THE MOLECULAR GENETICS OF HUMAN CANCER 
Andrew R Feinberg, M.D., M.RH., Assistant Investigator 
Dr. Feinberg is studying two aspects of the mo- 
lecular genetics of human cancer: 1) early events in 
neoplastic transformation and 2) the role of tumor- 
suppressor genes in carcinogenesis. 
L Early Events in Neoplastic Transformation. 
In previous work, Dr. Feinberg had discovered 
that widespread hypomethylation of the DNA of 
human cancers occurs early in multistep carcino- 
genesis, involving, for example, premalignant col- 
orectal adenomas. To understand the role of DNA 
methylation in the earliest stages of carcinogenesis, 
the laboratory has developed an in vitro model sys- 
tem to capture cells after treatment with an agent 
that causes DNA hypomethylation, but prior to 
neoplastic transformation. The mouse cell line 
C3H/10T1/2 was chosen for these studies because of 
its extremely low spontaneous transformation fre- 
quency and its long latency for transformation after 
carcinogen exposure. Remarkably, 5-a2acytidine an- 
alogues transformed the cells at a frequency of up 
to 10% and did not appear mutagenic. Single cells 
were treated at limiting dilution in microtiter well 
plates with 5-aza-2'-deoxycytidine (5-aza-dCyd) for 
24 h, grown to subconfluence, and each microcol- 
ony was trypsinized into two fractions; one fraction 
was cryopreserved for later study, and the other 
was replated to score for eventual transformation. 
In this manner ancestral pretransformed cells were 
isolated that are indistinguishable by both mor- 
phology and growth properties from non- 
transformed cells but that on continued growth 
eventually become morphologically transformed, 
anchorage independent, and tumorigenic in nude 
mice. This is the first time that phenotypically nor- 
mal cells have been captured prior to transforma- 
tion. Studies of these pretransformed cells will 
permit examination of the earliest events in carci- 
nogenesis and the role of DNA methylation in 
transformation. The laboratory is now preparing 
cDNA libraries from transformants and pre- 
transformants. Two techniques— differential screen- 
ing and subtraction hybridization— are being used 
to identify transcripts specific for commitment to 
transformation. 
As an additional strategy to identify the role in 
transformation of 5-aza-dCyd, high-molecular- 
weight DNA was transfected from 5-aza-dCyd-in- 
duced transformants into untreated recipient cells. 
The laboratory has developed a novel strategy to 
rescue a transfected gene from a syngeneic cell line; 
a cosmid clone that neoplastically transforms re- 
cipient cells has been isolated from 5-aza-dCyd- 
treated cells. This gene is now being characterized. 
II. Recessive Tumor Genes. 
Dr. Feinberg previously observed loss of 
heterozygosity on chromosome 11 in sporadically 
occurring Wilms' tumors, consistent with germline 
deletions of lip 13 in some patients and with 
Knudson's model of loss of two alleles of a reces- 
sive gene in Wilms' tumorigenesis. In the first re- 
port of a putative tumor-suppressor gene in a com- 
mon human malignancy, Dr. Feinberg and his 
co-workers found that bladder carcinomas also lose 
heterozygosity on lip. Recently the laboratory has 
been studying two malignancies, colorectal cancer 
and Wilms' tumor, for potential recessive tumor 
genes. Colorectal cancer offers the opportunity to 
time genetic changes during multistep carcinogene- 
sis, since normal mucosa, premalignant polyps, and 
carcinomas can be studied at progressive patho- 
logic stages. With collaborator Gilles Thomas (In- 
stitut Curie, Paris), the laboratory observed loss of 
heterozygosity on chromosomes 5, 17, and 18, 
often within the same tumor. Losses on chromo- 
somes 5, 17, and 18 were seen in early carcinomas, 
and chromosome 17 and 18 losses were also associ- 
ated with tumor progression. These data suggest 
that altered gene dosage at multiple loci, rather 
than a simple Mendelian model, may be involved in 
multistep colorectal carcinogenesis. 
It was also found that chromosome 17 and 18 
losses are not independent, and they are also signif- 
icantly associated with changes in ploidy. Cancers 
of the left side of the colon were much more likely 
to show allelic losses than cancers of the right side, 
even though the frequency of mutation at c-Ki-m5 
was constant. These data suggest a possible mecha- 
nistic association between losses on multiple chro- 
mosomes and suggest that the losses might account 
for the known biological differences between left- 
and right-sided colorectal tumors. The laboratory is 
now determining which molecular changes are spe- 
cifically associated with defined stages of tumor 
progression or prognostic outcome. 
The laboratory also reexamined Wilms' tumor 
(considered a paradigm of Knudson's model), in 
Continued 
191 
