Tankyrase 2 Poly(ADP-Ribose) Polymerase Domain-Deleted Mice Exhibit Growth Defects but Have Normal Telomere Length and Capping

ABSTRACT Regulation of telomere length maintenance and capping are a critical cell functions in both normal and tumor cells. Tankyrase 2 (Tnks2) is a poly(ADP-ribose) polymerase (PARP) that has been shown to modify itself and TRF1, a telomere-binding protein. We show here by overexpression studies that tankyrase 2, like its closely related homolog tankyrase 1, can function as a positive regulator of telomere length in human cells, dependent on its catalytic PARP activity. To study the role of Tnks2 in vivo, we generated mice with the Tnks2 PARP domain deleted. These mice are viable and fertile but display a growth retardation phenotype. Telomere analysis by quantitative fluorescence in situ hybridization (FISH), flow-FISH, and restriction fragment analysis showed no change in telomere length or telomere capping in these mice. To determine the requirement for Tnks2 in long-term maintenance of telomeres, we generated embryonic stem cells with the Tnks2 PARP domain deleted and observed no change, even upon prolonged growth, in telomere length or telomere capping. Together, these results suggest that Tnks2 has a role in normal growth and development but is not essential for telomere length maintenance or telomere capping in mice.

Regulation of telomere length and protection of chromosome ends are two critical telomere functions that are essential in preventing premature senescence and in maintaining genome stability (reviewed in references 22 and 41). Mammalian telomeres consist of TTAGGG repeats that are bound by telomeric DNA repeat binding proteins and their associated factors, which together regulate telomere length maintenance and chromosome end protection (reviewed in references 17 and 56). TRF1, a double-stranded telomere repeat binding protein (12), functions as a negative regulator of telomere length by acting in cis to control access of telomerase (2,58), a reverse transcriptase that uses an RNA template to add telomeric repeats (23,24). Tankyrase 1 (Tnks1), which was identified in a two-hybrid screen with TRF1 (55), acts as a positive regulator of telomere length (54). Tankyrase 1 is comprised of several domains: the amino-terminal HPS domain, which consists of homopolymeric tracts of histidine, proline, and serine repeats; the large ANK domain, which is made up of 24 ankyrin repeats and comprises five functional subdomains (19,50,51); the sterile alpha module (SAM) domain, which is involved in tankyrase multimerization (18,19,50); and a poly(ADPribose) polymerase (PARP) domain (PD). The PARP domain of tankyrase 1 places this protein in the superfamily of PARP proteins.
PARPs utilize NAD ϩ as a substrate to synthesize long linear or branched polymers of ADP-ribose on protein acceptors (reviewed in references 1 and 52). Tankyrase 1 poly(ADP-ribosyl)ates TRF1 in vitro, inhibiting its ability to bind to telomeric DNA (55). Upon overexpression of tankyrase 1 in the nucleus, TRF1 is removed from telomeres (13,54). The DNA-unbound form of TRF1 is ubiquitinated and degraded by the proteasome (10). Long-term overexpression of tankyrase 1 in human cells results in loss of TRF1 and telomere elongation, dependent on the catalytic PARP activity of tankyrase 1 (13,54).
Another function of tankyrase 1 was revealed by knockdown of tankyrase 1 expression. Tankyrase 1 short interfering RNA (siRNA) treatment in human cells resulted in mitotic arrest with aberrant chromosome configurations and abnormal spindle structures (20). Sister chromatids were able to separate at centromeres and arms but were unable to separate at their telomeres. Use of siRNA-resistant wild-type (WT), but not PARP-dead, tankyrase 1 rescued this phenotype, indicating a requirement for tankyrase 1 PARP activity in sister telomere resolution and mitotic progression (20). More recent studies characterized spindle defects in tankyrase 1 siRNA cells and found defects in bipolar spindle formation and supernumerary spindles (8).
In addition to its telomeric localization (via TRF1 binding), tankyrase 1 localizes to other subcellular sites, including mitotic centrosomes (53) and the Golgi apparatus (11), and has multiple binding partners. Tankyrase 1 has been shown to interact with IRAP (insulin-responsive aminopeptidase) in GLUT4 vesicles in the Golgi apparatus and further to be phosphorylated by mitogen-activated protein kinase upon insulin stimulation, suggesting a role for tankyrase 1 in mitogenactivated protein kinase-dependent regulation of GLUT4 vesicles (11). Tankyrase 1 has also been found to interact with and inhibit the Mcl-1 (myeloid cell leukemia 1) proteins, which function to regulate apoptosis (3). In addition, interactions of tankyrase 1 with FBP-17 (formin-binding protein 17) (21) and with TAB182 (tankyrase-binding protein of 182 kDa), a heterochromatin-and cortical actin-staining protein (51), have been observed. Finally, tankyrase 1 has been found to bind to and colocalize with NuMA (nuclear and mitotic apparatus protein) at mitotic centrosomes (49,53). Recent studies indicate that NuMA is a major acceptor of poly(ADP-ribosyl)ation by tankyrase 1 in mitosis (8,9). Tankyrase 2 (Tnks2; a closely related homolog) was identified by several groups through two-hybrid screens with IRAP (11); Grb14, an SH2 domain-containing adaptor protein that binds to the insulin receptor (IR) (40); and TRF1 (29), as well as by serological screens using a fetal brain cDNA library with meningioma patient sera (42) or using a breast tumor cDNA library with breast carcinoma patient serum (36). Studies indicate that there is twofold more tankyrase 1 than tankyrase 2 in DT40 chicken cells (19). The domains that comprise tankyrase 2 are highly homologous to those of tankyrase 1, with the main exception being that tankyrase 2 lacks the HPS domain. The tankyrase 2 ANK domain shares 83% identity with that of tankyrase 1, and as the ANK domain is responsible for proteinprotein interactions, tankyrase 2 interacts with many of the same proteins as tankyrase 1, including TRF1 (13,29), TAB182 (51), NuMA (49), and IRAP (11,50). The tankyrase 2 SAM domain shares 74% identity with that of tankyrase 1, allowing both selfoligomerization and oligomerization with tankyrase 1 (18,19,50). The PARP domain is the most conserved, with 94% identity. Tankyrase 2 has been shown to have PARP activity similar to that of tankyrase 1 (13,50) and can poly(ADP-ribosyl)ate itself and TRF1, releasing TRF1 from telomeres (13).
To further elucidate the role of tankyrase 2 at telomeres, we generated stable human cell lines overexpressing tankyrase 2 in the nucleus. We show that tankyrase 2 induces telomere elongation, dependent on its catalytic PARP activity. Our results confirm that, as expected from the high degree of homology between the two proteins, tankyrase 2 (like tankyrase 1) can influence telomere length in human cells. Next, to study the function of tankyrase 2 on telomere length in vivo, we generated a tankyrase 2 PARP domain-deleted mouse. Characterization of this mouse revealed a small mouse phenotype, and telomere analysis showed no change in telomere length and no defect in telomere capping. Thus, in mice, a PARP-active Tnks2 is not required for telomere maintenance but is necessary for normal growth.

MATERIALS AND METHODS
Plasmids. Full-length tankyrase 2 (amino acids 2 to 1166) with an aminoterminal Myc tag and nuclear localization sequence (MN-TNKS2.WT) was cloned as described previously (13). MN-TNKS2.HE/A was generated by replacing the conserved active site histidine and glutamic acid residues at positions 1031 and 1138, respectively, with alanine residues by site-directed mutagenesis of MN-TNKS2.WT using oligonucleotides 5Ј-GCCAATGAACGAATGCTATTT GCTGGGTCTCCTTTTGTGAATG-3Ј for H1031 and 5Ј-GGCCTAGCATTA GCTGCATATGTTATTTACAGAGGAGAAC-3Ј for E1138, with the Stratagene QuickChange site-directed mutagenesis kit. MN-TNKS2.WT and MN-TNKS2.HE/A were cloned into the pLPC vector, and HTC75 retroviral cell lines were generated as previously described (13).
Antibody generation. Anti-tankyrase 2 antibody 608 was raised and affinity purified against an Escherichia coli-expressed fusion protein containing human tankyrase 2 amino acids 811 to 896.
Construction of mTnks2 targeting vector. The targeting vector was designed to replace exons 25, 26, and 27 (PARP domain of Tnks2) with a neomycin (neo) resistance cassette. (Note that the mTnks2 exon nomenclature that is used in this paper follows that described in reference 29. The predicted exon structure for mTnks2 in the GenBank and Mouse Genome Informatics databases includes one additional exon upstream of exon 1.) The floxed phosphoglycerate kinase-neo cassette from the pPNTloxPneo vector (kindly provided by Alexandra Joyner, Skirball Institute) was cloned into pSP72 (from Promega, with a NotI site added to the P6 site). The 1.5-kb short arm was generated by PCR from mouse 129/SvEv genomic DNA using primers 5Ј-GTGGGAAAGATATACACACCG GAG-3Ј and 5Ј-GTGCACATCAGTCAAGCACTGAG-3Ј. The short arm began within exon 24 and extended to 90 bp upstream of exon 25 and was inserted 5Ј of the neo cassette. The long arm, which began at the XbaI site downstream of exon 27 and extended 7.5 kb, was generated from a phage clone isolated from a mouse 129/SvEv genomic library and was inserted 3Ј of the neo cassette.
Targeted disruption of the mTnks2 gene in ES cells. Ten micrograms of the targeting vector was linearized with NotI and electroporated into 129/SvEv ES cells. After selection in G418 (200 g/ml), surviving colonies were expanded, and PCR analysis was performed to identify clones that had undergone homologous recombination. To detect the wild-type allele, the following primer pair was used: primer 1, 5Ј-GTGGGAAAGATATACACACCGGAG-3Ј, and primer 2, 5Ј-AAA TTCCAGCTCCAAACATGCCAC-3Ј. For the mutant allele, the following primer pair was used: primer 3, 5Ј-GACAACTTACGGCAGTTGGTTCTG-3Ј, and primer 4, 5Ј-TGCGAGGCCAGAGGCCACTTGTGTAGC-3Ј. The PCR conditions were 94°C for 5 min; followed by 35 cycles of 94°C for 30 s, 62°C for 1 min, and 72°C for 2 min; and a final extension at 72°C for 7 min.
Generation of Tnks2 PD⌬ mice. Tnks2 PD⌬ heterozygous (ϩ/Ϫ) ES cell clones were microinjected into C57BL/6 blastocysts, and chimeric males were mated to C57BL/6 females for germ line transmission of the targeted allele. Genotyping was performed using the PCR primers above, and correct targeting was confirmed by Southern blotting. Homologous recombination of the targeting vector with the Tnks2 gene resulted in the introduction of a ScaI site, which was used to distinguish the 4.3-kb fragment in the wild-type allele from the 5.1-kb fragment in the mutant allele, by a 0.5-kb 5Ј-flanking probe. The probe was generated by PCR using the following primers: 5Ј-TTGTTCATACCAGCTATGGTT CTTG-3Ј and 5Ј-GTGTCTGTCTGAGTTCATACAC-3Ј.
Mouse breeding and housing. Tnks2 PD⌬ mice were maintained on a mixed 129/SvEv-C57BL/6 background; Tnks2 PD⌬neo⌬ mice were backcrossed twice to C57BL/6 mice. Mice were housed in a specific-pathogen-free facility, in accordance with institutional animal care guidelines.
Northern blotting. Total RNA was prepared from MEFs using the RNeasy Mini kit (QIAGEN). Blots were probed with DNA probes corresponding to amino acids 811 to 896 for human tankyrase 2 and 336 to 1163 for human tankyrase 1.
Telomere restriction fragment analysis. Genomic DNA was isolated from stable HTC75 cell lines expressing vector, MN-TNKS2.WT or MN-TNKS2.HE/A, and digested with HinfI and RsaI. Telomere restriction fragments were detected by Southern blotting with a 32 P-labeled TTAGGG probe, as previously described VOL. 26,2006 TANKYRASE 2 MUTANT MICE HAVE NORMAL TELOMERE LENGTH 2045 (13). The mean length of telomeric restriction fragments was determined using TELO software (Fox Chase Cancer Center). For primary or immortalized MEFs of the indicated genotypes, cells were embedded in agarose plugs and digested with HinfI and RsaI, DNA fragments were separated by pulsed-field gel electrophoresis (CHEF DR-II apparatus, Bio-Rad), and telomere restriction fragments were analyzed, as described previously (20).

Quantitative fluorescence in situ hybridization (Q-FISH).
Spleens from wildtype and Tnks2 PD⌬ littermate mice, or wild-type and Tnks2 PD⌬neo⌬ littermate mice, were used to prepare single-cell suspensions by mincing samples through a 70-m nylon cell strainer (BD Biosciences). Splenocytes were cultured in DMEM-10% FBS and activated with 1 g/ml concanavalin A (Sigma) and 1 g/ml phytohemagglutinin (Sigma) for 36 h before cells were arrested in metaphase with a 6-h treatment with 0.1 g/ml Colcemid (Gibco). Cells were collected, and metaphase spreads were prepared from splenocytes as described previously (20) and from ES cells as described previously (61). Images were acquired on a Zeiss Axioplan 2 microscope with a Photometrix SenSyn camera. Photographs were processed using OpenLab software. TFL-TELO software (kindly provided by P. Lansdorp, Terry Fox Laboratory, Vancouver, Canada) was used to quantify the telomere signal from 10 metaphases for each sample.
Flow-FISH. Spleen and thymus from wild-type and Tnks2 PD⌬ littermate mice or wild-type and Tnks2 PD⌬neo⌬ littermate mice were used to prepare single-cell suspensions of splenocytes and thymocytes by mincing samples through a 70-m nylon cell strainer. Erythrocytes were removed from the cell suspension by lysis with ACK lysing buffer, and samples were frozen at Ϫ80°C until processed for flow-FISH. Frozen samples were thawed and washed two times in DMEM. Viable cells were counted and resuspended in phosphate-buffered saline-0.1% bovine serum albumin (BSA) to the concentration of 1 ϫ 10 6 cells/ml. Cells were pelleted and resuspended in 1 ml of hybridization solution [70% formamide, 20 mM Tris, pH 7.5, 20 mM NaCl, 1% BSA, 0.3 g/ml fluorescein isothiocyanateconjugated (TTAGGG) 3 peptide nucleic acid (PNA) probe (Applied Biosystems)]. DNA was denatured at 86°C for 10 min, and the samples were hybridized for 2 h at room temperature. Cells were washed four times in wash solution 1 (70% formamide, 10 mM Tris, pH 7.5, 1% BSA, 1% Tween 20) and one time in wash solution 2 (5% glucose, 10 mM HEPES, 1% BSA, 1% Tween 20). For DNA counterstaining, cells were resuspended in 500 l staining solution (phosphatebuffered saline, 10 U/ml RNase A, 0.1% BSA, 0.1 g/ml propidium iodide) for 2 h before acquisition on the FACScan (Becton Dickinson) flow cytometer. Mean telomere fluorescence of cells gated at G 0 /G 1 was analyzed with CellQuest software.
Scoring of chromosomal abnormalities. The indicated number of metaphase images for each sample was scored for chromosomal abnormalities (end-to-end fusions, chromosome breakages, and signal free ends) by overlaying the telomere image on the chromosome image using Photoshop software.

RESULTS
Human tankyrase 2 is a positive regulator of telomere length. Previous studies indicated that tankyrase 2 shared a number of properties with tankyrase 1. Both were found to bind TRF1 and poly(ADP-ribosyl)ate TRF1 in vitro (13,29,55). Overexpression studies of tankyrase 1 showed that when it was expressed in the nucleus (via an amino-terminal nuclear localization signal) it removed TRF1 from telomeres (whereupon TRF1 was degraded by the proteasome) and, over the long term, induced telomerase-dependent telomere elongation (10,13,54). The activities of tankyrase 1 were dependent on its catalytic PARP domain, since a catalytically dead allele (tankyrase 1.HE/A) had no effect on TRF1 or telomere length maintenance (13). We showed previously that overexpression of a Myc epitope-tagged, nuclear localization signal-containing tankyrase 2 (MN-tankyrase 2) resulted in loss of TRF1 from telomeres (13). To determine if tankyrase 2 (like tankyrase 1) was capable of inducing telomere elongation, we generated stable HTC75 cell lines expressing MN-tankyrase 2 WT or a PARP-dead allele (HE/A) (generated in the same way as tankyrase 1.HE/A) by converting the highly conserved active site residues, histidine and glutamic acid at amino acid positions 1031 and 1138, respectively, to alanine residues by site-directed mutagenesis. As shown in Fig. 1A, overexpression of wild-type (lane 2) but not PARP-dead (lane 3) tankyrase 2 resulted in reduced levels of TRF1. Telomere length analysis showed that MN-tankyrase 2.WT induced telomere elongation (Fig. 1B). Telomeres showed progressive elongation at a rate of approximately 38 bp per population doubling (pd) (Fig. 1C), similar to tankyrase 1 (13,54). By contrast, overexpression of MN-tankyrase 2.HE/A had no effect on telomere elongation ( Fig. 1B and C) and was similar to a vector control (data not shown). Together these studies indicate that tankyrase 2 (like tankyrase 1) can act as a positive regulator of telomere length in human cells dependent on its catalytic PARP activity.
To determine if tankyrase 2 is required to maintain telomere length in vivo, we sought to delete the tankyrase 2 gene (Tnks2) in mouse. In order to monitor tankyrase 2 protein expression, we raised tankyrase 2-specific antibodies against an E. coli-expressed fusion protein containing human tankyrase 2 amino acids 811 to 896 (Fig. 1D). The specificity of the antibody was determined by immunoblotting against baculovirus-expressed human tankyrase 1 or tankyrase 2. As shown in Fig. 1E, anti-tankyrase 2 608 specifically detected tankyrase 2 (lane 4) but not tankyrase 1 (lane 3). By contrast, a previously generated antibody to tankyrase 1 (tankyrase 1 465) (55) Fig. 2A). We sought to disrupt the Tnks2 gene and create a null allele by deleting exons 25 through 27 (encoding carboxy-terminal amino acids 1032 through 1166 of mTnks2) and replacing them with the neomycin resistance gene (neo). However, as described below, the targeted disruption did not result in a null allele but rather in expression of truncated forms of Tnks2. The truncated proteins lack the catalytic PARP domain; hence, the targeted allele will be referred to as Tnks2 PD⌬ (PARP domain deleted) ( Fig. 2A).
Mouse ES cells (129/SvEv) were electroporated with the linearized targeting construct and selected for G418 resistance. Clones were screened using the PCR strategy described below. Chimeric mice (generated from injection of a positive ES cell clone into a blastocyst) were mated to C57BL/6 for germ line transmission to generate Tnks2 PD⌬ϩ/Ϫ mice.
Mating of Tnks2 PD⌬ϩ/Ϫ mice generated Tnks2 PD⌬Ϫ/Ϫ mice, as shown by Southern blotting and PCR analysis of tail DNA. For Southern blot analysis, to monitor the targeting of the Tnks2 locus, we made use of a ScaI site in the neo gene, which generates a 5.1-kb ScaI fragment that can be distinguished from the 4.3-kb ScaI fragment in the wild-type allele ( Fig. 2A  and B). The genotype of the mice was further confirmed using PCR analysis ( Fig. 2A and C). For the wild-type allele, PCR was performed with a 5Ј primer in exon 24 and a 3Ј primer in exon 25, which resulted in a 1.7-kb fragment specific to the wild-type allele. For the targeted allele PCR was performed with a 5Ј primer upstream of exon 24 (not found in the targeting vector) and a 3Ј primer in the neo gene, which resulted in a 2.0-kb fragment specific to the targeted allele.
Consistent with targeting of the Tnks2 locus, Northern blot analysis of primary MEFs showed no detectable Tnks2 RNA 2046 HSIAO ET AL. MOL. CELL. BIOL. (Fig. 2D). Moreover, we did not detect any major new transcripts in Tnks2 PD⌬ϩ/Ϫ or Tnks2 PD⌬Ϫ/Ϫ cells. As expected, immunoblot analysis of immortalized MEFs indicated that wild-type (fulllength) Tnks2 was not expressed in Tnks2 PD⌬Ϫ/Ϫ cells (Fig. 2E). However, alternative Tnks2 proteins (which were not found in wild-type cells) were expressed in Tnks2 PD⌬ϩ/Ϫ and Tnks2 PD⌬Ϫ/Ϫ cells, indicating that deletion of exons 25 through 27 did not result in complete ablation of Tnks2 protein expression. Our failure to detect alternative transcripts by Northern blotting could be due to their lower abundance or stability. We observed two classes of alternative Tnks2 proteins in Tnks2 PD⌬ϩ/Ϫ and Tnks2 PD⌬Ϫ/Ϫ cells, one migrating faster and the other slower than wild-type Tnks2 protein (Fig. 2E). The faster-migrating proteins likely corresponded to truncated Tnks2 proteins. The targeted Tnks2 allele lacks the carboxyterminal 134 amino acids. Thus, a truncated protein would be predicted to migrate with an apparent molecular mass of 114 kDa (compared to 128 kDa for wild-type Tnks2), consistent with the protein analysis in Fig. 2E. The truncated Tnks2 proteins likely contain the ANK and SAM domains, which are important for scaffolding functions and protein-protein interactions. However, the alternative Tnks2 proteins lack the catalytic PARP domain and are therefore PARP dead.
We reasoned that the slower-migrating class of proteins may have resulted from splicing of a truncated Tnks2 to downstream sequences, in particular neo. Other groups have re- ported generating neo fusion proteins when attempting targeted gene disruption with the neo cassette and have noted the presence of cryptic splice sites within the cassette (7, 28). As the neo gene in the targeted allele was flanked by loxP sites (Fig. 2A), we addressed this possibility by deleting the neo gene from the targeted allele using Cre recombinase. Thus, we crossed Tnks2 PD⌬Ϫ/Ϫ mice to C57BL/6 mice carrying a TK-Cre transgene.
The Tnks2 PD⌬neo⌬ϩ/Ϫ TK-Cre mice were backcrossed to C57BL/6 mice to remove TK-Cre. These Tnks2 PD⌬neo⌬ϩ/Ϫ mice were backcrossed again to C57BL/6 mice before breeding heterozygous mice to generate Tnks2 PD⌬neo⌬Ϫ/Ϫ mice. To monitor excision of the neo allele, we took advantage of replacement of the ScaI site in the neo gene with a downstream ScaI site which resulted in a 5.9-kb ScaI fragment, distinguishing the PD⌬neo⌬ allele from the wild-type allele ( Fig. 2A and F). The genotype of the mice was further confirmed using PCR analysis ( Fig. 2A and G). For the wild-type allele, PCR was performed (as described above) with a 5Ј primer in exon 24 and a 3Ј primer in exon 25, which resulted in a 1.7-kb fragment specific to the wild-type allele. For the neodeleted allele PCR was performed with a 5Ј primer upstream of exon 24 and a 3Ј primer just 3Ј to the remaining loxP site, which resulted in a 1.72-kb fragment specific to the neo-deleted allele. As shown by immunoblot analysis of brain tissue, the high-molecular-weight protein was no longer expressed in the Tnks2 PD⌬neo⌬Ϫ/Ϫ mice; only the low-molecular-weight, truncated products remained (Fig. 2H). Small size in Tnks2 PD⌬؊/؊ and Tnks2 PD⌬neo⌬؊/؊ mice. As described above, mating of Tnks2 PD⌬ϩ/Ϫ mice generated Tnks2 PD⌬Ϫ/Ϫ mice at the expected Mendelian ratio (37 Tnks2 PD⌬Ϫ/Ϫ mice, 69 Tnks2 PD⌬ϩ/Ϫ mice, and 35 Tnks2 PD⌬ϩ/ϩ mice). Tnks2 PD⌬Ϫ/Ϫ mice (both male and female) were viable and fertile. However, a small but significant fraction died within 20 days of birth (7/37 [19%] Tnks2 PD⌬Ϫ/Ϫ mice, 1/69 [1%] Tnks2 PD⌬ϩ/Ϫ mice, and 2/35 [6%] Tnks2 PD⌬ϩ/ϩ mice). The most striking feature of the Tnks2 PD⌬Ϫ/Ϫ mice was their smaller body size (Fig. 3A). We monitored the weight of the mice from birth to 8 weeks. As shown in Fig. 3B, Tnks2 PD⌬Ϫ/Ϫ pups showed a weight similar to that of their wild-type and Tnks2 PD⌬ϩ/Ϫ littermates 2 and 4 days after birth. However, by 6 days Tnks2 PD⌬Ϫ/Ϫ pups weighed less than their littermates. The difference was most dramatic at 20 days (Tnks2 PD⌬Ϫ/Ϫ mice were 40% smaller than Tnks2 PD⌬ϩ/Ϫ and Tnks2 PD⌬ϩ/ϩ mice) but decreased over time (at 56 days Tnks2 PD⌬Ϫ/Ϫ mice were 20% smaller than Tnks2 PD⌬ϩ/Ϫ and Tnks2 PD⌬ϩ/ϩ mice). The weights of heart, brain, lungs, testis, kidneys, and liver were proportionately reduced in size, but thymus and spleen were variable (Fig. 3C). We also observed a few cases of severely small Tnks2 PD⌬Ϫ/Ϫ mice that were 50 to 60% smaller than Tnks2 PD⌬ϩ/Ϫ and Tnks2 PD⌬ϩ/ϩ mice; however, these mice tended to die before weaning and were not included in the growth curve or organ size analysis.
Tnks2 PD⌬neo⌬Ϫ/Ϫ mice exhibited a similar small mouse phenotype ( Fig. 3D and E), indicating that the phenotype was consistent in both the Tnks2 PD⌬Ϫ/Ϫ and Tnks2 PD⌬neo⌬Ϫ/Ϫ mice. However, compared to the Tnks2 PD⌬ mice, we noticed an increase in death before 20   small Tnks2 PD⌬neo⌬Ϫ/Ϫ mice, suggesting that the Tnks2 PD⌬neo⌬ allele has a more severe effect on growth than does the Tnks2 PD⌬ allele. Normal telomere length in Tnks2 PD⌬؊/؊ and Tnks2 PD⌬neo⌬؊/؊ mice. We determined whether deletion of the PARP domain of Tnks2 affected the length of the telomeric repeats. First, we used flow-FISH, which measures the total telomere DNA content in a cell population (4,47). We determined the average telomere fluorescence (expressed in arbitrary units) in splenocytes and thymocytes from six sets of littermate Tnks2 PD⌬ϩ/ϩ versus Tnks2 PD⌬Ϫ/Ϫ mice. As shown in Table 1, there was no significant difference between Tnks2 PD⌬ϩ/ϩ and Tnks2 PD⌬Ϫ/Ϫ mice. To analyze telomere length distributions, we performed Q-FISH (45,62), a technique that measures the telomeric DNA signal at each chromosome end in individual metaphase cells (Fig. 4A). As shown in Fig. 4B, the telomere length distribution was similar in Tnks2 PD⌬ϩ/ϩ and Tnks2 PD⌬Ϫ/Ϫ splenocytes. As an alternative to fluorescence measurements, we used telomere restriction fragment analysis to measure telomere length. As shown in Fig. 4C, telomere restriction fragment analysis of primary and immortalized MEFs from Tnks2 PD⌬ϩ/ϩ , Tnks2 PD⌬ϩ/Ϫ , and Tnks2 PD⌬Ϫ/Ϫ mice showed no differences in telomere length.
As shown in Table 1 there was no significant difference between Tnks2 PD⌬neo⌬ϩ/ϩ (or ϩ/Ϫ) and Tnks2 PD⌬neo⌬Ϫ/Ϫ mice. In agreement, Q-FISH analysis (Fig. 4D) indicated that telomere length distribution was similar in Tnks2 PD⌬neo⌬ϩ/ϩ and Tnks2 PD⌬neo⌬Ϫ/Ϫ splenocytes. Together these data indicate that a catalytically PARP-active Tnks2 is not required for telomere length maintenance in mice.
Normal telomere capping in Tnks2 PD⌬ and Tnks2 PD⌬neo⌬ mice. To study the effect of the Tnks2 PARP domain deletion on telomere function, we analyzed 100 metaphase cells prepared from splenocytes from Tnks2 PD⌬ϩ/ϩ and Tnks2 PD⌬Ϫ/Ϫ mice by Q-FISH for chromosomal aberrations. We did not detect any change in signal free ends, end-to-end fusions, or chromosome breakage (Table 2). Similar analysis of splenocytes from Tnks2 PD⌬neo⌬ϩ/ϩ and Tnks2 PD⌬neo⌬Ϫ/Ϫ mice showed no chromosomal abnormalities ( Table 2).
Telomere length is maintained during prolonged culture of Tnks2 PD⌬؊/؊ ES cells. Previous studies have shown that, in ES cells deficient in telomerase, telomeres undergo shortening and show an increase in end-to-end fusions (37,38,43,44). This phenotype is more apparent as ES cells are continually passaged. To determine if a deficiency in Tnks2 might influence telomere function during long-term growth, we generated Tnks2 PD⌬Ϫ/Ϫ ES cell clones from a G418-resistant Tnks2 PD⌬ϩ/Ϫ ES cell clone by culture at increased G418 concentration. Clones were screened using the same PCR strategy described for Fig. 2C, and two Tnks2 PD⌬Ϫ/Ϫ ES cell clones (A3 and C10) were isolated (Fig. 5A). Immunoblot analysis indicated that (like Tnks2 PD⌬Ϫ/Ϫ mice) Tnks2 PD⌬Ϫ/Ϫ ES cells were deficient in wild-type Tnks2 protein but did express truncated (PARP-deleted) Tnks2 alone and fused to neo (Fig. 5B). We used Q-FISH to examine telomeres in Tnks2 PD⌬ϩ/ϩ versus Tnks2 PD⌬Ϫ/Ϫ ES cell clones. As shown in Fig. 5C and D we did not detect telomere shortening in Tnks2 PD⌬Ϫ/Ϫ ES cell clone A3 or C10 upon prolonged cell culture (from pd 42 to pd 212). In fact, the Tnks2 PD⌬ϩ/ϩ and the Tnks2 PD⌬Ϫ/Ϫ ES cell clones showed a slight increase in telomere length during long-term growth, as has been reported previously for wild-type ES cells (37). Note that Tnks2 PD⌬Ϫ/Ϫ ES cell clones had longer telomeres than Tnks2 PD⌬ϩ/ϩ ES cells even at pd 42. This is likely due to the additional pd's that the Tnks2 PD⌬Ϫ/Ϫ ES cell clones underwent during G418 selection. Regardless, all three lines showed a similar increase in telomere length over 170 pd's (WT, 28%; A3, 29%; C10, 26%), indicating that the Tnks2 PARP domain deletion had no effect on telomere length maintenance during prolonged cell culture.
To study the effect of the Tnks2 PARP domain deletion on telomere capping during prolonged cell culture, we analyzed 50 metaphase cells prepared from Tnks2 PD⌬ϩ/ϩ and Tnks2 PD⌬Ϫ/Ϫ ES cell cultures at 42 versus 212 pd's for chromosomal aberrations. We did not detect any change in signal free ends, end-to-end fusions, or chromosome breakage (Table 3).

Tankyrase 2 and telomere function in mice.
We have shown here that tankyrase 2 (like its closely related homolog tankyrase 1) can act as a positive regulator of telomere length when overexpressed in human cells (Fig. 1A to C). Telomere elongation by tankyrase 2 was found to be dependent on its catalytic PARP  activity. To determine if tankyrase 2 was required for telomere maintenance in vivo, we analyzed telomere function in mice containing a PARP domain-deleted Tnks2. Telomere length analysis using three different methods (Q-FISH, flow-FISH, and telomere restriction fragment analysis) revealed no change in telomere length in Tnks2 PARP domain-deleted mice ( Fig. 4; Table 1). Moreover, Tnks2 PARP domain-deleted ES cells, which were grown for 212 pd's, showed no telomere shortening (Fig. 5). Together these data indicate that a PARP-active Tnks2 is not required for telomere length maintenance in vivo in mice and during long-term growth of mouse cells.
How can we reconcile the lack of requirement for PARPactive tankyrase 2 in telomere length maintenance in mice with    PARP-1, the first identified member of the PARP family, is a DNA damage sensor and has also been implicated in many other cellular processes (reviewed in references 1 and 35). One report of PARP-1 Ϫ/Ϫ mice showed a 30% decrease in telomere length and end-to-end fusions in 25% of cells (15), whereas another group using a different PARP-1 Ϫ/Ϫ mouse found normal telomere length and chromosomal end capping (48). PARP-2, a homolog of PARP-1, has also been reported to play a role at telomeres. PARP-2 was shown to bind and poly(ADPribosyl)ate TRF2 (16), a telomeric DNA binding protein that has a protective function (59). PARP-2 Ϫ/Ϫ cells showed normal telomere length, but an increase in chromosome/chromatid breaks and signal free ends was observed (16). Another PARP with a potential telomeric function is vault PARP (VPARP) (34). VPARP is a component of the vault complex, a large (13-MDa), highly conserved, and ubiquitously expressed ribonucleoprotein complex (reviewed in reference 60). Other components of the vault complex include major vault protein, vault RNA, and telomerase-associated protein 1 (TEP1), which binds telomerase RNA (25). Studies of VPARP Ϫ/Ϫ mice and TEP1 Ϫ/Ϫ VPARP Ϫ/Ϫ mice, however, showed normal telomere length and structure (39). While it is possible that PARP-1, PARP-2, and VPARP could have functions redundant with tankyrases 1 and 2 in maintaining normal telomere length and capping of telomeres, we think it unlikely, since the only homology between these three PARPs and tankyrases 1 and 2 is in the PARP catalytic domain. Tankyrase 2 and postnatal growth in mice. Tnks2 PARP domain-deleted mice are generated at the expected Mendelian ratios, suggesting that Tnks2 PARP activity is not required for embryonic development. Indeed, Tnks2 PARP domain-deleted neonates are indistinguishable from WT littermates from birth until postnatal day 6. However, after day 6 growth retardation becomes apparent (Fig. 3B). The Tnks2 PARP domaindeleted mice were, on average, 20% smaller than WT or heterozygous littermates (Fig. 3A, B, and D). While we do not know the mechanism for this growth defect, we speculate (based on known tankyrase 2 binding partners) that tankyrase 2 could play a role in insulin-mediated effects on postnatal growth (see below).
Thus, tankyrase 2 has been show to bind to Grb14 (40), a member of the Grb7/10/14 family of adaptors (reviewed in reference 27). Grb14 binds activated IR (30) and inhibits the phosphorylation of IR substrates (6). Overexpression of Grb14 has been shown to inhibit the activation of downstream insulin signaling cascades (6,26,30). Consistent with the role of Grb14 as an inhibitor of IR signaling, Grb14 Ϫ/Ϫ mice display improved glucose tolerance, increased insulin sensitivity, and enhanced insulin signaling (14). Interestingly, Grb14 Ϫ/Ϫ mice also exhibit decreased body mass (5 to 10%), compared to wild-type littermates (14), although this is less than the 20% that we observed in our Tnks2 PARP domain-deleted mice (Fig. 3A, B, and D). Another difference in the Grb14 Ϫ/Ϫ mice is that the reduced body mass is not uniform in all tissues. Liver and spleen were significantly reduced (16% and 29%, respectively), while the heart size increased (18%) (14). This is in contrast to our Tnks2 PARP domain-deleted mice, where organ size was proportional to small body size ( Fig. 3C and E). Therefore, the Tnks2-Grb14 interaction may not completely account for the small mouse phenotype.
In addition to Grb14, IRAP is another protein in the insulin pathway that has been shown to bind to tankyrase 1 or 2 and to be an acceptor of poly(ADP-ribosyl)ation (50). IRAP colocalizes with GLUT4, an insulin-responsive glucose transporter in adipocytes and muscle (46). After insulin stimulation, both IRAP and GLUT4 translocate from vesicles in the cytoplasm and trans-Golgi network to the plasma membrane (reviewed in reference 32). In IRAP Ϫ/Ϫ mice, absence of IRAP resulted in decreased GLUT4 expression (33). It has been suggested that tankyrases 1 and 2 may have a role in targeting or maintenance of GLUT4 vesicles (11,50). Interestingly, Glut4 Ϫ/Ϫ mice (similar to our Tnks2 PARP-deleted mice) display growth retardation and are 20% smaller than controls (31). However, since IRAP Ϫ/Ϫ mice do not have decreased body weight (33), the role of the IRAP-tankyrase 2 interaction in the small mouse phenotype remains to be determined.
Our studies of Tnks2 PARP domain-deleted mice have shed some light on the different functions of Tnks2 in vivo. For telomere maintenance, Tnks1 and Tnks2 could share redundant functions, but for normal postnatal growth, Tnks1 is not sufficient, as our mice display growth retardation. Further studies of these mice and of mice deficient in both Tnks1 and Tnks2 will be valuable in distinguishing the functions of Tnks1 and Tnks2 in telomere length maintenance and growth.