Intratumoral Copper Modulates PD-L1 Expression and Influences Tumor Immune Evasion

Cancer immune evasion is recognized as a central hallmark of tumor development. One mechanism that cancer cells use to protect themselves from antitumor immune responses is overexpression of programmed death ligand 1 (PD-L1). The immune checkpoint protein programmed death receptor 1 (PD-1) expressed by lymphocytes negatively regulates T-cell effector functions, leading to reduced cytokine production and cytotoxic activity against target cells, including tumor cells (1). Several therapeutic mAbs targeting PD-L1/PD-1 have been approved by the FDA for adult melanoma and lung cancer (2). However, their efficacy, particularly for PD-L1/PD-1 blockade, is limited by cancer cell resistance in many patients (3) and immune-related adverse events in others (4). There is a need to better understand the mechanisms regulating the PD-L1/PD-1 axis to augment approaches that target this pathway.

Copper (Cu) homeostasis is finely regulated in mammals and Cu imbalance is associated with pathologies as diverse as Wilson disease (severe Cu overload) or, as recently demonstrated, cancer susceptibility (5). Cu is essential for key signaling pathways that govern at least three key properties of malignant cells: proliferative immortality, angiogenesis, and metastasis. Although Cu levels in cancer tissue is known to be highly heterogeneous, elevated Cu content has been reported in a wide spectrum of cancers including neuroblastoma (6), glioblastoma (GBM; ref. 7), breast, ovarian, lung, prostate, stomach, and leukemia (5), highlighting Cu homeostasis as an emerging therapeutic target in oncology. Interestingly, Cu also plays an important role in maintaining normal immunity, but its mechanism of action has not been determined (8). Early reports showed that dietary depletion of Cu increases the susceptibility of mice to bacterial pathogens (8). However, there are no studies investigating a potential role for Cu in regulating antitumor immune responses.

Given the high levels of Cu in cancer, and the previously reported role of Cu in regulating immune cell function, our aim was to understand how elevated intratumoral Cu levels modulate the molecular pathways responsible for cancer immune evasion. In this study, we describe a role for Cu in regulating the expression of PD-L1 in cancer cells, and we demonstrate that Cu-chelating drugs have the potential to be repurposed for immune checkpoint blockade therapy.

The retrospective patient cohort consisted of 180 patients (90 neuroblastoma and 90 brain tumor cases) who underwent diagnostic biopsy or surgical resection at the Children's Hospital at Westmead, Sydney, Australia (Human Ethics approval LNR/13/SCHN/389). In the tissue microarrays (TMA), each sample was present at least in duplicate.

IHC was performed using a BOND-RX automatic immunostainer (Leica Biosystem). The slides were scanned using an Aperio CS2 virtual microscope (Leica Biosystem). The slides were incubated with copper transporter 1 (CTR-1) antibody (Abcam, ab133385, 1:100 dilution), with PD-L1 antibody (ab80276, 1:50 dilution), with MT1X antibody (Proteintech, catalog no. 17172-1-AP, 1:100 dilution) and with phospho-EGFR antibody (Cell Signaling Technology, catalog no. 3777, 1:100 dilution). The percentage of tumor cells showing positive membrane staining was determined and the intensity of staining was judged on a semiquantitative scale of 0 to 3: no staining (0), weakly positive staining (1), moderately positive staining (2), and strongly positive staining (3). The H-score was derived as the product of the percentage of positive cells and the staining intensity to produce a score out of 300. Staining for CD3, CD4, and CD8 was used to estimate the cell density of tumor-infiltrating lymphocytes, while CD16, CD244, and CD335 staining was used to estimate the density of infiltrating natural-killer (NK) cells. Staining for CD25 and FOXP3 was used to estimate the density of T-regulatory cells (Treg).

Statistical analyses were undertaken using SPSS version 22.0 (SPSS Inc.). Bivariate relationships between CTR-1–positive and PD-L1–positive cases were analyzed using cross-tabulation and the association between expression was determined using Pearson correlation coefficient (Phi coefficient) while significance was tested using Fisher exact test.

Student t test was carried out for independent samples among untreated and Dextran-Catechin (DC)-treated (at 24 and 48 hours) Th-MYCN–derived neuroblastoma tumors for PD-L1 and CTR-1 H-score, as well as immune cell density. Similarly, Student t test was used to compare Th-MYCN–derived neuroblastoma tumors untreated or treated with tetraethylenepentamine pentahydrochloride (TEPA) for the expression of PD-L1 as well as the presence of infiltrating immune cells. Statistical significance was set at P < 0.05.

Neuroblastoma cell lines SK-N-BE(2)-C, SK-N-FI, SH-SY5Y and GBM cell line U87MG were obtained from ATCC and maintained in DMEM (Thermo Fisher Scientific) with 10% FBS. All cell lines used in this study were validated by using short tandem repeat (STR) profiling at the Children's Cancer Institute and routinely screened for Mycoplasma species. Children's Cancer Institute maintains a centrally managed Cell Bank of cell lines for use by internal researchers. These cell lines are maintained as a master stock, from which, working stocks are created and supplied as cryo-preserved vials upon request. Both the master stock and working stock are STR profile–validated, Mycoplasma-free at freeze down.

DC was synthesized and characterized as described previously (9). TEPA (catalog no. 375683), copper chloride hydrate (CuCl₂, catalog no. C3279), gefitinib (EGFR tyrosine kinase inhibitor, catalog no. SML1657), and MG-132 (proteasome inhibitor, catalog no. M7449) were purchased from Sigma-Aldrich. IFNγ was purchased from Roche (catalog no. 11040596001).

Cells were seeded in tissue culture–treated 6-well plates for 24 hours before addition of freshly diluted drug in a final volume of 1 mL (DC at 20 μg/mL, TEPA at 2 mmol/L, CuCl₂ at 1 mmol/L, and IFNγ or control). Cells were harvested after 24 hours of treatment with DC and TEPA. Treatment with 1 mmol/L of CuCl₂ were performed up to 8 hours, whereas the dose-response experiments with 20 μmol/L, 50 μmol/L, 100 μmol/L, and 200 μmol/L of CuCl₂ were performed at 24 hours of incubation. Cells were incubated with IFNγ up to 4 hours before being harvested and proteins were extracted. To study the effect of Cu-chelators on EGFR phosphorylation, cells were seeded in tissue culture–treated 6-well plates for 24 hours before addition of freshly diluted drug in a final volume of 1 mL (DC at 20 μg/mL, TEPA at 2 mmol/L, and gefitinib at 10 μmol/L) and incubated for 24 hours. Drugs concentrations and incubation times were optimized to keep cells viable.

qPCR analysis was performed using a SsoAdvanced Universal SYBR Green Super-mix (Bio-Rad, catalog no. 172-5270) and the CFX96 Real-time system (Bio-Rad). Quantifications were normalized using endogenous control GUSB. MT1X (F: GCTTCTCCTTGCCTCGAAAT and R: GCAGCAGCTCTTCTTGCAG), GUSB (F: TGGTGCGTAGGGACAAGAAC and R: CCAAGGATTTGGTGTGAGCG), PD-L1 (F: TTGTGGATCCAGTCACCTCTG and R: TTGATGGTCACTGCTTGTCC) primers were used in qPCR.

Cells were washed with cold PBS then lysed using scrapers in cold RIPA buffer (Thermo Fisher Scientific, catalog no. 89901) containing protease inhibitors and phosphatase inhibitors. Quantification of proteins was conducted using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, catalog no. 23225). Protein samples were mixed with 4× Laemmli buffer (Bio-Rad, catalog no. 161-0747, containing β-mercaptoethanol). Protein electrophoresis of whole cell lysates (15–30 μg) or tumor lysates (60 μg), was analyzed by Western blot analysis.

The siRNA oligonucleotides used for silencing CTR-1 were purchased from Qiagen (catalog no. 1027416) and were Hs-CTR1 catalog no. SI042273045 (siRNA#1) and Hs-CTR1 catalog no. SI04936512 (siRNA#2). Control cells were transfected with a scrambled siRNA duplex (Qiagen, catalog no. 1027280), which has no homology with any other human mRNAs. SK-N-FI cells were seeded in tissue culture–treated 6-well plates at 2.5 × 10⁵ and they were transfected with Lipofectamine LTX Transfection Reagent (Qiagen, catalog no. P/N94756) with 20 nmol/L siRNA. After 6 hours, media were diluted with fresh DMEM+10% FBS. Cells were harvested 24 hours posttransfection and analyzed by Western blot analysis.

For coculture experiments with NK cells, SK-N-FI cells were seeded in tissue culture–treated 96-well plates at 5 × 10³ and transfected with LTX Transfection Reagent (Qiagen) with 20 nmol/L siRNA. After 6 hours from transfection, media were diluted with fresh DMEM + 10% FCS. Forty-eight hours posttransfection, media were replaced with a ratio of 1:1 with NK cells (in NK media) and incubated for a maximum of 6 hours. Finally, NK cells were removed and live cancer cells were counted by Trypan blue exclusion.

SH-SY5Y at 7 × 10⁵, SK-N-FI at 6.5 × 10⁵, SK-N-BE(2)-C at 5.5 × 10⁵, and U87MG cells at 5.5 × 10⁵ were seeded in 25 cm² flasks and grown in DMEM with 10% FBS for 24 hours. The following day, the cells were treated with DC at 20 μg/mL or transfected with siRNA for CTR-1 (20 nmol/L siRNAs) for 24 hours or with Cu at 1 mmol/L for 1 and 4 hours. Cells were then lysed with scrapers using milli-Q water, incubated at least 10 minutes in ice. A small amount (5%) of lysate was saved for protein quantification. Samples were supplied in 15-mL falcon tubes to Mark Wainwright Analytical Centre, at UNSW, Australia. Samples were then open digested with 2% nitric acid solution. Samples were analyzed by ICP-MS using NexION 300D with Universal cell technology (Perkin Elmer).

SH-SY5Y cells at 2.5 × 10⁵ were seeded in tissue culture–treated 6-well plates and left growing for 24 hours before addition of freshly diluted drug (DC, TEPA, CuCl₂, and IFNγ) or control. Cells were harvested after 24 hours of drug treatment for DC, TEPA, and IFNγ or after 4 hours of CuCl₂. RNA was also extracted from tumors excised from Th-MYCN mice. A total of 30 mg of tumor tissue was homogenized. RNA was isolated using RNeasy Mini Kit (Qiagen, catalog no. 74104). DNA contamination was removed by performing on-column DNase digestion (Qiagen, catalog no. 79254). Quality control and library preparation were carried out by Ramaciotti Centre for Genomics (UNSW, Australia). Paired-end 100 bp sequencing was performed on the Illumina NovaSeq 6000 using TruSeq stranded mRNA for a total of 1.6B reads. The data have been deposited under Gene Expression Omnibus accession number GSE155031.

Paired-end RNA sequencing (RNA-seq) data were aligned to the human genome assembly (build hg38) using STAR (version 2.5a) with quantMode parameter set to “TranscriptomeSAM.” Raw gene counts and transcripts per million values were calculated using RSEM (version 1.2.31) command rsem-calculate-expression. Differential expression analysis was performed using edgeR in R (v.3.5.3).

Pan-cancer gene expression profiles derived from The Cancer Genome Atlas (TCGA database (https://portal.gdc.cancer.gov/) and normal samples from the Genotype-Tissue Expression (GTEx) project were used to build gene coexpression plots (https://gtexportal.org/home/index.html). Variance stabilizing transformation (VST) was applied to data before plotting, and correlation assessed using Spearman rank correlation method. Differential gene expression (DE) between treatments and control in SH-SY5Y cells was assessed with DESeq2 v1.22.2 using default parameters.

Effects of TEPA or DC in the presence of IFNγ when compared with IFNγ alone, were calculated using limma empirical Bayes moderated t statistics test (limma v3.38.3), and Benjamini–Hochberg correction applied. A corrected P-value ≤ 0.05 was considered significant.

Gene set enrichment analysis (GSEA) was performed using fgsea R package v1.8.0. Gene sets were derived from the Broad Institute's MSigDB collection (msigdbr v6.2.1), except for the STAT1 Targets gene set, which was built from the target list given in PMID 23645984. Immune deconvolution analysis using CIBERSORTx to impute cell fractions from the whole transcriptome expression tumor profile of the mice treated with TEPA, against the previously developed LM22 immune cells signature matrix (10, 11). Details on the master regulator analysis (MRA) are provided in the Supplementary Materials.

SH-SY5Y cells at 2.1 × 10⁶ were seeded in 75 cm² flask and grown in DMEM with 10% FBS for 24 hours. The day after, the cells were treated with DC at 20 μg/mL, TEPA at 2 mmol/L, and gefitinib at 10 μmol/L for 24 hours. Cells were scraped and lysed with ice-cold RIPA buffer containing proteinase and phosphatase inhibitors. Human Phospho-Kinase Array Kit (R&D System, catalog no. ARY003B) was used according to manufacturer's instructions.

SK-N-FI cells were seeded (∼3.5 × 10⁶/150 cm² flask) and grown in DMEM with 10% FBS for 24 hours. The following day, cells were treated with MG132 (3 μmol/L) to inhibit the proteasome, 1 hour prior to treatment with TEPA (2 mmol/L) or gefitinib (10 μmol/L) overnight. Cells were lysed in cold TNN buffer (50 mmol/L Tris-HCl, pH 8.0; 150 mmol/L NaCl; 5 mmol/L EDTA; 0.5% NP-40) supplemented with protease and phosphatase inhibitors (Roche). After removing cell debris, lysates were precleared with Protein A Dynabeads (Thermo Fisher Scientific) previously blocked in BSA. An aliquot of input lysates were reserved before incubation with Rabbit IgG Isotype Control (Invitrogen, catalog no. 10500C), anti-PD-L1 (Cell Signaling Technology, catalog no. 13684) or anti-Ubiquitin (Abcam, catalog no. ab7780) antibodies overnight on a rotating wheel at 4°C. Antibody-bound proteins were precipitated using blocked Protein A Dynabeads for 3 hours on a rotating wheel at 4°C. Washed beads were boiled in 1× Laemmli Sample Buffer (Bio-Rad) with 2.5% β-mercaptoethanol. Protein samples were separated using SDS-PAGE with anti-ubiquitin immunoprecipitation (IP) blots probed for PD-L1 and vice versa using the above antibodies. To prevent detection of heavy and light antibody chains, VeriBlot-HRP secondary antibody (Abcam, catalog no. ab131366) was used. Proteins were visualized using the Bio-Rad Chemi-Doc Touch Imaging system.

Phospho-STAT3 cellular localization was studied by fluorescence confocal microscopy. SH-SY5Y neuroblastoma cells were treated with IFNγ (2.5 ng/mL, 2 hours) or TEPA (2 mmol/L) and INFγ (2.5 ng/mL) combination for 2 hours. Cells were fixed in 4% formaldehyde and permeabilized with Triton X-100 0.4%. Cells were stained with Hoechst 33342 (1 μmol/L, 15 minutes) and phospho-STAT3 antibody (1:200; Cell Signaling Technology, catalog no. 9145) and Anti-rabbit IgG (H+L), F(ab')2 Fragment (1:2,000 Alexa Fluor 555 Conjugate; Cell Signaling Technology, catalog no. 4413). Imaging was performed with a confocal laser scanning microscope ZEISS LSM 880 with a Plan-Apochromat 63×/1.4 Oil DIC M27 objective (Zeiss). Airyscan z-stacks were captured with frame by frame sequential excitation using 405 nm and 561 nm lasers. Raw image data were deconvolved using the Airyscan processing algorithm in the Zen Black software package (SP3V2.1, Zeiss) and exported as orthogonal two-dimensional tagged image files (TIFs). Intensity profiles from representative single cells were calculated for (i) IFNγ and (ii) IFNγ + TEPA combination treatment, demonstrating channel localization and intensities along a defined trajectory. Intensity profiles calculated using ImageJ software.

Treatments of Th-MYCN mice (n = 4–6/group, homozygous for MYCN oncogene) commenced when a palpable tumor (3.5–5 mm in diameter, around 6 weeks of age) was detectable as previously described by Weiss and colleagues (12). Mice were treated with the following Cu-lowering agents: (i) DC at 300 μg/mL, single intravenous dose or (ii) TEPA at 400 mg/kg/day oral-gavage, every 24 hours. DC-treated mice were collected at 24 and 48 hours, while TEPA-treated mice were collected at 24, 48, 72 hours, and 7 days. Serum was isolated and INFγ analyzed using the Mouse Cytokine Array (Abcam, catalog no. ab197465) according to the manufacturer's instructions.

Animals used in the biological studies were approved by the Animal Care and Ethics Committee at UNSW Sydney (Sydney, Australia; ACEC, catalog no. 16/17B and ACEC, catalog no. 18/97B).

Treatments of Th-MYCN mice (n = 4/group) were commenced when a palpable tumor (3.5–5 mm in diameter) was detectable. Mice were treated to assess the short-term effect (72 hours) of the following Cu-lowering agents: (i) DC at 300 μg/mL, single intravenous dose, (ii) TEPA at 400 mg/kg/day, three doses by oral gavage, or (iii) PD-L1 antibody (BioXCell, InVivoPlus anti-mouse PD-L1, at 10 mg/kg/day, single intravenous dose). Seventy-two hours posttreatment, spleen (used for single control staining) and neuroblastoma tumors were excised and digested postmortem using a cocktail of 1 mg/mL collagenase type IV (Sigma-Aldrich) and 0.02 mg/mL DNase (Sigma-Aldrich). Cells were then analyzed by flow cytometry.

Treatments of Th-MYCN+/+ mice (7 mice/group) were started immediately after weaning (3 weeks of age). TEPA (400 mg/kg/day) treatment was administered daily for 5 consecutive days per week for a maximum of 3 weeks by oral gavage. Mice were humanely killed when the tumor size reached approximately 1 cm³ and blood, tumor, and organs were collected.

Animals used in the biological studies were approved by the Ethics Committee at UNSW Australia (ACEC, catalog no. 18/97B) and animals were obtained from the Australian Bio Resources Facility (Moss Vale, NSW, Australia). Female BALB/c-Fox1nu/Ausb, 6–8 weeks old were injected subcutaneously into the right flank with 5 × 10⁵ human SK-N-BE(2)-C cells suspended in 100 μL PBS and growth factor–reduced Matrigel (BD Biosciences) at a 1:1 ratio. At day 7 post cell inoculation, when tumors reached a size of 100–200 mm³, mice were randomized into treatment groups (4 mice/group) and treated daily by oral gavage for a total of 3 weeks with 200 mg/kg/day, 400 mg/kg/day, or 800 mg/kg/day of TEPA. Tumor size was measured using Vernier calipers and mice were euthanized once tumors reached 1,000 mm³.

NK cells, from healthy donor patients, were revived using NK media [X-vivo media (Lonza, catalog no. 04380Q), 10% FBS, 1% penicillin-streptomycin-glutamine, IL2 (Miltenyi Biotech, catalog no. 130-097-746 using 100 U/mL)]. Cancer cells were seeded in tissue culture–treated 96-well plates at 1 × 10⁴ for neuroblastoma cell lines and 1.5 × 10⁴ for GBM cell line. Then, the cells were incubated for 24 hours before addition of freshly diluted drug (DC, TEPA, and IFNγ) or vehicle control. Medium was then replaced with a ratio of 1:1 with NK cells (in NK media) and incubated for a maximum of 6 hours. Finally, NK cells were removed, and alive cancer cells were trypsinized and counted by Trypan blue exclusion method or analyzed by flow cytometry. For flow cytometry analysis, SH-SY5Y cells were seeded in tissue culture–treated 24-well plates at 1 × 10⁵ cells per well. Twenty-four hours post DC or TEPA treatment, NK cells were added with a ratio of 1:1 (without removing drugs) and incubated for a maximum of 6 hours. Finally, stained cells were analyzed using the BD FACSCanto II Flow Cytometer and with FlowJo V10 software.

All in vitro experiments were repeated at least three times, and the means ± SEM were calculated. Associations among gene expressions were determined using two-sided Fisher exact tests. The Mosaic plot was generated using R+ vcd package. Differences between two groups were determined with two-tailed Student t tests (unpaired or paired where specified). Differences between three or more groups were determined using one-way ANOVA followed by Bonferroni multiple comparison tests (for comparison with control means).

CTR-1 is a transmembrane pump responsible for Cu uptake in mammalian cells (13). Several reports have shown CTR-1 is highly expressed in different types of cancer that also feature elevated levels of Cu (5). In particular, we have previously shown that CTR-1 is highly expressed in neuroblastoma tumors (6). To evaluate a potential association between the expression of CTR-1 and PD-L1, we performed IHC staining for both proteins in TMAs, comprising biopsies from 90 patients with neuroblastoma and from 90 patients with brain tumor, including GBM (Fig. 1A and B). A total of 39% of patients with neuroblastoma were negative for both CTR-1 and PD-L1, 42% were positive for CTR-1 and negative for PD-L1, while 19% of patients with neuroblastoma were positive for both CTR-1 and PD-L1 (14). Whereas, 50% of patients with brain tumor were negative for both CTR-1 and PD-L1, 41% were positive for CTR-1 and negative for PD-L1, and 9% were positive for both proteins (Fig. 1A–C). We found a statistically significant association between CTR-1 and PD-L1 expression in both brain tumors (Phi = 0.31; P = 0.006) and neuroblastoma cases (Phi = 0.38; P < 0.0001). In the overall cohort, which includes 180 patients, 56% (100/180) showed CTR-1 positivity, 14% (25/180) were positive for both PD-L1 and CTR-1 and were positively associated (Phi = 0.36; P < 0.001).

Metallothioneins, such as MT1X (metallothionein 1X), are proteins that are upregulated in conditions of metal excess. MT1X is used as a surrogate measure of intracellular Cu, as its expression is proportional to intracellular Cu levels (15). For this reason, MT1X staining was performed on the same TMA as a positive control for the CTR1/Cu signaling axis. In summary, 13% of patients with neuroblastoma were negative for both MT1X and CTR-1, 4% were positive for MT1X and negative for CTR-1, while 67% of patients with neuroblastoma were positive for both MT1X and CTR-1. Furthermore, 28% of patients with brain tumor were negative for both MT1X and CTR-1, 11% were positive for MT1X and negative for CTR-1, and 47% were positive for both proteins (Fig. 1A–C). We found a statistically significant association between MT1X and CTR-1 expression in both brain tumors (Phi = 0.47; P < 0.001) and neuroblastoma cases (Phi = 0.47; P < 0.001). In the overall cohort, 64% (116/180) showed MT1X positivity, 57% (102/180) were positive for both CTR-1 and MT1X and were positively associated (Phi = 0.49; P < 0.001).

A similar association was calculated between MT1X and PD-L1. A total of 29% of patients with neuroblastoma were negative for both MT1X and PD-L1, 52% were positive for MT1X and negative for PD-L1, while 19% of patients with neuroblastoma were positive for both MT1X and PD-L1. Furthermore, 49% of patients with brain tumor were negative for both MT1X and PD-L1, 43% were positive for MT1X and negative for PD-L1, and 8% were positive for both proteins (Fig. 1A–C). We found a statistically significant association between MT1X and PD-L1 expression in both brain tumors (Phi = 0.25; P = 0.018) and neuroblastoma cases (Phi = 0.31; P = 0.006). In the overall cohort, 13% (24/180) were positive for both PD-L1 and MT1X and were positively associated (Phi = 0.29; P < 0.001).

The mosaic plots summarize the distribution of CTR-1, PD-L1, and MT1X protein expression in the combined neuroblastoma and brain cancer cohorts (Fig. 1C). Notably, PD-L1 expression was only detected in patients with positive CTR-1 or MT1X expression.

To evaluate whether this correlation was also present at the transcriptional level and in a wide range of cancers, we examined CTR-1 and PD-L1 mRNA expression in several cancers characterized by high Cu, including GBM, breast carcinoma, lung adenocarcinoma, and stomach adenocarcinomas (16–19). Analysis of TCGA showed a significant positive correlation between CTR-1 and PD-L1 in all the tumor types analyzed (Fig. 2A and B). To further examine the correlation between PD-L1 and intracellular Cu, we correlated the expression of PD-L1 with four Cu-responsive genes (20): the metallothionein MT2A, the metal-responsive transcription factor MTF1, the nuclear factor erythroid 2-like2 NFE2L2, and the cellular prion protein PRNP. TCGA analysis showed a significant positive correlation between each of those four Cu-responsive genes and PD-L1 in GBM and neuroendocrine tumors (Supplementary Fig. S1A). Subsequently, we wanted to determine whether the positive correlation of CTR-1 and PD-L1 is confined to tumor tissues. We therefore examined their mRNA expression levels in their corresponding normal tissues, using the GTEx database. Interestingly, we found no such association in normal adrenal gland (primary site for neuroblastoma and neuroendocrine tumors) lung, colon, stomach, pancreas, or spleen, while normal liver, breast, prostate, and adipose tissues displayed a negative correlation between CTR-1 and PD-L1 expression (Fig. 2A; Supplementary Fig. S1B). As GBM can arise from several parts of the brain (21), we analyzed the expression of CTR-1 and PD-L1 in normal cerebellum, frontal cortex, anterior cingulate cortex, substantia nigra, hypothalamus, basal ganglia, and cerebellar hemisphere, but again found no correlation in any region (Fig. 2B). Collectively, these findings suggest that the correlation between CTR-1 and PD-L1 is confined to tumor tissues.

Given the critical role of CTR-1 in regulating Cu levels, and the correlation between CTR-1 and PD-L1 in tumors, we hypothesized that intratumoral Cu levels could affect PD-L1 expression in tumor cells. We tested this hypothesis in neuroblastoma and GBM, which frequently display elevated Cu levels (6, 7). 1 mmol/L CuCl₂ was added to culture media of SH-SY5Y neuroblastoma and U87MG GBM cells and PD-L1 protein expression was monitored up to 8 hours (Fig. 3A and B). Moreover, we showed a dose-dependent effect of Cu in increasing PD-L1 expression. Supplementing media of the cells with increasing doses of CuCl₂ (concentration range 20–200 μmol/L) induced a proportional upregulation of PD-L1 (Supplementary Fig. S2A). These experiments clearly demonstrated that Cu supplementation induced PD-L1 upregulation in vitro in neuroblastoma and GBM cell lines. Inductively coupled plasma mass spectrometry (ICP-MS) data confirmed that Cu supplementation increased intracellular Cu levels in both neuroblastoma and GBM cell lines (Supplementary Fig. S2B). Interestingly, Cu supplementation also upregulated PD-L1 mRNA levels in both cell lines (Fig. 3C). In this experiment, IFNγ treatment was included as a positive control of PD-L1 stimulation (22). Again, MT1X levels were used as a surrogate measure of intracellular Cu.

As expected, MT1X mRNA was upregulated (>100-fold in SH-SY5Y, >10-fold in U87MG) when Cu was added in the culture media, confirming elevated intracellular Cu uptake at 4 hours (Supplementary Fig. S2C), as registered in ICP-MS data (Supplementary Fig. S2B). Conversely, the use of Cu-chelating agent TEPA, which reduces the availability of Cu in cells, caused a decrease in expression of MT1X (Supplementary Fig. S2D).

IFNγ is a proinflammatory cytokine produced by T cells and NK cells that induces PD-L1 upregulation in cancer cells enhancing immune evasion (23). To determine whether Cu could stimulate signaling pathways activated by IFNγ and involved in PD-L1–driven immune evasion, RNA-seq was performed in SH-SY5Y neuroblastoma cells incubated with CuCl₂. mRNA profiling clearly showed that Cu and IFNγ have similar effects on several pathways involved in PD-L1 expression and immune response to cancer. GSEA revealed that 326 pathways are upregulated and 350 are downregulated by either Cu or IFNγ stimulation (Supplementary Fig. S2E). Among these, we identified four signaling cascades known to be involved in tumor immunity: response to cytokines' regulation of cytokine production, IL6/JAK/STAT3 signaling (24), and TNFα signaling via NFκB, which are key regulators of PD-L1 expression (23). Differential expression (DE) analysis, comparing the changes in Cu-treated or IFNγ-treated cells against control cells, highlighted their strong overlap in regulating PD-L1 expression and other pathways driving immune evasion (Fig. 3D–F; Supplementary Fig. S2E and S2F).

Cu modulates oxidative phosphorylation in tumors and it activates EGFR phosphorylation (25), which in turn catalyzes STAT3 tyrosine phosphorylation (26, 27). This is consistent with our GSEA data showing that Cu activates IL6/Jack/STAT3 signaling as well as IFNγ, leading to increased expression of PD-L1 (Supplementary Fig. S2F). Moreover, our GSEA data suggested that exposure to Cu induces inflammation via NFκB-mediated response pathway (Fig. 3F). NFκB-mediated transcription is known to be a key regulator of the expression of numerous cytokines that influence the immunologic properties of the tumor microenvironment. This includes increased PD-L1 expression in response to INFγ and other cytokines such as IL12 and IL1β (28). Overall, we can speculate that the high levels of Cu in neuroblastoma cells can be important for transducing cytokine signals via NFκB, JAK/STAT, and PI3K/Akt/mTOR/S6K, thus triggering PD-L1 upregulation.

These findings prompted us to investigate whether reducing intracellular Cu could inhibit the induction of PD-L1 expression in response to IFNγ stimulation and thus potentially enhance antitumor immune response. Two approaches were adopted: (i) downregulation of Cu uptake using two independent siRNAs directed against CTR-1 and (ii) pharmacologically with either of two Cu-lowering drugs—DC, which inhibits Cu uptake (6), and TEPA, an analogue of the FDA-approved Cu chelator TETA (29). Each siRNA caused a significant reduction of both CTR-1 protein expression (Fig. 3G) and intracellular Cu in SK-N-FI neuroblastoma cells (Supplementary Fig. S2G). Consistent with our hypothesis a reduction in intracellular Cu led to PD-L1 downregulation in cancer cells (Fig. 3G). This experiment was optimized to obtain a reduction in intracellular Cu (Supplementary Fig. S2G). Cancer cells were next incubated either with 20 μg/mL of DC or 2 mmol/L TEPA for 24 hours (doses reducing Cu but not affecting cell viability; Supplementary Fig. S2H–S2J), either in the presence or absence of IFNγ, to determine whether Cu reduction could inhibit IFNγ-mediated PD-L1 induction. We demonstrated that both DC and TEPA downregulated PD-L1 protein in presence of IFNγ (Fig. 3H–J; Supplementary Fig. S2I). In support of this finding, our RNA-seq data also showed that DC and TEPA significantly inhibited the IFNα/β/γ- transcriptional signature in SH-SY5Y neuroblastoma cells (Fig. 3K; Supplementary Fig. S2K). Interestingly, only the Cu-chelating agent TEPA was able to decrease the immune response signature (including genes such as NFkB1A, KRAS, PIK3CD, CD38, and CD40) in neuroblastoma cells (Supplementary Fig. 2L). This may be because TEPA binds Cu directly and its effect could be faster than DC, which indirectly reduces Cu uptake via depletion of CTR-1 expression (6).

PD-L1 expression is induced by the activation of oncogenic pathways, such as the EGFR signaling pathway. EGFR stimulates three different signaling cascades: RAS/MEK/ERK, PI3K/mTOR/NF-kB, and JAK/STAT, which in turn modulate antitumor immunity, in part by driving PD-L1 upregulation (30). As intracellular Cu is involved in phosphorylation of EGFR and the RAS/MEK/ERK and PI3K/mTOR/NFκB pathways in cancer cells (31), we investigated whether pharmacologically reducing intracellular Cu may affect the expression of kinases that regulate PD-L1 expression. A human phospho-kinase array was performed on SH-SY5Y neuroblastoma cells incubated with DC, TEPA, or the phospho-EGFR inhibitor gefitinib (as positive control; Fig. 4A) and found that Cu-lowering drugs downregulated the phosphorylation of EGFR, AKT, TOR, ERK, and STAT3 (Fig. 4B). Subsequently, we validated these results and confirmed that treatment with DC and TEPA inhibited EGFR phosphorylation, which in turn reduced PD-L1 expression in SH-SY5Y, SK-N-FI neuroblastoma, and U87MG GBM cells (Fig. 4C and D; Supplementary Fig. S3A). As expected, cells exposed to gefitinib (10 μmol/L) also downregulated PD-L1 protein (Fig. 4C and D). Conversely, increased intracellular Cu induced EGFR phosphorylation and PD-L1 upregulation (Supplementary Figs S3B and S3C). It is well known that AKT activation by EGFR is correlated with membrane PD-L1 expression (32). Moreover, recent investigations demonstrated that activation of AKT by EGFR suppresses GSK3β activity through Ser9 phosphorylation and this induced PD-L1 destabilization (33). To further prove that Cu chelation downregulates EGFR signaling pathway we studied the effect of the Cu chelators TEPA and DC on AKT and GSK3 β phosphorylation. Consistent with our hypothesis, data showed that TEPA and DC inhibited pAKTS473 and GSK3betaS9 phosphorylation in two different cell lines (Fig. 4E and F).

A recent study found that inhibition of EGFR activation and signaling destabilizes PD-L1, possibly via a balance between mono-multi versus poly-ubiquitination status (Fig. 4G; ref. 34). To investigate whether Cu deprivation can increase the levels of ubiquitinated PD-L1 and PD-L1 degradation, we performed co-IP assay in SK-N-FI neuroblastoma cells treated with TEPA and gefitinib. We found that inhibition of EGFR phosphorylation due to Cu deprivation by TEPA in SK-N-FI cells caused an accumulation of ubiquitinated PD-L1, potentially inducing degradation via the proteasome (Fig. 4H). To further prove that Cu deprivation destabilized PD-L1, we examined the changes of PD-L1 protein half-life after treatment with DC. In these experiments, we pretreated cells with DC 20 μg/mL for 12 hours and then we added 200 μg/mL of cycloheximide and incubated the cells for 2, 4, 8, and 12 hours prior to preparing cell lysates for protein analysis. Cycloheximide exerts its effects by interfering with the translocation step in protein synthesis thus blocking eukaryotic translational elongation; however, it was well tolerated by cells up to 12 hours. Our results showed that under this experimental condition, the PD-L1 half-life in SK-N-FI cells was 23.28 hours, whereas when treated with DC PD-L1 half-life decreased to 13.76 hours (Supplementary Fig. S3D). Moreover, to confirm that Cu chelators affect the degradation rate of PD-L1, we performed a rescue experiment by using MG132, a potent and cell-permeable proteasome inhibitor. When cells were treated with MG132, both TEPA and gefitinib were not able to reduce the PD-L1 protein level. This is because the degradation of PD-L1, which is induced by TEPA and gefitinib, was abrogated by the proteasome inhibitor MG132 (Supplementary Fig. S3E and S3F).

Our phospho-array data revealed for the first time that Cu-lowering drugs reduced STAT3 phosphorylation (Fig. 4A and B). This result is particularly relevant, as phospho-STAT1 and phospho-STAT3 are involved in the activation of IFNγ-mediated PD-L1 expression (35). To validate this finding, SH-SY5Y neuroblastoma cells were preincubated with DC and TEPA and then stimulated with IFNγ. We found that both agents caused a reduction of STAT3 phosphorylation, signifying reduced cell responsiveness to IFNγ (Fig. 5A). To further demonstrate the impact of Cu chelation in STAT signaling, we used immunofluorescence staining and confocal microscopy and observed decreased nuclear accumulation of phosphorylated STAT3 in cells treated with TEPA compared with the control cells (Fig. 5B). Moreover, to study the effect of Cu chelation on STAT1 target genes, we performed a bioinformatics methodology called MRA (Fig. 5C). This analysis is an algorithm used to identify transcription factors whose targets are enriched for a gene signature (e.g., a list of differentially expressed genes). Both these experiments clearly showed that in presence of Cu chelation, we have a less translocation of STAT3 in the nucleus and a downregulation of the transcription of the gene network target of STAT1. To better understand the effect of Cu deprivation on the STAT signaling pathway, we interrogated a public database of chromatin immunoprecipitation sequencing (ChIP-seq) data for HeLa cells treated with IFNγ to obtain a comprehensive list of STATs target genes (36). When applied to our RNA-seq data, DC and TEPA were found to significantly reduce the expression of several STAT target genes (Fig. 5D and E). Conversely, Cu and IFNγ induced upregulation of the same target gene set (Fig. 5E). DE analysis, comparing the changes with untreated wild-type cells is represented as a heatmap for SH-SY5Y treated with Cu, IFNγ, DC, and TEPA with or without IFNγ (Fig. 5D and E). The heatmap clearly shows that both Cu and IFNγ have a similar stimulatory effect on the STAT target genes, whereas the Cu-lowering compounds DC and TEPA showed inhibitory effect.

Although we cannot exclude other indirect effects, our results show that Cu-lowering drugs reduced PD-L1 expression potentially through two modes of action: first, at the transcriptional level by downregulating the JAK/STAT signaling pathway which in turn suppressed PD-L1 upregulation in response to IFNγ stimulation; and second, at a posttranslational level by inhibiting EGFR signaling and promoting PD-L1 ubiquitination and degradation.

Our in vitro results prompted us to investigate the potential for Cu chelation to enhance the immune response to neuroblastoma. For these in vivo experiments, we used the immune-competent transgenic neuroblastoma mouse model, Th-MYCN (12). This model shares the major oncogenic driver with poor-outcome human neuroblastoma and it recapitulates many of the features of the human disease. When tumors reached approximately 3 mm in diameter, mice were randomized to experimental groups that received either saline (control), TEPA, or DC. TEPA was administered up to 7 days by oral gavage at a nontoxic dose typically used to control Cu accumulation in vivo (37). Alternatively, DC was given intravenously and tumors collected 24 and 48 hours later. DC dosing and end points for the experiment were selected according to our previous PET imaging study that confirmed reduced intratumoral Cu (6).

IHC showed that TEPA and DC caused a significant time-dependent downregulation of PD-L1 expression (Fig. 6A and F). The PD-L1 downregulation was directly associated with decreased EGFR phosphorylation (Fig 6B and G) increased tumor infiltration by CD8⁺ tumor-infiltrating lymphocytes (TIL), CD244⁺ and CD335⁺ NK cells (Fig. 6C–E and H–J) in the same tumor samples. Western blot analysis further confirmed PD-L1 downregulation in lysates of the excised tumors, following treatment with either TEPA or DC (Supplementary Figs. S4A and S5A, respectively). The level of IFNγ was found elevated in the sera of mice treated with TEPA for one week, which is consistent with an increased anticancer immune reaction (Supplementary Fig. S4B). Interestingly, this increased level of IFNγ was not able to induce PD-L1 upregulation in the tumors of mice treated with TEPA. This finding further supports the ability of Cu-chelating drugs to counteract the effect of IFNγ in inducing PD-L1 expression in the tumor.

DC-mediated CTR-1 downregulation in tumors was confirmed by IHC analysis (Supplementary Fig. S5B). TILs were also analyzed for their expression of CD4⁺, CD8⁺ (T cells), and CD16⁺ (NK cells) following treatment with either TEPA or DC (Supplementary Figs. S4C, S4D, S4G and S5C, S5D, and S5E, respectively). Flow cytometry analysis confirmed that TEPA and DC increased CD8⁺/CD4⁺ T- and CD244⁺ NK-cell infiltration (Fig 7A–C). PD-L1 mAb was injected in our neuroblastoma mouse model as a positive control confirming the tumors were responsive to increased tumor immune cell infiltration in response to PD-L1/PD-1 axis inhibition (Fig. 7A–C). Moreover, reducing Cu levels in tumor cells with the Cu-lowering drugs, DC and TEPA, did not affect the expression of PD-1 in the CD4⁺ and CD8⁺ tumor-infiltrating immune cells and of PD-L1 in the CD45⁺ (Fig. 7D–F). To further evaluate the effect of TEPA on composition of the tumor-infiltrating cell population, we performed RNA-seq analysis from tumors derived from mice treated with TEPA for 72 hours and one week. We then performed an immune deconvolution analysis using CIBERSORTx to impute cell fractions from the whole transcriptome expression tumor profile of the mice treated with TEPA, against the previously developed LM22 immune cells signature matrix (10, 11). This experiment showed that TEPA has changed the composition of the tumor microenvironment by increasing the number of macrophages, lymphocytes CD4⁺, CD8⁺, and NK cells (Fig. 7G). This profound (about 3.5 times) increase in the CD8⁺ and CD4⁺ tumor-infiltrating cells was followed by a mild increase of Tregs, as shown by the stain for CD25 and FOXP3 counts (Supplementary Fig. S4E and S4F). TEPA significantly increased the CD8:Treg ratio, indicating that this Cu chelator induces an immune landscape consistent with an antitumor immune response.

The presence of CD8⁺ TILs is known to signify a favorable clinical outcome in a variety of tumors. Whereas NK effector function is known to be suppressed by PD-L1/PD-1 signaling; however, a positive prognostic role for NK cells is far less clear (38). To examine the role of NK in our study, neuroblastoma and GBM cells were pretreated for 24 hours with DC or TEPA then cocultured with primary peripheral blood NK cells. When we evaluated cancer cell viability by trypan blue exclusion and flow cytometry analysis, we found that exposure to either TEPA or DC resulted in enhanced NK-mediated death of various cancer cell lines (Supplementary Fig. S6A–S6D). As shown in Supplementary Fig. 6E, pretreatment with DC or TEPA also reduced PD-L1 upregulation in response to IFNγ normally released by activated NK cells (39). Consistent with this finding, flow cytometry with 7-AAD staining showed increased NK-mediated cell death in SH-SY5Y cells pretreated with Cu-lowering drugs, compared with the same neuroblastoma cells cocultured with NK cells in standard media (Supplementary Fig. S6F). To further confirm that increased susceptibility to NK cells was related to reduced Cu in cancer cells, we cocultured NK cells with SK-N-FI neuroblastoma cells, in which CTR-1 expression had been transiently knocked down. Despite showing normal viability and proliferation under control conditions, susceptibility to NK-mediated lysis was again confirmed (Supplementary Fig. S6G). Overall, our findings strongly indicate that reduced Cu levels may relieve NK exhaustion (40), by downregulating PD-L1 expression in their target cells.

Given the ability of TEPA to reduce PD-L1 expression and to increase TIL infiltration in vivo, we investigated whether Cu chelation would affect tumor growth and progression when used prophylactically in immune-competent neuroblastoma-prone Th-MYCN transgenic mice. Mice (n = 7/group) commenced treatment immediately upon weaning at 21 days and were treated for 3 weeks with saline (control) or TEPA (400 mg/kg/day by oral gavage). As shown in Fig. 7H and 7I, TEPA significantly improved mouse survival compared with controls and reduced tumor growth. We were careful to avoid TEPA toxicity, which can typically arise with doses >550 mg/kg/day (41). Indeed, no toxicity was evident during treatment, nor was weight loss observed (Supplementary Fig. S7A), and no tissue abnormality was detected upon histopathologic examination. To further investigate potential off-target effects of the treatment, we examined the livers of neuroblastoma Th-MYCN mice treated with the Cu chelator TEPA and did not observe any effect of the drug on CTR-1 and PD-L1 expression in this organ (Supplementary Fig. 7B). We chose liver because it is the organ which is the physiologic storage of Cu in the body and can be the most affected by the use of Cu-lowering drugs. Moreover, analysis of publicly available genetic data of normal liver showed no correlation between the expression of CTR-1 and PD-L1 (Supplementary Fig. S1).

To obtain further proof of the important role of the immune cells in the anticancer activity of TEPA, we demonstrated that the same concentration which reduced tumor growth and increased survival in the transgenic neuroblastoma Th-MYCN mice (Fig. 7H), did not affect tumor growth when administered to an immunocompromised neuroblastoma mouse model (Supplementary Fig. S7C).

Together, our results support a key role of Cu homeostasis in cancer immune evasion.

The PD-L1/PD-1 axis is a major therapeutic target in cancer immune modulation therapy, and antibody-mediated blockade therapies have demonstrated good clinical results, mainly in adult melanoma and lung cancer. Unfortunately, this approach has not been as successful as expected for other solid tumors. Moreover, immune checkpoint inhibitor therapy is very costly and is also associated with several side effects, including rashes, diarrhea, colitis, inflammatory arthritis, and many others, frequently leading to treatment interruption (42, 43). Although we have seen rapid clinical translation of PD-L1/PD-1 inhibitors, there is still a dearth of knowledge about the mechanisms regulating PD-L1 expression in cancer cells, which limits our ability to design and develop new approaches to overcome resistance and reduce side effects.

Several mechanisms have been described to lead to aberrant PD-L1 activation, such as genomic alterations, constitutive activation of oncogenic signaling such as EGFR, mTOR, PI3K, AKT, and PTEN deletion (44–47), extrinsic factors such as proinflammatory cytokines IFNγ, TGF1β1, TNFα, and IL17 (48–50). Because PD-L1 expression in normal tissue regulates self-tolerance, it is also important to understand what makes its expression aberrant specifically in tumors; this will in turn allow us to develop therapies targeting PD-L1 in tumors potentially without affecting the normal mechanism of self-tolerance. Here, we identify a strong positive correlation between CTR-1 and PD-L1 present predominantly in malignant tissues, by comparison with normal matched tissues. This is consistent with the higher Cu levels present in neuroblastoma and GBM tumors, compared with normal tissues, and with our discovery that elevated Cu induces PD-L1 gene transcription and protein stabilization. Conversely, Cu-chelating agents decreased PD-L1 expression via inhibition of the response of the cancer cells to proinflammatory cytokines IFNγ, TNFα, and IFNα/β. Moreover, although we cannot exclude other potential indirect effects of Cu chelators, we demonstrated that Cu depletion in cancer cells downregulates STAT3, EGFR, AKT, and GSK3β phosphorylation, which inhibits the transcription of PD-L1 and regulates PD-L1 ubiquitination and stabilization/degradation, respectively.

Our results are consistent with recent work showing that the Cu ionophore disulfiram can induce stabilization of PD-L1 by overloading cancer cells with Cu (51). Although several studies and clinical trials have proved that Cu is important in cancer progression and that Cu chelators are effective in inhibiting tumor growth and angiogenesis, we are the first to show that chelation therapy enhances antitumor immune responses and could be repurposed for immune checkpoint inhibition. We have provided crucial evidence of using the Cu chelator TEPA to increase tumor-infiltrating immune cells and improve the survival in a neuroblastoma immunocompetent mouse model. Although some of the reduction in tumor size may be a consequence of decrease in angiogenesis or cell proliferation, this is an impressive result considering the aggressiveness of this model and that we used TEPA as a single agent. Moreover, results showing that TEPA did not affect tumor growth in an immunocompromised neuroblastoma xenograft model provides further evidence that the antitumor effect of TEPA and improved survival in the immunocompetent mouse model may be caused by the stimulation of an anticancer immune response. The increased number of TILs observed in response to Cu chelation could be a consequence of the PD-L1 blockade, which stimulates new T-cell clones to enter the tumor, as recently shown by Yost and colleagues (52).

Because there are many types of cancer whose progression has been demonstrated to be Cu-dependent, this new immune therapeutic approach has strong potential to be advanced to the clinic. Repurposing clinically available Cu-chelating drugs as immune checkpoint inhibitors is an appealing and attractive strategy, considering that they are not cytotoxic, have minimal effects on immune cells and that they seem to be more active in reducing PD-L1 expression specifically in tumors. The high correlation between Cu levels and PD-L1 expression specifically in transformed tissues, led us to hypothesize that this novel therapeutic approach may reduce the immune-related adverse events which are a major limitation of the current anti-PD-L1 blockade therapy. Moreover, the low cost of chelation therapy makes this novel therapeutic approach suitable for future combination with other immunotherapies, including the highly expensive and laborious CAR T-cell therapy.

In conclusion, this study describes a novel causal link between intratumoral Cu and the regulation of PD-L1, paving the way for clinical trials to evaluate Cu chelators as immune checkpoint inhibitors.

O. Vittorio reports grants from National Health and Medical Research Council (NHMRC), Cure Cancer Australia, Priority-driven Collaborative Cancer Research Scheme, Tour de Cure Grant, and The Ross Trust Foundation during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

F. Voli: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft. E. Valli: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. L. Lerra: Investigation, methodology. K. Kimpton: Investigation, methodology. F. Saletta: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. F.M. Giorgi: Data curation, software, formal analysis, writing-original draft, writing-review and editing. D. Mercatelli: Data curation, software, formal analysis, writing-original draft, writing-review and editing. J.R.C. Rouaen: Data curation, formal analysis, validation, investigation, methodology, writing-review and editing. S. Shen: Resources. J.E. Murray: Methodology. A. Ahmed-Cox: Investigation, methodology. G. Cirillo: Resources, writing-review and editing. C. Mayoh: Data curation, software, formal analysis. P.A. Beavis: Methodology, writing-original draft. M. Haber: Supervision, writing-original draft. J.A. Trapani: Supervision, writing-original draft. M. Kavallaris: Conceptualization, resources, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing. O. Vittorio: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing-original draft, project administration, writing-review and editing.

The authors acknowledge financial support from National Health and Medical Research Council (NHMRC) Career Development Fellowship (APP1164960 to O. Vittorio); Cure Cancer Australia, Priority-driven Collaborative Cancer Research Scheme (APP 1141582 to O. Vittorio); Cancer Institute NSW Career Development Fellowship (RG161875 to O. Vittorio); Tour de Cure Grant; The Ross Trust Foundation; NHMRC Program grant (APP1091261 to M. Kavallaris); NHMRC Principal Research Fellowship (APP1119152 to M. Kavallaris); ARC Centre of Excellence in Convergent Bio-Nano Science and Technology (CE140100036 to M. Kavallaris); NHMRC Program grant (APP1091261 to M. Haber); Cancer Institute New South Wales Program grant (14/TPG/1-13 to M. Haber). The authors also thank the Sydney Children's Tumour Bank Network (A/Prof D. Catchpoole and Ms A. Yuksel from the Children's Hospital at Westmead Tumour Bank) for providing the TMA slides and related clinical information for this study.

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.