Tankyrase inhibition sensitizes melanoma to PD-1 immune checkpoint blockade in syngeneic mouse models

The development of immune checkpoint inhibitors represents a major breakthrough in cancer therapy. Nevertheless, a substantial number of patients fail to respond to checkpoint pathway blockade. Evidence for WNT/β-catenin signaling-mediated immune evasion is found in a subset of cancers including melanoma. Currently, there are no therapeutic strategies available for targeting WNT/β-catenin signaling. Here we show that a specific small-molecule tankyrase inhibitor, G007-LK, decreases WNT/β-catenin and YAP signaling in the syngeneic murine B16-F10 and Clone M-3 melanoma models and sensitizes the tumors to anti-PD-1 immune checkpoint therapy. Mechanistically, we demonstrate that the synergistic effect of tankyrase and checkpoint inhibitor treatment is dependent on loss of β-catenin in the tumor cells, anti-PD-1-stimulated infiltration of T cells into the tumor and induction of an IFNγ- and CD8+ T cell-mediated anti-tumor immune response. Our study uncovers a combinatorial therapeutical strategy using tankyrase inhibition to overcome β-catenin-mediated resistance to immune checkpoint blockade in melanoma. Waaler et al. show that a tankyrase inhibitor, G007-LK, decreases WNT/β-catenin and YAP signaling, sensitizing tumors to anti-PD-1 immune checkpoint therapy in mice. This study suggests that a combinatorial therapy using tankyrase inhibition can be used to overcome β-catenin-mediated resistance to immune checkpoint blockade in melanoma.

C ancer immunotherapy is undergoing rapid advances. Treatment of patients using immune checkpoint inhibitors, such as antibodies against cytotoxic T-lymphocyteassociated protein 4 (CTLA-4), programmed cell death 1 (PD-1) and programmed death-ligand 1 (PD-L1) that enhance T cellmediated immune responses against cancer, is considered a major breakthrough 1,2 . However, many cancer patients, including 40-65% of melanoma patients, do not respond to checkpoint inhibitor treatment and the underlying mechanisms are not well understood [1][2][3][4] . Resistance mechanisms are currently being mapped, and cellular signaling pathways and components, such as epidermal growth factor receptor (EGFR), phosphoinositide-3kinase (PI3K)/AKT serine/threonine kinase (AKT), vascular endothelial growth factor (VEGF) and B-Raf proto-oncogene, serine/threonine kinase (BRAF)V600E, as well as Wingless-type mammary tumor virus integration site (WNT)/β-catenin signaling, emerge as promising targets for therapeutical intervention 1,5,6 . WNT/β-catenin signaling can play a central regulatory role in immune cell homeostasis, development and function as well as in peripheral T cell activation, differentiation and tumor-immune cell interplay 7 . β-catenin is the key transcriptional regulator of WNT/β-catenin signaling 8 . β-catenin-induced immune evasion is found in 13% of all tumors 9 and 42% of cutaneous melanoma 10 .
A recent study, using genetically engineered murine melanoma models, revealed that tumors expressing a dominant stable form of β-catenin showed negligible T-cell infiltration and were resistant to checkpoint blockade therapy 11 . In these β-catenin-positive tumors, production of C-C motif chemokine ligand 4 (CCL4) and additional chemokines was reduced. This, at least partly 12 , resulted in reduced recruitment of cluster of differentiation (CD) 103 + / basic leucine zipper transcriptional factor ATF-like 3 (BATF3)-lineage dendritic cells (DC) to the tumor microenvironment and finally defective host priming of antigen-specific T cells 11,13 . In contrast, in melanoma tumors with low levels of βcatenin, DCs and CD8 + T cells migrated into the tumor and the tumor-killing activities of CD8 + T cells could be unleashed or enhanced by the use of checkpoint inhibitors 11 . Hence, interventions that reduce WNT/β-catenin signaling may have the potential to broaden the anti-tumor spectrum of checkpoint inhibitors.
Although dysregulation of WNT signaling is a hallmark characteristic in a major fraction of cancers, anti-cancer therapy that targets this pathway is currently not available in clinical practice 8,[14][15][16] . Target identification and characterization of the small-molecular WNT/β-catenin signaling inhibitor XAV939 revealed that telomeric repeat factor (TRF1)-interacting ankyrinrelated ADP-ribose polymerases 1 and 2 (tankyrase 1 and 2, TNKS1/2) are key and druggable regulatory enzymes in the signaling pathway 17 . TNKS1/2 catalyze the post-translational modification poly(ADP-ribosyl)ation. AXIN1 and AXIN2 proteins are the main rate-limiting structural proteins that together with adenomatous polyposis coli (APC) control the formation of the β-catenin degradosome, which also contains the β-catenintargeting kinase glycogen synthase kinase 3 beta (GSK3β) 18 . TNKS1/2 poly(ADP-ribosyl)ate AXIN proteins to earmark them for degradation by the ubiquitin-proteasomal system 19 . Inhibition of TNKS1/2 can therefore lead to stabilization of AXIN proteins and hence the degradosomes. This, in turn, enhances the phosphorylation and degradation of the central transcriptional regulator β-catenin and inhibits WNT/β-catenin signaling 17,[20][21][22][23] . TNKS1/2 catalytic activity does not only regulate the stability of AXIN proteins, but also interferes with additional biological mechanisms and cell signaling pathways including telomere homeostasis, mitotic spindle formation, vesicle transport, and energy metabolism, as well as AKT/PI3K, AMPK and Hippo signaling 19,[24][25][26] . In Hippo signaling, tankyrase inhibitormediated stabilization of angiomotin (AMOT) proteins shifts the subcellular location of the transcription cofactors yes associated protein 1 (YAP) and tafazzin (TAZ), leading to a reduction of oncogenic YAP signaling 27,28 . Recent reports show that YAP signaling may support immune evasion in cancer and melanoma by inducing PD-L1 expression 29,30 , whereas others show that enhanced YAP signaling, due to loss of the regulating kinases large tumor suppressor kinase 1 and 2 (LATS1/2), may promote an anti-cancer immune response 31 .
Checkpoint inhibitor treatment, including blockade of the PD-1 receptor, has shown limited efficacy in the murine B16-F10 melanoma model, despite strong expression of the ligand PD-L1 on the tumor cells 32 ; a feature attributed to low tumor infiltration by effector CD8 + T cells 33,34 . G007-LK is a potent preclinical stage tankyrase inhibitor with a high selectivity towards tankyrase 1 and 2 and a favorable pharmacokinetic profile in mice with an oral bioavailability of 76% and a t 1/2 of 2.6 hours in female mice 23,35,36 .
Here, we describe G007-LK-mediated blockade of both WNT/ β-catenin and YAP signaling in B16-F10 cells in vitro and in vivo. We show that two murine melanoma models display resistance to monotherapy with either anti-PD-1 or G007-LK. A synergistic anti-tumor effect was observed upon combined anti-PD-1/G007-LK treatment. We show that the mechanistic basis for the synergy was not G007-LK-mediated enhanced release of the BATF3lineage DC-attracting chemokine CCL4 or increased tumor infiltration by CD8 + T cells. Instead, we find that alterations in T cell infiltration are mainly orchestrated by anti-PD-1 treatment alone. Next, we provide evidence that combined anti-PD-1/G007-LK treatment of B16-F10 tumors is effectuated by G007-LKinduced loss of β-catenin in the tumor cells and induction of an interferon-γ (IFNγ)-and CD8 + T cell-dependent anti-tumorimmune response. Finally, upon RNA sequencing of G007-LKtreated human melanoma cell lines and B16-F10 cells, we reveal a transcriptional response profile for a cell line subpopulation displaying high relative baseline YAP signaling activity and predisposition for reduced MITF expression upon tankyrase inhibition.

Results
G007-LK inhibits WNT/β-catenin and YAP signaling. Tankyrase inhibition can inhibit proliferation and viability in a subset of cancer cell lines in vitro 8,25 . When the anti-proliferative effect of G007-LK on cultured B16-F10 mouse melanoma cell line was monitored, only a limited cell growth reduction was observed ( Supplementary Fig. 1a, b). Efficacy of G007-LK treatment on WNT/β-catenin and YAP signaling in B16-F10 cells was then explored in vitro and in vivo. In cell culture, G007-LK-treated B16-F10 cells displayed stabilization of TNKS1/2 and AXIN1 proteins (Fig. 1a, Supplementary Fig. 2a and Supplementary  Fig. 27), as well as formation of cytoplasmic TNKS1/2-containing puncta ( Supplementary Fig. 3), indicating the formation and accumulation of β-catenin degradosomes 22,23,37 .
Next, C57BL/6 N mice with established B16-F10 tumors were treated with G007-LK for four days. This treatment destabilized TNKS1/2 and stabilized AXIN1 protein levels, similar to previous reports 23 , and decreased β-catenin protein levels as well transcription of WNT/β-catenin target genes in the tumors (Fig. 2a, b  These results show that tankyrase inhibitor treatment using G007-LK can attenuate WNT/β-catenin signaling and YAP signaling target gene expression in B16-F10 cells in vitro and in vivo. Synergistic tankyrase and PD-1 inhibition treatment effect. To test whether tankyrase inhibition can counteract resistance to immune checkpoint blockade, B16-F10 tumors were established subcutaneously in C57BL/6 N mice. Neither monotherapy with G007-LK, PD-L1 nor PD-1-blocking antibodies reduced tumor size. However, combined anti-PD-1/G007-LK treatment, but not the anti-PD-L1/G007-LK combination, reduced tumor volume and weight ( Fig. 3a and Supplementary Fig. 10a-e). In addition, the combined anti-PD-1/G007-LK treatment also reduced WNT/ β-catenin signaling and tumor volume of murine Clone M-3 Z1 melanoma in immunocompetent DBA/2 N mice ( To examine longer-term efficacy of combined anti-PD-1/G007-LK treatment, B16-F10-bearing C57BL/6 N mice were followed until the entire control group reached the endpoint criterion. In the three surviving anti-PD-1/G007-LK-treated mice (18.5%) ( Fig. 3c and Supplementary Fig. 12a-d), histopathological evaluation of immunostained tumor sections detected no viable tumor cells. Instead, the tumor implant site was infiltrated by macrophages loaded with melanin, presumably derived from B16-F10 cells (Fig. 3d).
In summary, the tested murine melanoma models are resistant to single-agent anti-PD-1 or G007-LK treatment. In contrast, a synergistic anti-tumor effect and eradication of a subset of the tumors was observed upon combined anti-PD-1/G007-LK treatment.
Tankyrase inhibition alters intratumoral cytokine composition. We next pursued the mechanistic basis for the observed synergy of anti-PD-1 and G007-LK treatment. Previous work in a genetically modified mouse melanoma model have indicated that decreased WNT/β-catenin signaling in the tumor cells promoted adaptive immune responses within the tumor by enhanced secretion of the cytokine CCL4 7,11 . The subsequent chemotaxis of dendritic cells to the tumor site was reported to support infiltration and activation of tumor-reactive CD8 + T cells 11 . To identify alterations in cytokine secretion upon treatment, conditioned supernatants from matrigel-embedded B16-F10 tumors 39 Luciferase-based reporter assay for measuring WNT/β-catenin signaling activity. B16-F10 cells transiently transfected with superTOPflash (vector with TCF promoter binding sites) or FOPflash (control vector with mutated TCF binding sites) along with Renilla luciferase (for normalization). All samples normalized to superTOPflash signal for wild-type control. For b, c Boxplots show median, first and third quartiles and maximum and minimum whiskers. One-tailed t-tests are indicated by *(P < 0.05) or **(P < 0.01). Background SuperTOPflash versus FOPflash activities were not significantly different, indicating low basal WNT/β-catenin signaling activation. One representative experiment of two repeated assays with three replicates is shown. c Real-time RT-qPCR analyses of WNT/β-catenin signaling target genes (Axin2 and Tcf7). One-tailed t-test is indicated by **(P < 0.01) and one-tailed Mann-Whitney rank sum tests are indicated by † (P < 0.05). Combined data from minimum three independent experiments with three replicates each are shown.
were screened using multiplex immunoassays while cell cultures were analyzed using ELISA assays. Few alterations were detectable in anti-PD-1-treated tumors, however, G007-LK treatment was associated with increased levels of three cytokines and decreased levels of five cytokines (>30% difference compared to control) ( Fig. 4a and Supplementary Fig. 13). Similar changes in cytokine secretion were apparent upon anti-PD-1/G007-LK treatment ( Fig. 4a and Supplementary Fig. 13). Notably, G007-LK treatment reduced CCL4 levels in B16-F10 tumors ( Fig. 4b) but no reduction was detected in cultured B16-F10 cells (Fig. 4c). Ccl4 transcript was not inversely correlated to its previously described negative regulator activating transcription factor 3 (Atf3) 11 in either wild-type B16-F10 cells or B16-F10 Ctnnb1KO cells when compared to wild-type cells (Fig. 4d, e).
In conclusion, the results suggest that G007-LK treatment, but not anti-PD-1 treatment, mainly alters the intratumoral cytokine composition. The herein observed anti-PD-1/G007-LK-induced anti-tumor effect cannot be attributed to enhanced CCL4 secretion.
The treatment effect depends on β-catenin in tumor, IFNγ, and CD8 + T cells. A knockout of β-catenin in B16-F10 cells can serve as a model to recapitulate G007-LK-mediated blockade of WNT signaling. To evaluate β-catenin-mediated immune evasion in the B16-F10 syngeneic mouse melanoma model, β-catenin was knocked out in B16-F10 cells (B16-F10 Ctnnb1KO , Supplementary  Fig. 5a, b) and one of the cell lines, B16-F10 Ctnnb1KO1 , was used to establish subcutaneous tumors in C57BL/6 N mice. Compared to vehicle control mice, anti-PD-1-treated mice displayed reduction in tumor size, indicating loss of anti-PD-1 resistance in β-catenindeficient tumors ( Fig. 5a and Supplementary Fig. 14a-d). The result suggests that the synergistic anti-PD-1/G007-LK treatment effect seen in wild-type B16-F10 tumors (Fig. 3a) is, to a considerable part, attributed to G007-LK-induced reduction of βcatenin levels in the melanoma cells of the tumors.
To evaluate if the observed effects of anti-PD-1/G007-LK treatment are mediated by an adaptive immune response, B16-F10 challenge experiments were repeated in recombinasedeficient (Rag2 −/− ) mice, which lack functional T and B cells, but possess functionally intact natural killer (NK) cells 40 . No significant effect of anti-PD-1/G007-LK treatment was observed in such mice ( Fig. 5b and Supplementary Fig. 15a, b). Although, a contributory role of NK cells and the innate immune system cannot be entirely excluded, the result indicates that the anti-PD-1/G007-LK treatment effect is orchestrated by an adaptive immune response.
Next, we further evaluated the adaptive immune response by performing selective elimination of either CD8 + T cells or IFNγ. Antibody-mediated CD8 + T cell depletion abrogated the antitumor effect of anti-PD-1/G007-LK treatment ( Fig. 5c and Supplementary Fig. 16a, b). Similarly, neutralization of IFNγ resulted in increased tumor growth comparable to anti-PD-1/G007-LK -treated mice ( Fig. 5c and Supplementary Fig. 16a, b). In summary, these results confirm a role of CD8 + T cells and IFNγ as mediators of the synergistic effect of anti-PD-1/G007-LK treatment.
To assess immune cell infiltration upon treatment, we next performed flow cytometry analysis using tumors of comparable size collected on day 7-17 (Supplementary Figs. 17a and 18a, b). No increase in tumor leukocyte abundance was observed in any of the treatment groups when compared to the control group ( Supplementary Fig. 17b). An increase in total T cell and CD8 + T cell infiltration was seen in both the anti-PD-1 and anti-PD-1/ G007-LK groups, whereas CD4 + T cells were similarly increased across all treatment groups ( Fig. 5d and Supplementary Fig. 17c). No differences in the abundance of CD4 + or CD8 + T cells expressing the memory marker CD44 were detected (Supplementary Fig. 17d). T regs constituted approximately 1% of infiltrating CD45 + cells in all groups ( Supplementary Fig. 17e). Myeloid DCs were decreased in the anti-PD-1 and anti-PD-1/ G007-LK-treated groups while CD103 + DCs were equally present in all treatment groups ( Supplementary Fig. 17f). Lymphoid DCs, myeloid-derived suppressor cells (M-MDSC) and neutrophils were present at comparable levels across all treatment groups ( Supplementary Fig. 17g).
Subsequently, infiltration of CD8 + T cells in tumors sections was scored using treated wild-type and β-catenin knock-out B16-F10 tumors. Only anti-PD-1 and anti-PD-1/G007-LK-treated B16-F10 Ctnnb1KO tumors showed an increase in CD8 + T cell infiltration when compared with controls ( Fig. 5e and Supplementary Fig. 19). The result suggests that only PD-1 inhibition, and not loss of β-catenin in the tumor cells, contributes to increased chemotaxis of CD8 + T cells.
To evaluate direct effects of G007-LK on T cell effector functions, we assayed the in vitro activation of MHC class I (H2-K b )-restricted and ovalbumin-specific CD8 + T cells. The presence of G007-LK moderately enhanced T cell proliferation . Two-tailed t-tests are indicated by *(P < 0.05) and two-tailed Mann-Whitney rank sum tests are indicated by ‡ (P < 0.01). b Real-time RT-qPCR analyses of WNT/β-catenin signaling target genes (Axin2 and Tcf7). One-tailed t-test is indicated by **(P < 0.01) and one-tailed Mann-Whitney rank sum test is indicated by ‡ (P < 0.01). Boxplots show median, first and third quartiles and maximum and minimum whiskers for combined data from two independent measurements with three replicates each are shown.
following cognate interaction with antigen-presenting cells, but did not affect proliferation following polyclonal activation by immobilized anti-CD3/anti-CD28 antibodies ( Supplementary  Fig. 20a). The presence of G007-LK did not affect the expression of T cell activation markers (CD26L, CD44, CD25, and CD69) or secretion of interleukin 2 (IL2) or IFNγ (Supplementary Fig. 20b-d). The addition of G007-LK induced a slight increase in intracellular granzyme B expression in CD8 + T cells following cognate or polyclonal activation, possibly indicating increased effector function (Supplementary Fig. 20e).
In summary, combined anti-PD-1/G007-LK treatment of B16-F10 tumors is dependent on tankyrase inhibitor-mediated loss of β-catenin in the tumor and induces an IFNγ and CD8 + T celldependent growth-inhibitory effect. Changes in myeloid DC and T cell infiltration are likely attributable to anti-PD-1 treatment, but do not alone seem to cause anti-tumor activity.
RNA sequencing reveals a subpopulation transcriptional response profile. The efficiency of tankyrase inhibitor-mediated inhibition of WNT/β-catenin and YAP signaling is known to be cell type and context-dependent. The beneficial synergistic anti-PD-1/G007-LK treatment effect seen on immunosurveillance in B16-F10 tumors may therefore vary between melanomas based on differences in cell signaling pathway activities and genetic background. In addition, the here utilized B16-F10 murine melanoma model, lacking the BRAF V600E mutation 41,42 , only partially recapitulates the genetic features of human melanoma. Thus, murine B16-F10 cells and a panel of 18 human melanoma cell lines, were exposed to G007-LK treatment followed by RNA sequencing and bioinformatic analyses.

Fig. 22) against overall gene expression (Supplementary
Upon treatment with G007-LK, samples in the MITF low group were predisposed for decreased MITF expression (MITF decreased ), while the MITF high group was oppositely predisposed for increased MITF expression (MITF increased )( Fig. 6b and Supplementary Fig. 24a). Notably, in B16-F10 tumors, transcription of Mitf was moderately reduced upon G007-LK-treatment (Fig. 6c). Attenuated WNT/β-catenin and/or YAP signaling activity was observed in nearly all samples upon G007-LK treatment (Fig. 6d). These changes in signaling pathway activities did not correlate with changes in MITF expression, nor act as a predictive marker for MITF regulation ( Fig. 6e and Supplementary Figs. 24b-e and 25a, b). In conclusion, YAP high is a marker for a melanoma subgroup that includes B16-F10 (Fig. 6e) and tracks with tankyrase inhibitor-induced reduction in MITF expression (Supplementary Fig. 26 and Supplementary Table 3).

Discussion
Here we show, for the first time to our knowledge, proof-ofconcept results for a previously unreported therapeutic strategy using tankyrase inhibitor-mediated blockade of WNT/β-catenin signaling to counteract β-catenin-supported immune evasion and resistance to checkpoint inhibition in syngeneic murine melanoma models.
The G007-LK treatment effect could be recapitulated in vivo by knockout of β-catenin in B16-F10 tumor cells. Although not further supported by a rescue experiment re-introducing Nterminal mutant β-catenin, the result suggests that loss of WNT/ β-catenin signaling in tumor cells is sufficient to trigger a gain of susceptibility and a synergetic relationship to checkpoint inhibition. Nevertheless, we cannot exclude the possibilities that For d, e Combined data from minimum three independent experiments with three replicates each are shown. Two-tailed t-test is indicated by **(P < 0.01), two-tailed Mann-Whitney rank sum test is indicated by ‡ (P < 0.01) and n.s. not significant. e Real-time RT-qPCR analyses of Atf3 and Ccl4 from cultured B16-F10 Ctnnb1KO cells compared to wild-type B16-F10 cells.
tankyrase inhibition impacts additional biological mechanisms in the tumor cells, in the tumor microenvironment, systemically or on immune cell subpopulations that can contribute to the treatment effect 7,19,24,25 . In addition to its effect on WNT/β-catenin signaling, G007-LK treatment stabilized AMOT proteins and reduced YAP signaling-mediated gene expression in B16-F10 cell culture and tumors. An absence of reduced nuclear YAP and TAZ protein levels, as also seen in HEK293 cells, along with the here observed aggregation of colocalized YAP, AMOT proteins and TNKS1/2-containing puncta, is inconsistent with previous reports indicating that further evaluation of tankyrase inhibitor-induced interference with YAP signaling is necessary 27,37,38 .
Our results provide evidence that the anti-tumor effect of combined tankyrase and checkpoint inhibitor treatment against B16-F10 tumors is dependent on both IFNγ and CD8 + T cells and occurs, at least in part, through direct suppression of WNT/ β-catenin signaling within the tumor cells. The precise mechanistic basis of this synergistic effect remains to be investigated. Previous work performed in genetically modified models indicate that lowered WNT/β-catenin signaling 11 resulted in increased intratumoral CCL4 cytokine release, which lead to enhanced influx of CD103 + /BATF3 + pDCs and subsequent activation of CD8 + T cell anti-tumor activity. Our results indicate that G007-LK treatment indeed could change the intratumoral cytokine composition; however, we found no evidence for increased CCL4 cytokine levels. In addition, assessment of the intratumoral immune response following G007-LK monotherapy did not yield significant quantitative alterations in T cell or antigen-presenting cell (APC) subsets in treated mice. While it cannot be excluded that qualitative or more subtle quantitative changes in particular tumor-specific CD8 + T cell subsets might be responsible for the observed therapeutic effects, an increase in overall CD8 + T cell infiltration was not observed. Only a moderate enhancement of proliferation and granzyme B expression was observed in G007-LK-treated MHC class I (H2-K b )-restricted and ovalbuminspecific CD8 + T cells, which could indicate enhanced effector function. In depth follow-up studies in the framework of tankyrase inhibition are required, including assessment of the effects of alterations in individual cytokine and chemokine levels on the T cell response 12 . The use of model systems with defined and traceable anti-tumor CD8 + T cell specificities would also be of value in determining the effects of tankyrase inhibition on the kinetics and qualitative characteristics of the ensuing immune response.
The RNA sequencing of 18 tankyrase inhibitor-treated human melanoma cell lines and B16-F10 cells shows that tankyrase inhibition context-dependently can influence WNT/β-catenin and YAP signaling, not only in murine B16-F10 cells, but importantly also in a subset of 18 human melanoma cell lines.
The results indicate that tankyrase inhibition may have a potential for treatment of a subset of patients with the human disease in combination with checkpoint inhibition. Analysis of the RNA sequencing data revealed a subgroup-specific transcriptional response profile. Upon tankyrase inhibition, the subgroup displaying elevated baseline YAP signaling activity was susceptible to reduced MITF expression. Presently, neither the meeting points between MITF, its regulation by YAP/ TEA domain transcription factor (TEAD), activator protein 1 (AP-1) Fig. 6 High activity of YAP signaling correlates with low baseline MITF expression and potential for decreased MITF transcription upon tankyrase inhibition. a Expression of YAP signaling target transcripts (Ccn1, Ccn2, and Amotl2) for untreated human and murine B16-F10 melanoma cell lines. Seven of 19 samples displayed high relative transcription of YAP signaling target genes (YAP high ) when compared to samples with less transcription (YAP low ). YAP high is highlighted by orange branches in the dendrogram and arrows. Scale bar indicates differences in Z-score (standard deviations < or > mean) values for log2 transcripts per millions (TPMs) within each row. Cancer stages: Metastatic in blue, radial growth phase (RGP) in pink and vertical growth phase (VGP) in green. b Baseline MITF expression in untreated samples (grey bars, log2-transformed TPMs × 10 −1 ) and change upon treatment with G007-LK (1 µM) for 24 h (black dots sorted descending from left to right, log2 values from treated versus untreated TPMs). Samples with increased (MITF increased , left) or decreased (MITF decreased , right) expression of MITF upon tankyrase inhibitor treatment are separated by scattered vertical line. *Indicates that no value for untreated sample is inserted. For b, d B16-F10 WNT3a = WNT3a + G007-LK relative to WNT3a-stimulated control. YAP high cell lines are highlighted by orange arrows. c Real-time RT-qPCR analysis of Mitf expression in G007-LK-treated B16-F10 s.c. tumors versus control tumors (n = 8). Two-tailed Mann-Whitney rank sum test is indicated by ‡ (P < 0.01). Boxplot shows median, first and third quartiles and maximum and minimum whiskers for data from two repeated measurements. d Changes in gene expression for WNT/β-catenin (Axin2 and Tcf7) and YAP (Ccn1, Ccn2, and Amotl2) signaling target genes for 18 G007-LK-treated (1 µM) human and murine B16-F10 melanoma cell lines. Downregulated signaling activity is indicated by ↓ (log2 < −0.2, highlighted in blue), upregulated activity is indicated by ↑ (log2 > 0.2, highlighted in pink), and inconclusive or lack of regulation is indicated by − . e Venn diagram depicting the cell lines intersecting profiles for pre-treatment (YAP high and MITF low ) and response to G007-LK treatment (MITF decreased and YAP or WNT decreased ). and tankyrase, nor the function in immune regulation and control of susceptibility to checkpoint-inhibitor therapy are well characterized [46][47][48][49][50][51] . Further molecular and functional evaluations are required for establishing a precise set of rules for contextdependent treatment efficiency using combined tankyrase and checkpoint inhibitor treatment in human melanoma. Translation of the combinatorial therapeutical strategy, using tankyrase inhibition to counteract β-catenin-induced resistance to immune checkpoint blockade, should also be evaluated for treatment efficacy against other cancers 6,7,9 . Tankyrase inhibitors have been suspected to cause intestinal toxicity 23 and bone loss 52 in certain mouse models. In this study, and similar to previous reports 22,36,53 , we observed no signs of toxicity, intestinal injury or body weight changes in G007-LKtreated mice. G007-LK is a preclinical stage tankyrase inhibitor without the properties needed to reach approval for clinical testing 35 . Currently, extensive research to identify and further develop safe drugs directed towards multiple biotargets in the WNT signaling pathway, including tankyrase, are ongoing 8,14-16 .
Collectively, the results presented here suggest that G007-LKinduced blockade of WNT/β-catenin signaling leads to improved efficacy of PD-1 immune checkpoint blockade, and in addition induction of an IFNγ and CD8 + T cell-dependent anti-tumorimmune response against B16-F10 tumors. The findings warrant a further in-depth preclinical and clinical evaluation of combining checkpoint inhibitors with tankyrase inhibition for the treatment of melanoma.
Proliferation assays. 1000 cells/well were seeded in 96-well plates in at least 6 replicates for each treatment tested. The day after, cell culture medium was changed to contain various doses of G007-LK or vehicle (DMSO, Sigma Aldrich) and the plates were incubated in an IncuCyte (FLR30140, Essen BioScience) at 37°C for real-time monitoring of cell confluency. At experiment endpoint (80-100% confluency after 5-7 days of cell growth), the cells were incubated for 1 h at 37°C with CellTiter 96 ® AQueous Non-Radioactive Cell Proliferation Assay (MTS, Promega) according to the supplier's recommendations. Abs 490 was measured spectrophotometrically (Wallac 1420 Victor2 Microplate Reader, Perkin Elmer) and compared to Abs 490 (t0) using the following formula to determine single well values relative to the DMSO vehicle control: (sample A 490 − average A 490 t0 )×100/ (average A 490 [for 0.01% DMSO controls] − average A 490 t0 ).
RNA isolation and real-time qRT-PCR. Total RNA was isolated from cell lines and tumor samples using GenEluteTM Mammalian Total RNA Miniprep Kit (Sigma Aldrich). The RNA concentration was measured using Nanodrop 2000c spectrophotometer (Thermo Scientific). cDNA was synthesized from the purified RNA using SuperScript™ VILO cDNA Synthesis Kit (Invitrogen). Real-time qRT-PCR (TaqMan®Gene Expression system, Applied Biosystems) was performed using Viia7 (Applied Biosystems). The following probes were used (all from Applied Immunofluorescence, SIM and confocal microscopy. Cells grown on coverslips pre-coated with poly-L-lysine (sc-286689, Santa Cruz Biotechnology) were fixed in 4% paraformaldehyde (P6148, Sigma Aldrich) for 15 min at room temperature and permeabilized with 0.1% Triton-X100/PBS (T8787, Sigma Aldrich, 15 min at room temperature) followed by 1 hour (at room temperature for SIM images) or 24 h (at 4°C for confocal images) incubations with primary and secondary (antibodies 1 h at room temperature) diluted in PBS with 4% bovine serum albumin. Nuclear counterstaining was performed with DAPI (D9542, Sigma Aldrich, 1 μg/mL, 5 min at room temperature) and coverslips were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher Scientific). The following primary antibodies were used: β-catenin (610153, 1:500, BD Biosciences), Tankyrase SIM images were acquired on a Zeiss Elyra PS1 microscope system using standard filters sets and laser lines with a Plan-APOCHROMAT 63 × 1.4 NA oil objective. SIM imaging was performed using 5 grid rotations with the 0.51 µm grid for 20 Z planes with 0.184 nm spacing between planes. SIM images were reconstructed with the following "Method" parameters in the ZEN black software (MicroImaging, Carl Zeiss): Processing: Manual, Noise Filter: -5, SR Frequency Weighting: 1, Baseline Cut, Sectioning: 100/83/83, Output: SR-SIM, PSF: Theoretical. The SIM images are displayed as maximum intensity projections rendered from all Z planes.
Confocal microscopy was performed on a Zeiss Meta 700 laser scanning confocal microscope using standard filters sets and laser lines with a 63 × oil immersion objective, images were acquired using Zen software (Zeiss). The confocal images were analyzed using Fiji software 57 . Confocal images are displayed as maximum intensity projections rendered from two Z-stacks at the nucleous region with 0.56 µm spacing between stacks.  35,53 . Compound delivery was monitored via food consumption. The chow was weighed thrice weekly to calculate G007-LK delivery per animal and day. Several animals had to be euthanized before experiment endpoint in all animal experiments performed, primarily due to skin ulcerations in the tumor area seen in all treatment groups, and less often due to anemia or excessive tumor size (tumor diameter ≥ 2 cm). The experiment endpoints were defined to a minimum of 8 or 9 surviving animals in any given treatment group.
For experiments using B16-F10 Ctnnb1 knock-out cells in C57BL/6 N mice: Intraperitoneal injections of anti-PD-1 were administered on day 11, 15, and 18. For experiments using B16-F10 cells in C57BL/6 N mice: Intraperitoneal injections of anti-PD-1 or anti-PD-L1 were administered on day 10, 13, 17 and every 3-4 days from day 21 until the end for the survival analysis. For experiments in Rag2 mice or for Clone M-3(Z1) in DBA/2 N mice: Intraperitoneal injections of anti-PD-1 were administered on day 8, 11, and 15. Primary tumors were measured by a caliper (OMC Fontana) and tumor sizes calculated according to the formula W 2 × L/2 (L = length and W = width).
Assay for the efficacy of G007-LK-mediated inhibition of tankyrase and WNT/ β-catenin and YAP signaling pathways using B16-F10 cells in C57BL/6 N mice: Mice with primary tumor volumes of 20-40 mm 3 were randomized into two groups (control or G007-LK for four nights, n = 8). Whole tumor protein extracts were prepared by sonication in an ultrasound bath (Bioruptor® Plus, Diagenode) in RIPA buffer (Thermo Fisher Scientific) containing phosphatase and protease inhibitors (PhosStop, 4906837001 and cOmplete™ Protease Inhibitor Cocktail, 4693116001, both from Roche). Tumor tissues were homogenized using MagNA Lyser Green Beads (Roche) and total mRNA isolated using GenEluteTM Mammalian Total RNA Miniprep Kit (Sigma Aldrich).
Tumor growth assay using B16-F10 cells in C57BL/6 N mice: On day 10, 72 tumor-bearing animals (primary tumors reaching 20-100 mm 3 ) were randomized into six groups (n = 12), and treatments (i)-(vi) were administered to the mice. On day 21, the experiment was terminated, and tumors from euthanized mice were dissected, weighed and volumes re-measured with caliper. For 6 euthanized animals from treatment groups (i), (ii), (iii), and (v): small intestinal was dissected, formalin-fixed (10%), paraffin-embedded in a Swiss roll configuration and sections were stained with H&E (115938 and 115935, Merck Millipore). 1, 2, 1, 4, 3, and 2 mice from treatment groups (i), (ii), (iii), (iv), and (v), respectively, were found dead or euthanized for ethical reasons during the experiment due to anemia or skin ulcerations in the tumor area.
Clone M-3(Z1) tumor growth assay: On day 8, 48 tumor-bearing animals (with primary tumors reaching~26 mm 3 ) were randomized into 4 groups (n = 12) and treatments (i), (ii), (iii), and (v) were administered to the respective mice. On day 18, the animals were euthanized and the tumors were dissected, weighed and volumes re-measured with caliper. 4, 1, 3, and 1 mice from treatment groups (i), (ii), (iii), and (v), respectively, were euthanized for ethical reasons during the experiment due to skin ulcerations in the tumor area.
Survival assay using B16-F10 cells in C57BL/6 N mice: 60 animals with primary tumors reaching 20-100 mm 3 were randomized into 2 groups (n = 30, control or G007-LK and anti-PD-1). The survival assay endpoint criterion was set to tumor volume >1000 mm 3 . Eleven mice in the control group and 14 mice in the treatment group were euthanized (for ethical reasons) or found dead due to anemia or skin ulcerations in the tumor area before reaching the endpoint criterion. Tumors were collected from surviving and treated animals at endpoint on day 38, and from control animals euthanized on day 26. The tumors were dissected, fixed (10% formalin), embedded in paraffin, sectioned and stained with H&E (115938 and 115935, Merck Millipore). For immunostaining, 2.5 µm sections were boiled for 20 min in 10 mM citrate buffer pH 6.0 and incubated with primary antibody anti-F4/ 80 (2 µg/mL, clone CI: A3, ab6640, Abcam) diluted in PBS with 1.25% BSA overnight at 4°C, and then incubated with fluorescently labeled secondary antibody (donkey anti-rat IgG Alexa Fluor 488, 5 µg/mL, A-21208, Molecular Probes) for 60 min at 37°C. Hoechst 33258 nuclear dye (0.5 µg/mL, Sigma Aldrich) was added to the final washing solution. Pictures were captured using a Nikon Eclipse model N i-U microscope (Nikon) equipped with Nikon Plan-Fluor objective lenses and an Infinity 2 digital camera (Lumenera Corporation).
Tumor growth assay using B16-F10 Ctnnb1 knock-out cells in C57BL/6 N mice: On day 11, 48 tumor-bearing animals (primary tumors reaching 20-100 mm 3 ) were randomized into 4 groups (n = 12), each treated with (i), (ii), (iii), or (v). On day 25, mice were euthanized, and the tumors were dissected, weighed and volumes re-measured with caliper. Three mice from treatment group (i) and two mice from treatment groups (ii), (iii) and (v) were euthanized for ethical reasons before experiment termination, due to skin ulcerations in the tumor area or excessive tumor size. For immunostaining, 2.5 µm frozen sections were incubated with primary antibody anti-CD8 (5 µg/mL, clone 4SM15, eBioscience) diluted in PBS with 1.25% BSA for 60 min at 37°C and then incubated with secondary antibody (donkey anti-rat IgG Alexa Fluor 488, 5 µg/mL, Thermo Fisher Scientific) for 60 min at 37°C. Hoechst 33258 nuclear dye (0.5 µg/mL, Sigma Aldrich) was added to the final washing solution. For semi-quantitative analysis, the three areas (high power fields) from each tumor section that contained the highest number of CD8 + T cells were selected and the number of T cells were counted in each area. The mean value of the three scored areas in each tumor was calculated and next used in statistical analyses.
Tumor growth assay using B16-F10 cells in B6.129S6-Rag2tm1Fwa N12 mice (Rag2 −/− ) 40 mice: On day 8, 24 tumor-bearing animals (with primary tumors reaching 18-61mm 3 ) were randomized into two groups (n = 12), and treatments (i) and (v) were administered to the respective mice. On day 20, the animals were euthanized. One mouse from treatment group (i) and two mice from treatment group (v) were euthanized for ethical reasons before experiment termination, due to skin ulcerations in the tumor area or excessive tumor size.
All animal experiment described in this section were performed by ProQinase GmbH, following approval by local animal experiment authorities (Freiburg, Germany) and in compliance with FELASA guidelines and recommendations.
Multiplex and ELISA immunoassay. Subcutaneous B16-F10 tumors were implanted (left flank, 0.2 × 10 6 B16-F10 tumor cells suspended in 200 µl [2:1] PBS: matrigel mixture) in 6-8 weeks old female C57BL/6 mice (Taconic). On day 6, animals were distributed into four groups and treatments (i) (n = 3), (ii) (n = 6), (iii) (n = 5), and (v) (n = 8) were administered to the mice (see paragraph describing tumor initiation, treatment and analysis of mouse melanoma in vivo). Intraperitoneal injections of anti-PD-1 were administered on day 6, 9, and 13. On day 14, the animals were euthanized and the matrigel-implanted tumors were dissected, treated with 1 mg/mL type IV collagenase and 0.3 mg/mL DNase I (both from Sigma Aldrich) in RPMI and incubated for 45 min at 37°C. The samples were next centrifuged (1500 × g) and the conditioned supernatants collected 39 . The animal experiment was approved by local animal experiment authorities (Norwegian Food Safety Authority, Norway) and in compliance with FELASA guidelines and recommendations. Bio-Plex Pro Mouse Chemokine Panel 33-plex (12002231, Bio-Rad) was used to screen the in vivo tumor conditioned supernatants for cytokine and chemokine expression according to the manufacturer's protocol. A Bio-Plex handheld magnetic washer (Bio-Rad) was used for the wash steps, and a Luminex-100 instrument with Bio-Plex Manager 4.1 software (Bio-Rad) was used for analysis. Cell supernatants from treated cells and conditioned B16-F10 tumor supernatants (see description for multiplex immunoassay) were assayed using Mouse CCL4/MIP-1 β Quantikine ELISA Kit (MMB00, R&D Systems) according to the manufacturer's protocol.
Tumor flow cytometry analysis. Primary tumor from treatment groups (i), (ii), (iii), and (v) was collected and processed for flow cytometry analysis to determine the presence of subpopulations of T cells and myeloid-derived suppressor cells (carried out by ProQinase 58,59 ). For analysis of T cells and myeloid-derived suppressor cells, animals were treated for 7-17 days to obtain similarly distributed primary tumor volumes ranging from 80-240 mm 3 (Supplementary Fig. 17a). The animal experiment was performed by ProQinase GmbH, following approval by local animal experiment authorities (Freiburg, Germany) and in compliance with Thereafter, 100 µl 1X permeabilization buffer (00-5523-00, eBioscience) was added and the cells were centrifuged at 400 × g. The cell pellet was resuspended in 1X permeabilization buffer containing anti-forkhead box P3 (FoxP3) antibody (FoxP3-PE [FJK-16s], 12-5773-82, eBioscience) and incubated for 30 minutes in the dark. After washing twice with permeabilization buffer, the cells were washed with FACS buffer and kept at 4°C in the dark until analysis. The samples were analyzed by flow cytometry using an LSR Fortessa (BD Biosciences) and the gating strategy is shown in Supplementary Fig. 18a, b.
Cells were harvested after 24 additional hours and radiolabeling was counted with a MicroBeta plate reader. Supernatant was harvested from the plates after 72 h of treatment. IL2 and IFNγ quantitation in culture supernatants was determined using ELISA Max Deluxe kits (431004 and 430804, Biolegend). For flow cytometric analysis of surface markers on APC/ SIINFEKL or ConA-activated CD8 + T cells, the cells were pooled from 8-12 wells from the setup described previously and the following antibodies were used: Anti-CD62L (1705-09 L, Southern biotech), anti-CD69 (1715-02, Southern biotech), anti-CD8a (553036, BD Biosciences), anti-CD25 (17-0251-82, eBioscience), anti-CD44 (103049, Biolegend), and anti-CD3e (35-0031, TONBO Biosciences). For intracellular staining, the cells were incubated for 4 h in protein-transport inhibitor before permeabilization and fixation, according to the manufactures' protocol (Fix/Perm kit with Golgistop 554715, BD Biosciences) and then stained with anti-granzyme B (12-8898-82, eBioscience). All antibodies were used at a working concentration of 2 µg/ml. The samples were analyzed using the Attune NxT flow cytometer (Thermo Fisher Scientific) and Flow Jo software (BD Biosciences). DNA sequencing. Kinome targeted re-sequencing of the 18 human melanoma cell lines was performed using the SureSelect Human Kinome kit (Agilent Technologies), with capture probes targeting 3.2 Mb of the human genome, including exons and untranslated regions (UTRs) of all known kinases and selected cancer-related genes (to a total of 612 genes). Library construction and in solution capturing was performed following Agilent's SureSelectXT library construction kit and SureSelect Target enrichment protocol, respectively. Sequencing was performed on an Illumina HiSeq2500 using the TruSeq SBS Kit v5 generating paired-end reads of 75 bp in length. Base calling, de-multiplexing and quality filtering was performed using Illumina's software packages SCS2.8/RTA1.8 and Off-line Basecaller-v1.8.
Bioinformatics. Transcripts were quantified with kallisto (v0.44) 60 using ensembl transcriptome release 91 for human (GRCh38) and release 92 for mouse (GRCm38) 61 . Ensemble BioMart was used to map human orthologs in mouse 60 . Differentially expressed genes (DEGs) were identified with sleuth (v0.29) 62 , limma (3.34.9) 63 (did not result in any comparisons with adjusted P values < 0.05) and DESeq2 64 in the R programming environment (The R Project for Statistical Computing). The R-package NMF (0.23.6) 65 was used to make hierarchical clusters with TPM values as input. For detection of probable driver mutations, the RNAseq data was aligned with HISAT2 (v2.1.0) 66 before VarDict (v1.2) 67 restricted to SNVs reported more than once in COSMIC (v82) 68 was applied. Due to the variable coverage in RNAseq data, additional SNVs found in an external unpublished gene panel sequencing experiment for the same cell lines, were included. Variants found in both datasets were annotated using ANNOVAR (2017-07-17) 69 . The data analysis was performed by the Bioinformatics Core Facility (Oslo University Hospital, Norway). DEGs analysis data, including log2-fold change and adjusted pvalues, were uploaded into Ingenuity Pathway Analysis (IPA) version 01-10 (Qiagen). The DEGs analysis data were analyzed using the core analysis function with the Ingenuity Knowledge Base (genes only) reference set and direct relationships, with no filters set for node types, data sources, confidence, species, tissues, and cell lines and mutations. For the IPA core analyses, the log2-fold and or adjusted p-value cutoffs are specified in the figure legends.
Statistics and reproducibility. No sample size calculation was performed. Sample sizes for both in vivo and in vitro experiments were determined based on experiment experience, pilots and preliminary experiments as well as what was reported in the literature. Samples sizes for each experiment and numbers of independent repeats are indicated in figure legends. All in vivo experiments included contain ≥8 independent biological replicates, except for Fig. 4a, b (≥3) and Fig. 5e (≥3). For all in vitro assays, all attempts at replication were successful through repeated experiments (two or more replications). Sigma Plot ® 12.5 (Systat Software Inc.) was used to perform statistical tests: Student's t-test for comparisons with homogeneous variances (Shapiro-Wilk test, P > 0.05) and Mann-Whitney rank sum tests for comparisons where the normality assumption was violated (Shapiro-Wilk test, P < 0.05). GrapPad Prism 7 was used for Kaplan-Meyer estimations and statistical analysis. Single outlier detections were identified by Dixon's and/or Grubb's tests (threshold, P < 0.05) using Con-trolFreak (Contchart software).
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The RNA sequencing data for this study are available from ArrayExpress with accession numbers E-MTAB-8438 (human cell lines) and E-MTAB-8101 (B16-F10 mouse cell line). For the mouse experiment, the data are deposited both as raw fastq files and processed as RNA abundance counts. For the human RNA experiment only abundance counts are deposited. Somatic mutations from DNA sequencing are available at https:// github.com/ous-uio-bioinfo-core/waaler-et-al-2020; 10.5281/zenodo.3703045. Fastq files for the human sequencing experiments and additional data generated and analyzed in this study are available from the corresponding author upon request. The source data underlying plots shown in main figures are provided in Supplementary Data 1. Additional data generated and analyzed in this study are available from the corresponding author upon request.

Code availability
Custom scripts used to process and analyze the sequencing data and to make most of the related figures and tables are available at https://doi.org/10.5281/zenodo.3703045.