The yes-associated protein (YAP) is associated with resistance to anti-GD2 immunotherapy in neuroblastoma through downregulation of ST8SIA1

ABSTRACT

Pediatric patients with high-risk neuroblastoma often relapse with chemotherapy-resistant, incurable disease. Relapsed neuroblastomas harbor chemo-resistant mesenchymal tumor cells and increased expression/activity of the transcriptional co-regulator, the Yes-Associated Protein (YAP). Patients with relapsed neuroblastoma are often treated with immunotherapy such as the anti-GD2 antibody, dinutuximab, in combination with chemotherapy. We have previously shown that YAP mediates both chemotherapy and MEK inhibitor resistance in relapsed RAS mutated neuroblastoma and so posited that YAP might also be involved in anti-GD2 antibody resistance. We now show that YAP genetic inhibition significantly enhances sensitivity of mesenchymal neuroblastomas to dinutuximab and gamma delta (γδ) T cells both in vitro and in vivo. Mechanistically, YAP inhibition induces increased GD2 cell surface expression through upregulation of ST8SIA1, the gene encoding GD3 synthase and the rate-limiting enzyme in GD2 biosynthesis. The mechanism of ST8SIA1 suppression by YAP is independent of PRRX1 expression, a mesenchymal master transcription factor, suggesting YAP may be the downstream effector of mesenchymal GD2 resistance. These results therefore identify YAP as a therapeutic target to augment GD2 immunotherapy responses in patients with neuroblastoma.

KEYWORDS:

• chemoimmunotherapy
• dinutuximab
• GD2
• neuroblastoma
• relapse
• YAP

Introduction

Clinical outcomes for children with the extracranial solid tumor neuroblastoma remain unsatisfactory. Following intensive multimodal treatment, greater than half of patients with high-risk neuroblastoma relapse with a substantially reduced chance for cure.^1–4 To improve outcomes for these patients requires a greater understanding and therapeutic targeting of pathways regulating disease recurrence. Relapsed neuroblastoma is characterized by an increased frequency of genomic alterations that activate the RAS-MAPK pathway, such as activating mutations in ALK, KRAS, NRAS, HRAS, PTPN11 and inactivating mutations of NF1 and PTPN14.^5–7 In its active state, PTPN14 inhibits the nuclear localization of the Yes-associated protein (YAP) to prevent YAP-mediated transcription.^6–10 Accordingly, the same genome-wide association studies of relapsed neuroblastomas also identified a significant increase in YAP transcriptional activity, suggesting a potential role for YAP in recurrent neuroblastoma.¹⁰

YAP is a transcriptional co-regulator that primarily binds to TEAD family transcription factors.^11,12 YAP and TEAD transcriptionally activate or repress downstream target genes, contributing to cell proliferation, self-renewal and survival in many cancers, including neuroblastoma.^13,14 YAP is highly expressed in neuroblastoma cells that demonstrate an undifferentiated mesenchymal phenotype, which is characteristically chemotherapy resistant.^10,15 Using paired high-risk neuroblastoma tumors derived from the same patient at diagnosis and at tumor recurrence following chemotherapy, we have previously shown increased YAP expression and transcriptional activity at relapse.¹⁶ Genetic inhibition of YAP delayed tumor growth and sensitized NRAS-mutated neuroblastoma xenografts to cytotoxic chemotherapy and MEK inhibitor treatment in vivo, yet failed to have the same effects in vitro, suggesting YAP plays a crucial role driving therapy resistance within the solid tumor microenvironment (TME).^16,17 RNA sequencing of neuroblastomas with and without YAP genetic knockdown revealed that YAP suppresses the BH3 pro-death gene, HRK, to attenuate chemotherapy and MEK inhibitor responses in vivo. ¹⁶ Therefore, YAP upregulation following chemotherapy and relapse promotes therapy resistance in high-risk neuroblastoma through transcriptional repression of genes that play a role in the TME.

A common approach to treating patients with chemotherapy resistant, relapsed neuroblastoma uses immunotherapies targeting neuroblastoma-specific tumor antigens. The glycosphingolipid GD2 is expressed on the surface of neuroblastomas,^18–21 and the introduction of humanized monoclonal antibodies targeting GD2 (i.e. dinutuximab) significantly improved survival for newly diagnosed patients with high-risk disease.^22,23 Anti-GD2 antibodies have also been combined with cytotoxic chemotherapy (“chemoimmunotherapy”), which demonstrated impressive response rates for relapsed neuroblastoma and resulted in GD2 chemoimmunotherapy becoming the most widely used salvage therapy for patients with refractory or relapsed disease.^3,24 Unfortunately, not all patients respond to GD2-targeting immunotherapies and robust biomarkers of response are so far lacking, leaving many to suffer toxicities with no clinical antitumor benefit.^23,25,26

Resistance to immunotherapy can be caused by lack of expression or downregulation of the cell surface target of interest.²⁷ Indeed, GD2 can become downregulated following therapy and neuroblastoma recurrence.^28–30 Recent studies also suggest that mesenchymal neuroblastomas resist GD2-targeted therapies via inhibition of GD2 synthesis, yet the role for YAP, a canonical mesenchymal marker, has not been explored.³¹ Given the increased expression and activity of YAP in relapsed neuroblastoma, and its role in mediating cytotoxic and targeted therapy resistance, we posited that YAP plays a role in GD2 immunotherapy response. Here, we demonstrate for the first time that YAP genetic inhibition sensitizes neuroblastomas to anti-GD2 antibody in vitro and in vivo. We further show that YAP transcriptionally suppresses ST8SIA1 that encodes GD3 synthase, the rate-limiting enzyme for GD2 synthesis, supporting that YAP inhibition can be leveraged therapeutically to enhance patient responses to immunotherapeutic approaches targeting GD2.

Materials and methods

Cell culture

Human-derived neuroblastoma cell lines, CHLA-255, NLF, and SK-N-AS were cultured in RPMI supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37°C, 5% CO₂. Cell lines were routinely STR genotyped and resulting identities were confirmed to match the COG cell line database (cccells.org). Cells were also verified to be free of Mycoplasma contamination using the MycoAlert contamination kit (Lonza).

Generation of stably transduced cell lines

YAP was stably inhibited genetically through short hairpin RNA (shRNA) as previously described.¹⁶ ST8SIA1 was genetically inhibited stably in shYAP SK-N-AS cells as previously published using two independent constructs expressing ST8SIA1-targeting shRNAs (Genecopoeia LVRU6H-b (shGD3S–1) and LVRU6H-c (shGD3S–2)) and a hygromycin selection marker.¹⁶ The equivalent non-targeting control vectors were transduced appropriately (Sigma SHC016 (control) and Genecopoeia CSI-neg-LVRU6H (LV control)). Cells with successful lentiviral transduction were selected with 2ug/mL puromycin (YAP constructs) and 150ug/mL hygromycin (ST8SIA1 constructs).

Western blot analysis

Neuroblastoma cells were harvested with versene (ThermoFisher Scientific) and lysed in CHAPS buffer (10 mM HEPES, 150 mM NaCl, 2% CHAPS) supplemented with 1% PMSF, 1% Protease Inhibitor Cocktail (Roche), and 4% sodium orthovanadate on ice for 2 hours. Debris was cleared from resulting lysates by centrifugation at 8000 rcf for 15 mins. Protein concentration was quantified by Bradford assay. 25 ug of total protein was loaded on 4–12% NuPage Bis-Tris gels (ThermoFisher Scientific) and electrophoresed at 200 V for 35 mins. Separated proteins were transferred onto polyvinylidene difluoride (PVDF) membranes at 30 V for 90 minutes. Primary antibodies were diluted in 5% blocking buffer (Bio-Rad) in tris-buffered saline with Tween 20 (TBST) overnight and secondary anti-rabbit or anti-mouse HRP for 2 hours as appropriate. Membranes were imaged by chemiluminescence using Pierce ECL substrate (Thermofisher Scientific). See Supplemental Table S1 for antibody information.

Gamma delta (γδ) T cell expansion

γδ T cells were expanded under our 12-day protocol as previously described with αβ T cell depletion on day 6 of culture from healthy donor peripheral blood mononuclear cells.³² The expanded γδ T cell population was profiled by flow cytometry with antibodies: CD3-BV421, CD56-APC-R700, CD16-BV480, and αβ-TCR-PE or γδ-TCR-PE and used between days 12 and 14 in the cytotoxicity assays described below.³² See Supplemental Table S2 for antibody information.

Cytotoxicity assays

Bioluminescence-based

GFP-luciferase-tagged neuroblastoma cell lines were plated at 34,000/well in RPMI supplemented with 10% heat-inactivated FBS in 96-well plates and allowed to adhere overnight. The following day, γδ T cells were added at increasing effector-to-target (E:T) ratios (0:1, 1:1, and 5:1), with and without 5ug/mL dinutuximab. Co-cultures were incubated for 4 hours prior to the addition of luciferin (75ug/mL, PerkinElmer) for detection of viable target (NB) cells. Luminescent signal was detected using the Promega GloMax™-Multi Detection System. The calculation of death was performed using the following formula: %specific lysis = 100 × (spontaneous death RLU – test RLU)/(spontaneous death RLU – maximal killing RLU) where RLU is an abbreviation for relative luminescence units.

Flow cytometry-based

Neuroblastoma cells were labeled with Violet Proliferation Dye 450 (VPD450, BD Biosciences) and plated in RPMI supplemented with 10% heat-inactivated FBS at 200,000 cells/well in 24-well plates and allowed to adhere overnight. The following day, fresh γδ T cells from expansion day 12 or 14 were added to neuroblastoma cells for co-culture at increasing E:T ratios (γδ T cells-to-neuroblastoma cells) (0:1, 1:1 and 5:1) in the presence and absence of dinutuximab (5 ug/mL, United Therapeutics). Cells were incubated together for 4 hours and then harvested with accutase (GeminiBio). Cells were washed with PBS and resuspended in Annexin V binding buffer (Biolegend), stained with Annexin V-APC antibody (Biolegend) and analyzed immediately on the Aurora Cytek spectral flow cytometer. Prior to acquisition, BD Via-Probe cell viability solution (BD Biosciences) was added to the cell suspension. Unmixing of flow cytometry data was performed at the cytometer with further data analysis and gating performed using FlowJo v10.8.1 (FlowJo, LLC) software. See Supplemental Figure S1 for gating strategy.

Detection of human IFNγ by enzyme-linked immunosorbent assay (ELISA)

Human IFNγ was detected using a commercial kit (Biolegend). Supernatants were harvested from cytotoxicity assays in which SK-N-AS control, shYAP1 and shYAP2 neuroblastoma cells were co-cultured with γδ T cells at E:T ratios of 0:1, 1:1, and 5:1 for 4 hours. Briefly, supernatants were centrifuged to remove cell debris. IFNγ standards were generated by reconstituting recombinant IFNγ (Biolegend) in sterile deionized water. Concentrations of IFNγ in samples and standards were determined per manufacturer’s instructions. The BioTek Synergy Mx Microplate reader was used to read absorbance at 450 nm.

Flow cytometry

GD2 staining of neuroblastoma cell lines and xenografts

Cells were harvested with versene (Gibco), washed in phosphate-buffered saline (PBS), followed by resuspension in FACs buffer (PBS, 10% FBS, 0.1% sodium azide, 5 mM EDTA), and then stained with the live-dead stain, fixable viability stain 780 (BD Biosciences), by incubation at room temperature protected from light. Cells were then washed and stained with Isotype-BV421/GD2-BV421 only for in vitro GD2 characterization and CD45-PerCP-Cy5.5, CD56-PE, CD81-FITC, Isotype-BV421/GD2-BV421, for in vivo GD2 characterization at room temperature, washed twice in FACs buffer and resuspended for data acquisition on the Cytek Aurora 5-laser spectral flow cytometer. Negative controls were fluorescence minus one (FMO) controls for NBx28r and SKNAS CDX, unstained for NBx14r, NBx27, NBx 34r (due to lack of tissue availability). All neuroblastoma patient-derived xenografts (PDXs) were passage 2 or less. Data were analyzed using FlowJo version 10.9.0. See gating strategy (in vitro) in Supplemental Figure S2. See Supplemental Table S3 for antibody information.

Determination of γδ T cell activation state (CD107a staining)

Neuroblastoma cells (SK-N-AS control, SK-N-AS shYAP1, and shYAP2) were co-cultured with γδ T cells at an E:T ratio of 1:1 as described in the flow cytometry-based cytotoxicity assay protocol above. CD107a-PE Cy7 antibody was added to each well 30 minutes after the cytotoxic assay was started. GolgiStop (BD Biosciences) was added one hour later at a final concentration of 0.7uL/mL. At the endpoint, γδ T cells were harvested and stained with CD3-BV421, CD56-APC-R700, and γδ TCR-PE, washed twice in FACs buffer and resuspended for data acquisition on the Cytek Aurora 5-laser spectral flow cytometer. See Supplemental Table S2 for antibody information. Data were analyzed using FlowJo version 10.8.1.

Extensive characterization of γδ T cells pre- and post-cytotoxicity assay

Neuroblastoma cells (SK-N-AS control, SK-N-AS shYAP1, and shYAP2) and γδ T cells were co-cultured at a 1:0 or 1:1 E:T ratio for 24 hours. The γδ T cells were harvested at 24 hours and profiled using our previously published extensive characterization panel.³² See Supplemental Table S4 for antibody information.

RT-qPCR

RNA was extracted from neuroblastoma cell lines using the TRIzol (Ambion)-chloroform (Millipore Sigma) extraction method and quantified using a NanoDrop 2000 (Thermo Scientific). cDNA was prepared from 2ug RNA by using the high-capacity cDNA reverse transcription kit (Applied Biosystems) per manufacturer’s protocol. For real-time qPCR, SYBR green reagent (Applied Biosystems) was used with the primers listed in Supplemental Table S5. Gene expression was normalized to GAPDH and HPRT using the CFX96 Touch Real-Time PCR Detection System software (Bio-Rad).

Mouse xenograft in vivo studies

All animal studies were conducted in accordance with policies set forth by the Emory University Institutional Animal Care and Use Committee (IACUC). Our protocol was approved by the Emory IACUC (PROTO201700089). Euthanasia was performed by asphyxiation with CO₂ and cervical dislocation. 4 × 10⁶ SK-N-AS cells were combined at a 1:1 ratio (by volume) with Matrigel (Corning) and injected subcutaneously into the flank of 4–6-week-old female and male NOD-scid IL2Rgamma^null (NSG) mice (The Jackson Laboratory). Tumor volume was calculated using the formula: length x width x height x p/6. When tumors grew to a volume of 100–200 mm³, mice were randomized to the treatment groups. Mice receiving the full regimen were treated on days 1, 4, 7 and 10 with 75 mg/kg cyclophosphamide (McKesson) intraperitoneally; on days 2, 8, and 14 with 100ug dinutuximab intravenously; on days 3, 6, 9, 12, 15, and 18 with 2.5 × 10⁶ γδ T cells (expanded from healthy human blood as described above) intratumorally. Mice were sacrificed when tumor burden reached IACUC-prescribed limit based on tumor volume (1500 mm³) and physical burden. Tumors were harvested and mechanically dissociated to extract RNA and protein to perform RT-qPCR, western blots, and flow cytometry as described above.

Statistical analyses

GraphPad Prism v9.4.1 was used to perform all statistical analyses. For pairwise comparisons throughout, unpaired t-tests with Welch’s correction were calculated. Kaplan–Meier survival plots were generated for in vivo investigations, and log rank test was performed to determine statistical significance.

Results

Neuroblastoma cell lines that express high YAP and low GD2 are resistant to dinutuximab and gamma delta (γδ) T cell treatment in vitro

Given the increased expression and activity of YAP in relapsed neuroblastoma and its influence on chemotherapy response, we sought to determine if YAP might also play a role in GD2 immunotherapy response. We first evaluated YAP and GD2 expression in three neuroblastoma cell lines: SK-N-AS, NLF, and CHLA-255. YAP protein expression was high in SK-N-AS (MYCN non-amplified) and NLF (MYCN amplified) while undetectable in CHLA-255 (MYCN amplified) (Figure 1A). GD2 cell surface expression was conversely low in neuroblastoma cells with high YAP expression, SK-N-AS and NLF, and high in CHLA-255 that expresses no YAP (Figure 1B). Gamma delta (γδ) T cells are an innate effector immune cell subset that can regulate antibody-dependent cell-mediated cytotoxicity (ADCC). γδ T cells have been shown by our group to synergize with dinutuximab against neuroblastoma models both in vitro and in vivo.^32,33 We therefore used ex vivo expanded γδ T cells as the immune effectors in combination with dinutuximab in these investigations. Agnostic of MYCN amplification state, GD2^low/YAP^high cell lines, SK-N-AS and NLF were resistant to γδ T cell-induced specific lysis with and without dinutuximab treatment (Figure 1C). Contrastingly, γδ T cells alone induced specific lysis of GD2^high/YAP^low cell line CHLA-255, with γδ T cell-mediated specific lysis significantly enhanced by the addition of dinutuximab at both 1:1 and 5:1 effector: target (E:T) ratios (Figure 1C).

Figure 1. YAP expression is high in neuroblastoma cell lines that are resistant to anti-GD2/γδ T cell immunotherapy. A, Western blot of YAP expression in the neuroblastoma cell lines, SK-N-AS, NLF, and CHLA-255. GAPDH is the loading control. B, Mean normalized fluorescence of GD2 cell surface expression by flow cytometry in SK-N-AS, NLF, and CHLA-255. C, Percentage specific lysis after 4-hour cytotoxicity assays between γδ T cells (effector) and the neuroblastoma cell lines (target), CHLA-255, NLF and SK-N-AS at effector: target (E:T) ratios of 0:1, 1:1, and 5:1 with or without the addition of the anti-GD2 monoclonal antibody, dinutuximab. For CHLA-255, 1:1, *p = 0.0235, 5:1, **p = 0.0072. All other differences are not statistically significant.

YAP inhibition sensitizes neuroblastomas to dinutuximab and γδ T cells in vitro and in vivo through upregulation of GD2 cell surface expression

Based on the inverse correlation of YAP and GD2 expression in neuroblastomas and differential dinutuximab responses, we evaluated the role for YAP in dinutuximab response through genetic knockdown. Using GD2^low and dinutuximab-resistant SK-N-AS that harbors an activating NRAS Q61K mutation, we generated stable YAP knockdown models using short hairpin (sh)RNA. Western blot analysis confirmed genetic inhibition of YAP in SK-N-AS cells selected to stably express YAP-silencing shRNA (shYAP1, shYAP2) compared to a non-targeting, scrambled control (Figure 2A). Increased cytotoxicity of γδ T cells alone was observed in SK-N-AS shYAP1 and shYAP2 cells at an E:T ratio of 5:1 compared to SK-N-AS control (Figure 2B). The addition of dinutuximab in the coculture further augmented cytotoxicity of YAP-inhibited SK-N-AS by the γδ T cells (Figure 2B).

Figure 2. Genetic inhibition of YAP increases in vitro response to anti-GD2/γδ T cell immunotherapy with corresponding upregulation of GD2 surface expression in SK-N-AS. A, Western blot of YAP expression in control- (SK-N-AS control) and shYAP-transduced cells (SK-N-AS shYAP1 and shYAP2). GAPDH is the loading control. B, Percentage apoptosis of neuroblastoma target cells, SK-N-AS control, shYAP1 and shYAP2 when co-cultured for 4 hours with γδ T cells, with (+ DIN) and without dinutuximab, 1:1 ±DIN: control vs shYAP1, *p = 0.0292, control vs shYAP2, *p = 0.0327; 5:1: control vs shYAP1, *p = 0.0146, control vs shYAP2, *p = 0.0185; 5:1 ±DIN: control vs shYAP1, **p = 0.0054, control vs shYAP2, **p = 0.0022; Data represent mean ± standard error of n = 3 independent experiments with two technical replicates per condition, student’s T-test with Welch’s correction. All other comparisons were not significant. C, Representative graph showing mean normalized fluorescence of GD2 cell surface expression by flow cytometry in SK-N-AS control, shYAP1 and shYAP2 cell lines. Lighter colors indicate isotype controls and darker colors indicate GD2-BV421 staining. The dotted line demarcates the GD2-bright population which was quantified in D and E. D, Quantification of median fluorescence intensity (MFI) of the GD2-bright population in SK-N-AS control, shYAP1 and shYAP2 cell lines: SK-N-AS control vs shYAP1: *p = 0.0215, SK-N-AS control vs shYAP2: *p = 0.0193. Data represent mean ± standard error of n = 3 independent experiments. E, Percentage of GD2-bright cells by flow cytometry in SK-N-AS control, shYAP1 and shYAP2 cell lines: SK-N-AS control vs shYAP1: *p = 0.0233; SK-N-AS control vs shYAP2: *p = 0.0142. Data represent mean ± standard error of n = 3 independent experiments.

To determine the mechanism of increased sensitivity of YAP-inhibited SK-N-AS to γδ T cells both alone and in combination with dinutuximab, we first evaluated for changes in the intrinsic killing mechanisms of γδ T cells imparted by YAP knockdown in the tumor.³⁴ We performed flow cytometric analysis of CD107a, a cell surface marker of early degranulation used as a surrogate for γδ T cell activation.^35–37 No detectable differences in early degranulation were observed between γδ T cells co-cultured with SK-N-AS control or SK-N-AS^shYAP1 cells ± dinutuximab (Supplemental Figure S3). Immunophenotyping of expanded γδ T cells before and after the 24-hour co-culture with SK-N-AS control, SK-N-AS shYAP1, or SK-N-AS shYAP2 cells showed no differences or changes in γδ T cell surface expression of common markers of activation (DNAM1, NKG2D), inhibition (KIR2DL1), or exhaustion (PD1, TIM3, CTLA4, TIGIT) (Supplemental Figure S4).³² γδ T lymphocytes harbor innate receptors that recognize and bind to stress ligands on the tumor cell surface, leading to T cell activation. In addition, they express FASL that binds to death receptors expressed on tumor cells, leading to perforin and granzyme release.^38,39 To elucidate whether the mechanism of increased death of SK-N-AS shYAP cells is due to changes in tumor cell surface markers or death receptors, we assessed the expression of NKG2D receptor ligands (MICA, MICB, and ULBP1/2/5/6), as well as death receptors (TRAIL-R1/2, CD95/FAS) and DNAM1 ligands (CD112/Nectin-2 and CD155/PVR). These markers did not change with YAP knockdown in SK-N-AS (Supplemental Figure S5).

In response to major histocompatibility complex (MHC)-independent activation by tumor cells, γδ T cells can produce IFNγ.⁴⁰ IFNγ can induce apoptosis in tumor cells.⁴¹ We therefore examined IFNγ production when γδ T cells were co-cultured with SK-N-AS shYAP or control cell lines with or without dinutuximab. In the absence of dinutuximab, we observed no difference in IFNγ concentrations when γδ T cells were co-cultured with SK-N-AS control, shYAP1 and shYAP2 cells (Supplemental Figure S6A). However, in the presence of dinutuximab, a statistically significant increase in IFNγ release was observed in the shYAP1 and shYAP2 co-cultures compared to control (Supplemental Figure S6B), corresponding with the increased cytotoxicity observed in the shYAP cells exposed to dinutuximab and γδ T cells at 1:1 and 5:1 (Figure 2B).

The presence of antigen or changes in antigen density at the cell surface are essential determinants of response to therapies that depend on ADCC.⁴² Given that intrinsic killing properties of γδ T cells are not significantly changed by differences in tumor YAP expression, we focused our attention on GD2 surface expression and its potential contribution to augment dinutuximab/γδ T cell combination effects. Wild-type SK-N-AS expresses low levels of GD2 on the cell surface (Figure 1B). The median fluorescent intensity (MFI) of the GD2-bright population (defined by GD2 MFI of >10⁴ based on the brightest point in the isotype staining – dotted line, Figure 2C) significantly increased for SK-N-AS shYAP1 (mean MFI = 44136) and SK-N-AS shYAP2 (mean MFI = 39032) following YAP knockdown compared to the SK-N-AS control (mean MFI = 2115) (Figure 2C,D). Additionally, the percentage of the GD2-bright cells was higher for SK-N-AS shYAP1 (mean = 30.2%) and SK-N-AS shYAP2 (mean = 38.7%) compared to the SK-NAS control (mean = 0.63%) (Figure 2E).

Given that YAP regulates chemotherapy response within the neuroblastoma TME and response to dinutuximab and γδ T cells in vitro, we ascertained whether YAP inhibition also influences tumor response to dinutuximab and γδ T cells in vivo.¹⁶ We have previously shown that dinutuximab and γδ T cells are more effective against tumors in vivo with the addition of a cytotoxic agent, in keeping with clinical trials showing dinutuximab in combination with chemotherapy is more effective in patients with relapsed neuroblastoma compared to dinutuximab alone.^3,16,43 We treated NSG mice harboring established SK-N-AS control or shYAP subcutaneous tumors with dinutuximab, γδ T cells, and cyclophosphamide and monitored tumors for growth (Treatment schema, Supplemental Figure S7A). SK-N-AS shYAP tumors had a significant prolongation of tumor regression and survival following treatment with dinutuximab, γδ T cells, and cyclophosphamide compared to mice with SK-N-AS control tumors (p = 0.0024) (Supplemental Figure S7B). We confirmed that YAP knockdown and lower expression of its canonical target, CYR61, were maintained in the SK-N-AS xenograft tumors at experimental endpoint after tumors recurred (Supplemental Figure S7C and S7D).

YAP inhibition increases cell surface expression of GD2 through upregulation of ST8SIA1

YAP can transcriptionally repress genes involved in therapy response.^13,16 Next-generation sequencing of paired diagnostic and relapsed high-risk neuroblastomas showed a significant decrease in expression of genes normally suppressed by YAP in relapsed tumors.¹⁰ We therefore examined genes in the biosynthetic pathway of GD2 (Figure 3A), using RNA sequencing data from SK-N-AS control versus SK-N-AS shYAP1 cells.¹⁶ ST8SIA1, which encodes for the critical rate-limiting enzyme GD3 synthase (GD3S) in the GD2 biosynthesis pathway, was found to be significantly increased (logFC = 2.62; p = 5.87 × 10⁻³) in SK-N-AS shYAP1 versus SK-N-AS control (Figure 3B). We validated this finding in both SK-N-AS shYAP1 and SK-N-AS shYAP2 models by RT-qPCR (Figure 3C). Reduced expression of YAP and its canonical target CYR61 were confirmed in shYAP1 and shYAP2 cells compared to control and corresponded to significantly increased expression of ST8SIA1 (>100-fold, p < 0.01) (Figure 3C). In the same models, other genes involved in the biosynthesis of GD2 (B4GALT5/6, ST3GAL5, ST8SIA5, B4GALNT1, B3GALT4, ST3GAL2) were either marginally changed or unchanged by RT-qPCR of shYAP1 and shYAP2 compared to control (Supplemental Figure S8). Notably, the gene encoding GD2 synthase (GD2S), B4GALNT1, was unchanged. Others have shown that forced expression of the master transcription factor PRRX1 causes adrenergic-to-mesenchymal transition, leading to epigenetic suppression of genes like ST8SIA1 in neuroblastoma.⁴⁴ Interestingly, the expression of PRRX1 significantly increased with YAP knockdown for both SK-N-AS shYAP1 and shYAP2 compared to control, yet ST8SIA1 expression and GD2 surface expression were not impacted (Figure 3C).

Figure 3. YAP inhibition mediates significantly increased gene expression of the GD2 biosynthetic enzyme, ST8SIA1. A, Schematic of ganglioside biosynthesis showing genes encoding enzymes in the biosynthetic pathway of GD2. Blue denotes genes downregulated, Red denotes genes upregulated, and Gray denotes genes unchanged in RNA sequencing data: SK-N-AS shYAP1 vs control. B, Volcano plot of -log(p-value) vs log(fold change(fc)) for GD2 biosynthetic genes from RNA seq data of SK-N-AS shYAP1 compared to control. Blue denotes genes downregulated, Red denotes genes upregulated, and Gray denotes genes not significantly changed. C, Normalized gene expression as determined by RT-qPCR of YAP: SK-N-AS control vs shYAP1: **p = 0.0011, SK-N-AS control vs shYAP2: **p = 0.0040, SK-N-AS shYAP1 vs shYAP2: *p = 0.0467; YAP canonical target, CYR61: SK-N-AS control vs shYAP1: **p = 0.0078, SK-N-AS control vs shYAP2: *p = 0.0102, SK-N-AS shYAP1 vs shYAP2: **p = 0.0010; ST8SIA1: SK-N-AS control vs shYAP1: **p = 0.0021; SK-N-AS control vs shYAP2: **p = 0.0043; SK-N-AS shYAP1 vs shYAP2: *p = 0.0349; and PRRX1: SK-N-AS control vs shYAP1: **p = 0.0081; SK-N-AS control vs shYAP2: **p = 0.0085; SK-N-AS shYAP1 vs shYAP2: **p = 0.0091. Data represent mean ± standard error of n = 3 independent experiments.

To confirm that GD2 cell surface changes were the result of YAP suppression of ST8SIA1 (Supplemental Figure S9), we genetically inhibited ST8SIA1 by shRNA in the SK-N-AS shYAP2 model. YAP knockdown was maintained in the control- and shST8SIA1-lentiviral transduced SK-N-AS shYAP2 cells (Figure 4A) and successful knockdown of ST8SIA1 was achieved using two separate shST8SIA1 constructs (Figure 4B and Supplemental Figure S9). Genetic inhibition of ST8SIA1 in the SK-N-AS shYAP2 cells led to significantly decreased median fluorescence intensity (MFI) of GD2 cell surface expression in SK-N-AS shYAP/shST8SIA1–1 and the SK-N-AS shYAP/shST8SIA1–2 compared to SK-N-AS shYAP/LV control, completely reversing the phenotype of increased GD2 surface expression upon YAP knockdown (Figure 4C,D). The percentage of GD2-positive cells in SK-N-AS shYAP shST8SIA1–1 and SK-N-AS shYAP shST8SIA1–2 was also >50-fold less than SK-N-AS shYAP control (Figure 4E). Furthermore, knockdown of ST8SIA1 in SK-N-AS shYAP cells reduced their in vitro sensitivity to γδ T cells in the presence of dinutuximab compared to SK-N-AS shYAP/LV control, with no difference in neuroblastoma killing by γδ T cells in the absence of GD2-targeting antibody (Figure 4F).

Figure 4. GD3S (ST8SIA1) inhibition reverses the phenotypes of increased GD2 surface expression and sensitivity to anti-GD2/γδ T cell immunotherapy when YAP is inhibited in SK-N-AS cells. A, Western blot of YAP expression in SK-N-AS shYAP lentiviral (LV) control, shYAP shST8SIA1–1, shYAP shST8SIA1–2, and SK-N-AS WT cells. GAPDH is the loading control. B, Normalized gene expression as determined by RT-Qpcr of ST8SIA1 in dually transduced cells, SK-N-AS shYAP LV Control, SK-N-AS shYAP shST8SIA1–1, and SK-N-AS shYAP shST8SIA–2, SK-N-AS shYAP LV Control vs shYAP shST8SIA1–1: *p = 0.0155, SK-N-AS shYAP LV Control vs shYAP shST8SIA–2: *p = 0.0157, SK-N-AS shYAP shST8SIA1–1 vs shYAP shST8SIA–2: *p = 0.0460. C, Representative graph showing mean normalized fluorescence of GD2 cell surface expression by flow cytometry in SK-N-AS shYAP LV Control, SK-N-AS shYAP shST8SIA1–1, and SK-N-AS shYAP shST8SIA–2 cell lines. Lighter colors indicate isotype controls and darker colors indicate GD2-BV421 staining. D, Quantification of median fluorescence intensity (MFI) of GD2 in SK-N-AS shYAP LV Control, SK-N-AS shYAP shST8SIA1–1, and SK-N-AS shYAP shST8SIA–2 cell lines: SK-N-AS shYAP LV Control vs shYAP shST8SIA1–1: *p = 0.0242, SK-N-AS shYAP LV Control vs shYAP shST8SIA–2: *p = 0.0221. Data represent mean ± standard error of n = 3 independent experiments. E, Percentage of GD2 cell surface expression by flow cytometry in SK-N-AS shYAP LV Control, SK-N-AS shYAP shST8SIA1–1, SK-N-AS shYAP shST8SIA–2 cell lines; SK-N-AS shYAP LV Control vs shYAP LV shST8SIA1–1: **p = 0.0088, SK-N-AS shYAP LV Control vs shYAP shST8SIA–2: **p = 0.0095. Data represent mean ± standard error of n = 3 independent experiments. F, Percentage apoptosis of neuroblastoma target cells, SK-N-AS shYAP LV control, shYAP shST8SIA1–1, and shYAP shST8SIA1–2 when co-cultured for 4 hours with γδ T cells, with (+ DIN) and without dinutuximab, 0:1 +DIN: shYAP LV control vs shYAP shST8SIA1–2, **p = 0.0056; 1:1: shYAP LV control without DIN vs +DIN: **p = 0.0076; 1:1 +DIN: shYAP LV control vs shYAP shST8SIA1–1, *p = 0.0392; shYAP LV control vs shYAP shST8SIA1–2, **p = 0.0018; Data represent mean ± standard error of n = 3 independent experiments with two technical replicates per condition, student’s T-test with Welch’s correction. All other comparisons were not significant.

YAP and ST8SIA1 or GD2 cell surface expression are inversely correlated in neuroblastoma primary tumors and patient derived xenografts

We queried publicly available gene expression datasets of primary neuroblastoma tumors to validate the clinical relevance of the regulation of GD2 by YAP through ST8SIA1. An inverse relationship between YAP and ST8SIA1 expression was demonstrated in two separate datasets with non-overlapping cohorts: in the TARGET (Asgharzadeh) dataset, which consists of 249 samples assessed by exon array, R=-0.233, p = 2.05×10⁻⁴ (Figure 5A), and for the Kocak dataset, which consists of 648 samples assessed by microarray, R=-0.132, p = 7.89×10⁻⁴ (Figure 5B). Additionally, Kaplan–Meier survival analysis shows that the overall survival probability is reduced when ST8SIA1 expression is lower (Figure 5C,D).

Figure 5. YAP and ST8SIA1 expression are negatively correlated in primary neuroblastoma tumors, low ST8SIA1 expression is associated with worse overall survival, and YAP and GD2 are inversely correlated in neuroblastoma patient derived xenografts (PDXs). YAP and ST8SIA1 expression in primary neuroblastoma tumors from patients: A, TARGET (Asghardazeh) dataset: n = 249, R=-0.233, p = 2.05 × 10⁻⁴. B, Kocak dataset: n = 648, R=-0.132, p = 7.89 × 10⁻⁴. Low ST8SIA1 expression in primary neuroblastoma tumors from patients is associated with worse overall survival: C, TARGET (Asghardazeh) dataset: n = 247, Bonferroni-corrected (bonf.) p = 0.052. D, Kocak dataset: n = 476, R=-0.137, bonf. p = 3.9 × 10⁻⁴. https://r2.amc.nl. E, Western blot of YAP expression in neuroblastoma patient-derived xenografts (PDXs), NBx-14r, NBx-27, NBx-28r, NBx-34r and the SK-N-AS neuroblastoma cell line-derived xenograft (CDX). GAPDH is the loading control. F, Mean normalized fluorescence of GD2 cell surface expression by flow cytometry in neuroblastoma patient-derived xenografts (PDXs), NBx-14r, NBx-27, NBx-28r, NBx-34r and the SK-N-AS neuroblastoma cell line-derived xenograft (CDX). Grey denotes isotype control staining and white denotes GD2-BV421 staining.

Since GD2 is a glycosphingolipid and thus, not genetically encoded, we sought to determine whether YAP protein expression and GD2 surface expression also inversely correlated by performing immunoblot and flow cytometry, respectively, in low-passage neuroblastoma patient-derived xenografts (PDXs). YAP expression was lower in NBx14r and NBx28r than in NBx27 and NBx34r (Figure 5E) and correspondingly, the MFI of GD2 on the surface of NBx14r and NBx28r was higher than that of NBx27 and NBx34r (Figure 5F).

Discussion

High-risk neuroblastomas that recur are notoriously chemotherapy resistant, leading to improvements in survival focused on immunotherapy approaches. Indeed, anti-GD2 antibodies in combination with chemotherapy have resulted in unprecedented response rates in relapsed patients.²² However, challenges to GD2 targeted immunotherapies remain, such as an incomplete understanding of biomarkers predicting response and mechanisms of resistance.^30,31 High-risk neuroblastoma tumors that relapse are enriched with mesenchymal cells as well as RAS pathway mutations,^6,15,44 leading many to investigate how these properties may influence immunotherapy resistance to identify new therapeutic targets.

YAP is a canonical mesenchymal gene that encodes for the YAP protein known to cooperate with hyperactivated RAS.¹⁷ Indeed, the YAP expressing neuroblastoma cell lines we investigated are RAS pathway mutated (SK-N-AS, NLF) and harbor the mesenchymal gene signature.^15,44,45 Our findings demonstrate that YAP genetic inhibition paradoxically leads to upregulated expression of PRRX1, one of the master transcription factors that can drive the mesenchymal phenotype.⁴⁴ Previously, it was shown that genetic induction of mesenchymal neuroblastoma via overexpression of PRRX1 induces similar transcriptional downregulation of ST8SIA1 with resultant decrease of cell surface GD2.³¹ PRRX1 converts an adrenergic neuroblastoma cell to a mesenchymal neuroblastoma cell (adrenergic to mesenchymal transition) with a decrease in adrenergic-differentiating genes like PHOX2B, GATA2, DLK1 and an increase in mesenchymal stem-like genes such as SOX2, SNAI2 and YAP.⁴⁴ We now show that YAP is sufficient to suppress the same glycosphingolipid biosynthesis pathway regardless of PRRX1 gene expression, suggesting that GD2 synthesis may be more directly regulated by YAP downstream of its mesenchymal driving forces such as PRRX1 and other master mesenchymal transcription factors. Further studies are warranted to validate the full functional roles for YAP within the mesenchymal neuroblastoma phenotype.

Despite the success of dinutuximab and other anti-GD2 antibodies, not all patients respond, and preclinical data suggest that it may be due to lack of GD2 on the tumor cell surface. Detection of GD2 in primary neuroblastoma tissue is limited since GD2 is a glycosphingolipid and its presence is therefore not detectable by immunohistochemistry on paraffin embedded tissue.¹⁸ Recent studies have therefore sought to determine and provide surrogate biomarkers for GD2 expression in an effort to triage patients most likely to benefit from GD2 immunotherapy.³¹ We found that neuroblastomas with low GD2 have high YAP gene and protein expression. Furthermore, this inverse correlation has functional relevance as we found YAP to suppress GD2 expression through inhibition of the GD3 synthase (GD3S) gene ST8SIA1. By suppressing GD3S and thus GD2 synthesis, YAP indeed serves as a mediator and potential biomarker of anti-GD2 antibody resistance. We validated this relationship through genetic knockdown studies, showing YAP inhibition restores response to dinutuximab and γδ T cells both in vitro and in vivo in SK-N-AS, with the therapy resistant phenotype restored upon knockdown of ST8SIA1 in the shYAP neuroblastoma cells. We also observed a trend toward inverse correlation between YAP protein and GD2 cell surface expression in low-passage high risk and relapsed neuroblastoma PDXs. In addition, primary neuroblastomas also show an inverse correlation between genetic expression of YAP and ST8SIA1. Based on these data, we hypothesize that high YAP expression in neuroblastoma tumors may predict GD2 immunotherapy resistance clinically. Immunohistochemical staining of YAP in primary neuroblastoma tumors is feasible and should be characterized prospectively to statistically correlate results to patient outcomes following anti-GD2 therapy to concretely define it as a biomarker of response.

A slight increase in neuroblastoma cell death was observed following coculture with γδ T cells alone in the SK-N-AS shYAP cells compared to control. We therefore investigated whether increased tumor cell death was due to paracrine effects of the YAP-inhibited neuroblastoma cells toward the γδ T cells. When SK-N-AS shYAP cells were co-cultured with γδ T cells, we observed no difference in γδ T cell markers of exhaustion, activation, or apoptosis, nor an increase in tumor-resident stress antigens or FAS/TRAIL receptors. Although there was an increase in IFNγ release when SK-N-AS shYAP cells were treated with the combination of γδ T cells and dinutuximab in vitro, this increased IFNγ release was not consistently observed in the absence of dinutuximab. This suggests that differences in cytotoxicity were not due to γδ T cell-intrinsic changes, increased release of cytotoxic mediators, or by increased T cell recognition ligands on the tumor cells. Further investigation is warranted to understand the effect of tumor YAP inhibition on increased γδ T cell activity.

YAP expression by immune cells themselves has been shown to promote immunosuppression and suppress immunotherapy activity.⁴⁶ For example, high YAP expression in regulatory T cells of hepatocellular carcinoma patients was found to facilitate an immunosuppressive TME and was an indicator of poor prognosis.⁴⁷ Since YAP has a well-established role in the TME, its role in the neuroblastoma TME-enacted resistance to anti-GD2 immunotherapy warrants evaluation by using immunocompetent murine models. Indeed, future studies in additional neuroblastoma models will identify whether therapeutic targeting of YAP may be beneficial by not only making tumor cells more vulnerable through upregulation of the immunotherapy target, but also through inhibition of immune cells contributing to the immune hostile TME.

Others have identified that mesenchymal master transcription factors epigenetically suppress ST8SIA1 expression, leading clinical efforts to evaluate the combination of GD2 antibodies with epigenetic modifying agents such as EZH2 inhibitors (tazemetostat) and the histone deacetylase inhibitor, vorinostat.^31,48 These epigenetic approaches to increase GD2 surface expression may be viable for improving targeting neuroblastoma cell populations that express low cell surface GD2 (GD2^low) in patients. Additionally, dual targeting of other highly expressed neuroblastoma antigens like B7-H3 or GPC2 with GD2 could ensure tumor specificity and optimize the targeting of GD2^low neuroblastoma.^49–51 Here, we demonstrate increased ST8SIA1 expression in SK-N-AS neuroblastoma cells upon genetic inhibition of YAP and an inverse relationship between YAP and ST8SIA1 in primary neuroblastoma tumors. Further mechanistic studies in additional models will be important to confirm our findings and determine the mechanism by which YAP suppresses ST8SIA1. If confirmed, these studies would lay the foundation for combining targeted YAP inhibition as a way to sensitize GD2^low neuroblastomas to anti-GD2 immunotherapy similar to epigenetic-modifying agents.

Overall, our findings have defined a role for YAP in downregulation of ST8SIA1, rendering GD2 antibody resistance in neuroblastoma cell lines. These results support that both ST8SIA1 (GD3 synthase) and YAP warrant further investigation with regard to their expression in relation to clinical response to anti-GD2 antibody immunotherapy. Currently, inhibitors of the YAP/TEAD interaction have shown preclinical promise in adult cancers with one agent in clinical application for adult mesothelioma,⁵² and investigations are ongoing to evaluate such inhibitors in neuroblastoma. Incorporation of YAP pharmacological inhibition with novel GD2 targeting immunotherapies, such as GD2-CAR T cells or novel anti-GD2 antibody combinations (γδ T cells with dinutuximab and chemotherapy; NCT05400603) may also improve outcomes for patients with high-risk and relapsed neuroblastoma.⁵³

Acknowledgments

We thank the Children’s Healthcare of Atlanta and Emory University’s Pediatric Flow Cytometry Core for support with flow cytometry experiments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data that support the findings of this study are available from the corresponding author, KCG, upon reasonable request.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/2162402X.2023.2240678

Additional information

Funding

This work was supported by the CURE Childhood Cancer Foundation to KCG and Atlanta Pediatric Scholars Program K12 Scholar supported by the National Institutes of Health, Eunice Kennedy Shriver National Institute of Child Health and Human Development grant K12HD072245 to JS.