The Molecular Genetics of Cancer 
Andrew P. Feinberg, M.D., M.P.H. — Associate Investigator 
Dr. Feinberg is also Associate Professor of Internal Medicine and Human Genetics at the University of 
Michigan Medical School. He received his B.A., M.D., and M.P.H. degrees from the Johns Hopkins Univer- 
sity. He received clinical training at the University of Pennsylvania and Johns Hopkins and did postdoc- 
toral research at the University of California, San Diego, and Johns Hopkins. Before moving to the Uni- 
versity of Michigan, Dr. Feinberg was Assistant Professor of Oncology and Medicine at Johns Hopkins. 
OUR laboratory is studying the molecular 
basis of human cancer. In particular we are 
attempting to elucidate the earliest events that 
convert a normal cell to a malignant cell and the 
role of gene inactivation in cancer development. 
The Earliest Events in Malignant 
Transformation 
We have found that human cancer involves al- 
terations of many genes, even within the same 
tumor. Most of these changes occur relatively late 
and probably do not play a causal role in initia- 
tion of malignancy. Cancer can require several 
decades to develop, including a long premalig- 
nant or benign phase. For example, in colorectal 
cancer, the time course of premalignant disease 
(adenomas or polyps) can be as long as 10 years. 
The only genetic change we have found in early 
adenomas is an alteration in DNA methylation — a 
tissue-specific modification of cytosine, one of 
the four DNA nucleotides. Hypomethylation, the 
loss of methyl groups from cytosine, appears to 
play an important role in normal gene activation 
or expression, and thus altered DNA methylation 
could contribute to the abnormal gene expres- 
sion that characterizes cancer. 
In addition to studying the problem of DNA 
methylation in cancer, we would like to know 
which genes are expressed specifically in cells 
that are transforming. By the time a cell is fully 
malignant, more than 1,000 genes have under- 
gone changes in their expression. If we could de- 
termine which genes are activated in premalig- 
nant cells, we might better understand how 
structural changes in the genome mediate the dra- 
matic changes in cell behavior that define malig- 
nancy, such as unregulated growth, tissue inva- 
sion, and metastasis. 
To approach these two problems, we have de- 
veloped a novel experimental system. We have 
captured cells in culture that have been treated 
with a carcinogen that causes hypomethylation of 
DNA. We have devised a way to tell which cells 
treated with this drug will become malignantly 
transformed, and which will not, before they 
have undergone any of the changes in growth 
properties or appearance that define malignancy. 
In this manner, we have isolated cells that look 
normal but are committed to malignant transfor- 
mation, as well as cells that have been handled in 
the same manner but will not become malignant 
(the ideal control). 
By preparing cDNA libraries (large collections 
of the expressed genes) , we have identified a rela- 
tively small number that are specifically asso- 
ciated with commitment to neoplastic transfor- 
mation. We have also identified several genes that 
are turned off early in that process. We are now 
determining the identity and function of these 
genes, one of which appears to be a novel onco- 
gene with the ability to transform cells when in- 
troduced into them. 
The Role of Gene Inactivation in 
Cancer Development 
One of the most important areas of cancer ge- 
netics is the ascertainment and characterization 
of tumor-suppressor genes — genes whose inacti- 
vation contributes to cancer. Since almost all 
genes are present in two copies in the cell, inacti- 
vation of one copy could be transmitted in fami- 
lies from parent to child. Individuals inheriting 
one bad copy of the gene would be at increased 
risk of developing cancer. Cancer would develop 
from deletion or inactivation of both copies of 
these suppressor genes. 
In the 1970s it was discovered that some chil- 
dren with Wilms' tumor, a childhood kidney 
cancer, are missing a large, microscopically visi- 
ble portion of chromosome 1 1 (in band 1 lpl3). 
Later we found that more subtle gene deletions 
on chromosome 1 1 could be detected in Wilms' 
tumors indirectly, using restriction fragment 
length polymorphisms (RFLPs) , which can distin- 
guish the maternal and paternal alleles, or copies, 
of a given gene (one on each chromosome) . 
During the past year, we have cloned 750 kilo- 
bases of DNA from band 1 lpl3 of chromosome 
1 1, using yeast artificial chromosomes (YACs), in 
collaboration with David Schlessinger and May- 
nard Olson (HHMI, Washington University) and 
131 
