Comprehensive Molecular Characterization of Pheochromocytoma and Paraganglioma

Abstract

We report a comprehensive molecular characterization of pheochromocytomas and paragangliomas (PCCs/PGLs), a rare tumor type. Multi-platform integration revealed that PCCs/PGLs are driven by diverse alterations affecting multiple genes and pathways. Pathogenic germline mutations occurred in eight PCC/PGL susceptibility genes. We identified CSDE1 as a somatically mutated driver gene, complementing four known drivers (HRAS, RET, EPAS1, and NF1). We also discovered fusion genes in PCCs/PGLs, involving MAML3, BRAF, NGFR, and NF1. Integrated analysis classified PCCs/PGLs into four molecularly defined groups: a kinase signaling subtype, a pseudohypoxia subtype, a Wnt-altered subtype, driven by MAML3 and CSDE1, and a cortical admixture subtype. Correlates of metastatic PCCs/PGLs included the MAML3 fusion gene. This integrated molecular characterization provides a comprehensive foundation for developing PCC/PGL precision medicine.

Keywords: CSDE1; MAML3; TCGA; expression subtypes; genomics; metastasis; molecular profiling; paraganglioma; pheochromocytoma; sequencing.

Figure 1. Germline and Somatic Genome Alterations

Genomic features in rows and primary tumors (n = 173) in columns; shading indicates the effect of a mutation on protein sequence. Significant somatically mutated genes (MutSig2, q < 0.05) indicated by an asterisk (*). See also Table S1 and S2 and S3.

Figure 2. Integrated Alterations in RET and in CSDE1

(A) Location of somatic and germline RET mutations within the protein sequence. (B) RET mRNA expression of mutation positive (+) and mutation negative (−) tumors. Boxplot horizontal lines indicate 25^th, 50^th, and 75^th percentiles, lines extend to the furthest point less than or equal to 1.5 times the interquartile range. Points indicate primary tumors, with horizontal jitter added to aid visualization. (C) Mutations within CSDE1 gene structure. (D) Association of CSDE1 mRNA expression versus CSDE1 DNA copy number, points represent primary tumors. See also Figure S1.

Figure 3. Detection of Fusion Genes

(A) Focal copy number amplifications and deletions from GISTIC analysis. (B) DNA copy number alterations at the TCF4, UBTF and MAML3 loci for tumors with MAML3 amplification; rectangles indicate DNA breakpoints with shading proportional to DNA copy number. mRNA or DNA fusion sequence positivity indicated by “+”. (C) Circos diagram of mRNA fusion genes. Color denotes fusion mates. (D) Exon expression diagrams for representative tumors from each MAML3 fusion gene species. Colors indicate relative differential expression across exons. Orange arrows indicate fusion breakpoints and exon number. See also Table S2 and S4 and Figure S2.

Figure 4. Molecular correlates of MAML3 fusion

(A) Differentially methylated probes among tumors by MAML3 fusion status. (B) Log2 ratios for select mRNA, miRNA and DNA methylation markers (false-discovery rate [FDR] < 0.05). Log2 ratios for select RPPA markers (Kruskal-Wallis tests: β-catenin p < 0.022, GSK3 p < 0.14, DVL3 p < 0.18). GSK3 refers to both GSK3α and GSK3β because the antibody used interacts with both. For display, RPPA expression were increased by the minimum value of each marker to provide positive values for the log2 ratio calculation. Log2 ratios calculated using primary tumors. Arrows indicate regulatory relationships, i.e. methylation within a particular gene region or a miRNA binding partner. (C) Expression scores based on published MAML3 signature (Heynen et al., 2016). (See Supplemental Procedures). Boxplot horizontal lines indicate 25^th, 50^th, and 75^th percentiles, lines extend to furthest point less than or equal to 1.5 times the interquartile range. Points indicate primary tumor values, with horizontal jitter added to aid visualization. See also Figure S3.

Figure 5. Integrated Molecular Subtypes

(A) mRNA subtypes. Primary tumors (n = 173) appear in columns, and clinical and genomic features displayed in rows. Categorical features analyzed using Fisher’s exact tests; continuous features were analyzed using Kruskal-Wallis tests. Select differentially expressed genes displayed below each subtype. (B) DNA copy number (Carter et al., 2012) clustering. Primary tumors are columns (n = 173). (C) DNA methylation clustering. Primary tumors (n = 173) appear in columns. Features tested for association with methylation subtypes by same method as in (A). (D) Ring plot displaying cross-platform subtype association. p prefers to chi-square tests on platform subtype vs mRNA expression subtype. See also Table S2 and Figures S4 and S5 and S6.

Figure 6. Recurrently Altered Pathways

Selected pathways recurrently altered by germline mutations, non-silent somatic variants and somatic fusion genes. Pathway heading percentages reflect alteration rate in the cohort (n = 173). Box shading reflects the alteration rate, with red – activating and blue – inactivating. Protein alteration frequencies and percentages displayed within the respective boxes. Grey boxes have alteration rates ≤ 1%. Succ – succinate; iso – isocitrate; fum – fumarate; 2OG – 2-oxogluterate.

Figure 7. Molecular Discriminants of Clinical Outcome

Primary tumors are columns (n = 173). Molecular and clinical features are rows. Somatic mutation total is the number of somatic mutations in a tumor. Marker and outcome associations were determined by log rank tests (p). See also Figure S7.