Identification of response signatures for tankyrase inhibitor treatment in tumor cell lines

Summary Small-molecule tankyrase 1 and tankyrase 2 (TNKS1/2) inhibitors are effective antitumor agents in selected tumor cell lines and mouse models. Here, we characterized the response signatures and the in-depth mechanisms for the antiproliferative effect of tankyrase inhibition (TNKSi). The TNKS1/2-specific inhibitor G007-LK was used to screen 537 human tumor cell lines and a panel of particularly TNKSi-sensitive tumor cell lines was identified. Transcriptome, proteome, and bioinformatic analyses revealed the overall TNKSi-induced response signatures in the selected panel. TNKSi-mediated inhibition of wingless-type mammary tumor virus integration site/β-catenin, yes-associated protein 1 (YAP), and phosphatidylinositol-4,5-bisphosphate 3-kinase/AKT signaling was validated and correlated with lost expression of the key oncogene MYC and impaired cell growth. Moreover, we show that TNKSi induces accumulation of TNKS1/2-containing β-catenin degradasomes functioning as core complexes interacting with YAP and angiomotin proteins during attenuation of YAP signaling. These findings provide a contextual and mechanistic framework for using TNKSi in anticancer treatment that warrants further comprehensive preclinical and clinical evaluations.

The master transcriptional regulator MYC proto-oncogene (MYC) is deregulated in >50% of human cancers, in line with a central function in controlling a multitude of oncogenic processes including differentiation, proliferation, and apoptosis (Chen et al., 2018). WNT/b-catenin, YAP, and PI3K/AKT signaling pathways are all promoters of MYC transcription in cancer cells (He et al., 1998;Huh et al., 2019;Kress et al., 2015;Neto-Silva et al., 2010).
The vast majority of studies on the antitumor effect of TNKSi focus on the impact on individual signaling pathways rather than examining the overall downstream biological effects of TNKSi. Here, we used the TNKS1/2-selective inhibitor G007-LK to screen 537 tumor cell lines for an antiproliferative effect and identified a subset highly TNKSi-sensitive cell lines originating from the colon, kidney, ovary, and lung. In this subset, functional and molecular analyses revealed that TNKSi can context-dependently antagonize the oncogenic signaling pathways WNT/b-catenin, YAP, and PI3K/AKT leading to an impediment of MYC-driven cell growth.

RESULTS
Proliferation screen identifies human tumor cell lines susceptible to growth inhibition by the selective tankyrase inhibitor G007-LK  (Lau et al., 2013) and in comparison with RKO control cells (blue). (C) Re-screening using endpoint MTS proliferation assay (Abs 492 ) for the indicated concentrations of G007-LK for 4-8 days relative to control (100%, 0.01% DMSO) and experiment time 0 values (t 0 , set to 0%). One-way ANOVA tests (Holm-Sidak method versus control) are indicated by *** (p < 0.001) and one-way ANOVA on ranks tests (Dunn's method versus control) are indicated by y (p < 0.05). Mean values GSD for one representative experiment of at least two repeated assays, each with six replicates, are shown. (D) Endpoint MTS proliferation assay GI 50-values.

OPEN ACCESS
iScience 24, 102807, July 23, 2021 3 iScience Article screened against a panel of 537 human tumor cell lines, including the NCI-60 tumor cell line panel. These human tumor cell lines originated from 29 different tissues bearing various primary diagnoses. The concentrations of G007-LK treatment that inhibited cell growth by 25% or 50% (GI 25 and GI 50 values) were determined. Out of the 537 tested tumor cell lines, 87 (16%) displayed GI 25 values < 1 mM G007-LK. These included >20% of the cancer cell lines originating from the kidney, ovary, stomach, liver, pancreas and lung ( Figure 1A and Table S1). The screening results suggest that TNKSi obstructed the growth of a broad range of cancer types in vitro.
The proliferation screen identified three cell lines that were particularly susceptible to the growth-inhibitory effect of G007-LK (TNKSi-sensitive) with GI 50 values < 200 nM. These three cell lines, UO-31 (renal cancer), OVCAR-4 (ovarian cancer), and ABC-1 (non-small-cell lung cancer), along with the previously identified benchmark TNKSi-sensitive cell line COLO 320DM (Lau et al., 2013) (colon cancer), were submitted to subsequent analyses to identify the mechanisms that render them particularly sensitive to TNKSi ( Figure 1B and Table S1). The TNKSi-insensitive colon cancer cell line RKO was included as a negative control (Mizutani et al., 2018;Solberg et al., 2018;Tanaka et al., 2017). To verify the screening data, the panel was subsequently rescreened. In the retested cell lines, G007-LK significantly decreased cell growth, as measured by colorimetric MTS viability (GI 50 values: 54-844 nM) and colony assays (42-66% reduction), in all TNKSi-sensitive cell lines, while control RKO cells remained unaffected by the treatment (Figures 1C,  1D, 1E, and S1A).
Next, cell cycle and apoptosis analyses were performed to further investigate TNKSi-induced cell growth inhibition. In the panel of TNKSi-sensitive cell lines, only ABC-1 cells exhibited significant G 1 cell cycle arrest and apoptosis upon G007-LK treatment ( Figures 1F, 1G, S1B, and S1C). By contrast, RNA sequencing and real-time qRT-PCR analyses revealed significantly reduced transcripts of the key cell-cycle-promoting genes MYC and cyclin D1 (CCND1), as well as MYC and cyclin D1 protein in all selected TNKSi-sensitive cell lines after G007-LK treatment, but not in RKO control cells, suggesting that TNKSi is unable to block MYC expression in RKO cells ( Figures 1H-1J and S1D). Finally, to examine whether decreased MYC expression can impair cell growth, the selected cell panel was transfected with siRNA against MYC. Knockdown of MYC, to recapitulate the G007-LK-mediated reduction of MYC protein, resulted in a significant inhibition of cell growth in all cell lines, also in RKO control cells ( Figure 5K and S1E).
The results indicate that all tested cell lines depend on expression of MYC for sustained cell proliferation.
In conclusion, the results suggest that TNKSi decreases MYC and cyclin D1 expression leading to induction of cytotoxic G 1 cell cycle arrest and apoptosis in ABC-1 cells, while overall slower cell cycle progression is the primary cause of the cytostatic cell growth inhibition observed in COLO 320DM, UO-31, and OVCAR-4 cells. Continued (E) Relative colony numbers (%) upon 7-11 days of treatment with G007-LK (1 mM) compared to DMSO (0.01%). Mean values GSD for combined data from a minimum of three independent experiments with three replicates each are shown. For E-G, I, and K, two-tailed t-tests are indicated by *** (p < 0.001), ** (p < 0.01), and * (p < 0.05) while Mann-Whitney ranksum tests are indicated by z (p < 0.01) and y (p < 0.05).
(F) Cell cycle alteration relative to control (%). G = gap 1 phase, S = synthesis phase, G2/M = gap 2 /mitosis phase. Mean values from combined data consisting of a minimum of four independent experiments are shown. For F and G, upon 72-h treatment with G007-LK (1 mM) compared control (set to 0%, 0.01% DMSO). The effect of TNKSi against tumor cell proliferation may depend on the tumor type, mutation load, the context in which the tumor cells are grown, and the intrinsic activities of various cellular pathways. To map changes in gene expression, proteome, and cell signaling pathways and to correlate them with oncogenic mutations, the selected cell line panel was exposed to G007-LK treatment followed by RNA sequencing, bioinformatic analyses, and proteomics analyses.
First, a mutation analysis of the RNA sequencing data set was performed by matching mutations identified in the selected cell lines against a set of previously defined driver oncogenes (Bailey et al., 2018). However, apart from the relative abundant mutations in TNKSi-insensitive RKO cells, no telltale mutational patterns unifying the 5 cell lines were identified (Table S2). Moreover, when comparing the RNA sequencing data, a principal component analysis revealed highly different pretreatment and post-treatment transcriptional profiles ( Figure 2A).  Table S3). For three out of the four TNKSi-sensitive cell lines, activities of four proteins were predicted to be upregulated upstream regulators after G007-LK treatment: Tumor protein p53 (TP53), nuclear protein 1, transcriptional regulator (NUPR1), tumor protein p73 (TP73), and BRCA1 DNA repair associated. Three proteins were predicted to be downregulated upstream regulators: MYC-associated factor X (MAX), YAP1, and Sp1 transcription factor (SP1) (Figures 2D and S2B and Table S3). Several of these identified upstream regulator proteins are known to control apoptosis and cell cycle, such as TP53 ( Transcription of MYC can be regulated by several signaling pathways that contain tankyrase target proteins including WNT/b-catenin, YAP, and PI3K/AKT signaling (He et al., 1998;Huh et al., 2019;Kress et al., 2015;Neto-Silva et al., 2010). Within these pathways, CTNNB1 (b-catenin) was predicted to be a downregulated upstream regulator in COLO 320DM and OVCAR-4 cells upon G007-LK treatment ( Figures 2D, 2E, and S2B and Table S3). Several of the upstream regulator proteins, predicted in the IPA core analysis, are associated with WNT/b-catenin signaling activity: SP1 is regulated by the b-catenin destruction complex (Mir et al., 2018), ESR1 is involved in cross talk with WNT/b-catenin signaling, while forkhead box O3 (FOXO3) can bind and interact with b-catenin (Kouzmenko et al., 2004;Valenta et al., 2012). YAP1 was predicted to be a downregulated upstream regulator in UO-31, OVCAR-4, and ABC-1 cells after exposure to G007-LK (Figures 2D and S2B and Table S3), while NUPR1 transcription is controlled by YAP signaling (Jia et al., 2016). FOXO3, a central effector of PI3K/AKT signaling (Stefanetti et al., 2018), was predicted to be an upregulated upstream regulator in UO-31 and ABC-1 cells ( Figures 2D, 2E, and S2B and Table S3). In a previous report, decreased NOTCH signaling was observed in a proteome analysis of TNKS1/2 knockout HEK293 cells (Bhardwaj et al., 2017). However, predictions of NOTCH1 activity were outside the threshold level used in the IPA core analysis for all cell lines (Table S3), and no distinct downregulation of NOTCH signaling target genes was observed in any of the cell lines ( Figure S2C).
Finally, an SILAC-based proteome analysis identified 590-847 statistically significant differently expressed proteins in the selected TNKSi-sensitive cells, while 501 proteins were found in RKO cells ( Figure S2D and Table S4). No robust clustering classifying the cell line's protein expression profiles was observed (Figure S2E). However, among upregulated proteins, the energy metabolism-regulating proteins transketolase, NADH:ubiquinone oxidoreductase subunit A8, and hydroxyacyl-CoA dehydrogenase trifunctional multienzyme complex subunit beta were identified in a minimum of four of the cell lines after G007-LK treatment (Table S4). Previous reports have shown TNKSi-mediated regulation of energy metabolism in mouse models Zhong et al., 2016aZhong et al., , 2016b.

OPEN ACCESS
iScience 24, 102807, July 23, 2021 5 iScience Article In conclusion, the analysis of transcriptional responses to G007-LK exposure indicates a repertoire of rather diverse regulation of signaling pathways in TNKSi-sensitive tumor cell lines. Nevertheless, TNKSi predominantly leads to cell type-dependent and primary inhibition of the WNT/b-catenin, YAP, and PI3K/AKT signaling pathways and, subsequently, counteraction of MYC-driven cell cycle progression and tumor cell growth ( Figure 2E). Hence, the effect of G007-LK against these three signaling pathways was further investigated in the selected cell line panel. To assess endogenous WNT/b-catenin signaling pathway activities, the cell line panel was transiently cotransfected with a vector containing WNT/b-catenin signaling-responsive promoter driving firefly luciferase expression (superTOPflash), or control vector (FOPflash), along with Renilla luciferase for normalization. COLO 320DM cells (APC mutated ) demonstrated high luciferase activity compared to RKO cells (APC wild-type ), indicating high endogenous WNT/b-catenin signaling activity ( Figure 3E). OVCAR-4 and ABC-1 cells showed moderate but significant increases in superTOPflash signal when compared to the FOPflash signal, suggesting rather low endogenous WNT/b-catenin signaling activities ( Figure 3E). In stable superTOPflash and Renilla luciferase transfectants, a decrease in WNT/b-catenin signaling activity was only seen in COLO 320DM cells exposed to various doses of G007-LK ( Figures 3F and S4C). Although transcription of AXIN2, a cell type-universal and negative-feedback-controlling target gene, was significantly reduced in COLO 320DM, OVCAR-4, and ABC-1 cells, RNA sequencing analyses revealed that transcription of a panel of WNT/b-catenin signaling target genes was reduced predominantly in COLO 320DM cells ( Figures 3E and 3G).
In APC-mutated colorectal cancer cells, TNKSi resulted in the accumulation of cytoplasmic puncta and bcatenin degradasomes containing TNKS1/2, AXIN1/2, APC, GSK3b, and b-catenin (Thorvaldsen et al., 2015). Hence, to gain further knowledge regarding b-catenin degradasome accumulation in the selected cell line panel, structured illumination microscopy imaging was performed to visualize TNKS1/2 and b-catenin upon G007-LK treatment. Decreased accumulation of nuclear b-catenin, and formation of distinct cytoplasmic puncta with colocalized TNKS1/2 and b-catenin (Lau et al., 2013;Thorvaldsen et al., 2015;Waaler et al., 2012), was only observed in APC-mutated COLO 320DM cells with high endogenous WNT/b-catenin signaling activity and expression of AXIN2 protein (Thorvaldsen et al., 2017) (Figure 4). In contrast, b-catenin localization, primarily found in the cell membrane, did not change in the other Figure 3. Continued controls (0.01% DMSO). Actin (cytoplasmic) and lamin B1 (nuclear) document protein loading. Representative data from two or more independent experiments are shown.
(E) Luciferase-based reporter assay for comparing baseline WNT/b-catenin signaling activity. The cells were transiently co-transfected with either a superTOPflash (vector with 7 X TCF promoter binding sites driving the firefly luciferase) or a FOPflash (control vector with mutated TCF binding sites) along with Renilla luciferase (for normalization). All samples are relative to normalized superTOPflash signal for RKO cells (= 1). Mean values GSD for combined data from 2-4 independent experiments with three replicates each are shown. Statistically significant differences between SuperTOPflash and FOPflash activities (TOP/FOP ratio) are indicated. iScience Article TNKSi-sensitive cell lines ( Figure 4). Instead, TNKS1/2 puncta were found in proximity to the cell membrane after treatment in UO-31, OVCAR-4, and ABC-1 cells (Figure 4). In RKO cells, TNKS1/2 accumulated in juxtanuclear puncta (Figure 4).
In conclusion, the results imply that only COLO 320DM and OVCAR-4 cells are dependent on WNT/b-catenin signaling for sustained cell proliferation, while UO-31 and ABC-1 cells show resistance to b-catenin knockdown. WNT/b-catenin signaling activity is robustly decreased by G007-LK treatment in COLO 320DM cells and modestly in OVCAR-4 and ABC-1 cells. TNKS1/2-and b-catenin-containing puncta are found in the cytoplasm in APC-mutated COLO 320DM cells, but close to the cell membrane in UO-31, OV-CAR-4, and ABC-1 cells.

G007-LK inhibits YAP signaling in the selected cell line panel and all cell lines depend on YAP for sustained proliferation
The IPA core analysis predicted YAP1 as a TNKSi-attenuated upstream regulator, suggestive for decreased YAP signaling in UO-31, OVCAR-4, and ABC-1 cells ( Figure 2). To assess whether decreased YAP signaling can impair cell growth, the selected cell panel was transfected with siRNA against YAP. Knockdown of YAP, to imitate G007-LK-mediated reduction of YAP signaling, resulted in a significant inhibition of cell growth in all cell lines ( Figure 5A).
To evaluate the effect of G007-LK on YAP signaling, the selected cell line panel was first examined by Western blot analysis. Treatment of each cell line with G007-LK stabilized AMOT, AMOTL1, and AMOTL2 proteins in both cytoplasmic and nuclear extracts ( Figures 5B and S5A), consistent with earlier reports using HEK293T cells , Nuclear YAP accumulation was enhanced in UO-31, OVCAR-4, and ABC-1 cells, similar to recent observations (Kierulf-Vieira et al., 2020;Waaler et al., 2020b), while no change in nuclear YAP levels were observed in COLO 320DM or RKO cells. Moreover, cytoplasmic YAP was not affected in any cell lines following TNKSi ( Figure S5A). The results are in contrast with previous publications showing lowered levels of nuclear YAP upon TNKSi (Wang et al., , 2016 (Figures 5B and S5A).
Although no reduction in nuclear YAP levels was observed in the selected cell line panel subjected to TNKSi, RNA sequencing analyses showed that transcription of a panel of YAP signaling target genes was decreased, in all TNKSi-sensitive cell lines and to a lesser extent in RKO cells ( Figure 5C). Real-time qRT-PCR analysis showed reduced transcription of the YAP signaling target genes CCN1 (previously named CYR61), CCN2 (previously named CTGF), and AMOTL2 in all cell lines ( Figure 5D). A moderate and significant reduction in YAP signaling luciferase reporter activity was seen in only COLO 320DM, UO-31, and ABC-1 cells ( Figure S5B).
Stabilization of AMOT proteins exposure to TNKSi (Troilo et al., 2016;Wang et al., 2015), and localization of YAP in the degradosome (Azzolin et al., 2014), have previously been described. To obtain additional information regarding the localization of AMOT proteins, YAP, and TNKS1/2 in the selected cell line panel, confocal imaging was next performed.
In summary, the results show that all cell lines are dependent on YAP signaling for sustained cell growth. G007-LK targets YAP signaling in all cell lines, through a mechanism involving TNKS1/2-mediated and AMOT protein-dependent sequestering and inactivation of transcriptionally active YAP protein. iScience Article G007-LK inhibits PI3K/AKT signaling in ABC-1 cells that depend on PI3K/AKT signaling for sustained cell growth The IPA core analysis predicted FOXO3 to be a TNKSi-augmented upstream regulator in UO-31 and ABC-1 cells, indicative of TNKSi-dependent decreased PI3K/AKT signaling ( Figure 2). However, the RNA sequencing analysis revealed that the transcriptional profile for activated FOXO signaling was most apparently regulated in ABC-1 cells ( Figure 7A). Moreover, Western blot analysis was performed to test the effect of G007-LK treatment on PI3K/AKT signaling in the selected cell lines. Reduced presence of the activated and phosphorylated forms of AKT, indicating blocked PI3K/AKT signaling, was only seen in ABC-1 cells ( Figures 7B and S7A).
In a previous report, TNKSi stabilized PTEN in colorectal cancer cell lines leading to downregulation of AKT phosphorylation and its signaling activity . In contrast, PTEN was not stabilized in any of the cell lines after treatment, suggesting the presence of a yet-to-be-explored and PTEN-independent mechanism behind TNKSi-induced AKT signaling inactivation ( Figure 7B). Earlier publications describe crosstalk signaling between the PI3K/AKT and AMPK signaling pathways (Han et al., 2018;Hawley et al., 2014), and TNKSi has been implicated in regulation of AMPK activation (Li et al., 2019). Hence, the activation status of iScience Article AMPK in the cell line panel upon G007-LK treatment was evaluated by Western blot analysis. In contradiction with the previous report (Li et al., 2019), our analysis showed that the activated phosphorylated form of AMPK was not upregulated in any of the cell lines by G007-LK exposure ( Figure S7B). Instead, AMPK activity was decreased in both UO-31 and ABC-1 cells after TNKSi ( Figure S7B). We could therefore not explain the TNKSi-induced AKT inactivation observed only in ABC-1 cells (Han et al., 2018;Hawley et al., 2014). To identify the precise mechanism for TNKSi-induced PI3K/AKT signaling inhibition will require further investigation.
To evaluate if PI3K/AKT signaling is essential for continued cell growth, ABC-1 cells were treated with PI3K and AKT inhibitors. Both inhibitors dose-dependently decreased the active and phosphorylated form of AKT while reducing cell growth ( Figure 7C). In conclusion, the data show that G007-LK can target PI3K/ AKT signaling in ABC-1 cells that are dependent on PI3K/AKT signaling for continuous cell growth.

DISCUSSION
Despite more than a decade of research, much remains unclear about the molecular profiles that render tumor cells sensitive or insensitive to the antiproliferative effect of TNKSi. Here, we describe a broad tumor cell line screen, classifying 84% of the cell lines to be TNKSi nonresponders (GI 25 values > 1 mM G007-LK) and 16% to be TNKSi responders (GI 25 values < 1 mM G007-LK) including 1.9% to be highly TNKSi-responding tumor cell lines (GI 50 values < 1 mM G007-LK). >20% of the tumor cell lines originating from the kidney, ovary, stomach, liver, pancreas, and lung were defined as TNKSi responders, suggesting that these types of tumor cell lines are most sensitive to the antiproliferative effect of G007-LK treatment. In general, the result indicates effectiveness of TNKSi against cell growth in subtypes of cancer across several tumor types. $85-90% of colon cancer cell lines contain mutations in APC resulting in aberrant activation of WNT/b-catenin signaling (Fearon, 2011). Thus, colon cancer was previously regarded a particularly relevant target for TNKSi (Lau et al., 2013). TNKSi-responsiveness in colorectal cancer has been shown to depend on the APC mutation genotype (Schatoff et al., 2019). Yet, only two out of 41 colon cancer cell lines tested in our screen displayed GI 25 values < 1 mM G007-LK.
From the 1.9% highly TNKSi-responding tumor cell lines, a panel of heterogeneous cell lines was selected for further analysis to identify mechanisms coupling TNKSi to attenuated proliferation. In these TNKSi-sensitive cell lines, we used bioinformatics analysis of RNA sequencing data and proteome profiles to characterize TNKSi effects and identified a variety of changes in response signatures. Numerous post-treatment upstream signaling regulators were predicted to be cell-type-dependently controlled by TNKSi according to the IPA core analysis, warranting detailed follow-up studies. While the involvement of additional signaling pathways cannot be excluded, the overall analysis suggests that the main primary events caused by TNKSi in the particular sensitive cell lines are a downregulation of WNT/b-catenin, YAP, and PI3K/AKT signaling pathways followed by decreased MYC expression.
Validation experiments showed that TNKSi (i) blocked WNT/b-catenin signaling in COLO 320DM, OVCAR-4, and ABC-1 cells; (ii) YAP signaling in all cell lines; and (iii) AKT signaling in ABC-1 cells ( Figure 7D). Moreover, TNKSi-mediated downregulation of these pathways correlated with lost expression of MYC and CCND1, suggesting that downregulation of these two proteins is a shared hallmark of all tested TNKSi-sensitive cell lines (He et al., 1998;Huh et al., 2019;Kress et al., 2015;Neto-Silva et al., 2010). In line with this notion, functional analyses of TNKSi-mediated cell cycle arrest and apoptosis revealed the induction of a cytostatic effect in all TNKSi-sensitive cell lines, with the exception of ABC-1 cells. In ABC-1 cells, TNKSi stimulated G 1 cell-cycle arrest, apoptosis and a cytotoxic antiproliferative effect. Moreover, (i) b-catenin knockdown could recapitulate the antiproliferative effect of TNKSi treatment on COLO 320DM and OVCAR-4 cells, (ii) YAP knockdown blocked the growth of all cell lines, whereas inhibition of (iii) PI3K/AKT signaling inhibited the proliferation of only ABC-1 cells (Figure 7D). Notably, combined TNKSi and PI3K/AKT inhibition showed additive antitumor effects in mouse colon cancer models (Arques et al., 2016;Solberg et al., 2018). Collectively, our results suggest that TNKSi-induced inhibition of WNT/b-catenin and YAP signaling, either individually or together, can cause a cytostatic effect, Figure 6. Effect of G007-LK treatment on the localization of YAP, TNKS1/2, and AMOTL2 in tumor cell lines Immunofluorescence staining and representative confocal images of YAP (red) and TNKS1/2 (green), YAP (red) and AMOTL2 (green), or AMOTL2 (red) and TNKS1/2 (green), along with nuclear DAPI staining (blue) upon vehicle control (0.01% DMSO) and G007-LK (1 mM) treatment (24 h) of the indicated cell lines. Red, antimouse antibody used. Green, antirabbit antibody used. Arrowheads indicate colocalizations. Scale bars = 20 mm. See also Figures 5, S5, and S6.

OPEN ACCESS
iScience 24, 102807, July 23, 2021 13 iScience Article while additional inhibition of PI3K/AKT signaling can provoke an additional cytotoxic effect. Whether similar combinatorial effects against signaling pathways can be projected onto a larger group of tumor cell lines remains to be investigated.
TNKSi stabilized AMOT, AMOTL1, and AMOTL2 in the cytoplasm and the nucleus of all cell lines. In parallel, YAP accumulated in nuclear extracts only in treated UO-31, OVCAR-4, and ABC-1 cells, but not COLO 320DM and RKO cells. YAP signaling target gene expression was reduced in all cell lines. Together, these observations are in line with recent reports that TNKSi induced accumulation of nuclear YAP correlating with reduced YAP target gene expression (Kierulf-Vieira et al., 2020;Waaler et al., 2020b). However, the observations are at odds with earlier reports that TNKSi induced a reduction of nuclear YAP leading to reduced YAP target gene expression (Wang et al., , 2016. When performing imaging, nuclear AMOTL2 levels were close to, or below the detection threshold, whereas the distribution of nuclear and cytoplasmic YAP was detected regardless of TNKSi. However, in iScience Article UO-31, OVCAR-4, and ABC-1 cells, puncta containing TNKS1/2-YAP, TNKS1/2-AMOTL2, and AMOTL2-YAP were found near the cell membrane following TNKSi. The results suggest that TNKS1/2-containing b-catenin degradasomes (Thorvaldsen et al., 2015) not only can physically interact with YAP, as previously suggested (Azzolin et al., 2014), but also function as complexes containing AMOT proteins . In contrast, the imaging of COLO 320DM cells revealed the formation of TNKSi-induced cytoplasmic TNKS1/2-puncta that colocalized with b-catenin, indicating b-catenin degradasome accumulation (Lau et al., 2013;Thorvaldsen et al., 2015;Waaler et al., 2012). However, in these cells, AMOTL2 and YAP colocalized with each other but not with TNKS1/2. Hence, similar to a previously proposed model , our observations suggest that AMOT proteins sequester YAP independent of TNKS1/2-containing b-catenin degradasomes in COLO 320DM cells. Notably, the APC-mutated cell line COLO 320DM, displaying a high endogenous WNT/b-catenin signaling activity, expresses higher levels of AXIN2 protein compared to the other non-APC-mutated cell lines in the selected panel. In a previous report, loss of expression of AXIN2, but not AXIN1, was associated with disintegration of TNKSi-induced cytoplasmic puncta (Thorvaldsen et al., 2017). The precise mechanism for TNKSi-dependent regulation of YAP signaling, and the association with the b-catenin degradasome, is currently under investigation.
In summary, the results provide evidence that TNKSi treatment is effective against subtypes of cancer cell lines across several tumor types. In four identified TNKSi-sensitive cell lines, TNKSi functions by contextdependent targeting of multiple signaling pathways including WNT/b-catenin, YAP and/or PI3K/AKT signaling, followed by loss of MYC expression and the induction of either cytostatic or cytotoxic effects, culminating in impaired tumor cell growth. These findings warrant further comprehensive preclinical and clinical evaluation of TNKSi as monotherapy or combination therapy for cancer.

Limitations of study
Our study identified several TNKSi-sensitive tumor cell lines, and the downstream in-depth analysis focused on only a small subset of highly sensitive cell lines originating from multiple tissues. These cell lines contain highly diverse oncogenic mutations, gene and protein expression profiles as well as cell signaling pathway activities, and as a consequence, prediction and identification of shared pretreatment and posttreatment markers was influenced. Numerous TNKSi-induced changes in activities of signaling pathways, in addition to WNT/b-catenin, YAP, and PI3K/AKT signaling pathways, were predicted and detailed follow-up studies are clearly needed. The experiments provide only a limited description of TNKSi-dependent regulation of YAP signaling, and the results contradict with previous descriptions of TNKSi-regulated PTEN/ PI3K/AKT and AMPK signaling pathways activities, indicating that further research is needed to identify the precise mechanisms involved.

Materials availability
Materials generated in this study can be made available upon request to the Lead Contact.

Data and code availability
Datasets and codes generated during this study are available at repositories indicated in the key resources iScience Article METHOD DETAILS

Human tumor cell line anti-proliferative screens
Compound screening and data analysis carried out by Genentech was performed similar to a previous description (Haverty et al., 2016). Briefly, a collection of cancer cell lines obtained from a variety of academic and commercial sources, such as ATCC and Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures, was used. Cell line identity was routinely verified by high-throughput single nucleotide polymorphism genotyping using multiplexed assays (Yu et al., 2015). All cell lines were cultured using standard tissue culture techniques and maintained in RPMI-1640 (31800, GIBCO), 2 mM glutamine (Kyowa Hakko Bio), 10% FBS (F4135, Sigma) for suspension cell lines, and 5% FBS for adherent cell lines, in a humidified incubator maintained at 37 C and 5% CO 2 . Cells were plated in 384-well plates (353962, Corning) at optimal seeding density to achieve 75% confluency at 96 hours monitored using Incucyte for live cell imaging (4647, Essen Bioscience). Optimal seeding for suspension and mix suspension/adherent cell lines was determined by 75% maximal signal at 96 hours using CellTiter-Gloâ Luminescent Cell Viability Assay (G7573, Promega). The day after, cell culture medium was changed to medium containing nine drug concentrations (using three to four replicates) or vehicle control (dimethylsulfoxide, DMSO, D2650, Sigma). After 72 hours, 25 mL CellTiter-Gloâ reagent was added to the wells and luminescence readout was measured using a 2104 EnVision Multilabel Plate Reader (2105-0010, PerkinElmer). Data were processed using R (R Core Team, 2012) and a Genentech-developed analysis package (singleAgentPlots, Dr. Richard Bourgon). Absolute GI 25 and GI 50 values were calculated relative to the corresponding vehicle control.
G007-LK was screened against the NCI-60 tumor cell line panel using their standard protocol: Briefly, all cell lines were grown in RPMI 1640 medium containing 5% FBS and 2 mM L-glutamine at 37 C and 5% CO 2 . 5,000-40,000 cells/well were seeded in 96-well plates depending on the cell's doubling speeds. The day after, cell culture medium was changed to medium containing five drug concentrations plus control. After an additional 48 hours, adherent cells were fixed in situ by the addition of 50 ml of cold 50 % (w/v) TCA (final concentration, 10 %) and incubated for 60 minutes at 4 C. The supernatant was discarded, and the plates were washed five times with water and air dried. Sulforhodamine B (SRB) solution (100 ml) at 0.4 % (w/v) in 1 % acetic acid was added to each well, and the plates were incubated for 10 minutes at room temperature. After staining, cells were rinsed five times with 1 % acetic acid and the plates were air dried. Bound SRB was subsequently solubilized with 10 mM trizma base, and the absorbance was read on an automated plate reader at a wavelength of 515 nm. For suspension cells, the methodology was the same except that the cells were allowed to settle to the bottom of the well before gently adding 50 ml of 80 % TCA (final concentration 16 %

Treatment with small-molecule inhibitors
All small-molecule inhibitors were dissolved in DMSO (D8418, Sigma-Aldrich) and kept as 10 mM stocks at 4 o C. General protocol for treatment: Cells were seeded one day before treatment to reach $20 or $80% confluence for a 72 hour or a 24 hour treatment, respectively. The medium was changed to medium containing vehicle (0.01% DMSO), 1 mM or various doses of the tankyrase inhibitor G007-LK (Mercachem), PI3K inhibitor BKM120 (CT-BKM120, Chemietek) or AKT inhibitor API-2 (2151, Tocris Bioscience).
Proliferation assays 5,000 (COLO 320DM), 2,500 (UO-31, OVCAR-4, ABC-1) or 1,000 (RKO) cells/well were seeded in 96-well plates in at least 6 replicates for each treatment tested, also for seeding of esiRNA transfected cells. For samples for treatment with small-molecule inhibitors, the cell culture medium was changed the day after to contain various doses of the indicated inhibitors or 0.01% DMSO. The plates were incubated at 37 C ll OPEN ACCESS iScience Article the human UniProt database (October 2014 version) supplemented with contaminants. The applied parameters were: enzyme: trypsin/P; variable modifications: oxidation (M), acetyl (protein N-term), Phospho (STY), hydroxyproline and deamidation (NQ); labels: Arg10, Lys6, max. peptide PEP: 0.1; min. peptide length: 7; min. unique peptides: 1; advanced: re-quantity, keep low-scoring versions of identified peptides, match between runs (0.7 min time window), label-free quantitation, and second peptide MS2 identification. Otherwise, the default parameters of MaxQuant were used. Normalized L/H abundance ratios were calculated in MaxQuant, and the ratios of treated versus control for each cell line were analyzed for differential protein abundance using NOISeq analysis in R. Volcano plots of differential protein abundance were generated in R.

QUANTIFICATION AND STATISTICAL ANALYSIS
Sigma Plotâ 12.5 (Systat Software Inc.) was used for all statistical analyses with the exception of bioinformatics analyses using NOISeq, IPA and R. Single outlier detections were identified by Dixon's and/or Grubb's tests (threshold, P < 0.05) using ControlFreak (Contchart software). Detailed descriptions of statistical tests used, description of the number of events (n) as well as depictions of mean and standard deviations can be found in the figure legends and figures. The minimum significance level was defined as P < 0.05. The two-tailed Student's t-test was used to test for significant differences (*** [P < 0.001], ** [P < 0.01] and * [P < 0.05]) between two samples with normally distributed parameters (Shapiro-Wilk test, P > 0.05). Mann-Whitney tests were used to test for significant differences (z [P < 0.01] and y [P < 0.05]) between two samples with parameters that were not normally distributed. One way ANOVA tests (Holm-Sidak method versus control) were used to test for significant differences (*** [P < 0.001], ** [P < 0.01] and * [P < 0.05]) between multiple samples with normally distributed parameters (Shapiro-Wilk test, P > 0.05). One way ANOVA on ranks tests (Dunn's method versus control) were used to test for significant differences (z [P < 0.01] and y [P < 0.05]) between multiple samples with parameters that were not normally distributed. NOISeq analysis was performed in R. Sample sizes were determined based on experiment experience, pilots experiments as well as what was reported in the literature. For the NOISeq analysis, the default probability value of > 0.8 was considered significant.