Oncogene (2002) 21, 9022 9032

2002 Nature Publishing Group All rights reserved 0950 9232/02 $25.00
www.nature.com/onc

Structure and function of nucleases in DNA repair: shape, grip and blade of
the DNA scissors
Tatsuya Nishino1 and Kosuke Morikawa*,1
1

Department of Structural Biology, Biomolecular Engineering Research Institute (BERI), 6-2-3 Furuedai, Suita, Osaka 565-0874,
Japan

DNA nucleases catalyze the cleavage of phosphodiester

bonds. These enzymes play crucial roles in various DNA
repair processes, which involve DNA replication, base
excision repair, nucleotide excision repair, mismatch
repair, and double strand break repair. In recent years,
new nucleases involved in various DNA repair processes
have been reported, including the Mus81 : Mms4 (Eme1)
complex, which functions during the meiotic phase and
the Artemis : DNA-PK complex, which processes a V(D)J
recombination intermediate. Defects of these nucleases
cause genetic instability or severe immunodeficiency.
Thus, structural biology on various nuclease actions is
essential for the elucidation of the molecular mechanism
of complex DNA repair machinery. Three-dimensional
structural information of nucleases is also rapidly
accumulating, thus providing important insights into the
molecular architectures, as well as the DNA recognition
and cleavage mechanisms. This review focuses on the
three-dimensional structure-function relationships of
nucleases crucial for DNA repair processes.
Oncogene (2002) 21, 9022 9032. doi:10.1038/sj.onc.
1206135
Keywords: DNA repair; nuclease; metal-dependent
cleavage; protein-DNA interaction; structure-function
relationships

Introduction
Quality control of genetic material is a function
conserved in all living organisms. DNA suffers from
many environmental stresses, including attacks by
reactive oxygen species, radiation, UV light, and
carcinogens, which modify the DNA. In addition,
there are intrinsic errors and unusual structures, which
are formed during replication or recombination, and
they must be corrected by the various repair protein
machineries to avoid alterations of the base sequences
or entanglement of the DNA. These DNA repair
proteins may function independently, but in many
cases, they form complexes to perform more efficient
repair reactions. In the repair complexes, nucleases
play important roles in eliminating the damaged or

*Correspondence: K Morikawa; E-mail: morikawa@beri.or.jp

mismatched nucleotides. They also recognize the

replication or recombination intermediates to facilitate
the following reaction steps through the cleavage of
DNA strands (Table 1).
Nucleases can be regarded as molecular scissors,
which cleave phosphodiester bonds between the sugars
and the phosphate moieties of DNA. They contain
conserved minimal motifs, which usually consist of
acidic and basic residues forming the active site.
These active site residues coordinate catalytically
essential divalent cations, such as magnesium,
calcium, manganese or zinc, as a cofactor. However,
the requirements for actual cleavage, such as the types
and the numbers of metals, are very complicated, but
are not common among the nucleases. It appears that
the major role of the metals is to stabilize intermediates, thereby facilitating the phosphoryl transfer
reactions. Cleavage reactions occur either at the end
or within DNA, and thus DNA nucleases are
categorized as exonucleases and endonucleases, respectively (Figure 1). Exonucleases can be further
classified as 5 end processing or 3 end processing
enzymes, according to their polarity of consecutive
cleavage.
This review describes the three-dimensional (3D)
structural views of the actions of various nucleases
involved in many DNA repair pathways. The rapidly
accumulating genomic, biochemical and structural data
have allowed us to classify various nucleases into
folding families. In general, the nucleases involved in
DNA repair recognize the damaged moiety through the
remarkably large deformation of DNA duplexes, and
thus in terms of their DNA recognition mode, they
apparently differ from the sequence-specific endonucleases, such as the restriction enzymes. The active sites
of DNA repair nucleases have some similarity with
other nucleases, including the metal-coordinating
residues; however, they also display pronounced
diversity.
Nucleases in various categories of DNA repair
Replication
DNA polymerase replicates a new strand of DNA, the
sequence of which is complementary to the template
DNA. Most DNA polymerases in prokaryotes and
eukaryotes are composed of two different enzymes, a

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9023
Table 1

Nucleases involved in DNA repair

Prokaryote/
Bacteriophage
Replication
Proofreading
Okazaki fragment processing

PolI, II
DnaQ
RNaseH

Replication fork cleavage

Base excision Repair
Abasic site processing
Mismatchrepair
Nucleotide excision repair
5 processing
3 processing
Short patch repair
Double strand break repair
End processing

Holliday junction resolvase

Archaea

Yeast

Mammals

PolB, D

Pold, e, g

Pold, e, g

RNaseHII
FEN1

RNaseH
FEN1
Dna2
Mus81
(+Mms4[Eme1])a,b

RNaseH

Hef

Mus81
Wrn

EndoV
EndolIV
ExoIII
MutH

APN1

UvrC(+UvrB)a
UvrC
Vsr

Rad1(+Rad10)a
Rad2

XPF(+ERCC1)a
XPG

Dna2
Mre11(+Rad50)a

Mre11(+Rad50)a

RecB(+RecCD)a
SbcD(+SbcC)a
RecJ
ExoVII[RecE]b
ExoI[SbcB]b
RuvC
RusA
T4 endoVII
T7 endoI

Mre11(+Rad50)a

HAP1[APE,APEX]b

Artemis(+DNA-PK)a

Ccell[Ydc2]
Hjc

Proteins in parenthesis form a complex. bProteins in brackets are a homolog or alternative name of the protein

Figure 1 Schematic diagram of the nuclease activity. The two

strands of DNA are schematically drawn. The cleavage made
by the nuclease is represented by arrowhead

polymerase and an exonuclease, encoded within the

same polypeptide, but sometimes they are formed by
different subunits. The exonuclease degrades misincorporated DNA strand in the 3 to 5 direction (Figure 2)
(reviewed in Shevelev and Hubscher, 2002). Deletion
of these proofreading nucleases results in lethal or
strong mutator phenotypes in bacteria (Fijalkowska
and Schaaper, 1996) and in yeast (Morrison et al.,
1993), and causes cancer in mice (Goldsby et al.,
2001).
The removal of Okazaki fragments is another
important process in replication. This DNA : RNA
hybrid is required to initialize DNA polymerization,
but once the replication starts, it is rapidly degraded.

Most of the Okazaki fragments are eliminated by

RNaseH, enzyme ubiquitously present in all living
organisms. RNaseH produces nicks in the RNA region
of Okazaki fragments (Figure 2). In eukaryotes and in
archaea, FEN1 endonucleases also participate in the
removal of Okazaki fragments (reviewed in Lieber,
1997). FEN1 is a multi-functional enzyme. In addition
to the 5 to 3 exonuclease activity to remove the
Okazaki fragments, the enzyme can also generate an
incision at the junction point of a 5 flap DNA
structure. This latter activity is required to eliminate
non-homologous tails in base excision repair and in
recombination intermediates.
The replication process is stalled by various modes
of DNA damage. Upon the halt of fork progression,
the DNA polymerase and other protein complexes
abandon the replication fork. The remaining fork must
be processed by various fork-specific protein machineries. The most notable protein among them is
Mus81, which was recently found as a new fork/
junction specific endonuclease (Boddy et al., 2000,
2001; Interthal and Heyer, 2000; Kaliraman et al.,
2001; Mullen et al., 2001). Genetic and biochemical
analyses have revealed that this endonuclease is
completely conserved in eukaryotes, while its homolog
has been found in archaea. The loss of Mus81 in yeast
causes UV or methylation damage sensitivity (Interthal
and Heyer, 2000) and defects in sporulation (Mullen et
al., 2001).
Oncogene

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9024

Base Excision Repair

Figure 2 Nuclease associated DNA repair pathways. The substrate DNAs are drawn schematically and the arrowheads denote
nuclease cleavage. RNA regions are drawn in bold line

Base excision repair

Abasic sites within DNA duplexes are frequently
produced by the actions of various DNA glycosylases
involved in the base excision repair pathway, in
addition to the spontaneous hydrolysis of bases. These
apyrimidine or apurine (AP) sites are removed by AP
endonucleases which cleave the phosphdiester bond
next to an abasic site (Figure 2) (reviewed in Mol et
al., 2000a). E. coli cells contain two AP endonucleases:
endonuclease IV (endoIV) and exonuclease III
(exoIII). Interestingly, these two enzymes show no
sequence similarity to each other; although their AP
endonuclease activities are quite similar. In eukaryotes,
there seems to be a single, major AP endonuclease
working in each organism. APN1, the yeast homolog
of E. coli endoIV, shows sequence and catalytic
activity similarity to endoIV. The absence of APN1
results in enhanced sensitivity to oxidative damage and
alkylating agents (Ramotar et al., 1991). Mammalian
organisms, including humans, bear Ape1, which shares
sequence similarity with E. coli exoIII but lacks the
intrinsic 3 to 5 exonuclease activity. In addition to
the AP endonuclease activity, Ape1 also plays a major
role in sensing the redox state of the cell (Xanthoudakis et al., 1992). The loss of Ape1 generates
embryonic lethality in mice (Wilson and Thompson,
1997).
Mismatch repair
In prokaryotes, mismatch repair is conducted mainly
by the MutSLH proteins, while the Vsr protein is
Oncogene

responsible for mismatches in certain sequences

(reviewed in Modrich and Lahue, 1996; Yang, 2000;
Tsutakawa and Morikawa, 2001). In the MutSLH
system, the MutS protein recognizes and binds
mismatched base moieties of DNA. MutL mediates
the interaction between the MutS and MutH proteins.
MutH recognizes a hemimethylated GATC sequence,
and cleaves next to the G of the non-methylated strand
(Figure 2). The cleavage activity of MutH is enhanced
by the MutL protein, although its mechanism remains
unclear. Vsr is a mismatch-specific endonuclease
involved in very short patch repair, and recognizes a
TG mismatch at the specific sequence CT(A/T)GG,
where the mismatch occurs at the second thymine,
upon spontaneous deamination. Vsr makes an incision
next to the mismatched base. In both cases, after the
nick has been introduced, these sites are degraded by
the RecJ, ExoVII, or ExoI nuclease and are resynthesized by the DNA polymerase.
Nucleotide excision repair
Nucleotide excision repair (NER) is primarily used to
process DNA damage that is not repaired by base
excision repair. These forms of damage involve those
generated by the UV radiation and the large adducts
produced by various chemicals. In the NER pathway, a
short stretch of DNA containing the damaged
nucleotide is removed. During this process, two
incisions, on the 5 side and the 3 side are made by
two different nuclease reactions (Figure 2) (reviewed in
Petit and Sancar, 1999; Prakash and Prakash, 2000; de
Boer and Hoeijmakers, 2000). In bacteria, this dual

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9025

incision is performed by the UvrB-UvrC complex. In

budding yeast, Rad2 and the Rad1-Rad10 complex
make the 5 and 3 incisions, respectively. The same
process in mammalian cells is conducted by their
homologs, XPG and XPF-ERCC1, respectively. Deletions or mutations introduced into these nucleases
cause sensitivity to UV damage, and result in cancer
formation. In addition, abnormalities of these proteins
cause defects in neural development.
Double strand break repair
Double strand breaks are generated by the accidental
halt of fork progression during replication or by
ionizing radiation and strand incision chemicals. They
are also generated as an intermediate state during
meiosis and V(D)J recombination. These double strand
breaks are repaired through the two main pathways of
non-homologous end joining and homologous recombination. In either case, the ends of the double strand
breaks must be processed to initiate the repair reaction
(Figure 2). Mre11 is a multi-functional nuclease
involved in the processing of the DNA ends or
hairpin structures (reviewed in DAmours and Jackson, 2002). While Mre11 itself exhibits a ssDNA
exonuclease activity, its complex with Rad50 processes
double strand break ends. Moreover, in the presence
of ATP, Rad50 activates the cleavage activity of
Mre11. Mutations introduced into Mre11 cause an
ataxia-telangiectasia-like disorder (Stewart et al.,
1999).
V(D)J recombination involves a reaction process, in
which hairpin DNAs are opened, and subsequently,
both ends are connected. Recently, the Artemis : DNAPK complex was shown to participate in this opening
reaction (Ma et al., 2002). Although Artemis alone
possesses a ssDNA exonuclease activity, its complex
formation with DNA-PK allows the processing of the
double strand break ends to open the hairpin structure.
Defects in each protein cause severe immunodeficiency
(Blunt et al., 1995; Kirchgessner et al., 1995; Moshous
et al., 2001).
In homologous recombination, two homologous
DNA strands are paired and are connected by D-loop
structures or Holliday junction intermediates. In
bacteria, the RuvC protein cleaves the Holliday
junction at two symmetrical sites near the junction
center to resolve the junction into two dsDNAs (Figure
2). Similar junction resolving enzymes have also been
found in other bacteria, bacteriophages, and archaea
(reviewed in Sharples, 2001). In eukaryotes, FEN1,
XPF-ERCC1, and Mus81 are known to cleave the Dloop structure, while Cce1/Ydc2 processes Holliday
junctions in mitochondria.
Structural classification of DNA repair nucleases
The primary sequences of nucleases are often poorly
conserved, except for the motifs related to catalytic
sites. The functional and biochemical properties of

many nucleases have been studied extensively.

However, in some cases, it is very difficult to identify
the actual functional targets of the nucleases, because
of their broad substrate specificity. Nevertheless, many
candidates for nucleases are available from various
genome sequences, and their functional properties can
be inferred by sequence comparisons with other wellstudied nucleases. For instance, Koonin and his
associates have successfully classified nucleases, phosphoesterases, and phosphatases into several families,
based on extensive data base analyses of the primary
sequences (Aravind and Koonin, 1998a,b; Aravind et
al., 1999). This classification has also revealed the
relationships between nucleases and identified several
new nuclease families.
In addition to the classifications of primary
sequences, 3D structural data have been rapidly
accumulating with respect to the proteins involved in
DNA repair, including nucleases. Most of their
structures were solved in the DNA-free states, although
a number of them were determined in complex with
cofactors or/and DNA (Table 2). The classification of
nucleases in terms of their 3D structures provides more
defined properties, since it is accepted that the 3D
structures are much less diverged and more closely
related to the functions than the primary sequences. As
a matter of fact, in the type II restriction endonucleases, all of the structures share the common core
motif, which includes the active sites, and thus could be
grouped into a single folding family, despite its primary
sequence diversity (reviewed in Pingoud and Jeltsch,
2001). In the following section, we describe each of the
folding families of the DNA repair nucleases, classifications based on the SCOP database (Figure 3 and Table
3) (Murzin et al., 1995).
RNaseH-like fold
The RNaseH-like fold, which is one of the most
ubiquitous architectures in the protein world, has been
found in RuvC, RNaseH, integrase, transposase, and
proofreading exonucleases (Figure 3a). The core
structure contains a five-stranded b-sheet flanked by
several a-helices. The strand order is 32145, with strand
2 anti-parallel to the others. The active site residues,
which are constituted according to the DDE motif, are
located on one side of the sheet. These three (or
sometimes four) acidic residues coordinate the metals,
which are essential for the catalytic reaction. For
instance, the crystal structures of RNaseHI exhibit one
(Katayanagi et al., 1993) or two (Goedken and
Marqusee, 2001) metals bound to the active site.
Similarly, the active site of the proofreading subunit
of DNA polymerase III coordinates two metals
(Hamdan et al., 2002). The cocrystal structure with
TMP revealed that the phosphate moiety is directly
coordinated between the two metals, as it mimics the
product DNA. A similar structure is also observed in
the DNA complexes of the Klenow fragment of
polymerase I (Beese and Steitz, 1991) and the RB69
DNA polymerase (Shamoo and Steitz, 1999).
Oncogene

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9026
Table 2 Structural analysis of DNA repair proteins
Free/partial +cofactor +DNA
Replication coupled repair proteins
Replicational polymerase
(+proofreading domain)a
Repair polymerase
Error prone/free polymerase
PCNA
FEN1a

*
*
*
*

*
*
*
*

*
*

Damage Reversal
Photolyase
Ada/Ogt
MutT

*
*
*

*
*
*

Base excision repair

Aag Glycosylase
AlkA
MutM/Fpg/EndoVIII
MutY/Ogg/EndoIII
UDG
TAG
ExoIII/Ape1a
EndoIVa

*
*
*
*
*
*
*
*

*
*
*
*
*
*
*
*

*
*
*
*
*
*
*
*

Mismatch repair
MutS
MutL/Pms2
MutHa
Vsra

*
*
*
*

Nucleotide excision repair

UvrB

*
*
*

Double strand break repair

Ku70-80
Mre11a
Rad50
Xrcc4-Ligase IV
Homologous recombination
RecA/Rad51
RecG
RecJa
RuvA
RuvB
Holliday junction
resolvase (RuvC etc.)a
a

*
*
*
*
*
*

*
*
*
*

*
*

*
*

Contain nuclease activity

Resolvase-like fold
This fold has been found in gd resolvase, 5 3
exonucleases, and FEN1 (Figure 3b). It is similar to
the RNaseH-like fold, with a five-stranded b-sheet.
However, it possesses a different strand order, which is
defined as 21345 with strand 5 anti-parallel to the
others. FEN1 possesses two acidic clusters formed by
four or three conserved aspartate/glutamate residues.
These clusters each coordinate a metal, and are
separated by 5 A from each other (Hwang et al.,
1998; Hosfield et al., 1998).

five-stranded b-sheet flanked by several a-helices. The

strand order is 12345, with strand 2, and in some cases,
strand 5, anti-parallel to the others. A conserved
PDXn(D/E)XK sequence is located on one side of
the b-sheet, and is involved in the formation of the
catalytic centers in most restriction endonucleases.
Similar sequences are also found in several DNA
repair nucleases, such as MutH, Hjc, and T7 endoI,
which are categorized into essentially the same folding
family. The Vsr endonuclease also shares a similar fold,
whereas the (D/E)XK sequence is replaced by FXH,
where histidine participates in catalysis (Tsutakawa et
al., 1999b). The active sites in endonucleases with the
restriction endonuclease-like fold coordinate up to
three metals depending upon the enzyme.
RecJ-like fold
This fold was recently identified by the determination
of the RecJ nuclease structure (Yamagata et al., 2002)
(Figure 3d). Previous sequence analyses have shown
that this family includes RecJ and the phosphoesterases, which contain conserved phosphoesterase
motifs (Aravind and Koonin, 1998a). The structure
revealed a novel fold, which consists of a five-stranded
parallel b-sheet flanked by six a-helices. The strand
order of the b-sheet is 21345. On one side of the bsheet, four phosphoesterase motifs form a cluster,
which contains five invariant aspartates and two
conserved histidines. The structure of the crystal,
grown in the presence of 100 mM MnCl2, exhibits a
strong metal peak coordinating three of the aspartates
and one of the histidines. These residues, which
constitute part of the active site, are likely to
participate in the cleavage reaction.
Metallo-dependent phosphatase fold
Mre11 and several phosphatases, including the purple
acid phosphatase and the ser/thr phosphatases, share
this fold (Figure 3e). The core structure contains two
b-sheets, which are sandwiched by a-helices to form a
four-layered structure. The primary sequence of this
family contains the conserved phosphoesterase motifs
usually constituted by six histidines, three aspartates,
and an asparagine, which form a cluster on one side of
the b-sheet. The cocrystal structure of Mre11 with Mn
and dAMP shows two manganese ions bound to the
active site, and these two metals are simultaneously
coordinated to the phosphate moiety, thus mimicking
the product-bound state (Hopfner et al., 2001). The
active sites of the ser/thr phosphatases bind two metals
(zinc and iron) with a similar coordination scheme
(Griffith et al., 1995).
DNaseI-like fold

Restriction endonuclease-like fold

The structures of restriction endonucleases revealed
that their catalytic domains share common fold
architecture (Figure 3c). The core fold comprises a
Oncogene

This fold is found in DNaseI, ExoIII, and Ape1

(Figure 3F). It is also observed in some phosphatases,
such as inositol 5-phosphatase. These nucleases share a
four-layered structure containing an a/b sandwich, as

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9027

Figure 3 Folding patterns of DNA repair nucleases. The core folding is drawn schematically. The yellow arrows indicate the core
b-sheet, where the strand orders are numbered on the top. a-helices are shown as blue cylinders. The positions of the bound metals
are marked by black circles. Representative repair nuclease of the folding is written in parenthesis

Table 3

Structural classification of DNA repair nucleases

RNaseH-like fold
RNaseH
RuvC
ExoI
proofreading (exonuclease domain)
Resolvase-like fold
FEN1
Restriction endonuclease fold
MutH
Vsr
T7 endoI
Hjc
RecJ fold
RecJ
Metallophosphatase fold
Mre11
DNaseI fold
ExoIII/Ape1
TIM b/a barrel fold
EndoIV
His-Me finger nuclease fold
T4 endoVII

found in the metallo-dependent phosphatases, although

the b-sheet topology and the environments around the
active sites are different. The active site is located on
one side of the b-sheet, which assembles several
conserved acidic residues. The crystal structures of
DNaseI (Suck et al., 1988) and ExoIII (Mol et al.,
1995) revealed a single metal ion bound to the active
site. On the other hand, one (Gorman et al., 1997) or
two (Beernink et al., 2001) metals were observed in the
free form of Ape1. The Ape1-DNA complex structure
revealed one metal, coordinated with the acidic
residues and the cleaved phosphate in the active site
(Mol et al., 2000b).

TIM b/a barrel fold

The TIM barrel was first observed in triosephosphate
isomerase, and is now known to be the most
ubiquitous fold adopted by various enzymes with
diverse functions (Farber and Petsko, 1990) (Figure
3g). It forms the a8/b8 barrel structure, where a barrellike parallel b-sheet is surrounded by eight a-helices. In
this fold, the key residues for the enzymatic activity are
usually located on the C-terminal side of the barrel.
The structure of E. coli endoIV was the first DNA
repair enzyme structure with the TIM barrel (Hosfield
et al., 1999). The active site contains a cluster of three
zinc ions coordinated by histidines and aspartates. The
endoIV-DNA complex structure revealed how these
zinc ions coordinate the cleaved AP site.
His-Me finger endonuclease fold
T4 endonuclease VII (T4 endoVII) and several other
nucleases, such as the colicin nucleases, Serratia
nuclease and I-PpoI intein, contain this folding motif
(Figure 3h). It is usually embedded as a constituent of
larger architectures. The core fold is a b-hairpin
flanked by two helices. Within the hairpin, several
histidines and acidic residues form a cluster and
coordinate a catalytically important divalent metal. In
the case of T4 endoVII, a single metal ion is
coordinated to aspartate, glutamate, and asparagines
(Raaijmakers et al., 1999). The I-PpoI-DNA complex
structure revealed that a histidine lies within the
distance of hydrogen-bond from the scissile phosphate
group in the metal-containing active site (Galburt et
al., 1999).
DNA recognition by DNA repair nuclease
The binding modes of DNA nucleases are roughly
divided into two categories, corresponding to nonspecific and specific associations. Both modes are
important for efficient and accurate recognition
Oncogene

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9028

between enzymes and DNA. Non-specific DNA

binding allows enzymes to scan for target sequences
or damage by a rapid diffusion process along the
DNA. Once the nuclease finds its proper target, specific
interactions are made to dock the active site residues
correctly to the chemical groups within the DNA for
cleavage. These two binding modes have been
visualized within the crystal structures of the type II
restriction endonucleases (reviewed in Pingoud and
Jeltsch, 2001). In the cases of EcoRV, BamHI, and
PvuII, the non-specific binding involves a weak
association, which is contributed by an electrostatic
interaction between the minimum surface area of the
protein and the DNA, and the overall shape of the
DNA remains in the canonical B-form, without serious
deformations. By contrast, in the specific complex, the
DNA is buried within the deep cleft of the protein in a
sequence-specific manner, accompanied by the remarkable deformation of the DNA duplex, which is
required for the cleavage by the enzyme.
This scheme can be generally applied to DNA repair
nucleases as well. The nuclease surfaces are rich in
basic residues, which form positive surfaces competent
for electrostatic interactions with DNA. Some
nucleases, such as MutH or Vsr, which both share
the restriction endonuclease fold, possess partial
competences for sequence-specific recognition, just like
restriction endonucleases. However, most DNA repair
nucleases recognize certain mismatches, forms of
damages, or particular backbone structures of DNA.
Therefore, they require additional and unique binding
mechanisms for specific interactions with DNA.
Although the information available for DNA repair
nuclease-DNA complexes is limited, they can still
provide considerable insights into such recognition
mechanisms.
Base flipping out
Base flipping out has been observed in many DNA
glycosylases and methyltransferases (reviewed in
Roberts and Cheng, 1998; Vassylyev and Morikawa,
1997; Mol et al., 2000a; Parikh et al., 2000). The
flipping-out of a base is defined as the local
conformational change of a DNA duplex, where a
base is swung out from inside of the helix into an
extrahelical position and is usually inserted into the
binding pocket of the protein. The space created by
this process of base pair disruption is occupied by
protein atoms, which are often involved in catalytic
reactions. This mechanism is observed in the two
crystal structures of the AP endonucleases, Ape1 and
EndoIV (Figure 4a,b), which were both complexed with
DNA duplexes containing an AP site in the middle.
These two structures showed a similar base flipping,
but different fitting modes, between the DNA and the
proteins.
In the Ape1-DNA complex, the abasic nucleotide
was flipped out into the enzyme pocket (Mol et al.,
2000b) (Figure 4a). The gap was filled on the minor
groove side by two methionines (Met270, Met271) and
Oncogene

on the major groove side by arginine (Arg177). These

insertions generate a sharp kink of the DNA duplex at
the abasic site. The comparison of the free form with
the complex revealed a small difference, suggesting that
the surface of the enzyme contains a preformed pocket
to be filled by the flipped out base. Thus, it is likely
that Ape1 searches its target by scanning for a possible
base flipping site. Once Ape1 finds the target, the base
flips out into the enzyme pocket, and the remaining
gap is occupied by the inserted arginine to stabilize the
protein-DNA complex. Biochemical experiments
confirmed the role of this arginine, which when
mutated to alanine, resulted in elevated enzyme
turnover. (Mol et al., 2000b).
In the endoIV-DNA complex, an abasic site is
similarly flipped out into the protein pocket (Hosfield
et al., 1999) (Figure 4b). However, the conformation of
the DNA duplex is drastically different from that of
Ape1-DNA. The orphan base opposite the abasic site
also occupies an extrahelical position. Consequently,
the DNA duplex is sharply bent (908) at the abasic site.
The gap made by both flipped out nucleotides is filled
by arginine (Arg37), tyrosine (Tyr72), and leucine
(Leu73) inserted from the minor groove. In contrast to
the preformed pocket of Ape1, the recognition loops of
endoIV undergo a drastic conformational change upon
DNA binding. The residues involved in base flipping
are located in this loop. It is likely that endoIV scans
the DNA duplex on the minor groove side by this
DNA recognition loop. Once the enzyme finds the
target, it inserts all of the DNA-penetrating residues,
and flips the two bases into extrahelical positions.
Insertion of aromatic side chains
Another important factor in the recognition between
repair enzymes and DNA is the insertion of aromatic
amino acids into DNA duplexes. This is different from
the insertion of amino acid side chains, which fill up
the gap created by a base-flipping out. A representative
case was observed in the Vsr-DNA complex (Tsutakawa et al., 1999a) (Figure 4c). Vsr recognizes a TG
wobble mismatch base pair located in a five base pair
long recognition sequence. In the close vicinity of the
mismatch, Vsr intercalates three conserved aromatic
amino acids (Phe67, Trp68, Trp86) from the major
groove. In addition to the inserted helix from the
minor groove, this insertion expands the space between
the TG mismatch and the adjacent base pair, while the
base pair itself is not disrupted. A similar insertion of
aromatic residues was observed in the MutS-DNA
complex, where the aromatic side chain of a conserved
phenylalanine was inserted next to a mismatched or
gapped base pair (Lamers et al., 2000; Obmolova et al.,
2000).
The exonuclease domain of DNA polymerase uses
aromatic residues for the correct positioning of the
nucleotides (Figure 4d). In the editing complex of
RB69 DNA polymerase with its substrate DNA, two
single-stranded nucleotides are located in the groove of
the exonuclease domain (Shamoo and Steitz, 1999).

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9029

Figure 4 DNA recognition by DNA repair nuclease. Ribbon diagram of the DNA repair nuclease. (Left panel) Overall structure of
the protein-DNA complex. (Right panel) Close-up view of the boxed region. Proteins are represented by a yellow ribbon diagram,
and the side chains involved in DNA recognition are displayed and numbered with a stick model. The bound DNA is shown as a
white stick model, and the flipped out nucleotides are colored red. The observed metals are shown as spheres. Blue, zinc; light blue,
manganese; red, magnesium; gray, calcium. (a) endoIV (b) Ape1 (c) Vsr (d) RB69 polymerase exonuclease domain

One of the nucleotides is held by forming a hydrogen

bond with the side chain of Arg260. Another
nucleotide, whose backbone is cleaved, is located more
deeply within the exonuclease pocket, and is segregated
from the remaining region by the insertion of two
aromatic side chains (Phe123 and Phe221) to separate
the two nucleotides. These phenylalanines create a wall,
and thus the base is correctly positioned within the
active site pocket.
In vivo and in vitro experiments, measuring the UV
sensitivity and probing with potassium permanganate,
have demonstrated that in E. coli RuvC, the aromatic
side chain of Phe69 plays a crucial role in specific
recognition with the Holliday junction (Yoshikawa et
al., 2001). Phe69 lies in the protruding loop and directs
its side chain into the catalytic cleft, which accommodates one of the DNA duplexes. A similar residue is
also present in the yeast structural homolog, Ydc2,
whereas it is absent in another yeast homolog, Cce1.
Consequently, the detailed structural view of recognition mechanism between RuvC and the junction DNA
is required to solve the complex directly.
Active site environments of DNA repair nucleases
All nucleases cleave the same phosphodiester bond, to
leave 5-phosphate and 3-OH groups at the produced
segments. Similar reactions are conducted by phosphatases and ribozymes, although their catalytic
mechanisms have not been clarified yet. The overall
aspect of this enzymatic scheme is that the attacking

water is activated by a general base in the nuclease

active center, which usually bears a metal cofactor.
This activation is performed by protein side chains or
divalent metals. The activated water is converted to a
hydroxide, which attacks the phosphate, thus forming
the transition state intermediate. There are two modes
for this nucleophilic substitution: associative and
dissociative. The associative mechanism involves the
formation of a pentacovalent intermediate with a
hydroxide, followed by the release of a leaving group.
In this mechanism, a general base is required to
generate the hydroxide, and a general acid is needed to
stabilize the leaving group. The dissociative mechanism, on the other hand, does not require this general
acid and general base, and they form a metaphosphate
intermediate, which requires more stabilization of the
transition intermediate. Many nucleases are assumed to
follow the associative mechanism, while alkaline
phosphatase uses the dissociative mechanism.
A large number of nucleases utilize metal cofactors
for the hydrolytic reaction. They are proposed to play
any one or a combination of the following roles
(Figure 5) (Jencks, 1969): (1) positioning the substrate
and/or the attacking nucleophile; (2) enhancing the
nucleophilicity of the phosphate at the scissile bond; (3)
activating the nucleophile; (4) neutralizing the negative
charge in the transition state; (5) facilitating the
departure of the leaving group. To examine these
roles, various metals are recruited to the nuclease
active sites. While the utilized metal may differ,
depending upon the nuclease, magnesium or manganese is the most common metal for catalysis, and in
Oncogene

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9030

Figure 5 Schematic diagram of cleavage by DNA repair nucleases. X, Y, Z-H denote general base, Lewis acid, and general acid,
respectively. Numbers in circles indicate reaction steps where the metal cofactors may be involved

rare cases, zinc is used. The magnesium ion appears to

be transiently recruited to the active sites, whereas zinc
and manganese are more tightly bound to the catalytic
centers.
EndoIV contains three zincs, which are coordinated
by five histidines, two glutamates, and two aspartates,
in addition to two water molecules (Figure 6a). These
metals are so tightly coordinated to the enzyme that
even EDTA cannot chelate them (Levin et al., 1991).
Two of the three zinc atoms are likely to be involved in
generating the attacking nucleophile, in cooperation
with the carboxyl side chain of Glu261. Furthermore,
all three of the metals coordinate the phosphate moiety
after cleavage (Hosfield et al., 1999).
Mre11 coordinates two manganese ions through five
histidines, two aspartates, and one asparagine (Figure
6b) (Hopfner et al., 2001). The two manganese ions
directly coordinate the phosphate moiety of the dAMP.
When magnesium is substituted for manganese, they
can only occupy one of the two metal binding sites,
and the nuclease is inactive. This indicates that both
metals are required for the nuclease activity.
As for the nucleases that require magnesium cations
for catalyis, the number of metals and their positions
in the active sites are more ambiguous. They are
coordinated with protein atoms in a more transient
manner. This relatively weak binding, and the fact that
the electron number of the magnesium cation is
comparable to a water molecule, make the clear
identification of the metal positions more difficult. In
addition, the number of bound metals may change,
depending upon different crystallization conditions.
The free form structure of the Ape1 crystal, obtained
under acidic conditions, revealed a single, bound metal
(Samarium) (Gorman et al., 1997), whereas the crystal
obtained at a neutral pH contained two metals (Lead)
in the active site (Beernink et al., 2001). These Ape1
data indicate that the metals occupy multiple sites,
which are affected by the protonation of the acidic
residues. It appears that two metals are required for
catalysis, since Ape1 is only active at a neutral pH.
However, the actual numbers and the role of each
metal cannot be clarified at the moment, because the
Oncogene

structure of the Ape1-DNA complex was obtained

under acidic conditions, and only one manganese ion is
bound to the product DNA cleaved at the abasic site
(Figure 6c) (Mol et al., 2000b). Similar ambiguity with
respect to the number of metals was reported for
RNaseHI, such as one magnesium (Katayanagi et al.,
1993) and two manganeses (Goedken and Marqusee,
2001). In the Vsr-DNA complex structure, two
magnesium ions are clearly observed in the active site,
with one of the metals holding both the 5 phosphate
and 3 OH groups (Figure 6d, Tsutakawa et al., 1999a).
Two metals are also found in the exonuclease domain
of polymerases (Calcium) (Figure 6e) (Shamoo and
Steitz, 1999) and in T7 endoI (Manganese) (Hadden et
al., 2002), although the two sites are not equivalent
between the two enzymes, and one of the two metals
shows partial occupancy.
Future perspectives
With the rapid accumulation of metal binding site
information, various catalytic mechanisms have been
proposed, including the classical two metal binding
mechanism (Beese and Steitz, 1991). However, it
appears to us that the actual numbers and positions
of the metals involved in catalysis are too broadly
varied from enzyme to enzyme to describe their
hydrolytic mechanisms by a unified catalytic scheme.
More detailed structural information, hopefully
combined with biochemical data, is essential to obtain
clear insights into the metal dependent nuclease
mechanisms. Meanwhile, the large diversity in nuclease
architectures suggests that they can specifically recognize DNA substrates by virtue of the large variety of
surface properties, which were adopted through
selection over an extremely long period. In particular,
the nucleases involved in DNA repair have acquired a
special damage recognition system. At present, much of
the structural information is based on that of
prokaryotic and archaeal proteins. Eukaryotic
nucleases obviously hold more complicated structures
and properties, because they must bear eukaryote-

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9031

Figure 6 Active site of DNA repair nuclease. Close-up view of the nuclease active site, shown in a stereo diagram. Residues involved in the nuclease activity and the metal coordination are drawn in stick models. The coloring scheme is same as in Figure
4. (a) endoIV (b) Mre11 (c) Ape1 (d) Vsr (e) RB69 polymerase exonuclease domain

specific regulatory mechanisms involving protein

protein and protein-DNA interactions. Further 3D
structural characterizations of the eukaryotic DNA
repair nucleases should provide additional variations or
conserved architectures of protein folding, while
structural analyses of their complexes with DNA
substrates will clarify the recognition mechanisms.

Acknowledgements
We regret that the limit of space may have not allowed us
to site all works in the field. We thank Kayoko Komori for
critical reading of the manuscript and helpful comments. T
Nishino is a research fellow of the Japan society for the
promotion of sciences. This research was partly supported
by NEDO (New Energy and Industrial Technology
Development Organization).

References
Aravind L and Koonin EV. (1998a). Trends Biochem. Sci.,
23, 17 19.
Aravind L and Koonin EV. (1998b). Nucleic Acids Res., 26,
3746 3752.
Aravind L, Walker DR and Koonin EV. (1999). Nucleic
Acids Res., 27, 1223 1242.
Beernink PT, Segelke BW, Hadi MZ, Erzberger JP, Wilson
III DM and Rupp B. (2001). J. Mol. Biol., 307, 1023 1034.
Beese LS and Steitz TA. (1991). EMBO J., 10, 25 33.
Blunt T, Finnie NJ, Taccioli GE, Smith GC, Demengeot J,
Gottlieb TM, Mizuta R, Varghese AJ, Alt FW and Jeggo
PA. (1995). Cell,, 80, 813 823.
Boddy MN, Gaillard PH, McDonald WH, Shanahan P,
Yates III JR and Russell P. (2001). Cell, 107, 537 548.

Boddy MN, Lopez-Girona A, Shanahan P, Interthal H,

Heyer WD and Russell P. (2000). Mol. Cell. Biol., 20,
8758 8766.
DAmours D and Jackson SP. (2002). Nat. Rev. Mol. Cell.
Biol., 3, 317 327.
de Boer J and Hoeijmakers JH. (2000). Carcinogenesis, 21,
453 460.
Farber GK and Petsko GA. (1990). Trends Biochem. Sci., 15,
228 234.
Fijalkowska IJ and Schaaper RM. (1996). Proc. Natl. Acad.
Sci. USA, 93, 2856 2861.
Galburt EA, Chevalier B, Tang W, Jurica MS, Flick KE,
Monnat Jr RJ and Stoddard BL. (1999). Nat. Struct. Biol.,
6, 1096 1099.

Oncogene

Structure and function of DNA repair nuclease

T Nishino and K Morikawa

9032

Goedken ER and Marqusee S. (2001). J. Biol. Chem., 276,

7266 7271.
Goldsby RE, Lawrence NA, Hays LE, Olmsted EA, Chen X,
Singh M and Preston BD. (2001). Nat. Med., 7, 638 639.
Gorman MA, Morera S, Rothwell DG, de La FortelleE, Mol
CD, Tainer JA, Hickson ID and Freemont PS. (1997).
EMBO J., 16, 6548 6558.
Griffith JP, Kim JL, Kim EE, Sintchak MD, Thomson JA,
Fitzgibbon MJ, Fleming MA, Caron PR, Hsiao K and
Navia MA. (1995). Cell, 82, 507 522.
Hadden JM, Declais AC, Phillips SE and Lilley DM. (2002).
EMBO J., 21, 3505 3515.
Hamdan S, Carr PD, Brown SE, Ollis DL and Dixon NE.
(2002). Structure (Camb), 10, 535 546.
Hopfner KP, Karcher A, Craig L, Woo TT, Carney JP and
Tainer JA. (2001). Cell, 105, 473 485.
Hosfield DJ, Mol CD, Shen B and Tainer JA. (1998). Cell,
95, 135 146.
Hosfield DJ, Guan Y, Haas BJ, Cunningham RP and Tainer
JA. (1999). Cell, 98, 397 408.
Hwang KY, Baek K, Kim HY and Cho Y. (1998). Nat.
Struct. Biol., 5, 707 713.
Interthal H and Heyer WD. (2000). Mol. Gen. Genet., 263,
812 827.
Jencks WP. (1969). Catalysis in Chemistry and Enzymology.
New York: McGraw Hill, pp 111 115.
Kaliraman V, Mullen JR, Fricke WM, Bastin-Shanower SA
and Brill SJ. (2001). Genes Dev., 15, 2730 2740.
Katayanagi K, Okumura M and Morikawa K. (1993).
Proteins, 17, 337 346.
Kirchgessner CU, Patil CK, Evans JW, Cuomo CA, Fried
LM, Carter T, Oettinger MA and Brown JM. (1995).
Science, 267, 1178 1183.
Lamers MH, Perrakis A, Enzlin JH, Winterwerp HH, de
Wind N and Sixma TK. (2000). Nature, 407, 711 717.
Levin JD, Shapiro R and Demple B. (1991). J. Biol. Chem.,
266, 22893 22898.
Lieber MR. (1997). Bioessays, 19, 233 240.
Ma Y, Pannicke U, Schwarz K and Lieber MR. (2002). Cell,
108, 781 794.
Modrich P and Lahue R. (1996). Annu. Rev. Biochem., 65,
101 133.
Mol CD, Hosfield DJ and Tainer JA. (2000a). Mutat. Res.,
460, 211 229.
Mol CD, Izumi T, Mitra S and Tainer JA. (2000b). Nature,
403, 451 456.
Mol CD, Kuo CF, Thayer MM, Cunningham RP and Tainer
JA. (1995). Nature, 374, 381 386.
Morrison A, Johnson AL, Johnston LH and Sugino A.
(1993). EMBO J., 12, 1467 1473.

Oncogene

Moshous D, Callebaut I, de Chasseval R, Corneo B,

Cavazzana-Calvo M, Le Deist F, Tezcan I, Sanal O,
Bertrand Y, Philippe N, Fischer A and de Villartay JP.
(2001). Cell, 105, 177 186.
Mullen JR, Kaliraman V, Ibrahim SS and Brill SJ. (2001).
Genetics, 157, 103 118.
Murzin AG, Brenner SE, Hubbard T and Chothia C. (1995).
J. Mol. Biol., 247, 536 540.
Obmolova G, Ban C, Hsieh P and Yang W. (2000). Nature,
407, 703 710.
Parikh SS, Putnam CD and Tainer JA. (2000). Mutat. Res.,
460, 183 199.
Petit C and Sancar A. (1999). Biochimie, 81, 15 25.
Pingoud A and Jeltsch A. (2001). Nucleic Acids Res., 29,
3705 3727.
Prakash S and Prakash L. (2000). Mutat. Res., 451, 13 24.
Raaijmakers H, Vix O, Toro I, Golz S, Kemper B and Suck
D. (1999). EMBO J., 18, 1447 1458.
Ramotar D, Popoff SC, Gralla EB and Demple B. (1991).
Mol. Cell. Biol., 11, 4537 4544.
Roberts RJ and Cheng X. (1998). Annu. Rev. Biochem., 67,
181 198.
Shamoo Y and Steitz TA. (1999). Cell, 99, 155 166.
Sharples GJ. (2001). Mol. Microbiol., 39, 823 834.
Shevelev IV and Hubscher U. (2002). Nat. Rev. Mol. Cell.
Biol., 3, 364 376.
Stewart GS, Maser RS, Stankovic T, Bressan DA, Kaplan
MI, Jaspers NG, Raams A, Byrd PJ, Petrini JH and Taylor
AM. (1999). Cell, 99, 577 587.
Suck D, Lahm A and Oefner C. (1988). Nature, 332, 464
468.
Tsutakawa SE, Jingami H and Morikawa K. (1999a). Cell,
99, 615 623.
Tsutakawa SE, Muto T, Kawate T, Jingami H, Kunishima
N, Ariyoshi M, Kohda D, Nakagawa M and Morikawa K.
(1999b). Mol. Cell, 3, 621 628.
Tsutakawa SE and Morikawa K. (2001). Nucleic Acids Res.,
19, 3775 3783.
Vassylyev D and Morikawa K. (1997). Curr. Opin. Struct.
Biol., 7, 103 109.
Wilson III DM and Thompson LH. (1997). Proc. Natl. Acad.
Sci. USA, 94, 12754 12757.
Xanthoudakis S, Miao G, Wang F, Pan YC and Curran T.
(1992). EMBO J., 11, 3323 3335.
Yamagata A, Kakuta Y, Masui R and Fukuyama K. (2002).
Proc. Natl. Acad. Sci. USA, 99, 5908 5912.
Yang W. (2000). Mutat. Res., 460, 245 256.
Yoshikawa M, Iwasaki H and Shinagawa H. (2001). J. Biol.
Chem., 276, 10432 10436.