Next-generation sequencing in thyroid cancer

Papillary carcinoma

Several studies have applied NGS to variable subgroups of PTCs. Nikiforova et al. analyzed FFPE or frozen tissue from 27 classic PTCs and 30 FVPTCs, using the ThyroSeq NGS panel targeting 12 genes with 34 amplicons on the Ion Torrent PGM sequencer [10]. The results showed that 70% of classic PTCs harbored mutated genes: BRAF (59%) was the most frequent, followed by PIK3A (11%), TP53 (7%), and NRAS (4%). In contrast, 83% of FVPTCs had mutated genes: RAS (73%) was the most frequent, followed by BRAF (7%) and TSHR (3%) [10]. Smallridge et al. performed RNA sequencing (RNA-Seq) using the Illumina HiSeq 2000 platform on frozen tissue from 12 BRAF V600E-mutated PTCs and 8 BRAF-wild type PTCs [11]. Among the 13,085 genes interrogated, 560 were differentially expressed between BRAF V600E-mutated PTCs and BRAF-wild type PTCs [11]. Among these 560 genes, 67 were related to immune function pathways, 51 were under-expressed in BRAF V600E-mutated PTCs, and HLAG, CXCL14, TIMP1, and IL1RAP were over-expressed. In BRAF-wild type PTCs, 4 immune function genes (IL1B, CCL19, CCL21, and CXCR4) were most significantly differentially expressed, and exhibited a high degree of correlation with lymphocytic infiltration [11]. In a study by Leeman-Neill et al. fresh frozen tissue from 62 radiation-associated PTCs and 151 sporadic PTCs was analyzed using RNA-Seq on the Illumina HiSeq 2000 platform. This identified an ETV6-NTRK3 rearrangement in 14.5% of radiation-associated PTCs and 2% of sporadic PTCs [12]. The authors suggested that an ETV6-NTRK3 rearrangement may be a key mechanism of radiation-induced carcinogenesis [12].

The Cancer Genome Atlas (TCGA) Network described the genomic characterization of 496 PTCs, and generated data using whole genome sequencing. This was done on the NGS platform and by a multiplatform analysis of SNP arrays, DNA methylation, and reverse phase protein arrays [13]. In PTCs that lacked known driver mutations, alterations of EIF1AX, PPMID, and CHEK2 were discovered as potential new tumor-initiating mutations. The TCGA project identified the TERT promoter mutation, which accounts for approximately 1% of PTCs, but shows association with a high risk of recurrence. Based on the multi-level molecular data, PTCs were separated into two groups of distinct downstream signaling pathways: the BRAF^V600E-like cohort and the RAS-like cohort. Genomic, epigenomic, and proteomic differences were revealed between these two groups, and most of the RAS-like PTCs were follicular variant PTCs (FVPTCs).

Regarding pediatric thyroid carcinoma, Picarsic et al. analyzed 17 pediatric PTCs (age range 8–17 years, median 13 years) from FNA, fresh frozen tissue, and FFPE samples. Mutation analysis with a 7-gene mutation panel using real-time PCR and ThyroSeq v2 on the Ion Torrent PGM sequencer showed that: (1) The detection rate of molecular alterations was increased by up to 87% by the ThyroSeq v2 NGS assay compared to an increase of 60% by the 7-gene mutation panel. (2) In pediatric thyroid carcinoma, chromosomal rearrangement (53%) was more common than point mutation (33%). (3) ETV6-NTRK3 fusion was identified in 18% of samples, and was associated with aggressive histologic features such as non-encapsulation, solid/insular/trabecular patterns, extensive glandular involvement, and thick tumor fibrosis [14]. Ballester et al. analyzed FFPE and FNA samples from 25 pediatric PTCs (age range 10–19 years, median 14 years) using the 50-gene Ion AmpliSeq Cancer Hotspot Panel v2 on the Ion Torrent PGM sequencer [15]. No additional mutations were detected using the NGS assay on pediatric PTCs that initially were negative for BRAF V600E mutation, RET/PTC1/3 fusion, and TERT promoter mutation [15].

Follicular carcinoma

Follicular carcinoma (FC) is a well-differentiated thyroid carcinoma and the second-most common thyroid cancer after PTC. It accounts for 10% of the total thyroid cancers [16]. Although it is not firmly established, FC is classified into the minimally invasive type and the widely invasive type, according to the microscopic tumor extent [17]. It can be classified into the conventional type and the oncocytic type (Hurthle cell type) on the basis of cell type [18].

A previous study retrospectively collected FFPE or fresh frozen tissue samples from 36 FCs to study 12 cancer genes and 34 amplicons using the ThyroSeq panel on the Ion Torrent PGM sequencer. The analysis detected mutations of NRAS (N = 9), KRAS (N = 2), HRAS (N = 1), TSHR (N = 4), TP53 (N = 4), and PTEN (N = 1) [10]. Interestingly, conventional type (N = 18) and oncocytic type (N = 18) samples showed distinguished genetic alterations. In the conventional type FCs, NRAS was the most frequently affected gene, followed by TSHR and KRAS. TP53 was the most commonly mutated gene in the oncocytic type FCs, followed by HRAS, KRAS, and PTEN [19].

Swierniak et al. analyzed 26 follicular adenomas, 22 FCs, and 34 paired normal thyroid tissue samples. Targeted NGS of 372 genes using the TruSeq kit on the Illumina HiSeq 1500 platform yielded the following results [20]: (1) Somatic alterations were identified in oncogenes (MDM2, FLI1), transcription factors and repressors (MITF, FLI1, ZNF331), epigenetic enzymes (KMT2A, NSD1, NCOA1, NCOA2), and protein kinases (JAK3, CHEK2, ALK). (2) Single nucleotide variants were the most common types of mutations, and large structural variants were the least frequent. (3) A novel translocation in DERL/COX6C was detected. (4) Somatic alteration affected non-coding gene regions and exhibited high penetrance. These results suggest that FC has significant molecular heterogeneity, since FC reveals far more complex somatic alterations than PTC, and each tumor harbors distinct somatic alterations.

Poorly differentiated carcinoma and anaplastic carcinoma

Poorly differentiated carcinoma (PDC) and anaplastic carcinoma (AC) are rare types of thyroid carcinoma, each with a prevalence of 10% [21, 22], and a 1–2% among all thyroid carcinomas [23]. PDC and AC have poor prognoses, with a 5-year survival rate of 51% and 0%, respectively [24]. Since these types of cancers respond poorly to conventional treatment options (including radioiodine therapy, chemotherapy, and radiotherapy), there are already clinical trials for molecular-targeted agents underway [25, 26]. Using NGS, it may be possible to identify targetable gene alterations that can improve the course of patient treatment.

In a study using the ThyroSeq panel on the Ion Torrent PGM sequencer, 12 genes with 34 amplicons were analyzed from FFPE or fresh frozen tissue from 10 PDCs and 27 ACs. The study showed that 30% of PDCs had mutations, whereas 74% of ACs had mutations [10]. The altered genes were NRAS, PIK3CA, GNAS, and BRAF in PDCs, and TP53, BRAF, RAS, PIK3CA, PTEN, and CTNNB1 in ACs.

Sykorova et al. analyzed fresh frozen tissue samples from 3 PDCs and 5 ACs using the TruSight Cancer Panel, targeting 94 cancer-related genes on the Illumina MiSeq sequencer [27]. All PDCs and ACs showed more than one genetic alteration, and TP53 mutations were identified in all but 2 cases [27]. CDH1, FANCD2, CHECK2, ADH1B, GPC3, TP53, and PTEN genes were altered in PDCs, and ATM, HNF1A, MET, NF1, TP53, PTEN, MSH2, RB1, NBN, NF1, MUTYH, TSC2, HRAS, and EGFR genes were altered in ACs [27]. However, the study could not assess mutation of larger genes or chromosomal rearrangements in a panel of 94 known cancer genes, nor could it distinguish germline mutations among the detected genetic changes.

Landa et al. performed NGS using the MSK-IMPACT cancer exome panel, and analyzed 341 genes from FFPE (N = 80) or fresh-frozen tissue (N = 37) samples from 34 PDCs and 33 ACs [28]. The analysis revealed the following: (1) ACs showed a higher mutation number than PDCs (6 ± 5 vs. 2 ± 3, median ± interquartile range), and harbored a higher frequency of mutation in TP35, TERT promoter, PI3 K/AKT/mTOR pathway effector, SWI/SNF subunit, and histone methyltransferase. (2) In PDCs, clinicopathologic features were different based on the genetic alterations: 92% of RAS mutations were found in PDCs met the Turin criteria, whereas 81% of BRAF mutations were found in PDCs met the Memorial Sloan-Kettering Cancer Center (MSKCC) criteria. PDCs harboring BRAF mutations were smaller and had frequent nodal metastasis, whereas RAS-mutant PDCs showed larger tumor sizes and a higher rate of distant metastasis. (3) The association of mutation between EIF1AX and RAS was notable in both PDCs and ACs. EIF1AX mutations have been reported in approximately 1% of PTCs, and known to be mutually exclusive for BRAF and RAS mutations. However, EIF1AX mutations were found in 11% of PDCs and 9% of ACs, and 93% were associated with RAS mutations. (4) Chromosomal rearrangements (including RET/PTC, ALK, and PAX8-PPARG fusions) were found in 14% of PDCs, but were absent in ACs.

Medullary carcinoma

Medullary carcinoma (MC) is a neuroendocrine tumor originating from C-cells, and accounts for approximately 5% of total thyroid cancers [29]. MC is composed of 75% sporadic form and 25% hereditary form, the latter resulting from a RET proto-oncogene mutation [30, 31].

Nikiforova et al. analyzed 12 genes and 34 amplicons using the ThyroSeq panel on the Ion Torrent PGM sequencer with FFPE or fresh frozen tissue samples from 15 sporadic MCs. Mutations were identified in 11 (73%) MCs, of which 7 (47%) were RET mutations, 3 (20%) were HRAS mutations, and 1 (7%) was a KRAS mutation [10]. Simbolo et al. analyzed 50 cancer-related genes using the Ion AmpliSeq Hot Spot Cancer Panel v2 on the Ion Torrent PGM sequencer. Out of 20 retrospectively collected FFPE samples of MCs, the study found that 85% of MCs harbored mutations as follows: 13 RET mutations (60%), 3 HRAS mutations (15%), 1 KRAS mutation (5%), 1 STK11 mutation (5%), and 3 samples where mutations were undetected (15%) [32]. RET status was evaluated with both the NGS and Sanger sequencing methods, and NGS showed higher sensitivity than the Sanger method. NGS identified an additional 3 RET mutations, which were undetected by the Sanger method. Although multiple mutations were found in MCs by the NGS assay, a relevant therapeutic target has not been identified, and further investigation is required to improve the treatment of MCs.