Gene Expression Profiles of Papillary Thyroid Microcarcinoma
The objective was to identify gene expression profile of papillary thyroid microcarcinoma. To help improve diagnosis of papillary thyroid microcarcinoma, we performed gene expression profiling and compared it to pair normal thyroid tissues. We performed microarray analysis with 6 papillary thyroid microcarcinoma and 6 pair normal thyroid tissues. Differentially expressed genes were selected using paired t test, linear models for microarray data, and significance analysis of microarrays. Real-time quantitative reverse transcription–polymerase chain reaction was used to validate the representative 10 genes (MET, TIMP1, QPCT, PROS1, LRP4, SDC4, CITED1, DPP4, LRRK2, RUNX2). We identified 91 differentially expressed genes (84 upregulated and 7 downregulated) in the gene expression profile and validated 10 genes of the profile. We identified a significant genetic difference between papillary thyroid microcarcinoma and normal tissue by 10 upregulated genes greater than 2-fold (P < 0.05).
Thyroid carcinomas less than or equal to 1 cm are almost exclusively papillary and are termed papillary thyroid microcarcinoma (PTMC).1 It is likely that the number of PTMC observed recently in many countries will continue to increase. Although some of the risk factors of PTMC include lymph node metastasis, extrathyroidal extension, and so on, PTMC has an indolent clinical course and generally has a better prognosis than papillary thyroid carcinoma (PTC). Recent studies showed that gene expression data from microarray analysis hold promise for understanding the molecular basis of thyroid carcinogenesis and thus improving tumor classification and predicting response to therapy.2,3 The first array investigation on PTC (including PTMC) was performed by Huang et al4 in 2001 using U95A Affymetrix GeneChips containing more than 12,000 transcripts. In their study of 8 PTC tumors, which were compared with normal surrounding tissues from the same 8 individuals, the authors specified 50 genes with the most distinct gene expression changes and identified a number of additional PTC-specific genes, many of which were associated with cell cycle or mitogenic control and some previously found in other cancers. However, to our knowledge, no study has identified the gene expression profile (GEP) of PTMC only and compared it with that of normal thyroid tissue of PTMC. We aimed to characterize the GEP of PTMC and compare it to that of pair normal thyroid tissues. We validated the microarray technique by selecting 10 upregulated genes and analyzing them by real-time quantitative reverse transcription-polymerase chain reaction (qRT-PCR).
Material and Methods
Tissue samples
The thyroid tissues were obtained from 6 PTMC patients who underwent total thyroidectomy between January 2010 and June 2011. Six PTMC fresh frozen tissues and 6 normal fresh frozen thyroid tissues from contralateral lobes of the patients were included in the study. Normal tissue samples were collected from the same patients for pairwise analysis. The medical records of the patients were reviewed to determine their clinical characteristics. Informed consent was obtained from all patients. This study was approved by the Institutional Review Board.
RNA extraction and microarray
Total RNA extraction was done with Trizol reagent (Life Technologies, Rockville, MD, USA) according to the manufacturer's instructions. After homogenization, 1 mL of solution was transferred to a 1.5-mL Eppendorf tube and centrifuged at 12,000 ×g for 10 minutes at 4°C to remove insoluble material. The supernatant containing RNA was collected, mixed with 0.2 mL of chloroform, and centrifuged at 12,000 ×g for 15 minutes at 4°C. RNA in the aqueous phase was transferred into a new tube, precipitated by mixing 0.5 mL of isopropyl alcohol, and recovered by centrifugation at 12,000 ×g for 10 minutes at 4°C. The RNA pellet was washed briefly in 1 mL of 75% ethanol and centrifuged at 7500 ×g for 5 minutes at 4°C. Finally, total RNA pellet was dissolved in nuclease water, and its purity and integrity were assessed with an Agilent Bioanalyzer 2100. Gene expression was analyzed with GeneChip Human Genome U133 plus 2.0 arrays (Affymetrix, Santa Clara, CA), which consist of 54,675 probe sets representing approximately 38,500 well-characterized human genes. For each gene, 11 pairs of oligonucleotide probes were synthesized in situ on the arrays. Biotinylated cRNA was prepared according to the standard Affymetrix protocol from 250 ng of total RNA (Expression Analysis Technical Manual, 2001, Affymetrix). After fragmentation, 15 μg of aRNA was hybridized for 16 h at 45°C on a GeneChip Human Genome array. GeneChips were washed and stained in Affymetrix Fluidics Station 450. GeneChips were scanned using the Affymetrix GeneChip Scanner 3000 7G. The data were analyzed with Microarray Suite version 5.0 (MAS 5.0) using Affymetrix default analysis settings and global scaling as normalization method. The trimmed mean target intensity of each array was arbitrarily set to 100. The normalized and log-transformed intensity values were then analyzed using GeneSpring GX 11.5.1 (Agilent Technologies, CA). Fold-change filters included the requirement that the genes be present in at least 200% of controls for upregulated genes and lower than 50% of controls for downregulated genes.
Statistical and bioinformatics analysis
Data files (.CEL) were generated in GeneChip Operating Software (GCOS; Affymetrix), using default settings for scaling and detection call parameters. The Robust Multichip Average (RMA) Express 1.0.4 program (Bolstad BM, University of California, Berkeley, CA) was used for background adjustment and quantile RMA normalization of the 54,675 probe sets encoding human genome transcripts. Differentially expressed genes (DEGs) between the 2 groups were identified using linear models for microarray data (LIMMA) in R using Bioconductor, and significance analysis of microarrays (SAM, version 4.0, 2011, www-stat.stanford.edu/∼tibs/SAM/) was used for overall and pairwise comparisons between the groups on RMA-normalized signal intensities. LIMMA uses Bayes method, which is a different algorithm for identification of significantly expressed genes. Both SAM and LIMMA methods yielded large numbers of DEGs. DEGs were ranked on moderated t statistics, and those with t scores >5 were followed up further. The moderated t-statistic score is based on the ratio of the log2 fold change to its standard error. Similar to traditional t, moderated t ranks genes on a combination of magnitude and consistency of differential expression but downgrades values where exceptionally small variance elevates the ratio in the absence of detectable difference in expression level. False discovery rates (FDRs) were also calculated for all comparisons. To screen for the genes that have larger effects on the traits and to reduce the number of less relevant genes, we compared the lists of top-ranked genes from the LIMMA analysis with the lists obtained from the SAM analysis. Genes significant (P < 0.01) from the LIMMA lists and ranking high on the SAM lists were selected. SAM was applied to log2 transformed data of the filtered data set to compare the PTMC and normal groups. The principal component analysis (PCA) is an exploratory tool to characterize predominant gene expression patterns. Through PCA, it is possible to identify whether samples from the same conditions have similar patterns of gene expression. Hierarchical clustering data were clustered groups that behave similarly across experiments using GeneSpring GX 11.5.1 (Agilent Technologies, CA). Clustering algorithm with Euclidean distance and average linkage was used. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genome (KEGG) pathway analyses were then carried out for DEGs using the Database for Annotation, Visualization, and Integrated Discovery (DAVID, version 6.7, david.abcc.ncifcrf.gov) to identify the most significant biological processes and signaling pathways.
Real-time quantitative PCR
To further validate the results of the microarray test, we used qRT-PCR to examine the relative expression of 10 genes (MET, TIMP1, QPCT, PROS1, LRP4, SDC4, CITED1, DPP4, LRRK2, RUNX2) among the DEGs selected by integration of the results from the LIMMA and SAM analyses. We designed and synthesized 2 oligonucleotide primers based on the published genomic sequence of the 10 selected genes: MET (hepatocyte growth factor receptor, GenBank accession no. BG170541), TIMP1 (TIMP, metallopeptidase inhibitor 1, NM_003254), QPCT (glutaminyl-peptide cyclotransferase, NM_012413), PROS1 (protein S [alpha], NM_000313), LRP4 (low-density lipoprotein receptor-related protein 4, AA584297), SDC4 (syndecan 4, NM_002999), CITED1 (CBP/p300-interacting transactivator, with Glu/Asp-rich carboxyl-terminal domain, 1, NM_004143), DPP4 (dipeptidyl-peptidase 4, M80536), LRRK2 (leucine-rich repeat kinase 2, AK026776), and RUNX2 (runt-related transcription factor 2, AL353944). Amplification of transcripts was performed with 3 μg/μL total RNA and reverse transcriptase-polymerase chain reaction (RT-PCR), using high-capacity cDNA reverse transcription kit (ABI Applied Biosystems, CA, USA). Real-time PCR primers were designed and synthesized by Macrogen Corporation (Seoul, Korea). The following gene-specific primers were used: MET, 5′-GAAGGAGGGA CAAGGCTGAC-3′ and 5′-TGTTATTGTG′; TIMP1, 5′-TCTGGCATCCTGT TGTTGCT-3′ and 5′-TGTGCATTCCTCACAGCCAA-3′; QPCT, 5′-TC CCAGACAGACTCGGAGAG-3′ and 5′-GCCCCTCCAA AGAGTGATCC-3′; PROS1, 5′-GTGTGCCTTCCCTTGAACCT-3′ and 5′-GGCAAGCATAACA CCAGTGC-3′; SDC4, 5′-GTCCTGGCAGCTCTGATTGT-3′ and 5′-GGACT CCACACAACATCCGT-3′; CITED1, 5′-G′ and 5′-TCCCGAGGAACTAGTGGGAG-3′; DPP4, 5′-TGTTTAACTCGG GGCCGAAA-3′ and 5′-CGTTTGGAGACCACCACAGA-3′; LRP4, 5′-GAG GACGACTGTGCAGACAA-3′ and 5′-GGTCGGGC CGTAGTTCATA G-3′; LRRK2, 5′-ATGCACTCACGAGCTTTCCA-3′ and 5′-CGGGACCTGGTAG GTACTGA-3′; RUNX2, 5′-GACCAGTCTTACCCCTC CTACC-3′ and 5′-CTGCCTGGCTCTTCTTACTGAG-3′; and β-actin, 5′-GCAGAAGGAGATC ACTGCCCT-3′ and 5′-GCTGATCCACATCT GCTGGAA-3′. For the quantification of gene amplification, real-time PCR was performed using an ABI 7300 Sequence Detection System with SYBR Green. Target sequences were amplified at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. β-Actin was used as an endogenous normalization control. Experiments were independently repeated 3 times and were calculated by the ΔΔCt method. The n-fold change in mRNAs expression was determined according to the 2−ΔΔCt method.
Results
To explore novel genes involved in PTMC, we investigated the genome-wide expression profiles of PTMCs and pair normal thyroid tissues. Using GeneChip Human Genome U133 plus 2.0 arrays (Affymetrix, Santa Clara, CA), we compared the expression profiles of PTMCs and pair normal tissues. After performing paired t test analyses, we identified 878 genes whose expression differed significantly at P < 0.05 with a fold change of >2 and <−2. Genes differing in expression were represented in a volcano plot (Fig. 1). Among these, 591 genes were upregulated and 287 genes were downregulated in PTMCs compared with pair normal tissues. Then, we used LIMMA software and SAM in screening the DEGs, and we selected those genes with adjusted P values <0.01 and fold change > 2 and <−2 as the threshold cutoff. We identified 91 DEGs in PTMC (Table 1). The expressions of 84 genes, including HMGA2, GABRB2, COL10A1, COMP, and LRP4, was increased more than 2-fold in PTMC tissues than that in pair normal tissues, and the expressions of 7 genes, including CITED2, FOXA2, LOC201651, and ANKRD37, decreased more than 2-fold than that in pair normal tissues.
Citation: International Surgery 102, 1-2; 10.9738/INTSURG-D-17-00049.1
As shown in Fig. 2, the relationship between the two groups is apparent in the PCA. The patterns that are grouped in similar areas within the 3-dimensional condition scatterplot demonstrate the similarity between members of each sample group, as well as differences between groups (Fig. 2), suggesting that the samples used in this study were appropriately prepared and selected. Genes differing in expression were subjected to hierarchical clustering analysis; the heatmap shows that PTMCs are distinctly different from pair normal tissues (Fig. 3). By using significantly upregulated and downregulated genes, the altered GO categories were identified from DAVID. Thus, we were able to show the processes significantly linking cell adhesion, biological adhesion, wound healing, regulation of body fluid levels, endochondral ossification, coagulation, endochondral bone morphogenesis, hemostasis, bone morphogenesis, cell–cell adhesion, apoptosis, programmed cell death, cell–substrate adhesion, focal adhesion, membrane organization, blood vessel morphogenesis, fat cell differentiation, glycosylation, biopolymer glycosylation, protein amino acid glycosylation, and skeletal system development. The genes examined were also submitted to the DAVID in an attempt to identify significantly dysregulated pathways. The KEGG pathway analysis for the 91 genes showed that the 2 significant pathways involved extracellular matrix (ECM)-receptor interactions and cell adhesion molecules.
Citation: International Surgery 102, 1-2; 10.9738/INTSURG-D-17-00049.1
Citation: International Surgery 102, 1-2; 10.9738/INTSURG-D-17-00049.1
To verify the data obtained from microarray analysis, we selected 10 genes (MET, TIMP1, QPCT, PROS1, LRP4, SDC4, CITED1, DPP4, LRRK2, RUNX2) for qRT–PCR analysis. The results showed that the gene expression of the detected 10 genes was in accordance with the microarray data, as shown in Fig. 4. The qRT–PCR results further confirmed the microarray data, with all of the tested genes showing statistically significant upregulation greater than 2-fold (P < 0.05).
Citation: International Surgery 102, 1-2; 10.9738/INTSURG-D-17-00049.1
Discussion
In this study, we performed GEP analysis using oligonucleotide microarray and identified 91 DEGs (84 upregulated and 7 downregulated) between PTMCs and pair normal thyroid tissues. The criteria of selection for DEGs were rather strict that, whereas other microarray studies on thyroid carcinomas used a P value of <0.05, we adopted an adjusted P value of <0.01 in LIMMA.
To know the similarity of GEP between PTMC and PTC, we compared them and we found significant genes of PTC in 2 meta-analysis studies. Griffith et al conducted a comprehensive meta-analysis of PTC gene expression profiling studies to identify meaningful biomarkers and reviewed 21 published studies, in which 34 comparisons were performed from 10 different expression platforms of microarrays; the analysis showed that 39 genes (23 upregulated and 16 downregulated in PTC) showed the same expression patterns among thyroid carcinoma, regardless of the tumor type, in an overlap of 3 or more studies.5 Among the 23 upregulated genes in PTC from their study, 13 genes (MET, TIMP1, QPCT, PROS1, FN1, LRP4, SDC4, CITED1, LGALS3, DPP4, TUSC3, P4HA2, CCND1) corresponded with those that were selected among the 84 upregulated genes in our study. In addition, Vierlinger et al adopted a meta-analysis approach for 4 publicly available microarray data sets on PTC.6 Among the 17 upregulated genes they identified, 13 genes (PROS1, LRP4, NPC2, LAMB3, DPP4, SDC4, QPCT, TIMP1, MET, CDH3, GGCT, CITED1, CHI3L1) overlapped with the genes identified in our study. Eight genes (MET, TIMP1, QPCT, PROS1, LRP4, SDC4, CITED1, DPP4) in this study were selected in the previous 2 meta-analyses. In the pathway analysis, the most obvious feature of PTMC, similar to PTC, is upregulation of extracellular activities, such as cell communication and adhesion and ECM-receptor interactions (Table 2). Up-regulation of ECM microenvironment genes has been described as an effect of BRAF mutations through phospho-ERK1/2 signaling.7 In addition, recent studies have shown the role of ECM in tumor progression and invasion through ECM remodeling and stiffening of the tissue stroma.8 These data and those from our study suggest dysregulation of the ECM-epithelial cell interactions as an early event, even before any evidence of invasion becomes visible in PTMC. Similarly, Birnie et al9 reported dysregulated cell adhesion and ECM-receptor interaction pathways in their comparison of prostate cancer stem cells and their normal and differentiated counterparts.
PTMC can remain biologically silent with minimal morbidity for decades. However, a significant percentage of PTMC may show lymph node metastasis and a small number may even have distant metastases. Clinical characteristics or parameters such as tumor size, age, sex, tumor multifocality, vascular or capsular invasion, extrathyroidal extension, lymph node metastasis, and histological variants of PTC need to be considered in risk-adapted management of PTMC to achieve minimal morbidity. In addition, molecular markers, combined with these clinical risk factors, should be taken into account. For example, the presence of the BRAF mutation in PTMC has been shown as associated with increasingly aggressive features, such as extrathyroidal extension, lymph node metastasis, multifocality, and high stage.10 The identification and investigation of GEP associated with risk factors of PTMC might improve the diagnosis, treatment, and prognosis of PTMC.
We identified 91 differentially expressed genes (84 upregulated and 7 downregulated) in the gene expression profile and validated 10 genes of the profile.
Conclusion
The present study presented a comprehensive analysis of gene expression profiling in PTMC with the use of microarray technology, which identified 91 DEGs (84 upregulated and 7 downregulated) between PTMCs and normal thyroid glands. To verify the reproducibility of these genes, we performed qRT–PCR using the same RNA used for microarray analysis. The qRT-PCR results were similar to those obtained from the oligonucleotide microarray. Ten representative genes of PTMC were analyzed; MET, TIMP1, QPCT, PROS1, LRP4, SDC4, CITED1, DPP4, LRRK2, and RUNX2.
Volcano plot of probe sets differing between PTMCs and normal tissues. Fold change (x axis) is plotted against statistical significance (y axis) for each probe set. Genes upregulated with a fold change >2 and P < 0.05 are depicted in red, and those downregulated with a fold change >2 and P < 0.05 are shown in blue. Gray represents genes in the arrays that were not found to differ significantly between 2 groups.
Principal component analysis (PCA) was applied to 12 thyroid samples that were characterized by the gene expression of all probes on the GeneChip Human Genome U133 plus 2.0 arrays (Affymetrix, Santa Clara, CA), including the control group (blue) and PTMC group (red). DEGs selected by LIMMA, SAM, and paired t test are shown.
Cluster analysis of the probes that were increased and decreased significantly in PTMC compared with normal tissue. A dendrogram of the cluster correlation is shown on the left. DEGs selected by LIMMA, SAM, and paired t test are shown. Pseudocolors indicate differential expression (red indicates transcript levels greater than the median; black indicates transcript levels equal to the median; green indicates transcript levels below the median; distance metric was Pearson centered and the average linkage rule was used).
Validation of array-based gene expression profiles by quantitative real-time PCR.
Contributor Notes