 (MN‐tankyrase 1), whose amino acid sequence showed 99% identity with the predicted sequence of rat paralog (XM_001059294). As shown in Fig. 5b, the tankyrase 1 foci induced by 3AB were not colocalized with rat telomeres. As seen in mice, this observation strongly suggests no interaction between tankyrase 1 and TRF1 in rats.

Why does murine TRF1 not interact with tankyrase 1?  In humans, the average telomere length is 5–20 kb, and telomerase activity is highly repressed in most somatic tissues. Strictly speaking, normal human fibroblasts activate telomerase at the S phase of the cell cycle.^{( 24 )} This transient activation of telomerase has been reported to be important for the integrity of the telomeric 3′‐overhang, a key structure for the end‐capping function of telomeres.^{( 24 )} Still, however, the level of such telomerase activation in normal somatic cells is too low to overcome the end replication problem of linear chromosome DNA. Thus, telomeres of normal human somatic cells erode as the cells proliferate. We have previously shown that when the level of telomerase catalytic activity is limited, this enzyme's accessibility to telomeres becomes a limiting factor for telomere elongation.^{( 19 )} In that sense, tankyrases, which enhance telomere access to telomerase, may be needed as an auxiliary player in human telomere maintenance.

On the other hand, mouse telomeres are extremely long, 20–50 kb, and many mouse somatic tissues exhibit high telomerase activity, although some strains, such as Mus spretus, retain exceptionally short telomeres (5–15 kb). One speculation is that mice do not need the tankyrase‐mediated auxiliary system for telomere maintenance simply because their telomeres are long enough, even without tankyrases. In fact, replicative senescence by telomere shortening is uncommon in mice. Neither did we obtain evidence for tankyrase 1/TRF1 interaction in rats, which also have very long telomeres and abundant telomerase activity in normal somatic tissues.^{( 25 )}

According to currently available DNA databases, TRF1 paralogs of other mammals, such as cattle, dogs, orangutans, chimpanzees, and rhesus monkeys, retain the tankyrase‐binding motif, RXX(P/A)DG [RG(C/R)ADG]. The telomere length of these species is similar to that of humans but is much shorter than those of mice and rats.^{( 26} , ²⁷ , ^{28 )} Furthermore, tissue distribution of telomerase in these species resembles that in humans. These facts suggest that TRF1 down‐regulation by tankyrases is conserved among most mammals, whereas mice and rats have long telomeres and high telomerase activity instead of a tankyrase‐mediated telomere elongation system.

Cross‐species conservation in other functions of tankyrases.  Tankyrases have a variety of binding partners other than TRF1, suggesting that they have additional, non‐telomeric functions.^{( 5 )} Recently, tankyrase 1 has been reported to play a role in sister chromatid separation, spindle pole formation, or both in mitotic cells.^{( 29} , ^{30 )} In mitosis, tankyrase 1 binds and poly(ADP‐ribosyl)ates NuMA, a centrosomal protein that is required for spindle formation and function.^{( 31 )} In contrast to TRF1, the tankyrase‐binding motif of NuMA (RTQPDG) is conserved in humans and mice.^{( 14 )} TAB182, another tankyrase‐binding protein, also has a conserved tankyrase‐binding motif (RPQPDG), and in this study we confirmed that tankyrase 1/TAB182 interaction occurred in mice. Thus, it is possible that humans and mice share non‐telomeric functions of tankyrases, which are mediated through interaction with binding partners other than TRF1. Meanwhile, highly contrastive tankyrase 1 expression in mouse and human skeletal muscle (Fig. 1b and Cook et al. ⁹ ) suggests another cross‐species difference in tankyrase 1‐mediated physiological events.

Identification of non‐telomeric tankyrase 1 foci.  In the present study, we found non‐telomeric tankyrase 1 foci in the nuclei of PARP inhibitor‐treated human, mouse, and rat cells. One possibility is that this focus formation might be due to aggregation or mislocalization of the ectopically expressed protein. Meanwhile, tankyrase 1 can form a specific multimer via its SAM domain,^{( 32 )} and ectopically expressed tankyrase 1 often forms intracellular vesicles by itself^{( 32 )} (our unpublished observations, 2004). However, the appearance of the nuclear foci in this study depended on treatment with PARP inhibitors. Moreover, when ligand recognition units ARCs III–V were deleted from the tankyrase 1 ANK domain, the foci disappeared in spite of the presence of the SAM domain. This suggests that tankyrase 1 has a nuclear partner other than TRF1 and regulates its function by poly(ADP‐ribosyl)ation. Identification of such a partner (or partners) would be of interest as a next goal of the study. From a drug development viewpoint, formation of tankyrase 1 foci is an indication of the enzyme's inhibition in intact cells and therefore could be applied as a screening strategy for PARP inhibitory compounds.^{( 20 )}

Tankyrases as targets for cancer therapy.  From a therapeutic perspective, tankyrase 1 is expected to be a novel target for cancer treatment. Telomerase inhibitors shorten telomeres and induce subsequent cancer cell crises.^{( 33 )} However, complete and durable inhibition of telomerase is thought to be technically difficult. As a result, residual telomerase activity in conjunction with partially shortened telomeres often causes resistance to telomerase inhibitors in target cancer cells:^{( 19 )} shortened telomeres have fewer TRF1 and therefore enhance access to the residual telomerase activity. Since tankyrase 1 decreases the TRF1 loading on a chromosome end, its blockade resumes telomere binding of TRF1, which in turn compromises telomere access to the residual telomerase activity. Thus, inhibition of tankyrase 1 increases the pharmacological effect of telomerase inhibitors.^{( 19 )} The present study suggests that a mouse model system may not be suitable for anti‐tankyrase 1 therapeutic trials, especially for telomere‐directed anticancer therapy. Also, this sort of cross‐species difference in a certain gene product must be taken into consideration when its functional annotation is incorporated into gene ontology databases.