Chromosomal aberration patterns. A, Among 31 primary choroidal melanomas as determined by cytogenetic characterization (GeneChip Human 250K NSPI Mapping Arrays; Affymetrix, Santa Clara, California) and copy analysis software (CNAT version 4.0.1, Affymetrix). Hierarchical clustering using 1 minus Pearson product moment correlation matrix sorting revealed 2 cytogenetic groups identified by chromosome 3 loss or by chromosome 6p gain. Red boxes indicate whole-arm loss or near–whole-arm loss. Green boxes represent whole-arm gain or near–whole-arm gain. Boxes labeled 2× indicate the detection of 4 or more copies. Four of the primary tumors highlighted in yellow resulted in liver metastasis. B, Among 4 primary tumors that have thus far resulted in liver metastasis. Each primary tumor had chromosome 3 loss and chromosome 8q gain. Losses in chromosomes 6q, 8p, and 16q were frequent aberrations.
McCannel TA, Burgess BL, Rao NP, Nelson SF, Straatsma BR. Identification of Candidate Tumor Oncogenes by Integrative Molecular Analysis of Choroidal Melanoma Fine-Needle Aspiration Biopsy Specimens. Arch Ophthalmol. 2010;128(9):1170-1177. doi:10.1001/archophthalmol.2010.180
To report integrative molecular analysis of choroidal melanoma fine-needle aspiration biopsy specimens to identify candidate tumor oncogenes.
Thirty-one choroidal melanoma fine-needle aspiration biopsy specimens were analyzed using cytopathologic diagnosis of melanoma, fluorescence in situ hybridization for chromosome 3, cytogenetic characterization (GeneChip Human 250K NSPI Mapping Arrays; Affymetrix, Santa Clara, California), and gene expression profiles (GeneChip Human Genome U133 Plus 2.0 Arrays, Affymetrix). These analyses were performed by clustering of cytogenetic aberrations, sorting by chromosome 3 loss and chromosome 6p gain, and comparing gene expression profiles in chromosome 3 loss– and chromosome 6p–gain tumors to identify genes with differential expression based on cytogenetic characteristics.
Of 31 choroidal melanoma biopsy specimens included in this study, 19 tumors had chromosome 3 loss, and 12 tumors without chromosome 3 loss had chromosome 6p gain. Comparative RNA analysis for these 2 groups revealed 49 genes with greater than 4-fold higher expression and 31 genes with greater than 4-fold lower expression in chromosome 3–loss tumors relative to chromosome 6p–gain tumors.
Molecular analysis of choroidal melanoma fine-needle aspiration biopsy specimens demonstrated 2 cytogenetically distinct groups characterized by chromosome 3 loss or chromosome 6p gain. In chromosome 3–loss melanomas relative to chromosome 6p–gain melanomas, integrative RNA analysis revealed genes with higher expression and lower expression and identified several genes that have not been reported in previous studies.
Genes differentially expressed between chromosome 3–loss and chromosome 6p–gain melanomas may provide new knowledge about the biologic nature of choroidal melanoma and may contribute to the development of targeted therapies.
Uveal melanoma, arising from the choroid, ciliary body, and iris of the eye, is the most common primary intraocular malignant neoplasm in adults. Despite improvements in diagnosis and local tumor control, almost half of the patients with uveal melanoma develop metastasis during long-term follow-up, and melanoma-related mortality rates have been virtually unchanged for decades.1,2
As predictors of metastasis, clinical and histopathologic factors have been superceded by cytogenetic and molecular characteristics of the primary tumor. Loss of heterozygosity in chromosome 3, which occurs in approximately 50% of uveal melanomas, is associated with an increased risk for the development of metastasis and for melanoma-related death.3- 6 Loss of chromosome 3 is usually associated with multiplication of chromosome 8, 8q, or parts of 8q,3,4 and additional copies of chromosome 8q are significantly related to reduced survival.4
Tschentscher et al5 reported that unsupervised hierarchical cluster analysis of gene expression data in primary uveal melanoma treated by enucleation identified 2 distinct entities characterized by the presence or absence of monosomy 3. In a larger series of primary uveal melanomas treated by enucleation, Onken et al6 also used hierarchical cluster analysis of gene expression to show 2 groups of uveal melanomas (class 1 and class 2) that predicted greater melanoma-related death in patients with class 2 melanomas. In subgroup analysis, gain of chromosome 6p was associated with the better prognosis of class 1 melanomas, while loss of chromosome 3 was associated with a poor prognosis.6
In a further series of choroidal melanomas treated by enucleation, van Gils et al7 combined clinical outcome, cytogenetic data, RNA hybridization, and unsupervised hierarchical cluster analysis to place tumors into 2 groups based on gene expression and prognosis. Chromosome 6p gain corresponded to a favorable prognosis, and chromosome 3 loss (often associated with 6p gain, 8p loss, and 8q gain) was associated with metastasis and shortened survival.
In 2007, our group described 59 choroidal melanomas subjected to fine-needle aspiration biopsy (FNAB) immediately before globe-conserving iodine 125 (125I) plaque brachytherapy (57 patients) or immediately after enucleation (2 patients).8 These patients were analyzed using fluorescence in situ hybridization (FISH) and high-density whole-genome mapping arrays to show that chromosomal aberrations were aligned into 2 discrete groups: tumors with chromosome 3 loss and tumors with chromosome 6p gain.
Focused on choroidal and ciliary body melanoma, this study presents an integrative molecular analysis of choroidal melanoma FNAB specimens. Our analysis combines hierarchical clustering of chromosomal aberrations, sorting by chromosome 3 loss and chromosome 6p gain, and comparing of RNA expression profiles to identify candidate oncogenes.
Among 82 patients with primary choroidal melanoma subjected to FNAB immediately before 125I plaque brachytherapy or immediately (<5 minutes) after enucleation between May 16, 2006, and July 22, 2008, 31 eyes (31 patients) met the following criteria for inclusion in the study: (1) voluntary consent to participate in the research study, (2) cytopathologic diagnosis of melanoma, (3) sufficient FISH material for the chromosome 3 centromere, (4) adequate material for nucleic acid analyses, (5) successful analysis by cytogenetic characterization (GeneChip Human 250K NSPI Mapping Arrays; Affymetrix, Santa Clara, California), (6) completed gene expression profiles (GeneChip Human Genome U133 Plus 2.0 Arrays, Affymetrix), and (7) a positive finding of cytogenetic aberration by copy number analysis of the mapping array data. Any patient whose biopsy sample was deficient in any 1 of these 7 criteria was excluded from the study. The most common reasons for exclusion were insufficient RNA (n = 20), inadequate material by FISH (n = 20), and undetectable cytogenetic abnormality in chromosome 3 or 6 (n = 12).
This research was approved by the Institutional Review Board of the University of California, Los Angeles (UCLA). Work was in compliance with the Health Insurance Portability and Accountability Act of 1996. Before treatment, evaluation of each patient included comprehensive ophthalmic examination, ultrasonography, photography, optical coherence tomography, and fluorescein angiography. All patients underwent systemic evaluation, usually by an oncologist at the Jonsson Comprehensive Cancer Center at UCLA, and were offered psychological support by a clinical psychologist or social worker with particular expertise in choroidal melanoma.9 Following treatment, patients were evaluated every 6 months or more frequently when indicated.
The technique of intraoperative transscleral FNAB and 125I plaque brachytherapy has been described elsewhere.8,10,11 In brief, FNAB was performed using a 30-gauge needle via a tangential transscleral approach. The biopsy specimen underwent processing for cytologic evidence of malignant melanoma, FISH for chromosome 3, nucleic acid analyses, and cell culture.
An aspirate was smeared on glass slides in the operating room. They were immediately fixed in 95% ethanol, stained with hematoxylin-eosin, and subsequently evaluated for cytologic evidence of melanoma by the pathologist.12
To corroborate our genome-mapping array findings, FISH was performed. In a standard manner described elsewhere,8,10,11 a Spectrum Orange–conjugated probe (Abbott-Vysis, Des Plaines, Illinois) specific for the centromeric region of chromosome 3 was used for interphase FISH. From 100 to 300 hybridization signals were manually counted in nonoverlapping nuclei of cells under a fluorescence microscope (Zeiss Axiophot; Zeiss, Jena, Germany) equipped with a triple filter (4′,6′-diamidino-2-phenylindole, fluorescein isothiocyanate, and Texas Red) at the UCLA Clinical Cytogenetics Laboratory.
In the operating room, biopsy aspirates were expelled directly into a cell reagent (RNAProtect Cell Reagent; QIAGEN Sciences Inc, Valencia, California) within 15 seconds of collection, and the needle hubs were rinsed with the same reagent to maximize recovery. Stabilized pooled aspirates were pelleted, and DNA and RNA were simultaneously isolated from the pooled sample using a kit (AllPrep DNA/RNA Mini Kit, QIAGEN Sciences Inc) per the manufacturer's instructions.
Isolated DNA was quantified using a commercial product (ND-1000; NanoDrop, Wilmington, Delaware). No DNA sample was subjected to whole-genome amplification techniques. The DNA copy number was assessed using mapping arrays (GeneChip Human 250K NSPI). Probe preparation, hybridization, and reading were performed by the UCLA DNA Microarray Core Facility according to the standard 96-well protocol provided by Affymetrix. Copy number variation was computed using commercially available software (CNAT version 4.0.1, Affymetrix).
Chromosomal aberration frequency analysis was performed using Fisher exact test. Chromosomal aberration clustering for each biopsy specimen was performed using 1 minus the Pearson product moment correlation matrix as an input of the mean linkage hierarchical clustering to arrive at a dendrogram clustering tree.
The RNA was quantified on a spectrophotometer (NanoDrop) and analyzed on a bioanalyzer (2100; Agilent, Santa Clara) for integrity. For inclusion in the study, biopsy specimen RNA had to meet the following criteria: (1) a yield of greater than 400 ng of total RNA, (2) a ratio of A260 to A280 (the optical spectrometer measurements of absorbance at the wavelengths of 260 nm and 280 nm) exceeding 1.9, and (3) an RNA integrity number of at least 8.5 as determined using the bioanalyzer. Most samples required 1 round of amplification, and prepared probes were hybridized to gene expression profiles (GeneChip Human Genome U133 Plus 2.0 Arrays) at the UCLA DNA Microarray Core Facility using the standard Affymetrix protocol.
To determine key overexpressed and underexpressed genes that were differentially expressed in chromosome 3–loss and chromosome 6p–gain melanomas, integrative comparative and statistical analysis was performed. RNA expression data CEL files were generated using Affymetrix software.13 R version 2.6 (http://www.r-project.org/) and Bioconductor (http://www.bioconductor.org/) were accessed to normalize the CEL files using the Celsius microarray database. Celsius was accessed to quantify and normalize each CEL file relative to 50 random samples of the same platform (GeneChip Human Genome U133 Plus 2.0 Arrays) using the default variables of robust multi-array average.
Normalized data were imported into dChip (http://biosun1.harvard.edu/complab/dchip/) for differential expression analysis.14,15 Two-group comparative analysis was performed using selection variables of at least 2-fold change and t test at P < .05. Samples were permuted 100 times through dChip to assess the false discovery rate.
The 31 eyes with primary choroidal melanoma were from 31 patients (22 men and 9 women), with a mean age of 64 years (median age, 66 years; age range, 26-86 years). Before treatment, the mean greatest basal diameter of choroidal melanomas was 13.2 mm (range, 6.5-19.3 mm), and the mean apical height was 6.7 mm (range, 3.0-12.4 mm). Twenty-nine melanomas were treated with globe-conserving brachytherapy, and 2 melanomas were treated with enucleation (Table 1).
Thirty-one patients with choroidal melanoma had no clinical or radiographic evidence of melanoma metastasis or other malignant neoplasm at the time of treatment. Follow-up of 31 patients after brachytherapy or enucleation ranged from 6 to 42 months (median follow-up, 30 months; mean follow-up, 26.1 months). During that interval, 4 patients developed clinical evidence of liver metastasis, and 2 of these patients died of liver failure.
In 31 FNAB specimens from primary choroidal melanoma, chromosomal aberrations were identified using mapping arrays (GeneChip Human 250K NSPI) and FISH for the centromere of chromosome 3. Tumors clustered into 2 distinct categories based on the fundamental aberration of chromosome 3 loss or chromosome 6p gain (Figure, A). Nineteen tumors had chromosome 3 loss, and 12 tumors had chromosome 6p gain in the absence of chromosome 3 loss. Among 19 melanomas with chromosome 3 loss, additional aberrations included 14 with chromosome 8q gain. Of these 14, 10 also had 8p loss; 3 had 6p gain in addition to chromosome 3 loss. Among 12 melanomas with chromosome 6p gain in the absence of chromosome 3 loss, 5 tumors had 6q loss, 5 had 8q gain, and 3 had 6p gain as the sole chromosomal aberration.
Four of 31 patients developed clinical evidence of metastasis during follow-up, and their cytogenetic profiles are shown in the Figure, B. All 4 tumors had chromosome 3 loss and chromosome 8q gain, 3 had 8p loss, 3 had 6q loss, and 1 had 6p gain. The CNAT version 4.0.1 analysis revealed that chromosome gains and losses throughout all samples seemed to be whole arm or almost whole arm in each instance.
dChip 2-group comparative analysis identified 903 genes that met the selection criteria, with a median false discovery rate of 43 (4.8%) and a 90th percentile of 341 (37.8%). Table 2 lists 49 genes with greater than 4-fold higher expression in chromosome 3–loss tumors relative to chromosome 6p–gain tumors. Table 3 lists 31 genes with greater than 4-fold lower expression in chromosome 3–loss tumors relative to chromosome 6p–gain tumors. Several genes are characterized as having robust expression in only 1 of 2 cytogenetic groups. Some of the genes with prominent expressional differences between the 2 groups are associated with the following pathways: G protein–coupled signaling, calcium response pathways, cell adhesion marker expression, retinoic acid response pathways, and regulation of palmitoylation (Table 4).
An increasing body of evidence shows that choroidal melanoma cytogenetic aberrations and gene expression profiles cluster into at least 2 groups characterized (1) by chromosome 3 loss, distinct gene expression profile, and increased risk of metastasis with tumor-related death and (2) by chromosome 6p gain in the absence of chromosome 3 loss, distinct gene expression profile, and decreased risk of metastasis.2- 7,10,16- 20 There may be additional clusters of melanoma with normal chromosome 3 disomy and normal chromosome 6p18 or with aberrations in chromosome 21 as the only cytogenetic change.2
On the premise that chromosome 3 loss or chromosome 6p gain is an early aberration and possibly the earliest manifested chromosomal aberration, we report integrative molecular analysis of 31 choroidal melanomas with chromosome 3 loss or with chromosome 6p gain in the absence of chromosome 3 loss sorted by this fundamental cytogenetic abnormality and analyzed by comparative gene expression; we identify 49 genes with greater than 4-fold higher expression and 31 genes with greater than 4-fold lower expression in chromosome 3–loss melanomas relative to chromosome 6p–gain melanomas (Tables 2 and 3). These genes may have potential roles in G protein–coupled signaling, calcium response pathways, cell adhesion marker expression, retinoic acid response pathways, and regulation of palmitoylation. As such, the differential expression of these genes may contribute to the greater propensity of chromosome 3–loss tumors to form metastatic lesions in patients and may implicate these pathways as targets for studying the progression of this disease.
Two genes with major overexpression and underexpression in chromosome 3–loss melanomas that have garnered our interest are tumor necrosis factor receptor superfamily member 19 (TNFRSF19, or TROY [GenBank NM_148957]) on chromosome 13q12.12 and hedgehog acyltransferaselike (HHATL [GenBank NM_020707.3]) on chromosome 3p22.1. TROY messenger RNA (mRNA) is highly expressed in most chromosome 3–loss tumors, including the 4 tumors with metastatic outcome mentioned herein, and is virtually unexpressed in chromosome 6p gain tumors. Other authors have shown TROY expression to be upregulated in metastasizing uveal7 and cutaneous21 melanomas and to have limited expression elsewhere in adults. TROY mRNA levels have been shown in a murine melanoma cell line to be responsive to retinoic acid.21 As a surface receptor protein, TROY may be a valuable target for clinical applications or for study of the underlying biologic nature of metastasis. HHATL is a structural analogue of an acyltransferase that appends a palmitic acid to sonic hedgehog, enhancing its activity; HHATL possesses the same functional domain but has an amino acid change that prevents it from binding palmitic acid and may serve as a negative regulator of palmitoylation.22 Although highly expressed in most chromosome 6p–gain melanomas, HHATL expression in chromosome 3–loss melanomas is low and is silenced in the metastatic-outcome tumors herein. HHATL resides on chromosome 3p22.1 and may represent an early expressional alteration.
Previous groups reporting about gene expression in choroidal melanoma have found similarities and differences when using various techniques and methods.5- 7,23 Therefore, direct comparisons with our data may be impossible. For example, gene expression profiles of tumors with known metastatic outcome have been compared with those without known metastasis. However, our study derives gene expression based entirely on the cytogenetic differences of chromosome 3 loss and chromosome 6p gain without chromosome 3 loss. Furthermore, our study used the technique of whole-genome single-nucleotide polymorphism array, which provides a higher degree of resolution than array comparative genomic hybridization, which may explain why Tschentscher et al,5 who also derived a list of genes based on cytogenetic differences, did not identify genes similar to those that we found. Perhaps most relevant to the understanding of our data is the 2008 study by van Gils et al,7 with whom we identified several genes in common (Table 5). In their gene list, van Gils et al used metastatic outcome and nonmetastatic outcome to arrive at their findings. All 3 studies contribute knowledge about potentially relevant genes toward understanding the biologic nature of choroidal melanoma.
Noteworthy in our series of 31 choroidal melanomas, patients with chromosome 3–loss tumors had greater risk of metastasis and tumor-related death. Four of 19 patients with chromosome 3–loss tumor developed clinical evidence of liver metastasis, and 2 of these patients died of liver failure within 2 years after primary melanoma treatment. During short-term posttreatment follow-up (range, 6-42 months to date), none of 12 patients with chromosome 6p–gain melanomas have developed clinical evidence of melanoma metastasis.
One limitation of this study is the small sample population of 31 patients with choroidal melanomas. Another limitation is the short posttreatment follow-up period (range, 6-42 months).
This study has several strengths. Our analysis is consistent in the use of FNAB to obtain tumor samples and in the rapid processing of tumor samples to minimize the interval before nucleic acid stabilization. The technique of FNAB allowed inclusion of small- and medium-sized melanomas managed using globe-conserving therapy, which may elucidate early changes in the cytogenetic and expressional characteristics of choroidal melanomas. Processing of the tumor samples in the operating room minimized the time between withdrawal of the sample and preservation of the transcriptomic state in stabilization medium. This may account for our discovery of genes (such as HHATL) that have not been previously reported. As an additional strength, the simultaneous isolation of DNA and RNA verified the presence of melanoma cytogenetic aberration in the sample used to detect RNA expression. With study of DNA and RNA in a single sample, we excluded samples in which the mapping array detected no aberration and avoided the possible inclusion of normal or benign tissue in our expressional analyses. A further strength of this study is the consistent use of mapping arrays (GeneChip Human 250K NSPI) for high-resolution chromosomal aberration detection. The consistent use of mapping arrays have additional analytic benefits through alternative integrative expression analyses involving examination of single-nucleotide polymorphisms, identification of isodisomies, and bioinformatic detection of focal gain and loss.
In summary, we used integrative molecular analysis of choroidal melanoma FNAB specimens to demonstrate 2 cytogenetically and biologically distinct melanoma groups characterized by chromosome 3 loss or by chromosome 6p gain. Comparative RNA analysis in chromosome 3–loss melanomas relative to chromosome 6p–gain melanomas revealed several unreported genes with differential expression that may provide growth and cell survival advantages to chromosome 3–loss melanomas; these results warrant further study.
Correspondence: Tara A. McCannel, MD, PhD, Department of Ophthalmology, David Geffen School of Medicine at UCLA, and Jules Stein Eye Institute, 100 Stein Plaza, Los Angeles, CA 90095-7000 (TMcCannel@jsei.ucla.edu).
Submitted for Publication: October 30, 2009; final revision received January 22, 2010; accepted April 20, 2010.
Author Contributions: Dr McCannel had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None reported.
Funding/Support: This study was supported by the George E. and Ruth Moss Trust and by an unrestricted grant from Research to Prevent Blindness (both to Dr McCannel).
Additional Contributions: Fei Yu, PhD, offered statistical expertise. Ascia Eskin, MSc, provided microarray expertise. Michael Gorin, MD, PhD, gave thoughtful insight in reviewing the manuscript.