In the current study, we observed that CD56 expression level was useful to differentiate PTC from FA and NIFTP using different H-score cut-off values, but the CD56 expression levels were significantly different depending on the histologic types of PTC. Most previous studies reported loss of CD56 expression in PTCs without further specifying their histologic variants [9–17, 23, 24]. In our study, classic, infiltrative follicular, and diffuse sclerosing variants of PTC showed a considerably reduced CD56 expression. However, tall cell variant PTC, classic PTC with tall cell features, Warthin-like PTC, and cribriform-morular PTC had significantly higher H-scores of CD56 expression than did classic PTC.
CD56 NCAM mRNA expression in TCGA data of PTC supports the differential expression of CD56 according to the histologic types. The reason for CD56 expression being different for each PTC variant has not yet been clarified. Many previous studies have focused on the usefulness of CD56 in differentiating between malignant, especially PTC, and benign thyroid tumors [9, 12]. Only a few studies have analyzed CD56 expression by specific PTC variants. Scarpino et al. [8] reported relatively high rates of CD56 expression in various PTC variants: 63% (12/19)% of the classic variant, 90% (19/21) of the follicular variant, 20% (1/5) of the diffuse sclerosing variant, 60% (3/5) of the encapsulated variant, 67% (4/6) of the oncocytic variant, and 80% (4/5) of the tall cell variant. Although the level of CD56 expression is lower in PTC than in adjacent normal tissue, the positive expression of CD56 may be more frequently found in the tall cell variant than in the classic variant. Since the studied cases are too few, conclusions based on these data could be a hasty generalization. Likewise, there are very limited subgroup analysis data on the CD56 in the other uncommon variants of PTC [13]. However, CD56 expression differs from subtype to subtype of PTC, as our data show, and more data on this subject will be needed.
CD56 regulates cell motility and affects the migration capacity of the carcinoma cells. Reduced expression of CD56 in tumor cells has been implicated for aggressive tumor behavior or poor prognosis in some other cancer types, such as colorectal cancer, cholangiocarcinoma, and pancreatic cancer [25–28]. In a transgenic mouse model, the loss of CD56 function increased lymph-node metastasis in pancreatic β cell tumors and also increased lymphangiogenesis [25, 26]. An in vitro study of the PTC cell line demonstrated that silenced CD56 is related to the increased adhesive capacity of PTC cells to the extracellular matrix, but the migratory capacity of the tumor cells was decreased [8]. These results enabled us to hypothesize that aggressive subtypes of the PTC, e.g., tall cell variant PTC and classic PTC with tall cell features [18], would have a lower expression of CD56 than did other subtypes. Instead, CD56 expression levels were significantly higher in tall cell variant PTC and classic PTC with tall cell features than in classic PTC. Also, our present study showed no significant correlation between reduced CD56 expression and aggressive clinicopathologic characteristics. Therefore, we suggest that the CD56 expression occurs early during PTC development but may not be associated with the progression of PTC.
In a recent study, researchers demonstrated that CD56 is a handy screening marker in thyroid cytology. CD56 showed high sensitivity (90%) and diagnostic accuracy (88.4%) in the diagnosis of the indeterminate lesion (Bethesda category III) [29]. Although the overall diagnostic performance of CD56 is good [16], our results reveal the possible pitfall when using CD56 for preoperative diagnosis. Expression of CD56 in the tall cell variant, classic PTC with tall cell features, Warthin-like, or cribriform-morular PTCs will not cause misdiagnosis in the surgical specimen, because the histological characteristics of these variants are evident. However, the preoperative prediction based on immunostaining of CD56 on cytology cell blocks or core-needle biopsy may require careful attention. Looking at our data, it will need careful consideration of the different expression levels between the subtypes of PTC to avoid misdiagnosis.
Diagnoses of follicular adenoma, NIFTP, and PTC are strictly morphology-based. But still, when the nuclear features of encapsulated follicular-patterned tumors are questionable for the determination of PTC, the differential diagnosis ranges from benign FA to malignant PTC depending on the pathologists’ subjective diagnostic thresholds. In this regard, our data discovered the excellent performance of CD56 immunostaining in distinguishing NIFTP from PTC and FA with high AUC values (0.866 and 0.885, respectively) using different cut-off values for the loss of CD56 expression. High H-scores of CD56 expression in NIFTP were also useful in discriminating NIFTP from FVPTC (AUC 0.935). These results are consistent with a previous study on distinguishing encapsulated follicular variant PTC from infiltrative follicular variant PTC or classic PTC [23]. The previous study had 90.9% sensitivity and 59.8% specificity in distinguishing infiltrative follicular variant PTC and encapsulated follicular variant PTC. Therefore, loss of CD56 expression could be useful for differential diagnosis of follicular-patterned neoplasm including NIFTP.
In our study, we used the H-scoring method to evaluate the degree of CD56 expression. Using this method has disadvantages that it is not routinely used in the pathologic diagnosis, it can be subjective to interpret the intensity, and the result can vary depending on the detection system. Despite the disadvantages, the derivation of the H-score using computer-aided image-analysis system retains consistency between the cases and can more objectively quantify the degree of CD56 expression when expression pattern is heterogeneous.