Identification of Biomarkers of Adrenocortical Carcinoma Using Genomewide Gene Expression Profiling

PAPER Identification of Biomarkers of Adrenocortical Carcinoma Using Genomewide Gene Expression Profiling

Gustavo G. Fernandez-Ranvier, MD; Julie Weng, BS; Ru-Fang Yeh, PhD; Elham Khanafshar, MD; Insoo Suh, MD; Christopher Barker, PhD; Quan Yang Duh, MD; Orlo H. Clark, MD; Electron Kebebew, MD

Hypothesis: The gene expression profiles of benign and Results: We found 37 genes differentially expressed by malignant adrenocortical tumors are different. 8-fold higher or lower. Fifteen genes were downregu- lated and 22 were upregulated in adrenocortical carci- Design: Genomewide gene expression profiling and vali- noma. Of the 37 genes, 29 differentially expressed by dation. microarray correlated with the gene expression levels by quantitative RT-PCR (PՅ.01). Of the 37 genes vali- Setting: Tertiary medical center. dated by RT-PCR, 22 were significantly differentially expressed between benign and malignant adrenocorti- Patients: Eighty-five patients with benign adrenocortical tumors (PϽ.05). Five of these 22 genes had an AUC cal tumors (n=74) and adrenocortical carcinoma (n=11). of 0.80 or greater (the AUC for IL13RA2 was 0.90; HTR2B, 0.87; CCNB2, 0.86; RARRES2, 0.86; and Intervention: Real-time quantitative reverse transcription–polymerase chain reaction (RT-PCR) in 89 adre- SLC16A9, 0.80), indicating high diagnostic accuracy for nocortical tissue samples (11 malignant and 78 be- distinguishing benign from malignant adrenocortical nign). The criteria for differentially expressed genes tumors. between benign and malignant adrenocortical tumors were a false discovery rate of less than 5% and an adjusted Conclusion: We identified 37 genes that are dysregu- PϽ.01. Genes differentially expressed by 8-fold higher lated in adrenocortical carcinoma, and several of the dif- or lower were validated by RT-PCR. ferentially expressed genes have excellent diagnostic accuracy for distinguishing benign from malignant Main Outcome Measures: The diagnostic accuracy adrenocortical tumors. of differentially expressed genes as determined by the area under the receiver operating characteristic curve (AUC). Arch Surg. 2008;143(9):841-846

DRENOCORTICAL TUMORS vasion or metastatic disease. However, are common, with a preva- most adrenocortical tumors are local- lence of approximately 4% ized, and there are no clinically reliable cri- in the US population. They teria to distinguish benign from malig- are discovered as a result nant localized adrenocortical tumors. ofA hormonal hypersecretion (Cushing syn- Consequently, sometimes patients are mis- drome, primary hyperaldosteronism, and diagnosed as having benign tumors based other conditions), because of local symp- on histologic examination but later de- Author Affiliations: toms (back or abdominal pain), and, most velop aggressive recurrent disease, even af- Departments of Surgery frequently, on abdominal imaging for ter initial complete resection.5 On the other (Drs Fernandez-Ranvier, Suh, another clinical indication (adrenal hand, many patients with adrenal inci- Duh, Clark, and Kebebew and incidentaloma).1,2 In contrast, adrenocor- dentaloma are subjected to adrenalec- Ms Weng), Epidemiology and tical carcinoma is rare, with an annual in- tomy based on the risk of adrenocortical Biostatistics (Dr Yeh), and cidence of 2 cases per million persons per carcinoma as estimated by tumor size, but Pathology (Dr Khanafshar), year, and it accounts for 0.2% of cancer histologic examination and long-term fol- J. David Gladstone Institutes deaths.3,4 Five-year survival for patients low-up show their tumors to be benign.6 Genomics Core Laboratory with adrenocortical carcinoma varies from Because of the clinical limitations of re- (Dr Barker), and UCSF Helen 3 Diller Family Comprehensive 32% to 45%. liably distinguishing between benign and Cancer Center (Drs Clark and It is relatively easy to distinguish be- malignant adrenocortical tumors, and to Kebebew), University of nign from malignant adrenocortical tu- gain some biological insight into the patho- California, San Francisco. mors when there is gross locoregional in- genesis of adrenocortical carcinoma, sev-

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 841

©2008 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 ing to confirm tissue diagnosis and type (adrenal cortex and me- Table 1. Clinical and Pathologic Features of Patients With dulla). Total RNA was extracted from homogenized frozen tis- Benign Adrenocortical Tumors or Adrenocortical Carcinoma sue using TRIzol reagent (Invitrogen, Carlsbad, California) and was purified using the RNeasy Mini Kit (Qiagen, Valencia, Cali- Benign Adrenocortical Adrenocortical fornia). We used 1 µg of total RNA for amplification and label- Tumor Carcinoma ing with a kit (MessageAmp aRNA Kit; Ambion Inc, Foster City, Feature (n=74) (n=11) California). Labeled and fragmented complementary RNA, 12 µg, Sex, No. (%)a was hybridized to a gene chip (Affymetrix Human Genome U133 Male 24 (34) 2 (29) Plus 2.0 GeneChip; Affymetrix Inc, Santa Clara, California) for Female 47 (66) 5 (71) 16 hours at 45°C. The gene chip arrays were stained and washed Male to female ratio 1:1.88 1:2.50 (Affymetrix Fluidics Station 400; Affymetrix Inc), according to Age at initial operation, 49.6 (11.8) [18-72] 49.7 (21.8) [18-77] the manufacturer’s protocol. The probe intensities were mea- mean (SD) [range], y sured using an argon laser confocal scanner (GeneArray Scan- Tumor size, mean 3.2 (2.0-6.5) 9.9 (3.5-20.0) ner; Hewlett-Packard, Palo Alto, California). (range), cm Functioning, No. 54 (24 with Cushing 7 (4 with virilizing syndrome and disease and REAL-TIME QUANTITATIVE RT-PCR 30 with 3 with Cushing hyperaldosteronism) syndrome) Genes differentially expressed in the microarray experiments Nonfunctioning, No. 20 4 were validated by means of real-time quantitative TaqMan (Ap- Follow-up, mean 25.7 (1-120) 25.6 (1-63) plied Biosystems, Foster City) RT-PCR in individual samples. (range), mo The same stock of total RNA used for the gene array experiments was used for the real-time quantitative RT-PCR. Total aSome patients had multiple tumor samples. RNA, 125 ng/µL, was reverse transcribed using the RT script complementary DNA synthesis kit (USB Corp, Cleveland, Ohio). eral investigators have used genomewide gene expres- Real-time quantitative PCR was used to measure messenger RNA sion profiling. Some of these studies7-10 have identified expression levels relative to glyceraldehyde-3-phosphate dehydrogenase messenger RNA expression. The gene expres- candidate diagnostic biomarkers of adrenocortical car- ⌬ ϫ cinoma and molecular profiles that distinguish between sion level is as follows: Ct=−2 (Ct of the gene of interest−Ct of glyceraldehyde-3-phosphate dehydrogenase), where Ct is the benign and malignant adrenocortical tumors. However, PCR cycle threshold. The PCR primers and probes for the genes the specific candidate genes found in these studies have were purchased from Applied Biosystems (Assay-on-Demand been discordant, possibly as a result of different ap- Kit). All the PCR experiments were performed in a final vol- proaches to data analysis. In addition, some studies have ume of 20 µL, with 1 µL of complementary DNA template on not validated microarray data using quantitative meth- a detection system (ABI PRISM 7900 Sequence Detection Sys- ods. Most important, to our knowledge, a formal analy- tem; Applied Biosystems). The PCR thermal cycler condition sis of the diagnostic accuracy of the proposed biomark- was 95°C for 12 minutes, followed by 40 cycles at 95°C for ers of adrenocortical carcinoma has not been performed. 15 seconds and 60°C for 1 minute. We, therefore, analyzed the genomewide expression profile of benign vs malignant adrenocortical tumors to iden- DATA ANALYSIS AND STATISTICAL ANALYSIS tify candidate markers. We then validated these markers using real-time quantitative reverse transcription– Raw microarray data were analyzed using the Affy package polymerase chain reaction (RT-PCR) and evaluated their (Affymetrix GeneChip Operating software; Affymetrix Inc) in diagnostic accuracy. the free statistical environment R/Bioconductor to generate an intensity value in log2 scale for each probe set using the ro- bust multiarray average method with default variables.11-13 For METHODS the class comparison (benign vs malignant), we used the limma package in R/Bioconductor to calculate the moderated t statis- TISSUE SAMPLES tics and the associated P values and the log posterior odds ratio (B statistic) that a gene is differentially expressed vs not dif- A total of 89 human adrenocortical tissue samples were ob- ferentially expressed.14 The P values were adjusted for multiple tained at surgical resection and were immediately snap frozen testing by controlling for the false discovery rate using the Ben- and stored at −80°C. Patient and tumor characteristics are sum- jamini-Hochberg method.15 marized in Table 1. Each tumor sample was reviewed again To evaluate the accuracy of candidate markers to distin- and was confirmed to be adrenocortical tissue. Approval for this guish benign from malignant adrenocortical tumors, we deter- study was obtained from the Committee on Human Research mined the area under the receiver operating characteristic curve at the University of California, San Francisco. (AUC). A stepwise regression analysis was used to determine Adrenocortical carcinoma was defined when gross local in- the AUC for the combination of markers. vasion or lymph or distant metastasis was present at diagnosis or developed during follow-up. The adrenocortical tissue samples were classified as benign if the tumor was localized at presen- RESULTS tation and there was no evidence of local or distant recurrent disease after mean follow-up of 2.1 years (range, 1-10 years). DIFFERENTIALLY EXPRESSED GENES BETWEEN BENIGN AND MALIGNANT RNA ISOLATION, PROBE PREPARATION, ADRENOCORTICAL TUMORS AND MICROARRAY HYBRIDIZATION Frozen adrenal tissue was sectioned for RNA isolation, and an Of the 89 adrenocortical tissue samples, 79 (74 cortical adjacent piece was sectioned for routine hematoxylin-eosin stain- adenoma/hyperplasia and 5 malignant) had arrays of ad-

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 842

©2008 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 equate quality for data analysis. We performed a class comparison of benign vs malignant adrenocortical carcino- Table 2. Genes Differentially Expressed Between Benign mas based on different criteria for differentially expressed and Malignant Adrenocortical Tumors by Class Comparison genes: (1) fold difference in gene expression levels and No. of Genes (2) a variety of statistical thresholds to account for mul- Statistical Criterion Differentially Expressed tiple comparisons. Table 2 summarizes the number of Ͼ ϫ differentially expressed genes based on these criteria. As 2 1930 Log posterior odds ratio (B statistic) Ͼ0 3800 expected, the number of differentially expressed genes False discovery rate Ͻ5% 9308 was lower when more stringent statistical criteria were Adjusted PϽ.01 1019 used to account for multiple comparisons. Thirty-seven genes (15 underexpressed and 22 overexpressed) were differentially expressed by 8-fold higher or lower (range, 8.0- to 24.4-fold change for underex- filtering criteria yielded 37 candidate diagnostic gene pressed genes and 8.0- to 24.2-fold change for overex- markers significantly differentially expressed between be- pressed genes), with a false discovery rate of less than nign and malignant adrenocortical carcinoma. Five of 5% and an adjusted PϽ.01. These 37 genes were vali- these genes (IL13RA2, HTR2B, CCNB2, RARRES2, and dated by means of real-time quantitative RT-PCR. Of the SLC16A9) had high diagnostic accuracy (AUC, Ն0.80). 37 genes, 29 showed strong correlation by RT-PCR gene Of the 5 genes that were the best biomarkers, IL13RA2 expression levels (PϽ.05) (Table 3). had the highest accuracy (overexpressed; AUC, 0.90). IL13RA2 overexpression has been identified in a set of DIAGNOSTIC ACCURACY genes that marks and mediates breast cancer metastasis OF DIFFERENTIALLY EXPRESSED GENES to the lungs.16 HTR2B encodes multiple receptor sub- types of serotonin neurotransmitters. Serotonin is known By real-time quantitative RT-PCR, the normalized gene to act as a growth factor for several types of nontumoral expression level of 22 genes was significantly different cells and has been proposed to contribute to cell prolif- between benign (54 adrenocortical adenoma, 20 adre- eration in aggressive tumors, such as small cell lung, pros- 17 nocortical hyperplasia, and 4 normal adrenocortical tis- tate, and colon carcinoma. On the other hand, the spe- sue) and malignant (11 adrenocortical carcinoma) tis- cific vasoconstrictive effect of serotonin or serotonin sue samples (Mann-Whitney test, PϽ.05). We determined receptor agonists might also be useful in inducing hy- the AUC for 22 of the significantly differentially ex- poxia in tumors, which could be used as a strategy using pressed genes, which ranged from 0.24 to 0.90 (Table 3). hypoxia-selective cytotoxins or hypoxia-selective gene 17 Of the 22 genes, 5 (IL13RA2 [NM_000640], HTR2B [NM therapy. CCNB2 is a member of the B-type family of cy- _000867], CCNB2 [NM_004701], RARRES2 [U77594], clins, which are essential components of the cell-cycle 18 and SLC16A9 [NM_194298, XM_166145]) had an AUC regulatory machinery. Cyclin B2 also binds to trans- ␤ of 0.80 or greater, indicating high diagnostic accuracy forming growth factor RII; thus, cyclin B2/cdc2 may ␤ for distinguishing benign from malignant adrenocorti- play a key role in transforming growth factor –medi- 18 cal tumors. The 5 genes in combination showed a slight ated cell-cycle control. This gene was also overex- 18 increase in diagnostic accuracy for distinguishing be- pressed in a recent microarray study and could be used nign from malignant tumors (Figure). We performed a as a reliable biomarker of lung adenocarcinoma. RARRES2 subset analysis excluding Conn syndrome, because these is a retinoid protein with potent growth inhibitory and 9 tumors are rarely malignant, and normal adrenocortical cell differentiation activities. SLC16A9 encodes a not well tissue samples. We found that 7 of the 22 differentially characterized protein involved in cell membrane trans- expressed genes (IL13RA2, HTR2B, CCNB2, RARRES2, portation. The potential implications of these genes in SLC16A9, ALDH1A1, and FREM2) had an AUC of 0.80 the pathogenesis of adrenocortical carcinoma need to be or greater, and the combination of the 5 highest AUCs further investigated. did not show improvement in diagnostic accuracy com- Frequent chromosomal loci of the most downregu- pared with the highest individual value (AUC, 0.913 for lated genes in adrenocortical carcinoma were chromo- IL13RA2 vs 0.907 for all 5 genes combined). Comparing somes 1, 2, 5, and 7. For upregulated genes, the most com- these markers with tumor size, one of the main current mon chromosomal loci were chromosomes 2, 8, 9, 12, clinical criterion used to assess risk of malignancy dem- 15, and 22. These results are in part consistent with those 7-10 onstrates that the differentially expressed genes are more of previous microarray studies. accurate (Figure). Among the most differentially expressed genes in adrenocortical carcinoma, we identified GIPC2 (Ͼ24-fold change), MGST1 (Ͼ22-fold change), IL13RA2 (Ͼ24- COMMENT fold change), and APOBEC3B (Ͼ17-fold change), al- though GIPC2, MGST1, and APOBEC3B were not found Because there are no reliable markers to distinguish be- to have high diagnostic accuracy. Expression of GIPC2 nign from malignant adrenocortical tumors, we studied messenger RNA has been shown to be significantly down- the genomewide expression profile of adrenocortical tu- regulated in a subset of kidney, colon, and rectal tumors to identify candidate diagnostic markers. We vali- mors.19 Downregulation of GIPC2 expression in human dated the microarray gene expression data by using real- primary tumors might lead to interference of transform- time quantitative RT-PCR. The application of stringent ing growth factor ␤ signaling.19 MGST1 is localized in the

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 843

©2008 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 Table 3. Results of Validation by RT-PCR and Diagnostic Accuracy of Genes Significantly Differentially Expressed by Microarray Between Benign and Malignant Adrenocortical Tumors

Spearman Correlation a b,c b,d c d Gene No. Gene Symbol UniGene ID Coefficient for C t P Value P Value AUC AUC Underexpressed 1 ALDH1A1 Hs.76392 Ͻ0.01 Ͻ.01 Ͻ.01 0.76 0.81 2 COL4A3 Hs.570065 Ͻ0.01 Ͻ.01 Ͻ.01 0.24 0.20 3 FLJ12993 Hs.199647 Ͻ0.01 .07 .01 0.67 0.73 4 GIPC2 Hs.13852 Ͻ0.01 .07 .07 0.62 0.68 5 GPR98 Hs.153692 Ͻ0.01 .72 .62 0.53 0.55 6 HSD3B2 Hs.654399 Ͻ0.01 Ͻ.01 .01 0.77 0.75 7 HTR2B Hs.421649 Ͻ0.01 Ͻ.01 Ͻ.01 0.87 0.88 8 LRRN3 Hs.3781 Ͻ0.01 .40 .64 0.65 0.63 9 MGST1 Hs.389700 Ͻ0.01 .01 .01 0.72 0.74 10 NRCAM Hs.21422 Ͻ0.01 .96 .86 0.50 0.48 11 PRLR Hs.368587 Ͻ0.01 .01 .01 0.74 0.75 12 RAPGEF4 Hs.470646 Ͻ0.01 .72 .54 0.52 0.63 13 RARRES2 Hs.647064 Ͻ0.01 Ͻ.01 Ͻ.01 0.86 0.88 14 SEMA6A Hs.156967 Ͻ0.01 .02 .01 0.29 0.25 15 SLC16A9 Hs.499709 Ͻ0.01 Ͻ.01 Ͻ.01 0.80 0.87 Overexpressed 1 APOBEC3B Hs.226307 Ͻ0.01 .29 .49 0.60 0.57 2 CCNB1 Hs.23960 Ͻ0.01 .06 .16 0.67 0.64 3 CCNB2 Hs.194698 0.66 Ͻ.01 Ͻ.01 0.86 0.84 4 CDCA1 Hs.234545 Ͻ0.01 .46 .67 0.62 0.66 5 CDCA2 Hs.33366 0.92 .02 .10 0.71 0.57 6 CDCA7 Hs.470654 Ͻ0.01 .01 .03 0.74 0.71 7 FREM2 Hs.253994 0.06 .006 Ͻ.01 0.70 0.80 8 GDF10 Hs.2171 0.33 .29 .25 0.64 0.61 9 IL13RA2 Hs.336046 Ͻ0.01 Ͻ.01 Ͻ.01 0.90 0.91 10 KIAA0101 Hs.81892 Ͻ0.01 .68 .19 0.70 0.61 11 LGR5 Hs.172176 0.10 .01 .008 0.74 0.76 12 MCM5 Hs.517582 Ͻ0.01 .01 .09 0.73 0.67 13 MCM7 Hs.438720 Ͻ0.01 .02 .10 0.71 0.66 14 MLF1IP Hs.481307 Ͻ0.01 .07 .19 0.67 0.63 15 PBK Hs.104741 Ͻ0.01 .04 .08 0.69 0.66 16 RACGAP1 Hs.505469 No amplification by PCR NA NA NA NA 17 PRG-3 Hs.382683 0.39 Ͻ.01 .01 0.76 0.75 18 SERPINE2 Hs.38449 Ͻ0.01 .75 .58 0.47 0.44 19 SHCBP1 Hs.123253 Ͻ0.01 .23 .41 0.61 0.58 20 TOP2A Hs.156346 Ͻ0.01 .03 .11 0.72 0.64 21 WIF1 Hs.284122 0.12 .03 .03 0.72 0.71 22 ZNF367 Hs.494557 0 Ͻ.01 .02 0.74 0.72

Abbreviations: AUC, area under the receiver operating characteristic curve; C t, PCR cycle threshold; NA, not applicable; RT-PCR, reverse transcription–polymerase chain reaction. a The log2 microarray gene expression vs the C t on RT-PCR was compared using the Spearman correlation coefficient. b P values were determined by Mann-Whitney test. c Analysis based on benign (54 adrenocortical adenoma, 20 adrenocortical hyperplasia, and 4 healthy adrenocortical tissue) and malignant (11 adrenocortical carcinoma) tissue samples. d Subset analysis of benign vs malignant adrenocortical tumors by RT-PCR gene expression, excluding Conn syndrome (n=30) and healthy (n=4) adrenocortical tissue samples.

endoplasmic reticulum and outer mitochondrial mem- for 2 underexpressed genes: ALDH1A1, an important en- brane, where it is thought to protect these membranes zyme of the major oxidative pathway of alcohol metabo- from oxidative stress and to serve as a cellular defense lism, and RARRES2.9 It is also the case for the overex- against carcinogens and xenobiotics.20 The APOBEC3B pressed gene KIAA0101, the regulator of cell proliferation gene encodes a protein that may be an RNA editing en- CCNB2, and the regulator of cell division TOP2A.7 The zyme that has roles in growth or cell-cycle control.21 overexpression of TOP2A and the downregulation of APOBEC3B has also been involved in carcinogenesis of HSD3B2 and RARRES2 gene expression have also been hepatocellular carcinoma through the generation of HBx documented in a microarray study23 of childhood adre- mutants, providing the hepatocytes with a selective clonal nocortical carcinomas. TOP2A encodes the topoisomer- growth advantage.22 ase II ␣, the direct molecular target of anthracyclines. This We identified several of the most differentially ex- gene is frequently coamplified with the HER2 gene in pressed genes that have been previously reported to be breast cancers, and there is agreement in different stud- dysregulated in adrenocortical carcinoma. This is the case ies that TOP2A has a predictive value for the efficacy of

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 844

©2008 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 Fernandez-Ranvier, Weng, Khanafshar, Barker, and 1.0 Kebebew. Analysis and interpretation of data: Fernandez- 0.9 Ranvier, Weng, Yeh, Khanafshar, Suh, Duh, Clark, and 0.8 Kebebew. Drafting of the manuscript: Fernandez- 0.7 0.6 Ranvier, Clark, and Kebebew. Critical revision of the manu- 0.5 script for important intellectual content: Fernandez- 0.4 Ranvier, Weng, Yeh, Khanafshar, Suh, Barker, Duh, and Sensitivity 0.3 Kebebew. Statistical analysis: Fernandez-Ranvier and 0.2 Multigene analysis Tumor size Kebebew. Obtained funding: Kebebew. Administrative, tech- 0.1 nical, and material support: Fernandez-Ranvier and 0.0 Kebebew. Study supervision: Duh and Kebebew. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Financial Disclosure: None reported. 1 – Specificity Funding/Support: This study was supported in part by grants from the UCSF Helen Diller Comprehensive Can- Figure. Area under the receiver operating characteristic curve (AUC) when the 5 biomarkers with the highest individual AUC values (IL13RA2, HTR2B, cer Center, the Mount Zion Health Fund, and the Ameri- CCNB2, RARRES2, and SLC16A9) were combined (based on 89 samples). can Cancer Society. The combined use of these biomarkers demonstrated a slight increase in Previous Presentation: This paper was presented at the diagnostic accuracy compared with the highest individual marker (IL13RA2 : AUC, 0.90), and all 5 genes were more accurate compared with tumor size as Pacific Coast Surgical Association 79th Annual Meet- a clinical variable for the diagnosis of adrenocortical carcinoma (AUC, 0.79). ing; February 16, 2008; San Diego, California; and is pub- An AUC of 1 represents the perfect diagnostic biomarker, without any lished after peer review and revision. The discussions that false-negative and false-positive results. follow this article are based on the originally submitted manuscript and not the revised manuscript. anthracyclines for the treatment of primary and meta- Additional Contributions: Pamela Derish, MA, helped static breast cancer.24 Overexpression of the IGF2 gene edit this manuscript. is one of the most common events seen in adrenocortical carcinoma.7-10,23 We did find several members of the insulinlike growth factor protein family to be upregu- REFERENCES lated in adrenocortical carcinoma, but their gene expres- 1. Lee J, Kebebew E. Adrenocortical tumors. In: Cameron JL, ed. Current Surgical sion levels were below the filtering criteria for differen- Therapy. 9th ed. Philadelphia, Pa: Mosby Elsevier; 2008:583-592. tial gene expression. 2. Latronico AC, Chrousos GP. Extensive personal experience: adrenocortical tumors. An important consideration when comparing genome- J Clin Endocrinol Metab. 1997;82(5):1317-1324. wide gene expression profiling studies in adrenocortical tu- 3. Icard P, Goudet P, Charpenay C, et al. Adrenocortical carcinomas: surgical trends mors is the type of microarray platform used because re- and results of a 253-patient series from the French Association of Endocrine Sur- geons study group. World J Surg. 2001;25(7):891-897. sults may vary with respect to the number of probe sets, 4. Favia G, Lumachi F, Carraro P, et al. Adrenocortical carcinoma: our experience. data quality assurance and validation of the microarray data, Minerva Endocrinol. 1995;20(1):95-99. type of data analysis that accounts for multiple compari- 5. Kebebew E, Reiff E, Duh QY, et al. Extent of disease at presentation and out- sons, and the number of samples analyzed. The focus of come for adrenocortical carcinoma: have we made progress? World J Surg. 2006; 30(5):872-878. this study was to select the most differentially expressed 6. Sturgeon C, Shen WT, Clark OH, et al. Risk assessment in 457 adrenal cortical genes because we were interested in identifying candidate carcinomas: how much does tumor size predict the likelihood of malignancy? diagnostic markers of adrenocortical carcinoma for clini- J Am Coll Surg. 2006;202(3):423-430. cal application. Such an approach obviously may be use- 7. Giordano TJ, Thomas DG, Kuick R, et al. Distinct transcriptional profiles of ad- ful for discovering biomarkers, but it compromises the abil- renocortical tumors uncovered by DNA microarray analysis. Am J Pathol. 2003; 162(2):521-531. ity to identify small but significant biological variability in 8. de Fraipont F, El Atifi M, Cherradi N, et al. Gene expression profiling of human gene expression and differentially involved molecular path- adrenocortical tumors using complementary deoxyribonucleic acid microar- ways in adrenocortical carcinoma. Therefore, these issues rays identifies several candidate genes as markers of malignancy. J Clin Endo- likely account for the discordances in chromosomal locus crinol Metab. 2005;90(3):1819-1829. 9. Vela´zquez-Ferna´ndez D, Laurell C, Geli J, et al. Expression profiling of adreno- and candidate genes between this study and previously pub- cortical neoplasms suggests a molecular signature of malignancy. Surgery. 2005; lished studies. 138(6):1087-1094. In conclusion, we identified 37 genes that are dys- 10. Slater EP, Diehl SM, Langer P, et al. Analysis by cDNA microarrays of gene ex- regulated in adrenocortical carcinoma, and several of these pression patterns of human adrenocortical tumors. Eur J Endocrinol. 2006; significantly differentially expressed genes have excel- 154(4):587-598. 11. Irizarry RA, Bolstad BM, Collin F, et al. Summaries of Affymetrix GeneChip probe lent diagnostic accuracy for distinguishing benign from level data. Nucleic Acids Res. 2003;31(4):e15 http://nar.oxfordjournals.org/cgi malignant adrenocortical tumors. /content/full/31/4/e15. Accessed April 1, 2008. 12. Gentleman RC, Carey VJ, Bates DM, et al. Bioconductor: open software devel- Accepted for Publication: April 6, 2008. opment for computational biology and bioinformatics. Genome Biol. 2004; 5(10):R80. Correspondence: Electron Kebebew, MD, Department 13. Bolstad BM, Irizarry RA, Astrand M, et al. A comparison of normalization meth- of Surgery and UCSF Helen Diller Family Comprehen- ods for high density oligonucleotide array data based on variance and bias. sive Cancer Center, University of California, San Fran- Bioinformatics. 2003;19(2):185-193. cisco, Box 1674, San Francisco, CA 94143-1674 14. Smyth GK. Linear models and empirical bayes methods for assessing differen- tial expression in microarray experiments. Stat Appl Genet Mol Biol. 2004; ([email protected]). 3:Article3. Author Contributions: Study concept and design: 15. Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. Controlling the false discovery Fernandez-Ranvier, Duh, and Kebebew. Acquisition of data: rate in behavior genetics research. Behav Brain Res. 2001;125(1-2):279-284.

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 845

©2008 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 16. Minn AJ, Gupta GP, Siegel PM, et al. Genes that mediate breast cancer metas- decisions? I agree with you that routinely we do not do fine- tasis to lung. Nature. 2005;436(7050):518-524. needle aspiration biopsy in our patients with an adrenal inci- 17. Vicaut E, Laemmel E, Stücker O. Impact of serotonin on tumour growth. Ann dentaloma to determine if it is malignant. This is because we Med. 2000;32(3):187-189. 18. Stav D, Bar I, Sandbank J. Usefulness of CDK5RAP3, CCNB2, and RAGE genes have not found fine-needle aspiration biopsy to be accurate for for the diagnosis of lung adenocarcinoma. Int J Biol Markers. 2007;22(2):108- distinguishing benign from malignant tumors. You also raised 113. the concern that fine-needle aspiration biopsy could lead to tu- 19. Kirikoshi H, Katoh M. Molecular cloning and characterization of human GIPC2,a mor seeding, which is possible but likely to be rare in the bor- novel gene homologous to human GIPC1 and Xenopus Kermit. Int J Oncol. 2002; derline cases of adrenal tumor that are 4 to 6 cm in size. As 20(3):571-576. you pointed out, these markers are more likely to be useful for 20. Siritantikorn A, Johansson K, Ahlen K, et al. Protection of cells from oxidative those that are indeterminate, so a large 8- or 10-cm tumor that stress by microsomal glutathione transferase 1. Biochem Biophys Res Commun. 2007;355(2):592-596. has suspicious imaging characteristics, such as irregular mar- 21. Bogerd HP, Wiegand HL, Doehle BP, et al. APOBEC3A and APOBEC3B are po- gins, necrosis, calcification, or high Hounsfield units, should tent inhibitors of LTR-retrotransposon function in human cells. Nucleic Acids just be resected and not biopsied, and maybe the risk of tumor Res. 2006;34(1):89-95. seeding with fine-needle aspiration biopsy could be a real con- 22. Xu R, Zhang X, Zhang W, et al. Association of human APOBEC3 cytidine deami- cern in these tumors. As far as applying the microarray analy- nases with the generation of hepatitis virus B x antigen mutants and hepatocel- sis to fine-needle aspiration biopsy, some groups have been able lular carcinoma. Hepatology. 2007;46(6):1810-1820. to use small amounts of RNA in a variety of tumor samples. 23. West AN, Neale GA, Pounds S, et al. Gene expression profiling of childhood adrenocortical tumors. Cancer Res. 2007;67(2):600-608. Having validated the candidate markers using quantitative real- 24. Arriola E, Rodriguez-Pinilla SM, Lambros MB, et al. Topoisomerase II alpha am- time RT-PCR, we would be able to use fine-needle aspiration plification may predict benefit from adjuvant anthracyclines in HER2 positive early biopsy samples that would yield enough RNA to use. For ad- breast cancer. Breast Cancer Res Treat. 2007;106(2):181-189. renocortical tumors, we have done this in an ex vivo fashion, and it yields enough RNA material. I should also mention that DISCUSSION for the quantitative PCR techniques, which we used to vali- date the results of the microarray analysis, only a minute amount Michael Bouvet, MD, La Jolla, California: As we have just heard, of RNA is required (50-125 ng of RNA). So, fine-needle aspi- you have identified 37 genes that are dysregulated in adreno- ration biopsy could be used. cortical carcinoma, and a number of differentially expressed Another situation besides the preoperative evaluation of ad- genes have an excellent diagnostic accuracy for distinguishing renocortical tumors that these markers may be useful for is in benign from malignant adrenocortical tumors. I have several the postoperative diagnosis of distinguishing benign from ma- questions for you. lignant tumors. We have had several patients who initially un- For this information to be useful in determining whether derwent complete surgical resection of a 2- to 3-cm adreno- or not to remove a nonfunctioning adrenal mass, a preopera- cortical tumor that was interpreted as an adrenocortical adenoma tive fine-needle aspiration biopsy would have to be per- on pathology but went on to develop locoregional recurrence formed. Using your current RNA isolation techniques, would as well as distant metastatic disease. So, I think that these mark- it be feasible to retrieve enough RNA from a needle biopsy to ers may be helpful as a postoperative diagnostic adjunct to his- run the microarray analysis? Have you done this yet? Tradi- tologic examination of the resected tumor. tionally, we tend to avoid fine-needle aspiration biopsy of ad- We and others have reported that, yes, tumor size is a use- renal masses. Is there any potential danger of tumor seeding if ful criterion for predicting the likelihood of malignancy. But, I one biopsies an adrenocortical carcinoma before removing it? think we could benefit and our patients could benefit from hav- Second, the average size of benign lesions in your series was ing more accurate adjunct diagnostic tests to predict which pa- 3.3 vs 9.9 cm for adrenocortical carcinoma. Your group has pre- tient is likely to have malignancy or not. Currently, if you con- viously published in 2006 in the Journal of the American Col- sider a patient with a 4-cm adrenocortical tumor, the likelihood lege of Surgeons that size is useful for predicting malignancy is 2-fold higher that the tumor is malignant. Assuming a preva- and that at a size threshold of greater than or equal to 4 cm, lence of malignancy of 5%, then 10% are malignant postopera- the likelihood of malignancy doubles to 10%, and it is more tively. That means you would have to resect 9 benign tumors than 9-fold higher for tumors greater than or equal to 8 cm, where the incidence of malignancy was 47%. The question then to catch 1 case of adrenocortical carcinoma. Moreover, earlier becomes, how much more information does microarray analy- diagnosis and treatment is the best way that we are going to sis really give the surgeon given that size is already an excel- improve the outcome of patients with adrenocortical carci- lent predictor of malignancy? Perhaps patients with intermediate- noma for which surgical resection is the primary effective treat- sized adrenal masses, say 4 to 6 cm, would benefit most from ment modality. Developing adjunct markers to better predict such an approach. which patient would benefit from early resection at a lower tu- Finally, we are in a new era of targeted therapeutics for many mor size threshold could benefit our patients. types of cancers. I think that your study does give excellent in- I do agree that these differentially regulated genes in adre- sight into potential molecular pathways that are involved with nocortical carcinoma may offer possible targets for therapy; many adrenocortical carcinoma and opens the door for further stud- of these genes have well-characterized functions, some with ies and may some day have potential therapeutic implications. agents currently available that may be effective at modulating Dr Kebebew: I think you have raised 2 important issues with their function. the questions you asked. One is, are we going to be able to trans- late these markers into clinical practice to make management Financial Disclosure: None reported.

(REPRINTED) ARCH SURG/ VOL 143 (NO. 9), SEP 2008 WWW.ARCHSURG.COM 846