Molecular Cell, Vol. 17, 657–670, March 4, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.molcel.2005.02.012
The Molecular Architecture of the Mammalian
DNA Repair Enzyme, Polynucleotide Kinase
Nina K. Bernstein,1R. Scott Williams,1
Melissa L. Rakovszky,1Diana Cui,1Ruth Green,1
Feridoun Karimi-Busheri,2Rajam S. Mani,2
Sarah Galicia,3C. Anne Koch,4Carol E. Cass,2
Daniel Durocher,3Michael Weinfeld,2
and J.N. Mark Glover1,*
1Department of Biochemistry
4–74 Medical Sciences Building
University of Alberta
Edmonton, Alberta T6G 2H7
2Experimental Oncology, Cross Cancer Institute,
and Department of Oncology
University of Alberta
Edmonton, Alberta T6G 1Z2
3Samuel Lunenfeld Research Institute
Mount Sinai Hospital
600 University Avenue
Toronto, Ontario M5G 1X5
4Department of Radiation Oncology
Princess Margaret Hospital
610 University Avenue
Toronto, Ontario M5G 2M9
initial genotoxic event, but also as an intermediate
product in several DNA repair pathways. As a result, all
organisms have evolved highly specialized DNA repair
polymerases and ligases to efficiently process strand
breaks. All DNA polymerases and ligases character-
ized to date are highly selective for the type of DNA
ends that can be utilized; all require 5#-phosphate and
3#-hydroxyl DNA termini. This requirement, however,
creates further complications for repair, as many DNA-
damaging agents and upstream repair processes often-
generate DNA ends that are incompatible with the
polymerases and ligases. DNA backbone breaks with
5#-hydroxyl termini can result from ionizing radiation,
DNase II action, and antineoplastic agents such as
camptothecin. Sources of 3#-phosphate at breaks in-
clude ionizing radiation, DNase II action, and Tdp1
cleavage of camptothecin-trapped topoisomerase I
DNA adducts (Wang, 1996) or 3#-phosphoglycolates at
some double-strand breaks (Zhou et al., 2005). Further-
more, 3#-phosphate termini are transiently generated
during repair via the base excision repair (BER) path-
way, the primary mechanism by which small base
lesions and single-strand breaks are repaired (Izumi
et al., 2003; Wiederhold et al., 2004). Mammalian poly-
nucleotide kinase (PNK) (Pheiffer and Zimmerman, 1982;
Habraken and Verly, 1983) possesses dual catalytic
activities, a 5# DNA kinase and a 3# phosphatase, and
is the principal enzyme responsible for restoring 5#-
phosphate and 3#-hydroxyl at DNA-strand breaks.
Although mammalian PNK has long been suspected
to play a key role in DNA repair (Teraoka et al., 1975),
only recently have conclusive experiments been per-
formed to implicate this enzyme in the repair of both
single-strand breaks (SSBs) and double-strand breaks
(DSBs). PNK has been shown to stimulate SSB repair
in both in vitro reconstitution experiments (Karimi-Busheri
et al., 1998; Whitehouse et al., 2001; Wiederhold et al.,
2004) and in vivo studies (Loizou et al., 2004). PNK has
also been shown to be an integral component of the
nonhomologous end-joining (NHEJ) pathway, which is
the major route for the repair of double-strand breaks
in mammalian cells (Chappell et al., 2002; Koch et al.,
2004). Consistent with a central role for PNK in both of
these processes, knockdown of endogenous PNK in a
human cell line leads to an increase in spontaneous
mutations and enhanced sensitivity to a broad range of
genotoxic agents (Rasouli-Nia et al., 2004). In S. pombe,
deletion of Pnk1, a homolog of PNK, produces similar
effects (Meijer et al., 2002). Not only does PNK play a
role in the repair of strand breaks, it functions in the
repair of base lesions as part of the BER pathway. The
recently discovered DNA glycosylases/AP lyases NEIL1
and NEIL2 initiate AP endonuclease-independent BER
and interact with PNK in a multiprotein BER complex
(Hazra et al., 2002; Rosenquist et al., 2003; Wiederhold
et al., 2004). These enzymes excise abasic (AP) sites
leaving breaks with 3#-phosphate termini. Although AP-
endonuclease was previously thought to process
3#-phosphates, it has now been shown to do so much
Mammalian polynucleotide kinase (PNK) is a key com-
ponent of both the base excision repair (BER) and
nonhomologous end-joining (NHEJ) DNA repair path-
ways. PNK acts as a 5?-kinase/3?-phosphatase to cre-
ate 5?-phosphate/3?-hydroxyl termini, which are a nec-
essary prerequisite for ligation during repair. PNK is
recruited to repair complexes through interactions
between its N-terminal FHA domain and phosphory-
lated components of either pathway. Here, we de-
scribe the crystal structure of intact mammalian PNK
and a structure of the PNK FHA bound to a cognate
phosphopeptide. The kinase domain has a broad sub-
strate binding pocket, which preferentially recognizes
double-stranded substrates with recessed 5? termini.
In contrast, the phosphatase domain efficiently de-
phosphorylates single-stranded 3?-phospho termini
as well as double-stranded substrates. The FHA do-
main is linked to the kinase/phosphatase catalytic do-
main by a flexible tether, and it exhibits a mode of
target selection based on electrostatic complemen-
tarity between the binding surface and the phos-
Scission of the DNA backbone is a common form of
damage that can arise not only as a direct result of an
more slowly than PNK, leaving PNK as the major cellu-
lar 3# phosphatase (Wiederhold et al., 2004).
Many DNA repair enzymes appear to be actively re-
cruited to specific repair complexes to ensure that po-
tentially mutagenic intermediates along the repair path-
way are not released prematurely (Mitra et al., 2002).
PNK contains, in addition to its catalytic kinase and
phosphatase domains, an N-terminal FHA domain (Cal-
decott, 2003), which has recently been shown to target
the enzyme to sites of BER and NHEJ repair. XRCC1
and XRCC4, central components of the BER and NHEJ
pathways, respectively, are phosphorylated at threo-
nine residues by CK2 (Koch et al., 2004; Loizou et al.,
2004). The FHA domain of PNK specifically recognizes
these phosphorylated forms of XRCC1 and XRCC4, di-
recting PNK to the site of repair. The PNK FHA exhibits
distinct preferences for peptide sequences N-terminal
to the phosphothreonine. In contrast, several other
well-characterized FHA domains show specificity for
the sequences C-terminal to the phosphothreonine.
The PNK FHA is highly similar to the FHA domain of
aprataxin (APTX), a protein associated with the neuro-
logical disorder ataxia-oculomotor apraxia and likely
also involved in DNA repair (Moreira et al., 2001;
Gueven et al., 2004). Indeed, aprataxin is found in vivo
in complexes with XRCC1/Lig III and XRCC4/Lig IV, in
parallel with but exclusive of PNK. A role for aprataxin
in regulating PNK activity has therefore been proposed
(Luo et al., 2004; Clements et al., 2004).
Another well-known bifunctional polynucleotide ki-
nase/phosphatase, T4 PNK, has been extensively char-
acterized structurally and functionally. In addition to a
battery of mutagenesis and activity experiments, crys-
tal structures have been determined for the kinase do-
main of T4 PNK, as well as for the full-length enzyme,
with and without bound oligonucleotide substrates
(Galburt et al., 2002; Wang et al., 2002a; Eastberg et al.,
2004). Aside from the similar catalytic activities, T4 and
mammalian PNK display profound differences. T4 PNK
is a tetramer in vivo, whereas mammalian PNK is a mo-
nomer (Midgley and Murray, 1985; Mani et al., 2001).
The T4 kinase is quite promiscuous in its substrate
choice, accepting RNA, DNA, as well as mono and oli-
gonucleotides (Kleppe and Lillehaug, 1979). The mam-
malian kinase is specific for DNA and requires a mini-
mal substrate length of 8 nucleotides (Karimi-Busheri
and Weinfeld, 1997). Furthermore, while the T4 kinase
clearly prefers protruding 5# ends, the mammalian ki-
nase shows a preference for recessed DNA ends (Kar-
imi-Busheri and Weinfeld, 1997). Finally, T4 PNK lacks
the FHA domain. These contrasts are consistent with
these enzymes’ different biological roles. T4 PNK does
not play a role in DNA repair, but instead participates
in the repair of a specifically cleaved bacterial lysine
tRNA as part of a phage-encoded system to evade the
host cell defense (Amitsur et al., 1987). T4 and mamma-
lian PNKs share virtually no sequence similarity, aside
from the core elements of the kinase and phosphatase
active sites, the P loop and the DxDGT motifs, respec-
tively. Moreover, the kinase and phosphatase are re-
versed in sequence in the two enzymes.
We have determined the crystal structure of full-
length mouse PNK (mPNK). In agreement with domain-
mapping experiments and sequence analysis, the pro-
tein comprises three domains: kinase, phosphatase
and FHA. We have refined the definition of substrate
preferences for the kinase and phosphatase using ac-
tivity and binding assays. Based on the biochemical
data, our structures, and structures of related proteins,
we have proposed the modes of substrate binding to
both the kinase and phosphatase domains of mPNK.
Furthermore, to characterize the role of PNK in multi-
protein assemblies, we have determined the crystal
structure of the mouse PNK FHA domain in complex
with the XRCC4-derived phosphopeptide. While the
mode of phosphothreonine binding is highly conserved
in PNK relative to the other FHA domains of known
structure, the elements of peptide recognition differ
significantly. This structure suggests that the PNK FHA
has an adaptable peptide binding specificity that al-
lows it to recognize highly negatively charged CK2-
phosphorylated targets within XRCC1 and XRCC4.
Results and Discussion
Structural Overview of PNK
The crystal structure of full-length mouse PNK was de-
termined by Se-MAD methods and refined at a resolu-
tion of 2.8 Å (Table 1). The structure reveals that PNK is
folded into three compact domains, an N-terminal FHA
domain (Ser 6–Ser 110), the phosphatase domain (Gly
145–Glu 336), and the kinase domain (Phe 340–Glu 521)
(Figures 1A and 1B). The kinase and phosphatase to-
gether form the catalytic domain. We observe two
NCS-related catalytic domains per asymmetric unit, but
only a single FHA domain with weaker electron density.
The second FHA domain is not visible in the electron
density and is presumably completely disordered.
Since the 30-residue linker between the FHA domain
and the catalytic domain is undefined in the structure,
we cannot assign the observed FHA to either of the
NCS-related catalytic domains. The interface between
the FHA domain and the closest phosphatase domain,
which buries only a small region of polar surface (less
than 850 Å2), likely results from crystal packing and
probably does not represent interactions that are stable
in solution. The flexible organization of the FHA and
catalytic domains relative to one another is consistent
with results of limited proteolysis experiments (Figure
1C). Trypsin digestion rapidly converts the full-length
PNK to a pair of fragments: the catalytic domain (40
kDa) and the FHA domain (14 kDa). SDS-PAGE analysis
of PNK crystals confirmed that they contained full-
length PNK and not proteolyzed fragments (Figure 1C).
The readily cleavable linker between the FHA and cata-
lytic domains accounts for the frequent recovery of
PNK as a 40 kDa protein from tissue extracts, (Bosdal
and Lillehaug, 1985; Karimi-Busheri and Weinfeld,
1997). The flexibility of the FHA domain relative to the
catalytic domain supports the idea that FHA has no di-
rect role in catalysis by PNK, and instead serves as a
In contrast, the kinase and phosphatase domains con-
tact one another through a large interface involving the
interdomain linker (Leu 337–Ala 339) and the C-terminal
residues Gln 518–Gly 522, which pack against helix 5
and the β12-α4 loop of the phosphatase (Figure 1A).
Structure of Mammalian PNK
Table 1. Crystallographic Data Collection, Phasing, and Refinement Statistics
Data Collection mPNKFHA
Resolution range (Å)
Data coverage total/final shella(%)
<I/σI> total/final shell
Rsymtotal/final shell (%)b
Resolution range (Å)
No. of selenium sites
FOM – solve
FOM – resolve
Resolution range (Å)
No. of refined atoms
Average B factors (Å2)
Data from wavelength λ1 were used during crystallographic refinement. Rfreecalculated with 5% of all reflections excluded from refinement
stages using the native data set. No I/σI cutoff was used in the refinement.
aFinal shell: λ1: 2.90 – 2.80 Å, λ2: 3.00 – 3.10 Å, λ3: 3.21 – 3.10 Å, FHA: 2.24 – 2.20.
bRsym= Σ |(Ihkl) − <I>|/Σ(Ihkl) where Ihklis the integrated intensity of a given reflection.
cRwork= Σh|Fo(h) − Fc(h)|/Σh|Fo(h)|, where Fo(h) and Fc(h) are observed and calculated structure factors.
The intimate association explains why the catalytic do-
mains could not be separated by proteolysis (Figure
1C). Despite the tight association of the kinase and
phosphatase domains, the two PNK molecules in the
crystallographic asymmetric unit display a moderate
degree of flexibility (10° rotation) in domain orientation.
While the flexibility does not perturb either active site,
it may enable PNK to adjust to different DNA-strand
Although the kinase and phosphatase domains are
reversed in order in the sequence of mammalian and
T4 PNK, in the three-dimensional structure their spatial
orientation is similar. The phosphatase is located on the
same side of the kinase in both molecules, but a signifi-
cant shift between the domains alters the relative orien-
tation of the kinase and phosphatase active sites. In T4
PNK the active sites of one monomer point in opposite
directions, whereas in mammalian PNK, the two active
sites face approximately the same direction. However,
the tetrameric organization of T4 PNK stabilizes the
relative orientation of the kinase and phosphatase and
allows the substrate a choice between several combi-
nations of active sites. Thus, a tRNA substrate was pro-
posed to interact simultaneously with kinase and phos-
phatase sites facing each other across the tetramer
interface, contributed by two monomers (Galburt et al.,
2002). Mammalian PNK, on the other hand, is a mono-
mer, where only the double interdomain linkage stabi-
lizes the domain orientation. Because in mPNK there is
only a single kinase and phosphatase active site, it is
convenient for them to face in the same direction, pre-
sumably from where all substrates approach.
The Kinase Domain
Early experiments on PNK revealed major differences
in substrate preference between the mammalian kinase
and the T4 kinase, consistent with the known biological
functions of the two enzymes (Karimi-Busheri et al.,
1998). T4 PNK was shown to prefer 5#-overhanging
substrates over either blunt or 5#-recessed termini,
whereas mammalian PNK was shown to preferentially
phosphorylate nicks and small gaps compared to sin-
Figure 1. Structure of Mouse PNK
(A) Ribbon diagram of mPNK, with kinase in yellow, phosphatase in blue and FHA domain in green. Catalytic side chains (Asp 170 and Asp
396 in the phosphatase and kinase, respectively) are in pink, the ATP binding P loop is in navy blue, and the sulfate bound at the P loop is in
orange and red spheres.
Structure of Mammalian PNK
gle-stranded substrates (Lillehaug et al., 1976; Karimi-
Busheri and Weinfeld, 1997).
To further define the kinase substrate preferences of
PNK, we assessed the ability of mPNK to phosphorylate
5#-hydroxyl groups in a series of DNA substrates with two
possible sites of phosphorylation, a 5#-hydroxyl at a
blunt, double-stranded end, and a second 5#-hydroxyl
at an internal site, corresponding to a gap or a nick
(Figure 2A). At low concentration, mPNK exhibited a
clear preference for the 5# recessed hydroxyl over the
blunt-ended substrate (Figure 2B), as well as over single-
stranded DNA (data not shown). However, as the enzyme
concentration increased, the selectivity for the recessed
5#-hydroxyl was lost. We next explored the optimal struc-
ture around the 5# recessed hydroxyl, varying the size
of the gap in our model substrate (Figure 2C). Our re-
sults show that varying the gap from 0 to 10 nucleo-
tides (Figure 2, substrates a–e) did not have any notice-
able effect on the amount of phosphorylation at the
internal 5#-hydroxyl. In addition, two-stranded sub-
strates with a 3# single-stranded overhang (Figure 2,
substrates f–h) exhibited the same extent of phosphor-
ylation as matched nicked or gapped substrates, sug-
gesting that the optimal kinase substrate is a recessed
5#-hydroxyl. Finally, we examined the minimal 3# over-
hang required by mPNK (Figure 2D). A set of DNA du-
plexes with 3# overhangs of 0, 3, 5, and 8 nucleotides
(Figure 2, substrates i–m) was tested for 5# phosphory-
lation. We found that selectivity for the recessed 5#-
hydroxyl appeared with 3# overhangs of between 3 and
To directly probe the affinity of mPNK for DNA sub-
strates, we utilized an assay which monitors the change
of intrinsic, mPNK tryptophan fluorescence upon bind-
ing to DNA substrates (Mani et al., 2003). Binding affini-
ties were measured for the duplex substrate containing
a 5 nucleotide 3# overhang, as well as for the constitu-
ent 13- or 8-mer oligonucleotides. The mPNK-DNA in-
teraction was accompanied by a partial quenching of
fluorescence with no change in the emission maximum.
Binding affinity (Kd) and stoichiometry were determined
from best fits to plots of fluorescence quenching as a
function of DNA concentration (Figure 2E). In each
case, the data indicate that mPNK binds its DNA sub-
strate at a single site with 1:1 stoichiometry. The over-
hanging duplex DNA was bound with an apparent Kdof
0.36 ± 0.04 ?M, while both the 13- and 8-mer DNAs
were bound with significantly higher Kds of 1.0 ± 0.1
and 1.4 ± 0.1 ?M, respectively.
Taken together, these experiments demonstrate that
mammalian PNK selectively binds and phosphorylates
5#-hydroxyl termini recessed by more than 3 nucleo-
tides. The length of the gap following the 5#-hydroxyl
appears to be immaterial. Indeed, the overhanging DNA
may be entirely single-stranded with no decrease in re-
cessed 5#-hydroxyl phosphorylation.
Structure of the Kinase Domain
The 189-residue kinase domain of mPNK is larger than
its similarly folded counterpart in T4 PNK, which con-
tains only 148 residues. This domain, which belongs to
the adenylate kinase family, consists of a 5-stranded
parallel β sheet, common to GTPases and P loop ki-
nases (Leipe et al., 2003), flanked by helices on both
sides. In addition, three helices (α12, 13, and 15) lie be-
tween the α/β sandwich and the phosphatase domain.
The T4 kinase is smaller, containing only a 4-stranded
parallel sheet with comparable topology.
The kinase active site is located in a long cleft, with
the ATP binding site at one end and the substrate bind-
ing site at the other (Figure 3A). While we have not yet
been able to crystallize a nucleotide-bound form of
mPNK, we can define the ATP binding site by the con-
served Walker A and B motifs present in various P loop
kinases (Leipe et al., 2003). The Walker A motif, the P
loop (Gly 371–Ser 378), binds the β- and γ-phosphates
of ATP and adopts a conformation essentially identical
to that found in T4 PNK. In the Walker B motif, Asp 421
forms a hydrogen bond with Ser 378 for proper posi-
tioning of the two motifs and coordinates Mg2+, which
is not observed in our structure but is required for ki-
nase activity. This motif is generally less conserved
among the P loop NTPases, and is replaced by a serine
in T4 PNK. The ATP binding site is completed by the lid
subdomain (helices 12 and 13), which folds over the P
loop. In our structure of mPNK a sulfate anion is located
at the position corresponding to the ADP β-phosphate
in the T4 structure (Galburt et al., 2002). The contacts
made by the sulfate and the ADP β-phosphate with the
P loop are the same in both structures, but the lid dif-
fers slightly in orientation. Sequence and structural
analyses of lid subdomains of P loop kinases identified
two conserved arginines (Rx(2-3)R) in the first lid helix
(α12 in mPNK), which interact with the bound ATP,
stacking with the adenine and forming salt bridges with
β- and γ- phosphates (Leipe et al., 2003). In T4 PNK Arg
122 and Arg 126 perform these roles. In mPNK, how-
ever, the corresponding residues in sequence, Arg 457
and Arg 461, are on the opposite face of helix 12, point-
ing away from the nucleotide binding site. Arg 463 from
helix 12 forms hydrogen bonds with the bound sulfate
(Figure 3A). When ADP from the T4 PNK structure is
superimposed, however, the side chain of Arg 463
(B) Sequence comparison of mouse PNK (mPNK) with human PNK (hPNK). The FHA, phosphatase and kinase domains are highlighted in
green, blue and yellow, respectively. In addition, the FHA domain is aligned with the human aprataxin FHA domain (APTX), and the phospha-
tase and kinase domains are structurally aligned with T4 PNK phosphatase and kinase (T4p and T4k). mPNK secondary structure is shown
above the sequence with every tenth residue of mPNK marked (|). Helices in the secondary structure are labeled sequentially, with α helices
indicated by ‘α’ and 310helices by ‘h’. In the FHA domain phosphothreonine- and peptide binding residues are marked with “P” and green
spheres, respectively. In the phosphatase, residues conserved at the active site are labeled with blue spheres, and residues predicted to bind
the substrate backbone, with purple spheres. In the kinase, ATP binding motifs (Walker A and B) are highlighted in purple. Yellow spheres
mark the catalytic Asp and residues predicted to bind the substrate backbone phosphate, and red spheres mark positive surface patches
predicted to interact with distal backbone phosphates.
(C) On the left, limited trypsin proteolysis of mPNK. On the right, SDS-PAGE gel of mPNK crystal (lane b) vs. a partially degraded mPNK in
solution (lane a). All figures showing structure were generated with PyMOL (http://www.pymol.org).
Figure 2. Elucidation of Kinase Substrate Specificity
(A) Substrates tested in kinase activity and binding assays. Circles at termini represent 5#-OH, and numbers above or below the individual
strands represent oligonucleotide length. (See Supplemental Data for sequences.)
(B) Phosphorylation of a fixed concentration (5 ?M) of substrate (b) by increasing concentrations of mPNK.
(C) Phosphorylation of model substrates with variable gap size (a–e) and the corresponding 5#-recessed OH (f–h) at 0.5 ?M by two concentra-
tions of mPNK.
(D) Determination of minimal length of 3# overhang. Substrates (i–m) are phosphorylated by increasing concentrations of mPNK. Strand 8*
runs faster on the denaturing gel due to sequence differences.
(E) Fluorescence titration of mPNK with substrate (k) (duplex) or its constituent oligonucleotides (13-mer and 8-mer). The inset figure shows
a sample raw data plot for the duplex.
clashes with the ribose ring, and would need to be ro-
tated upon ATP binding. Alternately, ATP may adopt dif-
ferent conformations in mammalian and T4 PNK or the
lid may refold upon ATP binding. CD studies of human
PNK showed a significant decrease in α-helical struc-
ture upon ATP binding (Mani et al., 2003).
The substrate binding end of the active site cleft also
shows gross differences between mammalian and T4
PNK, not surprising in light of their divergent substrate
preferences. The substrate binding site is more open in
mPNK than in T4 PNK. In the phage enzyme an inser-
tion (Gln 39–Ile 59) folded into a longer loop after helix
2 (α9 in mPNK) and a longer helix 3 (α10 in mPNK),
extends over the substrate binding site, enclosing the
active site in a tunnel narrow enough to admit only sin-
gle-stranded DNA (Figure 3B). In mPNK the active site
is more accessible to solvent and to substrate, which
is generally larger than the optimal T4 kinase substrate.
A Model for the Kinase/Substrate Complex
Based on our kinase assay experiments and previous
work (Karimi-Busheri and Weinfeld, 1997), we can de-
fine a minimal preferred kinase substrate as a DNA du-
plex with 8 base pairs upstream of the 5#-hydroxyl that
is recessed by 4–5 nucleotides. In an attempt to better
Structure of Mammalian PNK
Figure 3. The Kinase Domain
(A) mPNK active site cleft. The Walker A and Walker B motifs, forming the ATP binding site, are shown in green. Sulfate ions bound at the
active site are in orange and red. S1 corresponds to the β-phosphate of ATP, while S2 is proposed to mimic the binding of a DNA backbone
phosphate. The side chains of residues involved in catalysis or proposed to bind the DNA substrate are shown in yellow. Water molecule
W1(red sphere) is proposed to correspond to the substrate 5#-hydroxyl.
(B) Surface representations of the kinase domains of mPNK and T4 PNK, viewed from the substrate approach side. mPNK is shown with the
bound sulfate S2, and T4 PNK is shown bound to oligonucleotide 5#-GTCAC-3# (PDB ID 1RC8).
(C) Surface representation of mPNK with the modeled DNA substrate, which contains 8 base pairs and a 5#-hydroxyl recessed by 4 nucleo-
understand the structural basis for the substrate prefer-
ence of mPNK, we docked the modeled minimal pre-
ferred substrate to the kinase domain. Structural studies
of T4 kinase-DNA substrate complexes have revealed
particularly critical interactions with the 5#-hydroxyl
and 3#-phosphate of the 5# terminal nucleotide (East-
berg et al., 2004). By analogy, we used two reference
points, the 5#-hydroxyl and a backbone phosphate, to
position the double helical portion of the substrate at
the mPNK active site. The 5#-hydroxyl was aligned with
the water molecule W1 (Figure 3A), which corresponds
to the 5#-hydroxyl of the oligonucleotide substrate
bound to T4 PNK (Eastberg et al., 2004) and forms a
hydrogen bond with the catalytic Asp 396. Secondly, a
feature in the electron density interpreted as a bound
sulfate ion (S2 in Figure 3A) was assumed to define the
location of a backbone phosphate (3# to the second
nucleotide). In the complex with an oligonucleotide
substrate, T4 PNK binds the 3#-phosphate of the ter-
minal nucleotide between Thr 86, which corresponds to
Thr 423 in mPNK, and Arg 38, which is not conserved
in mPNK. The guanidinium of Arg 395 (mPNK) is in ap-
proximately the same location as the guanidinium of
Arg 38 (T4), but points away from Thr 423 and cannot
interact with a phosphate bound in that space. In
mPNK, sulfate ion S2, which is observed in both mole-
cules in the asymmetric unit and is located between the
guanidinium of Arg 395 and the edge of the Trp 401
indole ring (Figure 3A), is a likely candidate for a back-
bone phosphate mimic. Interestingly, using W1 and S2
as points of reference, a modeled minimal preferred
DNA substrate can be docked to mPNK such that both
the double-stranded portion and the single-stranded
overhang contact conserved positively charged resi-
dues on the surface of the protein. In the double-
stranded portion, the DNA backbone (positions 6-8 on
the complementary strand) approaches the protein
near a positively charged surface around Arg 403, while
4 nucleotides of the single-stranded 3# overhang can
be positioned near a second positive patch on the
other side of the cleft composed of Arg 482 and Lys
483. The model explains the minimal substrate require-
ments of 8 base pairs of double-stranded DNA (Karimi-
Busheri and Weinfeld, 1997) and at least 3 nucleotides
of 3# single-stranded overhang (Figure 2D).
phosphatase (Aravind et al., 1998; Kamenski et al., 2004)
(Figure 5A). All these enzymes hydrolyze a phosphate
in a Mg2+-dependent reaction via a phosphoaspartate
intermediate. Despite a generally low sequence sim-
ilarity, these proteins are characterized by the con-
served motif Dx(D/T)x(T/V), where the first Asp (170 in
mPNK) forms the covalent phosphoaspartate interme-
diate. Additional conserved motifs include a serine or
threonine (216), a lysine (259), and a pair of aspartates
(288 and 309) (Figure 5B). These conserved residues
are involved in catalysis or in binding the Mg2+and the
phosphate moiety of the substrate and phosphoaspar-
tate intermediate, and are all located in the HAD α/β
domain. Despite structural variations among the α/β
domains, the active site configuration is exquisitely
preserved (Figures 5A and 5B). The major structural dif-
ferences arise due to the nonconserved loops or cap-
ping structures that protrude above the active site (Fig-
ure 5A). While the α/β domain interacts with the
phosphate portion, the capping structure recognizes
the rest of the substrate, conferring substrate speci-
The mechanism of phosphate transfer catalyzed by
several HAD-family phosphotransferases has been elu-
cidated in a series of structural studies (Cho et al., 2001;
Lahiri et al., 2002; Wang et al., 2002b) and the conserva-
tion of key catalytic residues in mPNK suggests that
essential features of the mechanism are likely to be
maintained (Figure 5B). As the phosphorylated sub-
strate binds in the active site, its phosphate group is
stabilized by contacts with Lys 259 and Thr 216 as well
by the bound Mg2+. Although Mg2+is not observed in
the mPNK structure, its position is inferred from the
structure of BeF3-derivatized PSP (Cho et al., 2001).
The phosphate is transferred to Oδ1 of Asp 170 in an
in-line nucleophilic attack that forms the phospho-
aspartate intermediate. Asp 172 may provide general
acid assistance by protonating the alcohol leaving
group. The phosphoaspartate intermediate has been
observed in β-PGM and in PSP (simulated by a BeF3
adduct) (Cho et al., 2001; Lahiri et al., 2002). In the anal-
ogous intermediate envisioned for mPNK, the phos-
phate interacts with the side chains Lys 269 and Thr
216, with backbone NH of Leu 171 and Asp 172 and
with the Mg2+. The metal ion is coordinated by one of
the phosphate oxygens, the main chain of Asp 172, the
carboxylate groups of Asp 170 and Asp 288, and two
water molecules. The carboxylate of Asp 309 helps to
stabilize the active site configuration by forming hy-
drogen bonds with the backbone NH of Leu 171 and
Asp 172 and a salt bridge with Lys 259. In T4 PNK and
in β-PGM the carboxylate group at this position is con-
tributed by an Asp from a different part of the peptide
chain (Asp 277 in T4 PNK). The precise location of this
carboxylate is crucial, as mutation of the corresponding
Asp (218) in the S. cerevisiae homolog, Tpp1, to Glu
severely impairs phosphatase activity (Deshpande and
Wilson, 2004). Finally, the free enzyme is regenerated in
a dephosphorylation reaction, with Asp 172 activating
a water molecule for SN2 attack on the phosphorus
atom. This aspartate is conserved in all the phospha-
tases where the leaving group is an alcohol.
The Phosphatase Domain
To determine the minimal phosphatase substrate re-
quirements, we assessed the phosphatase activity of
mPNK on eight 3#-phosphorylated oligonucleotides (Fig-
ure 4). The dephosphorylation reaction proceeds equally
well on either nicked or gapped double-stranded sub-
strates, as well as a variety of single-stranded substrates
of various lengths. Efficient dephosphorylation is main-
tained in single-stranded DNAs as small as 3 nucleo-
tides. Further deletion to a dinucleotide substrate showed
a significant 3-fold decrease in activity, while a 5#-3#-
diphospho-mononucleotide resulted in a further de-
crease. Thus, in contrast to the kinase domain, the
phosphatase domain can accept a wide range of sub-
strates, but only requires a very short segment of oligo-
nucleotide chain 5# to the 3#-phosphate.
Structure of the Phosphatase Domain
The phosphatase domain encompasses residues 146
to 336, and has a fold typical of the haloacid dehalo-
genase (HAD) superfamily, which also includes β-phos-
phoglucomutase (β-PGM), phosphoserine phosphatase
(PSP), P-type ATPase and RNA polymerase II CTD
Structure of Mammalian PNK
Figure 4. Phosphatase Activity Assay
The amount of 3# dephosphorylation of vari-
ous deoxyribonucleotide substrates (shown)
is measured at different concentrations of
mPNK. The asterisk denotes the position of
radiolabeled 5#-phosphate on the substrates.
The single-stranded substrates, all bearing a
3#-phosphate, include a (pCnp) series where
(n = 1, 2, 3, 4), a 10-mer and a 21-mer (see
Supplemental Data for sequences).
A Model for the Phosphatase/Substrate Complex
Substrate specificity varies widely in the HAD phospho-
transferase family. To account for our finding that the
minimal substrate for the mPNK phosphatase is a sin-
gle-stranded DNA of at least 3 nucleotides in length,
we propose a model of the phosphatase:substrate
complex (Figure 5C). In the model, the 3#-phosphate
was positioned by overlaying C3# of the terminal ribose
with C1 of glucose-6-phosphate bound to β-PGM (La-
hiri et al., 2003), and a nearby sulfate ion located next
to Arg 258 in our structure was used as a guide for
placing the preceding backbone phosphate. The model
was minimized in REFMAC (Winn et al., 2003) to regu-
larize the geometry of the oligonucleotide and to allevi-
ate steric clashes. Arg 223 and Arg 258 form a positive
patch at the rim of the active site cleft, and in our sub-
strate model three backbone phosphates can bind at
this location. This is consistent with our phosphatase
activity assay, which shows that a trinucleotide is a bet-
ter substrate for mPNK phosphatase than the dinucleo-
tide, which in turn is much better than a mononucleo-
tide (Figure 4, rows f–h). The lack of improvement for
substrates longer than three nucleotides agrees with
the prediction that only three backbone phosphates
make significant contact with this surface patch (Figure
4, rows d and e). The requirement for at least one phos-
phate at the patch is further supported by the observa-
tion of Habraken & Verly that rat PNK does not work on
3#-monodeoxynucleotides, but removes the 3#-phos-
phate in dinucleotides (Habraken and Verly, 1988). The
phosphatase active site cleft appears to be too narrow
to admit a double-stranded DNA substrate, but the
edges of the cleft are formed by potentially flexible
loops that may adapt to different substrates.
located on adjacent β strands and the phosphopeptide
is bound by loops on the opposite side of the domain.
In the structure of full-length mPNK, the peptide bind-
ing loops of the FHA domain were poorly ordered, and
it was impossible to predict how this protein would in-
teract with a phosphopeptide. We therefore expressed
an isolated FHA domain, which we co-crystallized
with an XRCC4-derived phosphopeptide (Ac-YDES(pT)
DEESEKK-CONH2(Koch et al., 2004), Table 1). A num-
ber of structures of FHA domains bound to phos-
phothreonyl peptide ligands are available, and they all
display a conserved mode of binding to the phos-
phothreonine group (Durocher and Jackson, 2002). The
FHA domain of mPNK utilizes this standard mode of
phosphothreonine recognition, involving the invariant
Arg 35 and Ser 47, which is in some cases replaced by
Thr. Additional stabilization for the phosphate is pro-
vided by Arg 48, which corresponds to Lys or Asn in
many other FHA domains (Figures 1B and 6A).
Most FHA domains display additional specificity for
side chains 3 residues C-terminal to the phosphothreo-
nine (pT+3, (Durocher and Jackson, 2002)). In contrast,
peptide array binding studies showed that the FHA do-
main of mammalian PNK recognizes peptide positions
N-terminal to the phosphothreonine (Koch et al., 2004),
and a comparison of the natural PNK FHA binding
targets in both XRCC1 and XRCC4 suggests a prefer-
ence for acidic residues in the vicinity of the phos-
phothreonine (Loizou et al., 2004; Koch et al., 2004). In
our structure of the FHA:XRCC4 peptide complex, the
peptide is cradled between two surface loops, β2-β3
and β4-β5. The structure suggests that three positively
charged residues from the β2-β3 loop, Arg 44, Lys 45
and Arg 48, may be responsible for the recognition of
acidic residues not only N-terminal, but also C-terminal
to the phosphothreonine. Arg 48, in addition to its role
in phosphate recognition, is also close enough to pro-
vide electrostatic recognition of Asp at pT-3. Arg 44 dis-
plays different conformations in each of the three inde-
Phosphopeptide Recognition by the FHA Domain
The FHA domain possesses the typical FHA β sandwich
fold consisting of a 3-stranded and a 4-stranded anti-
parallel β sheet. The N- and C termini of the domain are
Figure 5. Structure of the Phosphatase Domain
(A) Comparison of mPNK and T4 PNK phosphatase domains, β-PGM and PSP in a common orientation, showing the variation in the active
site capping structures (blue). The common α/β domain elements are colored yellow and red, and the catalytic aspartate is shown in ball-
(B) Structure of the mPNK phosphatase active site (cyan) superimposed on the active sites of phosphorylated β-PGM (green, PDB 1LVH),
BeF3-derivatized PSP (pink, Mg2+in green, PDB 1J97), and T4 PNK (blue, PDB 1LTQ). The hydrogen bonding pattern for PSP is shown.
Residue numbering shown is for mPNK and in italics for T4 PNK.
(C) Surface representation of mPNK phosphatase with bound modeled tetranucleotide substrate. The green sphere represents the modeled
pendent complexes in the crystallographic asymmetric
unit, occupying a variety of positions between Glu at
pT-2 and Asp at the pT+1 position of the peptide (Figure
6B). Lys 45 could potentially provide additional electro-
static selectivity for both Asp at pT+1 and Glu and
pT+2. Interestingly, there is no interpretable electron
Structure of Mammalian PNK
Figure 6. The Basis of Phosphopeptide Recognition by the mPNK FHA Domain
(A) XRCC4 phosphopeptide (carbon atoms in green) bound to the mPNK FHA domain. The electrostatic surface potential indicates that the
peptide binding surface is predominantly positively charged.
(B) Three superimposed NCS-related FHA domains (thin lines labeled A, B, and C) are shown with the phosphopeptide from complex (A)
(C) Binding of the GST-tagged mPNK FHA domain (wild-type, R44A, K45A, and R48N) to the fluorescein-labeled peptide GGYDES-pT-DEESKK,
measured by fluorescence polarization. Dissociation constants for wild-type and the R48N mutant are shown. NB indicates no binding. Error
bars are derived from at least three independent measurements for each data point.
density for regions of the peptide C-terminal to pT+2.
This strongly suggests that the PNK FHA does not con-
tact these residues, and is consistent with previous
work that has shown that this domain, unlike all other
well-studied members of this family, does not exhibit
selectivity for the pT+3 residue.
To test the importance of Arg 44, Lys 45 and Arg 48
for phosphopeptide recognition, we mutated each of
these residues and used fluorescence polarization to
measure their affinities for the XRCC4-derived phos-
phopeptide, relative to the wild-type domain. Arg 44
and Lys 45 were mutated to alanine, while Arg 48,
which makes critical contacts to the phosphothreonine
as well as interactions with pT-3, was mutated to Asn,
which is found in several FHA domains and has been
shown to mediate phosphothreonine recognition. As
shown in Figure 6C, the R44A and R48N mutations ab-
lated the ability of the the FHA domain to recognize the
phosphopeptide, confirming an essential role for both
of these residues in peptide binding. In marked con-
trast, mutation of Lys 45 did not affect the binding affin-
ity, indicating that this residue only plays a minor role,
if any, in phosphopeptide recognition.
Sequence comparisons of the PNK and aprataxin
FHA domains show that all residues involved in phos-
phopeptide recognition are well-conserved (Figure 1B).
It is thus highly likely that the FHA domains of PNK and
APTX recognize the same phosphopeptide sequence,
in agreement with the observation that these proteins
are found in mutually exclusive complexes in vivo (Luo
et al., 2004). Curiously, unlike PNK, the pT+3 position
appears to be important in the APTX:XRCC1 interac-
tion, since mutation of Glu to Ala at this site in XRCC1
abolished the binding to APTX (Luo et al., 2004). APTX
has a single positively charged residue, Lys 75 (Pro 81
in PNK), which we predict to be close to the pT+3 resi-
due and could potentially provide sequence selectivity
at this position.
In spite of the striking differences in the mechanism
of sequence selectivity between the PNK FHA and
other members of the family, the backbone of the pep-
tide in the PNK structure is bound in a similar manner
to that found in other FHA-peptide complexes. The
PNK FHA contacts the peptide backbone through hy-
drogen bonding interactions between the conserved
Asn 70 and Arg 35, and the peptide backbone at pT+1
and pT-2, respectively. In addition, the pT-4 Tyr is con-
tacted by a stacking interaction with Pro 37.
stranded 5# ends will be encountered, which would be
poor substrates for mPNK. We suggest that such ends
might first be processed by the NHEJ-associated
nuclease Artemis (Ma et al., 2002; Pannicke et al.,
2004). Artemis has single-strand-specific 5#-3# exo-
nuclease and endonuclease activities that could re-
move nucleotides containing 5#-hydroxyl termini from
5# overhanging ends, leaving a blunt 5#-phosphate ter-
minus that would be suitable for ligation. Artemis also
cleaves long 3# overhangs, resulting in short 4–5 nucle-
otide 3# overhangs, the optimal substrate for mPNK.
Finally, we can envision an application for PNK to the
treatment of cancer. Radiation therapy and certain che-
motherapies work by damaging DNA in tumors. Since
various DNA repair processes may attenuate the effect
of therapy, it may be beneficial to down-regulate repair.
PNK, being essential in both SSBR and DSBR, may be
a suitable target for inactivation. Inhibitors may be de-
signed against either catalytic activity, or against the
Cloning, Expression, and Purification of mPNK Constructs
Two cDNA clones of mouse PNK (598211 and 2865792) were ob-
tained from the IMAGE consortium and cloned into the EcoRI and
XhoI sites of pET19b (Invitrogen). DNA sequencing of the two ex-
pression clones showed that clone 598211, derived from mouse
spleen, was alternately spliced, lacking exon 4, which encodes part
of the phosphatase active site. Clone 2865792, from mouse mam-
mary tumor, coded for the full length PNK, and was used in all sub-
sequent experiments. Mouse PNK was overexpressed in E. coli
BL21 Gold cells induced by 0.1mM IPTG and grown for 18 hr at
22°C. The protein was purified as described previously for human
PNK (Mani et al., 2003). After the last chromatographic step, the
buffer was exchanged to 150 mM KCl, 10 mM Tris (pH 8.5), 1mM
DTT. Selenomethionyl mPNK was produced as described in
(Doublie, 1997) and purified in the same manner as the native
The mPNK FHA domain (residues 1-110) was cloned using the
Gateway system (Invitrogen) into the pDEST15 expression vector
with an N-terminal GST tag followed by a Prescission protease site.
The GST-tagged protein was produced in E.coli BL21 Gold99 cells
(induced with 0.1mM IPTG and grown overnight at 22°C), and puri-
fied on glutathione (GSH) resin (Amersham). The GST tag was
cleaved off in solution (20 mM reduced GSH, 50 mM Tris (pH 7.5),
400 mM NaCl, 0.1% β-mercaptoethanol) by Prescission protease
(Amersham) at 4°C. The FHA domain was separated from GST by
size exclusion chromatography on a Superdex 75 column (Amer-
sham) in 50 mM Tris (pH 7.5), 400 mM NaCl, 1mM DTT.
The R44A, K45A and R48N point mutants of the FHA domain
were constructed by PCR mutagenesis and Gateway cloning. The
mutants were expressed and purified by the same method as the
wild-type mPNK FHA, but without removal of the GST tag.
Integration of mPNK within DNA Repair Pathways
A picture is now emerging of mammalian PNK as an en-
zyme that has evolved to play specific roles in XRCC1-
and XRCC4-mediated DNA repair. The specificity of
PNK for these processes is 2-fold. First, the enzyme is
initially recruited to the sites of repair through phos-
phorylation-dependent interactions with XRCC1/4 via
its FHA domain. The FHA is flexibly linked to the cata-
lytic domain, and does not appear to interact with, or
modulate its activity (not shown). Flexible attachment
of the catalytic portion enables either active site to ap-
proach the DNA end that requires processing.
The second level of specificity arises from the DNA
substrate preferences of the enzyme. Intriguingly, the
two catalytic domains have complementary and non-
overlapping minimal substrate requirements. This may
have evolved to ensure that the two catalytic activities,
while held together in a relatively rigid structure, never-
theless act independently, enabling the enzyme to pro-
cess a variety of substrates (eg: nicks, gaps of various
sizes, and double-strand breaks). While the two active
sites are on the same side of the protein, they are far
apart (w40 Å), and it seems unlikely that they will be
able to bind to the same substrate DNA, especially for
small nicks or gaps. In contrast, T4 PNK probably
evolved to bind simultaneously the 5#-hydroxyl and the
3#-phosphate of its nicked tRNA substrate (Galburt et
The kinase strongly prefers 5# recessed ends over
blunt or single-stranded ends, consistent with its role
in SSBR/BER, in which gapped and nicked substrates
are expected. For DSBR, however, blunt and single-
Kinase Activity Assays
Oligonucleotides (see Supplemental Data for sequences) were
purified on a Source Q column (Amersham) in 10 mM NaOH with a
gradient of 1 M NaCl, desalted using Sep-Pak cartridges (Waters),
and annealed by slow cooling from 95°C in 150 mM NaCl and 10
mM HEPES (pH 7.5). Each 10 µl assay reaction contained the re-
quired amounts of DNA and mPNK, 400 pmol ATP, 5 µCi γ32P-ATP,
80 mM succinate (pH 5.5), 10 mM MgCl2, and 1mM DTT. Reactions
were assembled on ice, incubated 4 min at 37°C, stopped with 10
µl urea loading buffer, heated 5 min at 95°C and loaded on a 12%
denaturing gel (containing 7M urea), run in TBE.
Steady-state fluorescence spectra were measured at 5°C on a Per-
kin-Elmer LS-55 spectrofluorometer (with 5-nm spectral resolution
Structure of Mammalian PNK
for excitation and emission). DNA ligands were added from stock
solutions to mPNK (0.15 µM at 5°C in 50 mM Tris, pH 7.5, 100 mM
NaCl, 5 mM MgCl2, and 1 mM DTT). The protein was excited at
295 nm, keeping the total absorption < 0.05, and the fluorescence
intensity was monitored at 335 nm as a function of oligonucleotide
concentration, correcting fluorescence intensities for dilution. Fluo-
rescence data were analyzed as described previously (Mani et al.,
Received: December 3, 2004
Revised: January 7, 2005
Accepted: February 2, 2005
Published: March 3, 2005
Amitsur, M., Levitz, R., and Kaufmann, G. (1987). Bacteriophage T4
anticodon nuclease, polynucleotide kinase and RNA ligase re-
process the host lysine tRNA. EMBO J. 6, 2499–2503.
Aravind, L., Galperin, M.Y., and Koonin, E.V. (1998). The catalytic
domain of the P-type ATPase has the haloacid dehalogenase fold.
Trends Biochem. Sci. 23, 127–129.
Bosdal, T., and Lillehaug, J.R. (1985). Purification and kinetic prop-
erties of polynucleotide kinase from rat testes. Biochim. Biophys.
Acta 840, 280–286.
Caldecott, K.W. (2003). DNA single-strand break repair and spino-
cerebellar ataxia. Cell 112, 7–10.
Chappell, C., Hanakahi, L.A., Karimi-Busheri, F., Weinfeld, M., and
West, S.C. (2002). Involvement of human polynucleotide kinase in
double-strand break repair by non-homologous end joining. EMBO
J. 21, 2827–2832.
Cho, H., Wang, W., Kim, R., Yokota, H., Damo, S., Kim, S.H., Wem-
mer, D., Kustu, S., and Yan, D. (2001). BeF(3)(-) acts as a phosphate
analog in proteins phosphorylated on aspartate: structure of a
BeF(3)(-) complex with phosphoserine phosphatase. Proc. Natl.
Acad. Sci. USA 98, 8525–8530.
Clements, P.M., Breslin, C., Deeks, E.D., Byrd, P.J., Ju, L., Biega-
nowski, P., Brenner, C., Moreira, M.C., Taylor, A.M., and Caldecott,
K.W. (2004). The ataxia-oculomotor apraxia 1 gene product has a
role distinct from ATM and interacts with the DNA strand break
repair proteins XRCC1 and XRCC4. DNA Repair (Amst.) 3, 1493–
Deshpande, R.A., and Wilson, T.E. (2004). Identification of DNA
3#-phosphatase active site residues and their differential role in
DNA binding, Mg2+ coordination, and catalysis. Biochemistry 43,
Doublie, S. (1997). Preparation of selenomethionyl proteins for
phase determination. Methods Enzymol. 276, 523–530.
Durocher, D., and Jackson, S.P. (2002). The FHA domain. FEBS
Lett. 513, 58–66.
Eastberg, J.H., Pelletier, J., and Stoddard, B.L. (2004). Recognition
of DNA substrates by T4 bacteriophage polynucleotide kinase.
Nucleic Acids Res. 32, 653–660.
Galburt, E.A., Pelletier, J., Wilson, G., and Stoddard, B.L. (2002).
Structure of a tRNA repair enzyme and molecular biology work-
horse: T4 polynucleotide kinase. Structure (Camb.) 10, 1249–1260.
Gueven, N., Becherel, O.J., Kijas, A.W., Chen, P., Howe, O., Ru-
dolph, J.H., Gatti, R., Date, H., Onodera, O., Taucher-Scholz, G.,
and Lavin, M.F. (2004). Aprataxin, a novel protein that protects
against genotoxic stress. Hum. Mol. Genet. 13, 1081–1093.
Habraken, Y., and Verly, W.G. (1983). The DNA 3#-phosphatase and
5#-hydroxyl kinase of rat liver chromatin. FEBS Lett. 160, 46–50.
Habraken, Y., and Verly, W.G. (1988). Further purification and char-
acterization of the DNA 3#-phosphatase from rat-liver chromatin
which is also a polynucleotide 5#-hydroxyl kinase. Eur. J. Biochem.
Hazra, T.K., Kow, Y.W., Hatahet, Z., Imhoff, B., Boldogh, I., Mokka-
pati, S.K., Mitra, S., and Izumi, T. (2002). Identification and charac-
terization of a novel human DNA glycosylase for repair of cytosine-
derived lesions. J. Biol. Chem. 277, 30417–30420.
Izumi, T., Wiederhold, L.R., Roy, G., Roy, R., Jaiswal, A., Bhakat,
K.K., Mitra, S., and Hazra, T.K. (2003). Mammalian DNA base exci-
sion repair proteins: their interactions and role in repair of oxidative
DNA damage. Toxicology 193, 43–65.
Kamenski, T., Heilmeier, S., Meinhart, A., and Cramer, P. (2004).
Structure and mechanism of RNA polymerase II CTD phospha-
tases. Mol. Cell 15, 399–407.
Karimi-Busheri, F., Lee, J., Tomkinson, A.E., and Weinfeld, M.
Phosphatase Activity Assay
The 3#-dephosphorylation of DNA substrates by mPNK was per-
formed as described previously (Karimi-Busheri et al., 1998). The
5# termini of the oligonucleotides (16 pmol in 10 ?l) were radiola-
beled by incubation with phosphatase-free T4 PNK (Roche) and 3.3
pmol γ32P-ATP in 50 mM Tris-HCl, 10mM MgCl2, 5 mM DTT, pH 7.5
for 30 s at 37°C. Reactions were stopped by incubation for 10 min
at 95°C. Double-stranded substrates were annealed by slow cool-
ing in 50 mM Tris-HCl, 10 mM MgCl2, 5 mM DTT, pH 7.5. The 5#-
labeled 3#-phosphorylated constructs (2.5 pmol/10 ?l) were used
to assess the 3#-phosphatase activity of mPNK in a reaction mix-
ture containing different quantities of mPNK incubated at 37°C for
20 min in the same buffer used for 5#-labeling. The labeled pro-
ducts were resolved by 20% denaturing PAGE for the mono and
dinucleotide substrates and by 12% gel for the longer oligonucleo-
tide and duplex substrates.
The binding of GST-tagged mPNK FHA domain (wild-type and the
R44A, K45A and R48N point mutants) to fluorescein-labeled pep-
tide GGYDES-pT-DEESKK was measured by fluorescence polariza-
tion as described previously (Koch et al., 2004).
Mouse PNK and the mPNK FHA domain were crystallized by vapor
diffusion in sitting or hanging drops. Native or selenomethionyl
mPNK (2.5 mg/mL) in 150 mM KCl, 10 mM Tris (pH 8.5), 0.1 mM
EDTA and 1 mM DTT was mixed with an equal volume of the reser-
voir solution (0.1 M Tris (pH 8.3 - 8.7), 18 - 20% PEG 5000 MME,
0.1 M Li2SO4and 5 mM DTT). Thin plates grew at room temperature
within 1-4 days.
The FHA domain was concentrated in the presence of a 1.5 molar
excess of XRCC4 peptide (Ac-YDES(pT)DEESEKK-CONH2, Alberta
Peptide Institute) to 60 mg/mL and a final peptide to FHA ratio of
1.17:1. The FHA/peptide solution was mixed with an equal volume
of reservoir solution (0.1 M Na citrate (pH 5.5), 25% PEG 4000, 0.2
M Li2SO4and 5 mM DTT). A cluster of plates grew at room temper-
ature within 2 weeks.
See Supplemental materials for details of the diffraction data col-
lection and processing as well as structure determination and re-
finement. Briefly, the structure of full-length mouse PNK was solved
by selenomethionyl MAD. The structure of the FHA:phosphopep-
tide complex was solved by molecular replacement with the FHA
domain from the full-length structure as the search model.
Supplemental Data include Supplemental Experimental Procedures
and one additional figure and can be found with this article online
We would like to thank members of the Glover laboratory, Mesfin
Fanta, and staff at the Alberta Synchrotron Institute for helpful dis-
cussions and data collection support. This research was funded by
grants from the Canadian Institutes for Health Research (J.N.M.G.,
D.D., and M.W.) and by the National Cancer Institute of Canada
(J.N.M.G.) J.N.M.G acknowledges with thanks the award of a Can-
ada Research Chair. N.K.B gratefully acknowledges the Terry Fox
Foundation for a postdoctoral fellowship.
Molecular Cell Download full-text
(1998). Repair of DNA strand gaps and nicks containing 3#-phos-
phate and 5#-hydroxyl termini by purified mammalian enzymes. Nu-
cleic Acids Res. 26, 4395–4400.
Karimi-Busheri, F., and Weinfeld, M. (1997). Purification and sub-
strate specificity of polydeoxyribonucleotide kinases isolated from
calf thymus and rat liver. J. Cell. Biochem. 64, 258–272.
Kleppe, K., and Lillehaug, J.R. (1979). Polynucleotide kinase. Adv.
Enzymol. Relat. Areas Mol. Biol. 48, 245–275.
Koch, C.A., Agyei, R., Galicia, S., Metalnikov, P., O’Donnell, P., Star-
ostine, A., Weinfeld, M., and Durocher, D. (2004). Xrcc4 physically
links DNA end processing by polynucleotide kinase to DNA ligation
by DNA ligase IV. EMBO J. 23, 3874–3885.
Lahiri, S.D., Zhang, G., Dunaway-Mariano, D., and Allen, K.N.
(2002). Caught in the act: the structure of phosphorylated beta-
phosphoglucomutase from Lactococcus lactis. Biochemistry 41,
Lahiri, S.D., Zhang, G., Dunaway-Mariano, D., and Allen, K.N.
(2003). The pentacovalent phosphorus intermediate of a phos-
phoryl transfer reaction. Science 299, 2067–2071.
Leipe, D.D., Koonin, E.V., and Aravind, L. (2003). Evolution and clas-
sification of P-loop kinases and related proteins. J. Mol. Biol. 333,
Lillehaug, J.R., Kleppe, R.K., and Kleppe, K. (1976). Phosphoryla-
tion of double-stranded DNAs by T4 polynucleotide kinase. Bio-
chemistry 15, 1858–1865.
Loizou, J.I., El-Khamisy, S.F., Zlatanou, A., Moore, D.J., Chan, D.W.,
Qin, J., Sarno, S., Meggio, F., Pinna, L.A., and Caldecott, K.W.
(2004). The protein kinase CK2 facilitates repair of chromosomal
DNA single-strand breaks. Cell 117, 17–28.
Luo, H., Chan, D.W., Yang, T., Rodriguez, M., Chen, B.P., Leng, M.,
Mu, J.J., Chen, D., Songyang, Z., Wang, Y., and Qin, J. (2004). A
new XRCC1-containing complex and its role in cellular survival of
methyl methanesulfonate treatment. Mol. Cell. Biol. 24, 8356–8365.
Ma, Y., Pannicke, U., Schwarz, K., and Lieber, M.R. (2002). Hairpin
opening and overhang processing by an Artemis/DNA-dependent
protein kinase complex in nonhomologous end joining and V(D)J
recombination. Cell 108, 781–794.
Mani, R.S., Karimi-Busheri, F., Cass, C.E., and Weinfeld, M. (2001).
Physical properties of human polynucleotide kinase: hydrodynamic
and spectroscopic studies. Biochemistry 40, 12967–12973.
Mani, R.S., Karimi-Busheri, F., Fanta, M., Cass, C.E., and Weinfeld,
M. (2003). Spectroscopic studies of DNA and ATP binding to human
polynucleotide kinase: evidence for a ternary complex. Biochemis-
try 42, 12077–12084.
Meijer, M., Karimi-Busheri, F., Huang, T.Y., Weinfeld, M., and Young,
D. (2002). Pnk1, a DNA kinase/phosphatase required for normal re-
sponse to DNA damage by gamma-radiation or camptothecin in
Schizosaccharomyces pombe. J. Biol. Chem. 277, 4050–4055.
Midgley, C.A., and Murray, N.E. (1985). T4 polynucleotide kinase;
cloning of the gene (pseT) and amplification of its product. EMBO
J. 4, 2695–2703.
Mitra, S., Izumi, T., Boldogh, I., Bhakat, K.K., Hill, J.W., and Hazra,
T.K. (2002). Choreography of oxidative damage repair in mamma-
lian genomes. Free Radic. Biol. Med. 33, 15–28.
Moreira, M.C., Barbot, C., Tachi, N., Kozuka, N., Uchida, E., Gibson,
T., Mendonca, P., Costa, M., Barros, J., Yanagisawa, T., et al. (2001).
The gene mutated in ataxia-ocular apraxia 1 encodes the new HIT/
Zn-finger protein aprataxin. Nat. Genet. 29, 189–193.
Pannicke, U., Ma, Y., Hopfner, K.P., Niewolik, D., Lieber, M.R., and
Schwarz, K. (2004). Functional and biochemical dissection of the
structure-specific nuclease ARTEMIS. EMBO J. 23, 1987–1997.
Pheiffer, B.H., and Zimmerman, S.B. (1982). 3#-Phosphatase activ-
ity of the DNA kinase from rat liver. Biochem. Biophys. Res. Com-
mun. 109, 1297–1302.
Rasouli-Nia, A., Karimi-Busheri, F., and Weinfeld, M. (2004). Stable
down-regulation of human polynucleotide kinase enhances spon-
taneous mutation frequency and sensitizes cells to genotoxic
agents. Proc. Natl. Acad. Sci. USA 101, 6905–6910.
Rosenquist, T.A., Zaika, E., Fernandes, A.S., Zharkov, D.O., Miller,
H., and Grollman, A.P. (2003). The novel DNA glycosylase, NEIL1,
protects mammalian cells from radiation-mediated cell death. DNA
Repair (Amst.) 2, 581–591.
Teraoka, H., Mizuta, K., Sato, F., Shimoyachi, M., and Tsukada, K.
(1975). Polynucleotide kinase from rat-liver nuclei. Purification and
properties. Eur. J. Biochem. 58, 297–302.
Wang, J.C. (1996). DNA topoisomerases. Annu. Rev. Biochem. 65,
Wang, L.K., Lima, C.D., and Shuman, S. (2002a). Structure and
mechanism of T4 polynucleotide kinase: an RNA repair enzyme.
EMBO J. 21, 3873–3880.
Wang, W., Cho, H.S., Kim, R., Jancarik, J., Yokota, H., Nguyen, H.H.,
Grigoriev, I.V., Wemmer, D.E., and Kim, S.H. (2002b). Structural
characterization of the reaction pathway in phosphoserine phos-
phatase: crystallographic “snapshots” of intermediate states. J.
Mol. Biol. 319, 421–431.
Whitehouse, C.J., Taylor, R.M., Thistlethwaite, A., Zhang, H., Kar-
imi-Busheri, F., Lasko, D.D., Weinfeld, M., and Caldecott, K.W.
(2001). XRCC1 stimulates human polynucleotide kinase activity at
damaged DNA termini and accelerates DNA single-strand break
repair. Cell 104, 107–117.
Wiederhold, L., Leppard, J.B., Kedar, P., Karimi-Busheri, F., Rasouli-
Nia, A., Weinfeld, M., Tomkinson, A.E., Izumi, T., Prasad, R., Wilson,
S.H., et al. (2004). AP endonuclease-independent DNA base exci-
sion repair in human cells. Mol. Cell 15, 209–220.
Winn, M.D., Murshudov, G.N., and Papiz, M.Z. (2003). Macromolec-
ular TLS refinement in REFMAC at moderate resolutions. Methods
Enzymol. 374, 300–321.
Zhou, T., Lee, J.W., Tatavarthi, H., Lupski, J.R., Valerie, K., and Po-
virk, L.F. (2005). Deficiency in 3#-phosphoglycolate processing in
human cells with a hereditary mutation in tyrosyl-DNA phospho-
diesterase (TDP1). Nucleic Acids Res. 33, 289–297.
Coordinates and structure factors for mPNK and the FHA:phospho-
peptide complex have been deposited in the Protein Data Bank
under ID codes 1YJ5 and 1YJM, respectively.