Thyroid cancers constitute the majority of endocrine malignancies. Differentiated thyroid cancers, arising from follicular cells, constitute more than 95% of all thyroid cancers and are histologically classified as either papillary thyroid carcinoma (PTC), follicular thyroid carcinoma, or poorly differentiated thyroid carcinoma [1
]. The 2017 World Health Organization Classification of Tumors of Endocrine Organs incorporates the newly defined entity noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP), which is a neoplasm with an unspecified, borderline, or uncertain clinical behavior, but not a benign or malignant tumor [1
]. Anaplastic thyroid carcinoma also arises from follicular cells but is a highly aggressive disease, while differentiated thyroid cancer is generally considered an indolent disease. Medullary thyroid carcinoma, a tumor derived from parafollicular C cells (neural crest origin), has biologic features that differ from those of follicular cell-derived cancers. The incidence of thyroid cancer has been increasing steadily worldwide over the last few decades [3
]. The most rapid rise in the incidence of thyroid cancer was observed in South Korea, where the rate increased nearly 15-fold from 1993 to 2011 [5
]. This increase is primarily attributed to the rising incidence of PTC, and specifically to that of low-risk PTC including subcentimeter-sized cancer or encapsulated follicular variant [4
]. PTC accounts for 85% of thyroid cancers in most countries, whereas in Korea PTC accounts for more than 95% of thyroid cancers (www.cancer.go.kr
]. Differentiated thyroid cancers are generally considered indolent; therefore, there is a need to develop preoperative diagnostic markers that can be used to identify thyroid cancers requiring surgical removal from thyroid nodules.
The cyclin D1 protein, coded by the CCND1
gene, is a gate-keeper regulating the transition from the G1 phase into the S phase of the cell cycle [9
]. Overexpression of cyclin D1 is observed in a variety of human cancers and is involved in tumorigenesis [9
]. The overexpression of cyclin D1 in human cancers can result from genetic alterations, changes in epigenetic regulation, gene transcription, and protein translation of CCND1
. We have shown that cyclin D1 is consistently overexpressed in PTC, and cyclin D1 immunostaining is useful for identifying the extent of tumor involvement [11
]. However, mutations and amplification of the CCND1
gene have rarely been found in the differentiated thyroid cancer [12
]. The CCND1
gene encodes two major splice variants: wild-type (CCND1a
mRNA) and an oncogenic isoform (CCND1b
consists of five exons containing a coding DNA sequence of 888 bp, which encodes a 295-amino acid protein (cyclin D1a). Failure to splice at the exon 4-intron 4 boundary of the CCND1
pre-mRNA generates the CCND1b
splice variant that contains intron 4 [9
]. Because intron 4 contains a translation stop codon, CCND1b
lacks exon 5 and encodes a 275-amino acid protein (cyclin D1b) with early termination of transcription at the intron 4 region [9
]. Degradation of cyclin D1 is regulated by C-terminal PEST domain and threonine residue 286 [10
]. However, the absence of a protein-destabilizing (PEST) domain and threonine residue 286 in cyclin D1b suggests that cyclin D1b is regulated by a different mechanism and is more stable than cyclin D1a [10
]. Polymorphism rs9344 (G870A) at the critical exon 4 splice junction is associated with the expression of CCND1b
]. The G allele at nucleotide 870 (codon 242) preferentially encodes the CCND1a
transcript, and the A870 allele preferentially encodes CCND1b
]. However, expression of CCND1b
can be found in tumors homozygous for G/G [15
]. Cyclin D1b expression is related to tumorigenesis, tumor progression, and poor outcomes in various human cancers such as those of the brain, esophagus, lung, breast, colon, prostate, and bladder, as well as in Ewing sarcoma and lymphoma [9
]. However, the roles of cyclin D1b expression and G870A polymorphism in thyroid cancer have not been fully defined.
In this study, we evaluated the diagnostic and clinical utility of mRNA and protein expression of CCND1 isoforms in thyroid tumors, and assessed the correlation between G870A polymorphism and expression of CCND1. We also investigated using the expression of cyclin D1b to differentiate NIFTP from benign thyroid tumors and PTC.
To the best of our knowledge, this study is the first to demonstrate the clinicopathologic significance of mRNA and protein expression of CCND1 isoforms in thyroid tumors. NIFTP was distinguished from PTC by low expression of CCND1b mRNA and protein, whereas the expression level of CCND1a mRNA and protein in NIFTP did not differ from that observed in PTC. Cyclin D1b expression, as assessed by immunohistochemistry, was significantly lower in NIFTP than in its closest mimic, which is invasive encapsulated follicular variant of PTC. In PTC, nuclear expression of cyclin D1b was associated with aggressive clinicopathologic features including lymph node metastasis, risk of tumor recurrence, and advanced stage.
The association of CCND1
rs9344 (G870A) polymorphism and risk of thyroid cancer has been reported in a few studies [27
]. In the Polish population, the AA genotype was more frequently found in patients with PTC than in the healthy population (23.1% vs. 18.5%). The AA genotype may be a risk factor for the development of this type of cancer (odds ratio, 1.452; 95% confidence interval, 1.059–1.989) [27
]. A study on the Turkish population showed that the frequency of the AA genotype was significantly higher in patients with PTC than in healthy individuals (37.3% vs. 28.7%) [28
]. In our study, the polymorphism rs9344 was associated with the AA genotype. Frequency of the AA genotype was significantly higher in patients with PTC than in those with nodular hyperplasia (33% vs. 13%), although we did not compare these results with a healthy control group. Nevertheless, the CCND1
rs9344 (G870A) polymorphism may also be a risk factor for developing PTC in the Korean population; this notion is supported by the agreement of our results with the data obtained from studies of the Polish and Turkish populations.
Previous studies showed that high expression of nuclear cyclin D1 is associated with lymph node metastases in PTCs [29
]. This association, however, is not consistent. In other studies, lymph node metastasis of PTC was not associated with the intensity or distribution of cyclin D1 immunostaining [31
]. In our previous studies, as well as in the present study, we observed that cyclin D1a was consistently overexpressed in PTC and there was no correlation between its overexpression and lymph node metastasis [11
]. The expression level of CCND1a
mRNA also had no impact on lymph node metastasis, as assessed using our study cohort and TCGA dataset. With respect to pathologic diagnosis of thyroid tumors, cyclin D1a immunostaining was useful for the differential diagnosis of non-neoplastic hyperplasia and thyroid neoplasms; this is because nodular hyperplasia was completely negative for the expression of cyclin D1a. However, the expression of cyclin D1a did not show clinically- or diagnostically-significant differences between NIFTP and other types of thyroid cancers, provided that all these tumors exhibited overexpression of cyclin D1a. Previous studies exploring the diagnostic role of cyclin D1a in tumors with a follicular pattern showed no difference in cyclin D1a immunostaining among follicular adenoma, follicular thyroid carcinoma, and follicular variant of PTC [31
]. Another study showed the expression of cyclin D1a in nodular goiter [31
]. These conflicting results described in the literature, as well as in our studies, may reflect variations in population diversity, cut-off values used for the evaluation of cyclin D1 expression, and various conditions used for performing immunohistochemistry.
With respect to the expression of cyclin D1b in thyroid tumor, no preexisting data were found in the literature. In this study, we showed that cyclin D1b was not expressed in follicular adenoma and rarely expressed in NIFTP. In PTC, 64.7% of the samples were positive for nuclear expression of cyclin D1b, which was associated with tumor metastasis, advanced stage, and increased risk of recurrence. These results are consistent with previous findings showing that as an oncogenic isoform, cyclin D1b plays a role in tumorigenesis and progression, and is correlated with poor outcomes in various non-thyroidal cancers [9
]. A recent study showed that knockdown of CCND1b
promoted apoptosis and suppressed cancer-cell stemness and epithelial mesenchymal transition in human bladder cancer cells [32
]. These results indicate that the cyclin D1b oncoprotein may play a role in thyroid cancer progression.
Cyclin D1 is a nuclear protein regulating cell cycle progression from the G1 to the S phase and has been implicated in tumor invasion and metastasis in human cancers [33
]. Phosphorylation of threonine residue 286 within the PEST domain enables cyclin D1 nuclear exportation and subsequent ubiquitin-dependent degradation in the cytoplasm [34
]. Altered ubiquitin-proteasome system is responsible for cyclin D1 overexpression in tumor cells [10
]. Cytoplasmic expression of cyclin D1 can control cancer cell migration, invasion, and metastasis, but not cell proliferation [33
]. In this study, cytoplasmic immunoreactivity of cyclin D1a was observed in most thyroid cancers, but rarely found in follicular adenoma and NIFTP. The cytoplasmic expression of cyclin D1b was observed only in thyroid cancers and was associated with lymph node metastasis and advanced stage in PTC patients. Further studies are necessary to elucidate the cytoplasmic cyclin D1-dependent mechanisms that control cell adhesion and migration in thyroid cancer.
In some cases, the initial histologic criteria for NIFTP have resulted in the misdiagnosing of encapsulated classic PTC with predominant follicular growth as NIFTP in some cases [26
]. Our previous study showed that several encapsulated follicular patterned tumors with nuclear features of PTC developed micro-metastases in regional lymph nodes or harbored the BRAF
V600E mutation, when criteria of less than 1% papillae was allowed [37
]. In another study, a 6.0-cm thyroid tumor, which met the initial criteria for NIFTP, had concurrent RAS
promoter mutations [40
]. As a result, the diagnostic criteria of NIFTP have recently been updated to avoid misdiagnosing these thyroid cancers as NIFTP [2
]. The revised diagnostic criteria recommend using the criterion of “no (0%) well-formed papillae” and thorough examination of the whole-tumor capsule to exclude the presence of capsular or vascular invasion [26
]. Furthermore, the entire tumor tissue should be submitted for histologic examination to exclude the presence of any papillae when the tumor has florid nuclear features (nuclear score of 3) of PTC [26
]. Exclusion criteria include the presence of BRAF
V600E and BRAF
V600E-like mutations, or that of high-risk mutations (such as those in TERT
), even if the tumor meets the histologic criteria for NIFTP. Molecular testing, however, is not mandatory for NIFTP diagnosis. In our present study, one case met the former microscopic criteria for NIFTP [2
], but was positive for the BRAF
V600E mutation and showed positive immunostaining for cyclin D1b and BRAF VE1. Our findings further support the recommended detection of BRAF
V600E by molecular testing or immunohistochemistry in order to differentiate classic PTC from NIFTP. Immunohistochemical staining for nuclear cyclin D1b can be helpful in diagnosing NIFPT, provided that nuclear cyclin D1b is rarely expressed in NIFTP but is highly expressed in PTC. We also observed a significantly higher nuclear expression of cyclin D1b in invasive encapsulated follicular variant of PTC than in NIFTP, as assessed in an independent cohort. These observations suggest that cyclin D1b immunostaining can be used in distinguishing NIFTP from its closest histologic mimic, invasive encapsulated follicular variant of PTC. In this context, the nuclear positivity for cyclin D1b immunostain indicates that the entire tumor tissue should be evaluated for evidence of malignancy such as true papillae, high-grade features, or invasion.
Various benign thyroid nodules can show atypical nuclear features, mimicking those of PTC. Hashimoto’s thyroiditis is the most common cause of false positive results in preoperative aspiration cytology. Immunostaining for cytokeratin-19, galectin-3, HBME1, and loss of expression l of CD56 are frequently used to diagnose PTC. However, the expression of cytokeratin-19, galectin-3, HBME1, and loss of expression of CD56 are also detected in 20%, 20–40%, 20%, and 7–90% of patients with Hashimoto’s thyroiditis, respectively [41
]. In the present study, cyclin D1b was not expressed in Hashimoto’s thyroiditis, as shown in Supplementary Figure S1
. However, immunostaining for cyclin D1a was strongly positive in the Hürthle cells of Hashimoto’s thyroiditis, as shown in Supplementary Figure S1
. These findings suggest that immunostaining for cyclin D1b, rather than for cyclin D1a, can be used to differentiate between Hashimoto’s thyroiditis and PTC in limited biopsy samples or cell blocks.
RNA extracted from formalin-fixed paraffin-embedded (FFPE) tissue blocks has often suffered degradation over time. The quality of RNA derived from FFPE samples is affected by pre-analytical procedures including time to fixation from tumor removal, tissue-processing and paraffin-embedding methods, and sample storage as well as RNA extraction methods [43
]. Nevertheless, RNA has been successfully extracted from stored FFPE specimens and used for quantitative measurement of mRNA levels by quantitative reverse transcription polymerase chain reaction (qRT-PCR), microarray analysis, and next-generation sequencing with successful results [43
]. In our study, we used standardized protocols for pre-analytical workflow, extraction of RNA from FFPE blocks, and RNA gene expression analysis. The majority of FFPE samples showed a RIN (RNA Integrity Number) between 2 and 4. Quantification cycle (Cq) values of the qRT-PCR amplifications were between 24 and 30 for GAPDH
mRNA (internal control). There was no significant correlation between RIN values and Cq values, which was consistent with the results from a previous study of RIN values and its effect on qRT-PCR in FFPE samples [43