Proc. Natl. Acad. Sci. USA
Vol. 92, pp. 4641-4645, May 1995
Biochemistry
Rolling replication of short DNA circles
ANDREW FIRE* AND SI-QUN XU
Department of Embryology, Carnegie Institution of Washington, 115 West University Parkway, Baltimore, MD 21210
Communicated by Donald D. Brown, Carnegie Institution of Washington, Baltimore, MD, February 10, 1995
ABSTRACT
Natural genes and proteins often contain
tandemly repeated sequence motifs that dramatically increase
physiological specificity and activity. Given the selective value
of such repeats, it is likely that several different mechanisms
have been responsible for their generation. One mechanism
that has been shown to generate relatively long tandem repeats
(in the kilobase range) is rolling circle replication. In this
communication, we demonstrate that rolling circle synthesis
in a simple enzymatic system can produce tandem repeats of
monomers as short as 34 bp. In addition to suggesting possible
origins for natural tandem repeats, these observations pro-
vide a facile means for constructing libraries of repeated
motifs for use in "in vitro evolution" experiments designed to
select molecules with defined biological or chemical properties.
Analysis of naturally occurring macromolecular sequence has
revealed repetitive structure at a variety of levels (1, 2).
Particularly relevant to gene expression and replication are sets
of short sequence motifs that often occur in multiple copies
around promoter/enhancer regions and replication origins (3).
The repetition of motifs within a control region has been
shown in many cases to allow individual trans-acting factors to
exert additive and/or cooperative effects; this design can
improve the specificity of a control mechanism by increasing
the signal of appropriate activity while decreasing the possi-
bility of fortuitous inappropriate activity (4). Requirements for
repeated sequence motifs have also been found in character-
izing the activities of specific RNA (5-8) and protein (9-11)
functions.
In investigating structure-function relationships in vitro and
in vivo, several researchers have used strategies that involve the
production of a large library of random sequences followed by
selection for sequences with a given property (12). These
schemes can produce experimentally useful reagents and pro-
vide a wealth of information about sequence requirements for
the selected activity (e.g., refs. 13-15). Application of such a
selection strategy depends on the ability to produce large
libraries of random sequences, efficient selection procedures,
and appropriate means for recovering and characterizing the
selected molecules. Frequently, the techniques for selection or
screening of molecules are insufficient to find active se-
quences. In particular, if several tandem copies of a functional
segment are required for activity, then the problem of recov-
ering an active sequence from a random pool becomes in-
creasingly more difficult.
To circumvent the insufficiency of available selection tech-
niques for many interesting biological and biochemical activ-
ities, we sought to produce libraries of random repeated
sequences: pools of molecules in which each member contains
tandem repeats of a different sequence element. The potential
usefulness of such concatamer libraries can be illustrated by
calculating the probability that a given 8-bp element will occur
independently in three positions in a single random 60-mer
sequence ("1 in 250 million). If we replace the random
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement" in
accordance with 18 U.S.C. §1734 solely to indicate this fact.
60-mers with a library of trimerized random 20-mers, then this
probability improves by -5 orders of magnitude.
We considered several different methods for constructing
concatamer libraries. Chemical synthesis can be used to gen-
erate random pools of DNA sequence oligomers (12), but
straightforward ligation to concatenate elements from these
pools would not produce the desired result, since there is no
means to ensure that ligation joins molecules of the same
sequence. A straightforward method for generation of a small
library would be to separately synthesize and concatenate a
number of separate oligonucleotides (16); unfortunately, this
would be cumbersome and expensive for large libraries.
As a more general procedure we chose a scheme based on
the rolling circle replication used by many plasmids and viruses
(17, 18). Rolling circle replication involves two simultaneous
processes: (i) DNA polymerase must synthesize sequences
complementary to a circular template. (ii) As this replication
proceeds, some mechanism must unwind the parental duplex
to allow the polymerase to advance. Models for physiological
rolling circle replication generally involve a template that is
predominantly double stranded, with a helicase or single-
strand DNA binding activity preceding the polymerase to allow
replication to continue (17, 18). Characterized rolling circle
replication mechanisms have been found to operate on tem-
plates on the order of kilobases and larger (18). Rolling circle
replication of templates smaller than 100 bp by previously
described mechanisms would be considered unlikely, since
formation of very short double-stranded circles would be
topologically obstructed (19). Although there was no prece-
dent, we chose to examine the ability of predominantly single-
stranded circles to act as templates for rolling circle synthesis.
MATERIALS AND METHODS
Reagents. Oligonucleotides were made on an Applied Bio-
systems DNA synthesizer (model 380B), desalted, precipitated
with ethanol, and used without further purification. Bacterio-
phage T4 DNA ligase, polynucleotide kinase, and DNA poly-
merase were from New England Biolabs. Escherichia coli DNA
polymerase was from Bethesda Research Laboratories; DNA
polymerase large fragment (Klenow) and Sty I were from
Boehringer Mannheim. Modified phage T7 DNA polymerase
(Sequenase) was from United States Biochemical.
Generation of Putative Concatamer Products. Step 1. Prim-
er/template was produced by dilute annealing followed by
ligation. Template oligonucleotide (zf42 or zf43) (0.8 nmol)
was phosphorylated with T4 polynucleotide kinase (60 Rich-
ardson units) in 250 gul of LK buffer (50 mM Tris-HCl, pH
7.8/10 mM MgCl2/10 mM dithiothreitol/1 mM ATP/25 ,ug of
bovine serum albumin per ml). After heating to 70°C for 10
min, the material was immediately diluted at 37°C into 12 ml
of LK buffer containing 0.7 nmol of primer oligonucleotide
(zf39). After 30 min at 23°C, the sample was transferred to
16°C and incubated for 4 hr with T4 DNA ligase (120 Weiss
units). Ligase reactions were stopped by addition of EDTA (to
10 mM), SDS (to 0.1%), NH4OAc (to 1 M), and 50 jig of
glycogen, extracted once each with 8 ml of phenol/chloroform
*To whom reprint requests should be addressed.
4641

4642
Biochemistry: Fire and Xu
(1:1) and chloroform and precipitated twice with ethanol.
Final samples were resuspended in 25 ,1 of TEN (10 mM
Tris-HCl, pH 7.5/1 mM EDTA/25 mM NaCI) and stored at
-70°C. In handling the annealed primer/template, care was
taken to avoid transient low-salt or high-temperature condi-
tions that could lead to denaturation.
Step 2. Primer extension reactions were carried out in 20 ,l
with 0.8 ,tl of primer/template. Extension reaction mixtures
contained 10 mM Bistris propane chloride; 10 mM NaCl; 1
mM dithiothreitol; 1 mM each dATP, dTTP, dGTP; 0.25 mM
dCTP (including [32P]dCTP to a specific activity of 0.5 Ci/
mmol; 1 Ci = 37 GBq); 0.1 mg of bovine serum albumin per
ml; and the indicated polymerase (in some experiments 50 mM
NaCl was added; this did not change the product electrophore-
sis pattern). Incubations were at 23°C for T4 and Klenow [large
fragment of E. coli DNA polymerase I (18)] and 37°C for E.
coli DNA polymerase I and coliphage T7 DNA polymerase
(Sequenase).
Although E. coli DNA polymerase was most efficient under
the reaction conditions used, longer exposures revealed similar
patterns for Klenow and T7; we have not extensively varied
reaction conditions to optimize output with these enzymes.
Similar reactions (not shown) with thermophilic DNA poly-
merases (Vent, Pfu, and Taq; reactions at 58°C) produced no
detectable rolling circle products but instead produced a series
of products indicative of multiround rolling hairpin replication
as described by Cavalier-Smith (20) (although we cannot rule
out the possibility that some rolling hairpin products were also
produced in reactions with the other polymerases, neither the
partial digestion products nor the structures of cloned products
were indicative of this as a major component of the E. coli
polymerase I reaction material).
Step 3. For further characterization by restriction digestion
and cloning, we used material from 4-hr extension products
produced with E. coli DNA polymerase I. This analysis re-
quired that the product be converted to a double-stranded
form. In initial experiments, an extra round of DNA synthesis
was performed to produce the second strand [this was done by
isolating the reaction product and allowing self-priming in a
standard hairpin reaction (20)]. Subsequently, we found that
much of the product was double stranded even without this
second round of synthesis [a plausible explanation would be
that self-priming hairpin structures (20) form at some fre-
quency during the initial 4-hr synthesis reaction, perhaps after
unraveling of the link between the elongating DNA terminus
and the rolling circle template].
Partial digestion was carried out with restriction endonu-
clease Sty I, which should cut just once in each tandem repeat.
To check the efficiency of the partial digest, samples were
resolved after denaturation on a 6% acrylamide sequencing
gel. Preparative samples were resolved in parallel without
denaturation; these produced a shifted ladder of bands con-
sistent with the double-stranded character of the material. The
nondenatured samples from the Sty I partial digest were
excised, eluted, ligated into a suitable Sty I-cut bacterial
plasmid vector, and transformed into a Rec- bacterial host.
Production of cloned libraries from the partial digest material
was relatively efficient; by using standard protocols for bac-
terial transformation (106 colonies per ,tg of control plasmid
DNA), several hundred clones were obtained with material
derived from 1 pmol of input oligonucleotides. Much larger
libraries could readily be obtained by optimizing and scaling up
the bacterial transformation.
RESULTS
The experimental design (Fig. 1) was based on a consideration
of possible behavior for an oligonucleotide primer that is being
extended by DNA polymerase after annealing to a single-
strand circular template. As the primer is extended, a double-
5'
zf43
3'
CTTGGTCTACT
GGAG
nnnnnnnnnnnnnnnnnnnnCTACGGATTGC
Polynucleotide
Kinase
5' ,zf43
3'
,CTTGGTCTACTGGAG
nnn
nnnnnnnnnnnnCTACGGATTGC
anneal zf39
3' <-ATGCCTAACQO^AACCAATctgacgtcagctgggaag
5'
..........::::::::::
::::
:
zf39
CTACGGATTGC CTTGGTCTAC
n
3'
5'
T
n
G
n
G
zf43
A
n
G
nnnnnnnnnnnnnnnnnnnnn
DNA Ligase
3' <-ATOCCTAACGQAeCCAQMTctgacgtcagctgggaag 5'
:::::::::::::::::::
zf39
CTACGGATTGCCTTGGTCTAC
n
T
n
G
n
G
n
zf43-cir
G
n
A
n
G
nnnnnnnnnnnnnnnnnnnnn
DNA polymerase
-QATGCCTAACCGAACCAGLTctgacgtcagctgggaag 5'
:::::::::::::::::::::
zf39
CTACGGATTGCCTTGGTCTAC
·*n
T
·-n
G
.n
G
zf43-cir
A
-.n
A
·-n
G
**nnnnnnnnnnnnnnnnnnnnn
:::::::::::
DNA polymerase
(at *)
and
Unraveling of duplex
(at
)
--GLTGCCTAACGGAACCACATGACCTC
nnnnnnnnnnnnnnn nnnn
/
:::::::::::::::::::::
CTACGGATTGCCTTGGTCTAC
·*n
T
·-n
G
n
f43-cir
G
·-n
A
·-n
G
*nnnnnnnnnnnnnnnnnnnnn
::::
3'
TQ
nnnCTCCAGTCTAGCCAAGKCAATCC
n
n nnnnnnnnnnnnnnnnnnGkTOCCTA
A
zf39->
C
5* gaagggtcgactgCAOTCTMAACCAMaG
FIG. 1.
Scheme for generation of concatamer library using rolling
circle replication. The template oligonucleotide (zf43) is circularized
using a partially complementary oligonucleotide (zf39) as a guide. This
primer template is then reacted with DNA polymerase in the presence
of deoxynucleoside triphosphate precursors. As synthesis proceeds,
the polymerase creates duplex DNA. At some point, the circle becomes
constrained so tightly that polymerization cannot proceed without some
relief of the restraint. One possible outcome is shown: the constraint on
the small circle might drive unwinding at the lagging end of the duplex.
If this process is combined with continued polymerization, then a long
repeating polymer would be produced. For this figure, the 52-base
oligonucleotide zf43 is shown as template; oligonucleotide zf42, used in
other experiments, is identical in the constrained regions but has 10 fewer
nonspecified residues (total length, 42 bases).
stranded region of the circle will be formed. At a certain point,
this double-stranded region will become sufficiently long to
"strain" the circular topology. At this point, one of three events
might occur: (i) the polymerase might stall or stop; (ii) the
template might be forced to unravel behind the polymerase; or
(iii) unwinding might occur at the site of polymerization-this
could lead to self-priming of the product strand (20).
If unraveling behind the polymerase occurs, then a rolling
circle mechanism could be set up; the production of new
double-stranded regions as the polymerase progresses 5' to 3'
would provide a steric force unwinding the DNA duplex
behind the active polymerase. Under these circumstances, the
energy for both polymerization and helicase activity would be
derived from utilization by DNA polymerase of nucleoside
triphosphates.
The substrates we used to test this rolling circle replication
scheme are shown in Fig. 1. A primer template pair is set up
Proc. Natl. Acad Sci. USA 92 (1995)

Proc. Natl. Acad. Sci. USA 92 (1995)
4643
by using two oligonucleotides with a permuted region of
complementarity. Annealing and ligation of these two oligo-
nucleotides should produce a primed, partially single-stranded
circle. To provide diversity in the template population, most of
the unpaired residues are nonspecified (i.e., an equimolar
mixture of A, G, C, and T). Rolling circle synthesis should then
generate a population of long single-stranded products, each
containing repeats complementary to an individual template
molecule.
In initial experiments with several available DNA poly-
merase enzymes, we found that E. coli DNA polymerase I was
particularly efficient in synthesizing high molecular weight
products from the primer template; after 3 hr, a large amount
of high molecular weight product had formed (Fig. 2A).
After conversion to a double-stranded product, the material
was further characterized by partial restriction digestion, with
an enzyme cutting once in each repeat. The resulting ladder of
bands (Fig. 2B) was consistent with a tandem array structure
derived from rolling circle synthesis. As expected, more com-
plete digestion with the restriction enzyme (Fig. 2C) produced
material that was predominantly monomer length. An estimate
of the yield from a 4-hr extension with E. coli polymerase I was
obtained from these experiments: -15 ng of Sty I-sensitive
high molecular weight material was synthesized per pmol of
input oligonucleotides.
As a critical test of the rolling circle model, it was essential
to show that the reaction indeed produced repeats derived
from individual input templates. This was demonstrated by
A
E. coli
T4 pol
Klenow
Pol
I
Sequenase
5' 30' 3hrl5'
30' 3hr
5'
30' 3hrl 5'
30' 3hrl
B
C
zf-42
zf-43 template
-do
Styl
1
23
6
lane
[['
180
150
120
90
60
30
10
168
126
012
Styl
1
2
3
4
5
lane
168
126
84
42
FIG. 2.
Electrophoretic separation of reaction products. (A) Time course of extension reactions. The annealed primer/template diagrammed
in Fig. 1 was extended with the indicated polymerase in the presence of [32P]dCTP. At the indicated time points, aliquots were removed, denatured,
and resolved on a sequencing-type gel. Base numbers on the right are sizes deduced from standard DNA sequencing reactions run on the same
gel (not shown). A delayed time course in the synthesis reaction with E. coli polymerase I (Pol I) (compare 30-min and 3-hr samples) has been
seen with a variety of primer/template combinations and polymerase preparations. The nature of this apparent lag has not been investigated. (B)
Partial digestion of reaction products with restriction enzyme Sty I. As above, samples were denatured and resolved on sequencing-type gels. Lanes:
1-3, digestion of zf42 extension product; 4-6, identical except that the template has been derived from oligonucleotide zf43, which is longer than
zf42 by 10 bases. The broadened bands observed in lanes 4-6 apparently result from the presence of shorter (internally deleted) oligonucleotide
molecules in the zf43 preparation. (C) Full digestion of reaction products with Sty I. Lanes: 1, undigested products from a 4-hr extension using
zf42 as template; 2-5, digestion with concentrations of Sty I in a range expected to yield full digestion. The major band at 42 nt presumably represents
completely digested unit-length material (the doublet band may represent the two different product strands). A small amount of material migrating
at 84 bp with the highest Sty I concentration is likely to result from a residual level of incomplete digestion. Other (minor) bands have not been
characterized.
Biochemistry: Fire and Xu
::.r::.::

4644
Biochemistry: Fire and Xu
cloning and sequencing of reaction products. Since the original
oligonucleotides contained 16 (or 26) nonspecified bases,
repetition of the identical sequence in several consecutive
members of a concatamer would be conclusive evidence that
the cloned multimer is indeed derived by a replication event
from a single template molecule.
From one set of synthesis reactions, we sequenced 18
different cloned products (Fig. 3). All gave different se-
quences. Ten of these sequences had the predicted structures,
with each having 3-5 tandem copies of a unique oligonucle-
otide from the original pool. Six of the characterized clones
had an alternating structure in which two of the initial oligo-
nucleotides alternate in a mixed tandem array. These products
could result from dimer circles (bridged by two primer mole-
cules) formed during annealing and ligation of the initial
primer template. The remaining two clones had repeated
structures that require more complex explanations but none-
theless evidence the occurrence of rolling circle replication
events.
Using other templates similar to those shown in Fig. 1, we
have observed replication products with tandem repeats of
A
cone
pPD74.33
pPD74.34
pPD74.36
pPD74.21
pPD74.20
pPD74.09
pPD74.24
pPD74.29
pPD74.23
pPD74.30
B
pPD74.15
pPD74.14
pPD74.10
pPD74.19
pPD74.26
pPD74.31
structure
44i41»
I
I
1
131»1
1444
???
???
???
111 12
1212
121
C
pPD74.12
1 1+ 1c
pPD74.25
?
*
Proc. Natl. Acad Sci USA 92 (1995)
monomers as short as 34 bp. In these and previous experi-
ments, there has been no indication of specific sequence
requirements for rolling circle synthesis. In particular, exam-
ination of the recovered sequences from Fig. 3 reveals no
evidence for bias in either base composition or secondary
structure.
DISCUSSION
We have shown that E. coli DNA polymerase I is capable of
carrying out a rolling circle type synthesis reaction on a very
short circular template. Although the mechanism of this
reaction has not been studied in detail, several theoretical
considerations of topology and scale are relevant. First, the
short templates used in these experiments would be unlikely to
form fully double-stranded circles without extreme topological
strain. As an alternative, it seems likely that the replicating
complex consists of a predominantly single-stranded circle,
with DNA polymerase working to extend a short double-
stranded region. At a certain point, extending the total length
of the double-stranded region would be energetically unfavor-
repeat unit(s)
zf43 complement=GCAATCCGTAG (N)
CTCCAGTAGACCAAG
1-GCAATCCG
TAaTCaCACCT
TACTACTCCTCAG-AGACCAAG
1-GCAATCCGTAGAAT TATATCTTCCCQQGTTC-T--CTCCAG-AGACCAAG
z
lGC
AATCCGTAAGGC
JIrAT
.CCTCCAGTAGACC
AA
G
|zf42 complement=GCAATCCGTAG (x) 6CTCCAGTAGACCAAG |
1- GCAATCCGTACGTTQAQCTCATAACACTCCAGTAGACCAAG
1- GCAATCCGTAGCTAATTQTATGACG0-CTCCAGTAGACCAAG
1- GCAATCCGTAGCCAQCCTATCTTQGCCTCCAGTAGACCAAG
1- GCAATCCGTAGATTTAATTAQAACTCCAGTAGACCAAG
1- GCAATCCGTAGTCAATTOQATTTCTCCAGTAGACCAAG
1- GCAATCCGTAGAQAACATAATOCCAACTCCAGTAGACCAAG
1= GCAATCCGTAGQACGTAATCTQAAATCCTCCAGTAGACCAAG
1- GCAATCCGTAGQATQTTCTTATACAC-CTCCAGTAGACCAAG
2- GCAATCCGTAGCQTATAA
CCTCCAGTAGACCAAG
1- GCAATCCGTAGAAGT
ATAATCTCTCCAGTAGACCAAG
2m GCAATCCGTAGgqQ
gC-------CTCCAGTAGACCAAG
1- GCAATCCGTAGTCTAACTCTAO--.-----CCAGTAGACCAAG
2- GCAATCCGTAGCCCCACGQTCTCTACTCCAGTAGACCAAG
1- GCAATCCGTAG----------
CTCCAGTAGACCAAG
2- GCAATCCGT-GCCATGTQ
AACCATCCTCCAGTAGACCAAG
1- GCAATCCGTAGACA
TCCOgCCCQCTCCAGTAGACCAAG
2- GCAATCCGTAGTATQCCCA
TACCCTCCAGTAGACCAAG
1- GCAATCCGTAGCCCTTTTTTCQQCTCCAGTAGACCAAG
2- GCAATCCGTAG ACaTCAgCAgCCCCCTCCAGTAGACCAAG
27bp
1+iGCAATCCGTAGACCCTCQMACACCTCCCTCCAGTAGAC
fr
39
1- GCAATCCGTAGCACTGQAACTCCCCTCCTCCAGTAGACCAAG
2= GCAATCCGTAGACOAAoTOTTCgACC-CTCCAGTAGACCAAG
FIG. 3.
Structures of 18 clones from E. coli DNA polymerase I extension of primer/templates. Sequences of DNA from 18 arbitrarily chosen
clones from putative rolling circle synthesis are shown. Plasmids pPD74.33, -74.34, and -74.36 were derived from zf43; all others were derived from
zf42. In three cases (pPD74.19, -74.21, and -74.23), the plasmids analyzed apparently contained two different inserts that had been joined during
ligation of insert to plasmid vector; in these cases, only the first insert is described. (A) The ten "simple" products each had the structure predicted
for rolling circle synthesis from the input primer/template. In several cases, the periodic repeat in these products differs in size by 1-4 bp from
the design of the input oligonucleotide; this probably reflects a population of internally deleted oligonucleotides in the original zf42 and zf43
preparations. (B) Six products had an alternating pattern derived from two different input molecules. These are assumed to result from formation
of dimeric primer/template circles in the initial ligation (two input molecules joined end to end to end). Consistent with this hypothesis, when the
initial ligation to form zf42 circles was performed at a 10-fold higher concentration of template and primer, a majority of clones (70%) exhibited
an alternating structure. (C) Two products had structures that were not simply explained, although each had a repeated structure consistent with
a rolling circle event.

Proc. NatL Acad. Sci. USA 92 (1995)
4645
able. At this point, continued polymerase activity (at the
leading end) could drive unwinding behind the polymerase.
A second constraint comes from a requirement for rota-
tional positioning of DNA polymerase on the template. The
twist in the local DNA helix at the site of synthesis necessitates
a rotation of polymerase relative to template. This must occur
without the polymerase physically passing through the cyclized
DNA template [DNA polymerase I is sufficiently large (from
the Klenow crystal structure; see ref. 21) that even with
maximum contortion, a nucleotide chain of <60 residues could
not encircle
it]. As an alternative, polymerase could act
essentially as a fixed surface while the template continually
twists inward on itself. The combined forward and twisting
motion of the template could provide the necessary constant
interface between polymerase and template.
The ability of isolated polymerase to carry out rolling circle
synthesis on a small DNA circle in vitro suggests that this
process could have played an evolutionary role [along with
mechanisms such as unequal recombination, transposition,
and replication slippage (2, 18)] in generating the plethora of
natural tandem-repeat structures that appear in coding and
noncoding sequences. The enzymatic activities used in our
scheme (kinase, ligase, and DNA polymerase) are readily
available in vivo. Template oligonucleotides might come from
a pool of replication intermediates or nucleic acid breakdown
products present in cells. The final requirement (permuted
complementary between primer and template oligonucleo-
tides) would be a relatively rare coincidence; nevertheless,
even a small genome would be expected statistically to contain
many suitable combinations of sequences.
Given the efficiency of the in vitro reaction, it is conceivable
that short circles might replicate in vivo as part of a concerted
physiological or pathological process. The smallest known
circular replicons are small viroid and satellite virus RNAs of
several hundred base pairs (22). It is conceivable that there are
(as yet undiscovered) shorter plasmids or viroid-like parasites
that replicate by using the single-stranded rolling circle mech-
anism described here.
From an experimental point of view, the ability to produce
large libraries of random or semirandom concatamers should
have numerous applications. We have begun using these
concatamer libraries to characterize cis-acting control se-
quences with defined activation patterns, with the goal of
identifying the corresponding developmentally regulated tran-
scription factors. Three other potential uses for concatamer
libraries are notable.
(i) Vergnaud et at (16) have used individually synthesized
and concatamerized random oligonucleotides as hybridization
probes to identify human DNA polymorphisms. The ability to
make large libraries of random concatamers should greatly
facilitate this approach.
(ii) In a variety of systems, sequences controlling translation
have been localized to short repeated sequence motifs in the
3' nontranslated leader of the mRNA (5-8). By constructing
a library of concatamers inserted into the 3' leader sequence
for an easily assayed reporter gene/expression vector, it should
be possible to identify control sequences producing different
patterns of translational regulation.
(iii) Repeated motifs in proteins have been used in evolution
to produce tight and very specific protein-protein interactions
(9-11). Construction of a concatamer library in a protein
coding context should allow in vitro or in vivo selection of
repeated peptide sequence motifs that allow specific binding to
a target protein.
Many of these selective schemes could be envisioned as a
two-step process. Once a molecule with desired characteristics
has been identified in any of the above screens, a second round
of selection could be carried out to isolate sequences with
optimal activity. This would be done by starting with material
in which each base is only partially randomized relative to the
initially recovered active sequence (e.g., an A residue in the
initial selected sequence might be replaced by a mixture of
90% A + 5% G + 3% T + 2% C).
We are grateful to D. Brown, J. Corden, J. Gall, N. Fedoroff, V.
Jantsch-Plunger, D. Kaiser, D. Koshland, M. Krause, P. Okkema, A.
Pinder, H. Rienhoff, R. Schleif, Z. Wang, an anonymous reviewer, and
members of the Department of Embryology for their help and
suggestions. This work was supported by the National Institutes of
Health (Grant R01-GM37706 to A.F.) and the Carnegie Institution.
A.F. is a Rita Allen Scholar.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Britten, R. J. & Kohne, D. E. (1968) Science 161, 529-540.
Beridze, T. (1986) Satellite DNA (Springer, Berlin).
Trifonov, E. N. & Brendel, V. (1986) Gnomic: A dictionary of
Genetic Codes (Balaban, Rehovot, Israel).
Ondek, B., Shepard, A. & Herr, W. (1987) EMBOJ. 6,1017-1025.
Theil, E. C. (1990) J. Biol. Chem. 265, 4771-4774.
Wharton, R. P. & Struhl, G. (1991) Cell 67, 955-967.
Goodwin, E. B., Okkema, P. G., Evans, T. C. & Kimble, J. (1993)
Cell 75, 329-339.
Wightman, B., Ha, I. & Ruvkun, G. (1993) Cell 75, 855-862.
Corden, J. L. (1990) Trends Biochem. Sci. 15, 383-387.
Bork, P. (1992) FEBS Lett. 307, 49-54.
Engel, J. (1992) Biochemistry 31, 10643-10651.
Szostak, J. (1992) Trends Biochem. Sci. 17, 89-93.
Bartel, D. P. & Szostak, J. W. (1993) Science 261, 1411-1418.
Beutel, B. A. & Gold, L. (1992) J. Mol. Biol. 228, 803-812.
Barbas, C. F., Amberg, W., Simoncsits, A., Jones, T. M. &
Lerner, R. A. (1993) Gene 137, 57-62.
Vergnaud, G., Mariat, D., Apiou, F., Aurias, A., Lathrop, M. &
Lauthier, V. (1991) Genomics 11, 135-144.
Gilbert, W. & Dressier, D. (1968) Cold Spring Harbor Symp.
Quant. Biol. 33, 473-484.
Baker, T. A. & Kornberg, A. (1992) DNA Replication (Freeman,
New York).
Koo, H.-S., Drak, J., Rice, J.A. & Crothers, D. M. (1990)
Biochemistry 29, 4227-4234.
Cavalier-Smith, T. (1974) Nature (London) 250, 467-470.
Ollis, D. L., Brick, P., Hamlin, R., Xuong, N. G. & Steitz, T. A.
(1985) Nature (London) 313, 762-766.
Diener, T. 0. (1991) FASEB J. 5, 2808-2813.
Biochemistry: Fire and Xu