Finding Patterns in Malignant Melanoma Gene Expression

Malignant melanoma appears as a darkly pigmented mole or tumor of the skin and takes a variety of forms and shapes, from skinny spindles to fairly flat, spreading cell masses. Melanoma is the most common human cancer and will affect about one in seventy-five Americans at some time in their lives. Although early diagnosis appears to improve the long-term cure rate, about 20% of those diagnosed with melanoma will die from it.

Predicting which melanoma patients will develop invasive, metastatic disease is difficult since metastatic melanomas are indistinguishable by physical or histopathologic examination. Melanoma's only reliable prognostic indicator—its transition from localized to invasive disease—is quantitatively evaluated using the Breslow thickness scale.

In addition, the different "taxonomic" forms of melanoma are considered part of a continuum of cancer types rather than discrete tumors. Molecular markers for melanoma invasiveness are therefore highly desirable, both scientifically and for their potential to improve patient outcomes.

Molecular mechanisms behind metastasis

A team of researchers headed by Dr. Jeffrey Trent at the Cancer Genetics Branch, National Human Genome Research Institute (NHGRI) of the National Institutes of Health in Bethesda, Maryland, is using DNA microarrays to hunt for genetic markers that might distinguish relatively benign melanomas from deadly, invasive cancers.

Lately, the Trent group has focused on the gene Wnt5a, a member of the Wnt family of 38-45-kDa cysteine-rich, hydrophobic signal proteins. Vertebrate Wnt genes emerge in overlapping patterns during embryonic development (gastrula stage) and are expressed in adults in various tissues. Signals from Wnt proteins are interpreted based on the particular Wnt protein expressed, the receptor on which it acts, and on other expressed proteins present when the signal is delivered. Moreover, multiple Wnt proteins can produce different signals than single Wnt proteins.

In tumorigenesis, abnormal Wnt signaling may be divided into three types of transforming ability: Wnt1, Wnt3a, and Wnt7a are highly transforming, while Wnt2, Wnt5b, and Wnt7b comprise the intermediately-transforming group. Wnt4, Wnt5a, and Wnt7b are considered non-transformers.

To explore the molecular mechanisms underlying the metastatic potential of different melanomas and to evaluate Wnt's potential role in melanoma, the Trent team examined expression profiles for 31 melanoma samples and 7 control cell lines that included fibroblast, ovarian adenocarcinoma, and breast epithelium cells.

Researchers isolated messenger RNA directly from melanoma biopsies or tumor cell cultures, prepared labeled complementary DNA from these samples, and hybridized these new structures to a microarray containing 8,150 cDNAs representing 6,971 genes. Hybridization gave quantitative and comparative measurements of the occurrence of each gene, which researchers hoped would allow them to uncover genes that invoke the machinery of metastasis. These genes, and the patterns in which they occurred, could then serve as predictors of tumor aggressiveness.

Microarrays reveal trends in expression

After isolating total RNA, scientists prepared fluorescently-labeled complementary cDNA from the mRNA and hybridized the cDNA to a homemade microarray containing probes for 8,150 cDNAs representing 6,971 unique genes. They then assessed the samples' relative genetic similarity using fluorescently labeled tumor cell mRNA as a reference probe. Relative hybridization intensities were obtained for each sample by measuring the fluorescence from the bound samples relative to the bound reference probe.

Comparing well-measured genes in every possible pair of samples makes it possible to calculate the samples' relative similarity. One test that yielded an immediately interesting result was the cluster affinity search technique (CAST), a method developed by Agilent scientists Zohar Yakhini and Amir Ben Dor. CAST allowed comparisons of the extents of similarity shown by the control samples with the extents of similarity shown by the 31 melanoma samples. Using two similar sets of controls to set a similarity threshold for grouping sample cells, researchers identified a group of 19 similarly-behaving samples and a number of smaller groups. The 31-dimensional dataset was eventually simplified and mapped into a 3D graph in which samples with similar expression patterns were positioned closely and relatively unrelated samples were far apart.

The analysis produced two distinct groupings of samples within the melanoma series. A group of 19 melanoma samples behaved more or less homogeneously with respect to the expression of 22 well-measured, highly differentially expressed genes. That is, they tended to express each gene at approximately the same level. The other group of 12 melanoma samples behaved non-homogeneously across the 22 well-measured genes. The most highly discriminating gene among the 22 is Wnt5a, which is also the best separator of the groups of 19 and 12.

The analysis produced two distinct groupings of samples. Wnt5a (top row) is the most highly discriminating gene within the melanoma series. Reprinted by permission from Nature magazine (Nature 406, 536-540, 03 Aug 2000), copyright 2000 Macmillan Publishers Ltd.

Through this analysis, the Trent group demonstrated that genes behaving homogeneously within one group of cells are part of a specific cellular behavior pattern (or phenotype) possessed by those cells, which is different than the phenotypes of cells outside that group.

Physical tests support microarray data

The behavior of the group of 19—different from the highly motile, aggressive phenotypes—suggested a conclusion: Because the 19 have an expression pattern dissimilar to the other phenotypes, they will not be as aggressive.

To test this conclusion, the team performed three physical tests of invasiveness on representative cell lines from the groups of 19 and 12. In all three tests, the two cell groups behaved quite differently, suggesting that it is possible to predict a phenotype based solely on gene expression. Because gene expression levels differentiate the two groups of cells, it’s logical that the phenotypic differences would be systemic, and that those differences arise from genes implicated in mobility and invasion.

Before this study, scientists knew that Wnt5a was important for melanocyte migration out of the neural crest of developing embryos, providing a clue that Wnt5a might also be implicated in melanoma invasiveness. The team is still trying to understand just how frequently Wnt5a drives motility and aggressiveness in melanoma. Much of how tumors develop appears to be related to residual control systems found in the cell type from which the tumor developed, which explains the Wnt5a-melanoma connection.

Agilent microarrays—reliable and reproducible en mass

When Agilent's relationship with NHGRI began (and throughout the period when the work described here was performed), Dr. Trent and coworkers spotted their own microarrays based on protocols developed at Stanford University by Prof. Pat Brown. These home-brewed microarrays could not be produced reproducibly or in large quantity. Realizing this, and the critical need for microarray products, the group entered its research collaboration with Agilent, with the expectation that both organizations would benefit from commercial, mass-produced microarrays.

Once Agilent learned how to manufacture microarrays in commercial quantities, Dr. Trent's group sent the company PCR product clones which Agilent printed onto microarrays through the identical SurePrint process used to create Agilent's catalog cDNA microarrays today. (Agilent does not normally spot custom cDNA clones, but did so under a five-year Collaborative Research and Development Agreement (CRADA) with Dr. Trent.)

NHRGI scientists continue to spot microarrays manufactured in-house for new projects but use Agilent printed arrays for the bulk of their work. Other Agilent products currently used by the Trent group include the Agilent bioanalyzer, custom deposition microarrays, hybridization buffer and the Agilent DNA microarray scanner.

For more information

For a detailed description of this research, please see "Microarrays Reveal Patterns in Malignant Melanoma Gene Expression" presented in the DNA microarrays section of our Web site. That article is based on "Molecular classification of cutaneous malignant melanoma by gene expression profiling," M. Bittner, et al., Nature 406, 536-540, 03 Aug 2000.

To learn more about Agilent's printed microarray solutions, please visit the DNA Microarrays & Scanner section of our Web site. For additional information about Agilent's full range of products and resources, please go to the Life Sciences/Chemical Analysis main page.