Imaging, genetic testing, and biomarker assessment of follicular cell-derived thyroid cancer

Abstract

Thyroid carcinoma is the most common endocrine malignancy worldwide, and its incidence continues to increase. As such the approach to a recently identified thyroid nodule is important to understand. The relevant imaging, examination, and need for fine-needle aspiration biopsy (FNA) are discussed. In approximately 25% of nodules, the diagnosis cannot be established with FNA-based cytology, and surgical excision is necessary for definitive diagnosis. Recent advances in genetic and molecular testing may increase the diagnostic accuracy of FNA in managing thyroid nodules.

• Key words:
• BRAF
• gene expression classifier
• MiRInform
• Ras
• RET/PTC

Key messages

• Ultrasound and FNA are important tools in the assessment of thyroid masses.

• Genetic testing of thyroid nodules can increase the specificity for diagnosis of a malignant thyroid nodule. Gene expression classifier testing has increased the sensitivity for detecting malignancy, and results can be used to support a clinical decision to monitor patients instead of proceeding with diagnostic thyroidectomy.

• Thyroid lobectomy is still required as a diagnostic measure for evaluation of nodules in which FNA cannot safely rule out malignancy.

Introduction

Since the 1970s there has been an increasing incidence of thyroid cancer diagnosed in most populations. The overwhelming majority of thyroid cancers diagnosed worldwide are papillary thyroid cancer (PTC) (1). In the United States more than 85% of thyroid malignancies are PTC, and the increase in incidence of thyroid cancer has largely been attributed to this histologic type (2,3). Follicular thyroid cancer (FTC), a well- differentiated thyroid cancer (WDTC), is the second most common thyroid cancer (1). Anaplastic or undifferentiated thyroid cancer is very aggressive with a very poor survival. In comparison, the 10-year survival for patients with PTC and FTC is 97.4% and 90.5%, respectively (4).

Once a thyroid nodule is discovered the next steps in evaluation of the patient are focused on appropriate diagnosis and treatment of the lesion. Imaging is an important initial step in diagnosis and management of a thyroid cancer. Ultrasonography can help decide which nodules need to be further evaluated. Current imaging techniques used to identify thyroid nodules and manage thyroid cancer will be discussed. Concerning lesions require further assessment with fine-needle aspiration biopsy (FNA).

FNA can indicate the need for surgical resection but cannot always rule out a malignancy (5). Currently, patients with such indeterminate nodules will undergo surgical excision to obtain a diagnosis. Recently, the use of genetic and molecular testing has been suggested as an adjunct to cytopathology to help establish an accurate diagnosis. While the list of biomarkers that have been identified and studied is prolific, some mutations and molecular variations have been more extensively reported on than others. A few of these have been incorporated into commercially available panels. This review will focus on the mutations as well as biomarkers which are incorporated by these panels.

Imaging

Imaging is an important diagnostic tool for evaluating thyroid nodules which are discovered in many different ways. Some patients will present with a complaint of a neck mass, or sometimes a nodule is found on routine physical examination. Thyroid nodules can also be discovered incidentally on neck ultrasound (US), computed tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET) scan performed for an unrelated condition (6). Unless there is invasion of surrounding structures or metastatic disease, it can be difficult to identify malignant features of an incidentally found nodule on CT or MRI (7). Thyroid nodules with evidence of calcifications, especially punctate or peripheral calcifications, on CT are more likely to be malignant (8–10). Approximately 2%–3% of patients undergoing PET have 2-deoxy-2[¹⁸F]fluoro-d-glucose (FDG) avid thyroid nodules. The risk of malignancy is approximately 33% in nodules with focal uptake, and there is a greater risk of malignancy in nodules with diffuse uptake (3,11). Regardless of the method of discovery, once a thyroid nodule is discovered, diagnostic thyroid US is the recommended next step in assessment of the nodule (3).

On ultrasound examination of thyroid nodules, there are various findings that raise the suspicion of malignancy. The shape of the thyroid nodule, specifically ‘tall rather than wide’ (the antero-posterior length is greater than medial-lateral length), is the most specific ultrasonographic feature of malignancy (Figure 1) (12–14). Nodules with irregular borders are associated with invasion into the surrounding parenchyma and are more often found in malignant nodules than benign lesions. The presence of microcalcifications also increases the risk of a nodule being malignant. Microcalcifications are small (less than 2 mm) echogenic punctate lesions which are generally present in papillary thyroid cancer secondary to the presence of psammoma bodies (13,15). On ultrasound examination, echogenicity refers to the intensity of a nodule compared to the surrounding parenchyma. Suspicious nodules tend to be hypoechoic and are darker than the surrounding thyroid (13). The use of Doppler US allows for the assessment of vascular flow through a mass; malignant lesions are hypervascular and have a unique flow pattern. Central flow in the nodule is particularly concerning for malignancy, as benign nodules tend to have either no increased flow or peripherally increased flow (12).

Figure 1. Thyroid ultrasound showing microcalcifications, poorly defined borders, and taller than wider shape.

If a nodule is found to be malignant, US examination of the lateral neck is recommended to evaluate for lymph nodes suspicious for metastatic disease. Similar to thyroid nodules, concerning findings on lymph node evaluation include hypoechogenicity and microcalcifications. Suspicious findings specific to lymph nodes are a round shape, loss of the fatty hilum, and peripheral vascularity. Per the American Thyroid Association (ATA) guidelines, for well-differentiated thyroid cancer (WDTC), US is the only recommended imaging modality for routine preoperative evaluation for metastatic disease (3).

Other imaging assessments such as diagnostic radioiodine whole-body scintigraphy (dxWBS), post-treatment iodine scan (PTS), ¹³¹I-SPECT/CT, and ¹⁸FDG-PET are useful for disease monitoring and detection of recurrences. DxWBS is primarily recommended by the American and European Thyroid Associations when residual thyroid tissue cannot be well assessed following surgery. The other circumstance in which dxWBS may be considered is if the results could potentially alter the radioactive iodine (RAI) treatment plan. If a dxWBS is going to be performed, an ¹²³I scan is recommended to prevent a blunted response to ¹³¹I treatment (3,16). PTS is generally performed 5–8 days after completion of RAI treatment and can allow for an assessment of response to treatment. Occasionally, PTS will identify metastatic disease not identified on pre-treatment scan. The results of the PTS can alter the treatment and surveillance approach for patients. Patients found to have no remaining disease will benefit little from future dxWBS and therefore are not recommended to have further surveillance dxWBS. On the other hand, patients found to have avid lesions on PTS may undergo further radioactive iodine treatment or need to have closer surveillance intervals (3).

Imaging plays a crucial role in the appropriate surveillance of patients with a history of thyroid cancer. Currently, US of the thyroid bed and central and lateral neck is recommended at 6–12 months following completion of treatment. Recombinant TSH-stimulated thyroglobulin (Tg) level combined with US has been shown to have increased sensitivity for detection of recurrence in comparison to dxWBS in WDTCs (17). DxWBS has a limited advantage, and its use remains controversial among clinicians (18). Some authors recommend its use because a small subset of patients have low stimulated Tg levels yet have evidence of metastatic disease on dxWBS (19).The limitation, however, of these studies is that most of the Tg-negative and dxWBS-positive metastatic disease involves cervical metastases which can be identified on neck US (19,20). The use of ¹⁸FDG-PET/CT is recommended in patients who have had a negative dxWBS but have elevated Tg levels (3,21,22). ¹³¹I-SPECT/CT and ¹⁸FDG-PET/CT have a higher sensitivity for detection of active disease than dxWBS in patients previously treated with radioactive iodine (22–24). ¹³¹I-SPECT/CT has been shown to provide accurate anatomic localization of avid lesions and is significantly advantageous for surgeons (23–25).

The use and limitations of fine-needle aspiration

As the discovery of thyroid nodules increases, clinicians are challenged to determine the most appropriate management of such nodules (26). Because a diagnosis of malignancy cannot be conclusively established from imaging alone, fine-needle aspiration (FNA) biopsy is recommended for patients who have pertinent positives on history or have nodules with concerning characteristics on US (3,12). Pertinent positives of the patient history include a history of exposure to radiation, family history of thyroid cancer or thyroid cancer-related syndromes, whole-body radiation for bone marrow transplant, and rapid changes in physical exam findings.

FNA has dramatically decreased the number of non-malignant thyroid nodules resected but cannot definitively rule out malignancy in about 25% of nodules (27,28). There are six recognized categories for reporting cytopathology of thyroid FNA according to the Bethesda System. Each category is stratified by risk of malignancy. Nodules classified as malignant are primarily papillary thyroid cancer, and the risk of malignancy is 97%–99% in this category. In tumors which do not exhibit papillary architecture the diagnosis cannot be made with the same certainty by FNA. In follicular lesions of undetermined significance (FLUS) repeat FNA is recommended for further evaluation. Lesions suspicious for follicular neoplasm carry a 15%–30% risk of malignant disease, and a diagnostic lobectomy is recommended despite a strong chance of the lesion being benign (27). Currently, the definitive diagnosis of the thyroid nodule can only be established histologically after the portion of the thyroid gland containing the nodule is removed and adequate examination of the entire nodule with its capsule can be performed (27).

More recent efforts in molecular testing are geared towards providing a method to diagnose thyroid nodules accurately by FNA. There have been multiple studies evaluating various genetic and molecular markers with the goal of finding a method of testing which is both sensitive and specific for the appropriate diagnosis. Several early studies focused on identifying mutations which lead to malignant transformation and using these as markers for diagnosis of malignancy (28). Other molecular markers that have been extensively studied include mRNA expression, microRNAs (miRNA), immunocytochemical markers, and epigenetic markers (28–31).

Genetics of thyroid carcinoma

In sporadic thyroid cancers, there are many somatic mutations which are key events in the malignant transformation of follicular cells (29). MAP (mitogen-activated protein) kinases and similar cell signaling pathways such as the PI3K-AKT play a vital role in controlling important cell functions like differentiation and apoptosis (Figure 2). As such, point mutations and chromosomal rearrangements involving the function of these two pathways are the most common genetic alterations in thyroid cancer (28).

Figure 2. Diagram of intracellular pathways commonly Involved in follicular cell carcinogenesis.

BRAF

BRAF, the most significant RAF family activator of MAP kinase kinases, is a serine-threonine kinase which is activated in response to extracellular signals (32). The most common BRAF mutation associated with carcinoma is a missense somatic mutation at position 1799 resulting in a valine to glutamic acid substitution at residue 600 (V600E). This change results in an increase in the basal activity of the protein.

The BRAF^V600E is commonly found in the classic and tall-cell variants of papillary thyroid carcinoma and rarely in the follicular variant (33). Prognostically, the presence of this BRAF mutation is associated with high-risk clinicopathological characteristics such as multifocal lesions, extrathyroidal invasion, presence of and more extensive lymph node metastases at the time of presentation. A shorter time to recurrence has also been reported with the presence of this mutation (34–42).

RET/PTC

Rearrangements involving RET are well recognized to have an association with thyroid carcinoma. RET is a proto-oncogene coding for a membrane receptor tyrosine kinase, and in its wild-type form RET is regulated by ligand binding (43,44). With the RET/PTC rearrangement the receptor becomes constitutively active in thyroid follicular cells, resulting in an increased transformation potential (44,45). There have been at least 15 RET/PTC rearrangements identified, the most common of which are paracentric inversions on chromosome 10q: RET/PTC1 and RET/PTC3. RET/PTC1 results from the fusion of RET to the coiled coil domain containing protein 6 (CCD6) and RET/ PTC3 is RET fused to ELE1 (also known as NCOA4 or ARA70), a transcriptional coactivator (46,47).

The presence of RET/PTC rearrangements in PTC is well established and shown to be more common in children and in patients with radiation-induced thyroid cancer (36,40–43). RET/PTC rearrangements have been found in cases of Hashimoto's thyroiditis and in follicular adenomas, but this variation may be more related to the detection method (48,49). In a recent review, Nikiforov suggests that methods of genetic testing identifying clonal RET/PTC rearrangements are specific to PTC, whereas non-clonal mutations are found in a more diverse group of nodules. Ultra-sensitive methods of detecting rearrangements may identify mutations that are present in only a few cells which are not representative of the entire nodule, resulting a high false positive rate (44).

PAX8/PPARγ

Another common genetic rearrangement in malignant thyroid nodules is the PAX8/PPARγ mutation. This is a 3p25 and 2q13 translocation which results in an inframe fusion of PAX8 from chromosome 2 to the peroxisome proliferator-activated receptor γ gene (PPARγ) (50). Translation of the translocated gene results instead in a novel PAX8/PPARγ fusion protein (PPFP) which alters the function of the individual gene products (51). The PAX8/PPARγ translocation is most commonly identified in FTC, but it has also been demonstrated in follicular variant papillary thyroid cancer (FVPTC) (50,52–54).

RAS

The RAS family of genes code for cell membrane-associated proteins involved in signal transduction of tyrosine receptor kinases as well as other G-protein coupled pathways. Point mutations of RAS occurring in codons 12, 13, and 61 of the RAS genes have been demonstrated in various malignancies. In thyroid carcinoma the most common RAS mutations have been identified in codon 61 of N-RAS or H-RAS (55). The result is a constitutively activated protein and unregulated stimulation of signaling pathways such as the MAPK and PI3K-AKT pathways (56).

RAS mutations are predominantly found in follicular carcinoma but also occur in 10%–20% of PTCs, usually the follicular variant (25). RAS mutations are also early events in tumor dedifferentiation and found in poorly differentiated and anaplastic thyroid carcinoma (57).

P53

P53 is a tumor suppressor gene commonly mutated in several human cancers. Normally, p53 plays a role in cell cycle regulation by causing cell cycle arrest in the G1 phase. Mutations of p53 have been found to cause overexpression of potentially non-functional proteins, or impaired transcription of the gene. The majority of ATCs have been found to have a p53 mutation (58–60). Until recently p53 mutations have rarely been reported in WDTC, but were found in 20% of oncocytic follicular carcinoma using next-generation sequencing (61).

PTEN and PIK3CA

PTEN is involved in regulation of the PI3K-AKT signaling pathway. Loss of function or decreased function of PTEN is associated with poorly differentiated thyroid carcinomas. The loss of function of PTEN has been related to silencing from DNA methylation, and increasing methylation is positively correlated to the amount of dedifferentiation (62).

PIK3CA, a catalytic subunit of PI3K activates the PI3K-AKT pathway, and increased expression of this gene is associated with tumorigenesis. Gain of function mutations of PIK3CA have been associated with dedifferentiated thyroid carcinoma (63). In addition to gain of function mutations, gene amplification has also been found in malignant thyroid nodules. Overall, alterations in this gene are more specific to carcinoma than to benign nodules. In one study 52% of FTCs harbored such an alteration versus 8% in follicular adenomas (Figure 3) (64–66).

Figure 3. Common mutations present in thyroid carcinoma. BRAF is most commonly seen in papillary thyroid cancer (PTC), but also found in anaplastic thyroid cancer (ATC). RET/PTC rearrangements are most often in PTC. PAX8/PPARγ is mostly associated with follicular thyroid cancer (FTC), but also found in follicular variant papillary thyroid cancer (FVPTC). RAS is more common in FTC.

Genetic testing as a diagnostic aid

The prevalence of the BRAF mutation varies in different populations and has been reported in 29%–83% in malignant nodules (67). In areas of high prevalence BRAF testing alone can play a major role in accurate diagnosis by FNA (68). Where BRAF mutations are not as prevalent, testing for RAS mutations or RET/PTC and PAX8/PPARγ rearrangements in addition to BRAF increases specificity for non-diagnostic cytology (69–73). Testing for these molecular markers has been shown to increase the accuracy of diagnosis when cytology is indeterminate; it is cost-effective through the reduction of diagnostic thyroid lobectomies (74). In a study of 1056 thyroid FNA samples, cytologically diagnosed as follicular lesion of undetermined significance, follicular neoplasm, and suspicious for malignant cells, the risk of malignancy was 88%, 87%, and 95% if genetic mutations were present, compared to 6%, 14%, and 28%, respectively, if the sample was negative for mutations (70). Cantara et al. report an increased total diagnostic accuracy in evaluating 235 samples when mutational analysis was incorporated (93% versus 83% with cytology alone) (71). Similar results have been corroborated by others (69,73). MiRInform (Asuragen, Austin TX, USA) is a commercially available testing panel that incorporates BRAF^V600E, N, K, and H Ras mutations as well as RET/PTC1, RET/PTC3, and PAX8/PPARγ rearrangements. The predicted increase in sensitivity for diagnosing malignancy from 44%–60% to 80%–90% when the panel is used with cytology is based on previously published clinical trials (75).

Traditional forms of genetic testing are very reliable due to stability of DNA and specificity of the mutations for thyroid carcinoma. A major limitation of genetic testing employing these methods of sequencing is the low incidence of identified mutations. The overall result is a low sensitivity for diagnosing thyroid cancer in indeterminate nodules. For example, in a previously mentioned study of indeterminate nodules, 967 samples were adequate for molecular analysis and only 87 nodules harbored any of the tested mutations (70).

Thyroseq (UPMC, Pittsburgh PA, USA), is a recently developed panel which relies on next-generation sequencing to identify mutations in a series of 12 genes. The selected genes are commonly involved in carcinogenesis: BRAF, N, K, and H RAS, PIK3CA, RET, CTNNB1, PTEN, TSHR, AKTI, p53, and GNAS. The method of sequencing employed by this panel can accurately detect mutations from smaller amounts of DNA than traditional methods. This allows Thyroseq to target 284 mutational hotspots in the 12 tested genes. The authors predict that if the panel is combined with testing for the commonly known genetic rearrangements such as RET/PTC, a mutational event may be identified in up to 80% of thyroid carcinomas. The increased detection of mutations may result in an increased sensitivity of genetic testing. Importantly, 51 of 52 samples obtained by FNA were sufficient to be sequenced, which is promising for the diagnostic utility of this test. Studies are still lacking on the overall diagnostic accuracy of FNA biopsy using Thyroseq to evaluate indeterminate nodules, but it is predicted to increase the sensitivity of gene testing (61).

Gene expression testing

Variable mRNA expression in thyroid nodules has also been discovered to be a viable option to distinguish between benign and malignant nodules. There are a few mRNA transcripts which are of value in isolation for accurately assessing the indeterminate nodule. Of these, the high mobility group AT-hook 2 (HMGA2) expression has been found to be increased in malignant tissue. HMGA2 expression testing has been shown to have a high specificity (95%) but a relatively low sensitivity (71%) for diagnosis of malignancy from FNA samples of indeterminate nodules (76).

Tissue-trained gene expression classifiers (GEC) based on expression of several mRNAs have also shown considerable potential (28,30,77,78). An initial study published by Chudova in 2010 was limited by sample size, but, soon after, the commercially available Afirma (Veracyte, CA, South San Francisco, USA) gene expression classifier was developed and has had promising results. Further, a recently published study of Hurthle cell neoplasms has established a unique profile for these tumors which is distinct from other FTCs and PTC (79).

Afirma tests for 167 separate RNA transcripts in FNA biopsy specimens. A multicenter trial of Afirma testing of 265 samples with indeterminate cytology resulted in high sensitivity (92%) for malignant nodules from those classified as FLUS and suspicious for follicular neoplasm by histopathology. This sensitivity results in negative predictive values of 95% and 94%, respectively (80). Afirma testing has been shown to change the practice of physicians monitoring patients with FLUS. Patients with negative Afirma testing are much less likely to be referred to undergo surgery for an indeterminate nodule (78,81). The greatest limitation of Afirma is its low specificity of 52% when evaluating indeterminate nodules, and it cannot be used to diagnose malignancy.

Another recently published study describing the use of whole-transcriptome profiling of thyroid nodules may increase the accuracy of diagnosis by FNA. This study involved the use of a non-commercial GEC (kNN249), composed of 249 individual markers. Many of these markers corresponded to non-protein coding sequences. They were able to show a PPV of 100%, and in contrast to Afirma (Veracyte, CA, South San Francisco, USA) this profile is more specific in diagnosing malignancy (100% versus 52%). The sensitivity of kNN249, however, is less than Affirma at 76.9% compared to 92%; the NPV for kNN249 is 85.7% (82). When compared to genetic testing, kNN249 offers equal specificity but a higher sensitivity for the diagnosis of malignancy.

Gene expression can be influenced by several factors, one of which is the presence of different miRNAs. MiRNAs are short ribonucleotide sequences that bind to mRNA and influence degradation or translation of these sequences. Similar to somatic mutations they can function as tumor suppressors or oncogenes depending on the original function (83). There have been many studies evaluating the expression profile of various different miRNAs, but overexpression of miR-221, 222, and 181b in PTC has been shown across multiple studies (84,85). Similarly, upregulation of miR-885 in follicular carcinoma in comparison to benign tissue is well established (84,85). Keutgen et al. describe the use of a statistical model developed based on expression of four different miRNAs (miR-222, miR-328, miR-197, miR-21) in FNA samples of previously excised nodules. The statistical model was then used to analyze the expression profile of those miRNAs from 72 indeterminate nodules, yielding a 100% sensitivity and 86% specificity for malignancy and a diagnostic accuracy of 90% (86).

Immunocytochemistry

Galectin-3 is a lectin that is involved in cell adhesion and cell cycle regulation, but it is not expressed in the cytoplasm of healthy follicular cells. It is often expressed in the cytoplasm of malignant thyroid cells. As such, galectin-3 makes a reasonable target for immunocytochemistry. Bartolazzi et al. evaluated galectin-3 as a molecular marker in thyroid nodules with indeterminate cytology, and found that testing for galectin-3 had a sensitivity of 78% and specificity of 93% for malignancy (87). Results vary among different studies, with sensitivity ranging from 55% to 83% and specificity varying from 81% to 100% for galectin-3 testing (88,89). Hector Battifora mesothelial-1 (HBME-1) is an antigen commonly found in villi of mesothelial cells discovered to be of diagnostic value for evaluating thyroid nodules (90). Franco et al. found the addition of HBME-1 to galectin-3 testing to increase the sensitivity for malignancy from 82.6% to 94.7% in cytologically indeterminate and suspicious nodules (89).

Cytokeratin-19 (CK19) is a protein expressed in epithelium that helps distinguish PTC from other hyperplastic lesions. In a recent study, immunocytochemistry with CK19 staining was more specific than galectin-3 for PTC (91). Ki-67 is a proliferation marker found in more cells of FTC than follicular adenomas or PTC (92). When looking at lesions concerning for follicular neoplasm a positive immunocytochemistry panel of HBME-1, CK19, and Ki-67 yielded a diagnostic accuracy of 91% (94).

Beta-catenin, another important cytologic marker, plays a role in cell adhesion complexes and is also involved in cell signaling in the Wnt pathway. Normally, Wnt signaling is important in stem cell proliferation, and activating mutations in this pathway can result in abnormal proliferation of progenitor cells (93). Beta-catenin is usually quickly degraded in the cytoplasm through phosphorylation, but mutations can result in abnormal stabilization of the protein. Increased beta-catenin has been shown in the nucleus and cytoplasm of poorly differentiated thyroid cancer cells. In comparison to follicular adenomas in which beta-catenin staining primarily localizes to the cell membrane (79% of nodules), in PTC, FTC, and FVPTC, beta-catenin primarily has cytoplasmic and nuclear staining (87%, 80%, and 71% of nodules) (94). A previously published meta-analysis has found a sensitivity and specificity of immunocytochemistry for beta-catenin of 85% and 90%, respectively, in distinguishing malignant nodules from benign lesions.

Future directions

Currently, the benefit of molecular testing remains confined to nodules with indeterminate cytology. A recent study evaluating the use of molecular testing for all nodules, regardless of cytology and radiographic studies, found the testing to be detrimental to appropriate patient management (95). Additionally, the NPV of Afirma in a population with a high thyroid cancer prevalence was found to be 89%, indicating a need for further trials to determine accuracy in different population types (96). While genetic and molecular testing continues to provide more information from FNA sampling, the eventual goal is to decrease or eliminate the diagnosis of ‘indeterminate nodule’ on FNA. Currently the ATA does not recommend for or against the commercially available GECs or genetic testing as expert panel review of the available literature is lacking (97). Studies evaluating the patient and cost benefits of using various panels of molecular testing would shed light on the best available methods to establish diagnosis. For many of the newer panels large multi-institutional studies examining the efficacy of the testing methods for diagnosis are lacking. Ultimately an ideal single method of testing should be cost-effective, sensitive, and specific enough to allow treatment catered to each individual at diagnosis.

Declaration of interest: The authors report no conflicts of interest.