Tankyrase, a Poly(ADP-Ribose) Polymerase at Human Telomeres

See allHide authors and affiliations

Science  20 Nov 1998:
Vol. 282, Issue 5393, pp. 1484-1487
DOI: 10.1126/science.282.5393.1484


Tankyrase, a protein with homology to ankyrins and to the catalytic domain of poly(adenosine diphosphate–ribose) polymerase (PARP), was identified and localized to human telomeres. Tankyrase binds to the telomeric protein TRF1 (telomeric repeat binding factor–1), a negative regulator of telomere length maintenance. Like ankyrins, tankyrase contains 24 ankyrin repeats in a domain responsible for its interaction with TRF1. Recombinant tankyrase was found to have PARP activity in vitro, with both TRF1 and tankyrase functioning as acceptors for adenosine diphosphate (ADP)–ribosylation. ADP-ribosylation of TRF1 diminished its ability to bind to telomeric DNA in vitro, suggesting that telomere function in human cells is regulated by poly(ADP-ribosyl)ation.

Human telomere function requires two telomere-specific DNA binding proteins, TRF1 and TRF2 (1,2). TRF2 protects chromosome ends (3), and TRF1 regulates telomere length (4). Overexpression of TRF1 in a telomerase-expressing cell line leads to progressive telomere shortening, whereas inhibition of TRF1 increases telomere length (4). TRF1 does not control the expression of telomerase itself but is thought to act in cis by inhibiting telomerase at telomere termini.

To identify additional telomere-associated proteins, we used a yeast two-hybrid screen with human TRF1 as bait (5, 6). This screen yielded two overlapping partial cDNAs (TR1L-4 and TR1L-12) (Fig. 1A). A full-length testis cDNA isolated with TR1L-4 encoded an open reading frame of 1327 amino acids, predicting a protein of 142 kD (Fig. 1A) (7). The central domain of this protein contains 24 ankyrin (ANK) repeats, a 33–amino acid motif that mediates protein-protein interactions (8), and its COOH-terminal region has homology to the catalytic domain of PARP, a highly conserved nuclear enzyme found in most eukaryotes (9). We therefore named the protein tankyrase (TRF1-interacting,ankyrin-related ADP- ribose polymerase).

Figure 1

Domain structure of tankyrase and two-hybrid interaction with TRF1. (A) Schematic representation of the structure of tankyrase and TRF1. Lines below the schematic indicate inserts of the two-hybrid plasmids (TR1L-4 and TR1L-12) and a plasmid used to generate recombinant protein for antibody production (ANK2). HPS, region containing homopolymeric runs of His, Pro, and Ser; ANK, ankyrin-related domain; SAM, homology to the sterile alpha motif; PARP, homology to the catalytic domain of PARP; Myb, Myb-type DNA binding motif; D/E, acidic domain rich in Glu and Asp. (B) Two-hybrid assay for the tankyrase interaction domain in TRF1. β-Galactosidase concentrations (Miller units; average of three independent transformations) were measured for strains expressing the indicated fusion proteins (6). GAD, GAL4 activation domain.

The tankyrase-interacting domain in TRF1 was identified by two-hybrid analysis (6) with TR1L-12 (Fig. 1A). The tankyrase fragment, consisting of 10 ANK repeats, interacted with full-length TRF1 but not with a TRF1 mutant lacking the NH2-terminal acidic domain of TRF1 (Fig. 1B). Consistent with this observation, significant interaction occurred with the isolated NH2-terminal 68 amino acids of TRF1, which encompass the acidic domain (Fig. 1B). These results indicate that the acidic domain of TRF1 is necessary and sufficient for interaction with tankyrase. This domain is absent from TRF2, and a two-hybrid analysis (6) indicated that tankyrase does not interact with TRF2.

Three observations suggested that tankyrase was a member of the ankyrin family, a group of structural proteins that link integral membrane proteins to the underlying cytoskeleton (10). First, tankyrase, like all ankyrins, contained 24 copies of the ANK motif, whereas other ANK repeat–containing proteins typically have 4 to 8 repeats. Second, the ANK repeats in tankyrase and the ankyrins shared characteristic sequence features, such as the presence of a hydrophobic amino acid at position 3 and an Asn or a Asp at position 29 (Fig. 2A). Third, the fifth ANK copy in tankyrase was notably shorter than all others, a feature also observed in ankyrins. Apart from the ANK repeat domain, however, there was no detectable homology between tankyrase and ankyrins. The ankyrin domain of tankyrase is flanked at the NH2-terminus by a region carrying homopolymeric His, Ser, and Pro tracts and at the COOH-terminus by a sterile alpha module (SAM) motif (Fig. 2B), which is postulated to function in protein-protein interaction (11).

Figure 2

Conserved domains in tankyrase. (A) Predicted amino acid sequence of tankyrase with an alignment of the 24 ANK repeats. Dashes indicate gaps, and sequences to the right indicate insertions that occur after the underlined amino acid in each line. Light shading indicates a match to the ANK repeat consensus, and darker shading is a match to the ankyrin-specific ANK repeat consensus (8). The SAM motif is doubly underlined, and the PARP domain singly underlined. (B) Alignment of the SAM motifs of human tankyrase (GenBank number AF082556),Drosophila Bicaudal-C (U15928), human diacyl glycerol kinase delta (D73409), and chicken embryo kinase 9 (U23783). (C) Alignment of the PARP-related domain of human tankyrase with aDrosophila expressed sequence tag (EST) LD10141(AA391467), the catalytic domain of human PARP (M32721),Drosophila PARP (D13806), and a PARP-related domain in a human EST KIAA0177 (D79999). The secondary structures underlined are based on the published crystal structure of chicken PARP. β Strands are indicated with c, d, e, f, g, m, and n. L denotes a conserved α helix. Asterisks indicate positions conserved in the prokaryotic ADP-ribosyltransferases, exotoxin A from Pseudomonas aeruginosa, and diphtheria toxin (12).

The most striking feature of tankyrase is the homology to PARP. In response to DNA damage, PARP catalyzes the formation of poly(ADP-ribose) onto a protein acceptor using nicotinamide adenine dinucleotide (NAD+) as a substrate (9). The catalytic domain of PARP consists of secondary structure units (multiple β strands and one α helix) (Fig. 2C) that form a cavity known as the NAD+-binding fold, a tertiary structure that is also present in all ADP-ribosylating toxins (12). Tankyrase has 28 to 30% amino acid identity with the catalytic domains of human and Drosophila PARP (Fig. 2C), including all critical amino acids implicated in NAD+ binding and catalysis. Other conserved aspects of the previously defined PARPs such as their automodification and DNA binding domains (9) are not represented in tankyrase, indicating that tankyrase is not just a PARP isoform but a substantially different protein.

Northern (RNA) blot analysis revealed that multiple tankyrase mRNAs (13) were ubiquitously expressed in human tissues, with the highest amounts detectable in testis (Fig. 3A). TRF1 and TRF2 transcripts show a similar ubiquitous expression pattern (1, 2). A single protein of ∼142 kD was detected by tankyrase immunoblot analysis of HeLa cells and rat testis, and this protein comigrated with the in vitro translation product of tankyrase cDNA (Fig. 3B) (14). A survey of mammalian cell lines suggested that tankyrase protein is ubiquitously expressed (15), consistent with the RNA data.

Figure 3

Expression and localization of tankyrase. (A) Northern blot of polyadenylated RNAs from human tissues (Clontech) probed with a tankyrase cDNA (TR1L-4) (13). Asterisks indicate tankyrase transcripts. The blot was rehybridized with a β-actin probe, and a double exposure of both signals is shown. Molecular size markers are indicated on the left in kilobases. (B) Immunoblot of the following protein samples: salt-extracted nuclear pellet from rat testis (Testis), whole-cell lysates from HeLa cells (HeLa), and products of a coupled in vitro transcription-translation (IVTL) of full-length tankyrase cDNA, probed with the indicated antibodies (14). Molecular size markers are indicated on the left in kilodaltons. (C) Colocalization of tankyrase and TRF1 at telomeres. Indirect immunofluorescence analysis of swollen, formaldehyde-fixed metaphase spreads from HeLa cells stained with anti-tankyrase (green) and anti-TRF1 (red) (16). “Merge” represents superimposition of the red and green images. DAPI staining of DNA is shown in blue. Scale bar, 5 μm.

Because TRF1 is predominantly associated with telomeres in human cells, including the telomeres of mitotic chromosomes, we used indirect immunofluorescence analysis of metaphase chromosomes to determine whether TRF1 positions tankyrase at chromosome ends. Metaphase spreads were dually probed with anti-tankyrase and antiserum to TRF1 (16). The results revealed that, like TRF1, tankyrase is located at or near the physical ends of metaphase chromosomes (Fig. 3C). Most of the tankyrase protein colocalized with TRF1, as evidenced by the merge of the two signals. These data suggest that tankyrase is a component of the human telomeric complex.

To investigate whether tankyrase has PARP activity, we tested baculovirus-derived recombinant protein in an assay that measures the addition of radiolabeled ADP-ribose to protein acceptors with [32P]NAD+ used as a substrate (17). Incubation of tankyrase in the presence of 1.3 μM radiolabeled NAD+ produced 32P-labeled species that comigrated with tankyrase, suggesting that tankyrase has the ability to ADP-ribosylate itself (Fig. 4A). Higher concentrations of NAD+ (0.04 to 1 mM) yielded much larger products, likely reflecting the addition of poly(ADP-ribose) to tankyrase. The generation of ADP-ribosylated tankyrase depended on the concentration of tankyrase (Fig. 4A), was eliminated by heat inactivation of the enzyme, and could be immunoprecipitated with anti-tankyrase (Fig. 4B) (18), indicating that the PARP activity was intrinsic to tankyrase.

Figure 4

Tankyrase is a PARP that inhibits TRF1 in vitro. (A) Tankyrase ADP-ribosylates itself and TRF1. Tankyrase was allowed to modify itself and TRF1 in the presence of [32P]NAD+, and the products were analyzed by Coomassie blue staining (left) and autoradiography (right) of SDS-PAGE gels (17). Reactions contained the proteins indicated above the lanes at the following amounts: TRF1 at 4 μg (+) and tankyrase at 4 μg (+) or at a range of 0, 0.8, and 4 μg (triangle). All reactions contained 1.3 μM [32P]NAD+ (+), and three reactions were also supplemented with increasing amounts of unlabeled NAD+(0.04, 0.2, and 1 mM, triangle). (B) ADP-ribosylation activity is intrinsic to tankyrase. Tankyrase was immunoprecipitated with preimmune serum or anti-tankyrase (α-Tankyrase) as indicated and incubated in a PARP assay with [32P]NAD+. The products were detected by autoradiography (18). (C) Tankyrase is inhibited by the PARP inhibitor 3AB. Reactions containing 4 μg of tankyrase (+), without (−) or with (+) 4 μg of TRF1, and 1.3 μM [32P]NAD+ were incubated without (−) or with (+) 1 mM 3AB and processed as in (A). (D) Tankyrase products contain poly(ADP-ribose). Tankyrase and TRF1 were added as in (C). Reactions for the left panel contained no NAD+ (−) or 1.3 μM [32P]NAD+ supplemented with 1 μM or 1 mM unlabeled NAD+ (triangle). Reactions for the right panel were identical to the reactions on the left but lacked labeled NAD+. Products were transferred to nitrocellulose and analyzed by autoradiography (left) or immunoblotted with monoclonal antibody 10H to poly(ADP-ribose) (17) (right). (E) Tankyrase inhibition of TRF1. A gel-shift assay for the TTAGGG repeat–binding activity of TRF1 was performed with a duplex [TTAGGG]12 DNA as a probe. Binding reactions contained the components indicated above the lanes. Tankyrase concentration was varied from 200 to 2.5 ng per 20-μl incubation in threefold dilution steps (triangle). TRF1 was either present at 13 ng (+) or varied from 120 to 13 ng in threefold dilution steps (triangle). NAD+ was at absent (−) or present at 0.2 mM (+). The asterisks indicate the position of TRF1-containing complexes as determined by antibody super-shift experiments.

Tankyrase also has the ability to modify TRF1. At low NAD+ concentration (1.3 μM), the ADP-ribosylated products comigrated with TRF1, whereas at higher NAD+ concentrations (0.04 to 1 mM), the slower and variable mobility of the labeled products suggested poly(ADP-ribosyl)ation of TRF1 (Fig. 4A). Inspection of Coomassie blue–stained SDS gels did not reveal a larger molecular weight species upon tankyrase-mediated TRF1 modification, indicating that only a small fraction of the TRF1 in the reactions was modified even at high tankyrase concentrations. Thus, tankyrase is likely to function as a processive PARP under these conditions. TRF2 is not a substrate for modification in vitro, as expected from the lack of protein-protein interaction between TRF2 and tankyrase.

To confirm that the labeling reaction with tankyrase was analogous to PARP-catalyzed poly(ADP-ribosyl)ation, we added the specific PARP inhibitor 3-aminobenzamide (3AB) to the reactions (19). Modification of both TRF1 and tankyrase was strongly inhibited by 3AB (Fig. 4C). Furthermore, modified tankyrase and TRF1 reacted with a monoclonal antibody to poly(ADP-ribose) (Fig. 4D) (17), consistent with their carrying ADP- ribose polymers. These data indicate that tankyrase is a genuine PARP with at least two specific substrates, TRF1 and tankyrase itself.

The effect of tankyrase on the telomeric DNA binding activity of TRF1 was determined by an in vitro gel-shift assay with the use of a double-stranded array of [TTAGGG]12 as a probe (20). TRF1 binds to DNA as a homodimer, and several such dimers can occupy one [TTAGGG]12 molecule at high TRF1 concentrations (6) (Fig. 4E). When TRF1 was incubated with baculovirus-derived tankyrase in the absence of NAD+, a slight stimulation of the TRF1 DNA binding activity occurred, resulting in the formation of higher order complexes, especially at high tankyrase concentrations. However, this stimulation of TRF1 also occurred with total insect cell protein and was therefore unlikely to represent a specific effect of tankyrase. A similar nonspecific enhancement of TRF1 was previously reported for β-casein and several other proteins (1). In contrast, when NAD+ was included in the TRF1-tankyrase mixtures, a reduction of the TRF1 activity resulted (Fig. 4E). This effect was dependent on the addition of active tankyrase (Fig. 4E), consistent with ADP-ribosylation being the cause of the TRF1 inhibition.

The identification of a telomeric PARP raises the possibility that the function of human telomeres is regulated by this type of protein modification. Because ADP-ribosylation usually inhibits protein activity (21), we favor the view that tankyrase is a negative regulator of another factor acting at telomeres. Although the in vivo targets of tankyrase remain to be established, TRF1 is a strong candidate, because it is a substrate for tankyrase in vitro and ADP-ribosylation inhibits the ability of TRF1 to bind to telomeric DNA. However, the PARP activity of tankyrase could also be directed at other telomere-associated factors, including telomerase, and ADP-ribosylation might enhance rather than inhibit the activity of the target protein (22). In vivo functional analysis will be required to determine whether tankyrase acts positively or negatively in the regulation of telomere length. PARPs have previously been implicated in the cellular response to DNA damage (9). The presence of a PARP activity at telomeres may also hint at a role for tankyrase in the protection of telomeres from inappropriate DNA damage processing activities.

  • * Present address: European Molecular Biology Laboratory–Heidelberg, Meyerhofstrasse 1, D-69117, Heidelberg, Germany.

  • To whom correspondence should be addressed. E-mail: delange{at}


View Abstract

Stay Connected to Science

Navigate This Article