miR-451a is underexpressed and targets AKT/mTOR pathway in papillary thyroid carcinoma

Papillary Thyroid Carcinoma (PTC) is the most frequent thyroid cancer. Although several PTC-specific miRNA profiles have been reported, only few upregulated miRNAs are broadly recognized, while less consistent data are available about downregulated miRNAs. In this study we investigated miRNA deregulation in PTC by miRNA microarray, analysis of a public dataset from The Cancer Genome Atlas (TCGA), literature review and meta-analysis based on a univocal miRNA identifier derived from miRBase v21. A list of 18 miRNAs differentially expressed between PTC and normal thyroid was identified and validated in the TCGA dataset. Furthermore, we compared our signature with miRNA profiles derived from 15 studies selected from literature. Then, to select possibly functionally relevant miRNA, we integrated our miRNA signature with those from two in vitro cell models based on the PTC-driving oncogene RET/PTC1. Through this strategy, we identified commonly deregulated miRNAs, including miR-451a, which emerged also by our meta-analysis as the most frequently reported downregulated miRNA. We showed that lower expression of miR-451a correlates with aggressive clinical-pathological features of PTC as tall cell variant, advanced stage and extrathyroid extension. In addition, we demonstrated that ectopic expression of miR-451a impairs proliferation and migration of two PTC-derived cell lines, reduces the protein levels of its recognized targets MIF, c-MYC and AKT1 and attenuates AKT/mTOR pathway activation. Overall, our study provide both an updated overview of miRNA deregulation in PTC and the first functional evidence that miR-451a exerts tumor suppressor functions in this neoplasia.


SUPPLEMENTARY DATA PTC tissue samples molecular analysis
Tumor samples were screened for the most common mutations and rearrangements reported in PTC. Hematoxylin and eosin stained slides of FFPE tissues were reviewed by an expert pathologist and used to guide the macro-dissection of adjacent unstained sections. Paraffin was removed by xylene extraction followed by ethanol wash. BRAF (exon 15), NRAS (exon 2) and HRAS (exon 2) mutations were tested on genomic DNA extracted by Qiamp FFPE DNA kit (Qiagen, Chatsworth, CA, USA) and amplified by PCR as previously described [1]. PCR products were subjected to direct sequencing using an ABI Prism 3500 DX Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) and then analyzed by ChromasPro software.
RET and NTRK1 rearrangements were tested on total RNA extracted by MasterPure RNA Purification Kit (Epicentre Biotechnologies, Madison, WI), including a DNaseI treatment step. cDNA was synthetized with random hexamers and Superscript™ III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). To detect rearrangements, samples were first analyzed for the expression of tyrosine kinase (TK) domain and then positive samples were further tested for the expression of extracellular (EC) domain; samples expressing TK domain but not EC domain were considered rearranged. The following primers were used: RET TK domain sense 5'-TGCAGCGAGGAGATGTACC-3' and antisense 5'-CCAGGTCTTTGCTGATGTCC-3'; RET EC domain sense 5'-CACCGGCCTCCTCTACCTTA-3' and antisense 5'-AGCCGCGGTTGCGG ACACTG-3'; NTRK1 TK domain sense 5'-CCTCCGATCCCATGGACCTGAT-3' and antisense 5'-CCTGCGTGATGCAGTCGATT-3'; NTRK1 EC domain sense 5'-ACCCTCTGCACTGTTC TTGT-3' and antisense 5'-ATTGGGCACCTGGACCTTC-3'.  Table 1 of the original publication independently of fold change (FC) and Local false discovery rates (Local FDR). To assign MIMAT accessions, we obtained probes sequences of the used miRNA chip [2,3] and directly verified the mature miRNA sequences in miRBase. Because in the same study the authors demonstrated in validation analyses the overexpression of hsa-miR-146b-5p (MIMAT0002809) rather than hsa-miR-146a-5p (MIMAT0000449), initially identified by microarray, we decided to include also this miRNA in the analyzed dataset.

Literature review and meta-analysis
Study 2 by Pallante et al. (Supplemental dataset 2): we selected the list of miRNAs reported in Table 1 of the original publication independently of FC and statistical significance (P). Data were ordered by decreasing FC. To assign MIMAT accession, we obtained probes sequences of the used miRNA chip [2] and performed alignments by BLAST (blast.ncbi.nlm.nih.gov/Blast) to identify within the probes the specific active site corresponding to the mature miRNA sequences. Once identified the mature miRNA sequences, these were directly verified in miRBase.
In the subsequent studies as more information was available about miRNAs identity and mature miRNA species (-3p/-5p), MIMAT accessions were assigned directly based on the ID reported in miRBase and/or in miRBase Tracker.
Study 3 by Nikiforova et al. (Supplemental dataset 3): we selected the list of top ten upregulated miRNAs reported in Table 1 Table 4 of the original publication. Because these miRNA were identifed in PTC samples classified according to variant and BRAF mutation (FV-PTC, wtC-PTC and mutC-PTC) but the authors showed that the three classes had very similar miRNA signatures displaying more quantitative than qualitative differences, we analyzed simultaneously the three classes. Data were ordered by decreasing log 2 (FC) and an absolute cutoff value of 0.7 was applied. Only miRNAs passing the cutoff value and common among the three classes were included in the final dataset.  Table 1 of the original publication, already filtered by statistical significance (P<0.05).
Study 11 by Wang et al. (Supplemental dataset 11): we selected the list of miRNAs reported in Table 2 Table 2 of the original publication. As for study 7, we analyzed simultaneously the two reported classes of PTC (FVPTC and Classic PTC). We filtered out miRNAs (n=4) differentially expressed between the two classes of PTC and selected only miRNA statistically significant (p-value <0.05) in both classes.
Study 13 by Swierniak et al.: as in this study multiple comparisons and approaches were used, we decided to include in our meta-analysis all the three reported analyses. 1) NGS analysis in PTC and paired normal tissues (Supplemental dataset 13_1): we selected the list of 44 miRNAs statistically significant (paired FDR T vs. N <0.05) reported in Table 1 of the original publication.
2) NGS analysis in PTC and unrelated noncancerous thyroid (Supplemental dataset 13_2): we selected among miRNAs reported in Table 1 Table 3 of the original publication only miRNA with concordant expression (fold T-N) between NGS and microarray data.
Study 14 by TCGA: as for study 13, we included in our meta-analysis the two NGS analyses reported in this study. 1) paired PTC/normal thyroid tissues (Supplemental dataset 14_1) and 2) PTC and normal thyroid tissues (Supplemental dataset 14_2). We selected the list of miRNAs reported in Supplemental Table S6 of the original publication. Data were filtered by FC (absolute FC >2) and statistical significance (dataset 14_1 BH adjusted p-value from the Wilcoxon paired test <0.05; dataset 14_2 BH adjusted p-value from the Wilcoxon unpaired test <0.05).
Study 15 by Mancikova (Supplemental dataset 15): although in this study were reported two distinct NGS analyses in two independent sets of PTC and normal thyroid, limited overlap between the two sets was found possibly due to different RNA extraction technique used, as proposed by the authors. Therefore, for our metaanalysis we selected only the list of miRNAs commonly deregulated in both the analyzed sets reported in Supplemental Table S4 of the original publication.