Original Article
Malignant Risk Stratification of Thyroid FNA Specimens With Indeterminate Cytology Based on Molecular Testing ´ , MSc1; Vladan Zivaljevic ´, MD, PhD2; Svetislav Tatic ´ , MD, PhD3; Svetlana Paskas, PhD1; Jelena Jankovic 4 5 5 ic ´, PhD ; Dragica Radojkovic ´ , PhD ; Svetlana Savin, PhD1; ´ , MD ; Aleksandra Nikolic Vesna Boz 1 ´, PhD and Dubravka Cvejic
BACKGROUND: Fine-needle aspiration (FNA) has been employed for many years for examining thyroid nodules, and the cytology of aspirates is the primary determinant for whether thyroidectomy is indicated. Fifteen to thirty percent of thyroid nodules, not being clearly benign or malignant, fall into an indeterminate category. The main goals of molecular diagnostics for thyroid nodules are to prevent unnecessary surgery in patients with benign nodules and to stop patients with malignant nodules from being subjected to repeated operations. METHODS: This study was designed to evaluate the diagnostic utility of 4 markers in thyroid FNA cytology via testing for the BRAF V600E mutation and the expression of microRNA-221, microRNA-222, and galectin-3 protein in FNA samples with indeterminate cytology. RESULTS: A predictor model distinguishing benign samples from malignant samples on the basis of the 4 aforementioned markers was formulated. This decision model provided a sensitivity of 73.5%, a specificity of 89.8%, and a diagnostic accuracy of 75.7%. The positive predictive value was 80.6%, and the negative predictive value was 85.5%; this suggested that the prediction had good reliability. CONCLUSIONS: One hundred twenty FNA samples were examined, and 62 nodules were classified as benign with the proposed diagnostic algorithm. This resulted in a reduction of the initial 120 patients to 58 and thus C 2015 decreased by half the number of persons undergoing surgery. Cancer (Cancer Cytopathol) 2015;123:471-9. V
American Cancer Society. KEY WORDS: BRAF; fine-needle aspirates; galectin-3; indeterminate cytology; microRNA 221 (miR-221); microRNA 222 (miR-222); thyroid cancer.
INTRODUCTION Thyroid cancer typically manifests as a thyroid nodule. Palpable nodules are found in approximately 4% to 7% of the population, whereas those that can be detected by sonography are present in 19% to 35% of the population, and nodules are discovered at autopsy in 8% to 65%.1 The incidence of malignancy is up to 13% for solitary nodules and up to 10% for multinodular goiters.2–4 Fine-needle aspiration (FNA) has been employed for many years to examine thyroid nodules. Thus, the cytology of aspirates is the primary determinant for whether thyroidectomy is indicated. In terms of decision making, the most problematic categories are the indeterminate ones (Bethesda categories III, IV, and V).5 Roughly 15% to 30% of thyroid nodules, not being clearly benign or malignant, fall into an indeterminate category. The risk of malignancy for these categories is 5% to 75%.5
Corresponding author: Svetlana Paskas, PhD, Institute for the Application of Nuclear Energy, Department of Endocrinology and Radioimmunology, University of Belgrade, Banatska 31b, 11080 Belgrade, Serbia; Fax: (011) 381 11 2618 724; [email protected] 1 Institute for the Application of Nuclear Energy, Department of Endocrinology and Radioimmunology, University of Belgrade, Belgrade, Serbia; 2Center for Endocrine Surgery, Institute of Endocrinology, Diabetes, and Diseases of Metabolism, University of Belgrade, Belgrade, Serbia; 3Institute of Pathology, Medical Faculty, University of Belgrade, Belgrade, Serbia; 4Department of Endocrine and Cardiovascular Pathology, Clinical Center of Serbia, Belgrade, Serbia; 5Institute of Molecular Genetics and Genetic Engineering, University of Belgrade, Belgrade, Serbia
Received: January 2, 2015; Revised: April 2, 2015; Accepted: April 2, 2015 Published online April 27, 2015 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/cncy.21554, wileyonlinelibrary.com
Cancer Cytopathology
August 2015
471
Original Article
The main goals of molecular diagnostics for thyroid nodules are to prevent unnecessary surgery in patients with benign nodules and to stop patients with malignant nodules from being subjected to repeated operations. Moreover, in cases of indeterminate cytology, the risk of watchful waiting might be too high for most patients.6 Performing molecular analyses such as somatic mutation testing, gene and microRNA (miRNA) expression, and immunocytochemistry enables patients to be separated into low- and high-risk categories. In this study, we evaluated a set of 4 markers and investigated their diagnostic potential for distinguishing benign thyroid nodules from malignant ones. We tested for the BRAF V600E mutation and the expression of microRNA-221 (miR-221), microRNA-222 (miR-222), and galectin-3 protein. A BRAF mutation is one of the most prevalent somatic genetic events in human cancer and has the hallmark of a conventional oncogene.7 The most common BRAF mutation found in thyroid cancer is V600E, which constitutes 95% of all known BRAF mutations.8 This mutation leads to constitutive activation of a BRAF kinase and is the most common genetic change in papillary thyroid carcinoma (PTC).8–10 It is also found in poorly differentiated and anaplastic carcinomas, and this suggests that these cancers are PTC-derived.11 Other rare BRAF mutations, such as K601E and small in-frame insertions or deletions surrounding codon 600, have been described,12 but they are present in a minority of papillary carcinomas (1%-2%).13 Somatic mutation tests have high specificity but poor sensitivity for differentiating benign indeterminate thyroid lesions from malignant ones. In addition, approximately 30% of thyroid cancers harbor none of the mutations tested in the aforementioned panels. Therefore, when the test is mutation-negative, an overall malignancy risk is still present. To expand the search for an optimal panel of markers, we introduced 2 miRNAs, miR-221 and miR222, already shown to distinguish PTC from benign tumors. When the expression levels of miRNAs in PTC were compared, miR-221 and miR-222 were among the 10 most upregulated miRNAs in PTC.14 miRNAs are suitable biomarkers because their expression is often dysregulated in cancer, and they are more stable than other RNAs; this makes isolation and quantification by quantitative polymerase chain reaction (PCR) more feasible.15 472
Several protein-based biomarkers, alone or combined in panels, have been used for the diagnosis of thyroid cancer by immunohistochemistry. Among the most promising, galectin-3 has been studied in normal and pathological thyroid tissue. Galectin-3 is one of the nonintegrin, b-galactoside–binding lectins. It has an affinity for carcinoembryonic antigen, immunoglobulin E, laminin, and other mucins.16 Overexpression of galectin-3 is suitable for the distinction of malignant lesions from benign ones.17,18 Galectin-3 expression is a complementary and useful diagnostic method for indeterminate thyroid FNA samples.19,20 The aim of this study was to establish whether such a combination of selected markers could be helpful in distinguishing benign tumors from malignant tumors and thus in reducing the number of thyroidectomies for benign follicular lesions. To achieve this, we developed an algorithm for the management of patients with indeterminate FNA based on the results of the combination of these 4 markers. The goal of this study was not to validate the role of each individual molecular marker or to decide which of the 4 markers is redundant because they reflect malignancy at different levels (ie, DNA, RNA, and protein).
MATERIALS AND METHODS Patients and FNA Samples
FNA samples were obtained from patients with ultrasonographically confirmed thyroid nodules. All clinical specimens examined in this study were approved by the Ethics Committee at the Center for Endocrine Surgery, Clinical Center of Serbia, Belgrade. Informed consent from patients was provided for all collected samples. These were aspirated preoperatively with a 25-gauge needle and transferred directly into 400 mL of RNAlater stabilization solution (Ambion; Life Technologies). On average, 3 to 5 FNA passes were performed for each thyroid nodule. Half of the aspirate was used for cytospin preparation for hematoxylin-eosin staining, and the other half was used for nucleic acid isolation. A cytopathologist reviewed all FNA specimens before surgery. The Bethesda criteria for the adequacy of FNA samples were adopted when this was possible: 6 groups with at least 10 cells per group were counted. Specimens that contained abundant colloid were considered adequate without the strict application of proposed numerical requirements. In addition, a specimen was considered satisfactory if the contamination of Cancer Cytopathology
August 2015
Malignant Risk Stratification/Paska s et al
thyrocytes with blood cells was less than 20% according to hematoxylin-eosin staining. The cytological reports were collected, and samples without a definitive benign or malignant local diagnosis, classified as diagnostic category III, IV, or V according to the Bethesda System for Reporting Thyroid Cytopathology, were taken for molecular testing. There were 56 patients in the atypia of undetermined significance/follicular lesion of undetermined significance category, 41 patients in the follicular neoplasia category, and 23 patients in the suspicious for malignancy category. In this way, we selected 120 patients with indeterminate cytology. The cytological diagnosis was established before molecular testing. An endocrine pathologist reviewed all surgery specimens, and the final histopathological diagnosis was used as a reference for molecular analysis. According to the final histopathological diagnosis, among 120 patients with indeterminate cytology, there were 30 diagnosed with PTC, 1 diagnosed with follicular thyroid carcinoma, 8 diagnosed with H€ urtle thyroid carcinoma (HTC), 2 diagnosed with anaplastic thyroid carcinoma (ATC), and 4 diagnosed with medullary thyroid carcinoma (MTC). The remaining patients had a benign diagnosis: 35 with goiters, 37 with follicular thyroid adenoma (FTA), and 3 with H€ urtle thyroid adenoma (HTA). Nucleic Acid Isolation
Total nucleic acids were extracted with the TRIzol reagent (Invitrogen, Carlsbad, Calif) according to the manufacturer’s instructions. The quantity of isolated DNA and RNA was assessed with a NanoVue spectrophotometer (GE Healthcare, Buckinghamshire, United Kingdom). Mutation Analysis
Samples were screened for the presence of BRAF point mutations V600E and K601E. BRAF exon 15 was amplified by standard PCR with forward primer TCATAATGCTTGCTCTGATAGGA and reverse primer GGCCAAAAATTTAATCAGT GGA. The amplification was performed under the following conditions: initial denaturation at 948C for 2 minutes; 40 cycles at 948C for 30 seconds, at 588C for 30 seconds, and at 728C for 30 seconds; and final elongation at 728C for 5 minutes. The obtained PCR products were purified with the PureLink PCR purification kit (Life Technologies, Carlsbad, Calif). PCR fragments were sequenced with an ABI-Prism BigDye terminator kit (Applied Cancer Cytopathology
August 2015
Biosystems, Foster City, Calif) on a 3130 genetic analyzer (Applied Biosystems). Sequences were evaluated with sequencing analysis software (Applied Biosystems). Stem-Loop Reverse Transcription (RT) and Real-Time PCR
Total RNA was reverse-transcribed with Revert Aid reverse transcriptase (Thermo Fisher Scientific, Waltham, Mass) and a 10-mM stem-loop RT primer. U6 small nuclear RNA was employed as a reference for miRNA expression. Stem-loop primers for human miR-221 (TCGTATCCAGTGCAGGGTCCGAGGTATTCGC (ACTGGATACGACGAAACCC) and human miR-222 (GTCGTATCCAGTGCAGGGTCCGAGGTATTCG CACTGGATACGACGAGACC) and an RT primer for U6 snRNA as an endogenous control (AAAATATGGAA CGCTTCACGAATTTG) were used for multiplex RT reactions under the following conditions: 30 minutes at 168C, 60 minutes at 428C, 15 minutes at 708C, and 5 minutes at 48C. Complementary DNA, obtained with stem-loop RT, was quantified by real-time PCR with a Kapa Sybr fast quantitative PCR kit (Kapa Biosystems, Wilmington, Mass) on an ABI-Prism 7500HT real-time PCR system (Applied Biosystems). The reaction was performed with forward and reverse primers for miR-221 (AGCTAC ATTGTCTGCTGG and GTATCCAGTG CAGGGTCC), miR-222 (AGCTACATCTGGCTA CTG and GTATCCAGTGCAGGGTCC), and U6 snRNA (CTCGCTTCGGCAGCACATATACT and ACGCTTCACGAATTTGCGTGTC). The reactions were run under the following conditions: 958C for 20 seconds and then 40 cycles at 958C for 3 seconds and at 568C for 1 minute. The level of miRNA transcripts was normalized to the level of the U6 snRNA transcripts and was quantified by the cycle threshold method. All quantitative RT-PCR was performed in duplicate and repeated at least twice. Data Assist software (Applied Biosystems) was used for relative quantification. Galectin-3 Immunocytochemistry
FNA samples were deposited onto slides by centrifugation at 250 rpm for 5 minutes. Cytospin-prepared slides were first stained with hematoxylin-eosin for cytological evaluation and then destained and processed for immunocytochemistry. For destaining, the slides were first immersed in xylene to remove traces of the mounting medium and 473
Original Article TABLE 1. Summary of BRAF V600E Mutations and Galectin-3 Immunostaining
BRAF V600E Wild type Mutant Galectin-3 Negative Positive
Benign
Malignant
Total
74 0
39 6
113 6
42 16
9 31
51 47
then rehydrated in ethanol. Hematoxylin was removed by the incubation of the slides in 1% HCl for 1 minute, and eosin was removed by incubation in 1.5% NH4OH for 1 minute. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide, and nonspecific binding of antibodies was blocked in 5% casein. Cytology slides were incubated with galectin-3 antibody (clone 194804; R&D Systems, Minneapolis, Minn) overnight at 48C. A streptavidin-biotin peroxidase detection system was used in accordance with the manufacturer’s instructions and developed with 3,30 -diaminobenzidine (Vector Laboratories, Burlingame, Calif). Sections were counterstained with 1% hematoxylin. Incubation without a primary antibody or incubation with an immunoglobulin G isotype was used as a negative control, and they exhibited no background staining. As positive controls, we used tissue sections from human papillary thyroid cancer previously shown to be galectin-3–positive by staining with the same antibody (clone 194804; R&D Systems) in our laboratory. In addition, positively stained macrophages were used as positive internal controls in cytospin preparations. A pathologist evaluated only galectin-3 cytoplasmic expression and used positive and negative staining scores. Having in mind that the cell number differs in FNA samples, we always evaluated at least 6 groups of cells and estimated thyrocyte staining as a percentage. Scoring was considered positive when more than 10% of thyrocytes were immunostained in accordance with Sapio et al.21 Statistical Analysis
Statistical Package for the Social Sciences (SPSS) software was used for the statistical analysis. An analysis of variance and a t test were used to determine the significance between groups. Receiver operating characteristic curves were employed to analyze the diagnostic utility of different markers. To evaluate the performance of diagnostic markers in combination, the area under the curve was determined. Sensitivity, specificity, positive and negative 474
predictive values, and diagnostic accuracy were calculated on the basis of the histological diagnosis with the SPSS software. Two-sided P values less than .05 were considered to indicate statistical significance. A classification model was built with the IBM SPSS Modeler program. Samples were classified as benign or malignant with a C5.0 algorithm.
RESULTS The BRAF V600E Mutation Was Detected in Malignant FNA Only
The samples were tested for BRAF mutations via PCR amplification with specific primers. The PCR products were sequenced. We detected a V600E mutation in 6 cases, all of which were malignant according to the final histopathological diagnosis: 4 PTC cases and 2 ATC cases (Table 1). None of the patients carried the K601E mutation. Up-Regulation of miR-221 and miR-222 in Thyroid Nodules With a Malignant Diagnosis
To evaluate miR-221/miR-222 expression in different FNA samples, we set up a real-time RT-PCR assay that allowed us to detect both miRNAs with high efficiency and specificity. After normalization to U6 small nuclear RNA, we sorted the miRNA expression data according to the final histopathological analysis into benign and malignant groups (Fig. 1). The expression of miRNAs demonstrated an increase in both miR-221 and miR-222 in malignant cases. Because of the high variability of expression, there was no statistically significant difference in the expression of miRNAs between benign and malignant patient groups. We decided to set cutoff points for miRNA expression and divided the patients into high- and low-expression groups. The value at which the sum of the sensitivity and the specificity was the highest was chosen as the cutoff point for miR-221/miR-222. Figure 1B demonstrates a greater participation of high miRNA expression in malignant cases and statistically significant high/low expression ratios between benign and malignant patient groups. Galectin-3 Expression Was Upregulated in Malignant Cases
The results of immunocytochemical staining are presented in Figure 2. FTAs (Fig. 2A) had small, regular Cancer Cytopathology
August 2015
Malignant Risk Stratification/Paska s et al
Figure 1. Expression of miR-221 and miR-222 in FNA samples. (A) Ct values have been normalized to U6 snRNA. Empty circles represent the microRNA values for each patient, and the lines represent group mean values. (B) Percentages of cases with high and low microRNA expression in B and M patient groups. *P
Malignant Risk Stratification of Thyroid FNA Specimens With Indeterminate Cytology Based on Molecular Testing ´ , MSc1; Vladan Zivaljevic ´, MD, PhD2; Svetislav Tatic ´ , MD, PhD3; Svetlana Paskas, PhD1; Jelena Jankovic 4 5 5 ic ´, PhD ; Dragica Radojkovic ´ , PhD ; Svetlana Savin, PhD1; ´ , MD ; Aleksandra Nikolic Vesna Boz 1 ´, PhD and Dubravka Cvejic
BACKGROUND: Fine-needle aspiration (FNA) has been employed for many years for examining thyroid nodules, and the cytology of aspirates is the primary determinant for whether thyroidectomy is indicated. Fifteen to thirty percent of thyroid nodules, not being clearly benign or malignant, fall into an indeterminate category. The main goals of molecular diagnostics for thyroid nodules are to prevent unnecessary surgery in patients with benign nodules and to stop patients with malignant nodules from being subjected to repeated operations. METHODS: This study was designed to evaluate the diagnostic utility of 4 markers in thyroid FNA cytology via testing for the BRAF V600E mutation and the expression of microRNA-221, microRNA-222, and galectin-3 protein in FNA samples with indeterminate cytology. RESULTS: A predictor model distinguishing benign samples from malignant samples on the basis of the 4 aforementioned markers was formulated. This decision model provided a sensitivity of 73.5%, a specificity of 89.8%, and a diagnostic accuracy of 75.7%. The positive predictive value was 80.6%, and the negative predictive value was 85.5%; this suggested that the prediction had good reliability. CONCLUSIONS: One hundred twenty FNA samples were examined, and 62 nodules were classified as benign with the proposed diagnostic algorithm. This resulted in a reduction of the initial 120 patients to 58 and thus C 2015 decreased by half the number of persons undergoing surgery. Cancer (Cancer Cytopathol) 2015;123:471-9. V
American Cancer Society. KEY WORDS: BRAF; fine-needle aspirates; galectin-3; indeterminate cytology; microRNA 221 (miR-221); microRNA 222 (miR-222); thyroid cancer.
INTRODUCTION Thyroid cancer typically manifests as a thyroid nodule. Palpable nodules are found in approximately 4% to 7% of the population, whereas those that can be detected by sonography are present in 19% to 35% of the population, and nodules are discovered at autopsy in 8% to 65%.1 The incidence of malignancy is up to 13% for solitary nodules and up to 10% for multinodular goiters.2–4 Fine-needle aspiration (FNA) has been employed for many years to examine thyroid nodules. Thus, the cytology of aspirates is the primary determinant for whether thyroidectomy is indicated. In terms of decision making, the most problematic categories are the indeterminate ones (Bethesda categories III, IV, and V).5 Roughly 15% to 30% of thyroid nodules, not being clearly benign or malignant, fall into an indeterminate category. The risk of malignancy for these categories is 5% to 75%.5
Corresponding author: Svetlana Paskas, PhD, Institute for the Application of Nuclear Energy, Department of Endocrinology and Radioimmunology, University of Belgrade, Banatska 31b, 11080 Belgrade, Serbia; Fax: (011) 381 11 2618 724; [email protected] 1 Institute for the Application of Nuclear Energy, Department of Endocrinology and Radioimmunology, University of Belgrade, Belgrade, Serbia; 2Center for Endocrine Surgery, Institute of Endocrinology, Diabetes, and Diseases of Metabolism, University of Belgrade, Belgrade, Serbia; 3Institute of Pathology, Medical Faculty, University of Belgrade, Belgrade, Serbia; 4Department of Endocrine and Cardiovascular Pathology, Clinical Center of Serbia, Belgrade, Serbia; 5Institute of Molecular Genetics and Genetic Engineering, University of Belgrade, Belgrade, Serbia
Received: January 2, 2015; Revised: April 2, 2015; Accepted: April 2, 2015 Published online April 27, 2015 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/cncy.21554, wileyonlinelibrary.com
Cancer Cytopathology
August 2015
471
Original Article
The main goals of molecular diagnostics for thyroid nodules are to prevent unnecessary surgery in patients with benign nodules and to stop patients with malignant nodules from being subjected to repeated operations. Moreover, in cases of indeterminate cytology, the risk of watchful waiting might be too high for most patients.6 Performing molecular analyses such as somatic mutation testing, gene and microRNA (miRNA) expression, and immunocytochemistry enables patients to be separated into low- and high-risk categories. In this study, we evaluated a set of 4 markers and investigated their diagnostic potential for distinguishing benign thyroid nodules from malignant ones. We tested for the BRAF V600E mutation and the expression of microRNA-221 (miR-221), microRNA-222 (miR-222), and galectin-3 protein. A BRAF mutation is one of the most prevalent somatic genetic events in human cancer and has the hallmark of a conventional oncogene.7 The most common BRAF mutation found in thyroid cancer is V600E, which constitutes 95% of all known BRAF mutations.8 This mutation leads to constitutive activation of a BRAF kinase and is the most common genetic change in papillary thyroid carcinoma (PTC).8–10 It is also found in poorly differentiated and anaplastic carcinomas, and this suggests that these cancers are PTC-derived.11 Other rare BRAF mutations, such as K601E and small in-frame insertions or deletions surrounding codon 600, have been described,12 but they are present in a minority of papillary carcinomas (1%-2%).13 Somatic mutation tests have high specificity but poor sensitivity for differentiating benign indeterminate thyroid lesions from malignant ones. In addition, approximately 30% of thyroid cancers harbor none of the mutations tested in the aforementioned panels. Therefore, when the test is mutation-negative, an overall malignancy risk is still present. To expand the search for an optimal panel of markers, we introduced 2 miRNAs, miR-221 and miR222, already shown to distinguish PTC from benign tumors. When the expression levels of miRNAs in PTC were compared, miR-221 and miR-222 were among the 10 most upregulated miRNAs in PTC.14 miRNAs are suitable biomarkers because their expression is often dysregulated in cancer, and they are more stable than other RNAs; this makes isolation and quantification by quantitative polymerase chain reaction (PCR) more feasible.15 472
Several protein-based biomarkers, alone or combined in panels, have been used for the diagnosis of thyroid cancer by immunohistochemistry. Among the most promising, galectin-3 has been studied in normal and pathological thyroid tissue. Galectin-3 is one of the nonintegrin, b-galactoside–binding lectins. It has an affinity for carcinoembryonic antigen, immunoglobulin E, laminin, and other mucins.16 Overexpression of galectin-3 is suitable for the distinction of malignant lesions from benign ones.17,18 Galectin-3 expression is a complementary and useful diagnostic method for indeterminate thyroid FNA samples.19,20 The aim of this study was to establish whether such a combination of selected markers could be helpful in distinguishing benign tumors from malignant tumors and thus in reducing the number of thyroidectomies for benign follicular lesions. To achieve this, we developed an algorithm for the management of patients with indeterminate FNA based on the results of the combination of these 4 markers. The goal of this study was not to validate the role of each individual molecular marker or to decide which of the 4 markers is redundant because they reflect malignancy at different levels (ie, DNA, RNA, and protein).
MATERIALS AND METHODS Patients and FNA Samples
FNA samples were obtained from patients with ultrasonographically confirmed thyroid nodules. All clinical specimens examined in this study were approved by the Ethics Committee at the Center for Endocrine Surgery, Clinical Center of Serbia, Belgrade. Informed consent from patients was provided for all collected samples. These were aspirated preoperatively with a 25-gauge needle and transferred directly into 400 mL of RNAlater stabilization solution (Ambion; Life Technologies). On average, 3 to 5 FNA passes were performed for each thyroid nodule. Half of the aspirate was used for cytospin preparation for hematoxylin-eosin staining, and the other half was used for nucleic acid isolation. A cytopathologist reviewed all FNA specimens before surgery. The Bethesda criteria for the adequacy of FNA samples were adopted when this was possible: 6 groups with at least 10 cells per group were counted. Specimens that contained abundant colloid were considered adequate without the strict application of proposed numerical requirements. In addition, a specimen was considered satisfactory if the contamination of Cancer Cytopathology
August 2015
Malignant Risk Stratification/Paska s et al
thyrocytes with blood cells was less than 20% according to hematoxylin-eosin staining. The cytological reports were collected, and samples without a definitive benign or malignant local diagnosis, classified as diagnostic category III, IV, or V according to the Bethesda System for Reporting Thyroid Cytopathology, were taken for molecular testing. There were 56 patients in the atypia of undetermined significance/follicular lesion of undetermined significance category, 41 patients in the follicular neoplasia category, and 23 patients in the suspicious for malignancy category. In this way, we selected 120 patients with indeterminate cytology. The cytological diagnosis was established before molecular testing. An endocrine pathologist reviewed all surgery specimens, and the final histopathological diagnosis was used as a reference for molecular analysis. According to the final histopathological diagnosis, among 120 patients with indeterminate cytology, there were 30 diagnosed with PTC, 1 diagnosed with follicular thyroid carcinoma, 8 diagnosed with H€ urtle thyroid carcinoma (HTC), 2 diagnosed with anaplastic thyroid carcinoma (ATC), and 4 diagnosed with medullary thyroid carcinoma (MTC). The remaining patients had a benign diagnosis: 35 with goiters, 37 with follicular thyroid adenoma (FTA), and 3 with H€ urtle thyroid adenoma (HTA). Nucleic Acid Isolation
Total nucleic acids were extracted with the TRIzol reagent (Invitrogen, Carlsbad, Calif) according to the manufacturer’s instructions. The quantity of isolated DNA and RNA was assessed with a NanoVue spectrophotometer (GE Healthcare, Buckinghamshire, United Kingdom). Mutation Analysis
Samples were screened for the presence of BRAF point mutations V600E and K601E. BRAF exon 15 was amplified by standard PCR with forward primer TCATAATGCTTGCTCTGATAGGA and reverse primer GGCCAAAAATTTAATCAGT GGA. The amplification was performed under the following conditions: initial denaturation at 948C for 2 minutes; 40 cycles at 948C for 30 seconds, at 588C for 30 seconds, and at 728C for 30 seconds; and final elongation at 728C for 5 minutes. The obtained PCR products were purified with the PureLink PCR purification kit (Life Technologies, Carlsbad, Calif). PCR fragments were sequenced with an ABI-Prism BigDye terminator kit (Applied Cancer Cytopathology
August 2015
Biosystems, Foster City, Calif) on a 3130 genetic analyzer (Applied Biosystems). Sequences were evaluated with sequencing analysis software (Applied Biosystems). Stem-Loop Reverse Transcription (RT) and Real-Time PCR
Total RNA was reverse-transcribed with Revert Aid reverse transcriptase (Thermo Fisher Scientific, Waltham, Mass) and a 10-mM stem-loop RT primer. U6 small nuclear RNA was employed as a reference for miRNA expression. Stem-loop primers for human miR-221 (TCGTATCCAGTGCAGGGTCCGAGGTATTCGC (ACTGGATACGACGAAACCC) and human miR-222 (GTCGTATCCAGTGCAGGGTCCGAGGTATTCG CACTGGATACGACGAGACC) and an RT primer for U6 snRNA as an endogenous control (AAAATATGGAA CGCTTCACGAATTTG) were used for multiplex RT reactions under the following conditions: 30 minutes at 168C, 60 minutes at 428C, 15 minutes at 708C, and 5 minutes at 48C. Complementary DNA, obtained with stem-loop RT, was quantified by real-time PCR with a Kapa Sybr fast quantitative PCR kit (Kapa Biosystems, Wilmington, Mass) on an ABI-Prism 7500HT real-time PCR system (Applied Biosystems). The reaction was performed with forward and reverse primers for miR-221 (AGCTAC ATTGTCTGCTGG and GTATCCAGTG CAGGGTCC), miR-222 (AGCTACATCTGGCTA CTG and GTATCCAGTGCAGGGTCC), and U6 snRNA (CTCGCTTCGGCAGCACATATACT and ACGCTTCACGAATTTGCGTGTC). The reactions were run under the following conditions: 958C for 20 seconds and then 40 cycles at 958C for 3 seconds and at 568C for 1 minute. The level of miRNA transcripts was normalized to the level of the U6 snRNA transcripts and was quantified by the cycle threshold method. All quantitative RT-PCR was performed in duplicate and repeated at least twice. Data Assist software (Applied Biosystems) was used for relative quantification. Galectin-3 Immunocytochemistry
FNA samples were deposited onto slides by centrifugation at 250 rpm for 5 minutes. Cytospin-prepared slides were first stained with hematoxylin-eosin for cytological evaluation and then destained and processed for immunocytochemistry. For destaining, the slides were first immersed in xylene to remove traces of the mounting medium and 473
Original Article TABLE 1. Summary of BRAF V600E Mutations and Galectin-3 Immunostaining
BRAF V600E Wild type Mutant Galectin-3 Negative Positive
Benign
Malignant
Total
74 0
39 6
113 6
42 16
9 31
51 47
then rehydrated in ethanol. Hematoxylin was removed by the incubation of the slides in 1% HCl for 1 minute, and eosin was removed by incubation in 1.5% NH4OH for 1 minute. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide, and nonspecific binding of antibodies was blocked in 5% casein. Cytology slides were incubated with galectin-3 antibody (clone 194804; R&D Systems, Minneapolis, Minn) overnight at 48C. A streptavidin-biotin peroxidase detection system was used in accordance with the manufacturer’s instructions and developed with 3,30 -diaminobenzidine (Vector Laboratories, Burlingame, Calif). Sections were counterstained with 1% hematoxylin. Incubation without a primary antibody or incubation with an immunoglobulin G isotype was used as a negative control, and they exhibited no background staining. As positive controls, we used tissue sections from human papillary thyroid cancer previously shown to be galectin-3–positive by staining with the same antibody (clone 194804; R&D Systems) in our laboratory. In addition, positively stained macrophages were used as positive internal controls in cytospin preparations. A pathologist evaluated only galectin-3 cytoplasmic expression and used positive and negative staining scores. Having in mind that the cell number differs in FNA samples, we always evaluated at least 6 groups of cells and estimated thyrocyte staining as a percentage. Scoring was considered positive when more than 10% of thyrocytes were immunostained in accordance with Sapio et al.21 Statistical Analysis
Statistical Package for the Social Sciences (SPSS) software was used for the statistical analysis. An analysis of variance and a t test were used to determine the significance between groups. Receiver operating characteristic curves were employed to analyze the diagnostic utility of different markers. To evaluate the performance of diagnostic markers in combination, the area under the curve was determined. Sensitivity, specificity, positive and negative 474
predictive values, and diagnostic accuracy were calculated on the basis of the histological diagnosis with the SPSS software. Two-sided P values less than .05 were considered to indicate statistical significance. A classification model was built with the IBM SPSS Modeler program. Samples were classified as benign or malignant with a C5.0 algorithm.
RESULTS The BRAF V600E Mutation Was Detected in Malignant FNA Only
The samples were tested for BRAF mutations via PCR amplification with specific primers. The PCR products were sequenced. We detected a V600E mutation in 6 cases, all of which were malignant according to the final histopathological diagnosis: 4 PTC cases and 2 ATC cases (Table 1). None of the patients carried the K601E mutation. Up-Regulation of miR-221 and miR-222 in Thyroid Nodules With a Malignant Diagnosis
To evaluate miR-221/miR-222 expression in different FNA samples, we set up a real-time RT-PCR assay that allowed us to detect both miRNAs with high efficiency and specificity. After normalization to U6 small nuclear RNA, we sorted the miRNA expression data according to the final histopathological analysis into benign and malignant groups (Fig. 1). The expression of miRNAs demonstrated an increase in both miR-221 and miR-222 in malignant cases. Because of the high variability of expression, there was no statistically significant difference in the expression of miRNAs between benign and malignant patient groups. We decided to set cutoff points for miRNA expression and divided the patients into high- and low-expression groups. The value at which the sum of the sensitivity and the specificity was the highest was chosen as the cutoff point for miR-221/miR-222. Figure 1B demonstrates a greater participation of high miRNA expression in malignant cases and statistically significant high/low expression ratios between benign and malignant patient groups. Galectin-3 Expression Was Upregulated in Malignant Cases
The results of immunocytochemical staining are presented in Figure 2. FTAs (Fig. 2A) had small, regular Cancer Cytopathology
August 2015
Malignant Risk Stratification/Paska s et al
Figure 1. Expression of miR-221 and miR-222 in FNA samples. (A) Ct values have been normalized to U6 snRNA. Empty circles represent the microRNA values for each patient, and the lines represent group mean values. (B) Percentages of cases with high and low microRNA expression in B and M patient groups. *P
