A 47-year-old Japanese woman consulted our hospital for the examination of a large goiter. Neck computed tomography (CT) showed multinodular lesions in both lobes (Figure 1A), and there were no abnormal findings on chest CT. She presented with subclinical hyperthyroidism (FT4: 1.53 ng/dl, FT3: 3.47 pg/ml, and TSH: 0.011 μIU/ml) and high levels of thyroglobulin (Tg; 742.5 ng/ml). Anti-thyroid peroxidase (TPO) antibodies and anti-Tg antibodies were negative. Scintiscan imaging showed radioiodine uptake in the non-nodular area of the thyroid isthmus, but absence in the nodular lesions of both lobes (Figure 1B). The radioiodine uptake at 3 h was 8.3% within the normal range.

Specimens obtained from the thyroid nodules by fine-needle aspiration were categorized as benign based on the Bethesda system. Seven months later, the diagnosis of Graves’ disease was made based on the patient’s overt hyperthyroidism (FT4: 3.17 ng/dl, FT3: 12.27 pg/ml, and TSH: <0.003 μIU/ml) and positive TRAb (7.9 IU/l). She entered remission 25 months after starting methimazole therapy, but the Graves’ hyperthyroidism relapsed thereafter. She underwent a total thyroidectomy that included the persistently large goiter, and the weight of the resected thyroid tissue was 209 g.

We performed a next-generation sequencing (NGS) analysis targeted to most of the coding exons of goiter-associated genes (Goiter Ampliseq panel, Table 1), using custom primers designed by Ion Ampliseq Designer (Life Technologies, Carlsbad, CA, USA), according to the manufacturer’s instructions. Briefly, genomic DNA was isolated from whole blood samples with a QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany), and a multiplex polymerase chain reaction (PCR) of a total of 429 amplicons (Table 1) was performed, followed by the addition of Ion Xpress Barcode Adaptors, with the Ion Ampliseq Library Kit 2.0 (Life Technologies).

After enrichment by clonal emulsion PCR on Ion Sphere particles (Ion PGM Template OT2 200 Kit), the barcoded libraries were loaded on an Ion 318 chip, and massively paralleled sequencing was carried out on an Ion Torrent PGM sequencer with the Ion PGM Sequencing 200 Kit version 2.

We analyzed the raw signal data of NGS by using Torrent Suite software (ver. 5.0.4), including adaptor trimming, read alignment to human genome 19 reference, coverage analysis, and variant calling. To determine whether the detected sequence variants are known pathogenic mutations or novel variants, we carried out variant filtration as well as annotation by using Ion Reporter software (ver. 5.0; Life Technologies). Sequence variant confirmation was performed by conventional Sanger methods with a BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA). This study was approved by the Ethics Committee of Kuma Hospital, and informed consent was obtained from the patient and her family members for the use of their blood and tissue samples for research purposes.

After surgical resection, the thyroid tissues were routinely fixed in 10% neutral buffered formalin, and specimens were embedded in paraffin. Serial sections (3-μm thick) were cut from each paraffin block. For the light-microscopic examination, the sections were stained with hematoxylin–eosin (HE). The immunostaining for human NRF2 (rabbit polyclonal, C-20, 1:50, Santa Cruz Biotechnology, Santa Cruz, CA, USA) was performed using the Leica Bondmax system (Leica Microsystems, Wetzlar, Germany) and a Bond refine kit (Leica Microsystems) according to the manufacturer’s instructions. Antigen retrieval was performed using Bond Epitope Retrieval Solution 1 (pH 6) at 100°C for 60 min. Control specimens were obtained from four unrelated patients with large multinodular goiters (>100 g) and negative anti-TPO and anti-Tg antibodies.

The subsequent histopathological findings of the thyroid specimens showed that both lobes of the thyroid was almost completely occupied by multinodular lesions that were partially encapsulated by fibrous tissue and concomitantly degenerative cystic structures without lymphoid infiltration in the stroma. The nodules were composed of various-sized thyroid follicles and showed largely high cellularity due to a microfollicular pattern with columnar epithelial cells (Figure 2A). The internodular thyroid parenchyma showed almost normal morphology, and malignant findings were not identified.

The patient’s father had presented with a marked goiter but had died earlier due to head injury. She had a family history of large goiter in two paternal aunts who were unavailable to undergo genetic testing (Figure 3). The patient’s mother, son, and daughter were euthyroid and had no goiter or nodular lesion on ultrasonography of the neck.

To identify any known germline mutation of goiter-associated genes, we first performed a targeted-NGS analysis of the patient’s genomic DNA from peripheral blood leukocytes with the Goiter Ampliseq panel (Table 1). The proband had 27 exonic variants identified in the Single Nucleotide Polymorphism database (dbSNP) and one heterozygous exonic variant in KEAP1 (c.1448 G > A, p.R483H) not registered in the dbSNP. According to the NCBI’s ClinVar database,¹ the detected 27 variants in dbSNP, including 16 missense polymorphisms, have not been reported as pathogenic mutations.

Although several exons of target genes were not completely sequenced with the Goiter Ampliseq panel (Table 1), conventional Sanger sequencing indicated no pathogenic polymorphism in those exons. Our subsequent analysis by Sanger methods validated the heterozygous mutation of c.1448 G > A in KEAP1 in the patient (Figure 4A).

Our sequence analysis of genomic DNA from the proband’s 16-year-old son showed an identical heterozygous guanine-to-adenine transition at position 1448; however, it was absent in the proband’s daughter and mother (Figure 4B). The R483H variant was evaluated as “disease causing or deleterious” by Mutant Taster,² PolyPhen2,³ PROVEAN,⁴ and PANTHER.⁵ This variant was not described previously in other individuals by the 1000 Genome Project⁶ or the Exome Aggregation Consortium.⁷

In this patient, NRF2 positive staining in the nuclei was detected in >30% of the cells in nodular lesions (Figure 2C) but in <10% of the cells in the non-nodular parenchyma (Figure 2B). In all four control specimens, a nuclear accumulation of NRF2 was detected in <10% of the cells in both the nodular lesions and the non-nodular parenchyma (Figure 2D and data not shown).