Clinical and Pathophysiological Implications of Chromosomal Alterations in Adrenocortical Tumors: An Integrated Genomic Approach
Olivia Barreau, Aurélien de Reynies, Hortense Wilmot-Roussel, Marine Guillaud-Bataille, Colette Auzan, Fernande René-Corail, Frédérique Tissier, Bertrand Dousset, Xavier Bertagna, Jérôme Bertherat, Eric Clauser, and Guillaume Assié
Institut National de la Santé et de la Recherche Médicale Unité 1016, Institut Cochin (O.B., H.W .- R., M.G .- B., C.A., F.R .- C., F.T., B.D., X.B., J.B., G.A.), 75014 Paris, France; Rare Adrenal Cancer Network COrtico-MEdullo Tumeurs Endocrines-Institut national du cancer (X.B.), 75014 Paris, France; Departments of Endocrinology (X.B., J.B., G.A.) and Pathology (F.T.), Unit of Digestive and Endocrine Surgery (B.D.), Center for Rare Adrenal Diseases (X.B., J.B.), and Oncogenetic Unit (M.G .- B., E.C.), Assistance Publique Hôpitaux de Paris, Hôpital Cochin, 75014 Paris, France; Université Paris Descartes (O.B., H.W .- R., F.T., B.D., X.B., J.B., G.A.), Sorbonne Paris Cité, 75006 Paris, France; Centre National de la Recherche Scientifique Unité Mixte de Recherche 8104 (O.B., H.W .- R., M.G .- B., C.A., F.R .- C., F.T., B.D., X.B., J.B., G.A.), 75014 Paris, France; and Programme Cartes d’Identité des Tumeurs (A.d.R.), Ligue Nationale Contre Le Cancer, 75013 Paris, France
Purpose: Diagnosing malignancy of adrenocortical tumors (ACT) and predicting prognosis in car- cinomas are often challenging. Transcriptome markers have recently emerged, providing prom- ising clinical relevance and improved pathophysiological knowledge. Whether tumoral chromo- somal alterations provide similar information is not known. The aim was to evaluate the diagnostic and prognostic value of chromosomal alterations in ACT and to identify genes associated with benign and malignant tumorigenesis.
Experimental Design: Chromosomal alterations of 86 adenomas and 52 carcinomas were identified by comparative genomic hybridization arrays and/or quantitative PCR.
Results: A larger proportion of the genome is altered in carcinomas compared with adenomas (44 vs. 10%, P = 2.10-10). In adenomas, the 9q34 region, which includes the steroidogenic factor 1 locus, is commonly gained and associated with an overexpression of steroidogenic factor 1 (SF-1). In carcinomas, recurrent gains include chromosomes 5, 7, 12, 16, 19, and 20 and recurrent losses chromosomes 13 and 22. Filtering the genes from these regions according to their expression profile identified genes potentially relevant to adrenocortical tumorigenesis. A diagnostic tool was built by combining DNA copy number estimates at six loci (5q, 7p, 11p, 13q, 16q, and 22q). This tool discriminates carcinomas from adenomas in an independent validation cohort (sensitivity 100%, specificity 83%). In carcinomas, the number of chromosomal alterations was not associated with survival (Cox P = 0.84). A prognostic tool based on tumor DNA was designed with a clustering strategy and validated in an independent cohort.
Conclusions: Chromosomal alterations in ACT discriminate carcinomas from adenomas and contain prognostic information. Chromosomal alterations alter the expression of genes important for tumorigenesis. (J Clin Endocrinol Metab 97: E301-E311, 2012)
A drenocortical tumors (ACT) are common (1-10% on radiological and autopsy series) and are often dis- covered incidentally (1). Most of these tumors are benign. Unlike adenomas, adrenocortical carcinomas are rare, with an estimated annual incidence of two per million (2). Their prognosis is poor, with a 5-yr survival rate not ex- ceeding 40% in most series (3, 4).
Pathological diagnosis of these tumors relies on several histological features and requires a high level of expertise. The histological Weiss score, which analyzes nine histo- pathological features, is most often used (5-8). Most often tumors can be classified as benign (Weiss score of 0 or 1) or malignant (Weiss score of 3 or more). However, the scoring of a tumor varies, depending on the pathologist, and tumors with a Weiss score of 2 cannot be unambig- uously diagnosed (9, 10). Molecular markers have been developed to improve ACT classification: IGF-II is over- expressed in 90% of carcinomas (11-13), and loss of heterozygosity in 17p13 is associated with a shorter dis- ease-free survival (14). More recently, transcriptome anal- yses of ACT have shown a clear discrimination between adenomas and carcinomas (11, 13, 15-18), and a molec- ular predictor of recurrence has been proposed (15).
The strongest prognostic factor of carcinomas is the tumor stage (19, 20). However, important variations oc- cur within stages. For tumors limited to the adrenal (stages I and II according to the European Network for the Study of Adrenal Tumors (ENSAT) classification), recurrence after complete surgery cannot be easily predicted (19). Metastatic tumors also show very variable survival (21). More recently, transcriptome analyses have identified two subgroups of carcinomas with different prognoses (15, 18), suggesting the existence of two types of carcinomas, and a molecular predictor of survival based on two genes has been proposed (15).
Knowledge of the pathophysiology of these tumors is also limited. The clear discrimination in transcriptome be- tween adenomas and carcinomas, and between two prog- nostic groups of carcinomas, must reflect different biol- ogy. Mechanisms behind this difference remain to be discovered.
In the current study, we move forward in the fine mo- lecular classification of ACT by studying the tumor ge- nome in detail. ACT have been studied by conventional comparative genomic hybridization (CGH) since the 1990s. An increased number of chromosomal gains and losses in the carcinomas compared with adenomas has been found (22-27). However, there is little consensus regarding regions of recurrent chromosomal alterations, and the resolution of conventional CGH is low. Two more recent studies with a high-density CGH array have shown
a link between some gains and losses and the overall sur- vival of carcinoma patients (28, 29).
The aim of this study was to assess the diagnosis and prognosis value of DNA alterations in the tumor genome and to identify genes of interest for the tumorigenesis from recurrent chromosomal alterations.
Materials and Methods
Patients
The adrenocortical tumors samples were prospectively col- lected at the Cochin Hospital between 1993 and 2005, snap frozen immediately after surgery, and kept in liquid nitrogen until use. Surgery was indicated for either potential malignancy and/or hormone excess. Metastases were diagnosed by system- atic imaging investigations, mainly abdominal and chest com- puted tomography scans and, when appropriate, bone scintig- raphy and magnetic resonance imaging. Tumor staging was performed using the ENSAT classification (19).
For each patient, diagnosis, tumor weight, size, and classifi- cation were determined by pathological examination. Malig- nancy was assessed according to Weiss criteria: for each tumor, a Weiss score (0-9) was determined by a single experienced pa- thologist (F.T.). Tumors with Weiss scores of 0 or 1 were con- sidered benign. Tumors with a Weiss score of 2 were considered benign in the absence of androgen secretion and of undetermined malignancy in the presence of androgen secretion, as previously reported (30-32). Tumors with Weiss scores of 3 or more were considered malignant. After surgery, patients were followed up at least twice a year for 2 yr and at least annually thereafter.
Informed signed consent for the analysis of the tumor and access to the data collected was obtained from all the patients, and the study was approved by the Institutional Review Board of the Cochin Hospital.
Tumor DNA preparation
Tumor samples (10-50 mg) were powdered under liquid ni- trogen. DNA was extracted and purified by cesium chloride gra- dient ultracentrifugation or proteinase K digestion and ethanol extraction, followed by a clean-up step on columns (QIAGEN, Courtaboeuf, France).
DNA concentrations were determined using Nano Drop ND- 1000 (Nyxor Biotech, Paris, France).
CGH arrays
Fifty-nine tumors (38 adenomas and 21 carcinomas) were studied by CGH array (Supplemental Fig. 1, published on The Endocrine Society’s Journals Online web site at http:// jcem.endojournals.org). Integra Chip bacterial artificial chro- mosome pangenomic arrays (IntegraGen, Evry, France) were used, covering the genome with 4434 bacterial artificial chro- mosome clones, with a median distance of 600 kb between clones. Tumor DNA (500-600 ng) was labeled by random priming with Cy5 (Bioprime labeling kit; Invitrogen, Carls- bad, CA) and then mixed to Cy3 control DNA matched for gender. DNA labeling and hybridization are described in Sup- plemental Information.
The arrays were scanned using a ScanArray 4000 (Packard Bioscience, Perkin-Elmer, Waltham, MA). Images were quanti- fied using UCSF Spot 2.1cc (33).
The full data set can be downloaded from the European Bioin- formatics Institute (http://www.ebi.ac.uk/arrayexpress/, exper- iment E-MTAB-659).
CGH data analysis
Determination of DNA copy number
Raw log2 ratio feature values were filtered using a signal-to- noise threshold of 2.0 for the reference channel. The remaining values were normalized using the lowess within-print tip group method (34).
Smoothing and segmentation was performed with the tiling Array R (Bioconductor) package (47).
The level of normal genomic ratios (copy number of 2) was determined for each sample as the first mode of the distribution of the smoothed log2 ratio values across all chromosomes. The SD of the normalized log2 ratio values in the smoothed log2 ratio segments are used as a threshold for determining gains or losses. The estimate of DNA copy number is explained in Supplemental Methods.
Mapping of gains and losses along the genome
The proportion of tumors with a gain or a loss was deter- mined for each chromosomal region. The relevance of each al- teration was further characterized by combining the frequency and the amplitude of the alteration by a method called GISTIC (Genomic Identification of Significant Targets in Cancer) (35). GISTIC was performed using DNA copy number estimates to allow comparison between samples.
Clustering analysis
To allow the validation of our results in an independent co- hort, independently of the array CGH technology, we used a common genomic scale: the genome was cut in 1 Mb segments with determination of the proportion of gains or losses for each segment. Hierarchical clustering analyses were performed using the stats R package (36), with various metrics (euclidean, Man- hattan, maximum, and binary), methods (ward, single, com- plete, average, mcquitty, median, and centroid), and various sub- sets of 1-Mb segments selected according to their variance among carcinomas (from the 5 to the 100% most variable segments).
Given a dendrogram and a related partition in two classes (clusters), each new sample could be assigned to a specific class by measuring its distance to the centroid of each cluster.
DNA copy number evaluation by quantitative PCR
One hundred thirty-eight tumors were tested by quantitative PCR, including the 59 tumors studied by CGH and 79 additional independent tumors (Supplemental Fig. 1). Quantitative PCR was performed in duplicate, using a LightCycler 480 (Roche Applied Science, Mannheim, Germany). Specific primers target- ing the regions of interest were chosen using PrimerBLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast) (Supplemen- tal Table 1). For each tumor, the DNA copy number was eval- uated by subtracting the cycle threshold in a DNA control from the cycle threshold measured in the tumor. The same control DNA was used for all the measures of all the tumors.
Gene expression
The gene expression data of 54 tumors of the training co- hort (16 carcinomas and 38 adenomas, Supplemental Fig. 1) were previously obtained by HG-U133 Plus 2.0 Affymetrix Gene Chip arrays (Affymetrix, Santa Clara, CA), as previously described (15).
Gene expression in carcinomas was determined in compari- son with adenomas. The 100 most over- and underexpressed genes in carcinomas were selected in comparison with adenomas. The P values were calculated by generating 104 sets of 100 ran- domly selected genes.
Statistical analysis
The comparison between adenomas and carcinomas was per- formed with Welch’s t test for quantitative variables (age, tumor size, Weiss score, ENSAT staging, and the number of events) and with Fisher’s exact test for qualitative variables (gender, presence or absence of secretion, and metastatic status). The correlation between tumor size or Weiss score and the number of chromo- somal alterations was determined by the Pearson correlation test.
Cox univariate models were used to study the link between survival and number of chromosomal alterations and to identify chromosomal gains and losses associated with overall survival.
In the cluster analysis, dendrograms were cut into two clus- ters, yielding partitions of the samples in two groups. For each partition, survival in the two groups was assessed with Kaplan- Meier estimates and compared with the log-rank test.
Analyses were performed using R statistical software [stats and survival packages (36)].
Results
Patient characteristics
The main clinical, hormonal, and pathological obser- vations of the 138 patients presenting with unilateral ad- renocortical tumors are summarized in Table 1.
Chromosomal alterations and malignancy
As expected, carcinoma’s genome is much more altered than adenoma’s, with carcinomas harboring chromo- somal alterations in 44% of the genome, against 10% in adenomas (P= 2.10-10, Fig. 1A). As previously described, the number of alterations is related to tumor size in ACT (Pearson coefficient = 0.62, P = 3.1.10-7, Fig. 1B). How- ever, considering separately adenomas and carcinomas, this correlation becomes less significant in adenomas (co- efficient = 0.5, P = 0.002) and inconsistent in carcinomas (coefficient = 0.01, P = 0.95). Similarly, the number of alterations is related to the Weiss score in ACT (coeffi- cient = 0.71, P = 3.10-10), but this correlation is linked to malignancy and becomes less significant in adenomas (coefficient = 0.54, P = 0.0006) and not significant in carcinomas (coefficient = - 0.13, P = 0.56) (Fig. 1A).
Despite the low number of chromosomal alterations in adenomas, a hot spot in 19q13 is gained in more than 50% of adenomas (Fig. 2). Common minimal region
| Training cohort (n = 59)ª | Validation cohort (n = 79)ª | |||||
|---|---|---|---|---|---|---|
| Adenomas (n = 38) | Carcinomas or tumors of undetermined malignancy (n = 21) | P | Adenomas (n = 48) | Carcinomas or tumors of undetermined malignancy (n = 31) | P | |
| Age (yr) | 0.93 | 0.23 | ||||
| Median | 48 | 44 | 48.5 | 41 | ||
| Range | 23-78 | 15-79 | 22-76 | 15-81 | ||
| Sex | 0.3 | 0.02 | ||||
| Male | 5 | 5 | 2 | 7 | ||
| Female | 33 | 16 | 46 | 24 | ||
| Hormone secretion | 0.24 | 0.006 | ||||
| Yes | 24 | 17 | 27 | 27 | ||
| No | 14 | 4 | 21 | 4 | ||
| Tumor size (cm)b | 7.3e-9 | 3.2e-14 | ||||
| ≤3 | 14 | 0 | 18 | 0 | ||
| >3, ≥6 | 22 | 4 | 28 | 5 | ||
| >6 | 2 | 16 | 2 | 26 | ||
| Weiss score | 5.8e-16 | <2e-16 | ||||
| 0-1 | 37 | 0 | 40 | 0 | ||
| 2 | 1 | 2 | 8 | 2 | ||
| >2 | 0 | 19 | 0 | 29 | ||
| ENSAT staging | 2.4e-5 | 7.8e-7 | ||||
| 1 or 2 | 38 | 12 | 48 | 18 | ||
| 3 or 4 | 0 | 9 | 0 | 13 | ||
| Metastasis | 0.001 | 9.4e-6 | ||||
| Yes | 0 | 6 | 0 | 11 | ||
| No | 38 | 15 | 48 | 20 | ||
| Metastasis or relapse | 1.3e-6 | 5.9e-12 | ||||
| Yes | 0 | 11 | 0 | 21 | ||
| No | 38 | 10 | 48 | 10 | ||
| Tumor-related death | 5.6e-6 | 5.6e-8 | ||||
| Yes | 0 | 10 | 0 | 15 | ||
| No | 38 | 11 | 48 | 16 | ||
| Time to death since surgery | ||||||
| (months) | ||||||
| Median | 10.5 | 21 | ||||
| Range | 1-86 | 2-58 | ||||
| Follow-up (months) | 0.96 | 0.12 | ||||
| Median | 54.5 | 55 | 58 | 34 | ||
| Range | 1-112 | 1-144 | 2-150 | 2-131 | ||
a No significant difference between the two cohorts was found.
b Size of one adrenocortical carcinoma was not available.
gained in this region is from 50,100 to 50,130 Mb, which include apolipoprotein E (APOE) and apolipoprotein C1 (APOC1). Other significant alterations in adenomas were identified by GISTIC, which combines the frequency and the amplitude of the alterations, on 9q34 and 17q25 for gains and 3p24 for losses (Supplemental Fig. 2). The 9q34 region includes the locus of steroidogenic factor 1 (SF-1), which expression is increased in adenomas (Supplemental File 1).
In carcinomas, the most common events included gains on chromosomes 5, 7, 12, 16, 19, and 20 and losses on chromosomes 13 and 22 (Fig. 2). Genes in minimal recur- rent regions identified by GISTIC have been filtered out by comparing their expression in carcinomas containing the gain/loss with the expression in normal adrenocortical samples (Supplemental Fig. 2 and Supplemental File 2).
The list of the selected genes includes the gain combined with overexpression of some well-known or supposed on- cogenes (fibroblast growth factor receptor 4 (FGFR4) in 5q35; cyclin-dependent kinase 2 (CDK2) and cyclin-de- pendent kinase 4 (CDK4) in 12q13; GINS complex sub- unit 2 (Psf2 homolog) (GINS2) in 16q24; TPX2, micro- tubule-associated, homolog (TPX2); cyclin E1 (CCNE1) in 19q13; ubiquitin-conjugating enzyme E2C (UBE2C) and v-myb myeloblastosis viral oncogene homolog (avi- an)-like 2 (MYBL2) in 20q11) and melanocortin receptor 1 (MC1R) (in16q24), which belongs to the same family as ACTH receptor (melanocortin receptor 2) and binds ACTH and the loss combined with the repression of tumor suppressor genes [large tumor suppressor, homolog 2 (Drosophila) (LATS2) in 13q12 and suppression of tu- morigenicity 13 (ST13) in 22q12].
Proportion of regions gained or lost (%) D
Proportion of regions gained or lost (%) OU
80
80
60
60
A
4
:
40
A
40
4
4
A
A
4
4
4
20
20
A
A
44
0
A
0
A
0 1 2 3 4 5 6 7 89
9
0
5
10
15
20
25
Pathology (Weiss score)
Tumor size (cm)
Benign
Malignant
Impact of chromosomal alterations on gene expression
Chromosomal alterations have an impact on gene ex- pression. Overall, genes in gained loci have a mean 1.1- fold increase in expression and genes in lost loci a 0.9-fold decrease (P < 10-16 compared with gene expression in loci neither gained nor lost). The top 100 most overex- pressed genes in carcinomas (vs. adenomas) are in loci that are gained in 47% of carcinomas (vs. 33% of carcinomas using gene sets randomly selected, P = 10-6). Conversely, the top 100 most underexpressed genes in carcinomas are in loci that are lost in only 10% of carcinomas (vs. 11% of carcinomas using gene sets randomly selected, P = 0.69).
In a tumor-by-tumor analysis, 28-70% of the top 100 most overexpressed genes are in gained loci. This proportion is greater than for gene sets randomly se- lected (P values ranging from 0.04 to < 10-4). This occurs in 15 of 16 carcinomas analyzed. In the remain- ing carcinoma, only 3% of the top 100 most overex- pressed genes are in gained loci. This tumor harbors very few gains. Considering the top 100 most underex- pressed genes, only four of the 16 carcinomas present a significant enrichment in genes from lost loci (16-58%, P values ranging from 0.03 to < 10-4).
DNA-based molecular marker of malignancy
The genome is much less altered in adenomas compared with carcinomas. Considering the chromosomal regions both frequently gained or lost in carcinomas but not al- tered in adenomas, the subtraction between the maximal DNA copy number and the minimal copy number of these regions is expected to be much higher in carcinomas than in adenomas. The three most commonly gained regions (5q22.1, 7p12.1, and 16q22.1) and the three most com-
monly lost regions (11p13, 13q31.1, and 22q12.1) were selected. The DNA copy number in these six regions was measured by quantitative PCR and combined by subtracting the maximal copy number of these six regions to the minimal copy number of these six regions. This measurement reflects the maximal difference in DNA copy number among the genome within each tumor and will be high in carci- nomas and low in adenomas. This measurement showed a good discrim- ination between carcinomas and ad- enomas (Fig. 3A), which was con- firmed on an independent cohort of 79 tumors (sensitivity 100%, speci- ficity 83%, Fig. 3, B and C).
The number of chromosomal gains and losses does not have a prognostic value in carcinomas
In tumors either malignant or of undetermined malig- nancy, there is no link between the number of chromo- somal alterations and the overall survival (P = 0.52 using Cox model) (Supplemental Fig. 3).
DNA-based molecular marker for the prognosis of carcinomas
We have tested for each probe of the genome the prog- nostic value associated with its gain or its loss (Supple- mental Fig. 4). Probes with significant prognostic value clustered in the regions 1q and 5q for the gains and 10p and 17q for the losses. However, only a small fraction of the prognosis is explained by these regions, and the com- bination of a limited number of regions did not meet suf- ficient prediction power (data not shown).
To overcome this issue, we considered the prognostic information contained in the whole genome using hierar- chical clustering (see Materials and Methods).
This cluster analysis separates carcinomas into two groups of seven and 14 tumors, with a lower overall survival in the first group (log rank P = 0.003, Fig. 4, A and B).
Validation of this clustering was tested in an indepen- dent cohort of 25 carcinomas with published CGH data (28). Each tumor of this set was assigned to either good or bad prognostic subgroups, depending on the distance with the good and bad prognostic subgroups established in our cohort. On this validation cohort, the difference of overall survival between the two groups is confirmed (log rank P = 0.04, Fig. 4C).
After stratification on the tumor stage (ENSAT) on our cohort, the prognostic value of the cluster is almost sig-
Adenomas
100
Proportion of tumors (%)
with a gain
50
0
with a loss
50
100
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 19 20 2122
X
Y
Carcinomas
100
Proportion of tumors (%)
with a gain
50
0
with a loss
50
100
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2122
X
Ý
Chromosomes
nificant with a Cox hazard ratio of 6.2 (95% confidence interval from 0.95 to 40.5, P = 0.056).
Discussion
In this study we identified the chromosomal gains and losses in carcinomas and adenomas with CGH arrays. As in previous conventional CGH studies (22-24, 26, 27), we found a significant difference in the number of events be- tween carcinomas and adenomas and a correlation be- tween the size and the number of events within the whole set of tumors. However, within the carcinomas only, this correlation is inconsistent and the largest carcinomas do not have more chromosomal alterations than the smallest. Therefore, also considering the bigger size of carcinomas compared with adenomas, the number of chromosomal alterations seems more related to malignancy than to the size of an individual tumor. In adenomas, however, a weak correlation was found. In addition, in adenomas a corre- lation between the number of chromosomal events and the
Weiss score was found. Adenomas with a Weiss score of 1 or 2 have more alterations than adenomas with a Weiss score of 0, and some adenomas have as many alterations as carcinomas. In ACTH-independent macronodular ad- renal hyperplasia (AIMAH), which is another type of be- nign tumor, a recent study (37) also found more chromo- somal gains in the larger nodules than in the smaller nodules from the same patients and identified different pathways associated with AIMAH, depending on the size (metabolic pathways in the smaller nodules, p53 signal- ing, and cancer genes in the larger nodules). This could suggest a stepwise progression from adenomas to carci- nomas, in addition to previous evidence, such as the de- scription of one heterogeneous tumor with a carcinoma developed within an adenoma (38). However, the chro- mosomal alterations in the adenomas with many altera- tions and the carcinomas are different (data not shown). Indeed, gain of the chromosomes 17, 21, and 22 are pres- ent in up to 83% of adenomas with many alterations, and only 5-10% of carcinomas. Therefore, it seems that the malignant transformation of adenomas is not common.
A
Training cohort
B
C
Validation cohort
2.5
2.5
Validation cohort
DNA-based molecular marker of malignancy
DNA-based molecular marker of malignancy
100
2.0
2.0
Disease-free survival (%)
80
1.5
1.5
60
A
A
40
1.0
1.0
1
4
20
4
Predicted adenomas
A
Predicted carcinomas
0.5
4
4
0.5
4
0
Log Rank, p=2.98e-05
0
50
100
150
Time (months)
0.0
0.0
Adenomas (n=38)
Carcinomas (n=21)
Adenomas (n=48)
Carcinomas (n=31)
We have built a predictor of malignancy based on DNA copy number evaluation by quantitative PCR of regions most commonly gained and lost. This predictor showed good dis- crimination between carcinomas and adenomas and was val- idated on an independent cohort. The advantage of DNA markers in comparison with the RNA markers recently de- veloped is the greater stability of DNA, making it more suit- able for clinical practice. Indeed, the sample collection can be performed up to several hours after surgery. We have not tested the use of paraffin-embedded samples, but genomic studies have already used such samples (39).
Chromosomal alterations have a weak but significant global impact on the expression, with a 1.1-fold increase of gained genes and a 0.9-fold decrease of lost genes. However, some re- gions combine frequency and amplitude of gain, and the gain can then be a mechanism for overexpression of an oncogene. A recent study on AIMAH (37) showed that half of the most over- expressed genes in their tumors are gained but that only a few of the most frequently underexpressed genes are lost. Our results are quite similar in our carcinomas, with loci of genes overex- pressed frequently gained but loci of genes underexpressed not frequently lost. Gains seem to have more impact than losses on gene expression.
We identified in adenomas recurring gains in 9q34 and 19q13. Previous works based on conventional CGH also identified some alterations in adenomas, but these alter-
ations vary among the studies (22-27). The gain in 9q is the only consensual event. This gain was also described in childhood adenomas and in both adulthood and child- hood carcinomas (40). This region contains the locus of the SF-1 (or NR5A1) gene, which plays an important role in the control of steroidogenesis and proliferation (41) and has been implicated in the prognosis of carcinomas (42). We found an overexpression of SF-1 in adenomas harbor- ing the 9q34 gain, compared with normal adrenocortical tissue. Another work identified SF-1 overexpression in both adulthood and childhood adenomas and carcinomas but with a greater frequency in children (43). Another gain found in our adenomas is a gain in 2p13, which is near but does not overlap the 2p16 region frequently gained in Car- ney complex tumors (44).
We identified many gains and losses in carcinomas, in agreement with previous reports (22-24, 26-28) (Supple- mental Table 2). Gains of chromosomes 5 and 12 are the most consensual, followed by gains of chromosomes 7, 16, and 20. Losses are both less frequent and less consensual. The discordance between studies might reflect a lower bi- ological significance of nonconsensual chromosomal al- terations, which might correspond to random passenger alterations. Conversely, recurring events might point to relevant genes. We have filtered out genes in these loci of interest by comparing gene expression in carcinomas con-
A
Specific death
Group 1 (n=7)
Group 2 (n=14)
5p
5q
16p, 16q
20p, 20q
4p, 4q
9p, 9q
10p, 10q
6p
6q
1q
15q
14q
1p
22q
11p, 11q
3p, 3q
13q
B
C
100
100
Overall survival (%)
80
Overall survival (%)
80
60
60
predicted good prognosis
group 2 (good prognosis)
predicted bad prognosis
40
group 1 (bad prognosis)
40
20
20
0
Log rank p=0.003
0
Log rank p=0.036
0
20
40
60
80
100
120
140
0
20
40
60
80
100
120
140
Time (months)
Time (months)
taining the gain/loss with the expression in normal adre- nocortical samples. This strategy reveals oncogenes and tumor suppressors, which might be of particular impor- tance for the tumorigenesis of carcinomas.
Regions of prognostic value cluster in 1q and 5q for the gains and 10p and 17q for the losses. However, we could not obtain a correct survival prediction with a limited num- ber of such regions. Similarly, Stephan et al. (28) had previ-
ously shown a link between some gains/losses and overall survival, but this was scattered among the genome. The clus- tering analysis based on our cohort has found prognostic information contained in the whole genome and was vali- dated on the cohort of Stephan et al. Thus prognostic infor- mation of our two data sets are in agreement.
We did not find any relation between the number of chromosomal alterations and the prognosis. In common cancers, a higher number of chromosomal alterations is associated with tumor progression, corresponding to an accumulation of additional events in tumors with poor prognosis (45). This is also described in childhood carci- nomas (46).
We and others previously described two subgroups of carcinomas, identified using an unsupervised clustering strategy based on gene expression (15, 18). This result unravels a major biological difference between these two groups of carcinomas. Considering survival, these two groups show very different prognoses. The question was raised whether these two groups were two stages of the same tumor type or two different types of carcinomas. A majority of our tumors have been studied both by CGH and transcriptome. We compared the number of chromo- somal alterations in the two groups of carcinomas iden- tified by the clustering based on gene expression (15) and found no difference (P = 0.69, Wilcoxon signed rank test). If the tumors of the bad prognosis group corresponded to advanced forms of tumors of the good prognosis group, one would expect a higher number of chromosomal alter- ations in tumors of the bad prognosis group. Indeed, these tumors would accumulate the alterations observed in tu- mors of the good prognosis group, plus the extra ones occurring during the tumor evolution. The lack of differ- ence in the number of chromosomal alteration would therefore indicate that the two groups of carcinomas are not two stages of the same tumor type but rather two distinct types of carcinomas. We compared the two groups of carcinomas of our chromosome-based cluster designed for the prognostic prediction, with the two groups of the transcriptome-based cluster (15). Interestingly, these two clusters meet consistent agreement (P = 0.01 with Fisher’s test).
In conclusion, chromosomal alterations have a strong diagnostic value and also some prognostic value that can be used in clinical practice. The confrontation of CGH and transcriptome reinforces the existence of two subgroups of carcinomas with different pathophysiology and outcome. Recurrent gained or lost regions point to oncogenes and tumor suppressors, which might be important for the tu- morigenesis of carcinomas.
Acknowledgments
We thank Franck Letourneur and Sebastien Jacques (the Genomic Platform) for their technical support; Jacqueline Me- tral and Jacqueline Godet (the Ligue Nationale Contre le Cancer) for the organization of the Cartes d’Identite des Tumeurs pro- gram; the members of our laboratories and the COrtico- MEdullo Tumeurs Endocrines-Institut national du cancer (COMETE-INCa), and the European Network for the Study of Adrenal Tumors (ENSAT) networks for their support and dis- cussions; and all of the staffs of the clinical departments of Co- chin Hospital who were involved in the care of the patients.
Address all correspondence and requests for reprints to: Professor Jérôme Bertherat, Department of Endocrinology, Metabolism, and Cancer, Centre Hospitalier Universitaire Cochin, 27 Rue du Fg. St. Jacques, 75014 Paris, France. E-mail : jerome.bertherat@inserm.fr.
This work was supported in part by the Conny-Maeva Foundation; the Plan Hospitalier de Recherche Clinique Grant AOM06179 (to the COrtico-MEdullo Tumeurs Endo- crines-Institut national du cancer Network) and the Recher- che Translationnelle Direction de l’hospitalisation et de l’organisation des soins (DHOS)/Institut national du cancer (INCa) 2009 Grant RTD09024; the FP7 Program (European Network for the Study of Adrenal Tumors-CANCER) (to the European Network for the Study of Adrenal Tumors net- work). O.B. is a recipient of an Institut National de la Santé et de la Recherche Médicale fellowship and was a recipient of a predoctoral grant support from the Société Française d’Endocrinologie. This work was also supported by the Carte d’Identité des Tumeurs program from La Ligue Contre le Can- cer (France).
Disclosure Summary: The authors indicated no potential con- flicts of interest.
References
1. Bertherat J, Mosnier-Pudar H, Bertagna X 2002 Adrenal inciden- talomas. Curr Opin Oncol 14:58-63
2. Libè R, Fratticci A, Bertherat J 2007 Adrenocortical cancer: patho- physiology and clinical management. Endocr Relat Cancer 14: 13-28
3. Abiven G, Coste J, Groussin L, Anract P, Tissier F, Legmann P, Dousset B, Bertagna X, Bertherat J 2006 Clinical and biological features in the prognosis of adrenocortical cancer: poor outcome of cortisol-secreting tumors in a series of 202 consecutive patients. J Clin Endocrinol Metab 91:2650-2655
4. Fassnacht M, Allolio B 2009 Clinical management of adrenocortical carcinoma. Best Pract Res Clin Endocrinol Metab 23:273-289
5. Schteingart DE, Doherty GM, Gauger PG, Giordano TJ, Hammer GD, Korobkin M, Worden FP 2005 Management of patients with adrenal cancer: recommendations of an international consensus conference. Endocr Relat Cancer 12:667-680
6. Weiss LM, Medeiros LJ, Vickery Jr AL 1989 Pathological features of prognostic significance in adrenocortical carcinoma. Am J Surg Pathol 13:202-206
7. Lau SK, Weiss LM 2009 The Weiss system for evaluating adreno- cortical neoplasms: 25 years later. Hum Pathol 40:757-768
8. Weiss LM 1984 Comparative histologic study of 43 metastasizing
and nonmetastasizing adrenocortical tumors. Am J Surg Pathol 8:163-169
9. Pohlink C, Tannapfe A, Eichfeld U, Schmidt F, Führer D, Paschke R, Koch CA 2004 Does tumor heterogeneity limit the use of the Weiss criteria in the evaluation of adrenocortical tumors? J Endo- crinol Invest 27:565-569
10. Volante M, Buttigliero C, Greco E, Berruti A, Papotti M 2008 Path- ological and molecular features of adrenocortical carcinoma: an update. J Clin Pathol 61:787-793
11. de Fraipont F, El Atifi M, Cherradi N, Le Moigne G, Defaye G, Houlgatte R, Bertherat J, Bertagna X, Plouin PF, Baudin E, Berger F, Gicquel C, Chabre O, Feige JJ 2005 Gene expression profiling of human adrenocortical tumors using complementary deoxyribonu- cleic acid microarrays identifies several candidate genes as markers of malignancy. J Clin Endocrinol Metab 90:1819-1829
12. Gicquel C, Raffin-Sanson ML, Gaston V, Bertagna X, Plouin PF, Schlumberger M, Louvel A, Luton JP, Le Bouc Y 1997 Structural and functional abnormalities at 11p15 are associated with the ma- lignant phenotype in sporadic adrenocortical tumors: study on a series of 82 tumors. J Clin Endocrinol Metab 82:2559-2565
13. Giordano TJ, Thomas DG, Kuick R, Lizyness M, Misek DE, Smith AL, Sanders D, Aljundi RT, Gauger PG, Thompson NW, Taylor JM, Hanash SM 2003 Distinct transcriptional profiles of adreno- cortical tumors uncovered by DNA microarray analysis. Am J Pathol 162:521-531
14. Gicquel C, Bertagna X, Gaston V, Coste J, Louvel A, Baudin E, Bertherat J, Chapuis Y, Duclos JM, Schlumberger M, Plouin PF, Luton JP, Le Bouc Y 2001 Molecular markers and long-term recur- rences in a large cohort of patients with sporadic adrenocortical tumors. Cancer Res 61:6762-6767
15. de Reyniès A, Assié G, Rickman DS, Tissier F, Groussin L, René- Corail F, Dousset B, Bertagna X, Clauser E, Bertherat J 2009 Gene expression profiling reveals a new classification of adrenocortical tumors and identifies molecular predictors of malignancy and sur- vival. J Clin Oncol 27:1108-1115
16. Slater EP, Diehl SM, Langer P, Samans B, Ramaswamy A, Zielke A, Bartsch DK 2006 Analysis by cDNA microarrays of gene expression patterns of human adrenocortical tumors. Eur J Endocrinol 154: 587-598
17. Velázquez-Fernández D, Laurell C, Geli J, Höög A, Odeberg J, Kjell- man M, Lundeberg J, Hamberger B, Nilsson P, Bäckdahl M 2005 Expression profiling of adrenocortical neoplasms suggests a molec- ular signature of malignancy. Surgery 138:1087-1094
18. Giordano TJ, Kuick R, Else T, Gauger PG, Vinco M, Bauersfeld J, Sanders D, Thomas DG, Doherty G, Hammer G 2009 Molecular classification and prognostication of adrenocortical tumors by tran- scriptome profiling. Clin Cancer Res 15:668-676
19. Fassnacht M, Johanssen S, Quinkler M, Bucsky P, Willenberg HS, Beuschlein F, Terzolo M, Mueller HH, Hahner S, Allolio B 2009 Limited prognostic value of the 2004 International Union Against Cancer staging classification for adrenocortical carcinoma: pro- posal for a revised TNM classification. Cancer 115:243-250
20. Macfarlane DA 1958 Cancer of the adrenal cortex: the natural his- tory, prognosis, and treatment in a study of fifty-five cases. Ann R Coll Surg Engl 23:155-186
21. Assié G, Antoni G, Tissier F, Caillou B, Abiven G, Gicquel C, Leb- oulleux S, Travagli JP, Dromain C, Bertagna X, Bertherat J, Schlum- berger M, Baudin E 2007 Prognostic parameters of metastatic ad- renocortical carcinoma. J Clin Endocrinol Metab 92:148-154
22. Dohna M, Reincke M, Mincheva A, Allolio B, Solinas-Toldo S, Lichter P 2000 Adrenocortical carcinoma is characterized by a high frequency of chromosomal gains and high-level amplifications. Genes Chromosomes Cancer 28:145-152
23. Kjellman M, Kallioniemi OP, Karhu R, Hög A, Farnebo LO, Auer G, Larsson C, Bäckdahl M 1996 Genetic aberrations in adrenocor- tical tumors detected using comparative genomic hybridization cor- relate with tumor size and malignancy. Cancer Res 56:4219-4223
24. Sidhu S, Marsh DJ, Theodosopoulos G, Philips J, Bambach CP,
Campbell P, Magarey CJ, Russell CF, Schulte KM, Röher HD, Del- bridge L, Robinson BG 2002 Comparative genomic hybridization analysis of adrenocortical tumors. J Clin Endocrinol Metab 87: 3467-3474
25. Zhao J, Roth J, Bode-Lesniewska B, Pfaltz M, Heitz PU, Kommi- noth P 2002 Combined comparative genomic hybridization and genomic microarray for detection of gene amplifications in pulmo- nary artery intimal sarcomas and adrenocortical tumors. Genes Chromosomes Cancer 34:48-57
26. Zhao J, Speel EJ, Muletta-Feurer S, Rütimann K, Saremaslani P, Roth J, Heitz PU, Komminoth P 1999 Analysis of genomic altera- tions in sporadic adrenocortical lesions. Gain of chromosome 17 is an early event in adrenocortical tumorigenesis. Am J Pathol 155: 1039-1045
27. Gruschwitz T, Breza J, Wunderlich H, Junker K 2010 Improvement of histopathological classification of adrenal gland tumors by ge- netic differentiation. World J Urol 28:329-334
28. Stephan EA, Chung TH, Grant CS, Kim S, Von Hoff DD, Trent JM, Demeure MJ 2008 Adrenocortical carcinoma survival rates corre- lated to genomic copy number variants. Mol Cancer Ther 7:425- 431
29. Szabó PM, Tamási V, Molnár V, Andrásfalvy M, Tömböl Z, Farkas R, Kövesdi K, Patócs A, Tóth M, Szalai C, Falus A, Rácz K, Igaz P 2010 Meta-analysis of adrenocortical tumour genomics data: novel pathogenic pathways revealed. Oncogene 29:3163-3172
30. d’Alva CB, Abiven-Lepage G, Viallon V, Groussin L, Dugue MA, Bertagna X, Bertherat J 2008 Sex steroids in androgen-secreting adrenocortical tumors: clinical and hormonal features in compari- son with non-tumoral causes of androgen excess. Eur J Endocrinol 159:641-647
31. Levine AC, Mitty HA, Gabrilove JL 1988 Steroid content of the peripheral and adrenal vein in Cushing’s syndrome due to adreno- cortical adenoma and carcinoma. J Urol 140:11-15
32. Terzolo M, Alì A, Osella G, Reimondo G, Pia A, Peretti P, Paccotti P, Angeli A 2000 The value of dehydroepiandrosterone sulfate mea- surement in the differentiation between benign and malignant ad- renal masses. Eur J Endocrinol 142:611-617
33. Jain AN, Tokuyasu TA, Snijders AM, Segraves R, Albertson DG, Pinkel D 2002 Fully automatic quantification of microarray image data. Genome Res 12:325-332
34. Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J, Speed TP 2002 Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res 30:e15
35. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, Vivanco I, Lee JC, Huang JH, Alexander S, Du J, Kau T, Thomas RK, Shah K, Soto H, Perner S, Prensner J, Debiasi RM, Demichelis F, Hatton C, Rubin MA, Garraway LA, Nelson SF, Liau L, Mischel PS, Cloughesy TF, Meyerson M, Golub TA, Lander ES, Mellinghoff IK, Sellers WR 2007 Assessing the significance of chromosomal aberrations in cancer: methodology and application to glioma. Proc Natl Acad Sci USA 104:20007-20012
36. Team RDC 2011 R: a language and environment for statistical com- puting R Foundation for statistical computing. Vienna: R Founda- tion for Statistical Computing
37. Almeida MQ, Harran M, Bimpaki EI, Hsiao HP, Horvath A, Chea- dle C, Watkins T, Nesterova M, Stratakis CA 2011 Integrated genomic analysis of nodular tissue in macronodular adrenocortical hyperplasia: progression of tumorigenesis in a disorder associated with multiple benign lesions. J Clin Endocrinol Metab 96:E728- E738
38. Gaujoux S, Tissier F, Groussin L, Libé R, Ragazzon B, Launay P, Audebourg A, Dousset B, Bertagna X, Bertherat J 2008 Wnt/B- catenin and 3’,5’-cyclic adenosine 5’-monophosphate/protein ki- nase A signaling pathways alterations and somatic ß-catenin gene mutations in the progression of adrenocortical tumors. J Clin En- docrinol Metab 93:4135-4140
39. Tuefferd M, de Bondt A, Van den Wyngaert I, Talloen W, Göhl-
mann H 2011 Microarray profiling of DNA extracted from FFPE tissues using SNP 6.0 Affymetrix platform. Methods Mol Biol 724: 147-160
40. Figueiredo BC, Cavalli LR, Pianovski MA, Lalli E, Sandrini R, Ri- beiro RC, Zambetti G, DeLacerda L, Rodrigues GA, Haddad BR 2005 Amplification of the steroidogenic factor 1 gene in childhood adrenocortical tumors. J Clin Endocrinol Metab 90:615-619
41. Doghman M, Karpova T, Rodrigues GA, Arhatte M, De Moura J, Cavalli LR, Virolle V, Barbry P, Zambetti GP, Figueiredo BC, Heck- ert LL, Lalli E 2007 Increased steroidogenic factor-1 dosage triggers adrenocortical cell proliferation and cancer. Mol Endocrinol 21: 2968-2987
42. Sbiera S, Schmull S, Assie G, Voelker HU, Kraus L, Beyer M, Ra- gazzon B, Beuschlein F, Willenberg HS, Hahner S, Saeger W, Ber- therat J, Allolio B, Fassnacht M 2010 High diagnostic and prog- nostic value of steroidogenic factor-1 expression in adrenal tumors. J Clin Endocrinol Metab 95:E161-E171
43. Almeida MQ, Soares IC, Ribeiro TC, Fragoso MC, Marins LV,
Wakamatsu A, Ressio RA, Nishi MY, Jorge AA, Lerario AM, Alves VA, Mendonca BB, Latronico AC 2010 Steroidogenic factor 1 over- expression and gene amplification are more frequent in adrenocor- tical tumors from children than from adults. J Clin Endocrinol Metab 95:1458-1462
44. Matyakhina L, Pack S, Kirschner LS, Pak E, Mannan P, Jaikumar J, Taymans SE, Sandrini F, Carney JA, Stratakis CA 2003 Chromo- some 2 (2p16) abnormalities in Carney complex tumours. J Med Genet 40:268-277
45. Kallioniemi A 2008 CGH microarrays and cancer. Curr Opin Bio- technol 19:36-40
46. Loncarevic IF, Hering A, Posorski N, Linden T, Hoyer H, Bucsky P 2008 Number of genomic imbalances correlates with the overall survival for adrenocortical cancer in childhood. Pediatr Blood Can- cer 51:356-362
47. Huber W, Toedling J, Steinmetz LM 2006 Transcript mapping with high-density oligonucleotide tiling arrays. Bioinformatics 22:1963- 1970
THE ENDOCRINE SOCIETY®
Challenge your diagnostic skills with a one-of-a-kind self-assessment resource, Diagnostic Dilemmas: Images in Endocrinology, edited by Leonard Wartofsky, M.D. www.endo-society.org/dilemmas