ORLANDO-Genetic "fingerprinting" of high-grade adult soft-tissue sarcomas by oligonucleotide array ("gene chip") analysis revealed a number of distinct tumor subsets and might help point to new therapeutic approaches, Robert G. Maki, MD, PhD, said at the 38th Annual Meeting of the American Society of Clinical Oncology (abstract 1611).
ORLANDOGenetic "fingerprinting" of high-grade adult soft-tissue sarcomas by oligonucleotide array ("gene chip") analysis revealed a number of distinct tumor subsets and might help point to new therapeutic approaches, Robert G. Maki, MD, PhD, said at the 38th Annual Meeting of the American Society of Clinical Oncology (abstract 1611).
"For example, samples that show high expression of certain human growth factor receptors may be targets for imatinib mesylate [Gleevec]," he commented.
Dr. Maki, Dr. Neil Segal, and their colleagues from Memorial Sloan-Kettering Cancer Center used gene chips to examine the activity patterns of 12,500 genes in 51 samples of various adult soft-tissue sarcomas. Sarcomas, which constitute less than 1% of all cancers, are very hetereogenous, he said. Tissue diagnosis is critical. "You want to make sure you’re not dealing with epithelial cancer or something else," Dr. Maki said, adding that treatment does vary among different types of sarcoma.
"Genetic fingerprinting of adult sarcomas will be useful in cases where pathologists disagree about a diagnosis or when the appearance of tumor cells does not conclusively link them to a particular subtype," Dr. Maki said.
His presentation focused extensively on malignant fibrous histiocytoma (MFH), which has been something of a catchall diagnostic category for ambiguous sarcomas. The genetic fingerprinting showed that certain MFH sarcomas are, indeed, a distinct tumor subtype.
The researchers prepared complementary RNAs (cRNAs) from 51 high-grade adult soft-tissue sarcomas. These were hybridized to U95A Affymetrix gene chips, and difference values were generated corresponding to levels of expression of approximately 12,500 genes.
Cluster analysis was done using hierarchical clustering, and data were visualized by multidimensional scaling.
Support vector machine analysis was also done, in which the expression data were used to predict the subtype of sarcoma. This produced a list of specific genes for different sarcoma subtypes.
Strong and Weak Clusters
"The high-grade adult sarcomas that had specific genetic changes tended to cluster very strongly and included synovial sarcoma, gastrointestinal stromal tumor (GIST), clear cell sarcoma, and round cell liposarcoma," he said. "They formed strong groups with multiple genes that were specific for each sarcoma subtype, some of which have therapeutic implications," Dr. Maki reported.
Interestingly, he said, a group of fibrosarcomas clustered, albeit more weakly, with the synovial sarcomas, but this was not the case for all of the fibrosarcomas. "Fibrosarcomas were not well classified using this method," he said. Tumors that did not have specific genetic changes clustered more weakly. Those included the leiomyosarcomas and MFHs.
Support vector machine analysis predicted sarcoma subtype based on the genes expressed. For the strongly clustering tumors, such as GIST, the analysis could easily classify all of the true positives and eliminate all the false positives.
There were other samples in which the true positives were easily classified, but there were still some false positives. "For these samples, adding more genes only added noise to the system," he said.
Another group, represented by fibrosarcomas and MFHs, were more easily analyzed using more genes. "There is not a specific characteristic gene for these subtypes, but once the computer analysis has the gestalt, it can tell you which type you are dealing with," Dr. Maki said.
The researchers also did a genomic clustering of MFH based on the one fibrosarcoma and the MFHs that clustered. "We found that we could use a relatively small number of genes to classify those MFHs," he said.
The investigators concluded that sarcomas with defined genetic changes have distinct expression profiles, and that previously unrecognized subsets of MFH and fibrosarcomas were recognized based on gene expression.
The support vector machine predicted some but not all subtypes, "so maybe we just need to examine a larger number of samples," Dr. Maki said. Finally, he said, the gene profiles provide insight into potential signaling pathways that may have therapeutic value.