Genomic and transcriptomic hallmarks of poorly differentiated and anaplastic thyroid cancers

Abstract

Background: Poorly differentiated thyroid cancer (PDTC) and anaplastic thyroid cancer (ATC) are rare and frequently lethal tumors that so far have not been subjected to comprehensive genetic characterization.

Methods: We performed next-generation sequencing of 341 cancer genes from 117 patient-derived PDTCs and ATCs and analyzed the transcriptome of a representative subset of 37 tumors. Results were analyzed in the context of The Cancer Genome Atlas study (TCGA study) of papillary thyroid cancers (PTC).

Results: Compared to PDTCs, ATCs had a greater mutation burden, including a higher frequency of mutations in TP53, TERT promoter, PI3K/AKT/mTOR pathway effectors, SWI/SNF subunits, and histone methyltransferases. BRAF and RAS were the predominant drivers and dictated distinct tropism for nodal versus distant metastases in PDTC. RAS and BRAF sharply distinguished between PDTCs defined by the Turin (PDTC-Turin) versus MSKCC (PDTC-MSK) criteria, respectively. Mutations of EIF1AX, a component of the translational preinitiation complex, were markedly enriched in PDTCs and ATCs and had a striking pattern of co-occurrence with RAS mutations. While TERT promoter mutations were rare and subclonal in PTCs, they were clonal and highly prevalent in advanced cancers. Application of the TCGA-derived BRAF-RAS score (a measure of MAPK transcriptional output) revealed a preserved relationship with BRAF/RAS mutation in PDTCs, whereas ATCs were BRAF-like irrespective of driver mutation.

Conclusions: These data support a model of tumorigenesis whereby PDTCs and ATCs arise from well-differentiated tumors through the accumulation of key additional genetic abnormalities, many of which have prognostic and possible therapeutic relevance. The widespread genomic disruptions in ATC compared with PDTC underscore their greater virulence and higher mortality.

Funding: This work was supported in part by NIH grants CA50706, CA72597, P50-CA72012, P30-CA008748, and 5T32-CA160001; the Lefkovsky Family Foundation; the Society of Memorial Sloan Kettering; the Byrne fund; and Cycle for Survival.

Figure 1. Cancer genome alterations in 84 PDTCs and 33 ATCs.

(A) Mutation density across the PDTC and ATC cohorts (n = 117), expressed as number of genetic alterations found in 341 genes present in MSK-IMPACT. (B) Clinicopathological features, including sample type, patients’ age (by decade), sex, metastasis site, survival status, tumor purity, cytological phenotype, growth pattern, and PDTC definition (PDTC-Turin vs. PDTC-MSK). Color keys are shown in the right outermost panel. (C–G) Oncoprints of PDTCs (left) and ATCs (right). Middle panel shows percentage of tumors altered for each event; *P < 0.05 between PDTCs vs. ATCs using Fisher’s exact test (see Supplemental Table 3 for extended information). Color key for genetic alterations is shown in the bottom panel. Mutations in drivers and other relevant genes (C); fusion events (D); TERT promoter mutations (E); mutations in TP53 and other tumor suppressor genes (F); and alterations in key pathways and functional groups: PI3K/AKT pathway (includes PIK3CA, PTEN, PIK3C2G, PIK3CG, PIK3C3, PIK3R1, PIK3R2, AKT3, TSC1, TSC2, and MTOR), SWI/SNF chromatin remodeling complex (ARID1A, ARID1B, ARID2, ARID5B, SMARCB1, PBRM1, and ATRX), HMTs (KMT2A, KMT2C, KMT2D, and SETD2), and MMR (includes MSH2, MSH6, and MLH1 genes) (G). See Figure 4 for detailed mutational information.

Figure 2. EIF1AX mutations and EIF1AX-RAS co-occurrence in thyroid cancers.

(A) Distribution of EIF1AX mutations in thyroid cancers and other tumors (modified from ref. 1). (B) Oncoprints showing the co-occurrence of EIF1AX with RAS mutations in PTC from TCGA (top, n = 401), PDTCs from our study (middle, n = 84), and ATCs from our series and from (21) (bottom, n = 55). (C) Kaplan-Meier graph showing significantly shorter survival in EIF1AX-mutated PDTCs (log-rank P = 0.048). See Supplemental Table 6 for detailed clinical correlations.

Figure 3. TERT promoter mutations in thyroid cancers.

(A) Location and overall frequency of TERT promoter mutations in PDTCs and ATCs. (B) Oncoprints of TERT promoter mutations vs. BRAF and RAS in (top) PTCs from TCGA (n = 381); (middle) PDTCs (n = 84); and (bottom) ATCs (n = 33). (C) Allelic frequency of TERT promoter mutations in thyroid cancers. Graph shows TERT mutant allelic frequency (MAF) corrected for tumor purity, determined based on allelic fraction of driver mutations (BRAF or RAS) for all three tumor types. (D) Kaplan-Meier survival in ATCs with log-rank P values. Top: TERT-mutant vs WT. Bottom: WT, TERT-mutant with or without BRAF/RAS mutations. See also Supplemental Table 7.

Figure 4. Pathways and novel functional groups mutated in advanced thyroid tumors.

Expanded oncoprints of genes belonging to the indicated functional categories, as defined in Figure 1G. Samples are divided by tumor type (ATC or PDTC) within each panel. Only altered cases, out of 117 tumors, are shown. Missense, truncating, and in-frame mutations are represented as green, black, and brown squares, respectively. (A) PI3K/AKT/mTOR pathway (includes PIK3CA, PTEN, PIK3C2G, PIK3CG, PIK3C3, PIK3R1, PIK3R2, AKT3, TSC1, TSC2, and MTOR); (B) SWI/SNF chromatin remodeling complex (ARID1A, ARID1B, ARID2, ARID5B, SMARCB1, PBRM1, and ATRX); (C) HMTs (KMT2A, KMT2C, KMT2D, and SETD2); and (D) MMR (MSH2, MSH6, and MLH1). (E) Percentage of tumors altered for each functional category and tumor type.

Figure 5. Recurrent MSK-IMPACT–derived CNAs found in 84 PDTC and 33 ATC.

Representation of arm-level regions recurrently gained or lost in PDTCs and/or ATCs. CNAs were corrected for tumor purity in each sample with known driver mutations (see Methods and Supplemental Figure 2). (A) IGV representation of the altered chromosomal regions, with approximate locations shown on the top panel (genome build hg19), expressed as red (gain) or blue (loss), with shading intensity proportional to the log-ratio (lr) values. Samples are grouped by tumor type and sorted by genetic driver alteration: BRAF, RAS, fusions (RET/PTC, PAX8-PPARG, and ALK), or none/unknown. Color key and annotations are shown on the left. (B) Frequencies of the indicated CNAs in PDTCs and ATCs. Copy number gains (red) or losses (blue) were defined using two lr thresholds: ±0.1 (lighter shading) and ±0.4 (darker shading). Asterisks denote significant differences expressed as Fisher’s exact test P values for ±0.4 threshold: PDTC, 0.06 for 1p loss; ATC, < 2 × 10^–4 for 8p loss, 17p loss, and 20q gain. (C) Kaplan-Meier survival curves for PDTCs harboring chromosome 1q gain (left, log-rank P values for ±0.1 and ±0.4 thresholds are 0.03 and 0.06, respectively) and for ATCs with 13q loss (middle, P = 0.07 and 0.02) or 20q gain (right, P = 0.01 and 0.06).

Figure 6. Principal component analysis (PCA) and BRS of 17 PDTCs and 20 ATCs.

(A) Two-dimensional PCA discriminates PDTCs (squares) from ATCs (circles). Color-coding for driver alterations is shown in B. Asterisks represent ATC and PDTC outliers; see text. (B) Heatmap generated by applying the 67-gene BRS signature to advanced thyroid tumors. Expression values are displayed as Z-scores after scaling the values of each gene across the 37 samples. The 26 most informative genes are shown; the complete 67-gene signature is shown in Supplemental Figure 5. Samples are sorted by ascending BRS score: (BRAF^V600E-like on the left and RAS-like on the right) and annotated for tumor type and driver alteration. (C) Detailed comparison of driver mutation vs. BRS in BRAF- and RAS-mutant PDTCs and ATCs. Paradoxically, RAS-mutant ATCs are primarily BRAF^V600E-like (Mann-Whitney U test, P = 0.003). Box plots were generated using the Tukey method: horizontal lines within each box represent median values, box heights symbolize the IQR (IQR = Q3–Q1); Q3 and Q1 quartiles correspond to the top and bottom boundaries of the box, respectively; whiskers represent values up to 1.5 times IQR greater than Q3 (top: Q3 + 1.5 × IQR) or smaller than Q1 (bottom: Q1 – 1.5 × IQR).

Figure 7. M2 macrophage signature and TDS of 17 PDTCs and 20 ATCs.

(A) Unsupervised clustering based on a 68-gene M2-macrophage signature in advanced thyroid tumors (35). Expression values are displayed as Z-scores after scaling the values of each gene across the 37 samples. The 11 most discriminatory genes (variance greater than 2) are shown; the complete 68-gene signature is shown in Supplemental Figure 6. ATCs clearly cluster apart from PDTCs consistent with their extensive macrophage infiltration. (B) Relative expression of the 16 genes of the TDS in 20 ATCs and 17 PDTCs, compared with 9 PTCs from He et al. (39) evaluated with the same mRNA array platform. ATCs have low TDS values for virtually all TDS genes, whereas PDTCs are comparable to PTCs. The 16-gene TDS signature discriminates ATCs and PDTCs (see unsupervised clustering in Supplemental Figure 7). (C) Correlation plots between TDS and BRS in PTCs from TCGA (top) and PDTCs and ATCs (bottom). Trend lines, Pearson’s correlation coefficients (r) and associated P values are shown in the graphs. TDS and BRS are positively correlated in PTCs (r = 0.74, P < 0.0001) and PDTCs (r = 0.72, P < 0.01); i.e., RAS-like tumors tend to be more differentiated than BRAF-like cancers. This relationship is lost in ATCs, which are profoundly undifferentiated (r = –0.43, P = 0.06). (D) Comparison of TDS values in BRAF- and RAS-mutated PDTCs and ATCs. Whereas ATCs are undifferentiated regardless of their driver alteration (Mann-Whitney U test, P = 0.21), BRAF-mutated PDTCs show a decrease in TDS compared with their RAS-mutant counterparts (P = 0.06). Box plots from B and D were generated using the Tukey method: horizontal lines within each box represent median values; box heights symbolize the IQR (IQR = Q3–Q1); Q3 and Q1 quartiles correspond to the top and bottom boundaries of the box, respectively; and whiskers represent values up to 1.5 times IQR greater than Q3 (top: Q3 + 1.5 × IQR) or smaller than Q1 (bottom: Q1 – 1.5 × IQR). Values outside these limits are considered outliers and are represented by dots.