Predominant DICER1 Pathogenic Variants in Pediatric Follicular Thyroid Carcinomas

Introduction

Thyroid cancer is uncommon in pediatric populations. Pediatric thyroid cancer differs from adulthood thyroid cancer in clinicopathological characteristics and prognosis (1). The link between childhood exposure to radiation fallout and the risk of thyroid cancer in adolescents in Chernobyl has been well documented (2), and the genomic landscape of post-Chernobyl radiation-induced thyroid cancers in children has been reported (3). Most genetic studies have been conducted in papillary thyroid cancer (PTC) cases (4 –10), with gene fusions found in a higher percentage of pediatric patients compared with the single-nucleotide variants (SNVs) frequently reported in adult patients (11).

This provides insight into the difference in the spectrum of genomic alterations and pathogenesis between pediatric and adult patients with histologically similar thyroid cancers (12).

Follicular thyroid cancer (FTC) is relatively uncommon during childhood, accounting for 5–10% of pediatric thyroid cancers (1,13). Pediatric FTC cases in turn account for 1.9% of all FTC cases, according to data from the National Cancer Institute Surveillance, Epidemiology and End Results (14) and a Japanese database (15). Although some studies reported that pediatric FTC patients exhibited poorer prognoses than those diagnosed between 20 and 45 years of age (16), prognoses similar to those of adult FTC patients aged ≥21 years have also been reported (15). The most frequently detected genetic alterations in adult Korean FTC cases are H/K/NRAS variants (30–50% of all such alterations), and PAX8-PPARγ rearrangements (∼2%) (17,18). However, little is known about the genetic characteristics of pediatric FTC patients. A previous study reported an NRAS variant and a PAX8-PPARγ fusion in five pediatric FTC cases (19). Another recent study, of 41 patients aged <21 years, reported a low prevalence of N/H/KRAS variants (12.2%), and the absence of PAX8-PPARγ fusions (20). Both studies evaluated a small number of candidate genes using direct (or pyro-) sequencing or a reverse transcription-polymerase chain reaction (RT-PCR) method. More comprehensive genetic assessments are therefore needed to enhance our understanding of the molecular characteristics of pediatric FTC patients.

In this study, we aimed to identify genetic alterations of pediatric FTCs using next-generation sequencing (NGS) methods. We also evaluated associations with the prognoses and clinicopathological characteristics of pediatric FTC patients according to genetic alterations.

Materials and Methods

Results

Genomic alterations and clinicopathological characteristics of pediatric FTCs

Eight (53.3%) of 15 patients had tumors with DICER1 variants (cases P1 to P8; Table 1; Supplementary Table S6); four (P1–3 and P5) had a germ line variant, and two showed loss of heterozygosity (LOH) with or without copy number gain or loss at multiple chromosomes (P5 and P8; Supplementary Fig. S2A). One patient had a pathogenic germ line PTEN frameshift variant (P9). One patient had a tumor with a PAX8/PPARγ rearrangement (P10). Multiple LOH with or without CNAs was found in five FTCs; two patients showed no other coexisting variant (P11 and P12; Supplementary Fig. S2B), while three had other variants (P5 with germ line and somatic DICER1 variants and P8 with somatic DICER1 variants, as described above, and P9 with a germ line PTEN variant; Supplementary Fig. S2B). The remaining three patients (P13–15), for whom we failed to construct an NGS library, were all negative for RAS and DICER1 variants based on Sanger sequencing (Table 1). TERT promoter or RAS variants were not present in any of the 15 tumors investigated.

All eight DICER1 variant-positive FTCs had a somatic missense variant involving metal binding glutamate or aspartate residues in the RNase IIIb domain (six at codon p.E1813 and two at codon p.D1709). Additional missense, nonsense, or frameshift variants in DICER1 were found in seven of the eight FTCs; three were somatic variants and four were pathogenic germ line variants (Fig. 1 and Table 2; Supplementary Table S6). Eventually, four patients (P1–3 and P5) were diagnosed with DICER1 syndrome. One patient with a pathogenic germ line PTEN frameshift variant (P9) was diagnosed with PTEN hamartoma tumor syndrome (Cowden syndrome). LOH was identified in chromosomes 7 and 16 without additional variants in her tumor. She had coexisting manifestations, including recurrent breast fibroadenomas, arteriovenous malformations, and hemangiomas in the brain.

OPEN IN VIEWER

FIG. 1.

Diagram of the DICER1 variants and their positional relationship with the RNase IIIb domain. Case numbers are indicated by squares (germ line variants), circles, or pentagons (somatic variants). Gray-filled areas represent the DICER1 syndrome patients (P1–3 and P5). The germ line truncating variants are widely distributed in the DICER1 gene, and the somatic missense variants of the DICER1 syndrome cases are located at codon p.D1709 or p.E1813. FA, follicular adenoma; FTC, follicular thyroid cancer.

OPEN IN VIEWER

Table 2.

Follicular Thyroid Cancer Patients with DICER1 Syndrome

Case No.	Sex	Age	Chief complaints	Blood (ACMG classification)	Tissue		Family members with germ line DICER1 variant	Family history of thyroid nodule or cancer	Previous or coexisting disease
					Normal thyroid	Cancer
P1	F	2.4	Palpable neck mass	p.S1169* (pathogenic)	p.S1169*	p.S1169*/p.E1813K	Father, brother	None	None yet
P2 a	F	9.3	Palpable neck mass	p.S1169fs (pathogenic)	N/A	p.S1169fs/p.E1813D	Mother	Mother, uncle, aunt, and grandmother on mother's side	PPB at age two years
P3	M	8.9	Palpable neck mass, MNG	p.E1793fs (pathogenic)	p.E1793fs	p.E1793fs/p.D1709N	Mother	Mother, uncle, aunt, and grandmother on mother's side	Bronchogenic cyst or pulmonary parenchymal cyst
P5	M	14.6	Palpable neck mass, MNG	p.S874* (pathogenic)	p.S874*	p.S874*/p.D1709N	Mother	Mother, two aunts, and grandmother on mother's side	PNET in lung at age five years

a

P2 was previously published by Shin et al. (42) and de Kock et al. (35).

ACMG, The American College of Medical Genetics and Genomics; MNG, multinodular goiter; N/A, not available; PNET, primitive neuroectodermal tumor; PPB, pleuropulmonary blastoma.

When clinicopathological characteristics were compared between DICER1-positive (n = 8) and DICER1-negative (n = 7) patients, NH and/or FA coexisted more frequently in DICER1 variant-positive FTCs (p = 0.026 using the Fisher exact test). All FTC patients aged <10 years had DICER1 variant-positive tumors. Otherwise, no differences were found in sex, age, tumor size, pathological findings, or prognoses between the two groups (Table 1).

Molecular and histological characteristics of eight DICER1-positive tumors

The four patients with DICER1 syndrome all presented with a combination of a truncating germ line DICER1 variant and a tumor-specific hotspot missense variant. While the germ line truncating variants were widely distributed in the DICER1 gene (p.S874*, p.S1169*, p.S1169fs, and p.E1793fs), the somatic missense variants were located at codon p.D1709 or p.E1813. The other four patients lacking germ line DICER1 variants had two (P4, P7, and P8) or one (P6) somatic DICER1 variant(s) in tumor tissue; all four patients had a variant in the same codon (p.E1813), and three had additional variants (p.D1822G, p.C1753fs, and p.E1425*) at different positions (Fig. 1 and Supplementary Table S6).

For two patients (P3 and P5) with DICER1 syndrome and multinodular goiters, FFPE tumor samples at different stages (normal, NH, FA, and FTC) were available. In patient 3 (P3), five well-demarcated nodules (two in the right lobe and three in the left lobe) were found in the resected thyroid specimen (Fig. 2A). The FTC in the right lobe, which was surrounded by a thick and irregular fibrous capsule, showed microfollicular proliferation and capsular invasion of tumor cells. Three FAs (one in the right lobe and two in the left lobe) showed a mixed microfollicular and normofollicular pattern, and were encapsulated by a thin fibrous band. The tumor cells of these four lesions (one FTC and three FAs) had round and hyperchromatic nuclei. However, nuclei of the follicular cells in the other hyperplastic nodule in the left lobe were less hyperchromatic compared with the FTC and three FAs, and a fibrous capsule was not evident. In patient 5 (P5), four nodular lesions (three in the left lobe and one in the right lobe) were seen (Fig. 2B). The FTC and FA in the left lobe showed microfollicular proliferation with a focal papillary pattern, although nuclei were round and papillary carcinoma-like nuclear features were not seen. The tumor cells of the FTC invaded the adjacent normal parenchyma. The other FAs (one in the left lobe and one in the right lobe) showed mixed microfollicular and normofollicular patterns.

OPEN IN VIEWER

FIG. 2.

Genetic alteration and histology of normal thyroid, nodular hyperplasia, follicular adenoma, and follicular thyroid carcinoma in two patients with DICER1 syndrome presenting with multinodular goiters (P3 in A and P5 in B).

We evaluated the genetic alterations of the DICER1 gene in normal, NH, and/or FA tissues and compared the results with the index FTC data. Notably, even within the same patient, tumors differing in histology or site had different DICER1 variants; all six coexisting FAs of two patients had additional somatic variants in codon p.E1705, p.D1709, p.D1810, or p.E1813, in addition to the germ line truncating variants, which were different from each other and from the indexed FTC tumor (Figs. 1 and 2). However, no additional variant was found in the tissues of the normal thyroid and NH (Fig. 2). These results suggested that variants at metal-binding sites in the RNase IIIb domain played important roles during the development of FAs or FTCs.

Discussion

We identified a genetic driver alteration in 80% of FTCs from patients younger than 20 years. Eight (53.3%) harbored a pathogenic DICER1 variant, including four with DICER1 syndrome. One (6.7%) had a germ line PTEN variant. A PAX8/PPARγ fusion was identified in one FTC (6.7%). Among five, multiple LOH with or without CNA, two (13.3%) had no coexisting variant, while the remaining three had coexisting DICER1 or PTEN variants mentioned above. Neither an RAS nor a TERT variant was detected in any FTC. Notably, all FTC tumors with a DICER1 variant had a variant at the metal binding sites within the RNase IIIb domain, and 87.5% had additional germ line or somatic variants across the DICER1 gene. Four patients (26.7% of 15 FTC patients; 100% in those diagnosed at age <10 years) were diagnosed with DICER1 syndrome with a pathogenic germ line DICER1 variant.

The DICER1 variant was predominantly detected in pediatric FTC patients in this study, whereas RAS variants and PAX8/PPARγ rearrangements were the most frequent genomic events in adult FTC patients (18,30). In addition, a germ line PTEN frameshift variant was found in one patient. A PAX8/PPARγ rearrangement was detected in one FTC, and no RAS or TERT variants were found; this suggests that the genomic architecture of pediatric FTCs is distinct from that of adult FTCs. A low frequency of RAS variants and/or PAX8/PPARγ fusions in pediatric FTCs was reported in two previous studies (19,20). However, the cited studies did not evaluate DICER1 status. Several pediatric studies on DICER1 variants have been conducted only in familial multinodular goiter patients and/or DTC in association with germ line DICER1 variants (31 –36). Recent NGS studies (17,37 –40) reported somatic DICER1 variants in 6.7% of FAs (n = 25) and 5.1% and 8% of FTCs (n = 30 and 39, respectively) (16,33), but not in PTCs (n = 125, 496, and 355, respectively) (17,39,40). However, these studies included patients of all ages, and only a few of the tumors were diagnosed in patients aged <20 years; thus, the findings seem to be representative of the frequency of adult tumors. Recently, pathogenic DICER1 variants have been identified in 3 (10%) of 30 adolescent-onset PTCs (41), suggesting a higher prevalence of DICER1 variants in the PTCs of young patients compared with adult patients. In the present study, although the number of patients is insufficient to draw any conclusions regarding the prevalence of genetic alterations in pediatric FTCs, we found that DICER1 variants played an important role as a major genetic alteration in pediatric FTCs.

Notably, our results suggest that the DICER1 syndrome should be considered in very young FTC patients, because all three children diagnosed with FTCs before the age of 10 years had DICER1 syndrome. Among them, a 9.3-year-old girl (P2), previously published by Shin et al. (42) and de Kock et al. (35), was diagnosed with PPB before FTC. However, a 2.4-year-old girl and 8.9-year-old boy initially presented only with FTCs but were eventually diagnosed with DICER1 syndrome. Regarding the DICER1 syndrome, recommendations for identifying at-risk individuals and surveillance strategies were recently published, with further updates to be expected (43). Because pathogenic germ line DICER1 variants are involved in >70% of PPB cases (44), genetic testing needs to be prioritized in all PPB patients (43). Based on our results, we suggest somatic DICER1 testing in children diagnosed with FTC, particularly if the child has a family history of multinodular goiter and/or thyroidectomy and/or history or coexisting history of an embryonal tumor. Two (P3 and P5) of our four patients with DICER1 syndrome in association with NH, FA, or FTC presented with multiple nodules at different stages, whereas nontoxic multinodular goiter was more frequently detected in elderly patients and the prevalence increased with age (45). Pediatric cases presenting with multinodular goiter should therefore be assessed for possible tumor development associated with DICER1 cancer predisposition syndrome.

Biological and molecular mechanisms explaining tumorigenesis in DICER1 cancer predisposition syndrome have been elusive. In addition to conferring increased risks of PPB and ovarian sex cord-stromal tumors, individuals with DICER1 syndrome may also develop lung cysts, cystic nephroma, renal sarcoma and Wilms tumor, genitourinary embryonal rhabdomyosarcoma, or embryonal brain tumors (43). Individuals with DICER1 syndrome typically show both truncating germ line DICER1 and second-hit hotspot missense variants. The DICER1 endonuclease is involved in microRNA (miRNA) production (31,46,47), and specific changes in DICER1 amino acids (p.E1705, p.D1709, p.G1809, p.D1810, and p.E1813) in the RNase IIIb domain result in defective cleavage of the hairpin loop structure of pre-miRNAs and a reduction of mature 5p miRNAs (36,48). Notably, all DICER1-positive tumors analyzed in this study (eight FTCs and six FAs) had one or two hotspot missense variants at metal binding glutamate or aspartate residues (49): p.D1709 in two FTCs and p.E1813 in six; p.E1705 in two FTCs and p.D1709 in one; and p.D1810 in one FA and p.E1813 in two (Fig. 1). In studies using murine Dicer knockout fibroblasts, inactivation of the RNase IIIb domain caused by the p.D1709A variant resulted in the elimination of 5p miRNAs, including let-7 (50). Although no specific miRNA or miRNA family has been identified as being involved in tumorigenesis, DICER1 mRNA expression showed an inverse relationship with let-7 and miR-345 levels (31,51), where let-7 is known to be necessary for thyroid follicular cell differentiation (52). All three human RAS genes contain multiple putative let-7 complementary sites in the 3′ untranslated regions, allowing let-7 to regulate RAS expression (53). Loss of the let-7 family can result in constitutive overexpression of the RAS oncogene and tumorigenesis (53,54).

To our knowledge, the c.5465A>G (p.D1822G) variant has not been reported previously; however, the other DICER1 missense variants identified in our study have been reported in thyroid lesions (32,36). We believe the novel variant (c.5465A>G, p.D1822G) is pathogenic because it has not been reported in a population database, including the 1000 Genomes Project and Exome Aggregation Consortium (55,56), and the variant was predicted to be deleterious (Sorting Intolerant From Tolerant [SIFT] score of 0) and a probably damaging (Polymorphism Phenotyping-2 [PolyPhen-2] score of 1 (57,58). These findings are shown in Supplementary Table S6. Similarly, another missense variant at the same amino acid codon (c.5465A>T p.D1822V) was inferred to be pathogenic in a patient with PPB (46) based on its location in the highly conserved residues of the RNase IIIb domain and the SIFT and PolyPhen-2 predictions. Furthermore, the in silico analysis predicted that transcripts with c.5465A>G had a higher chance of exon 25 skipping than those with the wild-type allele (59). Wu et al. reported that c.5429A>G and c.5438A>G not only function as missense variants (p.D1810G and p.E1813G, respectively) but also promote exon 25 exclusion from the DICER1 transcript (60), and that two additional variants (c.5441C>T and c.5428G>T) involved in skipping exon 25 resulted in a DICER1 protein incapable of producing 5p and 3p miRNAs (61). Although we could not perform functional studies of this novel variant (c.5465A>G), these findings suggest that it may be pathogenic, either as a missense variant (p.D1822G) or through skipping of exon 25.

The pathogenesis associated with progression of a normal thyroid to hyperplasia, and then to malignant transformation, has not yet been fully established. In two of our cases presenting with multinodular goiter, different second-hit missense DICER1 variants were present in each nodule diagnosed with FA and FTC, but not in NH in the same patient. Multiple distinct somatic hotspot variants have been reported in different thyroid nodules, including NH and thyroid cancer, in the same individual (32,34); however, we observed RNase IIIb hotspot variants in FA and FTC, but not in NH samples. It may be that germ line DICER1 variants predispose to thyroid hyperplasia, and the subsequent second hit of somatic DICER1 RNase IIIb hotspot variants leads to nodule development (32). DICER1 plays a role in terminating transcription by releasing polymerase II at sites of collision between transcription and replication, thus maintaining genome stability (62). Germ line truncating variants of the DICER1 gene may contribute to an environment in which second-hit variants can occur via adverse effects on genome stability or repair systems, which may in turn induce tumorigenesis. P7 and P8 in this study, who had two somatic DICER1 variants, also had a combination of one truncating and one missense (p.E1813K) variant; the former truncating variant may have occurred earlier than the p.E1813K variant due to the higher allele frequency, which tends to support the hypothesis delineated above (44). In particular, second-hit missense variants in all DICER1-related FTCs or FAs, regardless of whether there is a coexisting germ line DICER1 variant, occurred within sequences encoding the RNase IIIb domain, and especially at metal binding sites; this is consistent with previous reports (32,33,35,36,41). Along with the specific role of the RNase IIIb domain in processing miRNA, which is different from the RNase IIIa domain, the preference for somatic variants at the RNase IIIb domain suggests its importance in tumorigenesis. However, analysis of genetic alterations alone cannot fully explain the difference between FA and FTC development, so further studies are needed to better understand this distinct pathogenesis. In patients with a history of embryonal tumors (P2 and P5), antecedent chemotherapy may be associated with the pathogenesis of DICER1-associated FTC because the mutagenic effects of alkylating agents can accelerate tumorigenesis and reduce the latency of DICER1-associated thyroid cancer development (34).

Two cases (P11 and P12) had multiple LOH with or without CNAs, which were considered their main genetic alterations in the absence of other coexisting driver alterations. For the three FTC samples (P13–15) that failed to construct NGS libraries, we could only perform Sanger sequencing for hotspot regions of DICER, RAS, and TERT genes. If these FTC samples could be evaluated by more comprehensive methods such as NGS that cover multiple genes, it might be possible to identify the genetic driver alterations. The high success rate with respect to finding genetic driver alterations in this study (12 of 15 pediatric FTCs; 80%) also suggests that the genetic characteristics and pathogeneses of pediatric and adult FTCs differ.

Despite a large tumor size (>4 cm) in one-third of our patients, and coexisting angioinvasion in others, all patients showed favorable prognoses with NED during a median follow-up of 8.1 years, which may be attributable to the predominance of miFTC over eaFTC and wiFTC. While some reported that pediatric FTC patients present at more advanced stages, with larger tumors and lower rates of disease-free survival than adult FTC patients (16), no patients died from FTC (16,63,64). We found that the pediatric patients with FTC included in this study have a favorable course, regardless of the DICER1 variant. While a decrease in DICER1 gene expression may be associated with aggressive features in patients with PTC (51), those with DICER1 syndrome-associated DTC, which is predominantly diagnosed at a young age, may constitute a low-risk subgroup with low propensity for metastasis (36).

This study is limited by analyzing only 15 patients seen at a single center. However, the studied tumors could be analyzed with comprehensive genetic sequencing. Moreover, the FTC samples in this study were serially collected at a single center over 35 years, which minimizes selection bias. The predominance of DICER1 variants in pediatric FTCs needs to be validated in further studies with larger sample sizes.

In conclusion, pediatric FTCs have distinct genomic alterations and pathogenic mechanisms from adults, particularly characterized by prevalent DICER1 variants. Early-onset FTC in pediatric patients should raise suspicion for the presence of DICER1-positive tumors and DICER1 syndrome, especially in cases with multiple nodules, a family history of multinodular goiter, and/or a history of embryonal tumor.