P1: DPI/shd/rvt
May 16, 1998
P2: NBL/rpk
14:32
Annual Reviews
AR057-08
Annu. Rev. Biochem. 1998. 67:181–98
c 1998 by Annual Reviews. All rights reserved
Copyright BASE FLIPPING
Richard J. Roberts
New England Biolabs, Beverly, Massachusetts 01915; e-mail: [email protected]
Xiaodong Cheng
Department of Biochemistry, Emory University School of Medicine, Atlanta, Georgia
30322; e-mail: [email protected]
KEY WORDS:
DNA repair enzymes, DNA methyltransferases, DNA-modifying enzymes,
evolution
ABSTRACT
Base flipping is the phenomenon whereby a base in normal B-DNA is swung completely out of the helix into an extrahelical position. It was discovered in 1994
when the first co-crystal structure was reported for a cytosine-5 DNA methyltransferase binding to DNA. Since then it has been shown to occur in many systems
where enzymes need access to a DNA base to perform chemistry on it. Many
DNA glycosylases that remove abnormal bases from DNA use this mechanism.
This review describes systems known to use base flipping as well as many systems where it is likely to occur but has not yet been rigorously demonstrated. The
mechanism and evolution of base flipping are also discussed.
CONTENTS
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
KNOWN BASE-FLIPPING SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
HhaI DNA Methyltransferase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
HaeIII DNA Methyltransferase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Human Uracil DNA Glycosylase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
T4 Endonuclease V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
PROBABLE BASE-FLIPPING SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Amino-Methyltransferases M.TaqI and M.PvuII . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
E. coli DNA Photolyase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
E. coli Endonuclease III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3-Methyladenine DNA Glycosylase (AlkA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
E. coli Exonuclease III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
T4β-Glucosyltransferase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
E. coli Ada O6-Methylguanine DNA Methyltransferase . . . . . . . . . . . . . . . . . . . . . . . . . .
E. coli Mismatch-Specific Uracil DNA Glycosylase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
182
183
184
187
187
188
189
189
190
191
191
192
192
192
193
181
0066-4154/98/0701-0181$08.00
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
182
Annual Reviews
AR057-08
ROBERTS & CHENG
T7 ATP-Dependent DNA Ligase and T7 DNA Polymerase . . . . . . . . . . . . . . . . . . . . . . . . . 193
MECHANISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
THE ORIGINS OF BASE FLIPPING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
FUTURE PROSPECTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
INTRODUCTION
The binding of proteins to DNA is crucial to life. Whereas some proteins bind
to control transcription and replication, others both bind to DNA and catalyze
chemical reactions on the DNA. The latter proteins include nucleases, glycosylases, DNA methyltransferases, and various enzymes such as integrases
and recombinases, which rearrange DNA segments. For proteins such as transcription factors and repressors, whose function rests with DNA binding, some
common structural features have been identified, such as zinc fingers and helixturn-helix motifs (1). The binding to DNA by these proteins frequently involves
small deformations in the usual B helix, although in some cases extreme distortions can be found, such as the saddle structure on which the TATA-binding
protein sits (2, 3) or the bends induced by integration host factor (4). Some
nucleases, such as R.EcoRI (5) and R.EcoRV (6), which cleave the phosphodiester backbone, also induce conformational changes in the DNA helix, but
for the most part these changes are not dramatic. This is probably because the
target phosphodiester bonds lie on the outside of the helix and thus are readily
accessible to the enzyme. For proteins that need to interact with the bases rather
than the phosphodiester backbone and then perform chemistry on those bases,
accessibility is less clear. How might such proteins gain access to the interior of
the helix? Although it seems possible that the bases might become accessible
by distortion of the helix through bending and kinking, it is by no means simple
to envision. Obviously, a structure is needed for such a protein actually binding
to DNA.
In 1993 a structure was obtained for the cytosine-5 DNA methyltransferase,
M.HhaI, an enzyme that needed access to its target cytosine base to perform
the chemistry of methylation within the aromatic ring (7, 8). The crystal contained both the methyltransferase and its cofactor, S-adenosyl-L-methionine
(AdoMet), but no DNA. From the structure, and the accumulated biochemical knowledge about the enzyme, some clear predictions could be made about
where the DNA would be located. However, the structure gave little hint of
the rather dramatic conformational change in DNA that would take place upon
its binding. That change was revealed in 1994, when a ternary structure was
reported for a complex between M.HhaI, its DNA substrate, and the reaction
product, S-adenosyl-L-homocysteine (9). Surprisingly, the enzyme did not distort the DNA in some crude fashion by bending or kinking, but rather the target
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
183
cytosine had swung completely out of the helix and into the active-site pocket
of the enzyme (Figure 1, C-1, color section at end of volume). The base had
undergone a conformational shift of 180◦ , and the first example of base flipping
had been discovered. In retrospect, this shift makes sense, because it probably
provides the simplest method for a protein to access the interior of a DNA helix.
Because of the overall similarity at the primary sequence level of all cytosine-5
DNA methyltransferases (10), it seemed reasonable that base flipping was not
just an isolated quirk unique to M.HhaI but would be found in all members
of this set of methyltransferases. It was thus reassuring when a structure for
a second methyltransferase, M.HaeIII, appeared (11). It showed essentially
the same phenomenon, albeit with some differences that could be attributed
to differences in the recognition sequences. The big question, though, was
whether base flipping would show up in other situations. In our original paper we proposed that other classes of DNA methyltransferases and some DNA
glycosylases might also use base flipping to gain access to the bases within the
helix (9). Both of these predictions proved accurate, although in the case of T4
endonuclease V (12), the base flipping is not quite as we had envisioned (see
below). Many other examples have now appeared, either as co-crystal structures of DNA-protein complexes or from structures of the native enzymes with
or without cofactors. In the latter cases, it has often proved possible to infer
that base flipping may be involved because modeling of normal B-DNA into
these structures fails to bring the target base close to the active site—a situation
that is remedied if the base is flipped.
It has been argued that base flipping is an ancient process that may even have
preceded the use of DNA as the genetic material (13). If that is the case, then
one might expect to find other examples of base flipping among RNA enzymes
and in other proteins that interact with DNA. One might even imagine that base
flipping could nucleate the opening of a helix, as would be required during
the initiation of replication or transcription. In this review we summarize the
progress that has been made in studies of base flipping since its discovery in
1994 and discuss, somewhat speculatively, its mechanism and evolution. We
present the range of processes where base flipping has been proven to occur,
further cases where it is postulated to occur, and a few examples where it might
be worth looking for its involvement.
KNOWN BASE-FLIPPING SYSTEMS
Base flipping has been directly observed in four systems (Table 1): in the cocrystal structures for the cytosine-5 DNA methyltransferases M.HhaI (9) and
M.HaeIII (11), in a catalytically compromised form of T4 endonuclease V (12),
and in human uracil DNA glycosylase (14).
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
184
Annual Reviews
AR057-08
ROBERTS & CHENG
Table 1 Known base-flipping systems
Specific protein
Catalytic reaction
Reference
HhaI DNA methyltransferase
HaeIII DNA methyltransferase
T4 endonuclease V
Human uracil DNA glycosylase
Forms G-5mC-GC in DNA
Forms GG-5mC-C in DNA
Removes pyrimidine dimers from DNA
Removes uracil from DNA
9
11
12
14
HhaI DNA Methyltransferase
HhaI DNA methyltransferase (M.HhaI), a 327–amino acid residue protein,
methylates the internal cytosine of its recognition sequence 50 -GCGC-30 /30 CGCG-50 (15, 16). Despite its palindromic recognition sequence, the protein
functions as a monomer, methylating only one strand at a time. The methyl
donor AdoMet is required for enzymatic activity and is converted to S-adenosylL-homocysteine (AdoHcy) during methylation.
The structure of M.HhaI has been characterized extensively by X-ray crystallography in complexes with various forms of its DNA substrate (9, 17–19). The
protein has two lobes: a conserved AdoMet-dependent catalytic domain and
a DNA-recognition domain (7, 8, 20, 21). Nine ternary structures of M.HhaIDNA-cofactor have been determined, with either AdoMet or AdoHcy. Each
structure contains a 13-mer oligonucleotide duplex with a different combination of modifications at the position of the target cytosine within the recognition
sequence in the two DNA strands. Table 2 lists the oligonucleotides used in the
M.HhaI ternary complexes and the nomenclatures used.
5-FLUORO-20 -DEOXYCYTIDINE, 20 -DEOXYCYTIDINE, AND 5-METHYL-20 -DEOXYCYTIDINE The first three structures listed in Table 2 contain identical self-
complementary DNA sequences. The complex F13 contains 5-fluoro-20 -deoxycytidine (5FC or F) and was crystallized in the presence of AdoMet, which
resulted in the formation of a covalent linkage between the enzyme and DNA and
generation of 5-methyl-5-fluoro-20 -deoxydihydrocytidine and AdoHcy (9). The
N13 structure contains unmodified deoxycytidine, whereas the M13 contains
5-methyl-20 -deoxycytidine (5mC or M) (17). The structures are rather similar,
with one of the target nucleotides flipped out of the DNA helix and fitting snugly
into the active site of the enzyme (Figure 1, C-1, color section).
The 5mC residue in the fully methylated oligonucleotide in M13, the reaction
product, is flipped out of the DNA helix in the same manner as with the C in
N13 and 5FC in F13. This is surprising but consistent with biochemical data,
which suggest that the binding specificity for M.HhaI is asymmetric 50 -GXGC30 /30 -CGCG-50 and determined by the nucleotides neighboring the target (X)
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
185
BASE FLIPPING
Table 2 DNA sequences used in M.HhaI ternary structures
DNAa
Xb
Cofactor
Name
50 - TGATAGXG CTAT C -30
30 - C TATCG X GATAG T-50
F
C
M
AdoMet
AdoHcy
AdoHcy
F13
N13
M13
2.8
2.7
2.7
9
17
17
50 -TGTCAGX GCATGG -30
30 - CAGTCGMGTACC T-50
C
S
Z
AdoHcy
AdoMet
AdoHcy
HM13
S13
Z13
2.7
2.05
2.5
18
19
50 -TGTCAGXGC ATGG
0
Resolution (Å)
Reference
c
-3
30 - CAGTCGCGTACCT -50
AP
AdoHcy
AP13
2.4
d
50 -TGTCAGCGCATGG -30
30 - CAGTCG X GTACC T-50
A
U
AdoHcy
AdoHcy
GA13
GU13
2.8
2.7
d
d
a
The recognition sequence is in italics, and the flipped nucleotide is in bold and underlined.
A = adenine, C = cytosine, U = uracil, M = 5-methyl-20 -deoxycytidine, F = 5-fluoro-20 -deoxycytidine,
S = 40 -thio-20 -deoxycytidine, Z = 5,6-dihydro-5-azacytidine, AP = abasic sugar.
c
JK Christman, R Sheikhnejad, V Marquez, C Marasco, J Sufrin, M O’Gara, X Cheng, unpublished data.
d
M O’Gara, JR Horton, RJ Roberts, X Cheng, unpublished data.
b
nucleotide (22). In other words, the methyltransferase does not depend on the
flippable base for its binding specificity.
The HM13 structure (the fourth in Table 1) contained a hemimethylated DNA
substrate with a nonpalindromic sequence (18). The two strands of DNA are
clearly nonequivalent, because only the unmethylated target C flips out of the
DNA helix, whereas the 5mC on the complementary strand remains stacked in
the DNA helix. This seems to be an important aspect of DNA methylation; that
is, the ability of DNA methyltransferases to distinguish DNA substrates with
methyl groups on one strand (hemimethylated DNA) from those that carry no
methyl groups. However, many of the bacterial type II enzymes such as M.HhaI
are equally active on unmethylated and hemimethylated DNA, with an increased
affinity for asymmetrically methylated DNA. Consequently they function as
both maintenance and de novo methyltransferases. Both the N4 (NH2) group
and the methyl group of 5mC on the complementary strand interact with the
protein (18); these interactions may enable the de novo methyltransferase to
recognize both unmodified cytosine and methylated cytosine.
40 -THIO-20 -DEOXYCYTIDINE AND DIHYDRO-5-AZACYTIDINE That M.HhaI does
not show much binding specificity for the flippable base may reflect a need to
leave that base unencumbered by recognition contacts. This provides an opportunity to probe the structural and chemical interactions involved in sequencespecific recognition and catalysis using nucleotide analogs incorporated into
synthetic oligonucleotides in the position of the target nucleotide of M.HhaI. So
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
186
Annual Reviews
AR057-08
ROBERTS & CHENG
far, two nucleotide analogs—a 40 -thionucleoside and 5,6-dihydro-5-azacytosine
—have been used in our studies.
When 40 -thio-20 -deoxycytidine is incorporated as the target cytosine in the
recognition sequence for M.HhaI, binding to the 40 -thio–modified DNA is
almost identical to that of the unmodified DNA under equilibrium conditions
(19). In contrast, the methyl transfer was strongly inhibited in solution. Surprisingly, the flipped 40 -thio-20 -deoxycytidine in the S13 crystal structure was
partially methylated (19). These results show that 40 -thio-20 -deoxycytidine does
not disrupt DNA recognition, binding, or base flipping by M.HhaI. Instead they
suggest that it interferes with a step in the methylation reaction after flipping
but prior to methyl transfer.
5,6-Dihydro-5-azacytosine (DHAC) riboside was originally synthesized to
obtain a hydrolytically stable replacement for the antileukemic drug 5-azacytosine riboside (23, 24). DHAC contains a cytosine-like ring lacking aromatic
character with an sp3-hybridized carbon (CH2 group) at position 6 and an NH
group at position 5, resembling the transition state of a dihydrocytosine intermediate in the reaction of cytosine-5 DNA methyltransferases (25). The
Z13 structure, containing DHAC as the target (Table 2), showed that DHAC is
also flipped out of the DNA helix similarly to C, 5mC, and 5FC, but with no
covalent bond formed between the sulfur atom of nucleophile Cys81 and the
pyrimidine C6 carbon (JK Christman, R Sheikhnejad, V Marquez, C Marasco,
J Sufrin, M O’Gara & X Cheng, unpublished data). This result indicates that
the DHAC-containing DNA is sufficient to produce strong inhibition of the
DNA methyltransferase and that the DHAC moiety occupies the active site of
M.HhaI as a transition-state mimic.
MISMATCHED BASES (G:A, G:U, AND G:AP) Following the discovery of base flipping by M.HhaI, the effects of replacing the target cytosine by mismatched
bases, including adenine, guanine, thymine, and uracil, were investigated
(22, 26). By electrophoretic mobility shift analysis, M.HhaI and M.HpaII were
found to bind even more tightly to such mismatched substrates; the highest
affinity was for a gap formed by removal of the target nucleotide and both
phosphodiester linkages (22). Furthermore, the uracil can be enzymatically
methylated and converted to thymine at low efficiencies (22, 26), and the binding of these methyltransferases at the G:U mismatch prevented its repair by
uracil DNA glycosylase in vitro (26).
So far, three well-refined ternary structures of M.HhaI complexed with
AdoHcy and a nonpalindromic oligonucleotide containing a G:A, G:U, or
G:AP (AP = apurinic/apyrimidinic = abasic) mismatch at the target base pair,
respectively, have been determined (M O’Gara, JR Horton, RJ Roberts &
X Cheng, unpublished data). The mismatched adenine, uracil, and abasic site
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
187
are flipped out and located in the enzyme’s active site, respectively. It seems
likely that this active-site pocket is nonspecific for binding but specific for
methylation. In light of the nonspecific binding pocket, the DNA methyltransferase may be related more to repair enzymes such as 3-methyladenine DNA
glycosylase II and endonuclease III, which have broad substrate specificity. On
the other hand, the methylation reaction is specific in that catalysis occurs only
when the flipped base is cytosine or uracil (at low efficiency), more closely
resembling uracil DNA glycosylase.
HaeIII DNA Methyltransferase
HaeIII DNA methyltransferase (M.HaeIII), which also belongs to the monomeric type II bacterial methyltransferases, acts at its substrate site in palindromic
DNA and modifies the recognition sequence in two independent methyl transfer events—50 -GGC C-30 /30 -CC GG-50 —in double-stranded DNA. It methylates either of the underlined cytosines. In the structure of M.HaeIII bound to
hemimethylated DNA, the substrate cytosine is flipped out from the DNA helix
(Figure 2, C-1, color section), as observed for M.HhaI. M.HaeIII differs from
M.HhaI in that the flipping out of the cytosine is accompanied by rearrangement
of the remaining recognition bases in their pairing (11).
Between M.HhaI and M.HaeIII, the protein-base contacts in the recognized
sequence are expected to differ because of their different specificity. Indeed, the
folding of the corresponding DNA-recognition domains is different (11). However, these protein-base contacts share two important features (27). For both
methyltransferase-DNA complexes, six phosphates on the methylated strand
contact the protein: three on the 50 side and three on the 30 side of the target
cytosine, regardless of the position of the target nucleotide within the recognition sequence (second of four for M.HhaI, 50 -GCGC-30 , and third of four
for M.HaeIII, 50 -GGCC-30 ). From the protein side, a conserved arginine is
responsible for the recognition of the guanine 50 to the target cytosine. These
shared recognition patterns may be common to other 5mC methyltransferases
recognizing a guanine 50 to the target base (27, 28).
Human Uracil DNA Glycosylase
Human uracil DNA glycosylase (UDG) is responsible for the removal of uracil
residues within either single- or double-stranded DNA. The crystal structures
of the human (29, 30) and herpes simplex viral (31, 32) forms of the enzyme
have been solved. They revealed an extraordinarily specific binding pocket
that excludes normal DNA bases and uracil within RNA from specific binding.
Clearly these interactions could not form if the uracil was positioned inside a
B-DNA helix, so it seemed likely that base flipping could be involved. The
involvement of base flipping has now been confirmed with the description
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
188
Annual Reviews
AR057-08
ROBERTS & CHENG
of a co-crystal structure for human UDG complexed with a uracil-containing
double-stranded DNA (Figure 3, C-2, color section) (14). This structure,
which was solved using a double mutant of UDG (Leu272Arg and Asp145Asn)
showed that the uracil, the deoxyribose, and the 50 phosphate are rotated 180◦
from their starting structure within DNA, with Arg272 inserted into the position
previously occupied by uracil. Even though the catalytic efficiency of this double mutant was severely reduced, the N-Cl0 glycosidic bond had been broken
(Figure 3, C-2, color section).
The DNA conformation essentially maintained the B form, except for the
flipped nucleotide. The conformation of the flipped-out nucleotide is very
similar to that observed in the M.HhaI-DNA complex, and the protein-DNA
interactions are concentrated along the sugar-phosphate backbone of the strand
containing the flipped-out uracil (14). Thus, the four structurally characterized
base-flipping proteins, M.HhaI, M.HaeIII, human UDG, and T4 endonuclease
V (see below), have all shown that the DNA phosphate-protein contacts are
made mainly on one particular strand (reviewed in 27).
It was suggested that UDG uses a push-and-pull mechanism of base flipping
to ultimately achieve specific binding and catalysis (14). The push (protein
invasion) is contributed by Leu272: It displaces the target base through hydrophobic interactions with DNA. The pull (trapping the flipped nucleotide) is
derived by satisfying the specific binding pocket with uracil.
T4 Endonuclease V
T4 endonuclease V is a DNA glycosylase/AP lyase that can initiate repair of
cis-syn cyclobutane pyrimidine dimers in DNA by cleaving the glycosydic bond
of the 50 pyrimidine and then cleaving the phosphodiester backbone (33). Unlike M.HhaI and UDG, T4 endonuclease V kinks the dimer-containing DNA,
at an angle of approximately 60◦ (Figure 4, C-3, color section) (12). In contrast
to UDG, endonuclease V does not flip out the damaged bases into a binding
pocket. Instead, it moves the nucleotide opposite the 50 pyrimidine of the dimer
into a binding pocket on the surface of the enzyme (Figure 4, C-3, color section). In the co-crystal structure, a flipped adenine lies in proximity of and
sandwiched between two layers of protein atoms, both of which are arranged
parallel to the base plane. Vassylyev et al (12) suggested that the adenine was
stabilized by van der Waals interactions. Interestingly, the pocket into which
the adenine flips does not provide specific contacts that allow unambiguous
recognition of the base. This lack of specific recognition has been seen biochemically (34; unpublished data cited in 12) and shows that the glycosylase
activity is unaffected by the nature of the base opposite the 50 -lesion.
However, the key feature associated with endonuclease V’s base flipping is
that the hole in DNA, which is created by movement of the base, is filled by
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
189
BASE FLIPPING
the enzyme inserting its active-site residues into that hole. Thus, through the
change in the structure of the DNA, the enzyme is correctly positioned to carry
out a nucleophilic attack on Cl0 of the 50 pyrimidine of the dimer.
PROBABLE BASE-FLIPPING SYSTEMS
Several crystal structures have appeared for proteins that interact with DNA.
In these structures, it seems improbable that proper interaction could occur
if the DNA maintained its normal B conformation. However, in these cases
modeling with DNA containing a flipped base gives a much more reasonable
chance of interaction and suggests that base flipping is used in these systems.
Several recent reviews have appeared that discuss flipping in the context of
enzymes involved in the repair of DNA lesions (35–38). These systems are
described individually below and summarized in Table 3. Although definitive
proof of base flipping for those enzymes listed in Table 3 awaits successful cocrystallization studies, the locations of concave active sites within deep clefts
in these enzymes strongly suggest that a simple base flipping in the DNA takes
place rather than a complicated refolding of the protein.
The Amino-Methyltransferases M.TaqI and M.PvuII
One reason to flip out a DNA nucleotide is to subject it to chemical modification within a catalytic pocket (13, 27), as noted above for the cytosine-5
DNA methyltransferases (Figures 1 and 2, C-1, color section). Another class
of DNA methyltransferases, the amino-methyltransferases, methylate the exocyclic amino groups of adenine (N6) or cytosine (N4). None of these aminomethyltransferases have been structurally characterized in complex with DNA,
Table 3 Probable base-flipping systems
Specific protein
TaqI DNA methyltransferase
PvuII DNA methyltransferase
Escherichia coli photolyase
E. coli endonuclease III
3-Methyladenine DNA glycosylase
E. coli exonuclease III
T4 β-glucosyltransferase
E. coli Ada O6-methylguanine
DNA methyltransferase
E. coli mismatch-specific uracil
DNA glycosylase
a
Catalytic reaction
Forms TCG-6mA in DNA
Forms CTG-4mC-AG in DNA
Converts pyrimidine dimers to TT
Removes pyrimidine radiolysis
products from DNA
Removes 3-methyladenine from DNA
Cleaves 50 to Ap sites
Transfers glucose residues to T4 DNA
Removes Me from O6-Me-guanosine
Removes thymine from DNA
TE Barrett, R Savva, LH Pearl, unpublished data.
Reference
39
40
41
42
43, 44
45
46
47
a
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
190
Annual Reviews
AR057-08
ROBERTS & CHENG
but M.TaqI (adenine-N6 specific) (Figure 5, C-4, color section) and M.PvuII
(cytosine-N4 specific) (Figure 6, C-5, color section) have a catalytic-domain
structure with a concave active-site pocket very similar to that of M.HhaI (39, 40,
48–50). When these structures are modeled with normal B-DNA [Figures 5
(top) and 6 (bottom); see color section], the AdoMet is far away from the target
base, but by flipping this base out of the helix the two reactants are brought into
close proximity [Figures 5 (bottom), and 6 (bottom); see color section]. Thus,
both N6A-methyltransferases and N4C-methyltransferases almost certainly flip
out their target nucleotides.
In the case of M.EcoRI, which forms GA(6meA)TTC, biochemical evidence
in favor of base flipping has been obtained through a fluorescence-based assay
to detect conformational alterations in DNA induced by protein binding (51).
This method, which uses 2-aminopurine to replace the adenine in the substrate
DNA, may have general applications to probe potential base flipping in other
systems. Thus it has recently been shown that when M.HhaI and M.TaqI bind
to oligonucleotides containing 2-aminopurine as the target base within their
recognition sequences, the fluorescence intensity is greatly enhanced, consistent
with base flipping (B Holz, S Klimasauskas, S Serva & E Weinhold, unpublished
data). Biochemical evidence in favor of base flipping has also been developed
for another N6A-methyltransferase, M.EcoRV (52). Here it has been shown
that enhanced binding takes place when the target adenine, the first A in the
recognition sequence GATATC, is replaced by a modified base that weakens
base pairing. These results parallel those found for the known base-flipping
cytosine-5 methyltransferase, M.HhaI (22). In all of these studies it is believed
that destabilization of the base pair leads to an increased rate of base flipping
and the concomitant appearance of enhanced binding.
E. coli DNA Photolyase
In addition to DNA repair enzymes such as UDG that function within the
base excision repair pathway, DNA photolyase—an enzyme that acts through
a direct reversal pathway—also appears to incorporate nucleotide flipping in
its mechanism of action (41). E. coli photolyase uses a blue light–harvesting
chromophore 5,10 methenyl-tetrahydrofolylpolyglutamate (MTHF) to absorb
a photon and transfer the excitation energy to a catalytic chromophore, flavin
adenine dinucleotide (FAD) (53). The enzyme further transfers an electron
from FADH to the cyclobutane pyrimidine dimer to catalyze its fission to yield
the two original pyrimidines (53).
Examination of the solvent-accessible surface of the photolyase revealed
that the FAD cofactor is accessible to the dimer, only by way of a cavity in
the enzyme (Figure 7, C-6, color section). The dimensions of the putative
binding-site cavity are sufficient to bind the bases of the dimer but to exclude
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
191
the intradimer phosphate. The cavity is lined with polar residues on one side
and hydrophobic residues on the other. Park et al (41) point out that the asymmetric polarity of the cavity matches well with the asymmetric polarity of the
pyrimidine dimer, in which the cyclobutane ring is hydrophobic and the opposite edges of the thymine bases have nitrogens and oxygens capable of forming
hydrogen bonds.
E. coli Endonuclease III
The X-ray crystal structure of endonuclease III showed that this enzyme possesses an elongated bilobal structure with a deep cleft separating two distinct
domains linked by two loops (Figure 8, C-6, color section) (42, 54). The two
domains are α-helical in structure, with one containing a 4Fe-4S cluster and
the other containing a helix-hairpin-helix motif.
The cleft separating the two domains can accommodate B-form DNA; thus
both domains have been suggested to be involved in DNA binding. Endonuclease III substrate binding was investigated by soaking the enzyme crystals in
thymine glycol, a known inhibitor of glycosylase activity (54). The thymine
glycol binding site was located within a water-filled pocket in the cleft. Thus,
the damaged base is suggested to be flipped extrahelical and inserted into this
pocket for catalysis to occur, in a manner similar to that of M.HhaI and UDG
(Figure 8, C-6, color section).
3-Methyladenine DNA Glycosylase (AlkA)
AlkA is an N-alkylpurine DNA glycosylase that can efficiently remove
3-methyladenine, 7-methylguanine, 3-methylguanine, O2-methylcytosine, and
O2-methylthymine from double-stranded DNAs. The crystal structure was
published by Yamagata et al (43) and Labahn et al (44). These studies revealed
that AlkA is composed of three approximately equal-sized domains. Domain 1
has the topology and shape of the TATA-binding protein (2, 3, 55), whereas
domains 2 and 3 have the same topology and fold as endonuclease III (see
above). Domain 1 appears to serve as a platform for the other two domains.
The interface of domains 2 and 3 creates the base-binding and catalytic site,
and the flexibility of domain 3 permits the binding of various substrates.
Examination of the surface topography of AlkA revealed a cleft at the junction of domains 2 and 3, which is hydrophobic in nature and is dominated
by aromatic amino acid side chains (Figure 9, C-7, color section) (43). The
hydrophobic cleft also contains several charged residues, including the catalytically essential Asp. However, to position the methylated base correctly
within the active site, it was suggested (44) that AlkA uses the π electronrich aromatic rings to interact with the electron-deficient, positively charged
alkylated base, which is flipped out of the DNA helix. Thus, broad substrate
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
192
Annual Reviews
AR057-08
ROBERTS & CHENG
specificity can likely be achieved through these strong π donor-acceptor
interactions.
E. coli Exonuclease III
The only hydrolytic AP endonuclease that has its structure solved to date is
E. coli exonuclease III (45). The overall structure was shown to be similar to
that of DNase I (56), even though there is less than 20% sequence similarity.
It consists of two six-stranded β sheets, flanked by four α-helices forming a
four-layered αβ-sandwich motif. A ternary complex of exonuclease III with
Mn2+ and dCMP has also been reported (45). This complex showed the dCMP
bound at one end of the αβ sandwich, a region of the enzyme that contains a
high concentration of basic residues.
It was suggested that DNA-containing AP sites may be significantly distorted
such that the abasic sugar may protrude into the active site to allow the chemistry
to occur (45). As discussed above, M.HhaI is also able to flip an abasic sugar.
Alternatively, Mol et al (45) suggest that the nucleotide opposite the AP site
may be flipped into a pocket on the surface of the protein. If this flipping does
occur, then exonuclease III will be an example similar to T4 endonuclease V.
T4 β-Glucosyltransferase
Bacteriophage T4 uses 5-hydroxymethylcytosine in place of cytosine when
polymerizing its DNA. Following DNA synthesis, most (if not all) of these
hydroxymethylcytosines in duplex DNA are further modified by the addition of
glucose. This reaction is catalyzed by the β-glucosyltransferase, which transfers the glucose from uridine diphosphoglucose to the hydroxymethyl groups
in 5-hydroxymethylcytosine. The structure of the enzyme was solved in both
the presence and absence of the glucose donor, uridine diphosphoglucose (46).
The structure comprises two domains of similar topology (Figure 10, C-7, color
section). The two domains are separated by a cleft that contains a possible site
for duplex DNA binding. The catalytic site can be inferred from the position
of the UDP-glucose, which is deeply bound in a pocket at the bottom of the
cleft. When normal B-DNA is modeled into the structure, the glucose donor
and the target base are widely separated. However, as the authors suggest,
base flipping of the 5-hydroxymethylcytosine would nicely juxtapose the two
interacting moieties (46).
E. coli Ada O6-Methylguanine DNA Methyltransferase
O6-methylguanine DNA methyltransferase, otherwise known as the ada repair
enzyme, is a suicidal DNA-repair protein that removes the dangerous methyl
group from the promutagenic lesion O6-methylguanine and transfers it onto a
cysteine in the protein. The resulting self-methylation of the active-site cysteine
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
193
renders the protein inactive. A structure has been reported for the protein without DNA, but as noted (47), when B-DNA is modeled into the structure, a target
base would be more than 20 Å away from the active center! Thus a significant conformational change would be required to bring the O6-methylguanine
into close proximity with the buried catalytic cysteine that is the acceptor for
the methyl group (Figure 11, C-8, color section). Although Moore et al (47)
proposed a model in which the protein undergoes a drastic conformational distortion to bind DNA, it seems more likely that flipping the O6-methylguanine
out of the DNA helix would solve the problem of accessibility.
E. coli Mismatch-Specific Uracil DNA Glycosylase
Mismatch-specific uracil DNA glycosylase (MUG) removes uracil by a glycolytic mechanism from mismatches that contain uracil or thymine opposite
guanine, which would arise from deamination of cytosine or 5-methylcytosine.
The enzyme was originally discovered and characterized on the basis of its
amino acid similarity to human thymine glycosylase (57). Recently a crystal structure has been obtained for E. coli MUG binding to DNA (TE Barrett,
R Savva & LH Pearl, unpublished data). The structure shows great similarity
to uracil DNA glycosylase, especially around the active site. In the crystal,
the uracil base has been hydrolyzed and an abasic site is left. Interestingly, an
arginine residue is intercalated on the opposite strand, and this may be partly
responsible for the double-stranded DNA specificity of this enzyme.
T7 ATP-Dependent DNA Ligase and T7 DNA Polymerase
Finally, some proteins may mimic base flipping without actually flipping anything. The T7 ATP-dependent DNA ligase has a structure that broadly resembles
that of DNA methyltransferases (58). Subramanya et al (58) have proposed that
the reaction intermediate, which has AMP covalently bonded to the target nick
in the DNA, is formally analogous to the DNA methyltransferase intermediates:
The AMP is flipped out, even though no bases are missing from the target DNA.
A recent crystal structure for T7 DNA polymerase in complex with a primer
template, a nucleoside triphosphate, and its processivity factor (thioredoxin) has
been obtained (S Doublie, S Tabor, A Long, CC Richardson & T Ellenberger,
unpublished data). In this complex the template strand is sharply kinked such
that the base immediately 50 of the template base paired to the incoming nucleotide is flipped out of the active site. This conformation exposes the template
base for interactions with the polymerase and its nucleotide substrate.
MECHANISM
We have little hard data about the mechanism of base flipping. Two theories
are debated. In the first, it is suggested that base flipping is an active process
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
194
Annual Reviews
AR057-08
ROBERTS & CHENG
in which the base is pushed out of the helix by some appropriate residues on
the protein. Once flipped, it is pulled into the active site of the enzyme, where
it remains trapped during the reaction. A three-step pathway has been proposed for the DNA methyltransferases in which they first recognize the target
site and increase the interstrand phosphate-phosphate distance nearby. They
then initiate base flipping by protein invasion of the DNA and finally trap the
flipped DNA structure (27). An alternative possibility is that during the normal breathing of DNA, the bases naturally spend some time in the completely
flipped-out position, and it is this transient conformation in DNA that is recognized and caught by the protein. Although some biophysical measurements of
DNA motion have been made and lifetimes assigned to the flipped-out state, all
such measurements rely on assumptions about the states being modeled (59).
For instance, there is no direct experimental evidence showing that the lifetime
assigned to the flipped-out state is correct.
Most recently, nuclear magnetic resonance (NMR) techniques have been applied to study the interaction of M.HhaI with DNA containing 5-fluorocytosine
in solution (S Klimasauskas, T Szyperski, S Serva & K Wuthrich, unpublished
data). Surprisingly, the dynamics of base-pair opening appear unaffected by
binding M.HhaI alone. Only when the cofactor, or a cofactor analog, is added
can a clear signal be detected that is attributable to the flipped state of the target
base. Furthermore, these NMR studies suggest that at least three protein-DNA
complexes are involved: (a) an initial binding complex that forms between the
DNA and the protein in which the DNA maintains its normal stacked B conformation; (b) a complex, or series of complexes, in which the target base is
flipped from the helix but is not yet locked into its final flipped position observed
crystallographically; and (c) a final, catalytically competent complex in which
the base is locked into the active site. This last complex would correspond
to the complex seen in the crystal structure (9). Based on the NMR evidence,
the initial complex and the flipped-out complex are major equilibrium species
in solution, and all three types of complexes appear to be in rapid equilibrium
until the cofactor binds and shifts the equilibrium to favor the final complex.
This initial flipping is suggested to be an enzyme-assisted process even in the
absence of cofactor. Unfortunately, there is no crystallographic evidence about
the details of initial recognition. This evidence is needed to help unravel what
has become a complicated reaction pathway. In addition, T4 endonuclease V
has also been studied biochemically; evidence has been obtained for an initial
binding complex in which the adenine residue is not flipped (A McCullough,
ML Dodson, OD Scharer & RS Lloyd, unpublished data).
What is clear from the recent structure between M.HhaI and an abasic site
(M O’Gara, JR Horton, RJ Roberts & X Cheng, unpublished data) is that the
enzyme does not require a flipped-out cytosine residue as part of its initial
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
195
recognition mechanism. Similarly, T4 endonuclease V can flip bases other
than its normal substrate adenine (34; also cited in 12). Both observations
suggest that what is flipped is not the base, but rather the deoxyribose and/or its
flanking phosphates. Many of the other enzymes that use base flipping, such
as uracil DNA glycosylase, operate much more efficiently as enzymes than does
M.HhaI. In this case it seems unlikely that UDG simply waits for the natural
flip of the uracil residue.
THE ORIGINS OF BASE FLIPPING
It has been proposed that base flipping is an ancient process that may have
arisen even before the use of DNA as genetic material (13). The argument is
based on two lines of reasoning: Base flipping is used by many enzymes involved in DNA repair, and the DNA methyltransferase M.HhaI can bind tightly
to mismatches in DNA. In both cases abnormal base pairing is recognized with
dramatic consequences. A possible connection between DNA methylation and
repair has been suggested by Smith (60), who proposed that the human methyltransferase is a mismatch and DNA damage-recognition protein. Perhaps the
DNA methyltransferases have evolved from DNA mismatch binding proteins
by combining, into a single molecule, protein domains able to recognize mismatches in DNA, to perform sequence-specific recognition, to bind AdoMet,
and to carry out the methylation reaction. In this scenario, the key original
module is a mismatch recognition system that could accomplish base flipping.
Imagine an early stage in evolution when DNA first assumed its role as the
genetic material. Undoubtedly, the DNA polymerase of that time was less faithful than are current DNA polymerases, and incorrect bases would be inserted
with alarming frequency. In addition, spontaneous deamination of cytosine to
uracil could also lead to loss of fidelity, as would other chemical modifications
of the bases that might disrupt pairing. In short, a repair system would be essential if this early DNA were to be competent to serve as the genetic material.
How might this be accomplished?
A first step must be recognition that a mismatch was present, followed by
removal of the incorrect nucleotide and its replacement, most likely accomplished by the DNA polymerase having a second chance to copy faithfully.
Because it is inconceivable that a full-blown repair process would have arisen
immediately in a single enzyme, the repair process would require a series of
steps, each involving a separate catalytic molecule. Base flipping seems tailormade for the first step. One could imagine a small molecule, such as an RNA
or peptide, being able to recognize and bind a mismatch and, in the process,
flipping the mismatched base out of the DNA helix. Once the base is out of the
helix, it would then be simple for a second molecule such as an endonuclease to
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
196
Annual Reviews
AR057-08
ROBERTS & CHENG
cleave the phosphodiester chain and essentially create a new primer for the DNA
polymerase. Experimental evidence shows that at least one small molecule, a
metallo-porphyrin, can cause base flipping during intercalation into a B-DNA
helix (61). The real strength of this proposal is that it allows the separation
of mismatch recognition from the further steps of repair. Such stepwise accumulation of functions in a single molecule is undoubtedly how modern-day
proteins evolved.
If base flipping was indeed used for mismatch detection in the early DNA
polymers, one might reasonably ask how mismatch correction was dealt with in
the earlier RNA world. Presumably, exactly the same scenario could be painted
during the early course of RNA replication. Although there is no evidence for
base flipping by enzymes involved in RNA metabolism, there is also a dearth
of crystal structures for proteins involved in such processes.
FUTURE PROSPECTS
The theme common to all examples of base flipping is the requirement for
enzymes to perform chemistry on the bases normally embedded in a B-DNA
helix. As more proteins that perform chemistry on DNA are examined, further
examples of base flipping are likely to be discovered. Of particular interest will
be studies of the various mismatch repair systems, such as the methyl-directed
and the short patch-repair systems in E. coli, which might also be expected to
use base flipping.
Could there be other examples of base flipping within the realm of nucleic
acid metabolism that have so far been elusive? Perhaps! If base flipping truly
was an early evolutionary discovery, then one might anticipate its occurrence
in other processes where it might prove advantageous. One such process is the
local unwinding in DNA that is needed during the initiation of transcription
or replication. We do not know how this is achieved. Examination of any of
the DNA structures that contain flipped bases immediately suggests that base
flipping would provide an easy answer. With one base flipped out of helix and
the concomitant loss of both base pairing and stacking interactions, it would
be simple to disrupt an adjacent base pair and so begin the unzipping of the
helix. This is not the only mechanism that could be envisioned to initiate strand
separation, but it should be considered a possibility.
Within the area of RNA metabolism, all enzymes that perform chemistry on
RNA must also be considered candidates to employ base flipping. The enzymes
that modify tRNA or rRNA within base-paired regions, either by methylation or
by introducing more complex side chains, might use this mechanism. Another
likely candidate is double-stranded RNA adenosine deaminase. This enzyme
converts adenosine residues in RNA into inosine. It has been implicated in
ROBERTS & CHENG
C-1
Figure 1 M.HhaI complexed to its substrate DNA (pdb code 1mht) (9). DNA bases
are shown in blue, sugar-phosphate backbone in red, and protein in gold. AdoHcy is
shown in white.
Figure 2 M.HaeIII with DNA bound (pdb code 1dct) (11). DNA bases are shown in
blue, sugar-phosphate backbone in red, and protein in gold.
C-2
ROBERTS & CHENG
Figure 3 Human uracil DNA glysosylase complexed to DNA with a flipped uracil
(coordinates provided by C Mol and J Tainer) (14). The N-C1' glycosylic bond of the
flipped nucleotide has been cleaved, and the free uracil is bound in the specificity
pocket. DNA bases are shown in blue, sugar-phosphate backbone in red, and protein
in gold.
ROBERTS & CHENG
C-3
Figure 4 T4 endonuclease V complexed to DNA containing a thymine cyclobutane
dimer (pdb code 1 vas) (12). The adenine opposite the 5'-thymine residue of the
pyrimidine dimer is flipped. DNA bases are shown in blue, sugar-phosphate backbone
in red, and protein in gold.
C-4
ROBERTS & CHENG
Figure 5 M.TaqI-DNA docking model (coordinates provided by G Schluckebier and
W Saenger) (48) with (top) B-DNA and (bottom) DNA containing a flipped adenine.
DNA bases are shown in blue, sugar-phosphate backbone in red, and protein in gold.
AdoMet is shown in white.
ROBERTS & CHENG
C-5
Figure 6 M.PvuII-DNA docking model (40) with (top) B-DNA and (bottom) DNA
with a flipped cytosine. DNA bases are shown in blue, sugar-phosphate backbone in
red, and protein in gold. AdoMet is shown in white.
C-6
ROBERTS & CHENG
Figure 7 Ribbon diagram of Escherichia coli DNA photolyase (pdb code 1dnp) (41).
The α/β domain is shown in white, methenyl-tetrahydrofolylpolyglutamate (MTHF) in
light gray, and the helical domain in gold. The flavin mononucleotide group of flavin
adenine dinucleotide (FAD) is shown in green, and the AMP unit is in red.
Figure 8 Escherichia coli endonuclease III−DNA docking model with a flipped
nucleotide (coordinates provided by M Thayer and J Tainer) (42). DNA bases are
shown in blue, sugar-phosphate backbone in red, and protein in gold. The HhH motif
is shown in green.
ROBERTS & CHENG
C-7
Figure 9 Escherichia coli 3-methyladenine DNA glycosylase II−DNA docking
model with a flipped nucleotide (coordinates provided by Y Yamagata) (43). DNA
bases are shown in blue, sugar-phosphate backbone in red, and protein in gold. The
HhH motif is shown in green, and the catalytically essential Asp is in white.
Figure 10 T4 β-glucosyltransferase (pdb code 2bgu)-DNA docking model with a
flipped nucleotide (coordinates provided by P Freemont) (46). DNA bases are shown
in blue, sugar-phosphate backbone in red, and protein in gold. Uridine
diphosphoglucose is shown in white.
C-8
ROBERTS & CHENG
Figure 11 Ribbon diagram of ada O6-methylguanine DNA methyltransferase (pdb
code 1sfe) docked with a B-DNA. DNA bases are shown in blue, sugar-phosphate
backbone in red, and protein in gold. The side chain of the catalytic cysteine is shown
in white with a red border.
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
Annual Reviews
AR057-08
BASE FLIPPING
197
RNA editing and is suggested to resemble the DNA methyltransferases at the
sequence level (62).
In summary, the phenomenon of base flipping is appearing in a number
of systems following its initial discovery in the DNA methyltransferases, and
many additional enzyme systems must be considered candidates to use this
novel mechanism. Unfortunately, detailed mechanistic information is lacking,
and it remains to be proven whether base flipping is an active process in which
the protein pushes the base out of the helix or a passive one in which the protein
binds to a transiently flipped base. Most authors, including ourselves, favor
the active process, but further work is needed to settle the issue. Particularly
intriguing is the recent finding that when M.HhaI binds to an abasic site, it flips
the deoxyribose ring and its flanking phosphates into the same conformation
that is adopted during base flipping on its normal substrate. Thus if the process
is active, one might expect that the push will take place not on the base, but
rather on the sugar-phosphate backbone. Perhaps we should have termed the
phenomenon the less evocative “DNA backbone rotation.”
ACKNOWLEDGMENTS
We thank Drs S Kumar, S Pradhan, L Robow, and X Zhang for critical reading of
the manuscript; Dr. S Kumar and C Lin for help with the preparation of figures;
Drs T Barrett, P Freemont, C Mol, L Pearl, W Saenger, G Schluckebier, J Tainer,
M Thayer, and Y Yamagata for providing coordinates; and T Ellenberger,
S Klimasauskas, and E Weinhold for providing preprints. We also wish to
thank most warmly the members of our laboratories whose hard work was responsible for much of the methyltransferase story. Work in our laboratories is
supported by grants from the National Institutes of Health (GM46127 to RJR;
GM/OD52117 and GM49245 to XC).
Visit the Annual Reviews home page at
http://www.AnnualReviews.org.
Literature Cited
1. Pabo CO, Sauer RT. 1992. Annu. Rev. Biochem. 61:1053–95
2. Kim YC, Geiger JH, Hahn S, Sigler PB.
1993. Nature 365:512–20
3. Kim JL, Nikolov DB, Burley SK. 1993.
Nature 365:520–27
4. Travers A. 1997. Curr. Biol. 7:252–54
5. Kim YC, Grable JC, Love R, Greene PJ,
Rosenberg JM. 1990. Science 249:1307–9
6. Winkler FK, Banner DW, Oefner C,
Tsernoglou D, Brown RS, et al. 1993.
EMBO J. 12:1781–95
7. Cheng XD, Kumar S, Posfai J, Pflugrath
JW, Roberts RJ. 1993. Cell 74:299–307
8. Cheng X, Kumar S, Klimasauskas S,
Roberts RJ. 1993. Cold Spring Harbor
Symp. Quant. Biol. 58:331–38
9. Klimasauskas S, Kumar S, Roberts RJ,
Cheng XD. 1994. Cell 76:357–69
10. Kumar S, Cheng XD, Klimasauskas S, Mi
S, Posfai J, et al. 1994. Nucleic Acids Res.
22:1–10
11. Reinisch KM, Chen L, Verdine GL, Lipscomb WN. 1995. Cell 82:143–53
P1: DPI/shd/rvt
P2: NBL/rpk
July 25, 1998
10:34
198
Annual Reviews
AR057-08
ROBERTS & CHENG
12. Vassylyev DG, Kashiwagi T, Mikami Y,
Ariyoshi M, Iwai S, et al. 1995. Structure
4:1381–85
13. Roberts RJ. 1995. Cell 82:9–12
14. Slupphaug G, Mol CD, Kavli B, Arvai
AS, Krokan HE, Tainer JA. 1996. Nature
384:87–92
15. Roberts RJ, Myers PA, Morrison A, Murray K. 1976. J. Mol. Biol. 103:199–208
16. Mann MB, Smith HO. 1979. In Proc. Conf.
Transmethylation, ed. E Usdin, RT Borchardt, CR Creveling, pp. 483–92. New
York: Elsevier
17. O’Gara M, Klimasauskas S, Roberts RJ,
Cheng XD. 1996. J. Mol. Biol. 261:634–
45
18. O’Gara M, Roberts RJ, Cheng XD. 1996.
J. Mol. Biol. 263:597–606
19. Kumar S, Horton JR, Jones GD, Walker
RT, Roberts RJ, Cheng XD. 1997. Nucleic
Acids Res. 25:2773–83
20. Cheng XD. 1995. Curr. Opin. Struct. Biol.
5:4–10
21. Cheng XD. 1995. Annu. Rev. Biophys.
Biomol. Struct. 24:293–318
22. Klimasauskas S, Roberts RJ. 1995. Nucleic
Acids Res. 23:1388–95
23. Beisler JA. 1978. J. Med. Chem. 21:204–8
24. Creusot F, Acs G, Christman JK. 1982. J.
Biol. Chem. 257:2041–48
25. Wu JC, Santi DV. 1987. J. Biol. Chem.
262:4778–86
26. Yang AS, Shen J-C, Zingg J-M, Mi S, Jones
PA. 1995. Nucleic Acids Res. 23:1380–
87
27. Cheng XD, Blumenthal RM. 1996. Structure 4:639–45
28. Lange C, Wild C, Trautner TA. 1996.
EMBO J. 15:1443–50
29. Mol CD, Arvai AS, Sanderson RJ, Slupphaug G, Kavli B, et al. 1995. Cell 82:701–
8
30. Mol CD, Arvai AS, Slupphaug G, Kavli B,
Alseth I, et al. 1995. Cell 80:869–78
31. Savva R, McAuley-Hecht K, Brown T,
Pearl L. 1995. Nature 373:487–93
32. Savva R, Pearl LH. 1995. Nat. Struct. Biol.
373:752–57
33. Dodson ML, Michaels ML, Lloyd RS.
1994. J. Biol. Chem. 266:17631–39
34. McCullough AK, Scharer O, Verdine GL,
Lloyd RS. 1996. 271:32147–52
35. Vassylyev DG, Morikawa K. 1996. Structure 4:1381–85
36. Vassylyev DG, Morikawa K. 1997. Curr.
Opin. Struct. Biol. 7:103–9
37. Verdine GL, Bruner SD. 1997. Chem. Biol.
4:329–34
38. Cunningham RP. 1997. Mutat. Res. 383:
189–96
39. Labahn J, Granzin J, Schluckebier G,
Robinson DP, Jack WE, et al. 1994. Proc.
Natl. Acad. Sci. USA 91:10957–61
40. Gong W, O’Gara M, Blumenthal RM,
Cheng XD. 1997. Nucleic Acids Res. 25:
2702–15
41. Park H-W, Kim S-T, Sancar A, Deisenhofer
J. 1995. Science 268:1866–72
42. Thayer MM, Ahern H, Xing DX, Cunningham RP, Tainer JA. 1995. EMBO J.
14:4108–20
43. Yamagata Y, Kato M, Odawara K, Tokuno
Y, Nakashima Y, et al. 1996. Cell 86:311–
19
44. Labahn J, Scharer OD, Long A, EzazNikpay K, Verdine GL, Ellenberger TE.
1996. Cell 86:321–29
45. Mol CD, Kuo C-F, Thayer MM, Cunningham RP, Tainer JA. 1995. Nature 374:381–
86
46. Vrielink A, Ruger W, Driessen HPC,
Freemont PS. 1994. EMBO J. 13:3413–22
47. Moore MH, Gulbis JM, Dodson EJ, Demple B, Moody PCE. 1994. EMBO J. 13:
1495–501
48. Schluckebier G, Labahn J, Granzin J,
Schildkraut I, Saenger W. 1995. Gene 157:
131–34
49. Schluckebier G, O’Gara M, Saenger W,
Cheng XD. 1995. J. Mol. Biol. 247:16–20
50. Malone T, Blumenthal RM, Cheng XD.
1995. J. Mol. Biol. 253:618–32
51. Allan BW, Reich NO. 1996. Biochemistry
35:14757–62
52. Cal S, Connolly BA. 1997. J. Biol. Chem.
272:490–96
53. Sancar A. 1994. Biochemistry 33:2–9
54. Kuo CF, McRee DE, Fisher CL, O’Hadley
SF, Cunningham RP, Tainer JA. 1992. Science 258:434–40
55. Nikolov DB, Hu SH, Lin J, Gasch A, Hoffmann A, et al. 1992. Nature 360:40–46
56. Oefner C, Suck D. 1986. J. Mol. Biol. 192:
605–32
57. Gallinari P, Jiricny J. 1996. Nature 383:
735–38
58. Subramanya HS, Doherty AJ, Ashford SR,
Wigley DB. 1996. Cell 85:607–15
59. Guest CR, Hochstrasser RA, Sowers LC,
Millar DP. 1991. Biochemistry 30:3271–79
60. Smith SS. 1994. Prog. Nucleic Acid Res.
Mol. Biol. 49:65–111
61. Lipscomb LA, Zhou FX, Presnell SR, Woo
RJ, Peek ME, et al. 1996. Biochemistry
35:2818–23
62. Hough RF, Bass BL. 1997. RNA 3:356–70