APC Mutations as a Potential Biomarker for Sensitivity to Tankyrase Inhibitors in Colorectal Cancer

Loss-of-function mutations in adenomatous polyposis coli (APC) gene, a negative regulator of Wnt/β-catenin signaling, frequently occur in colorectal cancer (1, 2). These APC mutations promote tumorigenesis together with several common mutations, such as those in KRAS, SMAD4, and TP53 (3, 4). APC functions as a functional scaffold for the β-catenin destruction complex, which is composed of β-catenin, Axin, casein kinase 1α and glycogen synthase kinase 3β (GSK3β), and suppresses Wnt/β-catenin signaling (5–10). In the destruction complex, APC and Axin promote the ubiquitin-dependent proteasomal degradation of β-catenin via casein kinase 1α- and GSK3β-mediated phosphorylation of β-catenin (8, 10). In genetically engineered mouse models, APC mutation–derived activation of Wnt/β-catenin signaling was shown to be required not only for promotion of colorectal cancer but also for tumor maintenance (11, 12). These observations indicate that the Wnt/β-catenin signaling pathway is a rational therapeutic target for colorectal cancer. So far, however, this pathway lacks druggable molecular targets, which has hampered the development of therapeutic drugs targeting Wnt/β-catenin signaling.

Tankyrases (tankyrase-1/PARP-5a and tankyrase-2/PARP-5b) are members of the PARP family (13–15). This unique class of PARPs recognizes its binding partners via the multiple ankyrin repeat cluster domain and poly(ADP-ribosyl)ates (PARylates) them via the carboxyl terminal catalytic domain (13, 16, 17). Tankyrases enhance Wnt/β-catenin signaling by PARylation and subsequent destabilization of Axins (18, 19). Pharmacologic inhibition of tankyrases causes accumulation of Axins, which in turn facilitates degradation of β-catenin and suppresses the growth of several APC-mutated colorectal cancer cell lines (18, 20–23). These observations suggest that tankyrase inhibitors are promising drugs for colorectal cancer treatment. However, colorectal cancer cells are not equally sensitive to tankyrase inhibitors (22). Although appropriate stratification of patients has profound implications for new drug development, predictive biomarkers for sensitivity to tankyrase inhibitors remain elusive. In this study, we demonstrate that a class of APC mutations predicts the sensitivity of colorectal cancer cells to tankyrase inhibitors.

Human colorectal cancer cell lines COLO-320DM, SW403, SW480, and DLD-1 were maintained in RPMI1640 medium (Gibco, Life Technologies) supplemented with 10% FBS. RKO cells were maintained in DMEM low glucose (Nacalai Tesque) with 10% FBS. These cell lines were obtained from ATCC. HCT-15, HCC2998, HT-29, KM12, and HCT-116 cells were maintained as described previously (24). All cell lines were authenticated by short tandem repeat profiling analysis (BEX) in 2016. Cell proliferation was evaluated by the MTT assay (5-day drug treatment) as previously described (25) or by colony formation assay (10-day drug treatment). Normalization was done to cells treated with DMSO as vehicle, which were defined as 100%. Final concentration of DMSO was constant and the same for all the treatments.

Total RNA was purified from COLO-320DM cells using the RNeasy Kit (Qiagen) and used as a template to amplify the truncated APC [amino acid (aa) 1–811] fragment by reverse transcription-PCR. The primer sets were: 5′-ATCGGATCCGCTGCCACCATGGAACAAAAGCTGATTTCTGAAGAAGATCTGGAATTCGCTGCAGCTTCATATGATCAGTTGTTA-3′ (forward) and 5′-ACACTCGAGATATGGTGAAAGGACAGTCATGTTGC-3′ (reverse). The forward and reverse primers contain BamHI and XhoI sites, respectively. The PCR product was cloned into BamHI and XhoI sites of the pLPC vector to generate the pLPC-APC811 plasmid.

Cells were transiently transfected with the β-catenin-responsive reporter vectors pTcf7wt-luc (carrying seven repeats of the Tcf-binding consensus sequence) or pTcf7mt-luc (carrying seven repeats of a mutated Tcf-binding consensus sequence, to which Tcf cannot bind; provided by Dr. Kunitada Shimotohno, National Center for Global Health and Medicine, via Riken BioResource Center, Ibaraki, Japan) and phRLuc (Promega) by Lipofectamine 2000 reagent (Thermo Fisher Scientific) or by electroporation using the Neon Transfection System (Thermo Fisher Scientific) with two pulses for 20 ms and 1,300 V each time. Luciferase assays were performed using the Dual-Glo Luciferase Assay System (Promega) and firefly and Renilla luminescence were measured with a multifunctional reader Genios (Tecan).

Whole cell lysates were prepared using lysis buffer (1% Nonidet P-40, 150 mmol/L NaCl, 50 mmol/L Tris-HCl, pH 8.0) with 2% (v/v) of protease inhibitor cocktail (Nacalai Tesque) and 1% (v/v) of phosphatase inhibitor cocktail (Nacalai Tesque). Immunoblotting was performed as previously described (17). The primary antibodies were anti-tankyrase-1/2 (H350; Santa Cruz Biotechnology), anti-active β-catenin (8E7; Millipore), anti-phospho-β-catenin (Ser33/37/Thr41; Cell Signaling Technology), anti-Axin1 (C76H11; Cell Signaling Technology), anti-Axin2 (76G6; Cell Signaling Technology), anti-GAPDH (6C5; Fitzgerald), anti-APC (ALi 12-28; Santa Cruz Biotechnology) for detecting the short APC (aa 1–811), and anti-APC (2504; Cell Signaling Technology) for APC larger than 811 aa.

ON-TARGETplus (human CTNNB1 and nontargeting control) and Silencer Select [APC (s1433: siAPC#1; s1434: siAPC#2; s1435: siAPC#3) and negative control no. 1] siRNAs were purchased from GE Healthcare Dharmacon and Thermo Fisher Scientific, respectively. siRNAs were introduced into cells with reverse transfection method using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific).

Total RNAs were purified with the RNeasy Mini Kit (Qiagen) and cDNA were synthesized with the SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Thermo Fisher Scientific). The expression levels of AXIN2 or Naked1 (NKD1) mRNA were quantified by real-time PCR analysis with the LightCycler 480 Real-Time PCR System with Universal ProbeLibrary Probe #36 (Roche). The primers were as follows: AXIN2, 5′-CACACCCTTCTCCAATCCAA-3′ (forward) and 5′-TGCCAGTTTCTTTGGCTCTT-3′ (reverse); and NKD1, 5′-TCTCGCCGGGATAGAAAAC-3′ (forward) and 5′-TCTCGCCGGGATAGAAAAC-3′ (reverse).

Cells on glass coverslips were fixed with 2% formaldehyde/PBS and permeabilized with 0.5% Nonidet P-40/PBS. Immunofluorescence staining was performed as previously described (17) with antibody against nonphosphorylated (i.e., active) β-catenin (D13A1; Cell Signaling Technology). The nuclei were stained with VECTASHIELD Antifade Mounting Medium with 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories).

Surgical specimens of colorectal cancer from 16 patients were obtained between September 2013 and December 2015. Standard histopathologic analysis was performed to confirm the diagnosis of malignancy and the histological subtype. For all samples, total RNA and genomic DNA were isolated as described below. Electronic medical records were retrospectively reviewed to obtain clinical information in accordance with an institutional review board (IRB)-approved protocol. The patients submitted written informed consent for genetic and cell biological analyses, which were performed in accordance with protocols approved by the IRBs of the Japanese Foundation for Cancer Research (Tokyo, Japan).

A few pieces of tumors from colorectal cancer patients were obtained after surgical resection. Tumor pieces were immediately placed in ice-cold culture medium with antibiotic-antimycotic (Gibco). Tumor tissues were cut into small fragments, and washed with ice-cold PBS supplemented with antibiotic–antimycotic. Tumor pellets were enzymatically digested with collagenase/dispase (Roche) and DNase I in StemPro ESC culture medium (Invitrogen) for 30 to 60 minutes. After washing with antibiotic–antimycotic and 0.2% BSA-containing PBS, the cell pellets were cultured in the StemPro ESC medium supplemented with 10 μmol/L of Y-27632 to establish the patient-derived JC (JFCR-Colorectal) cell lines. Before subsequent experiments, the cells were subcultured until the coexisting stromal cells were scarcely detected under the microscope.

Genomic DNA was isolated from cell pellets or fresh frozen normal/tumor specimens using the DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer's protocol. The isolated DNA was processed on the MiSeq platform (Illumina, Inc.) with a Haloplex custom panel (Agilent), which is designed to detect well-known cancer-associated somatic mutations. Somatic variants were called using the Somatic Variant Caller (Illumina) and annotated using the SureCall software tool (Agilent). Total RNA was isolated from cell pellets or fresh frozen specimens using the RNeasy Mini Kit (Qiagen). Mutated APC cDNAs were PCR-amplified from cDNA synthesized from total RNA with oligo-dTs using the KOD Plus Neo Kit (Toyobo Co., Ltd.) and bidirectionally sequenced by the Sanger method.

Tukey–Kramer tests were performed to examine every combination of multiple experimental groups. Multiple regression analysis was performed with sensitivities to tankyrase inhibitors and existence of genetic mutations (APC, KRAS, PIK3CA, and TP53), which were defined as dependent and independent variables, respectively.

To identify molecular parameters that are associated with cellular sensitivity to tankyrase inhibitors, we first evaluated the growth-inhibitory effects of tankyrase inhibitors G007-LK, IWR-1 and XAV939 (18, 20, 22) on nine established colorectal cancer cell lines. COLO-320DM, SW403, HCC2998, DLD-1, HCT-15, HT-29, and KM12 cells are APC-mutated, whereas HCT-116 and RKO cells retain wild-type APC. As expected, none of the inhibitors repressed the growth of HCT-116 or RKO cells (Fig. 1A) because HCT-116 cells harbor a constitutively active mutation in β-catenin (CTNNB1, deletion of Ser45, a phosphorylation site of casein kinase 1α) and RKO cells have no Wnt/β-catenin signal-related mutations. Among the rest of the cell lines with APC-truncated mutations, COLO-320DM and SW403 cells showed high sensitivity to the three tankyrase inhibitors (Supplementary Table S1), whereas the other cell lines were resistant.

We next examined transcriptional activity of Tcf/LEF, a downstream signaling component of the Wnt/β-catenin pathway, using a Tcf/LEF luciferase reporter construct. Figure 1B shows that COLO-320DM and SW403 cells exhibit high Tcf/LEF transcriptional activity compared with the tankyrase inhibitor–resistant cells. These transcriptional activities roughly correlated with β-catenin dependency of the cells (Fig. 1C and D; Supplementary Fig. S1). Importantly, COLO-320DM and SW403 cell growths were sensitive to β-catenin depletion, ensuring the mode-of-action of tankyrase inhibitors. These observations so far indicate that tankyrase inhibitor-sensitive colorectal cancer cells possess three characteristic properties: (i) APC truncated mutation, (ii) highly activated Wnt/β-catenin signaling, and (iii) β-catenin–dependent cell growth.

Tankyrase promotes Axin degradation and thereby induces β-catenin stabilization (18). Therefore, the pharmacodynamic response of the tankyrase–Axin–β-catenin axis could be a clue to determine the cellular sensitivity to tankyrase inhibitors. When we examined the basal expression levels of tankyrase, Axins, and active β-catenin in the colorectal cancer cell lines used, we did not observe any correlation between their expression levels and sensitivity to tankyrase inhibitors (Supplementary Fig. S2A–S2C). β-Catenin is localized both in adherence junctions and in the nucleus (26). In APC-mutated cell lines, subcellular localization or expression levels of β-catenin did not correlate with tankyrase inhibitor sensitivity (Supplementary Fig. S2D). We next monitored the effect of tankyrase inhibitors on the tankyrase–Axin–β-catenin axis. When nine colorectal cancer cell lines were treated with G007-LK or IWR-1, accumulation of Axins (Axin2, especially) and subsequent decrease of nonphosphorylated β-catenin (i.e., active β-catenin) was detected in the inhibitor-sensitive COLO-320DM and SW403 cells (Fig. 2A; Supplementary Fig. S1). In contrast, in tankyrase inhibitor–resistant cell lines, protein levels of active β-catenin did not change after inhibitor treatment. Consistent with these observations, in the inhibitor-sensitive cells, tankyrase inhibitors efficiently suppressed the highly activated Tcf/LEF activity (Fig. 2B) and mRNA expression of AXIN2 and NKD1 (27, 28), two target genes of Wnt/β-catenin signaling (Fig. 2C). However, in the inhibitor-resistant colorectal cancer cells with APC mutations (DLD-1, HCT-15, HCC2998, and KM12 cells), the levels of Tcf/LEF activity and downstream gene expression were relatively low and suppressive effects of tankyrase inhibitors on these cells were less clear. Of note, in HCC2998, HCT-15, and DLD-1 cells, tankyrase inhibitors caused Axin accumulation but failed to downregulate β-catenin. These results indicate that the pharmacodynamic effect of tankyrase inhibition is not transmitted to β-catenin downregulation in these resistant cell lines.

Previous studies have shown that mutations of APC gene occur preferentially within the mutation cluster region (around aa 1282–1581) in colorectal cancers (2, 29–31). This results in APC gene products that lack several or all of the seven 20-amino acid repeats (20-AARs), which contribute to β-catenin binding and degradation. We next focused on the relationship between the positions of APC mutations and the cellular sensitivity to tankyrase inhibitors. As shown in Fig. 3, we noticed that the cell lines sensitive to tankyrase inhibitors harbored short, truncated APC mutants that lacked all seven 20-AARs (referred to as “short APC” hereafter), whereas the tankyrase inhibitor-resistant cell lines retained longer APC mutants containing two or more 20-AARs (referred to as “long APC” hereafter). We found that another colorectal cancer cell line, SW480, which retained a single 20-AAR, was resistant to tankyrase inhibitors (Supplementary Fig. S3A). Supplementary Table S2 shows statistically significant correlation between the short APC mutations and sensitivities to G007-LK (P = 0.015) and IWR-1 (P < 0.001). Moreover, PIK3CA mutations were correlated with sensitivities to IWR-1 (P = 0.044) but not G007-LK (P = 0.141). None of the other recurrent mutations in colorectal cancers showed correlation with the tankyrase inhibitor sensitivity. These observations suggest that APC mutation status could be a potential biomarker for prediction of tankyrase inhibitor sensitivity of colorectal cancer cells.

As shown earlier, tankyrase inhibitor-resistant colorectal cancer cell lines possess long APCs with two or more 20-AARs and fail to downregulate β-catenin upon tankyrase inhibition. To examine the functional involvement of the long APCs in the resistance of β-catenin to tankyrase inhibitor-mediated downregulation, we knocked down the expression of the long APCs in these cells. We particularly focused on the three cell lines expressing long APC with two 20-AARs, HCC2998, HCT-15 and DLD-1 cells, because the signaling from Axin2 accumulation to β-catenin downregulation was prevented in these cell lines. Depletion of the long APC resulted in elevation of active β-catenin levels (Fig. 4A; Supplementary Fig. S1) and downstream AXIN2 mRNA expression (Fig. 4B), although the effects on active β-catenin and AXIN2 expression in DLD-1 cells were not detected or detected only marginally, if any. These observations support the notion that the long APCs that partially retain several 20-AARs are hypomorphic mutants in terms of the ability to suppress intracellular β-catenin levels. Under long APC-depleted conditions, G007-LK–induced Axin2 accumulation and was able to downregulate β-catenin. Meanwhile, SW480, which retained a single 20-AAR, was able to downregulate β-catenin upon G007-LK treatment, and depletion of this mutant APC did not affect β-catenin stability (Supplementary Fig. S3B). Of note, in this cell line, the basal level of active β-catenin was very high and significant amounts remained even after the drug treatment, which would allow the drug-resistant cell growth (Supplementary Fig. S3A). Immunofluorescence staining confirmed that depletion of the long APC with two 20-AARs led to accumulation of active β-catenin in the cell nucleus, which was reversed by G007-LK (Fig. 4C). Similar results were obtained when other APC siRNAs and IWR-1 were used (Supplementary Fig. S4A and S4B). Knockdown of the long APCs did not affect the cellular sensitivities to tankyrase inhibitors (Supplementary Fig. S4C and S4D).

To investigate the functional difference of the short and long APCs, we evaluated the influence of short APC expression on tankyrase inhibitor–induced β-catenin degradation. Again, knockdown of the endogenous long APC in HCC2998 cells enabled G007-LK and IWR-1 to induce degradation of β-catenin (Fig. 5A; Supplementary Fig. S5A, lanes 1, 2, 5, 6). Exogenous expression of short APC (APC811, derived from COLO-320DM cells) in the absence of the endogenous long APC in HCC2998 cells did not affect tankyrase inhibitor–induced β-catenin degradation (Fig. 5A; Supplementary Fig. S5A, lanes 5–8). Ectopic expression of the short APC in HCC2998 cells did not affect the sensitivity to tankyrase inhibitors in the presence or absence of the long APC (Supplementary Fig. S5C and S5D). Because the long APC is a hypomorphic mutant, its depletion in HCC2998 cells increased Tcf/LEF transcriptional activity (Fig. 5B). Consistent with the results in Fig. 5A, ectopic expression of APC811 had little influence on this reporter activity. In contrast, overexpression or knockdown of APC811 in COLO-320DM cells did not affect the levels of active β-catenin or Tcf/LEF activity (Fig. 5C and D; Supplementary Fig. S5B). Tankyrase inhibitor–mediated degradation of β-catenin in long APC-depleted HCC2998 cells depended on Axin2 but not Axin1 (Fig. 5E). COLO-320DM cells also required Axin2 for tankyrase inhibitor–mediated degradation of β-catenin, on which APC811 (‘short’ APC) had no effect (Fig. 5F). Together, these observations indicate that the long APC, but not the short APC, disturbs tankyrase inhibitor–induced β-catenin downregulation (Supplementary Fig. S6).

To validate the short APC mutations as a potential biomarker to predict tankyrase inhibitor sensitivity in more clinically relevant cells, we established patient-derived cells (PDC) from surgical specimens of colorectal cancer patients and analyzed the sequence of the APC gene (Supplementary Fig. S7A). These PDCs exhibited differential sensitivities to G007-LK and IWR-1 (Fig. 6A). On the basis of the cell growth inhibition rate, we divided these cells into three groups: sensitive, intermediate, and resistant (Supplementary Fig. S7B). Integrated analysis of APC mutation status and tankyrase inhibitor sensitivity revealed that PDCs in the resistant group tended to possess long APC mutants compared with the sensitive and intermediate groups together (Supplementary Fig. S7C). Importantly, PDCs with short APC mutants (20-AAR = 0) were classified as sensitive/intermediate, without exception (Fig. 6B). On the other hand, PDCs with long APC mutants (20-AAR ≥ 1) were either sensitive/intermediate or resistant to tankyrase inhibitors with comparable tendency. Collectively, these observations indicate that the short APC lacking all 20-AARs could potentially be a predictive biomarker for sensitivity of colorectal cancer cells to tankyrase inhibitors. Colony formation assay of six PDCs gave essentially similar results with MTT assay, although the cells tended to be more sensitive to tankyrase inhibition in colony formation assay because of longer drug exposure (Supplementary Fig. S8A). Knockdown of β-catenin inhibited the growth of these PDCs to comparable extents with tankyrase inhibitors although the tankyrase inhibitor–sensitive JC-35 was resistant to β-catenin depletion presumably due to high basal level expression of β-catenin and incomplete knockdown of the gene (Supplementary Fig. S8B and S8C).

In this study, we demonstrated that the colorectal cancer cell lines that are sensitive to tankyrase inhibitors (COLO-320DM and SW403 cells) highly depend on β-catenin signaling: they exhibit (i) elevated Tcf/LEF transcriptional activities, (ii) higher expression of the downstream genes, and (iii) susceptibility to growth inhibition by β-catenin knockdown. As reported previously (18, 22), we confirmed that either the gain-of-function mutation of the CTNNB1 gene (HCT-116 cells) or the absence of mutations in the Wnt/β-catenin pathway (RKO cells) could be rendered as exclusion criteria for therapeutic application of tankyrase inhibitors in colorectal cancers.

Lau and colleagues classified colorectal cancer cells into either tankyrase inhibitor-“sensitive” or “resistant” according to the β-catenin signaling (i.e., TCF reporter activity and target gene expression) but not cell proliferation (22). They have demonstrated that response of β-catenin signaling to tankyrase inhibitors does not necessarily correlate with the cell sensitivity to the drug-induced growth inhibition or positions of APC mutations. In HCT-15 cells, for example, (i) tankyrase inhibitors blocked β-catenin signaling but did not inhibit the cell growth, (ii) APC knockdown upregulated β-catenin signaling. These results are consistent with our present results (Figs. 1A, 2B and C, 4B). Meanwhile, the level of β-catenin signaling in HCT-15 cells was not so high (Fig. 1B). We assumed that such a low level of β-catenin signaling would not significantly affect the growth of HCT-15 cells, if any. In fact, β-catenin knockdown did not inhibit the growth of HCT-15 cells (Fig. 1D). In total, the observations by Lau and colleagues (22) do not oppose our present conclusion that positions of APC mutations correlate with sensitivities of colorectal cancer cell growth to tankyrase inhibitors.

In colorectal cancer cells, activation of multiple driver pathways, such as Wnt/β-catenin, receptor tyrosine kinase (RTK)/RAS and PI3K/AKT, and p53 inactivation are not mutually exclusive (4). In this study, we did not observe any significant correlation between the sensitivity to tankyrase inhibitors and mutations in PIK3CA, KRAS, BRAF, and TP53, although PIK3CA mutations were correlated with sensitivity to one of the two inhibitors, IWR-1. However, tankyrase inhibitor sensitivities of APC-mutated colorectal cancer cell lines examined here were inversely correlated with their sensitivities to regorafenib, a multitargeted kinase inhibitor (32). Thus, COLO-320DM and SW403 cells are resistant to regorafenib, whereas tankyrase inhibitor–resistant HT-29 cells are sensitive to regorafenib. These observations suggest two hypotheses: (i) responders to tankyrase inhibitors and those to inhibitors of RTK pathways may not overlap; and (ii) colorectal cancer patients who do not respond to either drug may obtain benefit from combination therapy of tankyrase inhibitors and RTK pathway inhibitors.

Tankyrase inhibitor-resistant colorectal cancer cells that had mutant APCs with two or three 20-AARs did not allow Axin-dependent β-catenin degradation caused by tankyrase inhibitors. These long APC mutants would have a dominant negative effect on the Axin-dependent β-catenin degradation because the siRNA-mediated knockdown resumed β-catenin degradation upon exposure to tankyrase inhibitors (Supplementary Fig. S6). Of note, a previous study showed that the long APC mutant (APC1450) forms a hetero-oligomer with the wild-type APC and exerts a dominant-negative effect on APC interaction with a plus-end microtubule-binding protein (EB1) and thus blocks the appropriate dynamics of microtubules (33). As APC requires its carboxyl terminal 284 residues (codons 2560-2843) to interact with EB1 (34), all of the APC-mutated cell lines used in this study should be lacking the EB1–APC interactions. Thus, diverse numbers of retained 20-AARs in APC mutants may not give differential effects on EB1-mediated function of APC, if any.

Importantly, these long APC mutants occur most frequently in colorectal cancers and are hypomorphic (29, 31). In fact, colorectal cancers adopt the so-called “just right signaling” of the Wnt/β-catenin pathway and prefer partially inactivated APC mutants (29, 31). This unique property of APC mutations will provide a second explanation for why the depletion of long APCs restores β-catenin degradation by tankyrase inhibitors: the long APC knockdown derepresses the target gene, AXIN2, and increases its mRNA pool (Fig. 4B and Supplementary Fig. S4B), which further enhances accumulation of Axin2 protein upon treatment with tankyrase inhibitors and facilitates β-catenin degradation. Axins serve as a master scaffold for the β-catenin degradation complex and are the rate-limiting factor of β-catenin degradation (35). In fact, overexpression of Axin in APC-mutated SW480 cells causes a drastic reduction of the β-catenin level (36). These observations support that tankyrase inhibitor-mediated accumulation of Axin induces β-catenin degradation even in the absence of APC, although the degradation efficiency might be reduced by APC dysfunction (22).

COLO-320DM and SW403 cells, in which APC had no 20-AARs, increased Axins and downregulated β-catenin upon treatment with tankyrase inhibitors. Neither overexpression nor knockdown of the short APC affected the Tcf/LEF transcriptional activity or the Axin-dependent degradation of β-catenin. Therefore, these APC mutants are not hypomorphic or dominant-negative in terms of β-catenin regulation and response to tankyrase inhibitors, respectively. We propose that short APC mutants induce the highest activation of the β-catenin signaling and render susceptibility to Axin-dependent degradation of β-catenin, both of which contribute to the sensitivity to tankyrase inhibitors. Consistent with our observations, Waaler and colleagues reported that a tankyrase inhibitor JW55 efficiently blocks polyposis formation in conditional Apc knockout mice (21). These mice produce a short APC (1-580) mutant, which lacks all 20-AARs and is classified as tankyrase inhibitor–sensitive, according to our present study.

Our finding that PDCs with short APC mutants were either sensitive or responsive to tankyrase inhibitors corroborates the idea that APC with complete deletion of seven 20-AARs could be a predictive biomarker for the sensitivity to tankyrase inhibitors. Although the number of established cell lines examined might be small, we examined totally 25 different cells (9 established cell lines and 16 PDCs). Among them, 8 cells were classified as the short APC mutants without 20-AARs. Strikingly, all of these short APC cells responded to two different tankyrase inhibitors, G007-LK and IWR-1 (six were sensitive, two were intermediate, and none were resistant). The antiproliferative effect of tankyrase inhibitors on PDCs seems to be relatively weaker than the established colorectal cancer cell lines. This could be a result of relatively slower growth of the PDCs, particularly under the serum-free culture conditions optimized for PDC culture.

Meanwhile, depletion of the long APCs retaining two or more 20-AARs did not affect the cellular sensitivities to tankyrase inhibitors. These observations are consistent with that HCC2998 and HCT-15 cells were resistant to growth inhibition by β-catenin knockdown (Fig. 1D). In case of DLD-1 cells, tankyrase inhibitor–mediated degradation of β-catenin gave less deleterious effect than direct knockdown of β-catenin (Fig. 1D). This would be due to that tankyrase inhibitors only partially downregulated β-catenin in the mutant APC-depleted DLD-1 cells and the cells still maintained significant amounts of β-catenin (Supplementary Fig. S4A). Accordingly, we propose that APC mutations are potential biomarkers but not functional determinants of tankyrase inhibitor sensitivities of colorectal cancer cells.

We could not define the long APC mutants retaining two or more 20-AARs as exclusion criteria for application of tankyrase inhibitors. APC regulates not only the stability of β-catenin in the destruction complex but also regulates localization of β-catenin with its nuclear localization and nuclear export signals (37, 38). However, there was no general correlation between the nuclear localization of β-catenin and sensitivity to tankyrase inhibitors (H. Yoshida and N. Tanaka, unpublished observations). A recent report showed that APC2 contributes to the activity of the destruction complex upon tankyrase inhibition (39). Upon treatment with tankyrase inhibitors, tankyrase and Axin bind each other to form an oligomer complex, which is important for destabilizing β-catenin (40). These phenomena may provide a key to solve the tankyrase inhibitor sensitivity of PDCs with long APCs. Alternatively, tankyrase inhibitors may exert their therapeutic effect through the blockade of other functions of the proteins, including telomere maintenance, mitosis, downregulation of the tumor suppressor PTEN, and oncogenic YAP activation (41–44).

Although the Wnt/β-catenin pathway has been thought as a less-druggable target for colorectal cancers, tankyrase inhibitors are promising for blocking this signaling pathway. Our demonstration of APC mutations as a predictive biomarker for tankyrase inhibitor sensitivity has a profound implication, because mutations in this tumor suppressor gene are the most fundamental molecular alteration in the complicated etiology of colorectal cancers. Clarifying further mechanisms for the sensitivity to tankyrase inhibitors, especially in the long APC-mutated colorectal cancer cells, will enable the selection of appropriate patients for treating with tankyrase inhibitors and support the development of tankyrase inhibitors.

No potential conflicts of interest were disclosed.

Conception and design: H. Seimiya

Development of methodology: N. Tanaka, B. Gong, R. Katayama, H. Seimiya

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Tanaka, A. Mizutani, A. Sato, A. Aoyama, B. Gong, H. Yoshida, Y. Muramatsu, K. Nakata

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Tanaka, T. Mashima, B. Gong, M. Matsuura, R. Katayama, H. Seimiya

Writing, review, and/or revision of the manuscript: T. Mashima, R. Katayama, S. Nagayama, Y. Sugimoto, H. Seimiya

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Katayama, S. Nagayama, N. Fujita

Study supervision: Y. Sugimoto, H. Seimiya

The authors thank Dr. Fumiyuki Shirai (RIKEN) for kindly providing us with G007-LK. We also thank Dr. Kunitada Shimotohno (National Center for Global Health and Medicine) for providing us with pTcf7mt-luc and pTcfwt-luc reporter vectors.

This work was supported in part by a Grant-in-Aid for Scientific Research (B), Japan Society for the Promotion of Science (no. 16H04716 to H. Seimiya), and a Research Grant from Princess Takamatsu Cancer Research Fund (to H. Seimiya).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.