YAP1-TAZ/TEAD transcriptional networks restrain differentiation downstream of oncogenic Hedgehog-SMO activity

Disruption of the transcriptional activity of the Hippo pathway members YAP1 and TAZ has become a major target for cancer treatment. However, detailed analysis of the effectivity and networks affected by YAP1/TAZ transcriptional targeting are limited. Here, by comparing the effects of YAP1/TAZ knockdown with those resulting from TEAD blockage, we unveil the consequences of YAP1/TAZ transcriptional inhibition in cancer cells. We utilize TEADi, an inhibitor of the binding of YAP1 and TAZ with their main transcriptional target TEAD. In a mouse model of basal cell carcinoma (BCC) driven by the smoothened oncogene (SmoM2), TEADi and YAP1/TAZ knockdown lead to reduced proliferation and increased differentiation of tumor cells both in vitro and in vivo. We find that TEAD transcriptional networks inactivate differentiation in BCC by regulating KLF4. Furthermore, we determine YAP1/TAZ TEAD-independent effects in cancer cells that impact Stat3 and NF-κB gene networks. Our results reveal the TEAD dependent and independent roles of YAP1/TAZ in cancer and expose potential pitfalls for targeting TEAD transcription in tumors.


Introduction
The Hippo pathway is one of the main signaling networks altered in human cancer 1 , with genomic and signaling changes in Hippo components leading to activation and nuclear translocation of YAP1 and TAZ (WWTR1). YAP1 and TAZ are paralog proteins that interact with a wide range of cytoplasmic and nuclear transcriptional effectors, ultimately modulating the proliferative and differentiation status of cells in response to chemical and mechanical microenvironmental cues 2,3 .
In the skin, YAP1 and TAZ are activated in basal (BCC) and squamous (SCC) cell carcinoma 4 .
Interestingly, while YAP1 epidermal knockout leads to a reduction in tumor lesions in a mouse model of BCC 5 , a more profound effect is observed with the concomitant downregulation of both YAP1 and TAZ in mouse BCC and SCC tumors 6 . This evidence indicates that disruption of both YAP1 and TAZ signaling might be necessary to achieve significant responses for BCC and SCC treatment. One common YAP1 and TAZ downstream pathway is the activation of TEAD transcription factors and efforts are underway to develop YAP1/TAZ-TEAD interaction inhibitors that could be used for cancer and other hyperproliferative diseases. However, a major challenge in studying the effectivity of this approach is the lack of preclinical models to characterize the consequences of TEAD inhibition.
Precise evidence of the effectiveness of targeting YAP1/TAZ-TEAD has been limited to the use of verteporfin 7 and peptide inhibitors 8,9 . One problem with this approach is that verteporfin has YAP1/TAZ-TEAD independent effects 10,11 , and drugs and soluble peptide inhibitors do not provide cellular and tissue level information. On the other hand, studies of the role of YAP1/TAZ in cancer involve the knockout of these proteins, affecting TEAD-dependent events and other transcriptional and signaling components that interact with YAP1 and TAZ 12,13 . This complicates the precise understanding of TEAD functions and masks the potential differences between absence versus blockage of YAP1/TAZ proteins. An additional tool to study TEAD specificity are rescue experiments with the YAP1 S94/F95 mutant that does not bind to TEAD 14 , although YAP1 and TAZ have to be knocked-down or knocked-out for these assays and the rescue leads to YAP1 overexpression.
To circumvent some of the limitations to study TEAD inhibition in cells and tissues our group developed TEADi, a genetically encoded fluorescently traceable dominant negative protein that blocks nuclear interaction of TEAD with YAP1 and TAZ 15 . TEADi presents several advantages, including rapid inhibition of TEAD transcription and concomitant blockage of both YAP1 and TAZ without altering structural or cytoplasmic functions of these proteins. TEADi can be used to dissect in more detail the consequences of TEAD blockage and could serve as a resource to differentiate TEAD-dependent and independent effects, providing additional clues to suppress YAP1/TAZ activity in cancer.
Here, we utilize TEADi in a mouse model of BCC driven by oncogenic Hedgehog-Smoothened (SmoM2) activity to analyze the transcriptional and cell fate consequences of TEAD blockage in skin cancer. We find that TEAD inhibition in BCC triggers the rapid activation of differentiation programs, both in cell culture and in mouse skin. The activation of differentiation gene networks downstream of TEAD inhibition is dependent on the activation of KLF4. By comparing the effects of TEADi with those triggered by YAP1 and TAZ knockdown, our results reveal the TEAD dependent and independent roles of YAP1/TAZ in tumors and the potential issues for targeting TEAD transcription in cancer.

YAP1/TAZ-TEAD regulate differentiation gene networks in BCC
For our studies we utilized a mouse model of BCC driven by expression of the constitutive active While the role of YAP1 and TAZ in the development and progression of BCC lesions in K14-SmoM2 mice has been described before 5,6 , the precise effects of TEAD inhibition in BCC are not clear. To assess TEAD role in BCC, we utilized the TEAD-inhibitor TEADi and pooled-small interfering RNAs targeting YAP1 and TAZ (siYAP1/TAZ) in isolated skin keratinocytes from the tail of BCC mice 5 weeks following tamoxifen induction. As shown in Figure 1D, at this time point BCC lesions covered the entire tail epidermis. RNA sequencing (RNA-seq) of BCC keratinocytes with adTEADi or siYAP1/TAZ revealed the dysregulated expression of numerous transcripts ( Fig. 1E and Table S1). A significant overlap between differentially regulated genes in both conditions was observed (Fig. 1F), indicating common gene networks in TEADi and siYAP1/TAZ datasets. Both TEADi and siYAP1/TAZ led to the significant downregulation of the known TEAD targets Ccn1 and Ccn2 (also known as Cyr61 and Ctgf respectively), and Axl and Fst (Fig. 1F). siYAP1/TAZ lead to a significant reduction on Yap1 and Taz (Wwtr1) expression, while TEADi did not significantly altered these genes (Fig. 1F). Gene ontology (GO) analysis of genes regulated by both TEADi and siYAP1/TAZ indicated an upregulation of processes related to skin and keratinocyte differentiation, while downregulated gene networks were enriched for terms associated with extracellular matrix organization, cell adhesion and migration (Fig. 1G).
Although BCC is characterized by the dysregulation of Hedgehog signaling, we did not find significant alterations on the expression of Gli or Ptch or other Hedgehog targets by TEADi (Fig. 1H). Multiple alterations were present in genes central for keratinocyte differentiation, including Klf4, Foxn1, Notch3, Ivl and Krt10, among others (Fig. 1H). A broader analysis of differentiation gene expression revealed that both TEADi and siYAP1/TAZ lead to an activation of differentiation markers related to every differentiation stage (Fig. S1A) and a significant correlation with epidermal cell differentiation and keratinization terms Ingenuity pathway analysis (IPA) of upstream transcriptional regulators affected by siYAP1/TAZ and TEADi confirmed the downregulation of networks related to TEAD transcription factors (Fig. 1I). In addition, genes related to SRF/MRTF were downregulated (Fig. 1I). SRF and MRTF have been linked to therapeutic resistant in BCC 19 . On the other hand, networks related to ephrins (EFN) and peroxisome proliferator activated receptors (PPAR) were activated (Fig. 1I). TAZ can regulate PPARγ 20 and PPAR activation has been implicated in keratinocyte differentiation 21 . Ephrins have also been involved in reducing keratinocyte proliferation and increased differentiation 22 . Interestingly, IPA analysis revealed a similarity between the transcriptional networks activated by YAP1/TAZ knockdown and TEADi to those activated by cAMP and PKA signaling (Fig. S1C). This suggests that the transcriptional regulation downstream of PKA is partially mediated by YAP1/TAZ and TEAD blockage.
Overall, our results show that YAP1 and TAZ TEAD-dependent transcriptional networks are involved in the maintenance of the undifferentiated state of BCC cells and that TEAD inhibition leads to the rapid activation of differentiation genes.

KLF4 transcriptional networks are activated by YAP1/TAZ and TEAD inhibition in BCC
To better understand the dysregulation of differentiation pathways triggered by TEAD blockage in BCC we performed a transcription factor binding site enrichment analysis in proximal promoters of genes regulated by both TEADi and siYAP1/TAZ. Over-represented transcription factor binding sites in downregulated genes included TEAD and SRF ( Fig. 2A), supporting a direct role of TEAD in regulating expression of these transcripts. Binding sites in genes upregulated by YAP1/TAZ-TEAD inhibition showed a clear enrichment for KLF4 ( Fig. 2A). KLF4 is a central regulator of keratinocyte differentiation 23,24 and our group has previously shown that YAP1/TAZ-TEAD regulate the activity of this transcription factor in normal keratinocytes 15 . overexpression was sufficient to induce the expression of KRT10 in these cells (Fig. 2E). These results validate the necessity of TEAD activation and KLF4 inhibition for cell growth in BCC cells.
In order to validate the core function of KLF4 in BCC keratinocyte differentiation, we utilized available information on KLF4 chromatin immunoprecipitation analysis from mouse epidermis (ChIP-seq) 25 .
GO enrichment analysis demonstrated that these KLF4 bound genes were related to skin development and keratinocytes differentiation, while the set of genes without KLF4 binding did not show enrichment for differentiation terms (Fig. 2F). Our results indicate that repression of KLF4 transcriptional networks by YAP1/TAZ-TEAD is essential for maintaining basal cell identity and block differentiation in BCC.

YAP1 and TAZ regulate inflammatory related gene networks independently of TEAD
We next focused on the TEAD independent functions of YAP1/TAZ in BCC by evaluating pathways preferentially regulated by siYAP1/TAZ but not TEADi. Interestingly, YAP1 and TAZ knockdown lead to a significant upregulation of Gli1, the Hedgehog regulating gene Gas1 and Indian hedgehog (Ihh) (Fig.   1H), indicating the possibility of a TEAD-independent crosstalk between Hedgehog and Hippo signaling. siYAP1/TAZ also lead to an increase in Wnt5a and Wnt3, which could suggest an upregulation of the Wnt pathway, although global changes in this pathway were not clear (Fig. 1H). IPA analysis for YAP1/TAZ and TEADi regulated genes indicated preferential activation by YAP1/TAZ knockdown of several networks related to NF-κB signaling (Fig. 3A). YAP1 has been described to reduce NF-κB activation and the concomitant expression of proinflammatory cytokines IL-6, TNF-α and IL-1β 26 , and our analysis showed that siYAP1/TAZ but not TEADi increases IL-6, TNF and IL-1 gene networks (Fig. 3A). Other immunemodulatory pathways were also preferentially activated by YAP1 and TAZ knockdown, including interferon γ (IFNγ) and STAT3 signaling (Fig. 3A). Transduction of BCC keratinocytes with siYAP1/TAZ and TEADi confirmed the differential activation of Stat3 (Fig. 3B).
Interestingly, IPA analysis also indicated a stronger activation of KLF4 gene networks in the siYAP1/TAZ knockdown dataset (Fig. 3A). Analysis of over-represented transcription factor binding sites in genes differentially regulated by siYAP1/TAZ but not TEADi also showed an enrichment for KLF4 (Fig.   3C). It is worth noting that this TEADi independent set of genes did not present enrichment for TEAD binding sites (Fig. 3C). Cross-reference of the TEADi and siYAP1/TAZ differential regulated genes with KLF4-ChIPseq data indicated that KLF4 networks were activated in both conditions (Fig. 3D). Genes exclusively regulated by YAP1/TAZ that present KLF4 binding were not related to epithelial differentiation but to cell adhesion (Fig. 3D). These results suggest a TEAD dependency for the activation of KLF4 differentiation networks resulting from YAP1/TAZ inhibition.

TEADi leads to rapid elimination of tumor cells in BCC lesions
To study the effect of TEADi in BCC tumors we crossed K14-SmoM2 mice with mice carrying LSL-rtTA 27 and tetracycline-inducible TEADi (TRE-TEADi) 15 (Fig. 4A). The resulting animal model allowed us to trigger BCC formation with tamoxifen while controlling the expression of TEADi by feeding mice doxycycline chow at different points during tumor development (Fig. 4A).
When mice were induced with a single dose of tamoxifen and concomitantly fed doxycycline chow, we observed scattered cells in the epidermis positive for TEADi (GFP, Fig. 4B). Over time, TEADi expressing cells were eliminated from the epidermis and localized to the upper and differentiated layers of the skin by week 3 (Fig. 4C). Although BCC lesions still arise in these mice, the resulting tumors were smaller and TEADi negative (Fig. 4C and S2), indicating that cells in which TEAD was blocked did not proceed to form tumors. The appearance of BCC lesions negative for GFP in this model could be due to incomplete recombination of LSL-rtTA in a subset of cells that express SmoM2, allowing these cells to proceed with tumor formation. A similar scenario was reported in BCC studies with YAP1/TAZ knockout animals 5,6 , in which BCC tumors arise in YAP1/TAZ and YAP1 knockout mice that are positive for YAP1 due to incomplete recombination.
In another set of experiments, mice were treated with tamoxifen and TEADi expression was induced 5 weeks later. Although in this case BCC lesions were already developed before expression of the inhibitor, we observed a similar trend in which all TEADi positive cells were eliminated from BCC tumors over time (Fig. 4D). Analysis of differentiation and proliferation markers indicated that intramural TEADi cells were positive for the differentiation markers keratin 10 (KRT10) and KLF4 (Fig. 5A and B) and showed reduced labeling for the proliferation marker PCNA (Fig. 5C).
Our in vivo data shows that inhibition of TEAD during or subsequent to BCC tumor formation leads to the rapid activation of differentiation pathways and elimination of cells from tumor lesions and validate our conclusion that activation of TEAD is necessary for preventing differentiation downstream from Hedgehog signaling in BCC.

Discussion
In this study we use knockdown strategies and specific transcriptional inhibition to discern the TEADdependent and independent effects of YAP1/TAZ in cancer. We find that blockage of TEAD leads to rapid  28,29 . YAP1 has also been shown to dampen NF-kB activation by interacting directly with TRAF6 26 and TAK1 30 . Together, these data indicate that TEADindependent effects of YAP1 and TAZ might be mediated mainly through cytoplasmic rather than nuclear interactions with other proteins.

TEAD inhibition in BCC leads to differentiation and elimination of tumor cells, making TEAD inhibitors
candidates for differentiation therapy 31 . Nevertheless, our data indicates that cells that evade the activation of differentiation pathways could be less sensitive to TEAD blockage. Furthermore, potential tumor promoting and immune evasion signals mediated by cytoplasmic interactions of YAP1 and TAZ could pose a challenge on targeting this pathway, since blocking TEAD would not be able to prevent non-nuclear effects of YAP1/TAZ. Inhibitors that target cytoplasmic and nuclear functions of YAP1 and TAZ could prove to be more effective for cancer treatment. In this regard, we found that transcriptional networks activated by YAP1/TAZ knockdown present high similarity with those activated by cAMP and PKA signaling. Indeed, cAMP and PKA blockage can lead to BCC formation by inducing the cell-autonomous activation of GLI and YAP1 32 . Activation of PKA on the other hand can lead to blockage of YAP1 function [32][33][34] . Forskolin, which results in increased intracellular cAMP levels and PKA activation, has been shown to reduce BCC tumor formation in mouse models 35 , highlighting the potential of this pathway for YAP1/TAZ inhibition and BCC treatment.
We also find that KLF4 is central for the activation of differentiation gene networks following TEAD inhibition. Although KLF4 genomic alterations are not common in skin cancer, KLF4 is downregulated in BCC and SCC tumors and KLF4 knockout mice present increase sensitivity to tumor formation 23 . KLF4 is also responsible for reduced self-renewal and increased differentiation of skin cancer-initiating cells in mouse models 36   Statistics. All analyses were performed in triplicate or greater and the means obtained were used for ANOVA or independent t-tests. Statistical analyses were carried out using the Prism 7 statistical analysis program (GraphPad). Statistical analysis of intersections in Venn diagrams was performed by hypergeometric test (one tailed Fisher's exact test). Asterisks denote statistical significance.
Data availability. RNAseq primary and processed data generated in this manuscript is available from GEO database GSE156913 (BCC data). Processed RNAseq data is provided in Supplementary Tables S1.
RNAseq data from normal keratinocytes is from GSE136876 15 . KLF4 ChIP-seq data is from 25 .