DNA-sequencing Technologies 
Raymond F. Gesteland, Ph.D. — Investigator 
Dr. Gesteland is also Professor of Human Genetics at the University of Utah School of Medicine and Pro- 
fessor of Biology and Adjunct Professor of Bioengineering at the University of Utah. He received his B.S. 
degree in chemistry and his M.S. degree in biochemistry from the University of Wisconsin. He earned his 
Ph.D. degree in biochemistry from Harvard University, where he studied with J, D. Watson. He was an 
NIH postdoctoral fellow at the Institute de Biologic Moleculaire, Geneva, Switzerland. Dr. Gesteland served 
as Assistant Director and Investigator at Cold Spring Harbor Laboratory, New York, before assuming his 
present responsibilities. 
THE elegant and now widely used methods 
for sequencing DNA are precise and quite 
simple, but labor-intensive, expensive, and 
frustrating for large projects. The current auto- 
mated instruments based on fluorescence de- 
tection methods are clearly a step forward, but 
another enhancement of efficiency is needed to 
bring large projects like sequencing of the hu- 
man genome within range. The heart of se- 
quencing is the remarkably powerful size-frac- 
tionation of DNA chains into ladders with 
one-base resolution. In trying to evolve more 
efficient, ladder-based methods, we have pur- 
sued two approaches. 
Many of the labor-intensive steps of sequenc- 
ing, from growth of bacterial cultures and DNA 
preparation to fractionation of samples on the 
crucial gels, can be made parallel processes by 
multiplexing, as pioneered by George Church 
(HHMI, Harvard Medical School) using chemical 
sequencing and by us using chain termination se- 
quencing. Here, many vectors, each with its in- 
sert of DNA to be sequenced, are combined at an 
early stage and sequenced and fractionated as a 
mixture. Each vector in the mix has the same se- 
quence to prime DNA synthesis into its insert, so 
that all are done in one simple reaction, but each 
has unique sequences as tags just flanking the 
insert. 
The ladders of mixed sequences fractionated 
on a gel are transferred to membranes, which are 
then probed sequentially with radioactive or che- 
miluminescent probes that recognize each of the 
tag sequences, revealing one sequence pattern 
after another. The labor-intensive steps are sim- 
plified by the level of multiplexing (10- to 40- 
fold), and the readout is reduced to repetitive 
rounds of probing that can be readily automated. 
Various architectures for automation and signal 
detection are under development. 
The bottleneck for multiplex sequencing now 
shifts to other fronts. The "back end" problem of 
reading the DNA sequence off of the autoradio- 
grams that result from each daily cycle of probing 
quickly becomes rate-limiting. Automation of 
this crucial sequence-calling is difficult and has 
been getting much attention. We are trying a new 
approach based on communication and signal- 
processing theory. 
The "front end" problem of feeding appro- 
priate sequences into the system is especially 
crucial. In the past, multiplex sequencing has 
been done with random fragments of DNA in the 
hope of reassembling the large sequence by over- 
lapping information from 5- to 10-fold redundant 
sequencing. This becomes increasingly difficult 
with increasing project size, as the gaps are diffi- 
cult to fill. An organized front end seems neces- 
sary. Thus we are developing in vivo approaches 
to generate and map the input inserts. 
Clones to be sequenced are generated in bacte- 
rial cells by turning loose a transposon that in- 
serts itself quite randomly into the large DNA of 
interest. Quick-mapping methods allow assem- 
bly of the minimally overlapping set of clones to 
ensure coverage. The transposons are designed to 
have all the crucial elements for multiplex se- 
quencing in both directions out from the point of 
insert, thus generating twice the reading length. 
Small-scale tests are in progress and should tell us 
if this front end approach will have the expected 
simplification for multiplex sequencing. 
A second approach to sequencing that might 
provide increases in efficiency and speed is capil- 
lary gel electrophoresis. Recent developments 
here and in other laboratories have shown that 
microbore capillaries with an internal diameter 
of 50-70 microns can be used to fractionate DNA 
samples by size, just like the conventional slab 
gels. The virtue of capillaries is that very high 
voltages (10-30 kilovolts) can be used, since 
heat dissipation is so efficient. Molecules 300- 
400 in length can be analyzed in minutes, with a 
resulting increase in resolution. 
The difficulty is that the small gels can accept 
only small amounts of DNA, so that detection 
must be very sensitive. Working with Norm Dovi- 
chi (University of Alberta, Edmonton), we have 
adapted a sheath flow cuvette for detection of 
four fluorescent tags so that detection in the 
157 
