Gene Expression Profiling of Childhood Adrenocortical Tumors

Alina Nico West,” Geoffrey A. Neale,” Stanley Pounds,” Bonald C. Figueredo, Carlos Rodriguez Galindo,” Mara Albonei D. Pianovski,’ Antonio G. Oliveira Filho,” David Malkin,’ Enzo Lalli,” Raul Ribeiro,3,5 and Gerard P. Zambetti®

“Interdisciplinary Science Program, University of Tennessee Health Science Center; 2Hartwell Center for Biotechnology; 3International Outreach Program; and the Departments of ‘Biostatistics, “Oncology, and ‘Biochemistry, St. Jude Children’s Research Hospital, Memphis, Tennessee; ‘Department of Pediatrics, The Hospital for Sick Children, University of Toronto, Toronto, Ontario, Canada; $Center for Molecular Genetics and Cancer Research in Children and ‘Division of Pediatric Hematology and Oncology, Erasto Gaertner Hospital, and Department of Pediatrics, Universidade Federal do Parana, Curitiba, Brazil; 10Instituto de Pesquisa e Ensino Boldrini, Campinas, Brazil; and “Institut de Pharmacologie Moléculaire et Cellulaire Centre National de la Recherche Scientifique Unité Mixte de Recherche 6097, Valbonne, France

Abstract

Pediatric adrenocortical tumors (ACT) are rare and often fatal malignancies; little is known regarding their etiology and biology. To provide additional insight into the nature of ACT, we determined the gene expression profiles of 24 pediatric tumors (five adenomas, 18 carcinomas, and one undeter- mined) and seven normal adrenal glands. Distinct patterns of gene expression, validated by quantitative real-time PCR and Western blot analysis, were identified that distinguish normal adrenal cortex from tumor. Differences in gene expression were also identified between adrenocortical ade- nomas and carcinomas. In addition, pediatric adrenocortical carcinomas were found to share similar patterns of gene expression when compared with those published for adult ACT. This study represents the first microarray analysis of childhood ACT. Our findings lay the groundwork for establish- ing gene expression profiles that may aid in the diagnosis and prognosis of pediatric ACT, and in the identification of signaling pathways that contribute to this disease. [Cancer Res 2007;67(2):600-8]

Introduction

Pediatric adrenocortical tumors (ACT) are rare malignancies occurring at a rate of 0.3 to 0.4 annual cases per million children under the age of 18 years (1, 2). Signs and symptoms of ACT include virilization, acne, deep voice, facial hair, muscle weakness, facial hyperemia, hypertension, and other signs of Cushing syndrome. The tumor size and weight, disease staging, and selected histologic criteria have been used to classify ACT as either carcinoma or adenoma. Large tumors (>200 g), and locally invasive or metastatic tumors, have been associated with poor outcome. However, in many cases, clinical and pathologic features fail to identify patients with localized disease that eventually relapse. Current therapy for pediatric ACT primarily relies on surgical resection of the tumor, although mitotane (a DDT-related compound)-with or without DNA-damaging agents-has been used with some success (3). The overall 5-year disease-free survival

is 50%; however, patients with stage IV disease have less than a 10% chance of long-term survival (2).

The adrenal cortex synthesizes essential steroids (e.g., glucocor- ticoids, androgens, and mineralocorticoids) that regulate diverse biological processes such as blood pressure, glucose metabolism, immune surveillance, and sexual development (4, 5). During gestation, the cortex is subdivided into the outer-definitive and inner-fetal zones, which contribute to the maintenance of normal pregnancy through the production of dihydroepiandrosterone sulfate. As this function is no longer required after birth, the adrenal gland rapidly loses 50% of its volume within the first 2 weeks due to massive apoptosis. Subsequently, the adrenal cortex undergoes significant tissue remodeling and develops into three defined regions: outer zona glomerulosa, middle zona fasciculata, and inner zona reticularis. The zona glomerulosa is primarily responsible for the production of aldosterone, whereas the zona fasciculata and zona reticularis produce corticosteroids and androgens, respectively. Various genetic abnormalities, either acquired or inherited (see below), promote ACT development during childhood or late adulthood (6).

Pediatric ACT is frequently reported in families with Li-Fraumeni syndrome and Li-Fraumeni-like syndrome, which are usually associated with TP53 tumor-suppressor germ line mutations (7, 8). The most frequently observed tumors in Li-Fraumeni syndrome include soft tissue sarcomas, osteosarcomas, breast carcinomas, brain tumors, and adrenocortical carcinomas. Indeed, it has been proposed that pediatric ACT is almost diagnostic of a germ line TP53 mutation (9), but clearly alternative factors can contribute to this tumor type (e.g., Beckwith-Wiedemann syndrome, Carney’s complex, and multiple endocrine neoplasia type I; ref. 10). Beckwith-Wiedemann syndrome is characterized by the overgrowth of tissues and organs, including the adrenal gland. Beckwith-Wiedemann syndrome is usually sporadic; how- ever, it also occurs as a familial autosomal dominant form linked to the loss of imprinting at the insulin-like growth factor-II (IGF-II) locus on chromosome 11p15.5, resulting in the overproduction of IGF-II (6). The underlying genetic events responsible for the Beckwith-Wiedemann syndrome phenotype are complex, with multiple genes (e.g., KCNQ1 and CDKN1C) being implicated in its etiology (11, 12).

The cooperating factors and signaling pathways that promote the development of childhood ACT are not well defined. Animal studies implicate inhibin-a, a glycoprotein with homology to transforming growth factor-ß, as a suppressor of ACT development (13). Deletion of inhibin-a by gene targeting in gonadectomized mice causes fully penetrant ACT by 4 to 5 weeks of age. Consistent with the mouse model, mutation of inhibin-a with loss of

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

A.N. West and G.A. Neale contributed equally to this work.

Requests for reprints: Gerard P. Zambetti, Department of Biochemistry, St. Jude Children’s Research Hospital, 332 North Lauderdale, Memphis, TN 38105. Phone: 901- 495-3429; Fax: 901-525-8025; E-mail: gerard.zambetti@stjude.org.

@2007 American Association for Cancer Research. doi:10.1158/0008-5472.CAN-06-3767

heterozygosity at chromosome 2q33 was commonly observed in human pediatric ACT (13). Comparative genomic hybridization analysis of pediatric ACT also showed recurrent chromosomal alterations, such as the amplification of chromosome 9q34 (14). Localized within this region is the nuclear orphan receptor steroidogenic factor-1 (SF1, NR5A1), which is required for normal adrenal gland development. Subsequent studies showed that SF1 is amplified and overexpressed in ~ 90% of pediatric ACT (15, 16). Similarly, both pediatric and adult ACT express elevated levels of IGF-II (17, 18).

Due to the rarity of pediatric ACT, it becomes necessary to consolidate resources to maximize efforts in studying this disease in a comprehensive and thorough manner. We therefore estab- lished an International Pediatric Adrenocortical Tumor Registry and Bank at St. Jude Children’s Research Hospital.12 More than 250 subjects have enrolled in the registry component since 1990 (the adrenal tissue bank has been in existence since 2000). To identify key factors and signaling pathways that may be involved in adrenocortical tumorigenesis, we conducted an Affymetrix gene expression profiling analysis of pediatric ACT. As we report here, distinct expression signatures have been identified that discrimi- nate between normal adrenal cortex and ACT. In addition, our retrospective analyses identified profiles that may aid in the differential diagnosis of adenoma from carcinoma. Insight into the cell type of origin that gives rise to ACT has also been generated. Our findings provide the basis for identifying signaling pathways that are corrupted during adrenocortical tumorigenesis, with the goal of establishing new therapeutic targets that could be exploited in treating this often fatal disease.

Materials and Methods

Institutional review board approval. The institutional review board (IRB) of St. Jude and the ethics committees of the Hospital de Clinicas of the Federal University of Parana, Hospital Erasto Gaertner, and the Centro Infantil de Investigações Hematológicas Dr. Domingos A. Boldrini approved the genetic analysis of pediatric normal adrenal cortex and ACT. Informed consent was obtained for each subject.

Total RNA preparation. Tissue samples were classified according to established histopathologic criteria and verified by two independent pathologists. Total RNA was isolated from 50 to 100 mg of pediatric ACT using the RNeasy RNA Midi-Prep kit (Qiagen, Valencia, CA). Tumors were prepared in a 4℃ cold room, sliced into fine pieces using a sterile scalpel, and homogenized with 18- and 19-gauge needles in lysis Buffer RLT (Qiagen) containing B-mercaptoethanol. Total RNA was isolated by the Animal Tissues protocol following the manufacturer’s recommendations. The RNA was resuspended in diethyl pyrocarbonate-treated water, quantified by UV absorbance at 260/280 nm, and stored at -80℃.

cDNA amplification and real-time PCR analysis. cDNA was generated from 1 µg total RNA using the iScript cDNA amplification kit according to the manufacturer’s instructions (Bio-Rad Laboratories, Hercules, CA). cDNA was diluted 1:2 using sterile double-distilled water before real-time PCR analysis. The following genes were amplified by real-time PCR using the iQSybrGreen PCR amplification mix (Bio-Rad Laboratories; according to the manufacturer’s instructions) and 400 ng per primer: IGF-II, type II 3฿- hydroxysteroid dehydrogenase (HSD3B2), fibroblast growth factor receptor-4 (FGFR4), NURR1, NGF1-B, and nephroblastoma overexpressed (NOV). Ubiquitin was also amplified as a loading control. Each normal adrenal and tumor sample was amplified in triplicate via separate PCR conditions and compared with ubiquitin expression levels using the AAC, method (19).

Primer sequences and PCR conditions are described in Supplementary Table S1.

Western blot analysis. Protein was isolated from normal adrenal cortex and tumor tissues by homogenization in T-PER lysis buffer (Pierce Chemical, Rockford, IL) containing a protease inhibitor cocktail (Roche Diagnostics Corporation, Indianapolis, IN). Total protein (50 µg) was analyzed by SDS-PAGE using the Novex NuPAGE system (Invitrogen, Carlsbad, CA). Proteins were separated by electrophoresis and transferred to 0.45-um nitrocellulose membranes. Membranes were blocked in TBS-T buffer [10 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 0.1% Tween 20] containing 5% nonfat milk and probed with the following primary antibodies: goat polyclonal anti-human IGF-II (1:500; Sigma-Aldrich Chemical, St. Louis, MO), rabbit polyclonal anti-human HSD3B2 (1:500; gift from Dr. C. Richard Parker Jr., University of Alabama, Birmingham, AL), and mouse monoclonal anti-human actin (1:2000; Sigma-Aldrich Chemical). Membranes were washed with TBS-T and hybridized with the following horseradish peroxidase-linked antibodies diluted in TBS-T containing 5% nonfat milk: rabbit anti-goat (1:1,000; Calbiochem, San Diego, CA), donkey anti-rabbit (1:3,000; Amersham Biosciences, Piscataway, NJ), and sheep anti-mouse (1:2,000; Amersham Biosciences). The mem- branes were washed with TBS-T and developed using Supersignal West Dura chemiluminescence reagent (Pierce Chemical) according to the manufacturer’s protocol.

Microarray analysis. The Affymetrix U133A GeneChip was used to collect expression data for 22,215 probe sets on each of 31 samples (18 adrenocortical carcinomas, five adenomas, one undetermined ACT, and seven normal adrenal cortex). Microarray analysis was done in the Hartwell Center Affymetrix core laboratory at St. Jude. High-quality RNA, confirmed by UV spectrophotometry and an Agilent 2100 Bioanalyzer, was processed according to the Affymetrix one-cycle labeling protocol.13 In brief, 5 to 10 µg total RNA was annealed to an oligo-dT(24)-T7 primer to initiate cDNA synthesis. Purified double-stranded cDNA was used as a template to synthesize biotin-labeled cRNA using T7 RNA polymerase. Labeled cRNA (20 µg) was fragmented, added to a mixture containing blocking agents and array controls, and hybridized overnight at 45℃ to the gene chip array. Following hybridization, arrays were stringently washed, stained with streptavidin-conjugated phycoerythrin, and scanned using an Affymetrix GeneChip Scanner 3000. Relative expression signals for each gene was calculated using the Affymetrix GCOS software (version 1.4) using the global normalization method where the 2% trimmed mean signal was set to a target value of 500.

Statistical analysis. Microarray signals were summarized and normal- ized using Affymetrix GCOX software as described above. No probe set was excluded before subsequent statistical analysis because filtering has been found to be of questionable value (20). The Wilcoxon rank-sum test was used to compare median expression between normal and tumor tissues in each probe set (21). Likewise, the rank-sum test was used to compare the median expression level of each probe set between adrenocortical adenomas and carcinomas. To account for multiple testing in each of these analyses, we used a robust method to estimate the false discovery rate (22). These analyses were implemented using S-plus software, version 6.2 for Windows (Microsoft).14 The robust false discovery rate method was implemented using our freely available routines.15

To compare expression profiles in our pediatric ACT samples with data in other reports (23, 24), U133A probe sets were matched by either Genbank accession ID (23) or by the Affymetrix “best match” criteria (24). Fold- change point estimates were computed by exponentiation of the difference of means of log-transformed signals. This estimate of fold change can be interpreted as an estimate of the ratio median expression levels of the two groups. The t distribution was used to compute 95% confidence intervals for the difference of means of log signals; these intervals were transformed

12 www.stjude.org.

13 http://www.affymetrix.com/support/technical/manual/expression_manual.affx.

14 www.splus.com.

15 5 http://www.stjuderesearch.org/depts/biostats/robustfdr/index.html.

601

into confidence intervals for fold changes by exponentiation. The fold- change confidence intervals are not adjusted for multiple testing.

As measures of how fold changes observed in our study correlated with fold changes observed in other studies, we computed the number of probe sets with a directional agreement (i.e., the fold-change estimates from the two studies were in the same direction) and Kendall et al.’s method (25) with the two sets of fold changes as input. We used a permutation method to assess the statistical significance of the observed values of these measures of agreement. The permutation assessment was done by computing the fold changes on 1,000 data sets, derived by randomly reassigning group labels in our data set to the expression profiles in our data set, and then computing the agreement statistics. We counted the number of permuted data sets in which stronger values of the agreement statistics were observed to obtain the P values.

Estimates of overall and relapse-free survival were computed using the Kaplan-Meier method with SEs determined using the method of Peto et al. (26). Overall survival was defined as the duration from date of diagnosis to date of death with those living at last follow-up considered censored. Relapse-free survival was defined as the duration from date of diagnosis to date of relapse or death with those alive and relapse-free at last follow-up censored.

Results

Clinical information. Pediatric adrenocortical adenoma and carcinoma patients were enrolled on the International Pediatric Adrenocortical Tumor Registry and Bank protocol. Tumor speci- mens were harvested during surgery and snap frozen in liquid nitrogen to preserve tissue integrity. Data have been compiled for eight males and 15 females between 0 and 16 years of age. Table 1 summarizes the primary clinical information for each subject (excluding sample Unk1 with ACT of undetermined histology), including stage of the disease, tumor class, sex, age, relapse-free survival, and overall survival. Details regarding clinical features and treatment were also collected.

All subjects presented with virilization. Eleven patients had signs and symptoms consistent with an increased secretion of gluco- corticoids (Cushing syndrome) and eight patients were hyperten- sive at presentation. Normal adrenal glands were obtained with IRB approval as discarded tissue from seven cases of Wilms’ tumor. These patients, whose age ranged from 2 to 6 years, had not

Table 1. Clinical data of 24 pediatric adrenocortical cancer patients
Sample	Sample ID	Sex	Age* (y)	Tumor or normal	Histologic type (adenoma or ACC)	Tumor stage	Virilization	Cushing syndrome	Outcome	Treatment
1	ACC1	M	8	T	ACC	4	Y	N	CR	SC
2	ACA1	F	12	T	Adenoma	—	Y	Y	CCR	S
3	ACC2	F	5	T	ACC	2	Y	N	DD	SC
4	ACA2	F	<1	T	Adenoma	—	Y	Y	CCR	S
5	ACC3	M	2	T	ACC	3	Y	N	DD	SC
6	ACC4	F	2	T	ACC	1	Y	N	CCR	S
7	Unk1	Unk	Unk	T	Unk	Unk	Unk	Unk	Unk	Unk
8	ACC5	M	4	T	ACC	2	Y	N	CR	SC
9	ACC6	F	13	T	ACC	4	Y	Y	Deceased	SC
10	ACA3	F	3	T	Adenoma	—	Y	N	CCR	S
11	ACA4	F	2	T	Adenoma	—	Y	Y	CCR	S
12	ACA5	F	4	T	Adenoma	—	Y	N	CCR	S
13	ACC7	F	11	T	ACC	2	Y	Y	CR	SC
14	ACC8	M	9	T	ACC	2	Y	Y	DD	SC
15	ACC9	M	2	T	ACC	1	Y	Y	DD	SC
16	ACC10	F	<1	T	ACC	2	Y	Y	CCR	S
17	ACC11	M	2	T	ACC	1	Y	N	CCR	S
18	ACC12	M	12	T	ACC	3	Y	Y	CCR	SC
19	ACC13	M	4	T	ACC	3	Y	Y	CR	SC
20	ACC14	F	10	T	ACC	3	Y	N	CR	SC
21	ACC15	F	3	T	ACC	1	Y	Y	CR	SC
22	ACC16	F	6	T	ACC	2	Y	N		SC
23	ACC17	F	15	T	ACC	4	Y	N	DD	SC
24	ACC18	F	3	T	ACC	3	Y	N	CCR	SC
25	Nor001	Unk	Unk	N	NA	NA	NA	NA	NA	NA
26	Nor004	Unk	Unk	N	NA	NA	NA	NA	NA	NA
27	Nor006	Unk	Unk	N	NA	NA	NA	NA	NA	NA
28	Nor007	Unk	Unk	N	NA	NA	NA	NA	NA	NA
29	Nor009	Unk	Unk	N	NA	NA	NA	NA	NA	NA
30	Nor010	Unk	Unk	N	NA	NA	NA	NA	NA	NA
31	Nor011	Unk	Unk	N	NA	NA	NA	NA	NA	NA

NOTE: Adenomas were not staged based on standard ACT staging criteria (2).

Abbreviations: Unk, unknown; ACC, adrenocortical carcinoma; NA, not applicable; CR, complete remission; CCR, continuous complete remission; DD, died from disease; Deceased, died from unknown causes; S, surgical resection only; SC, surgical resection plus chemotherapy. *Ages were rounded to the nearest full year.

received chemotherapy before surgery, thus avoiding complications of chemotherapeutic effects. Normal adrenal cortex was subse- quently isolated by an American Board-certified pathologist and processed as described in the Materials and Methods.

Gene expression profiling distinguishes ACT from normal adrenal tissue. Gene expression profiles for the ACT and normal adrenal cortex samples were generated using the Affymetrix U133A gene chip, which recognizes 14,500 genes using 22,215 probe sets. We estimate that at least 33% of the probe sets on the array are differentially expressed between tumor and normal tissues; for 1,019 of the probe sets, we detected differences that were significant at P = 0.001 (see Supplementary Table S2). Furthermore, we estimate that 1.5% or fewer of the 1,019 detected differences are false discoveries. Hierarchical clustering analysis was used to visualize the variability between ACT and normal cortex (Fig. 1).

Validation of the gene expression data set. Among the 1,019 significant probe sets, we identified 25 with the greatest and least ratios of median expression in tumor samples to that of normal samples (Table 2). The median expression of FGFR4 in the ACT samples was 21 times that of the normal samples (95% confidence interval, 11.4-38.8) and represents the highest induced gene within the group. Previous studies implicate FGFR4 in breast cancer progression and other tumors (27).

The median expression of IGF-II in ACT samples was 18 times that of the normal samples (95% confidence interval, 7.8-42.7). Overexpression of IGF-II in the ACT samples was subsequently verified at the RNA and protein level by quantitative real-time PCR (qRT-PCR) assay and Western blot analysis, respectively (Fig. 2A and B). The qRT-PCR assay revealed higher expression values in the ACT samples than that determined by microarray analysis, most likely due to the larger dynamic range of the RT-PCR assay. IGF-II protein levels were also significantly higher in the tumors than in the normal adrenal cortex samples and correlated with changes in mRNA expression. However, the 7.5-kDa mature form of IGF-II was selectively expressed in the normal adrenal tissue, whereas multiple proforms of IGF-II, including the prominent 20-kDa form of the protein, were overexpressed in the ACT samples. These results are concordant with previous biochemical and microarray analyses of adult and pediatric ACT (18, 24, 28, 29). In addition, the expression of NOV, a member of the CCN gene family of secretory proteins that plays a role in cell adhesion, was significantly lower in the ACT samples compared with normal adrenal tissue (Table 2; data not shown). Loss of NOV expression in the pediatric ACT samples is also in agreement with those reported in adult ACT studies by Martinerie et al. (30), further corroborating our data set.

Cell origin of pediatric ACT. HSD3B2, a steroidogenic enzyme responsible for the conversion of pregnenolone to progesterone in the synthesis of glucocorticoids, mineralocorticoids, and andro- gens, is expressed at programmed times during adrenal develop- ment (for review, see ref. 31). During late embryogenesis, HSD3B2 is preferentially expressed in the adrenocortical definitive zone but not the fetal zone. After birth, HSD3B2 expression is restricted largely to the zona glomerulosa and zona fasciculata. Microarray analyses showed that the median expression of HSD3B2 in pediatric ACT samples is ~40-fold less than that of normal controls. This finding was confirmed by qRT-PCR and Western blot analysis (Fig. 2C and D).

Moreover, the expression of NURR1 (NR4A2) and NGF1-B (NR4A1), transcriptional regulators of HSD3B2 gene expression (32), were concomitantly lower in the ACT samples (Table 2; and data not shown). The expression of KCNQ1, which encodes a

Figure 1. Heat map and hierarchical clustering analysis comparing pediatric ACT and normal cortex. Relative expression signals of 1,019 unique probe sets: red, overexpressed; green, underexpressed. Differentially expressed genes were significant at P = 0.001. N, normal; ACC, adrenocortical carcinoma; Ad, adenoma; U, unknown. Bar, SD from the mean.

-3 -2 -1 0 1 2 3

Type

NOR001 N

NOR004 N

NOR009 N

NOR006 N

NOR011 N

NOR007 N

NOR010 N

ACC

ACC16 UNCC

ACC

ACC10 ACC

ACC

ACC11

ACC12

ACC17

ACC2

ACC14 ACC6

ACC7

ACC8

ACC15

ACC13

ACC18

=96

E97

ACA1

ACC9

ACA3

ACA5

Unk1

ACC4

ACC5

ACC3

ACA2

ACC1

ACA4

voltage-dependent potassium channel, was also lower (~85-fold) in the pediatric ACT samples than in normal adrenal cortex (Table 2). Murine Kcnq1 is preferentially expressed in the cortical zona glomerulosa (33), but not in the adrenal medulla. Taken together, these results suggest that pediatric ACT may arise from either the fetal zone or the more developmentally mature zona reticularis or zona fasciculata.

Comparison between adult and pediatric ACT. Giordano et al. (24) recently identified differences in gene expression pat- terns between adult ACT and normal tissue using the Affymetrix human U95A gene chip. Independently, Rainey et al. (23) compared the gene expression profiles of normal human fetal and adult adrenal cortex using a cDNA microarray approach. To our know- ledge, there have been no published studies to date compar- ing adult and pediatric ACT gene expression in a comprehensive manner.

To compare expression profiles across studies, we queried our microarray data set for the genes identified as significantly changed in the other two studies. We then used expression values relative to normal tissues within each study (log2 ratio) to compare gene profiles across studies. These analyses showed that the most significant differences identified in the comparison between adult adrenal tumors and normal adult adrenal cortex (24) were remarkably similar to our findings comparing childhood ACT (adenoma and carcinoma) to normal cortex (t = 0.56, P = 0.001; Fig. 3). Moreover, the observed direction of association was the

Table 2. Dysregulated genes in pediatric ACT
Probe Set ID	Gene symbol	Ratio of medians	95% Low	95% High
Increased
211237_s_at	FGFR4	21.1	11.4	38.8
204597_x_at	STC1	19.1	8.2	44.4
210881_s_at	IGF-II	18.3	7.9	42.7
202410_x_at	IGF-II	16.0	6.2	41.4
203213_at	CDC2	13.8	5.7	33.6
219918_s_at	ASPM	13.4	4.2	43.5
204285_s_at	PMAIP1	13.1	5.9	29.1
213562_s_at	SQLE	12.9	6.2	27.1
205345_at	BARD1	12.8	8.6	18.9
218755_at	KIF20A	12.0	5.6	25.7
204056_s_at	MVK	10.6	4.1	27.4
218009_s_at	PRC1	10.4	4.2	25.9
220091_at	SLC2A6	10.2	5.3	19.7
207414_s_at	PCSK6	10.0	3.7	26.9
218585_s_at	RAMP	9.6	3.7	24.6
203828_s_at	NK4	9.3	4.2	20.4
204641_at	NEK2	8.6	3.3	22.5
201292_at	TOP2A	8.6	3.2	23.0
201890_at	RRM2	8.6	3.0	24.8
201291_s_at	TOP2A	8.6	2.5	29.3
213126_at	MED8	8.2	3.8	18.0
203708_at	PDE4B	8.0	3.0	21.4
209218_at	SQLE	8.0	4.1	15.3
202779_s_at	UBE2S	7.9	2.3	27.5
213479_at	NPTX2	7.8	4.4	13.9
Decreased
204487_s_at	KCNQ1	85.5	49.2	148.6
206294_at	HSD3B2	41.1	16.1	104.7
204621_s_at	NR4A2	26.9	11.5	63.0
214630_at	CYP11B2	26.2	11.9	57.9
214321_at	NOV	24.4	10.8	55.3
209613_s_at	ADH1B	21.6	10.9	42.8
216248_s_at	NR4A2	19.2	8.5	43.6
204501_at	NOV	19.1	9.6	38.2
211959_at	IGFBP5	18.2	11.3	29.5
203523_at	LSP1	17.9	10.5	30.5
208606_s_at	WNT4	17.3	8.3	36.1
213764_s_at	MFAP5	15.7	4.3	56.6
204622_x_at	NR4A2	15.5	7.8	31.1
209496_at	RARRES2	14.3	6.7	30.2
202768_at	FOSB	14.0	4.8	41.2
217767_at	C3	13.7	7.2	26.2
211217_s_at	KCNQ1	13.6	6.4	29.1
202994_s_at	FBLN1	13.2	6.9	25.3
203131_at	PDGFRA	13.0	7.9	21.6
205969_at	AADAC	12.2	5.4	27.8
203798_s_at	VSNL1	12.0	6.0	23.8
203424_s_at	IGFBP5	11.1	6.6	18.5
204457_s_at	GAS1	11.0	5.3	22.9
213994_s_at	SPON1	10.8	3.9	30.0
211896_s_at	DCN	10.8	6.2	18.8

same for 147 of 153 probe sets in our study corresponding to their reported fold changes (P < 0.001).

Expression of IGF-II and HSD3B2 was dysregulated in a similar manner in both adult and childhood ACT; however, the degree of IGF-II expression seems to be greater in the adult tumors (200- versus 20-fold; Supplementary Table S3), possibly due to

the relatively lower levels of IGF-II in the normal adult adrenal cortex (23, 24). There was also a remarkable correlation in gene expression profiles between normal fetal adrenal tissue (23) and pediatric ACT (t = 0.34, P = 0.022; Fig. 3). Additionally, the direction of association agreed for 99 of 127 probe sets corresponding to genes for which they report fold changes

(P = 0.006; Supplementary Table S4). These results indicate that both adult and pediatric ACT resemble fetal tissue with respect to gene expression patterns. Furthermore, our findings suggest that adult and childhood ACT may select for common genetic and biochemical alterations and may be more physiologically related than previously considered.

Differences between pediatric adrenocortical carcinoma and adenoma. There are no definitive tests to predict ACT malignant potential. Tumor size is one of the most consistent prognostic indicators in children with completely resected ACT (1), although it is not uncommon for patients with small tumors to experience relapses. We therefore compared gene expression

profiles of ACT that were classified by histologic criteria as either adenoma or carcinoma to identify changes that may distinguish between these risk groups.

For 52 probe sets, we detected differences in expression between adrenocortical adenomas and carcinomas that were significant at the P = 0.001 level (Fig. 4; Supplementary Table S5). We estimate that 56% or more of the detected differences represent true discoveries. Among this set was a consistent and marked decrease in the expression of major histocompatibility class II genes. Specifically, the median expressions of HLA-DRB1, HLA-DPB1, HLA- DRA, and HLA-DPA1 mRNA levels were 6- to 8-fold lower in pediatric adrenocortical carcinomas than in adenomas. Similar findings

Figure 2. Dysregulation of IGF-II and HSD3B2 gene expression in pediatric adrenocortical carcinoma. A, IGF-II mRNA levels are significantly higher in tumors compared with normal tissue. Bars, SD. B, IGF-II protein is overexpressed in adrenal tumors but is incompletely processed. C, HSD3B2 transcripts are markedly lower in adrenal tumors than in normal tissue. Bars, SD. D, HSD3B2 protein levels are reduced in tumors compared with normal tissue.

Ave relative expression

Actin

IGF2

Ave relative expression

0.5

1.5

2.5

100

10000

NORMAL

ACC1

ACA1

Ad1

Normal

ACC2

Normal

Ad2

ACA2

ACC3

ACC4

Unk

Unk1

ACC5

ACC6

Ad3

ACC6

ACC1

Ad4

ACC1

ACA3

HSD3B2

IGF2

Ad5

ACA4

ACC3

ACC7

ACC3

ACA5

ACC4

ACC8

ACC7

ACC9

ACC4

Unk1

ACC10

ACC8

ACC5

ACC11

Unk1

ACC9

ACC6

ACC12

ACC10

ACC13

ACC5

ACC11

ACA3

ACC14

ACC12

ACA4

ACC15

ACC6

ACC13

ACC16

ACC2

ACC17

ACA3

ACC14

ACC18

ACC15

ACA4

ACC16

Mature

Unprocessed

ACC2

ACC17

ACC18

Research.

Cancer Res 2007; 67: (2). January 15, 2007

have been recently reported by Bornstein and coworkers (34, 35) in a study of adult ACT. HLA-class II expression may therefore serve as a marker for distinguishing between adrenocortical carcinoma and adenoma.

Discussion

We have established the first pediatric ACT gene expression profile database. Analysis of the ACT panel revealed a marked increase in FGFR4 and IGF-II expression, and a sharp decrease in KCNQ1, CDKN1C, and HSD3B2 gene expression in the ACT samples compared with normal adrenal cortex. In support of these results, qRT-PCR and Western blot analyses confirmed the differential expression of several of these factors (Fig. 2; and data not shown). Giordano et al. (24) also detected a similar pattern of IGF-II, KCNQ1, and CDKN1C expression in adult adrenocortical carci-

Figure 3. Comparisons of pediatric ACT gene expression profiles with those observed in adult ACT (top panel) and fetal adrenal cortex (bottom panel).

Adult ACT vs normal adrenal

(Log2Ratio)

-2

-4

-6

-6 -4 -2 0 2 4 6

Pediatric ACT vs normal adrenal (Log2Ratio)

Fetal adrenal vs adult adrenal

(Log2Ratio)

-2

-4

-6

-4

-2

Pediatric ACT vs normal adrenal (Log2Ratio)

nomas. All three of these genes are localized to an imprinted locus on chromosome 11p15, with IGF-II being normally expressed from the paternal allele and CDKN1C and KCNQ1 being expressed only from the maternal allele (for review, see refs. 11, 12). The coupled dysregulation of IGF-II, CDKN1C, and KCNQ1 (Table 2) implies an imprinting defect, similar to what has been observed in Beckwith- Wiedemann syndrome (11, 12).

Overexpression of IGF-II in pediatric ACT was anticipated based on previously published reports (Fig. 2A and B; Table 2; refs. 18, 24, 28, 36). However, the finding that the majority of the tumors grossly overexpress immature forms of IGF-II was surprising, but not unprecedented based on adult ACT studies (29). Pro-IGF-II must be posttranslationally modified by glycosyla- tion and proteolytic cleavage before its mature, active 7.5-kDa form is secreted (37). Here, we have detected, in the ACT samples, IGF-II proteins ranging from 14 to 22 kDa, but not the 7.5-kDa form, which was readily evident in normal adrenal cortex tissue. It is generally considered that the overexpression of IGF-II in ACT provides a growth advantage that drives tumorigenesis. Consistent with this hypothesis, transgenic mice engineered to express high levels of IGF-II develop adrenal hyperplasia (38) and recombinant IGF-II stimulates human fetal adrenocortical cell proliferation in culture (39). Because the IGF-I receptor is concomitantly up-regulated in the pediatric tumors analyzed here (Supplementary Table S2), it is reasonable to speculate that IGF-II may also play a role in pediatric adrenocortical tumorigenesis and therefore serve as a drug target. However, further consider- ation must be given as to whether these adrenal tumors secrete an active form of IGF-II that contributes to the growth and survival of these cells.

Interestingly, basic FGF-2 (bFGF-2) suppresses the processing of IGF-II in human ACT cells, thereby blocking its secretion, resulting in a marked accumulation of intracellular IGF-II (40). Consistent with the high levels of partially processed IGF-II pro- tein in the adrenal tumors, FGFR1 and FGFR4, both of which can be activated by bFGF-2, were found by microarray analysis to be significantly up-regulated in the ACT samples (Table 2; Supplementary Table S2). Moreover, because bFGF is a potent angiogenic factor and is mitogenic for fetal adrenal cortex cells (41, 42), the inhibition of the FGFR signaling pathway may represent a rational approach in developing new treatments for pediatric ACT. In support of this concept, 17 of the most signifi- cant genes dysregulated in pediatric ACT (Fig. 1; Supplementary Table S2) function within the mitogen-activated protein kinase pathway, including NRAS, an immediate downstream target of FGFR signaling.

The finding that the expression of KCNQ1, HSD3B2, and its corresponding transcriptional regulators NURR1 and NGF1B is markedly lower in pediatric ACT compared with normal adrenal cortex supports the thesis that the tumors originate from either the fetal zone during embryogenesis or the developing zona fasciculata or zona reticularis during the first few years of life. At the very least, the pediatric adrenal tumors share biochemical characteristics of these compartments. Because normal adult tissue is significantly different from the fetal adrenal cortex (23), the remarkable and somewhat unexpected similarity between adult and pediatric ACT implies (Fig. 3; Supplementary Table S3) the existence of an adrenal stem cell that may become corrupted to give rise to the developing tumor. Alternatively, the tumors, whether adult or pediatric, may undergo dedifferentiation as they develop (43).

Figure 4. Heat map of differentially expressed genes comparing pediatric adrenocortical carcinomas and adenomas. Median expression values calculated by the Wilcoxon rank-sum test generated data for 52 unique probe sets between adenoma and carcinoma significant at P = 0.001. U, unknown. Red, overexpressed; green, underexpressed. Bar, SD from the mean.

-2

-1

Type

ACC

ACC12

ACC15

ACC14

ACC10

ACC13

ACC17

ACC16

ACC18

ACC11

— 20

ACA3

ACA5

ACA1

ACA2

ACA4

Unk1

ACC4

ACC1

ACC7

ACC3

ACC6

ACC2

ACC8

ACC5

ACC9

-20

Gene Symbol

PLAGL1

-IPART1

EMCN

GIMAPS

BTN2A1

GRHL2

SLC25A24

TGFBR2

CYP11B1

APOE

HRC

KTN1

BTN2A1

GLS

KIAA0247

CYB5R3

DGAT1IDGAT1 /W LOC642255

CD7

PAH

GLUL

APOL3

HLA-DPA1

HLA-DRB1

HLA-DPB1

HLA-DRB1

C1OB

HLA-DRA

TNRC4

NLE1

C1orf135

C7orf24

SLC35C1

TRIPITRAIP

DLL3

FLJ22814

HOXB13

SAP130

DES

DDX56

LONRF3

-ILOD344450

CRSPB

WIT-1IWIT1

FAT

DUOX2

TCEB2

NUP58

PRPP3

PCTK1

In the present study, patterns of gene expression have been iden- tified that distinguish adrenocortical carcinomas from adenomas, which is often difficult to assess by standard histopathologic approaches. Interestingly, two adrenocortical carcinoma cases, which have not yet relapsed, segregated with the adrenocortical adenoma group (Fig. 4; Supplementary Table S5 and Fig. S1), underscoring the limitations of the histologic criteria to predict tumor malignant potential. Future prospective studies should determine the usefulness of gene expression analysis in the classification and prognosis of pediatric ACT.

Significant changes in the expression profiles between adreno- cortical adenomas and carcinomas included the MHC class II genes, which are largely restricted to hematopoietic lineages. Interestingly, the adrenocortical reticular zone also expresses MHC class II antigens after 4 years of age (34, 35). Based on the age of the patients diagnosed with adrenocortical adenoma, it is reasonable to speculate that the relatively high MHC class II expression reflects an infiltration of immune cells that limits tumor potential (B.F .; data not shown). Conversely, the association of low MHC class II expression in the carcinomas may represent a mechanism to evade immune surveillance, which could contribute to its malignant phenotype (35).

Little is known regarding the pathways and factors that promote pediatric ACT and there is no proven therapy for this rare malignancy other than surgery. Our findings identify potentially important components that may contribute to adrenocortical tumorigenesis. However, the establishment of genetically engi- neered mice, primary tissue culture cell lines, and/or human ACT xenografts will be required to explore new potential targets, such as FGFR4, IGF-II, and other dysregulated genes identified here. Only through these efforts can advancements in the treatment of pediatric ACT be made.

Acknowledgments

Received 10/12/2006; accepted 11/14/2006.

Grant support: NIH/National Cancer Institute grants CA63230 and CA71907 (G.P. Zambetti), Fondation pour la Recherche Médicale (E. Lalli), L’Association pour la Recherche sur le Cancer (E. Lalli), Cancer Center Core grant CA21765, National Cancer Institute of Canada with funds from the Canadian Cancer Society, and the American Lebanese Syrian Associated Charities.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Robert Lorsbach for his assistance in processing tissue samples, the Hartwell Center for the Affymetrix U133A GeneChip and biostatistical analyses, Dr. C. Richard Parker, Jr., for his generosity in providing the antihuman HSD3B2 probe, and Dr. Donald D. Samulack for his editorial assistance.

References

1. Michalkiewicz E, Sandrini R, Figueiredo B, et al. Clinical and outcome characteristics of children with adrenocortical tumors: a report from the International Pediatric Adrenocortical Tumor Registry. J Clin Oncol 2004;22:838-45.

2. Parkin DM, Kramarova E, Draper GJ, et al., editors. International Incidence of Childhood Cancer, vol. II, IARC scientific publication no. 144. Vol. 2. Lyon: IARC Press; 1998. p. 500.

3. Zancanella P, Pianovski MA, Oliveira BH, et al. Mitotane associated with cisplatin, etoposide, and doxorubicin in advanced childhood adrenocortical carcinoma: mitotane monitoring and tumor regression. J Pediatr Hematol Oncol 2006;28:513-24.

4. Else T, Hammer GD. Genetic analysis of adrenal absence: agenesis and aplasia. Trends Endocrinol Metab 2005;16:458-68.

5. Mesiano S, Jaffe RB. Developmental and functional biology of the primate fetal adrenal cortex. Endocr Rev 1997;18:378-403.

6. Gicquel C, Bertherat J, Le Bouc Y, Bertagna X. Pathogenesis of adrenocortical incidentalomas and gene- tic syndromes associated with adrenocortical neoplasms. Endocrinol Metab Clin North Am 2000;29:1-13, vii.

7. Malkin D, Li FP, Strong LC, et al. Germ line p53 mutations in a familial syndrome of breast cancer, sar- comas, and other neoplasms. Science 1990;250:1233-8.

8. Birch JM, Hartley AL, Tricker KJ, et al. Prevalence and diversity of constitutional mutations in the p53 gene among 21 Li-Fraumeni families. Cancer Res 1994;54: 1298-304.

9. Kleihues P, Schauble B, zur Hausen A, Esteve J, Ohgaki H. Tumors associated with p53 germline mutations: a synopsis of 91 families. Am J Pathol 1997;150:1-13.

10. Ribeiro RC, Figueiredo B. Childhood adrenocortical tumours. Eur J Cancer 2004;40:1117-26.

11. Li M, Squire JA, Weksberg R. Molecular genetics of Wiedemann-Beckwith syndrome. Am J Med Genet 1998; 79:253-9.

12. Robertson KD. DNA methylation and human disease. Nat Rev Genet 2005;6:597-610.

13. Longui CA, Lemos-Marini SH, Figueiredo B, et al. Inhibin «-subunit (INHA) gene and locus changes in paediatric adrenocortical tumours from TP53 R337H mutation heterozygote carriers. J Med Genet 2004;41: 354-9.

14. Figueiredo BC, Stratakis CA, Sandrini R, et al. Comparative genomic hybridization analysis of adreno- cortical tumors of childhood. J Clin Endocrinol Metab 1999;84:1116-21.

15. Figueiredo BC, Cavalli LR, Pianovski MA, et al. Amplification of the steroidogenic factor 1 gene in

childhood adrenocortical tumors. J Clin Endocrinol Metab 2005;90:615-9.

16. Pianovski MA, Cavalli LR, Figueiredo BC, et al. SF-1 overexpression in childhood adrenocortical tumours. Eur J Cancer 2006;42:1040-3.

17. Gicquel C, Bertagna X, Le Bouc Y. Recent advances in the pathogenesis of adrenocortical tumours. Eur J Endocrinol 1995;133:133-44.

18. Wilkin F, Gagne N, Paquette J, Oligny LL, Deal C. Pediatric adrenocortical tumors: molecular events leading to insulin-like growth factor II gene over- expression. J Clin Endocrinol Metab 2000;85:2048-56.

19. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-AAC(T)) method. Methods 2001;25:402-8.

20. Pounds S, Cheng C. Statistical development and evaluation of microarray gene expression data filters. J Comput Biol 2005;12:482-95.

21. Wilcoxon F. Individual comparisons by ranking methods. Biometrika 1949;1:80-3.

22. Pounds S, Cheng C. Robust estimation of the false discovery rate. Bioinformatics 2006;22:1979-87.

23. Rainey WE, Carr BR, Wang ZN, Parker CR, Jr. Gene profiling of human fetal and adult adrenals. J Endocrinol 2001;171:209-15.

24. Giordano TJ, Thomas DG, Kuick R, et al. Distinct transcriptional profiles of adrenocortical tumors uncov- ered by DNA microarray analysis. Am J Pathol 2003;162: 521-31.

25. Sheskin D. Handbook of parametric and nonpara- metric statistical procedures. 3rd ed. Boca Raton: Chapman & Hall/CRC Press; 2003. p. 1193.

26. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized clinical trials requiring pro- longed observation of each patient. II. Analysis and examples. Br J Cancer 1977;35:1-39.

27. Bange J, Prechtl D, Cheburkin Y, et al. Cancer progression and tumor cell motility are associated with the FGFR4 Arg(388) allele. Cancer Res 2002;62:840-7.

28. de Fraipont F, El Atifi M, Cherradi N, et al. Gene expression profiling of human adrenocortical tumors using complementary deoxyribonucleic Acid micro- arrays identifies several candidate genes as markers of malignancy. J Clin Endocrinol Metab 2005;90:1819-29.

29. Boulle N, Logie A, Gicquel C, Perin L, Le Bouc Y. Increased levels of insulin-like growth factor II (IGF-II) and IGF-binding protein-2 are associated with malig- nancy in sporadic adrenocortical tumors. J Clin Endocrinol Metab 1998;83:1713-20.

30. Martinerie C, Gicquel C, Louvel A, et al. Altered expression of novH is associated with human adreno- cortical tumorigenesis. J Clin Endocrinol Metab 2001;86: 3929-40.

31. Simard J, Ricketts ML, Gingras S, et al. Molecular

biology of the 3B-hydroxysteroid dehydrogenase/85-4 isomerase gene family. Endocr Rev 2005;26:525-82.

32. Bassett MH, Suzuki T, Sasano H, et al. The orphan nuclear receptor NGFIB regulates transcription of 3B- hydroxysteroid dehydrogenase. Implications for the control of adrenal functional zonation. J Biol Chem 2004;279:37622-30.

33. Arrighi I, Bloch-Faure M, Grahammer F, et al. Altered potassium balance and aldosterone secretion in a mouse model of human congenital long QT syndrome. Proc Natl Acad Sci U S A 2001;98:8792-7.

34. Marx C, Wolkersdorfer GW, Brown JW, Scherbaum WA, Bornstein SR. MHC class II expression-a new tool to assess dignity in adrenocortical tumours. J Clin Endocrinol Metab 1996;81:4488-91.

35. Marx C, Bornstein SR, Wolkersdorfer GW, et al. Relevance of major histocompatibility complex class II expression as a hallmark for the cellular differentiation in the human adrenal cortex. J Clin Endocrinol Metab 1997;82:3136-40.

36. Gicquel C, Raffin-Sanson ML, Gaston V, et al. Structural and functional abnormalities at 11p15 are associated with the malignant phenotype in sporadic adrenocortical tumors: study on a series of 82 tumors. J Clin Endocrinol Metab 1997;82:2559-65.

37. Duguay SJ, Jin Y, Stein J, et al. Post-translational processing of the insulin-like growth factor-2 precursor. Analysis of O-glycosylation and endoproteolysis. J Biol Chem 1998;273:18443-51.

38. Weber MM, Fottner C, Schmidt P, et al. Postnatal overexpression of insulin-like growth factor II in transgenic mice is associated with adrenocortical hyperplasia and enhanced steroidogenesis. Endocrinol- ogy 1999;140:1537-43.

39. Mesiano S, Mellon SH, Jaffe RB. Mitogenic action, regulation, and localization of insulin-like growth factors in the human fetal adrenal gland. J Clin Endocrinol Metab 1993;76:968-76.

40. Boulle N, Gicquel C, Logie A, et al. Fibroblast growth factor-2 inhibits the maturation of pro-insulin-like growth factor-II (Pro-IGF-II) and the expression of insulin-like growth factor binding protein-2 (IGFBP-2) in the human adrenocortical tumor cell line NCI-H295R. Endocrinology 2000;141:3127-36.

41. Hornsby PJ, Sturek M, Harris SE, Simonian MH. Serum and growth factor requirements for proliferation of human adrenocortical cells in culture: comparison with bovine adrenocortical cells. In Vitro 1983;19:863-9.

42. Crickard K, Ill CR, Jaffe RB. Control of proliferation of human fetal adrenal cells in vitro. J Clin Endocrinol Metab 1981;53:790-6.

43. Sell S. Cellular origin of cancer: dedifferentiation or stem cell maturation arrest ?. Environ Health Perspect 1993;101 Suppl 5:15-26.

Cancer Research The Journal of Cancer Research (1916-1930) | The American Journal of Cancer (1931-1940)

AAGR American Association for Cancer Research