Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Tankyrase 1 and tankyrase 2 are poly(ADP-ribosyl)ases that are distinguishable from other members of the enzyme family by the structural features of the catalytic domain, and the presence of a sterile α-motif multimerization domain and an ankyrin repeat protein-interaction domain. Tankyrases are implicated in a multitude of cellular functions, including telomere homeostasis, mitotic spindle formation, vesicle transport linked to glucose metabolism, Wnt–β-catenin signaling, and viral replication. In these processes, tankyrases interact with target proteins, catalyze poly(ADP-ribosyl)ation, and regulate protein interactions and stability. The proposed roles of tankyrases in disease-relevant cellular processes have made them attractive drug targets. Recently, several inhibitors have been identified. The selectivity and potency of these small molecules can be rationalized by how they fit within the NAD+-binding groove of the catalytic domain. Some molecules bind to the nicotinamide subsite, such as generic diphtheria toxin-like ADP-ribosyltransferase inhibitors, whereas others bind to a distinct adenosine subsite that diverges from other diphtheria toxin-like ADP-ribosyltransferases and confers specificity. A highly potent dual-site inhibitor is also available. Within the last few years, tankyrase inhibitors have proved to be useful chemical probes and potential lead compounds, especially for specific cancers.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Diphtheria toxin-like ADP-ribosyltransferases (ARTDs; EC 126.96.36.199) covalently modify acceptor proteins by transferring an ADP-ribose moiety to glutamates or lysines of acceptor proteins. In this process, the catalytic ARTD domain cleaves NAD+ to nicotinamide and ADP-ribose, which is covalently attached to an acceptor protein or to a growing ADP-ribose polymer. The ARTDs can be subdivided into those that form poly(ADP-ribose) (PAR) chains (ARTD1–ARTD6) and those predicted to catalyze only mono-ADP-ribosylation (ARTD7–ARTD18), two of which are putatively inactive .
Among poly-ADP-ribosyltransferases, tankyrase 1 (TNKS1/PARP-5a/ARTD5; 1327 residues) and tankyrase 2 (TNKS2/PARP-5b/ARTD6; 1166 residues) are unique, owing to their ankyrin and sterile α-motif (SAM) protein–protein interaction domains. The ankyrin repeats are responsible for recognizing and interacting with protein targets, whereas the SAM domain is thought to mediate multimerization of tankyrase itself (Fig. 1). In addition to these domains, TNKS1 contains an N-terminal histidine, proline and serine-rich (HPS) domain with a hitherto unclear function.
TNKS1 was initially discovered as a protein involved in telomere homeostasis . However, tankyrases have much broader cellular roles, which include functions in mitotic centrosomes, in vesicle transport linked to glucose metabolism, in the β-catenin destruction complex (DC), in proteosomal assembly, in viral replication, and in cherubism [3-7, 7a]. The proposed roles of tankyrases in disease-relevant cellular processes have made them attractive drug targets [8-11]. Indeed, mouse studies suggest an effect of tankyrase inhibition in some cancer models, fibrotic diseases, and herpes simplex virus (HSV) infection. Here, we review the structural features, the biochemical functions and the cellular roles of tankyrases. We then present the advances in the development of specific tankyrase inhibitors, and discuss their interactions with the catalytic domains of TNKS1 and TNKS2.
Structure of tankyrases
Tankyrases belong to the PAR-forming class of enzymes, together with ARTD1–ARTD4. Tankyrases differ from the other four poly(ADP-ribosyl)transferases by their overall domain structure and functions. Tankyrases contain two structural domains that are not present in other ARTDs but are found in some other protein families: ankyrin repeats [12, 13] and the SAM domain . The catalytic ARTD (PARP) domain of tankyrases differs from that of ARTD1–ARTD4 specifically by the lack of an α-helical regulatory domain next to the catalytic ARTD domain (Fig. 1) . Both tankyrases lack a 32-residue loop on the back side of the ARTD domain that is found in ARTD1 . The overall sequence identity between TNKS1 and TNKS2 is high (82%) and covers the entire length. A distinguishing feature between TNKS1 and TNKS2 is a short seven-residue insertion between the ankyrin repeats and the SAM domain in TNKS2. In addition, TNKS1 contains an N-terminal HPS domain. The C-terminal catalytic ARTD domains of both tankyrases show the same overall structure, with 89% sequence identity . Detailed structural information of full-length tankyrases is still lacking, but the structures of the catalytic domains of both enzymes have been solved (Fig. 1D) [15, 16].
The catalytic ARTD domain of tankyrases
The catalytic ARTD domain of tankyrases cleaves NAD+ to nicotinamide and ADP-ribose, which is covalently attached to an acceptor protein or to a growing ADP-ribose polymer. The core structure of the catalytic ARTD domain contains the same α/β-fold as other ARTDs. Specifically, the crystal structure of TNKS1 reveals two antiparallel β-sheets flanked by four α-helices (Fig. 1D) . In contrast to other ARTDs, the catalytic domain of TNKS1/TKKS2 also contains a structural CHCC-type zinc-finger motif. Currently, there are no functional data available to allow interpretation of the role of this metal-binding site. It could be speculated to be involved in interacting with other proteins, nucleotides, the PAR chain, or the SAM domain of tankyrase itself.
The catalytic center of both tankyrases contains a glutamate that is required for the elongation of the PAR chain, but differs significantly from the other polymer-forming enzymes by the lack of the α-helical regulatory domain, which, in the case of ARTD1, enables the catalytic domain to respond to the signal from DNA damage detection . The catalytic ARTD domain contains two binding sites: one for the substrate (donor) NAD+, and the other for the target protein. The NAD+-binding site, deduced from the NAD+-bound diphtheria toxin structure , is formed by the ARTD signature motif (β–α–loop–β–α) (Fig. 1). At present, no NAD+-bound structures are available for ARTDs, including tankyrase. Only TNKS2 bound to the byproduct nicotinamide has been studied with crystallography .
In tankyrases, the NAD+-binding site appears to be more accessible than in ARTD1–ARTD3, where it is flanked by the regulatory domain. In the absence of the substrate NAD+, the crystal structures of TNKS1 and TNKS2 show a closed conformation of the NAD+-binding site (Fig. 2A). In this closed conformation, a protein sequence called the D-loop lines the donor site in ARTD enzymes, and undergoes interactions with the α-helix on the opposite bank of the binding groove (Fig. 2A). The D-loops in TNKS1 and TNKS2 are identical in sequence but differ in conformation. Importantly, the D-loop lining the NAD+-binding site is a distinctive feature that adopts different conformations in various ARTDs.
The NAD+-binding site can be divided into subsites binding the nicotinamide and the adenosine moieties (Figs 1D and 2A). In TNKS1, the D-loop closes the binding groove near the nicotinamide subsite, mainly through hydrophobic interactions, whereas in TNKS2 it closes the binding groove near the adenosine subsite, mainly by His1048 (Fig. 2A). The D-loop therefore needs to open for TNKS1 and TNKS2 to bind the donor NAD+. Opening of the D-loop has been observed when certain tankyrase inhibitors bind to this site (Table 1; Fig. 2). As the NAD+-bound structure of any ARTD enzyme is not available, it is not clear exactly how the dynamics of the D-loop impact on the enzymatic activity of tankyrases. The conformation of the D-loop in crystal structures is altered by the presence of specific inhibitors, and it is sometimes completely disordered (Fig. 2).
TNKS1 has an autocatalytic activity with a kcat of 0.7 s−1, which is comparable to the activity of ARTD1 when measured without activation by DNA damage  (J. Rippmann, personal communication). Furthermore, the measured Km of 1.5 mm for the substrate NAD+ for TNKS1 is 10-fold higher than that of ARTD1 [17, 20]. As these data were determined on the basis of in vitro auto-poly(ADP-ribosyl)ation (PARsylation), we do not know the exact kinetics in a cellular context or the nature of potential activation of the ARTD domain. TNKS1 catalyzes the formation of, on average, 20-unit-long ADP-ribose polymers in vitro . The formation of branched ADP-ribose polymers, as is common for PARsylation by ARTD1, has not been detected with tankyrase.
The catalytic activity of TNKS1 and TNKS2 is potentially modulated by post-translational modifications that are not well understood. TNKS1 is phosphorylated on serines by mitogen-activated protein kinase following insulin or growth factor stimulation , and during mitosis by glycogen synthase kinase 3 (GSK3)  and Polo-like kinase 1 (Plk1) , although the precise functions of these modifications are not clear. Recently, hydroxylation of tankyrase on asparagine/histidines was shown, and detailed studies of TNKS2 revealed eight sites that were differentially sensitive to hydroxylation by hypoxia-inducible factor asparagine hydroxylase (FIH) . Again, the functional implications of this modification are poorly understood. Finally, auto-PARsylation of tankyrase provides a recognition signal for ubiquitinylation through the WWE domain of the E3 ubiquitin ligase RING finger protein 146 (RNF146). It has been proposed that RNF146-dependent protein degradation may be a major functional implication of tankyrases’ PARsylation activity [24, 25].
The SAM and ankyrin domains of tankyrases
It has been shown that TNKS1 and TNKS2 associate with each other in vivo . The SAM domain is important for the multimerization of tankyrases, but the molecular details of this interaction remain to be elaborated. The polymerization of tankyrases through the SAM domain is disrupted by auto-PARsylation, making tankyrases a class of proteins that can assemble and disassemble according to their PARsylation status . Although auto-PARsylation inhibits tankyrase polymerization through the SAM domain , the PARsylation sites in either tankyrase have not yet been mapped to specific residue(s). It is unclear whether other covalent modifications, such as phosphorylation or hydroxylation, affect the assembly and disassembly of tankyrases.
The SAM domain is not only important for multimerization, but may also affect the catalytic ARTD domain. Although the isolated catalytic domain of TNKS2 shows ADP-ribosyltransferase activity , a larger fragment that also includes the SAM domain has typically been used for inhibition assays for TNKS1 and TNKS2, as its automodification kinetics are comparable to those of the full-length enzyme .
Tankyrases contain a long stretch of 24 ankyrin repeats, a motif that defines a large family of ankyrin-containing proteins . The ankyrin domain is divisible into five ankyrin repeat clusters (ARCs), and provides multiple binding sites for interacting proteins (Fig. 1) . The current view is that the ankyrin domain is involved in the interactions between tankyrases and target proteins to facilitate context-dependent PARsylation of the latter by the catalytic ARTD domain of tankyrases. However, the binding of a protein to the tankyrase ankyrin domain does not necessarily lead to PARsylation; instead, it can modulate the activity of tankyrases, as seen in the cases of GDP-mannose-4,6-dehydratase (GMD) and Fanconi anemia group 2D protein, which bind and inhibit tankyrase [29, 30].
The crystal structures of the TNKS2 ARC4 domain with peptides from various tankyrase targets have revealed the molecular basis for substrate recognition  (Fig. 1C). Target proteins bind to the pocket that is formed on the concave side of the ankyrin repeat motif, whereby a specific amino acid sequence of the target is recognized. This sequence motif has a consensus pattern of RXXPXG, where the key residues are an arginine at position 1, a glycine at position 6, and a small and hydrophobic residue at position 4. Binding of a target protein to the ankyrin repeats of tankyrase may also be a mechanism that assists tankyrase dimerization, as seen in the structure of TNKS ACR2–ACR3 cocrystallized with a peptide from AXIN (Fig. 1B) .
Roles of tankyrases
Tankyrases are widely expressed in human tissues, with high amounts of mRNA in the testis . In mice, tankyrase transcripts were reported in adipose tissue, brain, endocrine pancreas, skeletal muscle, kidney, spleen, thymus, and testis [35, 36]. As tankyrases recognize a degenerate hexapeptide motif, they interact with a variety of potential target proteins [31, 33, 34]. A substantial number of tankyrase-binding proteins have been reported, including telomeric repeat binding factor-1 (TRF1) , AXIN1/2 , insulin-responsive aminopeptidase (IRAP) [4, 26, 33], nuclear mitotic apparatus protein (NuMA)1 [33, 37, 38], myeloid cell leukemia 1 , TNKS1-binding protein 1 [28, 33], Epstein–Barr nuclear antigen 1 [40, 41], formin-binding protein 1/formin-binding protein 17 , basic leucine zipper nuclear factor 1 , cancer susceptibility candidate 3 , and 3BP2 . Potential tankyrase-interaction motifs have also been identified in a multitude of other proteins, although not all candidate partners have been biochemically confirmed [6, 33].
The functional relevance of tankyrases has been demonstrated in mouse studies. Double knockout of both genes is embryonic lethal, as is TNKS1 haploinsufficiency in the absence of TNKS2 . However, TNKS1 and TNKS2 appear to have largely overlapping functions, as deletion of either gene leads to subtle phenotypes: Mice with a genetic deletion of the TNKS2 ARTD domain are viable, but show growth retardation and a reduction in body weight, particularly in males [44, 45]. TNKS1 knockout mice are also viable, but show increased food intake and energy expenditure, as well as decreased adipose depot and plasma leptin levels .
Tankyrase subcellular localization
Tankyrases show a dynamic pattern of subcellular localization across the cell cycle, directing and restricting their interactions with target proteins and their cellular functions . In HeLa cells, during interphase, TNKS1 is found at the nuclear envelope (particularly the cytoplasmic fibers of the nuclear pore complex) and in association with TRF1 at the telomeres in the nucleus . During mitosis and concomitant with nuclear envelope breakdown, perinuclear TNKS1 appears to relocate to the poles of the mitotic spindle, while nuclear tankyrase remains associated with the telomeres [2, 46]. In meiotic cells, it was proposed that TNKS1 is involved in telomere clustering, a process that gathers meiotic chromosome ends on the inner surface of the nuclear envelope proximal to the spindle pole body . In accordance with a role during meiosis, tankyrase expression is abundant in the testis .
In nonpolarized cells such as 3T3-L1 preadipocytes and adipocytes, TNKS1 is predominantly located in the perinuclear space, where it has been linked to glucose transporter 4 (GLUT4) storage vesicles in the Golgi region . In polarized Madin–Darby canine kidney cells, TNKS1 localizes to the lateral membranes. It has been shown that this lateral membrane association is promoted by cell–cell contact and extracellular calcium, but diminished by the PARsylation activity of TNKS1 .
Finally, tankyrases have been localized to the β-catenin destruction complex , which upon tankyrase inhibition is visible in the perinuclear space in SW480 colon carcinoma cells . In the subsequent paragraphs, we will discuss the roles of tankyrases in various subcellular localizations (Fig. 3).
Tankyrase involvement in telomere maintenance
Early studies established a central functional role of TNKS1 at the telomeres, which are repetitive nucleotide sequences that protect the ends of chromosomes in most eukaryotic cells. TNKS1 has been shown to bind and poly(ADP-ribosyl)ate (PARsylate) TRF1, a negative regulator of telomere length . PARsylation of TRF1 reduces its ability to bind telomeres. Consequently, active tankyrase increases the accessibility of telomeres to telomerase. In accordance with this, prolonged overexpression of tankyrase leads to a progressive elongation of telomeres [3, 7], and suppression of TNKS1 may cause telomere shortening [49, 50]. Cancer cells have been shown to have long telomeres, helping them to escape cellular senescence, and telomerase inhibition has therefore become a possible therapeutic strategy . Tankyrase inhibition by broad-spectrum inhibitors, such as 3-aminobenzamidine and PJ-34, enhances the telomere shortening caused by inhibition of the telomerase, and this has therefore been suggested as a therapeutic strategy . PARsylated TRF1 is ubiquitinated by either F-box family protein 4  or RING finger LIM domain-binding protein, and targeted for proteasomal degradation [53, 54]. The tankyrase–TRF1 interaction can be modulated by TRF1-interacting nuclear protein 2, (TIN2) which forms a complex with tankyrase to prevent PARsylation of TRF1 . Phosphorylation of TNKS1 by Plk1, on the other hand, appears to increase TNKS1 stability .
It has been shown that TNKS1 localizes to spindle poles during mitosis, and its knockdown by RNA interference leads to preanaphase arrest [37, 56, 57]. In particular, tankyrases have been implicated in two centrosome-related processes: centriole elongation and mitotic spindle formation.
During centriole elongation, centrosomal P4.1-associated protein (CPAP) regulates the formation of elongated procentrioles and centriole multipolarity. CPAP stability and levels are regulated through PARsylation by TNKS1, a reaction that limits centriole elongation [56, 57].
During mitotic spindle formation, several proteins have been shown to be PARsylated. PARsylation of MIKI by TNKS1 during late G2 and prophase is required for the translocation of MIKI from the Golgi apparatus to the mitotic centrosome, where it anchors the scaffold protein CG-NAP . TNKS1, MIKI or CG-NAP downregulation leads to a prometaphase defect associated with impairment of microtubule aster formation . It has been suggested that PARsylation of MIKI by TNKS1 is a key initial event promoting prometaphase . TNKS1 also PARsylates NuMA, a large coiled-coiled protein that shuttles between the nuclear matrix in interphase and the spindle poles in mitosis, where it organizes microtubules [59, 60]. It has been proposed that PARsylation contributes to the dynamics of protein complex clustering at spindle poles, as the long PAR chains formed on the TNKS1–NuMA complex would interact with PAR-binding proteins to ensure the bipolarity of spindle assembly . Although small interfering RNA (siRNA) knockdown of tankyrase does not impair NuMA localization to the spindle pole, NuMA depletion abolishes the presence of tankyrase at the spindle poles. The binding of tankyrase to NuMA appears to be inhibited by GMD, whereby GDM blocks the catalytic activity of tankyrase and also sequesters an inactive tankyrase pool to prevent mitotic entry . Interestingly, another member of the ARTD superfamily, ARTD3/PARP-3, appears to be a critical player in mitotic spindle formation as well, as it associates with both NuMA and TNKS1. ARTD3 has been shown to stimulate NuMA PARsylation in a DNA-dependent manner. In addition, ARTD3 increases the automodification of TNKS1 and its ability to modify NuMA .
During mitosis, tankyrase is phosphorylated by GSK3β. Although the functional significance of this event remains unknown, it has been speculated to enhance the interaction of tankyrase with NuMA or other spindle-associated components . Plk1 has also been connected to TNKS1 in the spindles. Phosphorylation by Plk1 increases TNKS1 stability and its ADP-ribosyltransferase activity, whereas targeted inhibition of Plk1 decreases the stability and activity of TNKS1, leading to distorted mitotic spindle pole assembly .
An integrated model for the multiple interactions between tankyrase and the dynamic mitotic spindle apparatus remains to be established. Furthermore, it remains to be studied how the engagement of tankyrase at the spindle apparatus impacts on its other cellular functions, including the regulation of the β-catenin DC and β-catenin levels, which show cell cycle-dependent oscillation .
GLUT4 vesicle translocation
GLUT4 plays a central role in glucose homeostasis . The regulation of its subcellular localization has been extensively investigated in cultured 3T3-L1 adipocytes. In this in vitro system, GLUT4 in the basal state is sequestered to intracellular vesicles, and thus cannot facilitate glucose uptake. Following insulin stimulation, these vesicles translocate to the plasma membrane, and exocytosis enables GLUT4 to take up extracellular glucose. An earlier study showed that GLUT4 vesicles carry the tankyrase-binding protein IRAP , and that GLUT4 translocation is impaired when either IRAP or TNKS1 is knocked down by siRNA . Like TNKS siRNA, the generic ARTD inhibitor PJ34 also impairs GLUT4 translocation after a 45-min treatment . This is in contrast to the enhanced GLUT4 translocation reported in a subsequent study in adipocytes that were treated for 24 h with a potent tankyrase inhibitor, XAV939 . This study showed that tankyrase forms a complex with AXIN and KIF3A (a kinesin-like motor) to promote GLUT4 translocation. The formation of this ternary complex and its effect on GLUT4 are enhanced by XAV939 and insulin, both through inhibiting the catalytic activity of tankyrase . Given the paradox that GLUT4 translocation in this context is enhanced by XAV939 but attenuated by PJ34, it is not possible to predict the in vivo effects of tankyrase inhibitors on the basis of the phenotypes of existing tankyrase knockout mice. For this purpose, studies using tankyrase-selective inhibitors and mouse models with more subtle alterations, e.g. within the catalytic ARTD domain of TNKS1 or TNKS2, are required.
Tankyrase involvement in the β-catenin DC
The β-catenin DC is a dynamic multiprotein complex that assembles around β-catenin and presumably other related proteins of the armadillo protein family [65, 66]. Key proteins in the DC are the structural proteins adenomatous polyposis coli (APC) and AXIN1/2. GSK3β, cyclin-dependent kinase (CK)1α, CK1δ and CK1ε are recruited to the structural core of the DC. This leads to sequential phosphorylation of β-catenin near the N-terminus to induce recognition by the E3 ubiquitin ligase β-transducin-repeat-containing protein, followed by the degradation of β-catenin by the proteasome . Tankyrases play a key role in the DC by regulating the stability of the rate-limiting AXIN proteins, RNF146 and of tankyrase itself [10, 24, 48]. The E3 ubiquitin ligase RNF146 recognizes tankyrase-mediated PARsylation and eartags AXIN, tankyrase and itself for proteasome-mediated degradation [24, 25]. Thus, tankyrase controls the protein stability and turnover of key components of the DC, and consequently the cellular levels of β-catenin. Inhibition of the catalytic activity of tankyrases by small molecules induces robust stabilization of AXIN in the DC, leading to visible accumulation of DC protein aggregates in the perinuclear space and, in a context-dependent manner, to a reduction in Wnt–β-catenin signaling [10, 48, 68].
Interestingly, AXIN and tankyrases also interact in other multi-protein complexes, as AXIN can undergo dynamic interactions with various tankyrase partners, including the kinesin-like motor KIF3A, that is involved in insulin-stimulated GLUT4 translocation [63, 69]. AXIN also binds to Dishevelled in the Wnt signalosome, although a role of tankyrase in this context has not been shown [62, 70, 71]. Finally, it is noteworthy that GSK3β kinase activity is modulated through mono-ADP-ribosylation by ARTD10/PARP-10 . However, it is unclear whether this GSK3β modification has functional implications in the context of the β-catenin DC.
Tankyrase and proteosomal activity
Recently, tankyrase was identified as a regulator of proteasomal activity in both Drosophila and mammalian cells. A model was suggested where TNKS-mediated ADP-ribosylation provides a signal for local assembly of 26S proteasomes possibly around proteins targeted for degradation. TNKS-mediated ADP-ribosylation reduces the affinity of the proteasome inhibitor PI31 to the 20S proteasome alpha subunit and thereby relieves [7a] the repression of the 20S subunit. In the next step, ADP-ribosylated PI31 sequesters the dp27 and dS5b assembly chaperones from the 19S particles that can now assemble with the 20S subunit to form a functional 26S proteasome [7a]. Since the rate of intracellular protein degradation is dependent on the metabolic state of a cell, a possible link between cellular NAD+ levels, tankyrase activity and proteasomal activity has been proposed [7a]. Tankyrase inhibitors may provide novel tools to attenuate metabolism dependent proteasomal activity.
Tankyrase involvement in diseases
Altered tankyrase expression or activity is implicated in various disease states; in particular, altered expression of TNKS1 and/or TNKS2, as well as genetic alterations in the tankyrase locus, have been detected in multiple tumors (Table 2). Increased expression of TNKS1 and TNKS2 has been observed in many different cancers, including fibrosarcoma , ovarian cancer , glioblastoma , pancreatic adenocarcinoma , transitional cell carcinoma of the bladder , gastric cancer , and breast cancers [2, 8, 77], including estrogen receptor-negative and progesterone receptor-negative tumors, where TNKS1 mRNA expression shows a grade-dependent increase . Furthermore, a significant association between TNKS1 upregulation and the pathological grade of astrocytomas was recently reported . Increased expression has been found in colon cancer [80, 81], in which discordant changes in TNKS1 and TNKS2 expression are detected, as TNKS2 tends to be upregulated whereas TNKS1 tends to be downregulated as compared with normal tissue . Curiously, one study associated lower TNKS1 mRNA levels in colon cancer with reduced survival . Moreover, in subsets of gastric and colorectal cancers, deletion or duplication in exons of TNKS1 and TNKS2 were found .
Recent work points towards context-dependent effects of tankyrase inhibition on tumor development and progression. Tankyrase inhibition by G007-LK, a reagent with high target selectivity, reduces growth and induces differentiation of selected APC mutant colorectal carcinomas (CRCs) in vitro and in vivo, although, in several APC and KRAS-mutant CRC cell lines, this tankyrase inhibitor did not show single-agent activity . Interestingly, in these CRC cell lines, G007-LK potentiates the effect of a MEK inhibitor . Further evidence is accumulating that tankyrase inhibition can be enhanced by other inhibitors or altered components of the signaling pathway. In CRC models, for instance, the tankyrase inhibitor XAV939 can unmask the proapoptotic effect of phosphoinositide 3-kinase and AKT inhibitors . Furthermore, in non-small cell lung cancers, simultaneous inhibition of tankyrase (with small hairpin RNA or small molecules) and epidermal growth factor receptor showed synergistic efficacy . Importantly, it has been demonstrated that, in tumor cells with decreased expression of either breast cancer type 1 susceptibility protein (BRCA)1 or BRCA2, tankyrase inhibition can be synthetically lethal, and can exacerbate the centrosome amplification phenotype associated with BRCA deficiency . Clearly, more research is needed in order to understand the context dependency of tankyrase inhibitors in cancer cells.
Tankyrases are not only important in the context of tumors; they also appear to impact on viral infections. In HSV infection, it was shown that the virus cannot replicate efficiently in cells that with depletion of both TNKS1 and TNKS2. The tankyrase inhibitor XAV939 was shown to decrease HSV viral titers significantly . In contrast, tankyrase has been proposed to inhibit the replication of the Epstein–Barr virus origin of plasmid .
A striking emerging field is a connection between tankyrase and glucose metabolism. It has been reported that DNA polymorphism in a chromosomal region including the sequence encoding tankyrase/methionine sulfoxide reductase A is robustly associated with early-onset obesity . A link to obesity would also be consistent with the findings of reduced fat pads in TNKS1 knockout mice . Nevertheless, further work is needed to clarify the connections between tankyrase and obesity.
Tankyrase may also play a role in tissue fibrosis. In experimentally induced fibrosis, tankyrase inactivation with XAV939 or siRNA leads to antifibrotic effects . Recently, a causative connection was established between tankyrase and cherubism, a syndrome characterized by the progressive loss of the jawbone and the accumulation of fibrous tissue, and systemic inflammation. It was shown that the mechanistic basis of cherubism is the loss of tankyrase interaction with the substrate protein 3BP2 [31, 43].
As in the case of other ARTDs, tankyrase inhibitor discovery has so far been focused on the catalytic ARTD domain. In the long term, the protein interfaces between ARCs and substrate proteins may provide further interference points for targeting specific tankyrase substrates.
The enzymatic characterization of tankyrases shows inhibition by 3-aminobenzamide, a general ARTD inhibitor . Compounds that show cross-inhibition of tankyrases have also been identified during ARTD1 inhibitor development . The first published tankyrase-selective inhibitors, XAV939 and IWR-1, were discovered by searching for compounds that block Wnt–β-catenin signaling [10, 90]. Wnt reporter-based screens also uncovered the specific tankyrase inhibitors WIKI4, JW55, and JW74, and its derivative G007-LK [68, 91, 92].
Direct biochemical approaches, recently combined with fragment based ligand design, have been utilized for screening and characterizing compounds inhibiting tankyrases [93-96]. The methodologies include enzymatic assays, thermal stabilization of the recombinant protein, and alleviation of the toxicity of tankyrase overexpression in yeast cells.
Structural basis of tankyrase inhibition
The nicotinamide subsite
The search for ARTD inhibitors has been focused on the nicotinamide subsite of the donor NAD+-binding site . This site is conserved among members of the human ARTD family, and has also provided the anchoring site of many tankyrase inhibitors. Some of the small-molecule inhibitors binding to this site can be considered to be general ARTD inhibitors, and structural information is currently available for TIQ-A [Protein Data Bank (PDB): 4AVW] and phenanthridinone (PDB: 4AVU). These small inhibitors bind to the catalytic site of ARTDs, overlapping well with the binding site of nicotinamide, a byproduct of the enzymatic reaction (Fig. 2A). The nicotinamide subsite is also targeted by the potent tankyrase inhibitor XAV939, which modestly cross-inhibits ARTD1 and ARTD2 (Table 1 and Fig. 3B; IC50 of 2294 nm for ARTD1 and 114 nm for ARTD2) [10, 16, 98]. Like most ARTD inhibitors, XAV939 forms the typical hydrogen bonds with Gly1032 and Ser1068 of TNKS2 (Gly1185 and Ser1221 for TNKS1) at the bottom of the pocket (Fig. 2G). Another general feature of inhibitors binding to this site is the π–π stacking with Tyr1071 of TNKS2 (Tyr1224 in TNKS1) (Fig. 2G). Interestingly, flavones, a class of natural products found in many plants, have been shown to be potent tankyrase inhibitors [93, 95]. These molecules bind to the nicotinamide subsite much like synthetic inhibitors, despite lacking hydrogen bonding with the carbonyl of the glycine in TNKS . The strength of synthetic inhibitor binding to the nicotinamide subsite can be increased by additional interactions, mainly hydrophobic, with adjacent residues, such as Phe1035 of TNKS2 (Phe1188 in TNKS1), that are unique to tankyrases (Fig. 2G).
The adenosine subsite
IWR-1 was discovered as a Wnt–β-catenin signaling inhibitor, and was later shown to directly inhibit tankyrases [10, 90] with remarkable specificity [10, 19]. Importantly, instead of binding to the nicotinamide subsite, like most other known ARTD inhibitors, IWR-1 binds to the adenosine subsite of tankyrase and causes opening of the D-loop (Fig. 2D) . Its norbornyl moiety binds between three tyrosines of the active site, and its adenosine moiety binds between the active site helix and a histidine in the D-loop (Fig. 2H).
Additional tankyrase inhibitors that do not mimic nicotinamide but instead bind to the lower part of the donor NAD+-binding cleft include WIKI4 , JW55 , and JW74 . Alterations of the JW74 scaffold have improved both the pharmacokinetics and the potency towards tankyrase, and, in particular, G007-LK shows significantly improved pharmacokinetics in mice [92, 100]. These compounds bind to the adenosine subsite and form π–π stacking with His1048 in the binding groove (Fig. 2E). This opens the D-loop much like IWR-1 does (Fig. 2H). G007-LK also undergoes interactions with Tyr1060 and Tyr1071 of TNKS2 and T-shaped π–π stacking with Phe1035 . Also, recently discovered highly potent oxazolidinone compounds bind to the adenosine site of tankyrases .
The general ARTD inhibitor PJ34 was shown to bind as expected to the nicotinamide subsite. Interestingly, a second molecule of the inhibitor was found to bind to the adenosine subsite (Fig. 3C) , stacking between the α-helix and His1201 of TNKS1, similarly to inhibitors binding exclusively to this site. It also forms a hydrogen bond with the backbone amide of Asp1198 of TNKS1.
The discovery of the second distinct druggable binding site in TNKS1 and TNKS2 has resulted in efforts to design inhibitors that block these two sites simultaneously. This generated a quinazolinone binding to the nicotinamide subsite with a long extension reaching the adenosine subsite (Fig. 2F) . This compound shows excellent potency towards tankyrases and selectivity over ARTD2.
Common features influencing the potency and selectivity of tankyrase inhibitors
The apo crystal structures of the TNKS1 and TNKS2 catalytic domains both show a closed, inactive conformation of the D-loop lining the donor NAD+-binding site (Fig. 2A). This D-loop conformation does not preclude inhibitors from binding to the nicotinamide subsite, because the conformation is not significantly altered by tankyrase inhibitors, despite their high potency in solution (Fig. 2A,B). This is also the case when the crystal form would allow the D-loop to open without disturbing the crystal contacts. The crystal structures can therefore be used to guide inhibitor development while taking into account potential conformational changes of the D-loop induced by inhibitors. Thus, when inhibitors bind to the adenosine subsite, the D-loop partially opens to form a groove that accommodates the adenosine mimic (Fig. 2). In contrast, without specific interactions with an inhibitor, the D-loop often appears to be disordered in the crystal structures (Fig. 2C).
The increasingly well-characterized adenosine subsite of TNKS1 and TNKS2 allows the design of new chemical scaffolds with improved specificity towards tankyrases over other ARTDs. It is not yet clear whether such selective compounds will have off-target effects on other NAD+-utilizing enzymes, such as cytochrome P450 .
Tankyrases use the ankyrin domain to bind to the RxxPxG motif on dozens of established partners, and presumably numerous other proteins bearing this sequence motif. Many of these partners are recruited to the ARTD domain to become PARsylated, whereby their stability in the protein complex and degradation are regulated. The catalytic function of tankyrase presumably enables the modulation of many cellular processes beyond the well-characterized telomere elongation and AXIN degradation. Adding further complexity to the possible biological response of tankyrase blockers, evidence is emerging that AXIN stabilization by tankyrase inhibitors does not always reduce Wnt–β-catenin signaling. Instead, it can increase Wnt–β-catenin signaling, as has been shown in the late primitive streak of mouse embryos . Moreover, as tankyrases share the utilization of NAD+ with many other NAD+-consuming enzymes in cells, including other ARTDs and the sirtuins, tankyrase inhibition could redirect the NAD+ cosubstrate towards other consumers, leading to enhancement of other NAD+-dependent protein modifications. In addition, tankyrase inhibition may attenuate the cellular redox state by increasing the NAD+/NADH ratio.
Many of the cellular pathways that are modulated by tankyrases, such as telomere maintenance and Wnt–β-catenin signaling, impact on decision points between stem cellness and differentiation. Indeed, a frequently seen phenomenon in cells that are treated with specific tankyrase inhibitors is induction of differentiation. As a consequence, tankyrase inhibition may elicit short-term or long-term effects on natural stem cell or progenitor populations, and affect aging. In consequence, a multitude of potential side effects of the inhibition need to be considered as therapeutic protocols are developed. Dose-limiting intestine toxicity, including reduction in intestinal crypt proliferation, has been observed in mice treated with high doses of a tankyrase-specific inhibitor . However, it is not yet clear whether the effect is target-specific, how tankyrase inhibition affects crypt stem cells, or whether the observed toxicity is reversible.
Evidently, more functional studies will be required to increase our understanding of tankyrases and their roles in cellular processes on the path towards therapeutic protocols. There appears to be a general understanding in the cancer field that tankyrase inhibitors may have to be combined with other pathway-attenuating agents to potentiate the effect of, for example, MEK, epidermal growth factor receptor and phosphoinositide 3-kinase inhibitors. Also, combined inhibition of various selected ARTD isoenzymes is an interesting option that needs to be explored further .
Despite these considerations, several possible applications for tankyrase inhibition have been suggested, and initial promising results in selected cellular and in vivo models show that tankyrase inhibitors will provide a valuable new tool for the therapeutic arena.
The work was supported by Biocenter Oulu (L. Lehtiö), the Research Council of Norway (S. Krauss), and the Department of Veterans Affairs of the USA (N.-W. Chi). The authors would like to thank M. Costa for critical reading of the manuscript.