Differential miRNA expression profiles in variants of papillary thyroid carcinoma and encapsulated follicular thyroid tumours

Background: Recent studies showed a significant upregulation of distinct microRNAs (miRNAs) in papillary thyroid carcinoma (PTC). The objective of this study was to explore whether this upregulation could also be assigned to distinct histomorphological variants of PTC, especially the follicular variant and other encapsulated follicular thyroid tumours. Methods: We used total RNA of 113 formalin-fixed paraffin-embedded tissues of 50 PTCs ((10 conventional type (PTC-CT), 10 tall cell variants (PTC-TCVs), 30 follicular variants (PTC-FVs)), 10 follicular adenomas (FAs), 10 multinodular goitres (MNGs), 21 follicular thyroid carcinomas and 22 well-differentiated tumours of unknown malignant potential (WDT-UMP) to analyse the miRNA expression pattern of five selected miRNAs (146b, 181b, 21, 221 and 222) using RT–PCR TaqMan miRNA assay to explore the diagnostic utility of this method. Results: The mean values of the expression pattern of all miRNAS in PTCs show a statistically significant difference from those in MNG and FA with fold changes up to 90 for miRNA 146b, P<0.001. No differences in expression pattern could be showed between MNG and FA. The PTC-FVs differ significantly from FA in all five miRNAS, from MNG in three and from WDT-UMP in one miRNA with fold changes between 1.7 and 21.2, but failed to be of diagnostic value regarding individual cases with substantial overlaps. Conclusion: We conclude that analysis of a set of five selected miRNAS distinguish common variants of PTC from FA/MNG but failed to be a useful diagnostic method in individual and doubtful cases, especially in the differential diagnosis of encapsulated follicular thyroid tumours.

Papillary thyroid carcinomas (PTCs) represent the most common thyroid malignancy; its diagnosis is based on the demonstration of characteristic nuclear features such as enlargement, overlapping, irregularity of nuclear contours, ground glass nuclei, grooves and pseudoinclusions (Rosai et al, 1992;DeLellis and Williams, 2004). However, PTCs comprise a morphologically heterogeneous group covering distinct variants that are classified on the basis of the occurrence of predominantly papillary structures (conventional type (PTC-CT)), a distinct growth pattern (follicular variant (PTC-FV) or cell type (e.g., tall cell variant (PTC-TCV)), among other features (DeLellis and Williams, 2004;LiVolsi and Baloch, 2004). This heterogeneity is also reflected in variable prevalences of the three most common genetic alterations, RET/PTC rearrangements (Nikiforov et al, 1997;Adeniran et al, 2006), BRAF Xing, 2005) and RAS mutations (Vasko et al, 2003;Castro et al, 2006) that can be shown in approximately 70% of all PTCs (Nikiforova et al, 2008).
Since its original description by Crile and Hazard (1953) in 1953 and confirmation by Lindsay in 1960(Lindsay, 1960 the PTC-FV represents a diagnostic challenge . Differential diagnostic problems are caused by the encapsulated form of PTC-FV that essentially has to be distinguished from other encapsulated lesions. With regard to therapeutic consequences, it is more important to differ PTC-FV from follicular adenoma (FA) than PTC-FV from minimally invasive follicular thyroid carcinoma (FTC). Moreover, pathologists are frequently faced with encapsulated thyroid tumours having 'questionable' PTC-type nuclear changes, as it has been pointed out by Rosai, (2005). Those tumours are referred to as 'well-differentiated tumours of uncertain malignant potential' (WDT-UMP) in the literature.
MicroRNAs (miRNAs) are endogenous, non-coding, small RNAs that regulate gene expression. A large number of miRNAs are involved in almost every major cellular function (Cowland et al, 2007) and as a consequence, deregulation of miRNAs has also been linked to a broad variety of cancers (Calin et al, 2002Michael et al, 2003;Takamizawa et al, 2004;Iorio et al, 2005;Lu et al, 2005;Mattie et al, 2006;Murakami et al, 2006;Volinia et al, 2006). Recently, a few studies reported on deregulated miRNAs in PTC using miRNA microarrays (He et al, 2005;Pallante et al, 2006;Tetzlaff et al, 2007;Nikiforova et al, 2008) and RT -PCR TaqMan miRNA assay (Tetzlaff et al, 2007;Chen et al, 2008;Nikiforova et al, 2008), identifying a limited number of miRNAs that are significantly upregulated in PTC compared with normal thyroid tissue (He et al, 2005;Pallante et al, 2006;Chen et al, 2008;Nikiforova et al, 2008), hyperplastic nodules (Chen et al, 2008;Nikiforova et al, 2008) and multinodular goitre (Tetzlaff et al, 2007; for review, see Table 1), suggesting miRNA analysis as a promising tool in diagnostic thyroid pathology.
With regard to both the morphologic and genetic differences between PTC variants, we asked whether analysis of a distinct set of miRNAs is able to reliably distinguish common variants of PTC (PTC-CT, PTC-TCV and PTC-FV) from multinodular goitre (MNG) and FA and whether miRNA expression profiling is a useful tool in the differential diagnosis of encapsulated follicular thyroid tumours.

Patients and tumour samples
For this study we selected the following cases from the files of the Institute of Pathology and Neuropathology, University Hospital of Essen, Germany: 10 cases with an (almost) exclusive papillary architecture, the characteristic nuclear features outlined by the WHO classification (2004), and particularly lacking both the cellular and nuclear features of the tall cell variant of PTC that belong to the conventional-type group (PTC-CT). Another 10 cases were categorised as the tall cell variant of PTC (PTC-TCV). These tumours showed tumour cells at least twice high than wide with an abundant eosinophilic cytoplasm and typical nuclear characteristics, including eosinophilic pseudoinclusions. To explore differences in miRNA expression, especially in encapsulated follicular thyroid tumours, we selected 30 cases composed of 495% of follicular structures and characteristic nuclear features corresponding to the follicular variant of PTC (PTC-FV). Out of 50 PTCs, seven had been previously analysed for miRNA analysis Sheu et al, 2009). A total of 21 minimally invasive FTCs with either capsular (n ¼ 6) or vascular invasion (n ¼ 13) or both (n ¼ 2) were also included. In addition, we selected 22 encapsulated follicular tumours with questionable nuclear changes without vascular/capsular invasion; these tumours are categorised as WDT-UMP. A total of 10 encapsulated thyroid FAs and 10 MNGs were also included. All patients gave informed consent and the study was strictly performed according to the Declaration of Helsinki.

Macrodissection of tumour tissue
Macrodissection from paraffin-embedded specimens to obtain 'pure' tumour tissues was performed as described before (Sheu et al, 2007). From all cases, at least three tissue blocks were available and 'morphologic homogeneity', especially in variants of PTC, was proven in all blocks. Clinicopathologic data of all cases are summarised in Table 2.

RNA extraction
RNA was extracted using the RNeasy FFPE Kit (Qiagen, Hilden, Germany). In brief, tissue sections were deparaffinised by xylene/ ethanol treatment. Tissue pellets were resuspended in 150 ml buffer PKD, 20 ml proteinase K and incubated overnight on a shaker incubator at 56 1C. Further processing of the samples was performed according to the recommendations of the supplier.  (Nikiforova et al, 2008) and multinodular goiter (Tetzlaff et al, 2007). This set of miRNAs was analysed using the real-time RT -PCR scheme for miRNA quantification according to the protocol of Applied Biosystems (P/N: 4364031); this two-step protocol consists of reverse transcription with a miRNA-specific primer, followed by real-time PCR with TaqMan probes. The TaqMan miRNA assays used were also provided by Applied Biosystems. In brief, for each RT -PCR 50 ng RNA was reverse transcripted to cDNA using 3 ml specific looped RT primers (

miRNA expression pattern in variants of PTC
The miRNA patterns of different variants of PTC are depicted in Figure 1. Whereas PTC-CT and PTC-TCV did not differ significantly in mean DCt values and consecutively fold changes of gene expression for all types of miRNA, the miRNA patterns of both PTC-CT and PTC-TCV differed from PTC-FV. When comparing mean values of PTC-CT and PTC-FV, a significant difference was exclusively found for miRNA 146b (fold change 8.0; P ¼ 0.043). Mean DCt values of PTC-TCV differed significantly from PTC-FV in three miRNAs (146b, 21 and 222; Po0.028) that correspond to fold changes between 2.9 and 9.9. Regarding individual cases of PTC-FV, we observed a broad variability in every miRNA analysed. This variability covered the whole range of DCt values of benign (MNG and FA) and also TCV and CT of PTC. However, there was no overlap between every

miRNA expression pattern in encapsulated follicular tumours
Taking out PTC-FV as a group, we looked for differences in expression pattern compared with benign thyroid lesions ( Fig. 1 and Table 4). As it could already be observed for all PTCs (Table 3) Table 1). Considering the mean DCt values of all miRNAs analysed in our study, we showed highly significant changes between PTC and MNG/FA. However, the most common variants of PTC show a different miRNA expression pattern with similar profiles of PTC-CT and PTC-TCV in contrast to the follicular variant of PTC. As for diagnostic purposes, only miRNA 146b reliably distinguish the conventional and tall cell variant from benign thyroid lesions in individual cases, whereas all other miRNAs show substantial overlap. This is in accordance with a study by Chen et al (2008) who found miRNA 146b to be most consistently overexpressed in both conventional and follicular variants when compared with 'borderline' follicular lesions, although the number of five follicular variants in their study is rather small.
The functional relevance of overexpressed miRNA 146b and its effect on PTC tumourigenesis had been elucidated in two studies (Taganov et al, 2006;Jazdzewski et al, 2008) so far. Taganov et al, (2006) identified TNF receptor-associated factor 6 (TRAF6) and IL-1 receptor-associated kinase (IRAK1), which represent potential molecular targets of miRNA 146, as modulating the immune response in a NF-kB-dependent manner. As far as NF-kB is one of the key factors controlling anti-apoptotic response in thyroid cells, it is also modulated by activated MAPK (Namba et al, 2007) in PTC. This pathway, in turn, is involved in downstream effects of RET/PTC rearrangements, RAS and BRAF mutations, the latter being most frequently verifiable in PTC. Therefore, it seems not surprising that upregulation of miRNA 146 is more distinctive in the conventional and tall cell variant of PTC, as these variants are significantly associated with the common V600E BRAF mutation. However, in a previous study we asked for a possible correlation between the occurrence of V600E BRAF mutation and miRNA expression profile and found no significant differences in miRNA expression between PTC harbouring the BRAF mutation and wildtype BRAF, implicating that this mutation has no regulatory influence on the expression pattern of these 5 miRNAs ). Our results are in contrast to another study by Nikiforova et al (2008) who founded a strong relationship between miRNA expression and mutational status (BRAF, RET/PTC and PAX8-PPARg). The reported differences might be due to the variable number of tissue samples (28 vs 6) harbouring V600E mutation and probably different statistical analysis and illustration (raw data presenting DCt values vs principal component analysis (PCA)). This possible relationship should be validated in a larger cohort of PTCs.
The miRNA pattern of PTC-FV in this study differed from PTC-TCV in three (miRNAs 146b, 21 and 222) and from PTC-CT in 1 miRNA (miRNA 146b). In our previous study  we normalised fold changes in PTC variants to adjacent normal thyroid tissue in a pairwise manner and found that follicular variants showed 3/5 upregulated miRNAs (146b, 221 and 222), whereas the conventional type differed in 4/5 (146b, 181b, 221, and 222) and tall cells in all examined miRNAs, indicating a heterogeneous regulatory role of certain miRNAs within PTCs. Although PTCs as a group of tumours sharing cytologic similarities showed a distinct upregulated miRNA pattern, differences in various genetic alterations among PTC variants are quite common. The V600E BRAF mutation has been shown in approximately 43% of PTC (Lupi et al, 2007), ranging from 12% in PTC-FV to 77% in PTC-TCV (Xing, 2005). Interestingly, a distinct BRAF mutation (K601E) had been exclusively found in PTC-FV (Trovisco et al, , 2005 (Lima et al, 2004). The genetic and morphologic overlap of PTC-FV is also supported by the results of Castro et al (2006) who found similar frequencies of activating point mutations of the RAS genes and PAX8-PPARg rearrangement in PTC-FV, FA and FTC; both genetic alterations are absent (PAX8-PPARg; Kroll et al, 2000;Nikiforova et al, 2002) or exceedingly rarely (RAS; Vasko et al, 2003) found in 'non follicular variant' of PTC.
However, the comparison of miRNA expression profile of PTC-FV and other encapsulated follicular thyroid lesions revealed a broad variability among individual cases with substantial overlap. We observed a similar broad DCt range especially in PTC-FV and WDT-UMPs, reflecting that both tumours not only share distinct morphologic characteristics but also similarities in miRNA regulation. As for practical purposes, the determination of miRNA expression profile of our types analysed does not contribute to clarify the biological significance of those tumours assuming that there might be other factors than the characteristic nuclear features in the 'majority of the tumour' of PTC-FV as outlined in the WHO classification. Regarding clinical and therapeutical consequences, especially the discrimination between FA and PTC-FV and FA vs FTC, respectively, the miRNA expression analysis also failed to be of diagnostic value, although highly significant mean DCt values and consecutively fold changes up to approximately 21 (for 146b) between PTC-FV and FA indicate remarkable differences. However, these changes could only be showed in mean values but not in individual cases. Surprisingly and in addition to the identified miRNAs (197, 328, 346 and 192) by Weber (Weber et al, 2006) we found differences in miRNA expression pattern between FA and FTC with fold changes between 3.1 and 3.5 for miRNAs 146b and 21, and in accordance with their study, not for miRNAs 221 and 221, indicating the latter two being essentially involved in PTC pathogenesis, as it has previously been shown by Pallante (Pallante et al, 2006). This points towards the role of miRNA 146b as being generally involved in both (papillary and follicular) phenotypes of thyroid carcinogenesis, reflecting other genetic alterations that partly result in characteristic nuclear features. In our study, and as for the distinction between FA and FTC, miRNA 21 seem to have a regulatory role, as an upregulation of miRNA could recently be shown in RAS-transformed FRTL-5 thyroid cells (Talotta et al, 2009). Two targets of miRNA 21, the tumour suppressor genes PTEN and PDCD4, are downregulated in a novel autoregulatory loop mediated by miRNA 21 through the transcription factor AP1 in response to RAS, thus indicating a tumourigenetic role for miRNA, but they failed to be of diagnostic value in every single case in our study. We have shown that an analysis of a set of five selected miRNAs distinguish common variants of PTC from follicular adenoma and multinodular goitre but failed to be a useful diagnostic method in individual and doubtful cases, especially in the differential diagnosis of follicular thyroid tumours (PTC-FV, FA, WDT-UMP and FTC). In addition, miRNA expression profiling confirms the so far 'intermediate' position of PTC-FV between conventional and tall cell variants of PTCs on one hand, and on the other hand, the follicular thyroid tumours with partly nuclear features of unknown malignant potential and minimal invasive FTC.