YAP-Mediated Repression of HRK Regulates Tumor Growth, Therapy Response, and Survival Under Tumor Environmental Stress in Neuroblastoma

Despite intense multimodal therapy, more than 50% of patients with high-risk neuroblastoma relapse with aggressive chemotherapy-resistant disease, stressing an exigent need for therapies targeting pathways that drive recurrence (1). Following treatment and relapse, high-risk neuroblastoma tumors display significantly more RAS/RAF/MAPK genetic alterations than at diagnosis (2, 3). Gene set enrichment analyses (GSEA) also show increased transcriptional activity of the Yes-associated protein (YAP) in posttherapy recurrent neuroblastoma tumors (4). YAP is a transcriptional coactivator that binds with TEA domain (TEAD) family transcription factors (TF) to regulate genes promoting organ growth, cell self-renewal, and survival (5–8). YAP, and its paralog TAZ, are negatively regulated by proteins of the Hippo pathway that phosphorylate YAP/TAZ, leading to their cytoplasmic retention and inactivation. When YAP is dephosphorylated, it translocates into the nucleus to bind TEAD and mediate transcription. YAP's aberrant transcription functionally contributes to many aspects of cancer formation and survival (8–12). YAP's transcriptional regulation of downstream target genes such as EGFR, SOX2, and OCT4 has been shown to promote the mesenchymal phenotype and resistance to therapy (8–13). In addition, YAP regulates genes associated with epithelial-to-mesenchymal transition, thus promoting metastasis in many cancers (14). In an in vivo metastatic mouse model of neuroblastoma, neuroblastoma tumors that had metastasized to the brain displayed increased levels of YAP, suggesting a potential microenvironmental and metastatic role for YAP (15). While YAP “genetic signatures” are associated with tumor recurrence and independently predict poor neuroblastoma patient outcome, the functional role for YAP itself in neuroblastoma has not been characterized (3, 15). Furthermore, given the lack of YAP mutations or upstream Hippo pathway mutations in relapsed neuroblastoma tumors, it is unclear how YAP transcriptional activity increases at or before tumor recurrence and whether it contributes to neuroblastoma relapse. We now show that standard cytotoxic chemotherapy induces both YAP expression and transcriptional activity in murine models of neuroblastoma. We also demonstrate that YAP promotes a mesenchymal phenotype to affect in vivo neuroblastoma tumor growth and therapy response, and further inhibits apoptosis in response to nutrient deprivation through a novel regulation of the proapoptotic protein, Harakiri (HRK).

Human-derived neuroblastoma cell lines were obtained from the Children's Oncology Group Childhood Cancer Repository and the ATCC. Neuroblastoma cell lines were cultured in RPMI1640 (Sigma) with 10% FBS (Gemini) and 1% penicillin–streptomycin (Gemini) and incubated at 37°C with 5% CO₂ (16). All cells used were maintained at low passage, which did not exceed 15 passages. Cell lines and transduced cell models underwent short tandem repeat–based genotyping (Texas Tech Cancer Cell Repository) and identities were verified using the Children's Oncology Group cell line genotype database (www.cccells.org). Cells were routinely tested for Mycoplasma with MycoAlert Mycoplasma Detection Kit (Lonza). Patient-derived xenografts (PDX) were created from fresh deidentified human tumor tissue obtained from patients at diagnosis or at relapse under a Children's Healthcare of Atlanta Biorepository (Atlanta, GA) protocol approved by the institutional review board (IRB00034535). Written informed consent was obtained from patients/families for tumor tissue procurement, including cell line and PDX derivation, and deidentification for research investigations. Studies were conducted in accordance with recognized ethical guidelines (Declaration of Helsinki).

RNA was purified from neuroblastoma cell lines and snap-frozen xenograft tumors using the RNeasy Mini Kit (Qiagen) and TRIzol (Ambion) and Chloroform (Sigma), respectively. cDNA was created using the high-capacity cDNA Reverse Transcription Kit (Applied Biosystems). Power SYBR Green PCR Master Mix (Applied Biosystems) and the Bio-Rad System were utilized for quantitative PCR, with samples run in triplicate. Relative gene expression was normalized to GAPDH and HPRT using the CFX Manager Software (Bio-Rad). Primer sequences are detailed in Supplementary Table S1.

Snap-frozen xenograft tumors were dissociated with mortar and pestle. Neuroblastoma cell lines were consistently harvested at 80% confluency, as YAP expression can be influenced by cell contact (17). Neuroblastoma cells were lysed with CHAPS buffer (10 mmol/L HEPES, 150 mmol/L NaCl, and 2% CHAPS) with protease inhibitors (Roche Protease Inhibitor Cocktail 1×, 2 mmol/L sodium orthovanadate, and 2 mmol/L PMSF), run on Nu-PAGE 4%–12% Bis-Tris Gels (Invitrogen), transferred to polyvinylidene difluoride Membranes (Invitrogen), and detected for proteins by chemiluminescence as described previously (18).

Antibodies from Cell Signaling Technology used include: YAP (#4912, RRID:AB_2218911), phospho-YAP (Ser127) (#4911, RRID:AB_2218913), TAZ (V386) (#4883, RRID:AB_1904158), pan-TEAD (D3F7L) (#13295, RRID:AB_2687902), EGFR (C74B9) (#2646, RRID:AB_2230881), phospho-EGFR (Tyr1068) (#3777, RRID:AB_2096270), CYR61 (E5W3H) (#39382, RRID:AB_2799154), GAPDH (14C10) (#2118, RRID:AB_561053), BCL-XL (#2762, RRID:AB_10694844), BIM (#2933, RRID:AB_1030947), and cytochrome c (#11940, RRID:AB_2637071). Other antibodies used were: β-Tubulin (Santa Cruz Biotechnology, #sc-53140, RRID:AB_793543), BCL2 (Agilent, t# M0887, RRID:AB_2064429), and MCL1 (Enzo Life Sciences, #ADI-AAP-240, RRID:AB_10997659).

YAP gene expression was stably knocked down in SK-N-AS (RRID:CVCL_1700) and NLF (RRID:CVCL_E217) using MISSION YAP shRNA Lentiviral Particles (Sigma, SHCLNV-NM_006106). Cells were transduced with shYAP or control vector (CV) containing lentiviral particles with 8 μg/mL polybrene for 48 hours, and then, transfectants were selected for and maintained in puromycin. For YAP expression in IMR5 (RRID:CVCL_1306) and NGP (RRID:CVCL_2141), pGAMA-empty (Addgene, plasmid #74755; RRID: Addgene_74755) and pGAMA-YAP (Addgene, plasmid #74942; RRID:Addgene_74942) were gifts from Miguel Ramalho-Santos (University of Toronto, Toronto, Canada). HEK293T (RRID:CVCL_0063) cells were transfected with 3 μg of plasmid DNA, pMD2.G (gift from Didier Trono, EPFL, Lausanne, Switzerland; Addgene, plasmid #12259; RRID:Addgene_12259) and psPAX2 (gift from Didier Trono; Addgene, plasmid # 12260; RRID:Addgene_12260) using Fugene 6 (Promega) transfection reagent protocol at ratio of 1:3 (plasmid DNA:Fugene 6). Plasmid containing viral particles were collected from the supernatant and used to transduce neuroblastoma cells.

Cells were plated in triplicate at 10,000 cells/well and CellTiter-Glo (Promega) was used to quantify live cells at different timepoints.

Cells were plated at 2,000 cells/well and cultured for 7 days. Cells were fixed with formaldehyde (1%) and stained with crystal violet (0.1%, w/v) in 20% methanol. Wells were imaged with a Lionheart FX Microscope (BioTek) and quantification was performed with ImageJ (NIH; RRID:SCR_003070).

Cells were plated at 15,000 cells/mL in neurosphere media (50/50 F12/DMEM, Thermo Fisher Scientific; 1× B27, Thermo Fisher Scientific; 1× N2, Thermo Fisher Scientific; 0.1 mmol/L BME, Sigma; 2 μg/mL heparin, Stemcell Technologies; 1% penicillin–streptomycin, Gemini; 20 ng/mL EGF, Corning; and 40 ng/mL FGF, Corning) and cultured in Ultra-Low Attachment Plates (Corning). Neurospheres were harvested after 7 days for qRT-PCR. For quantification assays, cells were plated in triplicate at a density of 2,000 cells/100 μL in each well and grown for 7 days, and then imaged.

Cells were plated in triplicate at 10,000 cells/well. After 24 hours, cells were treated with vehicle (DMSO), etoposide (Sigma), irinotecan (Sigma), topotecan (Sagent Pharmaceuticals), and melphalan (Sigma) for 48 hours, or trametinib (Selleck Chemicals) for 72 hours and evaluated by CellTiter-Glo (Promega).

Harvested cells were suspended in subcellular fractionation buffer (250 mmol/L sucrose, 20 mmol/L HEPES, 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 1 mmol/L EDTA, and 1 mmol/L EGTA) with protease inhibitors (Roche Protease Inhibitor Cocktail 1×, 2 mmol/L sodium orthovanadate, and 2 mmol/L PMSF) and lysed. Lysate was incubated on ice for 20 minutes, and then centrifuged at 720 × g for 5 minutes. The resulting supernatant was centrifuged at 10,000 × g for 5 minutes to produce cytosolic (supernatant) and mitochondrial (pellet) fractions. The mitochondrial fraction was washed and resuspended in subcellular fractionation buffer supplemented with 10% glycerol and 0.1% SDS. The resulting fractions were immunoblotted for cytochrome c.

For serum starvation experiments, cells were grown in RPMI with 10% FBS for 24 hours, media were replaced with RPMI containing 10% FBS or 0.1% FBS, and cells were incubated for 48 hours. For etoposide (Sigma) treatments, cells were exposed to either vehicle (DMSO) or etoposide (5 μmol/L) for 48 hours. BD ViaProbe/7-AAD (Thermo Fisher Scientific) and APC Annexin V (BioLegend) were added and detected by Flow Cytometry (Cytek Aurora). Analysis was performed using FlowJo software (v10, RRID:SCR_008520). Caspase Glo 9 (Promega) was used per protocol to quantify caspase-9 activity.

All animal studies were conducted in accordance with policies set forth by the Emory University (Atlanta, GA) Institutional Animal Care and Use Committee (IACUC). Our protocol was approved by the Emory IACUC (PROTO201700089). Euthanasia was performed by CO₂ and cervical dislocation.

Mice were approximately 6 weeks of age at time of injection and male/female mice were distributed equally among experimental arms. For all tumor growth and drug treatment studies, mice were randomized on the basis of tumor size and gender into the different treatment arms to maintain equivalent median tumor sizes per arm of each study. We measured binary endpoints (i.e., tumor growth or tumor regression), 4–6 mice per group were used to detect 50% differences in group means, assuming sample SDs < 50% of the mean (α = 0.05 and power = 0.8). For PDX studies, low passage (≤3rd) PDX tumors were removed from one NOD scid gamma (NSG; The Jackson Laboratory; IMSR catalog no., ARC:NSG, RRID:IMSR_ARC:NSG) mouse, dissociated, mixed 1:1 in Matrigel (Corning) as a single-cell suspension with 4 × 10⁶ cells/injection, and reinjected into the subcutaneous flank of additional NSG mice. Tumor volumes were measured with calipers and calculated using formula L × W × H × π/6. When tumors reached approximately 150 mm³, mice were treated with vehicle control (PBS) or 0.05 mg/kg topotecan (Sagent Pharmaceuticals) and 20 mg/kg cyclophosphamide (Jiangsu Hengrui Medicine) via intraperitoneal injections once daily for 5 days. Mice were sacrificed on day 6 and tumor was harvested and evaluated. For tumor formation studies, SK-N-AS CV, shYAP#1, and shYAP#2 cells (4 × 10⁶ cells/injection) were injected into the subcutaneous flanks of NSG mice (n = 4 per arm). Mice were sacrificed when tumor volumes reached 1,500 mm³. For cyclophosphamide treatment studies, SK-N-AS xenografts were established as above in nu/nu athymic mice (The Jackson Laboratory; MGI #3848172, RRID:MGI:3848172). Mice (n = 5 per arm) were treated with vehicle (PBS) or 75 mg/kg cyclophosphamide by intraperitoneal injection twice weekly for four doses and sacrificed when tumor volumes reached 1,500 mm³. For trametinib studies, SK-N-AS xenografts established subcutaneously in NSG mice (n = 6 per arm) were treated with vehicle (35% Kollisolv PEG E 400, 60% Phosal, and 5% DMSO) or 3 mg/kg Trametinib (Selleck Chemicals) for 14 days once daily via oral gavage and mice were sacrificed when tumor volumes reached 1,500 mm³.

SK-N-AS CV, shYAP#1, and shYAP#2 pellets in triplicates were harvested, RNA was isolated, and quality checked at the Emory Integrated Genomics Core. RNA-sequencing (RNA-seq; poly-A selection, 20 M paired-end 150 bp reads on the Illumina NovaSeq Platform) was performed at Novogene, with data analyzed by the Emory Integrated Computational Core. Sequencing data were first checked by FastQC (RRID:SCR_014583) for quality control, and then aligned to the human “indexed” reference genome (hg38) using STAR (RRID:SCR_015899; ref. 19). Gene quantification was performed using HTSeq-count (RRID:SCR_011867; ref. 20) and gene count data were filtered before analysis to remove lowly expressed genes. Counts per million (CPM) data were generated after normalization using trimmed method of M-values, as implemented in the Bioconductor package (RRID:SCR_006442), EdgeR (RRID:SCR_012802; ref. 21). To determine genes important for a case–control, a t test followed by FDR (Benjamini–Hochberg method) on normalized log₂-transformed CPM was performed. A gene was defined as differentially expressed gene (DEG) if its FDR ≤ 0.05. To characterize expressed genes, a preranked permutation-based GSEA was performed with weighted Kolmogorov–Smirnov statistical test (22). A gene set from the Molecular Signatures Database (v7.0) was used as input data for this analysis. The identified DEGs were used in pathway analysis.

A Student t test or a one-way ANOVA was used to assess statistical significance. A log-rank (Mantel–Cox) test was performed on the Kaplan–Meier survival curves for xenograft studies. Statistical significance was defined as P < 0.05. GraphPad Prism (v8.0; RRID:SCR_002798) was used for all analyses.

Given previous reports demonstrating increased YAP transcription in primary recurrent neuroblastomas, we first sought to determine YAP expression in neuroblastoma PDXs from the same patient's prechemotherapy diagnostic (NBX-4) and post-chemotherapy recurrent (NBX-4R) tumor. Results demonstrate that YAP expression and transcriptional activity increased in the relapsed PDX compared with the pretherapy PDX from the same patient (Fig. 1A). In keeping with this finding, paired isogenic cell lines derived from the same patient at diagnosis prechemotherapy (SK-N-BE1) and at relapse following chemotherapy (SK-N-BE2) showed increased YAP expression in the relapsed SK-N-BE2 (Supplementary Fig. S1). To gain a sense for the timing of YAP increase during or after therapy, we established in vivo tumors using chemotherapy-naïve NBX-4 cells and subsequently treated mice with 5 days of vehicle or topotecan and cyclophosphamide at doses and regimen comparable with a standard cycle of chemotherapy used for neuroblastoma patient induction (23). The tumors grew through chemotherapy or vehicle treatment and were harvested on day 6. Notably, YAP gene and protein expression increased in the chemotherapy-treated tumors and remained absent in the vehicle-treated control and expression of YAP transcriptional targets, CTGF and EGFR, increased in the chemotherapy-treated tumor, but not in the vehicle control (Fig. 1B and C). We treated a second neuroblastoma PDX derived from another patient's diagnostic chemotherapy-naïve tumor (NBX-1a) with topotecan and cyclophosphamide. Tumors grew through the topotecan and cyclophosphamide treatment, and again we observed that YAP protein expression increased by day 6 in the chemotherapy-treated PDX, but not in the vehicle control (Fig. 1D). We then evaluated a panel of human-derived neuroblastoma cell lines and found YAP protein to be heterogeneously expressed (Fig. 1E). Phosphorylated and inactive YAP (pYAP) was present in some cell lines. TAZ was neither always coexpressed with YAP nor was it increased in YAP-null cells. TEAD protein was present throughout all cell lines. Expression of known downstream targets of YAP-mediated transcription, EGFR and CYR61 (24, 25), was increased in neuroblastomas with YAP expression, suggesting that YAP transcription may be active despite the concurrent presence of pYAP in certain cells. We also found that YAP expression does not correlate with MYCN amplification status and YAP was commonly seen in neuroblastoma cell lines harboring RAS/MAPK mutations, but not ALK mutations (Fig. 1E; Supplementary Table S2).

A prime role for YAP in both normal development and in cancer is to promote growth through transcriptional as well as posttranslational regulation of genes involved in cell proliferation (26, 27). To test the effects of YAP on neuroblastoma growth, we used both gain- and loss-of-function approaches to perturb YAP expression. Western blot analysis of these cell lines confirmed YAP knockdown (Fig. 2A) and overexpression (Fig. 2B; Supplementary Fig. S2A). TAZ and TEAD expression were not altered by YAP knockdown (Fig. 2A). While EGFR and phosphorylated EGFR (pEGFR) did not significantly change, YAP targets, CTGF and CYR61, decreased with YAP knockdown and increased with YAP overexpression, suggesting these models efficiently modulate YAP transcriptional activity (Fig. 2A and B). In vitro, YAP perturbation does not significantly alter cell proliferation rates (Fig. 2C; Supplementary Fig. S2B). Yet, in the in vivo setting, time to sacrifice occurred later in the SK-N-AS shYAP xenografts compared with the SK-N-AS CV tumors (Fig. 2D). Importantly, YAP remained knocked down at time of sacrifice (Fig. 2D). These results suggest that YAP affects neuroblastoma proliferation within the heterotypic tumor microenvironment (TME) and may promote the initiation of neuroblastoma growth.

In other tumor models, YAP regulates tumor cell growth by mediating contact-independent cell proliferation and expression of stem-like genes (9, 10, 28). We, therefore, evaluated the colony formation potential of NLF shYAP cells and show a significant decrease in the number of colonies formed compared with NLF CV cells (Fig. 3A). To further investigate for mesenchymal properties of YAP, we grew human-derived neuroblastoma cell lines as neurospheres in neurobasal media, thus mirroring the stem-like environment. YAP gene expression was significantly increased upon neurosphere formation (Fig. 3B). OCT4 and SOX2, known mesenchymal genes regulated by YAP (28, 29), increased under neurosphere conditions, compared with regular culture conditions (Fig. 3B). We evaluated for changes in mesenchymal gene expression in the NLF shYAP model versus NLF CV cells grown in neurobasal media and noted that YAP knockdown leads to decreased levels of these genes (Fig. 3C). Furthermore, RNA-seq of SK-N-AS shYAP models compared with SK-N-AS CV shows genes associated with a mesenchymal and stem-like state in other tumor types, such as JAK1, IL6ST, TBX3, and PPP2R2B (30, 31), decreased when YAP was knocked down in neuroblastoma (Supplementary Fig. S3). We also noted fewer neurospheres in the NLF shYAP cells, and an increased number of neurospheres in the YAP-overexpressed cells (Fig. 3D). Together, these results suggest that YAP promotes a dedifferentiated mesenchymal phenotype.

Neuroblastoma cells with a mesenchymal phenotype represent the most chemotherapy-resistant population (29, 32–34). Given YAP is increased at relapse and under stem-like media conditions, we evaluated whether YAP expression and transcriptional activity affected neuroblastoma response to chemotherapy. We treated neuroblastoma cells harboring YAP knockdown or CV in vitro with cytotoxic agents used in high-risk neuroblastoma first-line therapy or in relapsed salvage therapy, melphalan, irinotecan, topotecan, and etoposide, and evaluated for cell death. YAP genetic knockdown or YAP overexpression neither sensitizes nor promotes resistance to cytotoxic therapy, respectively, in vitro (Fig. 4A; Supplementary Fig. S4A–S4D). Short hairpin RNA (shRNA) inhibition of YAP also fails to augment apoptosis in NLF shYAP cells treated with etoposide in vitro (Supplementary Fig. S5). YAP has been shown to transcriptionally upregulate prosurvival multidomain BCL2 family members (BCL2 and BCL-XL) in other cancers (35, 36), however, YAP knockdown does not affect baseline BCL2 family prosurvival protein expression in neuroblastoma (Supplementary Fig. S6A). Although YAP knockdown leads to an increase in proapoptotic BIM protein expression, it does not alter BIM binding to prosurvival BCL2 members that are known to prevent apoptosis (Supplementary Fig. S6B; ref. 37).

Given the heterogeneity of solid tumors due to their TME (38), we ascertained whether YAP affects cytotoxic response in vivo. We treated SK-N-AS CV and shYAP xenografts with cyclophosphamide and monitored tumor growth. SK-N-AS shYAP xenografts were significantly more sensitive to cyclophosphamide in vivo compared with control xenografts (Fig. 4B). Because of strong evidence in other cancers for an interaction between YAP and oncogenic RAS (39), we queried whether YAP expression also affects MEK inhibitor therapy response. We treated the NRAS-mutated SK-N-AS shYAP and CV cells with the MEK inhibitor, trametinib, both in vitro and in vivo. Similar to chemotherapy responses, there was no change in trametinib-induced cell death in vitro with YAP genetic inhibition in neuroblastoma cell lines harboring NRAS (SK-N-AS) and NF1 (NLF) mutations (Fig. 4C). In contrast, SK-N-AS shYAP xenografts were significantly more sensitive to trametinib in vivo with more tumor regression in YAP shRNA–transduced cell line xenografts compared with SK-N-AS CV xenografts (Fig. 4D). Importantly, for both the cytotoxic and trametinib xenograft studies, the decrease in tumor growth in response to therapy in the YAP-knockdown models remained significant after considering the delay in normal tumor growth imparted by YAP knockdown alone. Therefore, YAP knockdown has an increased effect on cytotoxic and trametinib responses in situ, supporting a potential role for YAP-influencing responses to stresses imparted by the 3-dimensional microenvironment in neuroblastoma.

To understand genes regulated by YAP that may be contributing to the more profound effects seen on tumor growth and treatment response in vivo, we performed RNA-seq on SK-N-AS shYAP#1 and #2 models and SK-N-AS CV cells. Hierarchical clustering analysis revealed that the gene expression profiles of the SK-N-AS shYAP#1 and SK-NAS shYAP#2 were more similar than SK-N-AS CV cells (Fig. 5A). While the direction of gene expression (increased or decreased) in the two YAP-knockdown models was not identical, this was likely due to the higher degree of YAP genetic inhibition in shYAP#1 compared with shYAP#2 (Fig. 5B). We, therefore, focused on the shYAP#1-knockdown model and compared it with SK-N-AS CV RNA-seq results to prioritize our evaluation of genes affected by changes in YAP expression, and then confirmed those genes were expressed similarly in SK-N-AS shYAP#2. As expected, YAP was one of the downregulated genes in the Hippo signaling pathway analysis, confirming effective knockdown in this model (Fig. 5C). Hippo pathway genes that are YAP transcriptional targets, such as SMAD1, and the stemness gene, PPP2R2B, were downregulated with YAP knockdown, again supporting inhibition of YAP transcription (Fig. 5C). We evaluated for changes in apoptosis genes and confirmed that BCL2 antiapoptotic members were not altered upon YAP knockdown and that proapoptotic BIM (BCL2L11) expression increased, as was seen at the protein level (Fig. 5D; Supplementary Fig. S6A).

Given that tumor environmental factors, such as hypoxia and nutrient deprivation, can contribute to solid tumor therapy resistance in situ, and that increased BIM in shYAP neuroblastoma cells failed to enhance in vitro apoptosis from chemotherapy, we honed in on other YAP-regulated genes that mediate apoptosis in response to TME stress. Notably, HRK, a BH3-only prodeath protein that activates mitochondrial apoptosis in response to cytokine deprivation (40, 41), had one of the highest fold increases in expression of all genes altered between the YAP shRNA neuroblastoma models compared with CV (Fig. 5D; Supplementary Tables S3 and S4). We confirmed upregulated HRK in both shYAP#1 and #2 models compared with CV (Fig. 5E). We queried a high-risk neuroblastoma primary tumor gene expression dataset (R2 database: http://r2.amc.nl) and found that patients with high-risk neuroblastoma with decreased tumor expression of HRK had significantly worse overall survival, supporting a potential tumor suppressive role for HRK in neuroblastoma. In that same dataset, high YAP expression portended a worse prognosis, supporting the reciprocal relationship (Fig. 5F). HRK tumor expression did not correlate with MYCN status, but was noted to have lower expression in higher staged tumors (Supplementary Fig. S7).

To confirm RNA-seq data, we evaluated SK-N-AS and NLF shYAP models for HRK expression and found that HRK gene expression was indeed upregulated when YAP was inhibited genetically. We complemented this analysis by showing HRK gene expression decreases significantly in the IMR5 cell line with YAP overexpression (Fig. 6A). HRK is a BH3-only prodeath protein that is activated by environmental stresses such as cytokine depletion and severe hypoxia in neurons, conditions that are often found in aggressive solid tumors such as neuroblastoma (40, 41). We, therefore, evaluated neuroblastoma cells with and without YAP knockdown for their survival in both hypoxic and normoxic conditions and found no difference in apoptosis (Supplementary Fig. S8). Given the role of metabolic stress in the tumor environment, we evaluated neuroblastoma cells with and without YAP knockdown for their responses to serum starvation. Upon 48 hours of serum starvation, NLF shYAP showed increased apoptosis in low serum (0.1% FBS) conditions compared with YAP-expressing NLF CV cells in low serum conditions (Fig. 6B). Importantly, under serum starvation, HRK gene expression increased significantly in both NLF shYAP and SK-N-AS shYAP cells compared with their control counterparts (Fig. 6C). Furthermore, cytochrome c release and caspase-9 activation were increased in serum-starved shYAP neuroblastoma cells, confirming that apoptosis proceeds through the intrinsic apoptosis pathway, where HRK-mediated apoptosis occurs (Fig. 6D; Supplementary Fig. S9; ref. 42). Most importantly, apoptosis in response to etoposide treatment in vitro was significantly increased in NLF shYAP and SK-N-AS shYAP cells, but not in control cells treated with etoposide under serum-starved conditions (Fig. 6E).

We initially showed that YAP expression increases in relapsed PDXs compared with matched diagnostic tumors and YAP also increases in response to a single cycle of chemotherapy in vivo (Fig. 1A and B). We, therefore, evaluated for HRK expression in these same models. The inverse correlation for YAP and HRK holds true in vivo, as HRK expression was significantly decreased in NBX-4R PDX compared with the NBX-4 PDX (Fig. 6F). Furthermore, following topotecan and cyclophosphamide treatment in vivo, not only did YAP expression increase, but HRK gene expression concurrently decreased in the chemotherapy-treated NBX-4 PDX (Fig. 1B and F). Thus, YAP may be driving the aggressive nature of recurrent neuroblastoma through modulation of in vivo stress conditions, promoting tumor growth and therapy resistance via inhibition of HRK.

Despite the finding of increased YAP transcriptional activity (4), the role of YAP in relapsed neuroblastoma has not been well-characterized. Here, we describe a role for YAP in regulating neuroblastoma tumor growth and therapy response in vivo that may, in part, be due to YAP repression of the proapoptotic protein, HRK. We have confirmed that YAP transcriptional activity increases in relapsed neuroblastoma models, and now show that YAP expression and transcriptional activity immediately increase in response to standard high-risk neuroblastoma chemotherapy in vivo. Given the brief period of drug exposure and the continued growth of tumors throughout chemotherapy or vehicle treatment, our findings suggest that increased YAP expression in the chemotherapy-treated tumors may represent a cell intrinsic response to treatment rather than a therapy-induced cell death and selection of a YAP-expressing drug-resistant clone.

YAP is necessary for organ growth during the embryonic stages of development (43) and is involved in normal neural crest migration and fate (44). YAP also contributes to the mesenchymal state in pediatric solid tumors like rhabdomyosarcoma (39). Our results concur with such findings in neuroblastoma, with both neurosphere studies and RNA-seq showing the transcription of stem-like genes to be dependent on YAP expression in neuroblastoma. Importantly, we have uniquely shown that mesenchymal properties such as colony formation, tumor growth, and stem cell gene expression are negatively affected upon YAP genetic knockdown, supporting a potential tumor-initiating role for YAP in neuroblastoma.

A hallmark of the mesenchymal phenotype in neuroblastoma and other cancers is extreme therapy resistance (29, 45, 46). This may be attributed to YAP's transcriptional regulation of mesenchymal genes that are also known to impart therapy resistance, such as EGFR. For example, YAP transcriptionally upregulates EGFR in esophageal cancer to mediate resistance to chemotherapy (13). Our results also show that EGFR is overexpressed in YAP-expressing neuroblastoma cells and upregulated in PDXs in response to chemotherapy treatment. Yet, EGFR expression does not change in response to YAP knockdown or overexpression, perhaps because EGFR can be influenced by many additional pathways (47). Despite the persistence of EGFR in SK-N-AS shYAP xenografts, the shYAP tumors regress more in response to trametinib treatment compared with control, suggesting that alternative YAP-regulated genes are responsible for in vivo therapy resistance, such as the suppression of HRK. YAP has also been shown to promote therapy resistance through inhibition of mitochondrial apoptosis through upregulation of prosurvival BCL2 family genes (35, 36). In neuroblastoma, however, we show that YAP knockdown does not affect prosurvival BCL2 family gene or protein expression and it does not restore apoptosis in response to multiple cytotoxic agents in vitro. Despite the increased expression of BIM with YAP knockdown, BIM remains sequestered to and inactivated by MCL1, which may explain the lack of in vitro chemosensitization. BIM overexpression has also been a marker of trametinib sensitivity in other cell models (48, 49), yet the lack of in vitro sensitization to trametinib with YAP knockdown again negates BIM as the etiology for in vivo trametinib sensitization by YAP genetic inhibition.

YAP regulates MEK inhibitor resistance in multiple adult cancers (50, 51). Coggins and colleagues recently demonstrated that YAP transcription is activated in response to trametinib therapy in RAS-mutated neuroblastoma cell lines in vitro and YAP reciprocally induces MEK inhibitor resistance through transcriptional activation of MYCN and E2F in an NF1-mutated MYCN-amplified neuroblastoma cell treated with trametinib (52). We extend this to show that in neuroblastoma PDXs without RAS-hyperactivating mutations, YAP expression and transcriptional activity also increase in response to cytotoxic chemotherapy. Coggins and colleagues also show that complete knockout of YAP in NLF sensitized the cells to trametinib in vitro (52). While complete knockdown of YAP in NLF was not achieved by us, we provide data showing that complete knockdown of YAP in NRAS-mutated SK-N-AS and overexpression of YAP in NF1-mutated IMR5 and NGP did not significantly affect cell proliferation, chemotherapy response, or trametinib response in vitro. Importantly, we show enhanced tumor regression to trametinib in MYCN-nonamplified NRAS-mutated SK-N-AS xenografts when YAP was genetically inhibited in vivo. Our results, therefore, both build upon others' findings and add new dimensions by confirming that this combination holds true in vivo and highlighting the therapeutic potential of YAP and MEK coinhibition in all RAS/RAF/MAPK-mutated neuroblastomas, regardless of MYCN status.

The differences in our cell culture and xenograft responses to YAP knockdown lead us to conclude that YAP's influence on cytotoxic and targeted therapy resistance in vivo may be due to its effect on tumor environmental factors that notably contribute to solid tumor therapy resistance in situ (38). Apoptosis pathway analyses showed that gene expression of HRK significantly increased upon YAP knockdown. Rather than being activated in response to chemotherapy or radiation, HRK instead is activated and promotes apoptosis in response to environmental stresses such as cytokine withdrawal, nutrient deprivation, and hypoxia, tumor properties that have all been shown to promote therapy resistance in vivo (38, 40, 41). Therefore, we focused on HRK and its potential effects on neuroblastoma therapy response. Under serum deprivation conditions, we demonstrate that HRK expression increases, mitochondrial apoptosis ensues, and cells are more sensitive to chemotherapy in vitro, but only in neuroblastoma cells with YAP knocked down. These environmental stresses may explain the more profound differences in tumor growth and xenograft response to therapy seen in vivo, but not in vitro (in normal media conditions) when YAP is genetically inhibited. Importantly, the reciprocal relationship of HRK inhibition in the setting of YAP expression was maintained in vivo. When YAP expression was increased, HRK was downregulated in the relapsed compared with diagnostic PDX paired tumors. HRK also decreases in response to chemotherapy treatment of a chemotherapy-naïve PDX when YAP increases.

While HRK has not been extensively investigated in neuroblastoma or other cancers, primary tumor gene expression data confirm a significantly negative impact on high-risk neuroblastoma patient survival when HRK is low, legitimizing HRK as a potential tumor suppressor in neuroblastoma. HRK is inactivated in other cancers via epigenetic silencing through promoter methylation (53, 54). Preliminary methylation analysis of the HRK promoter region following PCR amplification reveals that this region is not methylated at baseline in YAP-expressing neuroblastoma cells (Supplementary Fig. S10). We have also interrogated TF binding site databases (JASPAR, TFBIND, and ALGGEN), which do not show putative TEAD binding sites (CATTCCA/T-3′) on the HRK promoter. YAP and TEAD1–4 family members have been described to bind to other transcriptional cofactors (such as MEF2, AP-1, and SRF) that can bind to the promoter and enhancer regions of target genes and recruit YAP/TEAD to regulate these genes (55). While TEAD might not bind to the HRK promoter directly, there may be a potential for other transcriptional cofactors to directly interact with HRK and recruit YAP/TEAD to suppress HRK transcription (55, 56). Therefore, investigations are ongoing to deduce the direct versus indirect regulation of HRK by YAP/TEAD and other potential coregulators.

Overall, these data demonstrate a novel role for YAP in suppressing HRK to restrict stress-induced apoptosis in vitro and inhibit therapy response and promote tumor growth in vivo. This strongly supports YAP as a logical therapeutic target to combine with MEK inhibitors or chemotherapy, and HRK as a novel tumor suppressor worthy of further study in neuroblastoma and perhaps other aggressive solid malignancies.

No potential conflicts of interest were disclosed.

J. Shim: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. J.Y. Lee: Data curation, formal analysis, methodology. H.C. Jonus: Data curation, formal analysis, methodology. A. Arnold: Data curation. R.W. Schnepp: Conceptualization, writing-review and editing. K.M. Janssen: Data curation. V. Maximov: Data curation. K.C. Goldsmith: Conceptualization, resources, supervision, funding acquisition, validation, visualization, methodology, project administration, writing-review and editing.

This work was supported by CURE Childhood Cancer Foundation and Hyundai Hope on Wheels (to K.C. Goldsmith) and Hyundai Hope on Wheels and Emory Pediatrics Fellow Research Fund (to J. Shim). We are grateful for the thoughtful advice and reviews on our work by Drs. Lawrence Boise and Anna Kenney.

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.