Upregulation of SLC2 (GLUT) family genes is related to poor survival outcomes in papillary thyroid carcinoma: Analysis of data from The Cancer Genome Atlas

ARTICLE IN PRESS

Upregulation of SLC2 (GLUT) family genes is related to poor survival outcomes in papillary thyroid carcinoma: Analysis of data from The Cancer Genome Atlas Young Jun Chai, MD,a,b Jin Wook Yi, MD,b,c So Won Oh, MD, PhD,d Young A Kim, MD, PhD,e Ka Hee Yi, MD, PhD,f Ju Han Kim, MD, PhD,g and Kyu Eun Lee, MD, PhD,b,c Seoul, Korea

Background. The Warburg effect describes increased glucose uptake in cancer cells, and glucose transporter proteins are overexpressed in many tumors. In this study, we evaluated the expression of 14 SLC2A genes encoding glucose transporter proteins in papillary thyroid carcinoma patients. Methods. Clinical information and gene expression data from 499 papillary thyroid carcinoma patients were downloaded from The Cancer Genome Atlas database. Correlations between SLC2 gene family (SLC2A1–14) mRNA expression levels and clinicopathologic factors were analyzed. Results. There were 14 mortalities during follow-up (median, 21.6 months). Patient overall mortality was associated with age $45 years, extrathyroidal extension, higher TNM stage, and increased expression of SLC2A1, SLC2A3, and SLC2A14 mRNA. Greater SLC2A1, SLC2A3, and SLC2A14 expression was associated with increased mortality (odds ratio: 11.692, 95% confidence interval: 3.362–36.938; odds ratio: 12.725, 95% confidence interval: 4.247–40.187; and odds ratio: 13.768, 95% confidence interval: 4.208–61.710, respectively). Kaplan–Meier survival analysis indicated that overall survival was shorter in patients with high rather than low SCL2 expression (SLC2A1, P = .003; SLC2A3, P < .001; and SLC2A14, P < .001). Conclusion. Upregulation of the SLC2A1, SLC2A3, and SLC2A14 genes was associated with increased mortality in papillary thyroid carcinoma patients, and SLC2 gene expression levels are potentially useful prognostic indicators. (Surgery 2016;j:j-j.) From the Department of Surgery,a Seoul National University Boramae Medical Center, the Cancer Research Instituteb and the Department of Surgery,c Seoul National University Hospital and College of Medicine, the Department of Nuclear Medicine,d Department of Pathology,e and Department of Internal Medicine,f Seoul National University Boramae Medical Center, and the Division of Biomedical Informatics,g Systems Biomedical Informatics Research Center, Seoul National University College of Medicine, Seoul, Korea

Y.J.C. and J.W.Y. contributed equally to this study. Supported by a multidisciplinary research grant-in-aid from the Seoul Metropolitan Government Seoul National University Boramae Medical Center (02-2015-8), and the Korean Health Technology R&D Project, Ministry of Health and Welfare (HI13C2164). Presented at the 37th Annual Meeting of the American Association of Endocrine Surgeons, April 10–12, 2016, Baltimore, MD. Accepted for publication April 10, 2016. Reprint requests: Kyu Eun Lee, MD, PhD, Seoul National University Hospital & College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul, 110–744, Korea. E-mail: kyu.eun.lee.md@ snu.ac.kr. 0039-6060/$ - see front matter Ó 2016 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2016.04.050

UNLIKE NORMAL SOMATIC CELLS, which oxidize glucose to produce energy, rapidly proliferating cancer cells obtain energy by glycolysis in preference to oxidation, even when oxygen is available.1 This phenomenon, known as the Warburg effect or aerobic glycolysis, provides the theoretical basis for the use of 18F-fluorodeoxy-glucose positron emission tomography (FDG-PET) to detect highly proliferating malignant tumors. Excessive glucose influx is required for cancer cell proliferation and is facilitated by glucose transporter (GLUT) proteins. GLUT proteins are categorized into 3 classes based on sequence similarity: Class 1 comprises GLUT proteins 1–4 and 14; Class 2 comprises GLUT proteins 5, 7, 9, and 11; and Class 3 comprises GLUT proteins 6, 8, 10, 12, and 13.2 They are encoded by the solute SURGERY 1

ARTICLE IN PRESS 2 Chai et al

carrier family 2 (SLC2) genes, with 14 GLUT proteins (GLUT1–14) corresponding to 14 SLC2 genes (SCL2A1–14).3 Among the 14 GLUT subtypes, GLUT1 has been most widely investigated in various cancers. In breast cancer, GLUT1 expression is associated with aggressive tumor grade and decreased survival.4,5 The GLUT1 protein is highly expressed in invasive ovarian carcinoma and fallopian tube adenocarcinoma,6 and associations with hepatocellular carcinoma and non–small cell lung cancer have also been reported.7,8 In thyroid carcinoma, the association between GLUT expression and prognosis remains controversial. There are studies reporting that overexpression of the SLC2A1 gene is related to adverse prognostic factors, such as poor differentiation and advanced tumor stage9,10; however, another immunohistochemical study reported no association between GLUT1 protein expression and clinicopathologic characteristics in thyroid carcinoma patients.11 These previous studies were limited by small sample sizes and part of the results was based on subjective interpretation of immunohistochemical staining. Moreover, only a few of the SLC2 genes or GLUT proteins were evaluated in the previous studies. In this study, we analyzed mRNA expression of all 14 SLC2 genes using data from The Cancer Genome Atlas (TCGA) database and investigated associations with survival outcomes in papillary thyroid carcinoma (PTC), thus evaluating the prognostic value of SLC2 gene expression levels in PTC. PATIENTS AND METHODS Data acquisition. From the TCGA data portal (https://tcga-data.nci.nih.gov/tcga/tcgaDownload. jsp), we downloaded mRNA expression counts for SLC2 family genes, the data on somatic mutation, such as BRAF and RAS, and corresponding clinical information. The patient information is anonymized and deidentified. According to TCGA publication guidelines (http://cancergenome.nih.gov/ publications/publicationguidelines), these somatic mutation and mRNA sequencing data have no restrictions on publication and no specific permission is needed for investigators to publish using these data. Somatic mutation data were provided as a mutation call file by the Broad Institute and the Baylor College of Medicine. The Illumina Genome Analyzer was used as the platform for DNA sequencing (Illumina Inc, San Diego, CA). RNA sequencing data were obtained by Illumina HiSeq

Surgery j 2016

Table I. Association between the clinicopathologic characteristics and overall survival status Variables Age <45 0 $45 14 Sex Female 9 Male 5 Thyroiditis Absent 14 Present 0 Tumor size #2 cm 3 >2 cm 11 Extrathyroid extension No 6 Yes 8 AJCC TNM Stage I/II 2 III/IV 12 BRAF status Wild-type 5 V600E 9 mutation Other than 0 V600E mutation RAS status Wild-type 14 Mutation 0

Alive

P value

(0%) (100%)

228 (47.0%) 257 (53.0%)

<.001

(64.3%) (35.7%)

356 (73.4%) 129 (26.6%)

.540

(100%) (0%)

319 (65.8%) 166 (34.2%)

.019

(21.4%) (78.6%)

197 (40.6%) 288 (59.4%)

.176

(42.9%) (57.1%)

336 (69.3%) 149 (30.7%)

.044

(14.3%) (85.7%)

331 (68.2%) 154 (31.8%)

<.001

(35.7%) (64.3%)

253 (52.2%) 227 (46.8%)

.375

Dead

(0%)

(100%) (0%)

5 (1.0%)

432 (89.1%) 53 (10.1%)

.381

2000 RNA Sequencing Version 2 analysis and provided by the University of North Carolina. Expression counts of mRNA were expressed as RNA-Seq by Expectation Maximization values. Two authors (YJC and JWY) independently reviewed every original scanned pathology report and revised incorrect or missing clinical information. Patient information. We collected mRNA expression values of SLC2 family genes for each patient, along with their clinicopathologic characteristics, including age, sex, histologic subtype, thyroiditis, tumor size, extrathyroidal extension (ETE), American Joint Committee on Cancer (AJCC) stage, BRAF and RAS mutation status, and overall survival (Supplemental Table I). Follicular and tall cell variants of PTC were diagnosed when tumors exhibited >99% follicular pattern or contained >50% tall cells, respectively.12 Statistical analyses. All statistical analyses and plots were generated using R software (version 3.0.2; R Foundation for Statistical Computing, Vienna, Austria). For categorical variables, the Fisher exact or v2 tests were applied according to

ARTICLE IN PRESS Chai et al 3

Surgery Volume j, Number j

Table II. Summary of the correlation between the clinicopathologic characteristics and mRNA expression count of SLC2 family genes

Age ($45) SLC2A1 SLC2A2 SLC2A3 SLC2A4 SLC2A5 SLC2A6 SLC2A7 SLC2A8 SLC2A9 SLC2A10 SLC2A11 SLC2A12 SLC2A13 SLC2A14

Male sex

FVPTC

Thyroiditis

Size >2 cm







+++

+ ++

+

ETE

AJCC TNM stage (III/IV)

BRAFV600E mutation

RAS mutation

Mortality

++

++

+++






++

+++ +++

+ +

+++







+++

++ +++

+++

+++ +++


+ +++ +

++
















++ +++





+++ +++











+




+++






+++

+, positive correlation with P value < .05; ++, positive correlation with P value < .01; +++, positive correlation with P value < .001.
, negative correlation with P value < .05;

, negative correlation with P value < .01;


, negative correlation with P value < .001. FVPTC, Follicular variant papillary thyroid carcinoma.

Table III. Linear regression analysis between clinicopathologic characteristics and mRNA expression count of SLC2A1, SLC2A3, and SLC2A14 genes SLC2A1

SLC2A3

SLC2A14

Variables

t value

P value

t value

P value

t value

P value

Age $45 Male sex FVPTC Thyroiditis Size >2 cm Extrathyroid extension AJCC stage (III/IV) BRAFV600E mutation RAS mutation Mortality

0.278 1.32
6.86
1.020 1.107 3.214 3.271 8.602
4.004 2.804

.781 .187 <.001 .308 .269 .001 .001 <.001 <.001 .005


1.396
0.791
7.738
1.520 2.258 5.665 2.235 7.266
5.923 4.963

.163 .429 <.001 .129 .024 <.001 .026 <.001 <.001 <.001


1.007
1.258
7.086
1.461 1.851 5.755 2.418 7.504
5.240 4.541

.315 .209 <.001 .145 .065 <.001 .016 <.001 <.001 <.001

t values indicate the correlation between the 2 variables. FVPTC, Follicular variant papillary thyroid carcinoma.

sample size. Linear regression was used to evaluate associations between clinical variables and SLC2 gene expression counts. For correlation coefficients, t values were extracted. To determine the optimal cut-off value of SLC2 gene expression for prediction of survival, receiver operating characteristic curves and regression tree–based classification analyses were applied. Gene expression counts were transformed into natural logs for determination of cut-off values. Logistic regression and Kaplan-Meier survival analyses were performed based on calculated cut-off values.

RESULTS A total of 499 PTC cases were included in this study, consisting of 355 classical type PTC and 101 follicular, 35 tall cell, 4 diffuse sclerosing, 2 cribriform morular, 1 columnar cell, and 1 mixed papillary and follicular variant case. The median follow-up was 21.6 months (range, 0.03– 171.7 months). Among clinicopathologic characteristics, mortality was associated with age $45, absence of thyroiditis, ETE, and AJCC stage III/IV (Table I).

ARTICLE IN PRESS 4 Chai et al

Surgery j 2016

Table IV. Correlation between mortality and mRNA expression count of SLC2A1, SLC2A3, and SLC2A14 genes AUC* Cut-off valuey Below cut-off Above cut-off Logistic regression Odds ratio 95% confidence interval P value

SLC2A1

SLC2A3

SLC2A14

0.677 7.471 472 27

0.760 7.840 445 54

0.763 3.629 386 113

11.692 3.362–36.938 <.001

12.725 4.247–40.187 <.001

13.768 4.208–61.710 <.001

*Area Under the Receiver Operating Characteristic Curve. yGene expression counts are represented by the values of the natural log.

Fig. Kaplan-Meier survival analysis of 499 papillary thyroid carcinoma patients stratified by (A) SLC2A1, (B) SLC2A3, and (C) SLC2A14 gene expression levels.

Table II shows correlations between clinicopathologic characteristics and mRNA expression values of SLC2 family genes. AJCC TNM stage III/IV was associated with expression of SLC2A1, SLC2A3, SLC2A4, and SLC2A14 (details in Supplemental Table II), and mortality was associated with SLC2A1, SLC2A3, and SLC2A14 expression. Linear regression analysis demonstrated that mortality was significantly associated with greater mRNA expression of SLC2A1 (t = 2.084, P = .005), SLC2A3 (t = 4.963, P < .001), and SLC2A14 (t = 4.541, P < .001) (Table III). The expression of these 3 genes was also positively correlated with ETE, AJCC TNM stage III/IV, and BRAFV600E mutation, whereas it was inversely correlated with follicular variant PTC and RAS mutation. In survival prediction analyses using receiver operating characteristic curves, the area under the curve (AUC) values for SLC2A1, SLC2A3, and SLC2A14 were 0.677, 0.760, and 0.763, respectively (Table IV). When the cases were categorized into low and high mRNA expression groups based on cut-off values that maximized the AUC for each

gene, SLC2A1, SLC2A3, and SLC2A14 were associated with increased mortality (odds ratio [OR]: 11.692, 95% confidence interval [CI]: 3.362– 36.938; OR: 12.725, 95% CI: 4.247–40.187; and OR: 13.768, 95% CI: 4.208–61.710, respectively). Kaplan–Meier survival analyses indicated that groups with high SLC2A1, SLC2A3, and SLC2A14 mRNA expression had shorter overall survival than those with low expression (P = .003, P < .001, and P < .001, respectively) (Fig). DISCUSSION Adverse effects of GLUT overexpression on survival outcomes in PTC have been implied by studies reporting associations between GLUT1 overexpression and tumor aggressiveness or dedifferentiation.9,13,14 In addition, high glucose uptake measured by FDG-PET scan is associated with low expression of sodium-iodine symporter and decreased radioactive iodine uptake, which are indicators of dedifferentiation in thyroid tumors.15,16 Therefore, the 2015 American Thyroid Association Guidelines recommend considering FDG-PET scan

ARTICLE IN PRESS Surgery Volume j, Number j

for high-risk patients or those with poorly differentiated carcinomas.17 In this study, we demonstrated that increased mRNA expression of 3 Class 1 GLUT genes, SLC2A1, SLC2A3, and SLC2A14, is independently associated with ETE, advanced tumor stage, BRAFV600E mutation, and mortality in PTC patients. Firstly, GLUT1 is overexpressed in various human cancer tissues3 and more highly expressed in thyroid cancer than in normal thyroid tissue.18 In addition, some studies have demonstrated an association between GLUT1 overexpression and cancer dedifferentiation; SLC2A1 gene and GLUT1 protein expression levels (determined by RT-PCR and immunohistochemical staining, respectively) were greater in less differentiated tumors than in those that were well differentiated.9,14 Overexpression of the SLC2A1 gene and GLUT1 protein are also reported to be associated with advanced thyroid tumor stage and poor prognosis.10,19 GLUT3 is mainly expressed in brain and testis20; however, little is known about its expression in thyroid tumor. GLUT3 protein was not detected in the thyroid gland using immunohistochemical methods.19,20 In contrast, mRNA expression analyses demonstrated that levels of SLC2A3 were greater in thyroid cancer cells than those in normal cells,10,18 and high SLC2A3 mRNA expression is also associated with advanced thyroid tumor stage.10 Similarly, we found that SLC2A3 expression was associated with indicators of tumor aggressiveness, including tumor size >2 cm, ETE, advanced tumor stage, BRAFV600E mutation, and mortality. To date, the role of GLUT14 in tumorigenesis remains unclear, and this is the first study reporting its overexpression in advanced PTC. SLC2A14, which encodes GLUT14, shares 95% sequence homology with SLC2A3; however, SLC2A14 gene expression is largely confined to the testis, whereas GLUT3 is expressed in various tissues.21 Like SLC2A1 and SLC2A3, SLC2A14 upregulation was associated with ETE, advanced tumor stage, BRAFV600E mutation, and mortality in this study. Further investigation is required to understand the role of SLC2A14 in PTC. We also demonstrated that the expression of 3 SLC2 genes associated with mortality was independently associated with BRAFV600E mutation, a wellknown, adverse prognostic factor in PTC.22 However, BRAFV600E mutation was not associated with mortality in this study. This might be because a large sample size with long-term follow-up is essential to show a survival difference between PTC patients with BRAFV600E mutation and those with wild-type PTC,

Chai et al 5

because BRAFV600E mutation is also common in nonaggressive PTC as well as in aggressive PTC. In fact, studies with a small sample size or shortterm follow-up are unlikely to demonstrate that BRAFV600E mutation is a poor prognostic factor.23 Taken together, BRAFV600E mutation might be involved in early tumorigenesis and acts to regulate expression of genes expressed later in this process, such as SLC2 family genes. Therefore, the short follow-up period of this study (median, 21.6 months) might be insufficient to reflect the effect of the BRAFV600E mutation on survival, whereas SLC2 overexpression, which may arise late in PTC evolution, was associated with survival. A major limitation of this study is the short follow-up period, and the small number of deaths might weaken the clinical applicability of the present results. A more comprehensive analysis will be possible if TCGA data are regularly updated in the future. In addition, it could be argued that expression of SLC2 mRNAs may not correspond to that of GLUT proteins; however, a study concerning this possible discrepancy confirmed significant correlation between the overexpression of GLUT1 and GLUT3 proteins and high levels of SLC2A1 and SLC2A3 mRNA in thyroid carcinoma.10 Therefore, mRNA expression analysis can be useful to predict GLUT protein expression in PTC. Although this study proved the hypothesis that overexpression of GLUT is associated with poor prognosis in PTC, which was generated by previous studies, further validation using a different patient group or a different modality, such as a microarray or immunohistochemical assay with a large volume, might also be necessary to support the results of the current study. SLC2 expression analysis might also be clinically useful in the near future. SLC2 expression analysis of a cytologic or surgical specimen can help to select cases that are expected to have a poor prognosis. Accordingly, more aggressive treatment or more frequent follow-up for such cases might be suggested. In addition, there is a demand for new drugs for PTCs refractory to standard treatment with radioactive iodine, as the impact of tyrosine kinase inhibitors on survival is not satisfactory and their side effects are often fatal.24 Drugs inhibiting GLUT or SLC2 genes could have great therapeutic potential by inducing direct starvation of tumor cells, as demonstrated by in vitro experiments using a GLUT1 inhibitor.25 In conclusion, upregulation of the SLC2A1, SLC2A3, and SLC2A14 genes was associated with decreased overall survival in patients with PTC. SLC2 gene expression analysis is potentially useful

ARTICLE IN PRESS 6 Chai et al

for prognostication and as a marker to predict survival in PTC. SUPPLEMENTARY DATA Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.surg.2016.04.050.

REFERENCES 1. Koppenol WH, Bounds PL, Dang CV. Otto Warburg’s contributions to current concepts of cancer metabolism. Nat Rev Cancer 2011;11:325-37. 2. Mueckler M, Thorens B. The SLC2 (GLUT) family of membrane transporters. Mol Aspects Med 2013;34:121-38. 3. Yamamoto T, Seino Y, Fukumoto H, Koh G, Yano H, Inagaki N, et al. Over-expression of facilitative glucose transporter genes in human cancer. Biochem Biophys Res Commun 1990;170:223-30. 4. Ravazoula P, Batistatou A, Aletra C, Ladopoulos J, Kourounis G, Tzigounis B. Immunohistochemical expression of glucose transporter Glut1 and cyclin D1 in breast carcinomas with negative lymph nodes. Eur J Gynaecol Oncol 2003;24:544-6. 5. Kang SS, Chun YK, Hur MH, Lee HK, Kim YJ, Hong SR, et al. Clinical significance of glucose transporter 1 (GLUT1) expression in human breast carcinoma. Jpn J Cancer Res 2002;93:1123-8. 6. Rudlowski C, Moser M, Becker AJ, Rath W, Buttner R, Schroder W, et al. GLUT1 mRNA and protein expression in ovarian borderline tumors and cancer. Oncology 2004; 66:404-10. 7. Amann T, Maegdefrau U, Hartmann A, Agaimy A, Marienhagen J, Weiss TS, et al. GLUT1 expression is increased in hepatocellular carcinoma and promotes tumorigenesis. Am J Pathol 2009;174:1544-52. 8. Younes M, Brown RW, Stephenson M, Gondo M, Cagle PT. Overexpression of Glut1 and Glut3 in stage I nonsmall cell lung carcinoma is associated with poor survival. Cancer 1997;80:1046-51. 9. Kim S, Chung JK, Min HS, Kang JH, Park DJ, Jeong JM, et al. Expression patterns of glucose transporter-1 gene and thyroid specific genes in human papillary thyroid carcinoma. Nucl Med Mol Imaging 2014;48:91-7. 10. Jozwiak P, Krzeslak A, Pomorski L, Lipinska A. Expression of hypoxia-related glucose transporters GLUT1 and GLUT3 in benign, malignant and non-neoplastic thyroid lesions. Mol Med Rep 2012;6:601-6. 11. Kaida H, Hiromatsu Y, Kurata S, Kawahara A, Hattori S, Taira T, et al. Relationship between clinicopathological factors and fluorine-18-fluorodeoxyglucose uptake in patients with papillary thyroid cancer. Nucl Med Commun 2011; 32:690-8. 12. Cancer Genome Atlas Research Network. Integrated genomic characterization of papillary thyroid carcinoma. Cell 2014;159:676-90. 13. Grabellus F, Nagarajah J, Bockisch A, Schmid KW, Sheu SY. Glucose transporter 1 expression, tumor proliferation, and iodine/glucose uptake in thyroid cancer with emphasis on poorly differentiated thyroid carcinoma. Clin Nucl Med 2012;37:121-7. 14. Kim YW, Do IG, Park YK. Expression of the GLUT1 glucose transporter, p63 and p53 in thyroid carcinomas. Pathol Res Pract 2006;202:759-65.

Surgery j 2016

15. Moon SH, Oh YL, Choi JY, Baek CH, Son YI, Jeong HS, et al. Comparison of 18F-fluorodeoxyglucose uptake with the expressions of glucose transporter type 1 and Na+/I- symporter in patients with untreated papillary thyroid carcinoma. Endocr Res 2013;38:77-84. 16. Chung JK, So Y, Lee JS, Choi CW, Lim SM, Lee DS, et al. Value of FDG PET in papillary thyroid carcinoma with negative 131I whole-body scan. J Nucl Med 1999;40:986-92. 17. Haugen BRM, Alexander EK, Bible KC, Doherty GM, Mandel SJ, Nikiforov YE, et al. 2015 American Thyroid Association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer; the American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer. Thyroid 2016;26:1-133. 18. Matsuzu K, Segade F, Matsuzu U, Carter A, Bowden DW, Perrier ND. Differential expression of glucose transporters in normal and pathologic thyroid tissue. Thyroid 2004;14: 806-12. 19. Sch€ onberger J, R€ uschoff J, Grimm D, Marienhagen J, R€ ummele P, Meyringer R, et al. Glucose transporter 1 gene expression is related to thyroid neoplasms with an unfavorable prognosis: an immunohistochemical study. Thyroid 2002;12:747-54. 20. Haber RS, Weinstein SP, O’Boyle E, Morgello S. Tissue distribution of the human GLUT3 glucose transporter. Endocrinology 1993;132:2538-43. 21. Wu X, Freeze HH. GLUT14, a duplicon of GLUT3, is specifically expressed in testis as alternative splice forms. Genomics 2002;80:553-7. 22. Xing M, Alzahrani AS, Carson KA, Viola D, Elisei R, Bendlova B, et al. Association between BRAF V600E mutation and mortality in patients with papillary thyroid cancer. JAMA 2013;309:1493-501. 23. Walczyk A, Kowalska A, Kowalik A, Sygut J, Wypi orkiewicz E, Chodurska R, et al. The BRAF(V600E) mutation in papillary thyroid microcarcinoma: does the mutation have an impact on clinical outcome? Clin Endocrinol (Oxf) 2014; 80:899-904. 24. Marotta V, Ramundo V, Camera L, Del Prete M, Fonti R, Esposito R, et al. Sorafenib in advanced iodine-refractory differentiated thyroid cancer: efficacy, safety and exploratory analysis of role of serum thyroglobulin and FDGPET. Clin Endocrinol (Oxf) 2013;78:760-7. 25. Ulanovskaya OA, Cui J, Kron SJ, Kozmin SA. A pairwise chemical genetic screen identifies new inhibitors of glucose transport. Chem Biol 2011;18:222-30.

DISCUSSION Dr Sareh Parangi (Boston, MA): I was wondering if you could comment on 2 things. And these are just thoughts. I do not think they would have been part of your study. But do you think that these changes have something to do with the PET positivity of more aggressive lesions? Do you think these genes are altering the glucose uptake in more aggressive thyroid cancers and making them more likely to be PET positive? Dr Young Jun Chai: There are several studies to correlate PET positivity of more aggressive cancers to GLUT expression. Interestingly, the results are

ARTICLE IN PRESS Surgery Volume j, Number j

controversial although both PET positivity and overexpression of GLUT gene are associated with tumor aggressiveness. I think it might be because there are various types of GLUT genes and they function differently. And BRAF mutation might also play a significant role as a confounder in this correlation between PET positivity and GLUT expression. Dr John A. Olson (Baltimore, MD): Thank you for a very nice use of publicly available data to move the bar forward. I think that is a great study. My question for you is related to that. These databases are full of genes that could be examined. And in the case of a glucose transporter, that is just the beginning of the pathway that presumably is involved with the underlying biology. My question is, did you examine and did you find other

Chai et al 7

genes that their expression would be involved in the Warburg effect that you begin your idea with for a similar correlation that provides validity to your findings? Dr Young Jun Chai: As you mentioned, we also evaluated genes involved in glucose metabolism including GLUT, hexokinase, HIF, EPAS, EPO, and EPO. However, those genes had no correlation with mortality except GLUT genes. So we focused on these genes. Dr John A. Olson (Baltimore, MD): Let me push a little bit. What do you make of that? That bothered me a little bit that you do not see consistency through a pathway. Dr Young Jun Chai: As the number of samples in TCGA is limited, we cannot say conclusively. Further study should be necessary.