Reprogramming neuroblastoma by diet-enhanced polyamine depletion

Abstract

Neuroblastoma is a highly lethal childhood tumor derived from differentiation-arrested neural crest cells^1,2. Like all cancers, its growth is fueled by metabolites obtained from either circulation or local biosynthesis^3,4. Neuroblastomas depend on local polyamine biosynthesis, with the inhibitor difluoromethylornithine showing clinical activity⁵. Here we show that such inhibition can be augmented by dietary restriction of upstream amino acid substrates, leading to disruption of oncogenic protein translation, tumor differentiation, and profound survival gains in the TH-MYCN mouse model. Specifically, an arginine/proline-free diet decreases the polyamine precursor ornithine and augments tumor polyamine depletion by difluoromethylornithine. This polyamine depletion causes ribosome stalling, unexpectedly specifically at adenosine-ending codons. Such codons are selectively enriched in cell cycle genes and low in neuronal differentiation genes. Thus, impaired translation of these codons, induced by the diet-drug combination, favors a pro-differentiation proteome. These results suggest that the genes of specific cellular programs have evolved hallmark codon usage preferences that enable coherent translational rewiring in response to metabolic stresses, and that this process can be targeted to activate differentiation of pediatric cancers.

Keywords: Protein translation; amino acid restriction; arginine; diet; difluoromethylornithine; glutamine; ornithine; ornithine aminotransferase; polyamines; proline.

Figure 1:. MYCN-driven neuroblastoma tumors are characterized by high proline and a functionally disconnected proline and arginine metabolism dependent on uptake from circulation.

a) Primary neuroblastoma tumor tissue undergoing liquid chromatography-mass spectrometry-based metabolomics. b) Differential abundance of 303 metabolites. Proline is the most significantly increased metabolite in MYCN amplified primary human neuroblastoma relative to non-amplified tumors. Dotted line marks significance threshold, with p-values corrected for false discovery rate of 0.05, *q < 0.05. n = 10. c) Relative levels of proline, glutamine, arginine and ornithine in contralateral xenografts from MYCN amplified and non-amplified neuroblastoma cell lines. *P < 0.05, **P < 0.01, two-tailed paired t-test. Mean ± s.e.m., n = 4. d) Proline concentration in MYCN driven neuroblastoma tumors is significantly increased in the TH-MYCN mouse model, whereas glutamine, arginine and ornithine levels across organs are within physiological range. Mean ± s.e.m., tumor proline n = 31, glutamine n = 29, arginine n = 27 and ornithine n = 24; other organs n = 8–31. e) Gene expression of the metabolic network producing the polyamine precursor ornithine highlights low OAT (ornithine amino transferase). Color of labels indicates relative expression in patients (MYCN amplified n = 93 / non-amplified n = 551). f) In vivo stable isotope tracing elucidates the precursors of intratumoral metabolites. Labelling is given normalized to the serum of each infused [U-¹³C]metabolite in TH-MYCN mice. Mean ± s.e.m., n = 4–9. g) Whole-body flux model of sources and interconversions between circulating metabolites. Exchange fluxes are given for circulating proline, ornithine and arginine and their exchange with glutamine. Flux in nmol C/min/g. Mean, n = 4–9. h) Direct circulating nutrient contributions to tumor tissue metabolite pools of proline, arginine and ornithine in TH-MYCN mice. The color indicates the respective circulating nutrient source. Mean ± s.e.m., n = 4–9. i) Schematic showing tumor metabolite sources in neuroblastoma. The non-essential amino acids proline and arginine are primarily taken up from circulation. Tracing identifies the polyamine precursor ornithine to be primarily derived from circulation and not from intratumoral biosynthesis from either arginine or from proline/glutamine through OAT.

Figure 2:. A proline/arginine free diet enhances tumor growth suppression by DFMO in MYCN-driven neuroblastoma.

a) Schematic of 2 factor intervention including a combined proline and arginine amino acid free diet (day 21) and DFMO drug treatment via the drinking water (1 %, day 0) in the TH-MYCN genetically modified mouse model. b) Kaplan-Meier curve of tumor free survival upon treatment where control diet (CD) or a proline and arginine free diet (ProArg-free) are combined with DFMO (difluoromethylornithine). Log-rank test p-value compared to CD. c) Tumor growth defined as tumor weight at death normalized by day of life. Two-tailed t-test compared to CD. Mean ± s.e.m.. For b and c: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. CD n = 13, CD DFMO n = 14, ProArg-free n = 13, ProArg-free DFMO n = 14.

Figure 3:. Dietary intervention causes substrate depletion to enhance polyamine biosynthesis inhibition by DFMO.

a) Schematic of arginine and proline metabolism and its direct link to polyamines via ornithine. b) Differential serum metabolite levels comparing ProArg-free diet to CD. Blue highlights metabolites that are significantly depleted (FDR < 0.05) and rose upregulated compared to CD. n = 7–8. c) Serum arginine, proline, glutamine and ornithine across groups. Statistics comparisons to CD. Mean ± s.e.m., n = 7–10. d) Tumor arginine, proline, glutamine and ornithine levels reveals dysregulation of arginine and proline metabolism under combined treatment. Average age at sacrifice is 8 weeks. Statistics comparisons to CD. Mean ± s.e.m., n = 4–7. e) A ProArg-free diet enhances polyamine depletion in tumor tissue induced by DFMO treatment in prolonged treatment. Average age sacrifice 12 weeks. Comparison treatment to CD. Zooms display differences in polyamine levels between induced by ProArg-free on top of DFMO. Mean ± s.e.m., n = 4–6. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, two-tailed t-test. Abbreviations: CD, control diet; ProArg-free, proline arginine free diet; DFMO, difluoromethylornithine.

Figure 4:. Ribo-Seq reveals defective decoding of adenosine-ending codons upon polyamine depletion.

a) For functional evaluation of translation, tumors were lysed under the presence of a translation inhibitor for preparation of RNA- and Ribo-Seq libraries. Ribosome protected RNA fragments were isolated and sequenced for translation evaluation. b) Normalized ribosome density across transcripts. Combining a proline/arginine free diet with DFMO affects the global ribosome distribution at initiation/early elongation (left + insert) and causes a termination defect (right). c) Normalized ribosome depth at positions encoding for >= 3 prolines in a row. Decoding of these polyproline tracts is affected by combining DFMO with proline and arginine free diet. d) Proline translation defects are codon specific. Relative ribosome density centered around proline codons across treatment groups relative to control diet (zero-line). The left panels denote density of ribosomes at poly proline tracts and right codon occupancy on proline codons outside of poly-proline tracts. Increased occupancy manifests at CCA and less at CCC. e) Adenosine-ending codons show specific translation defects induced by the combined ProArg-free DFMO treatment when the transcriptome-wide relative ribosome occupancy is compared to control diet. f) Schematic showing two mechanisms of polyamine depletion therapy. Only combined treatment induces hallmarks of eIF5A hypusination deficiency and boosts previously unknown codon specific translation defects induced by polyamine depletion. As described in figure 3, data in B-F are from TH-MYCN mouse model. In vivo treatment with DFMO is 1% in the drinking water. For all Mean, n = 5. Abbreviations: CD, control diet; ProArg-free, proline and arginine depleted diet; DFMO difluoromethylornithine.

Figure 5:. Regulation of translation by polyamine depletion is driven by fractional codon content

a) Venn diagram displaying the number of significant changes per treatment group across the layers of protein biosynthesis: gene expression (RNA-Seq), gene translation (RNA-Seq) and protein levels (Proteomics). Each group was compared to CD by differential analysis. Corresponding enrichment analysis across omics layers comparing ProArg-free DFMO to CD using the Reactome gene sets. All pathways are depicted, ranked by significance and signed. The most down- and upregulated sets are ‘cell cycle’ and ‘neuronal system’ the protein level, respectively, and are connected across the omics layers. b) Average fraction (right) of codons ending with adenosine (A-ending) across all genes or selected pathways or their enrichment based on ranking genes by the fraction of A-ending codons across all Reactome pathways (left). Pathways positively enriched (positive p-adj) have higher A-ending fraction whereas negatively enriched have a lower fraction. c) The percentage of A-ending codons correlates with the average protein levels across Reactome pathways. Pathways with an increasing fraction of A-ending codons have lower protein levels in ProArg-free DFMO compared to CD (log2 fold change). d) Fold change across omics layers of top downregulated cell cycle proteins indicates predominant change on the protein level. e) Percentage of codons with the respective nucleotide at the ending position in the Itgb3pg gene (CENPR protein) compared to the transcriptome background. f) Gene set enrichment analysis across omics layers in all three treatment groups using the Hallmark gene set. Only in the ProArg-free DFMO combined treatment group a significant effect is achieved when compared to CD. The major effect is on the translation and protein level. Displayed are the five top enriched sets (complete in Extended Data Fig.10c). Size denotes significance and color normalized enrichment score, with red showing enriched in the intervention group (CD DFMO, ProArg-free or ProArg-free DFMO) and blue in CD. g) Western blot analysis of MYCN in tumors from CD and ProArg-free DFMO arms. NC negative control: CHLA20 neuroblastoma cell line (non-amplified) and PC positive control: IMR5 neuroblastoma cell line (MYCN amplified). GAPDH loading control. i) Representative H&E-stained sections. CD and ProArg-free show undifferentiated primitive neuroblasts, absent neuropil, and prominent mitotic figures. CD DFMO shows poorly differentiated primitive neuroblasts with scant neuropil (arrowhead) and foci of cytodifferentiation (<5% differentiating, arrow). ProArg-free DFMO tumors show high fractions of differentiating neuroblasts (>5% differentiating) with increased cytoplasmic to nuclear ratio (arrow) and abundant neuropil (arrowhead). j) Summary of treatment effects. While cell cycle and MYCN programs are downregulated on the proteome due to translation inhibition, immature cancer cells are driven into neuronal differentiation. For a,c,d and f: RNA-Seq: ProArg-free DFMO n = 5; CD n = 4, Ribo-Seq: n = 5, Proteomics: ProArg-free DFMO n = 6; CD n = 5. Abbreviations: CD, control diet; ProArg-free, proline arginine free diet; DFMO, difluoromethylornithine. Scale bar= 50μm.