The Transcriptional Coactivator TAZ Is a Potent Mediator of Alveolar Rhabdomyosarcoma Tumorigenesis

Rhabdomyosarcoma is the most common soft tissue sarcoma of childhood. Originating from primitive mesenchymal tissue and associated with the skeletal muscle lineage (1, 2), RMS is composed of two histologic subtypes: embryonal (eRMS), which accounts for the majority of cases, and alveolar (aRMS), which accounts for 20% of cases. Both eRMS and aRMS are treated with multimodal therapy consisting of chemotherapy, surgery, and/or radiation. Standardization of management through collaborative clinical trials improved RMS survival rates from 25% in the 1970s to 70% in the 1990s. Vincristine, dactinomycin, and cyclophosphamide (VAC), a combination first identified in the 1970s, became the accepted chemotherapy regimen. However, subsequent clinical trials have failed to identify other regimens or targeted agents superior to VAC (3). While the 5-year overall survival rates for eRMS and fusion-negative (PF-neg) aRMS are approaching 80% (4), survival for patients with fusion-positive aRMS remains stagnant at less than 50%. Most aRMS tumors are characterized by signature t{2;13} or t{1;13} chromosomal translocations that encode the PAX3-FOXO1 (P3F) and PAX7-FOXO1 (P7F) fusion proteins, respectively. These oncogenic chimeric transcription factors are usually the sole DNA mutation in an otherwise quiet genome, and thought to be the main drivers of aRMS tumorigenesis (5, 6). Metastasis, chemoresistance, and/or relapse are common (4, 7), underscoring the need for novel effective therapies.

Our previous investigations into proteins that collaborate with P3F in aRMS tumorigenesis revealed a role for Hippo/MST1 kinase inactivation (8). The Hippo pathway plays a vital role in tissue growth and homeostasis, organ size control, and tumor suppression (9, 10). The core of the Hippo pathway is an MST1/2-LATS1/2 kinase cascade, which phosphorylates and inactivates two downstream effectors known as YAP (Yes-associated protein) and its less studied paralog TAZ (transcriptional co-activator with PDZ-binding motif). YAP and TAZ are encoded by the YAP1 and WWTR1 (WW domain containing transcriptional regulator 1) genes, respectively. Phosphorylation of YAP and TAZ, which occurs at five (YAP) and four (TAZ) serine residues, respectively, leads to YAP/TAZ cytoplasmic retention through the binding of 14-3-3 proteins at phospho-S127 (YAP) or phospho-S89 (TAZ), as well as β-TRCP-dependent proteasomal degradation (11). When unphosphorylated, YAP and TAZ localize to the nucleus and coactivate progrowth transcription factors (12, 13), most notably the TEAD family (14, 15). Functionally, YAP/TAZ are essential for cellular proliferation, amplification of tissue-specific progenitor cells during tissue regeneration, and ultimately control of organ size (11, 16). In many contexts, YAP and TAZ have overlapping roles. However, they share only 50% homology and have divergent functions in development. For example, YAP knockout mice are embryonic lethal, while TAZ knockout mice are viable but frequently develop polycystic kidney disease (17, 18). In skeletal muscle homeostasis, YAP inhibits myogenesis (19), while TAZ enhances myogenic differentiation by associating with and activating MyoD-induced gene expression (20).

The roles of YAP/TAZ in epithelial malignancy have been widely studied. For example, in breast cancer, TAZ binds to TEADs to potentiate invasion and metastasis (21, 22) as well as cancer stem-like properties and chemoresistance (23). Similarly, in hepatocellular carcinoma and malignant glioma, TAZ promotes tumorigenesis, supports stemness, and mediates epithelial to mesenchymal transition (24, 25). However, an understanding of the roles of YAP and TAZ in mesenchymal cancers, including RMS, is just beginning. In eRMS, higher YAP/TAZ expression at the IHC level correlates with reduced patient survival (8, 26, 27), and a subset of tumors have copy-number gains in the YAP1 and/or WWTR1 loci (26, 27). YAP contributes to eRMS tumorigenesis by supporting proliferation and stemness, and opposing myogenic differentiation (8, 26, 28), potentially at the early steps of tumorigenesis based on a human myoblast model of eRMS (28). Similarly, TAZ contributes to eRMS by supporting proliferation, colony formation, and increasing the expression of select cancer-related genes (27). Expression of TAZS89A (a constitutively active TAZ mutant) transforms C2C12 myoblasts (27), again suggesting that YAP/TAZ exert oncogenic effects early during tumorigenesis.

Less is known about the roles of YAP/TAZ in aRMS. We had previously shown that YAP is highly abundant in P3F-aRMS cells, supporting proliferation and evasion of senescence (8). Given this, we expected to find in our previously established myoblast-based model of P3F-initiated tumorigenesis that YAP1 would be upregulated. Instead, WWTR1 was increased at the mRNA level in this model, suggesting that TAZ has a specific role in aRMS tumorigenesis. A potential functional role for TAZ in aRMS is further suggested by studies showing that TAZ is essential to the transcriptional activity of wild-type PAX3 (29, 30) and that the binding of TAZ to PAX3 occurs via domains that are retained in the P3F fusion (5). The aim of this study was to elucidate the oncogenic activity of TAZ in P3F-aRMS sarcomagenesis.

Human RMS cell lines Rh28 (31) and Rh30 (32) were gifts from Tim Triche (Children's Hospital of Los Angeles, Los Angeles, CA) in 2005; Rh3 (33), Rh41 (34), and CW9019 (35) were gifts from Brett Hall (Columbus Children's Hospital, Columbus, OH) in 2006. All cell lines tested negative for mycoplasma (using Lonza MycoAlert PLUS test at the Duke University Cell Culture Facility) and were also authenticated by short tandem repeat analysis (Promega Powerplex 18D at Duke University DNA analysis facility) in 2014; Rh28 and Rh30 were reauthenticated in 2016. Human skeletal muscle myoblasts (HSMMs) and 293T cells were obtained from Lonza and the ATCC through the Duke University Cell Culture Facility, respectively, and cultured as described previously (36). Knockdown and overexpression constructs described below were stably expressed using established lentiviral and selection methods resulting in polyclonal cell lines (37). TAZ shRNA oligos (23, 38) or nontargeting (NT) scrambled controls (Supplementary Table SI) were annealed and ligated into pLKO.1-puro or Tet-pLKO-puro (Addgene plasmids #8435 and #21915, respectively). Lenti-EF-ires-blast and pLenti-EF-FH-TAZ S89A-ires-blast (Addgene #52084) vectors were gifts from Yutaka Hata (Tokyo Medical and Dental University, Tokyo, Japan). pQCXIH-Flag-YAP-S127A (Addgene 33092) was a gift from Kun-Liang Guan (University of California San Diego, CA; ref. 39).

Standard growth curves using manual and automated hemocytometer counting, as well as MTT and bromodeoxyuridine (BrdUrd) assays, were performed as described (40). Cell-cycle analyses were performed and analyzed by the Duke University Flow Cytometry core as described previously (41).

Differentiation assays and MF20 (DSHB Hybridoma Product MF20) staining were performed as described previously (37) except that on day 0, Rh28 and Rh30 cells were plated at a density of 5 × 10⁵ and 2.5 × 10⁵/6-cm dish, respectively. Positively and negatively staining cells (five images per condition) were counted manually with the aid of cell counting software (ImageJ, NIH, Bethesda, MD).

To establish and propagate aRMS cell spheres, we modified a protocol developed for eRMS cells (42). Briefly, aRMS cells were grown in ultra-low attachment plates or flasks (Corning) in Neurobasal media (Gibco) supplemented with 1× B27 (2× B27 for Rh28 cells; Invitrogen), 80 ng/mL bFGF (Corning), 40 ng/mL EGF (Sigma), and 50 μg/mL insulin. Limiting dilution assays were based on sphere formation, and assessed 48 wells per condition. Wells were scored positive (≥1 sphere/well) or negative (0 spheres/well) for sphere formation after seven days in culture. Sphere-forming frequency and statistics were calculated using ELDA software (43).

Rh28 (1 × 10⁵ cells/well) or Rh30 (5 × 10⁴ cells/well) cells stably expressing TAZ shRNA, TAZS89A, or NT vector were transiently transfected with TEAD luciferase reporter (8× GTIIC plasmid) or empty vector (pGL3-E-P) and 5 ng of Renilla reporter (pHRC TR1) for 48 hours in triplicate in 24-well plates. Treatment (see Drug Treatment section below) with verteporfin (VP) or protoporphyrin IX (PPIX) was performed for 24 hours prior to reading. Luciferase activity was assayed using the Dual Luciferase Reporter Assay (Promega) as per manufacturer's protocol in a luminometer (Turner Biosystems Modulus). Data are presented as Firefly/Renilla luciferase activity.

For cell fractionation, 1 × 10⁶ cells were plated in a 10-cm dish; after 48 hours in culture they were trypsinized, washed, and fractionated into subcellular components using a detergent-based kit (Cell Signaling Technology, #9038). Quantitative real-time PCR (qRT-PCR) was performed as described previously (8) using primer sets listed in Supplementary Table S1. Immunoblotting was performed as described previously (40) using anti-TAZ, anti-PARP, anti-cleaved PARP (Asp214), anti-phospho-histone (Cell Signaling Technology, #4883, #9542, #9541, #9701), anti-β-tubulin, anti-actin (Sigma #T8328, #A5441), anti-histone H3 (Abcam ab1791), and anti-MF20 (DSHB Hybridoma Product MF20) antibodies.

Xenograft studies utilized 1 × 10⁷ Rh28 cells resuspended in Matrigel (BD Biosciences) and implanted subcutaneously into the flanks of immunodeficient SCID/beige mice as done previously (8, 44). For the TAZ shRNA study, Rh28 cells stably expressing doxycycline-inducible TAZ shRNAs were used. The drinking water was supplemented with 1 mg/mL doxycycline (Sigma-Aldrich) in 5% w/v sucrose or 5% w/v sucrose (control). Mice were monitored twice weekly, and upon observing palpable tumors, randomly assigned to the doxycycline or sucrose group. We had previously demonstrated no difference in tumor growth dynamics in Rh28 cell xenografts with a NT vector control in mice treated with doxycycline versus sucrose (8, 44). Tumors were measured using calipers and tumor volume calculated as [((width × length)/2)³]/2. Mice were sacrificed upon reaching an Institutional Animal Care and Use Committee (IACUC)-defined maximum tumor burden (2,000 m³) or decline in health (TAZ shRNA study), or after 28 days from development of a palpable tumor (VP studies). Portions of tumors were preserved in RNAlater (Qiagen) for PCR or formalin-fixed and paraffin-embedded (FFPE) for IHC.

Human RMS tissue microarrays (TMAs) were obtained through the Children's Oncology Group (COG) from the Biopathology Center at Nationwide Children's Hospital (Columbus, OH) and contained cores of 63 (in the TAZ analysis) or 64 (in the YAP analysis) unique aRMS tumors from 217 tissue samples that were suitable for analyzing. TAZ analysis included 34 P3F, 13 P7F, and 13 PF-neg aRMS tumors. YAP analysis included 36 P3F, 10 P7F, and 15 PF-neg aRMS tumors (See Supplementary Table S2). Three aRMS tumors from each TMA were not annotated for fusion status; these were included in the overall analysis (Supplementary Fig. S3F) but were excluded from the fusion status subset analyses. TAZ and YAP immunostaining were performed as described previously (8, 28) using TAZ (Sigma #HPA007415) and YAP (Cell Signaling Technology #4912) antibodies per the manufacturer's protocols. Specifically, we performed antigen retrieval by heating slides for 4 minutes at 125°C in a pressure cooker, followed by incubation with primary antibody at a 1:50 dilution overnight at 4°C. Expression was scored by two blinded observers using a semiquantitative method (8) based on the relative amount of nuclear staining. Scoring standards are shown in Supplementary Fig. S1. Of note, we had previously reported YAP staining of eRMS and aRMS cases (8) prior to aRMS subtype annotation from the COG. Therefore, the YAP TMA was reanalyzed taking into consideration the annotation and reclassification, which resulted in two eRMS tumors being reclassified as aRMS, and two aRMS tumors being reclassified as eRMS. Details of the TMA cases are in Supplementary Table S2.

In separate studies, FFPE xenograft tumors samples were similarly immunostained and scored for YAP and TAZ expression (standards in Supplementary Fig. S2) and also H&E (Sigma #HHS16, HT110316), Ki67 (Dako #M7240), and TUNEL (Trevigen #4810-30-K) per the manufacturer's protocols (8, 28). Slides were photographed, positively and negatively stained cells were counted manually with the aid of cell counting software (ImageJ, NIH, Bethesda, MD), and four to five images counted per condition. Tumor samples were examined by a pathologist (R.C. Bentley) with experience in pediatric sarcomas.

For in vitro studies, VP (Proactive Molecular Research P17-0440), PPIX (Sigma P8293), and vincristine sulfate (VCR; Hospira) were dissolved in DMSO: PPIX and VP at 100 mg/mL stock, and VCR at 10 μmol/L stock. In cell culture media, VP and PPIX were further diluted to 10 μmol/L, and VCR to 100 nmol/L. For in vivo experiments, two studies were evaluated: (i) DMSO vs. VP and (ii) DMSO vs. VP vs. VCR vs. VCR plus VP. Drugs were diluted in PBS to a concentration of 10% DMSO, 10 mg/mL VP, or 1 mg/mL VCR and administered by intraperitoneal injection. Upon palpable tumors, mice were randomly assigned to treatment groups of equal number and were dosed with DMSO, 100 mg/kg VP, and/or 1 mg/kg VCR. DMSO and VP were administered every other day for eight total treatments. VCR was administered weekly for four total treatments. Tumors were harvested at day 28 following tumor formation. A 28-day duration of therapy was predetermined in the study design as prior xenografts studies using 1 × 10⁷ Rh28 cells had reached maximum tumor burden by this point (8, 44).

Microarray analysis of transcriptional changes in HSMMs stably expressing P3F was described previously (8, 36). The associated dataset was deposited in the Gene Expression Omnibus database (accession GSE40543).

All animal studies were performed under Duke University IACUC-approved protocols. Human TMAs obtained from the COG had been generated from deidentified patient tissue collected with informed consent and were approved for use by the Institutional Review Board of Duke University (Durham, NC).

Unless noted, experiments were performed in at least three experimental replicates, and data are presented as the mean and SE. Statistical analysis was performed using GraphPad Prism (GraphPad). One-way ANOVA, two-way ANOVA, log-Rank (Mantel–Cox) test, unpaired t test, and Pearson correlation coefficient were used as appropriate. P values were considered significant under 0.05, with *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

To gain insight into genes regulated by P3F, our laboratory developed a cell-based model of aRMS based on P3F-initiated transformation of HSMMs. In brief, forced expression of P3F in combination with p16^INK4A loss in HSMMs enabled senescence bypass (37); additional expression of hTERT and MycN yielded cells that formed tumor xenografts histologically mimicking human aRMS (36). Microarray analysis of P3F-initiated transformation of HSMMs revealed upregulated expression of RASSF4, which is a transcriptional target of P3F and promotes aRMS tumorigenesis by restraining the Hippo tumor suppressor kinase MST1 (8). As the transcriptional coactivator YAP is a critical downstream effector of oncogenic Hippo signaling, we anticipated finding that YAP1 would also be upregulated. However, analysis of the microarray dataset showed that YAP1 expression was unchanged, while WWTR1 (encoding TAZ) was upregulated (Fig. 1A). These findings were validated by qRT-PCR using cDNA generated from the original cells represented in the microarray (Fig. 1B and C). While the increase in expression was only about twofold, other genes known to be targets of P3F, such as IL4R (45, 46), were also only upregulated about twofold (Fig. 1D), suggesting that the increase in TAZ could be important.

After confirming that TAZ was upregulated in the HSMM-based model, we examined WWTR1 mRNA and TAZ protein expression in a panel of P3F or P7F fusion-positive human aRMS cell lines. In contrast to quiescent mature skeletal muscle, both WWTR1 and TAZ expression were significantly higher in aRMS (Fig. 1E and F). To determine whether this occurred in human aRMS tumors, we queried the NIH Pediatric Oncology Branch Oncogenomics database (47) and found that WWTR1 mRNA levels are higher in human aRMS tumors than in skeletal muscle (Fig. 1G). Using the same human RMS TMA block that we had previously used to analyze YAP (8), we examined TAZ protein expression by IHC. Compared with skeletal muscle, where TAZ is membrane-bound or cytoplasmic, in aRMS tumors TAZ was predominantly nuclear (Fig. 1H). This held true for P3F, P7F, and PF-neg tumors. Interestingly, TAZ abundance is not specific to P3F aRMS and was also observed in P7F aRMS, PF-neg aRMS, and eRMS cases (Fig. 1I, Supplementary Fig. S3A–S3C), suggesting that TAZ is important for RMS tumorigenesis across the mutational landscape. Because TAZ was upregulated in our P3F-initiated model, and because TAZ is a critical coactivator for wild-type PAX3 in other cell types (29, 30), we focused our studies on P3F-positive aRMS.

As TAZ and YAP have overlapping functions, and YAP is also abundant in aRMS (8), we hypothesized that individual tumors might preferentially express one or the other, as seen in malignant peripheral nerve sheath tumors (48). To determine whether TAZ and YAP abundance inversely correlated, we reanalyzed the TMA previously immunostained for YAP (8) and observed a similar degree of YAP expression among the P3F, P7F, and PF-neg aRMS groups (Fig. 1J and K, Supplementary Fig. S3D). In the nine P3F-aRMS tumors where YAP was scored as 1 (<25% nuclear staining), two had a TAZ score of 2, and seven had a TAZ score of 3. However, using a Pearson correlation coefficient, TAZ and YAP levels do not significantly correlate in P3F-aRMS samples, in the other subgroups, or overall in aRMS (Supplementary Fig. S3E and S3F). This suggests that most aRMS tumors coexpress TAZ and YAP, which was also observed in synovial sarcoma (48).

After determining TAZ abundance in human aRMS cell lines and tumor tissue, we sought to determine the phenotypic consequence of TAZ loss of function in Rh28 and Rh30 (both P3F-positive) human aRMS cell lines. However, because prior reports had noted mostly cytoplasmic TAZ in aRMS IHC studies (27), to be sure that there was a significant nuclear (active) TAZ pool in human aRMS cells, we took a complementary biochemical approach and fractionated aRMS cells into cytoplasmic, nuclear, and membrane-bound compartments (Fig. 2A). As expected (20), TAZ was predominantly cytoplasmic or membrane-bound in undifferentiated C2C12 myoblasts. On the other hand, TAZ was predominantly nuclear in both aRMS cell lines, corroborating our TMA IHC studies. We thus tested the effect of TAZ loss of function using RNAi, generating five independently targeting, lentiviral-delivered, constitutive shRNAs to TAZ. Of these, two (sh2 and sh5) were selected for further study as they consistently demonstrated TAZ knockdown. Rh28 and Rh30 cells were studied in tandem, and in response to TAZ-directed shRNAs both showed WWTR1/TAZ suppression at the mRNA and protein levels (Fig. 2B and C). Decreased expression of TAZ target genes CTGF and CYR61 verified functional knockdown (Fig. 2D and E). We next analyzed the phenotypic response to TAZ inhibition, and in cell counting assays found that, compared with the NT vector, the population growth of both cell lines declined over time (Fig. 2F and G).

To determine the mechanism through which TAZ suppression reduced aRMS cell growth, we examined effects on proliferation, apoptosis, and differentiation. As assessed by BrdUrd incorporation, TAZ suppression inhibited proliferation in both Rh28 and Rh30 cells (Fig. 2H and J). As assessed by immunoblot for cleaved PARP, TAZ suppression also induced apoptosis (Fig. 3L and M), explaining our prior observation of an increased number of floating cells during the culture of P3F-aRMS cells undergoing TAZ suppression. Finally, as assessed by morphologic changes seen under light microscopy, immunocytochemistry, and immunoblot for MF20 (sarcomeric myosin), and qRT-PCR for expression of MYOD1, MYOG, and MYF6, TAZ suppression also increased myogenic differentiation (Fig. 3A–F). Of note, increased myogenic differentiation was observed only in aRMS cells cultured in fusion media; aRMS cells cultured in normal growth media did not undergo morphologic elongation and had no or minimal increase in myogenic markers (Supplementary Fig. S4A and S4B). These findings were in contrast to earlier reports in C2C12 myoblasts and skeletal muscle, which showed that TAZ has a prodifferentiation effect (20, 27). Given these contrasting findings, we also knocked down TAZ in C2C12 myoblasts and found that, as others had observed (20, 49, 50), differentiation of C2C12s is TAZ dependent (Supplementary Fig. S4C–S4E). To begin to understand the differential response to TAZ suppression in transformed (aRMS) versus nontransformed (C2C12) cells, we hypothesized that the cellular response to TAZ expression might depend upon whether a cell is “poised” for differentiation, and that in a heterogeneous population we might see both phenotypes. To investigate this, we turned to a complementary gain-of-function approach and ectopically expressed a constitutively active TAZ mutant (TAZS89A, Supplementary Fig. S4F) in Rh28 and Rh30 cells. While a small proportion of the cells initially underwent a morphologic elongation reminiscent of myotubes, the majority had higher proliferation, as shown by increased BrdUrd incorporation (Fig. 2I and K), suggesting a heterogeneous response to TAZ expression that is cell-context dependent. In summary, TAZ suppression decreases cell proliferation, induces apoptosis, and supports myogenic differentiation in P3F-aRMS cells.

To determine additional mechanisms by which TAZ suppression blocks aRMS cell growth, we performed cell-cycle analysis of aRMS cells stably expressing TAZ shRNAs. Compared with the NT control vector, TAZ-deficient cells were enriched in the G₂–M phase of the cell cycle (Fig. 3G and H), with an increase from 19% to 32%–34% in Rh28 cells and an increase from 16% to 21%–22% in Rh30 cells. To determine whether cells were arrested in G₂ or M, we blotted for phospho-histone H3 (p-HH3), which is specifically phosphorylated during mitosis. We did not see an increase in p-HH3 expression with TAZ suppression in Rh28 cells (Supplementary Fig. S4G) but did see a modest increase in p-HH3 in Rh30 cells (Fig. 3I). These data suggest that TAZ depletion results in an accumulation of cells in G₂–M; however, a primary reduction of cells in G₁–S is not ruled out and needs further study.

As TAZ confers stem-like properties and chemoresistance in cancer cells of epithelial origin (21, 23, 24), we hypothesized that TAZ might also be important for stemness in RMS. In addition, although fusion-positive aRMS is typically sensitive to initial chemotherapy, acquired resistance and/or recurrence are common, occurring in up to 50% of cases (3). This implies that within a population of aRMS cells, some may be initially quiescent but have the ability to self-renew. To determine whether TAZ mediates P3F-aRMS stemness, we developed a three-dimensional (3D) sphere culture system that is a well-established method for studying cancer stem cell biology (23, 24), including in eRMS cells (42). We adapted an existing eRMS sphere culture protocol to permit growth of aRMS and showed that aRMS cells are also capable of forming and being propagated as spheres (Fig. 4A and C). Passaging of these aRMS spheres resulted in a 3- to 12-fold increase in WWTR1 expression as measured by qRT-PCR (Fig. 4B and D), suggesting that TAZ may be important for stem cell enrichment.

We then examined whether the expression of stem cell markers SOX2, OCT4, and NANOG were TAZ dependent in aRMS cells. As demonstrated via qRT-PCR, TAZ suppression led to decreased expression of these markers (Fig. 4E–H). To determine the functional role of TAZ in aRMS cancer cell stemness, we performed limiting dilution assays (LDA) using aRMS rhabdospheres (Fig. 4I and J). We stably expressed the TAZ shRNAs or NT control in Rh28 and Rh30 cells, then plated them at varying densities. While the estimated sphere-forming frequency of the NT control for Rh28 and Rh30 cells was 1/24 and 1/21, respectively, it was reduced to <1/44 in both TAZ shRNA groups, in both cell lines. These data suggest that TAZ is functionally required for P3F-aRMS stemness and cancer cell self-renewal.

After determining that suppression of TAZ decreases proliferation, induces apoptosis, supports myogenic differentiation, and decreases stemness in P3F-aRMS cells in vitro, we used Rh28 cell xenografts to examine the role of TAZ in aRMS sarcomagenesis in vivo. As constitutive suppression of TAZ in Rh28 cells inhibited cell viability, we used a conditional doxycycline-inducible shRNA system to suppress TAZ expression after tumor formation. After validating the constructs in vitro (Supplementary Fig. S5 and S6), we generated SQ xenografts and found that inducible TAZ knockdown decreased tumor growth and prolonged survival (Fig. 5A and B). By day 21, IACUC-defined maximum tumor burden was reached in all of the mice in the control groups, compared with only one tumor in the doxycycline groups, which had a mean day 21 tumor volume of 801 mm³. Using qRT-PCR (Fig. 5C, left) and IHC (Fig. 5D, second column), we validated that TAZ suppression, including decreased nuclear TAZ expression, (Fig. 5E, left) was maintained throughout the duration of the study at the mRNA and protein levels. We then evaluated potential mechanisms through which TAZ inhibition caused in vivo growth delay. We performed IHC analysis of Ki67 and TUNEL (Fig. 5D, third and fourth columns) to evaluate proliferation and apoptosis, respectively. Similar to our findings in vitro, TAZ suppression in vivo led to decreased proliferation, as well as a slight increase in apoptosis (Fig. 5F and G).

While we observed tumor growth inhibition with TAZ suppression, by day 21 the mice receiving doxycycline began to exhibit similar growth as the control groups, suggesting the P3F-aRMS cells either developed resistance to TAZ suppression or mechanisms to compensate for TAZ inhibition. Anticipating that YAP upregulation could be a compensatory mechanism, we examined YAP1 expression via qRT-PCR (Supplementary Fig. S7A), as well as TAZ/YAP-TEAD targets CTGF and CYR61 (Fig. 5C). While YAP1, CTGF, and CYR61 mRNA expression was slightly higher in the two doxycycline-treated tumors that had grown most rapidly, YAP1 expression did not change and targets CTGF and CYR61 were overall suppressed, indicating these were not mechanisms of overcoming TAZ inhibition. This result is similar to the in vitro observations made with TAZ suppression in C2C12 myoblasts and RD eRMS cells, where no compensatory increase in YAP expression was seen (27), and suggests that the P3F-aRMS xenografts developed alternate means for overcoming TAZ suppression.

As TAZ is a transcriptional coactivator, it does not directly bind DNA but rather exerts tumorigenic activities through binding to and activating a host of oncogenic transcription factors, including TEADs (14). TEAD activity was shown to be TAZ dependent in skeletal muscle and RD eRMS cells (27). To verify this was the case for P3F-aRMS, we transiently transfected constitutively active TAZS89A along with an 8xGTIIC TEAD reporter in Rh28 and Rh30 cells and observed a 7- to 11-fold increase in TEAD activity (Fig. 6A; Supplementary Fig. S7B).

Porphyrin compounds were identified in a drug screen to disrupt the association between YAP and TEADs (51). Although not used clinically as an anticancer therapy, verteporfin (VP) is FDA approved for the treatment of macular degeneration. Both VP and protoporphyrin IX (PPIX), another porphyrin derivative, are commonly used in preclinical studies of the Hippo pathway (28, 51, 52). We tested these compounds in the TEAD reporter assay in P3F-aRMS cells, and both drugs decreased TEAD activity (Fig. 6B and C; Supplementary Fig. S7C–S7F). In addition, VP treatment partially abrogated constitutive TAZS89A activity (Fig. 6A, Supplementary Fig. S7B), and enhanced the effects of TAZ shRNA suppression (Fig. 6B). These studies suggest TEADs are activated by TAZ in P3F-aRMS, but targeting both TAZ and YAP may be necessary to completely abolish TEAD activity.

After validating target inhibition using porphyrin compounds in aRMS cells, we evaluated Rh28 cell viability following treatment with PPIX and VP. We determined that 10 μmol/L of either drug inhibits cell growth (Fig. 6D and E), which is consistent with the doses used in other cancer studies (28, 51, 52). We then tested these drugs on P3F-aRMS cell and tumor viability. PPIX treatment dramatically reduced Rh28 cell growth in vitro as measured by cell counting over 5 days (Fig. 6F). Delayed tumor growth was also seen with VP treatment of Rh28 SQ xenografts in vivo (Fig. 6G). As tumors became palpable, mice were randomly assigned to treatment with VP or DMSO vehicle via intraperitoneal injection every other day for eight doses. Similar to what was seen with VP treatment in eRMS (28), responses were variable. While a near-complete response was seen in two of the mice, one tumor grew to maximal tumor burden within the treatment observation period (Supplementary Fig. S7G and S7H). Although VP decreased TEAD-reporter activity in vitro, we did not see suppression of TEAD target genes CTGF and CYR61 when examining mRNA from tumors harvested after VP treatment (Supplementary Fig. S7I), which is consistent with other observations of in vivo VP treatment, including in eRMS xenografts (28). Our studies suggest that targeting TAZ/YAP/TEAD pathways may be an effective treatment strategy in P3F-aRMS. However, as was observed in the TAZ shRNA xenograft studies, aRMS tumors may be able to overcome inhibition of TEAD activity, reinforcing a need for rational combination therapies.

Because TAZ suppression leads to a G₂–M arrest in P3F-aRMS cells, and TAZ mediates resistance to antitubulin drugs in other malignancies (23, 52–55) in vitro, we hypothesized that TAZ might confer resistance to VCR (the antimicrotubule agent in VAC; ref. 3). In an MTT assay, we treated Rh28 and Rh30 cells expressing either constitutively active TAZS89A or control with varying doses of VCR and found that TAZS89A shifts the IC₅₀ from 2.2 to 4.7 (Fig. 6H; Supplementary Fig. S7J), suggesting that TAZ promotes VCR chemoresistance in P3F-aRMS in vitro. We then assessed whether VP might cooperate with VCR to increase its therapeutic effect. As measured via MTT assay, an additive effect is seen in Rh28 cells treated with the combination of VCR and either 0.3 or 1 μmol/L VP (Fig. 6I). To test whether TAZ inhibition could potentiate the effects of VCR in vivo, we evaluated the combination of VP and VCR in Rh28 SQ xenografts. VP was again administered every other day for eight doses and VCR was injected weekly for four doses. While efficacy was seen in monotherapy of both drugs, the combination of VP plus VCR was most effective (Fig. 6J), suggesting that TAZ-TEAD inhibition augments the activity of anti-tubulin drugs in fusion-positive aRMS.

Although more than 20 years have passed since the PAX3-FOXO1 (P3F) fusion was discovered as the principal mutation responsible for otherwise karyotypically simple aRMS tumors, survival rates for patients with P3F-aRMS remain dismal. VAC was first discovered to have activity against RMS in the 1970s, but no subsequent clinical trials have identified a more effective treatment regimen. Today, most children with P3F-aRMS will become chemo-refractory or relapse, underscoring an urgent need for novel agents that target resistance mechanisms and CSC renewal. In adult epithelial cancers, TAZ promotes metastasis, CSC-like properties, and chemoresistance (reviewed in refs. 56–58). By interrogating the role of TAZ in P3F-aRMS through in vitro and in vivo genetic and pharmacologic approaches, we discovered that TAZ also mediates many of these tumorigenic traits in this mesenchymal cancer.

We have found that TAZ is highly expressed in human aRMS cells and tissue samples, with a significant fraction in the nucleus. We also found that WWTR1 mRNA and nuclear TAZ expression are higher in aRMS tumors compared with skeletal muscle, mirroring a smaller study reporting high levels of nuclear TAZ in six of 10 aRMS samples (48). Interestingly, a third analysis (using the same anti-TAZ antibody) showed TAZ staining in aRMS to be negative or cytoplasmic in the majority of tumors evaluated (27). We do not know the reason for these differences, but posit that technical differences in antigen retrieval or primary antibody incubation conditions could be responsible. Mechanistically, we find that TAZ depletion decreases aRMS cell viability and proliferation and induces myogenic differentiation in vitro. TAZ is also protumorigenic in vivo, as in murine xenograft assays, TAZ suppression decreased tumor growth and proliferation, increased apoptosis, and prolonged survival. Our finding that TAZ opposes myogenic differentiation was not unexpected—indeed a signature phenotype of RMS is its loss of the ability to differentiate (59). However, as in normal skeletal myogenesis, TAZ associates with MyoD and promotes differentiation, we speculate that TAZ behaves differently depending upon whether the cellular milieu (epigenetic state) is permissive for differentiation signals. This has been noted with other proteins such as mTOR, which antagonize myogenic differentiation during conditions favoring proliferation but facilitate myogenic differentiation if activated during conditions favoring differentiation (60).

Regarding the molecular relationship between TAZ and PAX3-FOXO1, it does not appear that TAZ (or for that matter YAP) are direct downstream targets of PAX3-FOXO1. As noted in our prior work, while the PAX3–FOXO1–RASSF4 axis restrains Hippo/MST1 kinase activity and YAP1 is upregulated, RASSF-suppressed phenotypes cannot be rescued with constitutively active YAP (8). Neither do we understand why WWTR1/TAZ, but not YAP1, was uregulated in our transcriptome analysis of P3F-initiated transformation of HSMMs. However, based on studies in melanocytes, where TAZ is essential to the transcriptional activity of wild-type PAX3 (29), and the fact that binding of TAZ to PAX3 occurs via domains that are retained in the P3F fusion, we hypothesize that TAZ might be a required transcriptional coactivator of P3F, and proteomic studies examining the relationship between TAZ and P3F are ongoing.

The most important translational finding from this work is that TAZ mediates aRMS cancer cell phenotypes, including supporting stemness and promoting resistance to VCR. Using novel culture conditions, we show that the functional ability of P3F-aRMS cells to form spheres is TAZ dependent, suggesting that TAZ is critical for aRMS cell self-renewal. Although we do not know the mechanism, studies in C2C12 myoblasts show that ectopic expression of TAZS89A causes the upregulation of Myf5 (27), a cancer stem cell gene in a zebrafish model of eRMS (61). As most patients with aRMS who relapse are insensitive to VCR, and as TAZ mediates resistance to anti-tubulin drugs in other malignancies (23, 53–55), we evaluated whether TAZ-TEAD signaling could modulate sensitivity to VCR. We found that TAZS89A decreased, while VP increased, the sensitivity of aRMS cells to VCR.

While no direct TAZ inhibitors are commercially available, TAZ is an attractive drug target due to its role in promoting stem-like properties and chemoresistance in several common adult carcinomas. Thus, academic and pharmaceutical groups are actively investigating ways to target TAZ and/or the TAZ–TEAD interaction (57, 58, 62, 63). While problems with solubility, toxicity, and off-target effects limit their utility in patients, the efficacy of porphyrin compounds in preclinical cancer studies provides proof-of-principle that interfering with TAZ–TEAD and YAP–TEAD interactions is a promising approach. We predict that dual inhibition of TAZ and YAP may be more efficacious than monotherapy, as there is feedback between these paralogs, and inhibition of one may induce the upregulation of the other. For example, combining TAZ suppression with 5-fluorouracil paradoxically increased hepatocellular carcinoma cell growth through upregulation of YAP protein (64). In the current work, the xenograft tumors eventually grew despite TAZ RNAi or VP treatment, reinforcing the need to target TAZ in the context of rational combination therapies. On the other hand, VP combined with VCR was effective at abrogating tumor growth.

While the P3F transcription factor is currently not therapeutically tractable, inhibiting TAZ may be a means of attenuating its activity. In melanocytes, TAZ is essential to the transcriptional activity of wild-type PAX3 (29, 30), and the binding of TAZ to PAX3 occurs through domains that are retained in the P3F fusion (5). TEAD1 and AP-1, progrowth transcription factors that depend on TAZ/YAP activation (10), are among the top enriched motifs in P3F-binding sites (46), suggesting that these proteins may work together to control P3F-mediated transcription, a future research direction in our laboratory.

Finally, our study of TAZ biology in P3F-aRMS may shed light on the role of TAZ in sarcomas in general. A recent IHC analysis of TAZ in a range of sarcomas showed that TAZ may be associated with more undifferentiated and/or higher-grade tumors (48). For example, 87% (13/15) of high-grade dedifferentiated and pleomorphic liposarcomas stained positive for nuclear TAZ, while only 6% (1/17) of low-grade myxoid and well-differentiated liposarcomas were positive. When analyzing all sarcomas, nuclear YAP and TAZ were expressed in 50% and 66% of 159 samples, respectively, and higher YAP or TAZ expression independently correlated with higher tumor grade and worse survival (48).

In conclusion, understanding TAZ biology may provide insight into molecular mechanisms of sarcomagenesis and treatment resistance (48). Here, we show that TAZ is abundant in human aRMS tumor samples, and that TAZ suppression decreases proliferation, promotes differentiation, and inhibits cancer cell stemness. TAZ-deficient aRMS cells are also enriched in G₂–M, suggesting that TAZ may be important for G₂–M cell-cycle progression. Constitutive activation of TAZ diminishes the efficacy of VCR, and combining VP with VCR is more effective than either agent alone in blocking aRMS xenograft tumor growth. Inhibiting TAZ is a promising adjunctive therapy for targeting the aRMS CSC population and reducing chemoresistance.

No potential conflicts of interest were disclosed.

Conception and design: M.D. Deel, A.R. Hinson, L.E.S. Crose, C.M. Linardic

Development of methodology: M.D. Deel, K.K. Slemmons, L.E.S. Crose, C.M. Linardic

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.D. Deel, K.K. Slemmons, A.R. Hinson, K.C. Genadry, B.A. Burgess, N. Kuprasertkul, K.M. Oristian, C.M. Linardic

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.D. Deel, K.K. Slemmons, A.R. Hinson, B.A. Burgess, N. Kuprasertkul, R.C. Bentley, C.M. Linardic

Writing, review, and/or revision of the manuscript: M.D. Deel, K.K. Slemmons, A.R. Hinson, R.C. Bentley, C.M. Linardic

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.K. Slemmons, K.M. Oristian, C.M. Linardic

Study supervision: M.D. Deel, C.M. Linardic

We thank Drs. Dan Wechsler, David Kirsch, Gerard Blobe, Oren Becher (Duke University School of Medicine) and Erin Rudzinski (Seattle Children's) for helpful discussions and experimental suggestions. We also thank Drs. Yi-Tzu Lin, Julie Kephart, and Waitman Aumann, as well as Stefan Riedel, Margaret DeMonia, and Elaine Justice (Duke University) for technical support. The Duke Microarray and Flow Cytometry cores (Duke National Cancer Institute and Duke Genomic and Computational Biology shared resources) provided microarray and cell cycle analyses, respectively. This research was supported by NIH K12 HD043494, a St. Baldrick's Foundation Fellowship, a Derfner-Children's Miracle Network Award, and the Norman & Bettina Roberts Foundation (to M.D. Deel), a St. Baldrick's Summer Fellowship (to K.C. Genadry), NIH R01 CA122706, Alex's Lemonade Stand Innovation Award (to C.M. Linardic), and a Hyundai Hope on Wheels Pediatric Cancer Research Grant (to C.M. Linardic and A.R. Hinson).

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