CKLF instigates a “cold” microenvironment to promote MYCN-mediated tumor aggressiveness

Abstract

Solid tumors, especially those with aberrant MYCN activation, often harbor an immunosuppressive microenvironment to fuel malignant growth and trigger treatment resistance. Despite this knowledge, there are no effective strategies to tackle this problem. We found that chemokine-like factor (CKLF) is highly expressed by various solid tumor cells and transcriptionally up-regulated by MYCN. Using the MYCN-driven high-risk neuroblastoma as a model system, we demonstrated that as early as the premalignant stage, tumor cells secrete CKLF to attract CCR4-expressing CD4⁺ cells, inducing immunosuppression and tumor aggression. Genetic depletion of CD4⁺ T regulatory cells abolishes the immunorestrictive and protumorigenic effects of CKLF. Our work supports that disrupting CKLF-mediated cross-talk between tumor and CD4⁺ suppressor cells represents a promising immunotherapeutic approach to battling MYCN-driven tumors.

CKLF attracts CCR4-expressing CD4⁺ cells to foster immune repression and tumor progression in MYCN-driven cancer.

INTRODUCTION

The MYCN oncoprotein drives the initiation, progression, and treatment resistance of a broad spectrum of human cancers (1, 2). Solid tumors with aberrant MYCN activation, particularly high-risk neuroblastoma, are immunologically “cold” and respond poorly to immunotherapy (1, 3–6). These cold tumors are characterized by a low mutational burden and few tumor-infiltrating lymphocytes (TILs), particularly cytotoxic CD8⁺ cells, which is in stark contrast to the immunologically “hot” tumors (7–10). Strategies that turn cold tumors into immunologically active tumors are heavily sought after to improve the success rate of immunotherapy.

Neuroblastoma is the most common extracranial solid tumor among children, and increased MYCN activity contributes to ~50% of high-risk neuroblastoma (1, 11–14). MYCN-driven neuroblastoma metastasizes rapidly with only a 40% 5-year survival rate, representing a clear unmet medical need (13, 15–17). MYCN can induce an immunosuppressive tumor microenvironment (TME) by inhibiting the secretion of CXCL10, which, in turn, impedes the recruitment of CD8⁺ T cells to the TME (4). In addition, MYCN represses the expression of CCL2, a key chemokine responsible for the chemoattraction of natural killer (NK) T cells (18). MYCN-driven tumors also exhibit reduced activity of the interferon (IFN) pathway, a central component of antitumor immune response (4). Despite this knowledge, it remains unknown whether MYCN can induce immunosuppression by up-regulating a chemokine, which can represent an ideal therapeutic target to reverse the protumor milieu to an antitumor TME.

The zebrafish model of MYCN-driven neuroblastoma, in which tumor development intertwines with a gradually maturing immune system, recapitulates the genetic and molecular features of human neuroblastoma (19–23). To uncover previously inaccessible mechanisms of MYCN-mediated immunosuppression, we integrated analyses of human solid tumors with live tracking and functional studies of the TME in this zebrafish cancer model. We demonstrate that MYCN transcriptionally up-regulates chemokine-like factor (CKLF) to promote the cross-talk between tumor and CCR4-expressing CD4⁺ cells, thereby cultivating immunosuppression and tumor aggression.

RESULTS

CKLF is the most expressed chemokine and predicts patient prognosis across diverse solid tumor types

Most solid tumors harbor a T cell–poor environment, posing a major challenge for immunotherapy (7). One class of molecules that modulates the composition of the TME is chemokines, which tumors often secrete to influence the movement and dynamics of immune cells (24). Here, we used the Cancer Cell Line Encyclopedia (CCLE) database to screen for tumor cell–expressing chemokines that may contribute to an immunorestrictive TME (25). Analysis of 1108 cell lines across 84 types of solid tumors uncovered CKLF as the most highly expressed chemokine (Fig. 1, A and B, and fig. S1) (26, 27). Next, we leveraged four online platforms (the R2: Genomics Analysis and Visualization Platform, Survival Genie, cBioPortal, and Kaplan-Meier plotter) to assess the association of CKLF and two other top-expressed chemokines with patient survival (28–32). The median expression of CKLF, CXCL16, and CXCL1 was used to categorize patients into low- and high-expression groups. Kaplan-Meier analysis revealed that unlike CXCL16 or CXCL1, elevated CKLF expression was primarily linked with poor patient survival in immunologically cold tumors for which survival data were available (Fig. 1C and tables S1 to S3). Among these tumors, the association of CKLF expression with poor patient survival was most apparent for neuroblastoma and glioma (Fig. 1, C and D, fig. S2A, and table S1), in which MYCN drives tumor aggressiveness (1, 2). Higher CKLF expression in neuroblastoma was linked to MYCN amplification, advanced disease stage, older patient age, and high-risk disease (fig. S2, B to G). We also observed the association of high CKLF expression with poor survival of MYCN nonamplified patients (Fig. 1E and fig. S2H). Patients with MYCN amplification tend to have the poorest survival (33); CKLF expression was able to somewhat further segregate the prognosis of these patients (fig. S2I). On the other hand, CXCL16 and CXCL1—the second and third most highly expressed—showed the opposite or no association with neuroblastoma patient survival, respectively (Fig. 1C and tables S2 and S3). Collectively, these analyses demonstrate that CKLF is the most expressed chemokine across various solid tumor types and predicts poor patient prognosis in multiple cold tumors.

Fig. 1. CKLF is the most expressed chemokine and predicts patient prognosis across diverse solid tumor types.

(A and B) Normalized relative expression (A) and quantification (B) of all chemokines annotated in the CCLE database, containing 1108 human cancer cell lines across 84 types of solid tumors (25). The ratios of the average expression of CKLF to other chemokines were used to rank order each cancer type. Dots denote the mean expression value of each chemokine in multiple cell lines from the same cancer type. AU, arbitrary units; SCLC, small cell lung cancer; ATRT, atypical yeratoid/rhabdoid tumor; PNET, primitive neuroectodemal tumor; SCCOHT, small cell carcinoma of the ovary, hypercalcemic type; NSCLC, non-small cell lung cancer; MMMT, malignant mix Müllerian tumor. (C) Kaplan-Meier analysis of the association of the top three expressed chemokines, CKLF, CXCL16, and CXCL1, with patient overall survival across multiple solid tumors. Patients were dichotomized into high-expressing versus low-expressing groups based on the median expression value of each chemokine. The overall survival was compared between the two groups of patients using a log-rank test, with the P values depicted as −log₁₀(P values). Red or purple color indicates the association of the high chemokine expression with worse patient survival, while blue color denotes better survival. Data were obtained from the four online platforms: R2: Genomics Analysis and Visualization Platform, Survival Genie, cBioPortal, and Kaplan-Meier plotter (tables S1 to S3) (28–32). HER2, Human Epidermal Growth Factor Receptor 2. (D and E) Kaplan-Meier curves showing the association of CKLF expression with overall (left) and event-free (right) (i.e., no occurrence of cancer progression, relapse, or patient death) survival of all (D) or MYCN nonamplified (Nonamp) (E) patients with neuroblastoma, using the R2 Kaplan-Meier scan method to categorize patients (GSE49710) (81). Data in (B) are presented as means ± SEM and compared for statistical significance by an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

CKLF is transcriptionally up-regulated by MYCN and secreted by tumor cells to attract CCR4-expressing CD4⁺ cells and forkhead box P3 (FOXP3)⁺ regulatory T cells in particular

Given the critical role of both MYCN and MYC in driving tumor aggressiveness (34), we assessed their relationship with CKLF using RNA sequencing data from 498 human primary neuroblastoma samples (GSE49711) (35). We applied the 87-gene MYCN signature score and the 18-gene MYC signature score that were previously published for this analysis (36, 37). Consistent with the importance of both oncogenes in neuroblastoma genesis (38, 39), the CKLF score was positively correlated with both MYCN and MYC signature score (Fig. 2A) (36, 37). Since the contribution of MYCN to immunosuppression is well established in neuroblastoma (4–6, 18, 40), we focused our study on the molecular connection between MYCN and CKLF. To understand whether MYCN transcriptionally up-regulates CKLF, we examined chromatin immunoprecipitation sequencing (ChIP-seq) data from human SHEP neuroblastoma cells in which exogenous MYCN expression is induced (on) or inhibited (off) by tetracycline (tet) (41, 42). We found that MYCN binds to the promoter regions of CKLF and increases histone acetylation signals, indicating transcriptional activation of CKLF by MYCN (Fig. 2B). To validate this prediction, we measured CKLF transcript levels in human SHEP–MYCN–estrogen receptor (ER) neuroblastoma cells, in which 4-hydroxytamoxifen (4OHT) conditionally activates MYCN (43). MYCN activation upon 4OHT treatment significantly up-regulated the expression of CKLF at the same time points as a known MYCN target gene NPM1 (Fig. 2C and fig. S3).

Fig. 2. MYCN transcriptionally up-regulates CKLF in human neuroblastoma cells to attract CD4⁺ cells, especially FOXP3⁺ T_reg cells.

(A) Scatter plots depicting the Pearson correlation between CKLF and MYCN (top) or MYC (bottom) signature scores in neuroblastoma patient samples (red: Amp, MYCN-amplified; yellow: NA, information not available; and black: Nonamp, MYCN nonamplified). (B) Reanalysis of ChIP-seq database for the MYCN and active chromatin acetylation (H3K27ac) signals at the CKLF promoter in human Tet-on or Tet-off MYCN SHEP neuroblastoma cells (GSE80151) (42). (C) Quantitative real-time polymerase chain reaction (qRT-PCR) to detect transcript levels of CKLF in human SHEP-MYCN-ER cells with or without 4OHT treatment for 8 or 12 hours (n = 6). (D) Enzyme-linked immunosorbent assay (ELISA) detection of CKLF in the medium secreted by human SHEP-MYCN-ER cells with or without 4OHT treatment for 48 hours (n = 6). (E) Schematic of the transwell migration assay. The sequences of two CKLF1 C-terminal peptides (C27 and C19, blue) are shown. (F and G) Percentages of human CD4⁺, T_reg, and CD8⁺ T cells migrated to the lower chamber of the plate, containing medium of human SHEP-MYCN-ER cells treated with or without 4OHT for 48 hours (F) (n = 9) or supplemented with CKLF1 synthetic C27 or C19 peptides (G) (n = 8). (H) Representative histological and immunohistochemical staining (brown) and quantification of MYCN (nuclear), CKLF (cytoplasm), and FOXP3 (nuclear) in primary neuroblastoma patient samples (excluding immune niches; n = 16 and 14 for MYCN-low and -high samples and n = 17 and 13 for CKLF-low and -high samples). H&E, hematoxylin and eosin. Inserts: Zoom-in views. Arrowheads: FOXP3⁺ T cells. Scale bars, 200 and 50 μm (inserts). n indicates experimental [(C) to (G)] or biological replicates (H). Data [(C) to (H)] are presented as means ± SEM and compared with unpaired two-tailed t tests. Representative of two [(C), (D), (F), and (G)] experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Next, we measured CKLF protein levels in the culture medium of the SHEP-MYCN-ER cells and detected a significant increase upon MYCN activation (Fig. 2D), indicating that MYCN activation in these tumor cells leads to increased CKLF secretion. Since neuroblastoma formation requires the recruitment of CD4⁺ cells to the TME, we applied a coculturing system to determine how CKLF affects T cell recruitment to MYCN-activated tumors (Fig. 2E). The CKLF-containing medium from 4OHT-treated SHEP-MYCN-ER cells recruited more CD4⁺ and regulatory T cells (T_reg cells) but not CD8⁺ cells (Fig. 2F and fig. S4, A and B). The C-terminal peptides of CKLF1 (i.e., a secreted isoform of CKLF), C27 and C19, bind to the chemokine receptor CCR4 (44), which is expressed on the surface of CD4⁺ cells, including T_reg cells (45). To assess whether the CKLF-CCR4 interaction contributes to the recruitment of CD4⁺ cells, we supplemented the medium with C27 and C19 in the lower chamber of transwell plates and quantified lymphocytes that migrated to the lower chamber (Fig. 2E). Notably, both C27 and C19 peptides significantly enhanced the migration of CD4⁺ and T_reg cells, but not CD8⁺ T cells, with the longer peptide C27 showing better chemotaxis (Fig. 2G and fig. S4, C and D). Moreover, the chemotaxis effect of C27 on CD4⁺ and T_reg cells was dampened by pretreating lymphocytes with the Food and Drug Administration–approved CCR4 antagonist mogamulizumab (fig. S5).

To assess the relationship between MYCN, CKLF, and T_reg cells in primary patient samples, we performed immunohistochemical staining to detect protein expression of MYCN, CKLF, and the T_reg marker FOXP3. As expected, MYCN protein levels were significantly higher in tumors with MYCN amplification, with CKLF following the same trend (fig. S6 and table S4). In addition, CKLF expression was significantly elevated in all primary tumor samples with high MYCN levels but barely detectable in most tumor samples with low MYCN expression (Fig. 2H, fig. S7, and table S4). Occasionally, CKLF expression was detected in patient neuroblastoma cells without concurrent MYCN expression (table S4), suggesting the existence of additional regulators of CKLF other than MYCN. Consistent with our findings in the in vitro coculturing assay, FOXP3⁺ T cells were significantly enriched in tumors with high MYCN and CKLF expression compared to those with low expression, indicating the recruitment of T_reg cells by CKLF (Fig. 2H, fig. S7, and table S4). Together, our findings demonstrate that human neuroblastoma cells up-regulate CKLF through MYCN and secrete this chemokine to attract CD4⁺ and FOXP3⁺ T_reg cells through CCR4.

TME tracking in zebrafish demonstrates similarities to human disease and reveals tumor stage–specific and developmental oscillations

The zebrafish model of MYCN-driven neuroblastoma, referred to as MYCN;EGFP, in which human MYCN and EGFP genes are coexpressed under the zebrafish dopamine β hydroxylase (dβh) promoter, develops tumors in the interrenal gland (the analogous location of the human adrenal medulla) (19). To assess the molecular relationship between MYCN and CKLF in the zebrafish neuroblastoma model, we studied tumors from MYCN;EGFP siblings that developed either localized or metastatic disease (Fig. 3A). An approximately fourfold increase in MYCN protein levels was detected in primary tumors from zebrafish with metastatic tumors (PT-FM) compared to those from sibling fish with localized disease (PT-FL) (Fig. 3B). This result faithfully reproduces what is observed in human disease, in which high MYCN expression is associated with advanced disease and poor prognosis (46). Furthermore, ChIP–polymerase chain reaction (PCR) assays using PT-FM revealed that MYCN binds to the promoter regions of cklf and npm1 significantly more than cxcl12a, a non–MYCN target gene (fig. S8A) (47). Correspondingly, quantitative real-time PCR (qRT-PCR) showed an increase in cklf transcript levels in PT-FM with elevated MYCN expression compared to the PT-FL (Fig. 3C).

Fig. 3. The zebrafish model of MYCN-driven neuroblastoma resembles human disease and harbors a cold TME enriched with Cd4⁺ Foxp3⁺ T_reg cells as tumors progress.

(A) Overlay of brightfield and enhanced green fluorescent protein (EGFP) images of representative control Tg(dβh:EGFP) and Tg(dβh:MYCN;dβh:EGFP) zebrafish, referred to as EGFP and MYCN;EGFP, respectively, at 4 months old. Representative images of 4-, 6-, and 8-month-old MYCN;EGFP fish with metastatic tumor are shown. EGFP signals enable visualization and quantification of tumors, and red arrowheads point to metastatic tumors. Scale bars, 1 mm. (B) Western blotting analysis (left) and quantification (right) of MYCN protein levels in the primary tumor isolated from individual MYCN;EGFP sibling fish that developed either PT-FL or PT-FM at 4 months of age (n = 6). Actin serves as the loading and normalization control. (C) Relative cklf expression, detected by qRT-PCR, in PT-FL versus PT-FM (n = 6). (D to G) Flow cytometric analysis of TILs (D), Cd8⁺ (E, left), NK (F), and Cd4⁺ (G, left) TILs in PT-FL or PT-FM that express low or high MYCN and cklf, respectively (n = 4 or 5). qRT-PCR to detect relative expression of ifng in Cd8⁺ TILs (E, right) and foxp3a, il-10, ifng, il-4/13a, and ccr4 expression in Cd4⁺ TILs (G, right) from the above two groups (n = 6). (H and I) Flow cytometric analysis of total TILs (H), Cd4⁺, Cd8⁺, and NK cell percentages (I) in the TME of primary tumors from MYCN;EGFP fish at the indicated age (n = 6 or 7). n indicates biological [(B) and (D) to (I)] or experimental (C) replicates. [(B) to (G)] Representative of two experiments as means ± SEM and compared with unpaired two-tailed t tests. *P < 0.05, **P < 0.01, and ****P < 0.0001.

Human MYCN-driven neuroblastomas are immunologically cold and characterized by few TILs (4). To understand how the TME is modulated during tumor development and progression, we characterized the PT-FL and PT-FM from sibling zebrafish to minimize age-induced immune changes. We detected a significantly lower percentage of TILs, as well as Cd8⁺ T and NK cells, in PT-FM compared to PT-FL (Fig. 3, D to F, and figs. S8B and S9, A to C). Consistent with data from human tumor samples (4), Cd8⁺ TILs from PT-FM also exhibited decreased ifng expression. Although the frequency of Cd4⁺ T cells is slightly higher in PT-FM, they are immunosuppressive, as demonstrated by increased expression of T_reg marker genes, foxp3a and il-10 (48), as well as decreased expression of ifng (Fig. 3G and figs. S8C and S9B). Next, we analyzed MYCN;EGFP fish of different ages, with older fish developing more metastases (Fig. 3A and fig. S10). The percentages of TILs declined as fish aged and tumors progressed, with a significant decrease in 6- and 8-month-old fish compared to younger fish (Fig. 3H and fig. S9D). Among the TILs, Cd4⁺ cells comprised a larger proportion than Cd8⁺ and NK cells, peaking at 6 months and declining at 8 months (Fig. 3I and fig. S9, E and F). As expected, frequencies of both Cd8⁺ and NK cells in the TME were low: Cd8⁺ cells peaked at 4 months, while NK cells dropped initially and then rose moderately with age (Fig. 3I). Together, our findings in zebrafish reveal disease stage–specific and developmental oscillations of TME and demonstrate the similarities of our zebrafish model to patient neuroblastoma (4), with aggressive tumors expressing high MYCN/CKLF and harboring a cold TME enriched with CD4⁺ T_reg cells.

CKLF collaborates with MYCN to recruit CD4⁺ cells, drive immunosuppression, and promote tumor aggressiveness

To characterize the functional importance of CKLF in MYCN-mediated immunosuppression, we overexpressed cklf (cklf_OE) in zebrafish neural crest cells to mimic high CKLF expression in aggressive patient tumors (49). Specifically, we injected dβh:cklf construct into the one–cell stage embryos from the EGFP or MYCN;EGFP fish crossed with the Tg(cd4-1:mCherry) fish (Fig. 4A and fig. S11A). The differential fluorescence labeling of immune cells and neuroblasts with mCherry and EGFP, along with the optical transparency of zebrafish, enabled us to monitor the dynamic interactions between immune and malignant cells at different ages of fish (50). We found that Cd4⁺ cells did not migrate toward the neural crest regions in fish at 1 week of life (figs. S11B and S12, A and B). However, both macrophages and neutrophils had already infiltrated the MYCN-overexpressing premalignant mass (fig. S12, A and B), when imaging the progeny from Tg(mepg1:mCherry) or Tg(lysc:mCherry) fish crossed to EGFP or MYCN;EGFP fish (51, 52). By 2 weeks, Cd4⁺ cells began to infiltrate the premalignant mass in both MYCN;EGFP;mCherry and MYCN;EGFP;cklf_OE;mCherry fish (fig. S11C). Ectopic expression of cklf did not increase the frequency of Cd4⁺ cells in the premalignant mass of 1- or 2-week-old fish (fig. S11, B and C). By 3 weeks, significantly more Cd4⁺ cells encompassed and penetrated the premalignant mass of MYCN;EGFP;mCherry fish compared to the age-matched EGFP;mCherry fish (Fig. 4B). We detected twice as many Cd4⁺ cells in the TME of MYCN;EGFP;cklf_OE;mCherry fish than was seen in MYCN;EGFP;mCherry fish at 3 weeks of age (Fig. 4B). Hence, tracking the TME by live imaging enabled us to monitor the initial infiltration of immune cells and uncover the ability of Cklf in recruiting Cd4⁺ cells before tumor formation.

Fig. 4. CKLF collaborates with MYCN to recruit CD4⁺ cells, induce an immunosuppressive TME, and promote neuroblastoma aggressiveness.

(A) Schematic of experimental flow. T, thymus; IRG, interrenal glands. (B) Overlay of EGFP and red fluorescent protein images at the interrenal gland of Tg(dβh:EGFP;Cd4-1:mCherry), Tg(dβh:MYCN;dβh:EGFP;Cd4-1:mCherry), and Tg(dβh:MYCN;dβh:EGFP;dβh:cklf;Cd4-1:mCherry) fish, referred to as EGFP;mCherry, MYCN;EGFP;mCherry, and MYCN;EGFP;cklf_OE;mCherry, respectively. Scale bars, 100 and 25 μm (inserts). Quantification of Cd4⁺ cells surrounding or infiltrating EGFP⁺ neural crest or premalignant mass (n = 4 or 16). (C) Quantification and characterization of Cd4⁺ TILs from primary or metastatic tumors of 6-month-old MYCN;EGFP and MYCN;EGFP;cklf_OE fish (n = 6). qRT-PCR to detect foxp3a, il-10, ifng, il-4/13a, and ccr4 expression in Cd4⁺ TILs (n = 6). (D to F) Flow cytometric analysis of TILs (D), NK cells (E), and Cd8⁺ TILs (F) of the primary or metastatic tumors from 6-month-old MYCN;EGFP versus MYCN;EGFP;cklf_OE fish (n = 6). qRT-PCR to detect relative ifng expression of Cd8⁺ TILs purified by fluorescent-activated cell sorting (FACS) (F) (n = 6, right) from MYCN;EGFP and MYCN;EGFP;cklf_OE primary tumors. (G) Overlay of brightfield and EGFP images of 9-week-old MYCN;EGFP (left) and MYCN;EGFP;cklf_OE (right) fish. Scale bars, 2 mm. (H) Kaplan-Meier analysis of rates of neuroblastoma development in MYCN;EGFP (black line; n = 47), MYCN;EGFP;cklf_OE (red line; n = 36), and cklf_OE fish (purple line; n = 22). (I) Quantification of tumor burden over time based on EGFP intensity of MYCN;EGFP (n = 7 to 17) and MYCN;EGFP;cklf_OE fish (n = 15 to 17). (J) Metastases quantification in MYCN;EGFP and MYCN;EGFP;cklf_OE fish (n = 10). n indicates biological replicates [(B) to (J)]. Representative of two [(C) to (F) and (I)] or pooled [(B) and (H)] experiments. (H) was analyzed using log-rank Mantel-Cox test. [(B) to (F) and (I)] Data are presented as means ± SEM and compared with unpaired two-tailed t tests. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Next, we asked whether CKLF could increase the frequency of Cd4⁺ cells in established tumors. Despite similar MYCN levels, overexpression of cklf significantly enriched Cd4⁺ cells in primary tumors from MYCN;EGFP;cklf_OE fish, compared to those from their sibling MYCN;EGFP fish (Fig. 4C and figs. S11D and S13A). Molecular characterization revealed high expression of T_reg marker gene foxp3a and il-10 in Cd4⁺ TILs purified from MYCN;EGFP;cklf_OE primary tumors, indicating the enrichment of T_reg cells in the TME. We also detected a higher level of ccr4 and il-4, along with decreased expression of ifng (Fig. 4C). Consistent with their enrichment of T_reg cells, the percentage of TILs in MYCN;EGFP;cklf_OE primary tumors was lower compared to those in the MYCN;EGFP fish (Fig. 4D and fig. S13B). In particular, the frequencies of cytotoxic NK and Cd8⁺ cells in primary tumors overexpressing cklf were significantly lower than those in MYCN;EGFP fish (Fig. 4, E and F, and figs. S11E and S13, A and C). In addition, Cd8⁺ TILs from cklf-overexpressing primary tumors had lower levels of ifng (Fig. 4F), further indicating a role of CKLF in enhancing immunosuppression. Next, we characterized lymphocyte infiltration in primary versus metastatic tumors from individual MYCN;EGFP and MYCN;EGFP;cklf_OE fish. Metastatic tumors had an increased percentage of Cd4⁺ TILs yet decreased Cd8⁺ TILs and NK cells compared to primary tumors in MYCN;EGFP fish (Fig. 4, C to F, and fig. S13, A and C). Unexpectedly, there were no differences in Cd4⁺ TILs but a slight decrease in Cd8⁺ TILs and NK cells in metastatic tumors compared to primary tumors from MYCN;EGFP;cklf_OE fish (Fig. 4, C to F, and fig. S13, A and C). The lymphocyte infiltration property of primary tumors from the MYCN;EGFP;cklf_OE fish closely resembled that of metastases in the MYCN;EGFP fish (Fig. 4, C to F), indicating that increased cklf expression may induce a TME favoring metastatic spread.

Last, we asked whether CKLF can potentiate MYCN-mediated tumor aggressiveness. Although ectopic expression of cklf alone was insufficient to induce tumors, MYCN;EGFP;cklf_OE fish exhibited faster and more frequent tumor development than MYCN;EGFP fish, despite similar levels of MYCN expression in their neuroblasts (Fig. 4, G to I, and fig. S11D). By 4 weeks of life, ~15% of MYCN;EGFP;cklf_OE fish developed tumors, whereas none of their MYCN;EGFP siblings did (Fig. 4H). By 17 weeks, tumors arose in 89% of MYCN;EGFP;cklf_OE fish, in contrast to only 68% of MYCN;EGFP fish (Fig. 4H). Quantification of the EGFP intensity demonstrated a significantly heavier tumor burden in MYCN;EGFP;cklf_OE fish compared to MYCN;EGFP fish (Fig. 4, G and I), indicating the ability of CKLF to advance MYCN-driven tumor progression. As expected, overexpression of cklf in neural crest cells significantly increases metastatic spread and burden (Fig. 4J), demonstrating the ability of CKLF in promoting tumor aggressiveness at both primary and metastatic sites. Together, live-tracking and functional studies of zebrafish demonstrate that CKLF collaborates with MYCN to recruit and enrich CD4⁺ cells in the TME, an event that occurs as early as the premalignant stage, leading to immunosuppression and tumor aggressiveness.

Depleting T_reg cells abolishes the protumor and immunosuppressive effects of CKLF

Considering our demonstration that CKLF markedly accelerates tumor initiation and progression while capable of attracting CD4⁺ FOXP3⁺ T_reg cells to the TME, we next asked whether T_reg cells contributed to CKLF-induced tumor aggressiveness and immunosuppression. The transcription factor FOXP3 is crucial for the development and immunosuppression of T_reg cells (53), and its expression in neuroblastoma is undetectable (Fig. 2H and figs. S7 and S14A). In zebrafish, allelic loss of foxp3a reduces T_reg-like cells without causing obvious developmental defects (fig. S14B) (54). Therefore, we overexpressed cklf through transgenesis in neural crest of MYCN;EGFP and MYCN;EGFP;foxp3a^+/− fish. Consistent with our earlier findings, cklf collaborated with MYCN to promote tumor initiation and progression in sibling fish without foxp3a loss (Fig. 5, A to C). Notably, disrupting the development of T_reg cells through foxp3a loss was sufficient to abolish the ability of Cklf to promote neuroblastoma aggressiveness, as MYCN;EGFP;foxp3a^+/−;cklf_OE fish could no longer expedite tumor initiation or progression (Fig. 5, A to C). While 83% of MYCN;EGFP;cklf_OE fish developed tumors at 15 weeks of life, only 38% of MYCN;EGFP;cklf_OE;foxp3a^+/− fish had tumors (Fig. 5B). Depleting T_reg cells in MYCN;EGFP fish also delayed tumor onset and progression, despite similar MYCN expression in tumor cells (Fig. 5, A and B, and fig. S14C). By 15 weeks, 60% of MYCN;EGFP fish had developed tumors, compared to 30% of MYCN;EGFP;foxp3a^+/− fish (Fig. 5B). In addition, fluorescence microscopy imaging documented that tumors in MYCN;EGFP fish grew faster than those in their MYCN;EGFP;foxp3a^+/− siblings (Fig. 5C).

Fig. 5. Depleting T_reg cells abrogates the effects of cklf on enhancing neuroblastoma aggressiveness and immunosuppression.

(A) Representative overlay of bright-field and EGFP images of MYCN;EGFP (top left), MYCN;EGFP;cklf_OE (top right), MYCN;EGFP;foxp3a^+/− (bottom left), and MYCN;EGFP;foxp3a^+/−;cklf_OE (bottom right) fish at 7 weeks old. Scale bars, 2 mm. (B) Rates of tumor onset for MYCN;EGFP (black line; n = 24), MYCN;EGFP;cklf_OE sibling fish (red line; n = 25), MYCN;EGFP;foxp3a^+/− (dark blue line; n = 20), and MYCN;EGFP;foxp3a^+/−;cklf_OE (light blue line; n = 21). (C) Quantification of tumor burden based on EGFP intensity of neuroblastoma from the four fish groups at 14 and 15 weeks old (n = 22 to 40). (D and E) Flow cytometric analysis and quantification of total TILs (D), Cd4⁺ TILs, Cd8⁺ TILs, and NK cells (E) from the TME of neuroblastoma dissected from the above fish groups at 6 months of age (n = 5 to 7). (F) Schematic image depicting the proposed working model to describe our findings. n indicates the number of biological replicates [(B) to (E)]. Data are representative of two experiments [(B), (D), and (E)] or pooled from two experiments (C). Survival curves in (B) were compared in a pairwise manner using the log-rank Mantel-Cox test. Data in [(C) to (E)] are presented as means ± SEM and compared to an unpaired two-tailed t test. *P < 0.05, **P < 0.01, and ***P < 0.001.

Next, we asked whether T_reg cells contributed to CKLF-induced immunosuppression, as reflected by the reduced Cd8⁺ and NK cells in the TME. To address this question, we first quantified the frequency of total TILs in primary tumors from all four fish groups and found that depleting T_reg cells through foxp3a loss was sufficient to completely restore TILs in both fish groups regardless of cklf overexpression (Fig. 5D and fig. S15A). In addition, significantly higher frequencies of Cd4⁺ TILs were detected in the TME of MYCN;EGFP;foxp3a^+/−, MYCN;EGFP;cklf_OE, and MYCN;EGFP;cklf_OE;foxp3a^+/− fish than MYCN;EGFP fish (Fig. 5E and fig. S15B). Cd4⁺ TILs from MYCN;EGFP;foxp3a^+/− primary tumors were more immunoreactive than those from MYCN;EGFP tumors, as demonstrated by reduced il-10 and increased ifng expression (fig. S16B). Moreover, there are more Cd8⁺ TILs and NK cells in the TME of MYCN;EGFP;foxp3a^+/− and MYCN;EGFP;cklf_OE;foxp3a^+/− fish than MYCN;EGFP and MYCN;EGFP;cklf_OE fish (Fig. 5E and figs. S15, C and D, and S16A). Cd8⁺ TILs from MYCN;EGFP;foxp3a^+/− tumors showed up-regulated ifng expression, indicating improved cytotoxicity (fig. S16C). In contrast to the TME, the percentages of Cd4⁺ TILs were similar in the kidneys among these two groups of fish, indicating that the enhancement of immune responses is tumor-specific (fig. S16D). Together, our findings establish the critical role of CD4⁺ FOXP3⁺ T_reg cells in mediating CKLF-elicited tumor aggressiveness and immunosuppression.

DISCUSSION

Our research used high-risk neuroblastoma as a model to unravel the mechanisms by which solid tumors cultivate an immunorestrictive TME to favor cancer growth. Our data indicate that MYCN transcriptionally up-regulates CKLF, which is secreted into tumor parenchyma to repress the infiltration and cytotoxicity of CD8⁺ and NK cells through CD4⁺ FOXP3⁺ T_reg cells, thus promoting immunosuppression and disease aggression (Fig. 5F). The induction of such a tumor-favoring milieu by the MYCN-CKLF-T_reg axis may be achieved through (i) the increased chemotaxis of CCR4-expressing T_reg cells by tumor-secreted CKLF, (ii) the recruitment of CCR4-expressing naïve CD4⁺ cells by CKLF to the TME where they differentiate into T_reg cells, (iii) the enhanced immunosuppressive activity of T_reg cells through the CKLF-CCR4 interaction, and (iv) the cross-talk between T_reg cells and myeloid-derived suppressor cells/dendritic cells within the TME (55–58). In contrast to earlier findings that MYCN inhibits chemokine expression to restrict lymphocyte recruitment (4, 18, 40), our work provides the first example that MYCN can activate the chemokine CKLF to attract CD4⁺ immune cells to the TME and induce immunosuppression.

It is well established that MYCN drives tumor initiation and progression through multiple mechanisms, including reprogramming of cellular metabolism (59), suppression of estrogen receptor alpha to block neural differentiation (60), and inducing an immunosuppressive TME (4, 5, 18). Hence, it is expected that cklf overexpression alone is insufficient to induce tumorigenesis, a cumulative outcome of abnormal intracellular and intercellular events (61). Our pan-cancer cell line analyses demonstrated the broad expression of CKLF across 84 solid tumor types, including those not driven by MYCN. In addition, high CKLF expression predicted poor patient prognosis in both MYCN-driven and non-MYCN–driven cancers, such as kidney renal clear cell carcinoma. These data suggest that genes other than MYCN could up-regulate CKLF to induce immunosuppression in non-MYCN–driven cancers, a direction for future study.

In patients with MYCN-amplified neuroblastoma, low immune cell infiltration including that of CD4⁺ cells is a characteristic feature, although the underlying mechanisms are not clearly understood (6). Patients with high-risk disease typically have bone marrow involvement and likely reduced overall immune cell production (62). Studies of animal models allow us to untangle the complex events during the formation and evolution of the TME. Besides its pathological and genetic resemblance to human disease (19–23), the zebrafish model of MYCN-driven neuroblastoma provides unparalleled accessibility to the TME through live imaging. We show that Cklf-high neuroblastoma recruited Cd4⁺ cells to tumor parenchyma as early as the premalignant stage, yet the rate of Cd4⁺ cell infiltration decreased as tumors progressed to a much more advanced stage in older fish. Recent murine studies have provided supporting data indicating the presence of oncogenic CD4⁺ cells in the TME (63). We observed the recruitment of neutrophils and macrophages to the MYCN-overexpressing neural crests before the arrival of Cd4⁺ cells. These findings unveil the sequential order of immune cell recruitment and CKLF-driven interactions between innate and adaptive immune cells during the emergence of the TME.

In addition to its advantage in live imaging, the zebrafish serves as a superior model for the study of T_reg cells. Notably, the FOXP3 gene lies on the X chromosome in mammalian systems (64), making it difficult to study a true monoallelic FOXP3 loss. In female mammals, X chromosomes are inactivated randomly in cells throughout the body, leading to both wild-type and mutant T_reg cells (65). The function of wild-type T_reg cells dominates in female FOXP3 heterozygous mice, resulting in a phenotype indistinguishable from true wild-type animals (65–67). Zebrafish offer a unique advantage studying T_reg cells due to the analogous foxp3a gene being located on a somatic chromosome (68).

The presence of T_reg cells in the TME is a characteristic feature of many human tumors (69, 70); however, their role is context dependent. Our studies have demonstrated the critical role of T_reg cells in MYCN-driven cancer initiation and progression mediated through the CKLF-CCR4 interaction. Despite remarkable efforts to develop T_reg-specific therapies for multiple cancers, directly targeting T_reg cells often triggers systemic inflammation and has not been clinically applicable (71, 72). Our discovery of the crucial role of CKLF, rather than the known T_reg chemokines CCL17 and CCL22 (73), in driving T_reg cell recruitment and immunosuppression will inspire alternative therapeutic developments. Instead of directly targeting T_reg cells, disrupting the CKLF-CCR4 interaction by antibodies or small-molecule inhibitors could uncouple the cross-talk between MYCN-activated tumors and T_reg cells, thus relieving tumor-specific immunosuppression while sparing normal functions of T_reg cells.

MATERIALS AND METHODS

CCLE analysis

Data extraction, data analysis, and figure generation for the CCLE database were performed using Python version 3.9.13. The expression of 43 chemokines was extracted from 1108 cell lines across 84 solid tumor types in the CCLE database. Each specific chemokine expression was averaged across all available cell lines for each cancer type. Cancer types were ranked on the basis of the ratio of CKLF expression compared to the average expression of the other chemokines. The higher the ratio, the more forward the cancer was listed to the left in Fig. 1A and fig. S1.

Patient sample analysis

Publicly available datasets from Gene Expression Omnibus or The Cancer Genome Atlas were reanalyzed using R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl), Survival Genie (https://bbisr.shinyapps.winship.emory.edu/SurvivalGenie/), cBioPortal (www.cbioportal.org), and Kaplan-Meier plotter (https://kmplot.com/analysis/) online platforms (28–32). The datasets are detailed in tables S1 to S3. The median expression value of CKLF, CXCL16, and CXCL1 was used to categorize patients into low- and high-expression groups. Overall and event-free (for neuroblastoma patients only) survival was estimated using Kaplan-Meier scan methods and compared using the log-rank Mantel-Cox test. The R2 Kaplan scan method optimizes the survival cutoff based on statistical testing and generates a Bonferroni correction, separating the dataset into two groups based on gene expression.

Derivation of gene signature scores and correlation analysis

Of the 157 MYCN signature identified by Valentijn et al. (37), 87 genes positively correlated with MYCN expression were ranked on the basis of their expression levels in each tumor. The percentiles were derived for each gene, and the average percentile value was calculated to determine the MYCN signature score (37). In the same way, 18 genes from Jung et al. (36) were used to calculate the MYC signature score. One CKLF gene was used to determine the CKLF score for each tumor.

The correlation analysis for MYCN signature, MYC signature, and CKLF score was performed using the R corrplot package. Pearson’s correlation coefficients were calculated for all pair-wise correlations among all parameters using a correlogram. The degree of association between parameters was created and visualized using the sizes and colors of the circles.

ChIP-seq analysis

The ChIP-seq signal (revolutions per minute per base pair) graph files were obtained through the National Center for Biotechnology Information Gene Expression Omnibus database (GSE80151) (24). The USCS genome browser was used to process the files to produce graphs for each cell line analyzed. An ln(1 + x) transformation of the signal was applied to graphs at the CKLF promoter region for all cell lines upon different treatments as described (42).

Cell culture and enzyme-linked immunosorbent assay

SHEP-MYCN-ER cells were obtained from M. C. Simon’s laboratory at the University of Pennsylvania (74), and cultured at 37°C in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and 5% CO₂. For enzyme-linked immunosorbent assay (ELISA), SHEP-MYCN-ER cells were cultured in the presence of ethanol or 500 nM 4OHT (Sigma-Aldrich, H6278) for 48 hours. The amounts of secreted CKLF (i.e., CKLF1 as it is the only secreted isoform by SHEP-MYCN-ER cells) in the medium were detected using a human CKLF ELISA kit (Novus, NBP2-75289).

Transwell migration assay

For testing the ability of CKLF to attract T_reg cells, medium from human SHEP-MYCN-ER cells treated with or without 4OHT for 48 hours or supplemented with synthetic C27 and C19 peptides of the secreted CKLF1 were placed in lower chambers of the transwell plates with 5-μm pore membrane (Corning). Peripheral blood mononuclear cells (PBMCs) were isolated from human blood (Research Blood Components LLC) through centrifugation with the Ficoll Plague reagent (Cytiva) at 400g for 30 min, with acceleration/deceleration rates at 9/0. PBMCs were briefly incubated in RPMI 1640 medium supplemented with 10% FBS for 2 hours. To examine the effect of CCR4 blockade, PBMCs were cultured in the presence of mogamulizumab (1 μg/ml) or phosphate-buffered saline for 1 hour. Nonadherent mononuclear cells were then added to the upper chamber of the transwell plates at a density of 10⁶ cells/100 μl. The plates were incubated at 37°C supplemented with 5% CO₂. After 4 hours, cells from both upper and lower chambers were collected and stained with BV421-conjugated CD4 (1:100; clone RPA-TA, BioLegend, 300501), allophycocyanin (APC)/CY7-conjugated CD8 (1:100; clone SK1, BioLegend, 344702), and phycoerythrin-conjugated anti-CD25 (1:100; clone M-A251, BioLegend, 356101). These cells were subsequently fixed and permeabilized with a human FOXP3 buffer set (BD Biosciences) and then stained with Alexa Fluor 647–conjugated anti-FOXP3 monoclonal antibodies (1:100; clone 259D/C7, BD Biosciences, 560044) to identify CD4⁺CD25⁺FOXP3⁺ T_reg cells. The percentages of migrated cells were calculated by dividing the number of migrated cells in the lower chamber by the total cells of the same type.

Immunohistochemical staining and scoring

Formalin-fixed, paraffin-embedded tumor tissue was obtained from patients with neuroblastoma that had targeted multigene sequencing at Brigham and Women’s Hospital/Dana-Farber Cancer Institute (75). Patient samples were analyzed under Dana-Farber/Harvard Cancer Center Institutional Review Board protocol (#10-417) and Mass General Brigham Institutional Review Board approval (#2019P000017), which is a secondary use of excess human material protocol for which consent was waived. Four-micrometer-thick tissue sections were used for hematoxylin and eosin staining and immunohistochemical staining, the latter of which was performed using the Leica Bond III automated staining platform and Leica Biosystems Refine Detection Kit. The primary antibodies include anti–N-MYC (1:160 dilution; clone D4B2Y, Cell Signaling Technology, 51705; citrate antigen retrieval), anti-CKLF (1:400 dilution; clone EPR11985[2], Abcam, ab180512; EDTA antigen retrieval), and anti-FOXP3 (1:50 dilution; clone D2WE8, Cell Signaling Technology, 98377). Pancreatic tissue was used as a positive control for anti-CKLF staining and brain tissue as the negative control. The staining was analyzed by board-certified pathologists who were blinded to the samples’ molecular status. Nuclear MYCN and cytoplasmic CKLF staining were assessed under ×200 magnification, using a modified H score (ranging from 0 to 300), in which the percentage of tumor cells with positive staining was multiplied by the staining intensity [graded from 0 (none) to 3 (strong)] as described (76). The score is based on tumor areas only, excluding areas with immune niches. H score of 50 for both MYCN and CKLF was used as the cutoff to categorize patient samples into high and low groups, respectively. The number of FOXP3⁺ T cells was quantified at ×200 magnification, divided by the field-of-view area to obtain cell density, and then averaged across multiple fields of view.

Zebrafish husbandry

Zebrafish husbandry and handling were performed in the aquatic facility at Boston University School of Medicine under approved protocols from the Institutional Animal Care and Use Committee. The Tg(dβh:EGFP) and Tg(dβh:MYCN;dβh:EGFP) fish in the AB wild-type background were obtained from the A. Thomas Look Laboratory at Dana-Farber Cancer Institute (20). The Tg(foxp3a:EGFP) and foxp3a^+/− fish in the AB background were obtained from C. Ceol’s laboratory at the University of Massachusetts Worcester (54). The Tg(Cd4-1:mCherry) fish in the AB background were obtained from A. Hurlstone’s laboratory at the University of Manchester (48).

Subcloning

To generate the dβh:cklf construct, zebrafish cklf cDNA was amplified using phusion high-fidelity enzyme (New England Biolabs) and cloned into the pDONR221 gateway donor vector (Invitrogen, 12536017). The expression construct was generated by combining three entry clones—p5E-dβh, pDONR221-cklf, and p3E-polyA—with a modified destination vector containing I-Sce I recognition sites (20). The primer sequences used for subcloning include cklf: 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTAATGGAAGTTGATTTGGCCTTGTTG-3′ (forward) and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTTTACTCTTCCCCTGTTTGGGCT-3′ (reverse).

Zebrafish genotyping and tumor surveillance

The Tg(dβh:MYCN;dβh:EGFP) fish were crossed with Tg(Cd4-1:mCherry) transgenic or foxp3a^+/− mutant zebrafish to generate one–cell stage embryos, which were coinjected with dβh:cklf and cmlc2:EGFP constructs. The Tg(dβh:MYCN;dβh:EGFP) embryos were also coinjected with the dβh:mCherry construct as the controls. Cointegration of the cmlc2:EGFP construct (77), which expresses EGFP under the cardiac myosin light chain 2 promoter (cmlc2), served as a marker to identify the Tg(dβh:cklf) fish. Progenies were raised and screened at 5 days after fertilization under a fluorescent dissecting microscope (Olympus) for EGFP-expressing cell masses at the interrenal gland to confirm the presence of MYCN transgene. At 4 weeks of life, the sorted fish were fin-clipped to extract genomic DNA for PCR amplifications using gene-specific primers and Taq DNA polymerase (New England Biolabs). The following primers were used to identify fish harboring foxp3a mutations: 5′-CAGTTCTGAAGGCAAAGGG-3′ (forward) and 5′-AATCGGTGCTTATGCGTC-3′ (reverse) as previously described (54). Fish were monitored and imaged weekly for tumor development as described (19, 20). The EGFP mass away from primary tumors were considered as sites of metastasis, which were manually counted under a fluorescent microscope after surgical dissection of fish (Olympus, MVX10). Tumor burden from the primary and metastatic tumors was quantified using ImageJ software [National Institutes of Health (NIH)].

Frozen sectioning and confocal microscopy

Fish were euthanized as described in Institutional Animal Care and Use Committee protocols, subsequently fixed in 4% paraformaldehyde (Thermo Fisher Scientific) at 4°C overnight with gentle agitation, washed with phosphate-buffered saline containing 0.1% Tween 20 (Thermo Fisher Scientific), equilibrated in 30% sucrose at 4°C overnight, and frozen at −80°C. The frozen fish were embedded in optimal cutting temperature compound (Sakura Finetek) and sectioned at a 14-μm thickness using a cryostat (Thermo Fisher Scientific).

Live embryos and fish embedded in 1% low-melting agarose (Life Technologies) or fixed samples were imaged using LSM 710-Live Duo Confocal microscope (Zeiss). Images were processed using ImageJ software (NIH).

Flow cytometry

Both primary and metastatic tumors were dissected under a fluorescent microscope (Olympus), dissociated in a mixture of RPMI 1640 (Corning), Liberase (0.025 μg/ml; Roche), deoxyribonuclease I (0.6 μg/ml; Thermo Fisher Scientific), and 1× penicillin/streptomycin (Corning), washed in RPMI 1640 supplemented with 10% FBS (Sigma-Aldrich), and filtered with a 40-μm filter (Falcon). Fish kidneys were dissected, dissociated, and filtered in RPMI 1640 supplemented with 10% FBS. Single-cell suspensions were stained with the primary antibodies: rat anti-fish CD4 (1:100; clone 6D1, Bio Cosmo, CAC-NIH-NA-01), CD8 (1:100; clone 2C3, Bio Cosmo, CAC-NIH-NA-02) (78–80), or anti-zebrafish Ncam1 antibody (1:100; Institute for Protein Innovation) in phosphate-buffered saline containing heparin (10 U/ml; Sigma-Aldrich), 10% FBS, and 1× penicillin/streptomycin for 30 min at 4°C. Cells were then stained with the secondary antibody, goat anti-rat APC (1:500; Invitrogen, A10540) or goat anti-human (1:200; Jackson ImmunoResearch, 109-136-170) for 30 min at 4°C and counterstained with 4′,6-diamidino-2-phenylindole (1:5000; Thermo Fisher Scientific, 62248). Analysis was performed on an LSRFortessa flow cytometer (BD Biosciences), and fluorescent-activated cell sorting (FACS) was performed on a FACSAria II (BD Biosciences). Purified cells were used for RNA extraction and qRT-PCR analysis. FlowJo software 10.4 (Tree Star) was used to analyze all flow cytometric data.

Quantitative real-time PCR

RNA was extracted from cells using TRIzol (Thermo Fisher Scientific) and/or the RNeasy Micro Kit (QIAGEN) and subjected to genomic DNA clean-up and cDNA synthesis with QuantiTect Reverse Transcription Kit (QIAGEN). SYBR Green PCR master mix (Applied Biosystems) and a Step-One PCR instrument (Applied Biosystems) were used for the qRT-PCR reaction. The qRT-PCR primers are listed in table S5. ACTB or act expression levels were used to normalize the differences in RNA input. All reactions were performed in triplicates.

Western blotting

Tumors were lysed in radioimmunoprecipitation assay buffer [1% NP-40, 0.1% SDS, 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.5% sodium deoxycholate, and 1 mM EDTA] supplemented with 1× or 2× Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Primary antibodies included anti-MYCN (1:1000; Santa Cruz Biotechnology, sc-53993) and anti-actin (1:1000; Santa Cruz Biotechnology, sc-47778). Secondary antibodies included horseradish peroxidase–conjugated anti-mouse or anti-rabbit immunoglobulin G (1:2000 or 1:5000; Thermo Fisher Scientific, 31430 or 31460). Proteins were detected by enhanced chemiluminescence detection kits (Thermo Fisher Scientific), and autoradiographs were obtained with a G:BOX Chemi XT4 (Syngene) and a charge-coupled device camera. Quantification analysis was performed using Syngene GeneTools software (Syngene).

Statistical analysis

Statistical analysis was performed with GraphPad Prism 8.0 using unpaired two-tailed t tests unless otherwise stated. Patient survival and zebrafish tumor onset curves were estimated using Kaplan-Meier methods and compared using the log-rank Mantel-Cox test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Supplementary Materials

This PDF file includes:

Supplementary Materials and Methods

Figs. S1 to S16

Tables S1 to S6

Legend for table S7

References

Other Supplementary Material for this manuscript includes the following:

Table S7