Full text of DOI:10.1093/nar/28.13.2585

© 2000 Oxford University Press
Nucleic Acids Research, 2000, Vol. 28, No. 13 2585–2596
DNA double-strand break repair in cell-free extracts from
Ku80-deficient cells: implications for Ku serving as an
alignment factor in non-homologous DNA end joining
Elke Feldmann, Viola Schmiemann, Wolfgang Goedecke, Susanne Reichenberger and
Petra Pfeiffer*
Institut für Zellbiologie des Universitätsklinikums Essen, Virchowstraße 173, D-45122 Essen, Germany
Received February 18, 2000; Revised and Accepted May 18, 2000
ABSTRACT
sequence at the break; (ii) single-strand annealing (SSA), a
process that leads to the formation of deletions; or (iii) non-
homologous DNA end joining (NHEJ) that joins two broken
ends directly. While HRR and SSA involve the members of the
Rad52 epistasis group and strictly require extensive sequence
homology, the Rad52-independent NHEJ process can dispense
with sequence homology (1–4). Although increasing evidence
indicates that vertebrate cells are also quite proficient at
repairing DSB by HRR (5) during S and G2, NHEJ is still
considered the major pathway of DSB repair.
Non-homologous DNA end joining (NHEJ) is considered
the major pathway of double-strand break (DSB) repair
in mammalian cells and depends, among other
things, on the DNA end-binding Ku70/80 hetero-
dimer. To investigate the function of Ku in NHEJ we
have compared the ability of cell-free extracts from
wild-type CHO-K1 cells, Ku80-deficient xrs6 cells and
Ku80-cDNA-complemented xrs6 cells (xrs6-Ku80) to
rejoin different types of DSB in vitro. While the two
Ku80-proficient extracts were highly efficient and
accurate in rejoining all types of DNA ends, the xrs6
extract displayed strongly decreased NHEJ efficiency
and accuracy. The lack of accuracy is most evident in
non-homologous terminus configurations containing
3′-overhangs that abut a 5′-overhang or blunt end.
While the sequences of the 3′-overhangs are mostly
preserved by fill-in DNA synthesis in the Ku80-
proficient extracts, they are always completely lost in
the xrs6 extract so that, instead, small deletions
displaying microhomology patches at their break-
points arise. In summary, our results are consistent
with previous results from Ku-deficient yeast strains
and indicate that Ku may serve as an alignment
factor that not only increases NHEJ efficiency but
also accuracy. Furthermore, a secondary NHEJ
activity is present in the absence of Ku which is
error-prone and possibly assisted by base pairing
interactions.
First reports on NHEJ date back to the early eighties (6–8).
In the following years, the mechanisms of NHEJ were studied
in detail in mammalian cells (9–12), extracts from Xenopus
eggs (13,14) and later in yeast (15–17). NHEJ includes (i) the
simple ligation of complementary (cohesive) protruding single
strands (PSS) or blunt (bl.) ends which leads to the precise
restoration of the broken sequence and (ii) the joining of non-
complementary (non-homologous) DNA termini. Studies in
Xenopus egg extracts showed that NHEJ is a highly accurate
process that tends to preserve the sequences of interacting non-
homologous DNA ends by generating two major types of
products (overlap and fill-in). The type of product formed depends
on the structure of ends being joined; overlap junctions typically
arise during the joining of DNA ends containing anti-parallel
PSS (5′/5′ and 3′/3′) while fill-in junctions are formed between
abutting DNA ends (bl./5′, bl./3′ and 5′/3′). In the first case, the
ends form by pairing of single fortuitously complementary
bases, incompletely matched overlaps whose structure determines
the patterns of subsequent repair reactions (Fig. 1; 18,19). In
the second case, the sequences of participating 5′- and/or 3′-PSS
are fully preserved by fill-in DNA synthesis in a process in
which the ends are transiently held together by non-covalent
interactions while the 3′-hydroxyl group of the 5′-PSS or blunt
end is used as a primer to direct repair synthesis of the 3′-PSS
(Fig. 1B; 14). Based on these data, we postulated in 1990 that
an alignment factor may function to maintain the two ends in
alignment to facilitate their biochemical reconfiguration into a
ligatable structure (14). Similar joining events have been
observed in vivo and in vitro in mammalian cells (10,11,20,21)
indicating that the mechanisms found in the Xenopus system
also apply for mammalian systems. The identity of the putative
alignment factor, however, remained unknown.
INTRODUCTION
Double-strand breaks (DSB), probably the most disruptive
type of lesion in DNA, may arise spontaneously in the cell or
after exposure to DNA-damaging agents, such as ionizing
radiation (IR). If left unrepaired, DSB leads to broken chromo-
somes and cell death; if repaired improperly, mutations,
chromosome rearrangements and oncogenic transformation
may arise. Repair of DSB is achieved by at least three different
mechanisms: (i) homologous recombination repair (HRR), a
highly accurate process that usually restores the precise DNA
*To whom correspondence should be addressed. Tel: +49 201 723 3385; Fax: +49 201 723 5904; Email: petra.pfeiffer@uni-essen.de

2586 Nucleic Acids Research, 2000, Vol. 28, No. 13
‘fill-in’ junctions (Fig. 1), is also significantly decreased in the
xrs6 extracts. This becomes most evident in terminus configu-
rations containing 3′-PSS. Our results indicate that Ku80, in
conjunction with Ku70, may indeed function as the previously
postulated alignment factor (14).
MATERIALS AND METHODS
Materials
All three cell lines (CHO-K1, xrs6 and xrs6-Ku80) used in this
study were purchased from the European Collection of Cell
Culture (Wiltshire, UK), cell culture media and all additives
from Biochrom (Berlin, Germany). The V3 (43) and XR-1 (44)
mutant CHO cell lines used in one control experiment were
generous gifts from E. Dikomey (University of Hamburg,
Germany) and S. Critchlow and S. P. Jackson (Cambridge
University, UK), respectively. Enzymes and reagents for
immunoblotting [nitro blue tetrazolium chloride (NBT)/5-
bromo-4-chloro-3-indoyl-phosphate (BCIP)] were obtained
from Roche Diagnostics (Mannheim, Germany), proteinase K
from Sigma (Munich, Germany), N-ethyl-N-nitroso urea
(ENU) from Serva (Heidelberg, Germany), kits for gel
extraction of DNA and plasmid minipreparation from Qiagen
(Hilden, Germany), [α-32P]dCTP (3.000 Ci/mM), random
priming kit and T7-dideoxy sequencing kit from Amersham/
Pharmacia Biotech (Eschborn, Germany), microdialysis filters
from Millipore. Polyclonal anti-mouse-Ku86 and -Ku70
antisera were purchased from Santa Cruz Biotech (Heidelberg,
Germany) and the secondary alkaline phosphatase-coupled anti-
body from Jackson Immuno Research Laboratories (Baltimore,
MD).
Figure 1. Major pathways of accurate NHEJ as observed in vitro in Xenopus
eggs (14,19) and mammalian cells (12,21). (A) DNA ends with anti-parallel
5′- or 3′-PSS form short mismatched overlaps at positions of complementary
base pairs (black boxes) which determine the patterns of DNA fill-in synthesis
(open and black triangles; 18,19). (B) Sequences of 5′- and 3′-PSS in abutting
terminus configurations are preserved by fill-in synthesis (14). While fill-in of
5′-PSS (open triangles) can be primed at the recessed 3′-OH group of the same
end, fill-in of 3′-PSS (black triangles) can be primed only at the 3′-OH of the
abutting terminus which may be a blunt (bl.) end or a 5′-PSS.
Analysis of hamster cell lines hypersensitive to IR and
defective in DSB repair led to the identification of four
complementation groups involved in NHEJ (22,23). The corre-
sponding genes (XRCC4–7) encode the XRCC4 protein, an
essential co-factor of DNA ligase IV (24–26), and the three
components of the DNA-dependent protein kinase (DNA-PK)
represented by the 86 and 70 kDa subunits of the heterodimeric
Ku complex (Ku70/80) and the catalytic subunit of the protein
kinase (DNA-PKCS), respectively (27–29). Due to its property
of binding to a variety of end structures, including blunt ends
and 5′- and 3′-PSS (for review see 30), the highly abundant
Ku70/80 heterodimer is a particularly attractive candidate for
the putative alignment factor (23,31). This assumption is
supported by the analysis of IR-sensitive hamster cells bearing
mutations in the XRCC5 gene (encodes Ku80) such as the xrs6
cell line (27,32) which is devoid of Ku end-binding activity
and does not express detectable levels of the Ku70 subunit,
whose stability requires the presence of Ku80 (33,34). These
mutants are defective in general DSB repair and the joining of
broken DNA ends created during V(D)J recombination
(27,35,36). In accordance with a function of Ku in alignment,
Saccharomyces cerevisiae rad52 mutant strains defective in
the yeast Ku70/80 homolog not only display strongly
decreased efficiency of NHEJ but also decreased accuracy of
NHEJ as expressed by elevated levels of deletions (37,38).
This led to the suggestion that Ku might be involved in
protecting DNA ends from degradation thereby enhancing the
accuracy of NHEJ. However, in hamster cells, this issue is still
controversial (39–41).
To investigate whether the Ku heterodimer is able to exert
alignment function and enhance the accuracy of NHEJ, we
have examined the ability of cell-free extracts prepared from
the Ku80-deficient xrs6 cell line (27,32) to join complementary
and non-complementary DNA ends using a plasmid rejoining
assay employed previously to study NHEJ in Xenopus egg
extracts (13,14). As controls, cell-free extracts from the
parental wild-type chinese hamster ovary (CHO)-K1 cell line
and a Ku80-cDNA-complemented xrs6 cell line (xrs6-Ku80;
42) were used. We show that NHEJ activity is strongly
decreased in the xrs6 extracts. Furthermore, the accuracy of
NHEJ, as expressed by the formation of precise ‘overlap’ and
Cell culture
The xrs6 cell line, originally derived from the CHO-K1 line on
the basis of sensitivity to IR (32), is deficient in XRCC5 and
fails to express functional Ku80 protein because of a splice site
mutation that results in a 13 nt insertion and a frame shift (42).
Control cell lines included the parental wild-type CHO-K1 cell
line and the xrs6-Ku80 cell line, which is derived from xrs6 by
stable transfection of the hamster wild-type XRCC5 cDNA
(42). All three cell lines were grown at 37°C in a humidified
5% CO2 atmosphere in Ham’s F-12 medium enriched with
10% fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin
and 100 µg/ml streptomycin.
Cell-free extracts
For all experiments described in this study, whole cell extracts
were used which were prepared exactly as described
previously (21). Approximately 5 × 108 cells of each cell line
were used in each preparation to yield 0.5–1 ml of extract with
a protein concentration ranging between 6 and 10 mg/ml. For
most reproducible conditions, extracts from all three cell lines
were prepared on the same day using the same freshly made
solutions. Stored in 50 µl aliquots in liquid nitrogen, extracts
remained active for 6–12 months. Prior to use in NHEJ reactions,
the extract volume needed was dialyzed against freshly
prepared M-buffer [50 mM 3-(N-morpholino)-2-hydroxy-
propane sulfonic acid (MOPSO)-NaOH pH 7.5, 40 mM KCl,
10 mM MgCl2, 5 mM 2-mercaptoethanol] on microdialysis
filters (0.025 µm) for 30 min at 4°C (45).

Nucleic Acids Research, 2000, Vol. 28, No. 13 2587
DNA substrates
DNA was purified by three washes with 2 ml TE in centricon
tubes (10 kDa MW cut off, Amicon, Witten, Germany). To
measure preferential repair synthesis, 50 ng of ethylated pSP65
(3 kb) together with 50 ng of untreated pSP65/λ (4.2 kb) were
incubated for the times indicated at 25°C in a total volume of
10 µl containing 36 µg of whole cell extract protein in M-
buffer supplemented with 2 µCi [α-32P]dCTP, 0.8 µM dCTP,
50 µM each dATP, dTTP, dGTP, 1 mM ATP, pH 7.5, and 50
ng/µl BSA. After adjustment to 10 mM Tris–HCl pH 7.5, 10
mM EDTA and 0.5% SDS, reactions were terminated by incu-
bation at 65°C. Samples were purified as described for NHEJ
assays and split into two equal aliquots, one of which was
cleaved with EcoRI while the other remained untreated. Differ-
ential 32P incorporation was analyzed by separation of total
EcoRI-digested and untreated samples in 1% agarose gels
containing 1 µg/ml EthBr and subsequent autoradiography.
Thereafter, gels were hybridized to a pSP65-specific 32P-
labeled probe as described for NHEJ-assays to control the
DNA content per lane.
The three substrates for cohesive and blunt end ligation were
derived from the pSP65 (3 kb; Promega, Mannheim, Germany)
by linearization with a single restriction enzyme (RE) (Bam,
5′-PSS; Pst, 3′-PSS; Sma, blunt ends). The eight substrates for
NHEJ were derived from a 4.2 kb modified pSP65 containing
a 1.2 kb λ-DNA insert between the restriction sites used for
substrate preparation (pSP65/λ; 19). Generation of 3 kb linear
plasmid substrates containing two non-homologous ends was
controlled by quantitative excision of the 1.2 kb λ-insert. Each
substrate was named after the pair of RE used in its preparation
[(i) antiparallel ends: Bam/Asp and Bam/Sal, 5′/5′; BstX/BstX and
Kpn/Pst, 3′/3′; (ii) abutting ends: Sma/Sal, bl./5′; Sma/Pst, bl./
3′; Bam/Pst and Sac/Sal, 5′/3′ and 3′/5′, respectively].
Although obtained from cleavage with BstXI (CCAN5/NTGG)
only, the BstX/BstX substrate contains non-complementary
termini, due to the presence of different linkers upstream (5′-
CCACTAAG/GTGG) and downstream (5′-CCACTTTG/
GTGG) of the λ-insert. All substrates were gel purified.
Western blots and gel shift assays
Assay for NHEJ and analysis of products
For immunoblotting, Ku-specific antisera were used (1:100
dilution) and visualized with the NBT/BCIP reagents as
specified in the supplier’s manual. DNA end binding activity
of Ku was assayed with the help of gel shift experiments essen-
tially as described previously (47,48); 1 ng of a 5′-terminally
phosphate-labeled double-stranded 69mer oligonucleotide was
incubated in a total volume of 5 µl in M-buffer with varying
amounts (0.5–15 µg) of K1- and xrs6-Ku80 extract protein in
the presence or absence of 250 ng circular plasmid competitor
for 15 min at room temperature. Samples were separated by
4.5% native PAGE.
In the K1-positive control, ligation and NHEJ was found to
occur efficiently within 6 h at 25°C in the presence of 1–2 ng
DNA/µl and 4–10 µg protein/µl. In standard reactions, 10 ng of
each substrate were incubated for the times indicated at 25°C
in a total volume of 10 µl containing 6–8 µg/µl of extract
protein (always adjusted to the extract with the lowest protein
concentration) in M-buffer supplemented with 1 mM ATP
pH 7.5, 200 µM dNTPs (50 µM each) and 50 ng/µl bovine
serum albumin (BSA). After adjustment to 20 mM Tris–HCl
pH 7.5, 10 mM EDTA, 1% SDS, reactions were stopped by
incubation at 65°C for 5 min. Samples were digested for
30 min at 37°C with 2 mg/ml proteinase K and further purified
by phenol extraction and ethanol precipitation. Equivalents of
2 ng substrate DNA were electrophoresed in 1% agarose gels
containing 1 µg/ml ethidium bromide (EthBr) to separate open
circle (oc) from covalently closed circle (ccc) products and
visualized by in situ gel-hybridization (14) using a pSP65-
specific probe labeled with [α-32P]dCTP by random priming.
For quantification of reaction products in gels, a phosphor-
imaging facility (Fuji, Düsseldorf, Germany) was used.
Circular joined products were cloned by transformation of 4 ng
equivalents of substrate DNA of each NHEJ sample in
Escherichia coli strain DH5α. Cloned NHEJ products were
purified by miniscale extraction and subjected to restriction
analysis (where applicable) or T7-dideoxy sequencing.
RESULTS
General assay for ligation and NHEJ in vitro and control of
extract quality
Plasmids linearized with a single RE represent homologous
substrates with cohesive 5′- or 3′-PSS or blunt ends to measure
the efficiency of ligation. Non-homologous substrates, generated
by cleavage with two different RE, contain non-complementary
DNA ends and are used to measure the efficiency of NHEJ.
Extract-mediated end joining converts both substrate types into
monomeric oc and ccc and various linear multimers which are
separated in agarose gels. Isolation of single NHEJ events for
analysis of junctional sequences is achieved by cloning of the
circular products in E.coli. Using a total of three homologous
and eight non-homologous substrates, we have examined
whole cell extracts from the Ku80-deficient xrs6-cell line (32),
the wild-type CHO-K1 (K1) parent and the hamster Ku80-
cDNA-complemented xrs6 cell line (xrs6-Ku80; 42) to
investigate the role of Ku in NHEJ.
Since xrs6 cells in vivo are defective in DSB repair (27), cell-
free extracts from these cells should display decreased NHEJ
efficiency when compared to extracts from CHO-K1 or xrs6-
Ku80. To verify that observed differences in NHEJ are not
caused by fluctuations in extract quality, we have compared
the quality of different extract batches in an independent assay
that measures excision-repair-mediated DNA synthesis on a
plasmid containing alkylation damage. For this, a supercoiled
Assay for preferential repair synthesis as a control of
extract quality
For preparation of the ethylated plasmid substrate to
measure excision repair synthesis, a 6 mg/ml solution of ENU
in 2-(N-morpholino)-ethane sulfonic acid (MES) buffer (1 mM
MES–HCl pH 6.0, 60 mM NaCl, 0.5 mM EDTA), diluted from
100 mg/ml stock in dimethyl sulfoxide (DMSO), was mixed
with an equal volume of 100 mg/ml pSP65 in alkylation buffer
[50 mM 3-(N-morpholino)-propane sulfonic acid (MOPS)-NaOH
pH 7.4, 60 mM NaCl, 0.5 mM EDTA]. The resulting ENU
concentration of 3 mg/ml generates about four to five 6O-ethyl-
guanines per plasmid molecule and a spectrum of other alkyla-
tion adducts (46). After incubation at 37°C for 60 min, pSP65

2588 Nucleic Acids Research, 2000, Vol. 28, No. 13
from the ENU-treated plasmid leaves single-stranded gaps that
are filled by repair DNA synthesis as detected by incorporation
of radiolabeled dNTPs. Time-dependent incorporation of radi-
olabel occurs preferentially in the 3 kb ENU-treated plasmid
while the 4.2 kb native control plasmid is labeled to a signifi-
cantly lower extent (Fig. 2, top). The controls at the bottom
show that all lanes contain the same amounts of DNA to verify
that preferential labeling of the 3 kb plasmid is not due to
higher amounts of input substrate. This experiment shows that
all three extracts are capable of performing preferential repair
synthesis on the damaged plasmid. The fact that the efficiency
of repair synthesis is nearly the same in all three extracts
demonstrates that the quality of the extract preparations is
comparable and that observed differences in NHEJ can indeed
be ascribed to the different properties of the three cell lines. All
experiments described below were performed with extracts
whose quality had been checked in this assay.
Efficiency of ligation and NHEJ is decreased in xrs6 extracts
A representative example of the different joining properties of
the three extracts is shown in Figure 3. As reflected by the high
levels of circle formation, most efficient ligation of cohesive
5′- (Bam) and 3′-PSS (Pst), and blunt ends (Sma), and NHEJ of
non-homologous 5′/5′-PSS (Bam/Asp) is observed in the K1
extract. By contrast, circle formation is strongly decreased in
xrs6 indicating that the absence of Ku affects the joining of all
substrate types, however, to varying degrees depending on the
type of ends being joined. Generally, the strongest impairment
is observed for Bam (13.8% of wild-type), Sma (14%) and
Bam/Asp (15.7%) while Pst is much less affected (46%)
(Table 1). In the Ku80-complemented xrs6-Ku80 extracts, the
inhibition of ligation and NHEJ is reverted confirming that the
observed defect in xrs6 is indeed caused by the absence of
Ku80. Still, ccc product formation in xrs6-Ku80 reaches on
Figure 2. Control of extract quality. Top, kinetics (times in min below each lane)
at 25°C of incorporation of [α-32P]dCTP in supercoiled 3 kb ENU-treated
(asterisk) and 4.2 kb native plasmid substrates in extracts from CHO-K1 (K1),
xrs6 and xrs6-Ku80 (Ku80) cells. Lanes 1–9, ccc and oc forms of the plasmid
substrates. Lanes 10–18, the same samples after linearization with EcoRI
show preferential accumulation of radiolabel in the 3 kb band. Bottom, same
lanes as at top but after hybridization to a 32P-labeled plasmid probe for control
of DNA content. Shift of ccc bands at time 0 (lanes 1, 4 and 7) results from
different levels of positive supercoil induced by EthBr in the gel which is
caused by residual negative supercoil (relaxed upon extract treatment) plasmids
purified from E.coli.
ENU-treated plasmid (3 kb) was incubated together with a larger
native control plasmid (4.2 kb). Excision of the ethyl-adducts
Table 1. Efficiency of ligation and NHEJ in the different extracts
Terminus
CHO-K1
% ccc
21.7
12.6
11.4
15.2
8.3
xrs6
% ccc
3.0
5.8
1.6
3.5
1.3
1.7
1.2
1.5
1.6
1.1
1.5
1.0
1.4
xrs6-Ku80
configurations
Bam (5′-coh.)
Pst (3′-coh.)
% Di
7.9
% wt
13.8
46.0
14.0
22.8
15.7
17.5
14.5
31.3
18.8
45.9
15.3
24.4
20.0
% Di
5.6
%wt
70.9
93.8
103.2
90.5
100.0
96.5
90.2
105.3
106.3
74.0
82.6
49.2
87.8
% ccc
12.3
7.2
3.4
7.6
4.3
2.5
2.9
1.7
2.9
3.8
5.9
5.3
3.7
% wt
56.8
57.2
29.9
50.0
51.8
25.8
35.0
35.5
34.1
158.7
60.2
130.0
52.9
% Di
5.7
% wt
72.2
11.3
9.3
10.6
9.6
8.7
77.0
Sma (bl.)
10.3
8.2
110.8
86.8
Average lig.
9.5
8.6
Bam/Asp (5′/5′)
Bam/Sal (5′/5′)
BstX/BstX (3′/3′)
Kpn/Pst (3′/3′)
Sma/Sal (bl./5′)
Sma/Pst (bl./3′)
Bam/Pst (5′/3′)
Sac/Sal (3′/5′)
Average NHEJ
12.1
11.4
11.2
11.4
11.2
9.6
12.1
11.0
10.1
12.0
11.9
7.1
9.6
79.3
9.7
10.3
9.5
90.3
8.3
84.8
4.8
12.7
12.8
13.5
13.8
12.1
11.8
111.4
114.3
140.6
126.6
91.7
8.5
2.4
9.8
10.9
13.2
11.4
9.0
4.1
6.5
7.0
10.0
103.5
The efficiency of formation of ccc monomers (ccc) and linear dimers (Di) from the three homologous (ligation, lig.) and eight non-homologous substrates (NHEJ)
after 6 h at 25°C was measured by phosphorimager analysis of gels as radioactivity per product band. Corrected for background and product size (dimers are
twice as large as circles) and expressed as percentage of total radioactivity per lane (% ccc, % Di) to compensate for the differences in the amount of substrate
input, these values are set in relation to the corresponding values obtained in the CHO-K1 wild-type extract (% wt) which is considered to be 100%.

Nucleic Acids Research, 2000, Vol. 28, No. 13 2589
Table 2. Fidelity of ligation and NHEJ
Terminus
CHO-K1
xrs6
Σ jcts
105
46
xrs6-Ku80
configurations
Σ jcts
84
Σ acc.
75
% acc.
89
Σ acc.
59
27
18
14
9
% acc.
56
59
78
39
7
Σ jcts
84
Σ acc.
73
% acc.
87
92
83
86
41
84
2
Ligation cohesive/blunt
5′/5′
25
22
88
24
22
3′/3′
23
18
78
23
24
20
bl./bl.
36
35
97
36
36
31
NHEJ antiparallel ends
133
69
81
61
121
49
123
58
50
5′/5′
60
87
9
18
0
49
3′/3′
64
21
33
72
0
65
1
NHEJ abutting ends
133
36
80
60
121
35
18
18
0
15
51
0
120
36
63
53
69
50
43
57
bl./5′
bl./3′
5′/3′
28
78
25
24
9
37
23
24
12
73
43
59
63
0
0
60
26
Total
350
236
67
347
86
25
327
186
The ‘accuracy’ of ligation and NHEJ was derived from a total of 1024 junctional sequences and determined individually for each extract and each group of terminus
configurations (underlining marks groups comprising two different substrates) in absolute numbers of total junctions (Σ jcts) and ‘accurate’ junctions (Σ acc.), and
the corresponding percentages (% acc.).
average only 50% of the wild-type level (Table 1). Possible
reasons for this difference are discussed below. Thus we may
summarize that the efficiency of ligation and NHEJ is
decreased by about a factor of 5 in xrs6 and a factor of 2 in
xrs6-Ku80. Furthermore, the average efficiency of cohesive
and blunt end ligation is about twice as high as the efficiency
of NHEJ in all three extracts reflecting the higher complexity
of the NHEJ reaction.
In contrast to circle formation, the formation of linear
multimers is much less strongly affected in xrs6 (dimers,
average 90% of wild-type; Table 1). This unexpected result
could be explained in two ways: either the formation of linear
multimers is not or is less dependent on Ku than the formation
of circles or a secondary NHEJ mechanism exists that is
mainly active in the absence of Ku and preferentially promotes
the formation of linear dimers. The first possibility appears
unlikely because in another cell-free system, the formation of
linear multimers was shown to be dependent on Ku (49). On
the other hand, a study performed in Xenopus egg extracts
showed that the formation of linear multimers is even
increased upon immuno-depletion of Ku (50). A detailed inter-
pretation of these seemingly conflicting results together with
our own data is presented in the Discussion.
Accuracy of DNA end joining is decreased in xrs6 extracts
To investigate the hypothesis that Ku enhances the accuracy of
DNA end joining (37,38,41) it is important to define the term
‘accuracy’. While it is obvious that ‘accurate ligation’ of
complementary ends restores the original restriction site used
Figure 3. Efficiency of NHEJ in the different extracts. Kinetics (times in min)
to create the DSB, the definition of ‘accurate NHEJ’ is not self-
at 25°C of ligation (Lig.) of cohesive 5′- (Bam) and 3′-PSS (Pst), and blunt
evident because the joining of non-complementary ends
necessarily causes a change in the original sequence. Still,
general rules for NHEJ were established because extracts from
Xenopus eggs (14,18,19) and mammalian cells (21) generate
highly reproducible spectra of junctions by two main NHEJ
pathways, designated as ‘overlap’ and ‘fill-in’ mechanisms
ends (Sma) and NHEJ of non-complementary 5′/5′-PSS (Bam/Asp) in extracts
from CHO-K1 (K1), xrs6 and xrs6-Ku80 (Ku80) cells. Terminus configurations
are shown at the top of each panel with base matches as white letters. Accurate
NHEJ of Bam/Asp is known to involve formation of a mismatched overlap
whose nicks are closed by ligation (19). Reaction products as marked on the
left side: lin., linear; oc, open circle; ccc, covalently closed circle; M, monomer,
3 kb; Di, dimer, 6 kb; Tri, trimer, 9 kb.

2590 Nucleic Acids Research, 2000, Vol. 28, No. 13
with the former joining anti-parallel and the latter abutting
ends (Fig. 1). In the following, all junctions formed by the
‘overlap’ or ‘fill-in’ pathway are therefore defined as ‘accurate’.
In this study, we have analyzed the sequences of 1024 cloned
junctions obtained by end joining of 11 DNA substrates in the
three extracts. The 273 junctions derived from ligation of cohesive
5′- (Bam), 3′- (Pst) and blunt ends (Sma) are shown in
Figure 4A. NHEJ junctions are subdivided in 377 junctions
derived from substrates containing anti-parallel ends (5′/5′,
Bam/Asp, Bam/Sal; 3′/3′, BstX/BstX, Kpn/Pst) which are
joined by ‘overlap’ formation (Fig. 4B) and 374 junctions
derived from substrates containing abutting ends (bl./5′, Sma/
Sal; bl./3′, Sma/Pst; 5′/3′, Bam/Pst; 3′/5′, Sac/Sal) which are
joined by ‘fill-in’ (Fig. 4C). To facilitate direct comparison, the
accuracy values for ligation and NHEJ are summarized in
Table 2.
have formed by complete loss of both 3′-PSS followed by
blunt end ligation (Fig. 4B, IIIc and d, IVc and d). Since blunt
end ligation was found to be inefficient and inaccurate in xrs6
(see Table 1 and Fig. 4A, III) it cannot be ruled out that the
single base matches present at the junctional breakpoints of
IIIc, and IVc and d (Fig. 4B) stimulate ligation by stabilizing
the joining intermediate.
Strikingly, the accuracy of anti-parallel 3′-PSS joining is not
restored to wild-type levels in xrs6-Ku80 (2%) so that the
spectra of 3′/3′ junctions resemble more closely the ones found
in xrs6 which strictly contrasts the results obtained from
antiparallel 5′-PSS joining. Together with the fact that the
overall joining efficiencies are decreased in xrs6-Ku80 (50%
wild-type, Table 1) this indicates that the features of xrs6-Ku80
are not identical to those of the CHO-K1 parent so that the
complemented cell line takes an intermediate position between
the mutant and the wild-type. To investigate the reason for the
observed differences between K1 and xrs6-Ku80, we
performed western blots to detect potential differences in the
amount of Ku in the two extracts, and gel shift experiments to
assay for the potential differences in DNA end binding activity
of Ku (47,48). While no differences in Ku expression levels
were observed in western blots, a minimal reduction of DNA
end binding activity (~1.5-fold) was detected in the xrs6-Ku80
extract (not shown). Whether this difference is sufficient to
account for the observed differences in general NHEJ efficiency
and accuracy of anti-parallel 3′-PSS joining is unknown but
may indicate that Ku is not able to develop its activities to the
full in xrs6-Ku80. This could be due to the fact that the xrs6-Ku80
cell line is complemented with the Ku80 cDNA but not the
genomic sequence which may contain additional elements
important for the regulation and/or post-translational modification
of Ku80.
Spectra of junctions formed by ligation of cohesive and
blunt ends
The spectra of ligation junctions (Fig. 4A) show that the accuracy
of 5′-PSS and blunt end ligation is significantly decreased in
xrs6 and restored to wild-type levels in xrs6-Ku80 (Table 2).
Interestingly, the accuracy of 3′-PSS ligation remains
unchanged in xrs6 indicating that the dependence of ligation
on Ku varies with the type of ends being joined with blunt ends
(39%) < cohesive 5′-PSS (59%) < 3′-PSS (78%, no difference
to wild-type; note that wild-type accuracy is slightly reduced
with respect to the other two substrates). In a first approximation,
ligation accuracy appears to be related to ligation efficiency
because blunt ends and 5′-PSS yielded much lower ccc product
levels (∼14% of wild-type) than 3′-PSS (46%; Table 1). This
may be due to the substrate specificity of ligase IV which,
together with the XRCC4 protein, is known to be involved in
NHEJ in vivo (51,52).
In an attempt to achieve complete restoration of the accuracy
of antiparallel 3′-PSS joining we performed mixing experiments
between the xrs6 extract and extracts from two other mutant
CHO cell lines, V3 and XR-1, which have normal levels of Ku
but are defective in DNA-PKCS and XRCC4, respectively
(24,43,44,53). Under these conditions, all proteins are derived
from their genomic sequences. Both the V3 as well as the XR-1
extract alone exhibit not only strongly decreased efficiency
(not shown) but also accuracy of NHEJ as shown in Table 3 for
BstX/BstX (compare Fig. 4B, III). The decrease in accuracy is
less pronounced in V3 than in XR-1 indicating that DNA-PKCS
is less important for NHEJ accuracy than XRCC4. Remarkably,
mixing of both mutant extracts with the xrs6 extract results in
restoration of the accuracy of BstX rejoining to wild-type
levels. This indicates that the lack of accuracy of 3′-PSS
joining in the xrs6-Ku80 cell line could be due to slightly
different interactions between the proteins required for the
trimming of 3′-PSS and the Ku protein expressed from its
cDNA with respect to the one expressed from its genomic
sequence.
Spectra of junctions formed by NHEJ of anti-parallel ends
Substantial fractions of the junctional spectra formed in the K1
extract display the typical features of ‘overlap’ and ‘fill-in’
junctions and are therefore defined as ‘accurate’ [Fig. 4B, (b)
junctions; Fig. 4C, (a) junctions]. This result confirms that the
rules for ‘accurate’ NHEJ, originally established for the
Xenopus system, are generally applicable.
The ‘accuracy’ of anti-parallel 5′-PSS joining by ‘overlap’
formation is highest in both K1 (87%) and xrs6-Ku80 (84%)
(Fig. 4B, Ib and IIb). In xrs6, however, the fraction of these
‘accurate’ junctions is much smaller (18%) so that accuracy is
decreased by about a factor of 5 (Table 2). Instead, NHEJ in
xrs6 has shifted from ‘overlap’ formation to another pathway
(Ia, 26%; IIa, 27%) in which both 5′-PSS are blunted by fill-in
followed by blunt end ligation. In some cases (not shown
explicitly in Figure 4 but summarized in the fraction of ‘other
junctions’), the 5′-PSS have lost some nucleotides while the
rest is conserved by fill-in and blunt end ligation (I, 4%; II,
23%), a phenomenon also observed with ligation of cohesive Bam
5′-PSS (Fig. 4A, I, 14%). This indicates that, in the absence of Ku,
‘overlap’ formation is competed by fill-in of 5′-PSS to yield
blunt ends, which are subsequently ligated.
Spectra of junctions formed by NHEJ of abutting ends
As expected, the joining of abutting terminus configurations in
K1 is achieved mainly by the ‘accurate’ ‘fill-in’ pathway
which conserves the entire sequences of both 5′- and 3′-PSS by
fill-in DNA synthesis (Fig. 4C). The accuracy of this pathway
varies with the type of ends being joined with Sma/Sal (I) and
Sac/Sal (IV) displaying the highest accuracy (78 and 80%,
The ‘accuracy’ of anti-parallel 3′-PSS joining is generally
strongly decreased in K1 (33%, Table 2; Fig. 4B, IIIb and
IVb), but in xrs6, not a single junction fulfills the criteria of
‘accurate’ ‘overlap’ formation (0%). Instead, most junctions

Nucleic Acids Research, 2000, Vol. 28, No. 13 2591
Figure 4. Spectra of junctions generated by DNA end joining in extracts from
CHO-K1, xrs6 and xrs6-Ku80 cells. (A) Ligation of cohesive and blunt ends;
(B) NHEJ of anti-parallel ends by overlap formation; (C) NHEJ of abutting
ends by fill-in. Terminus configurations including flanking double-strand
sequences are shown at the top of each panel with complementary bases used
in junction formation highlighted by white letters. Junctions are listed below as
top strand sequences with vertical lines marking breakpoint (middle) and blunt
positions (left and right). Note that only the main junctions are shown while
less frequent events are summarized under ‘other junctions’, which comprise
all three possible types of interactions (between two PSS, one PSS with the
flanking duplex of the other terminus and two flanking duplexes). ‘Larger
deletions’ designate junctions not determined (n.d.) due to failure of primer
annealing during sequencing. Sequences between broken lines highlight ‘accurate’
junctions of which the ones formed by overlap from Bam/Asp, BstX/BstX and
Kpn/Pst contain mismatches that yield two different sequences (boxes with
two bases) after segregation in E.coli (18,19). Total numbers of bases deleted
from each junction are listed below the open triangles with asterisks marking
the ‘accurate’ junction. Underlining indicates a restored restriction site used for
quicker analysis (e.g. BamS). Microhomology patches (white letters) found at
breakpoints are ascribed arbitrarily to the left end although the actual origin of
the affected bases cannot be determined unambiguously. Frequencies of
junctions obtained in the different extracts are shown on the right. Σ indicates
total numbers of junctions analyzed, bold numbers and underlined percentages
frequencies of ‘accurate’ junctions, numbers in brackets the fraction of ‘other
junctions’ displaying microhomologies of 1–3 bp at their breakpoints.

2592 Nucleic Acids Research, 2000, Vol. 28, No. 13
Table 3. Restoration of accuracy in mixed extracts
K1
xrs6
Xrs6-Ku80
0 of 30
0%
V3
XR-1
1 of 21
5%
xrs6 × V3
14 of 24
58%
xrs6 × XR-1
12 of 24
50%
13 of 27
48%
0 of 36
0%
8 of 24
33%
Given are the fractions of clones and corresponding percentages of ‘accurate’ joining of anti-parallel 3′-PSS (BstX/BstX; see Fig. 4B, III)
in the wild-type K1 extract, the complemented xrs6-Ku80 extract, the mutant xrs6, V3 and XR-1 extracts and mixtures of xrs6 with V3
(xrs6 × V3) and xrs6 with XR-1 extract (xrs6 × XR-1), respectively.
respectively) followed by Sma/Pst (II) with intermediate
(38%) and Bam/Pst (III) with lowest accuracy (17%). This
shows that NHEJ in K1 is able to fill in 3′-PSS by DNA
synthesis that is primed at the partner terminus, a feature an
alignment factor is supposed to have (14). Although slightly
reduced, accuracy patterns obtained in xrs6-Ku80 resemble
those of K1, reflecting the same substrate specificity (Fig. 4C
and Table 2).
The difference in accuracy of the two 5′/3′-terminus configu-
rations Sac/Sal and Bam/Pst observed in K1 (80 versus 17%)
and xrs6-Ku80 (64 versus 13%) is striking and not easily
explained but suggests that, apart from the structural configu-
ration of the ends, the sequences of involved overhangs may
also be important. In this context, it may be of relevance that an
A is filled in at the ultimate position of the 5′-PSS in Sac/Sal.
As many DNA polymerases are capable of adding single
nucleotides to blunt ends, preferentially A or a nucleotide
identical to the last one filled in (54,55), it is possible that
untemplated addition of a second A to the 3′-OH of the blunted
5′-PSS facilitates pairing with the ultimate T of the partner 3′-PSS
which would enhance fill-in of the 3′-PSS and thus result in the
high accuracy achieved with this substrate (56,57). In Bam/Pst,
by contrast, a C is filled in the ultimate position of the 5′-PSS
and a T would have to be added in order to facilitate pairing, a
configuration expected to reduce the accuracy of this joining
reaction.
With the exception of Sma/Sal, the spectra of junctions
produced in xrs6 are completely different from the ones
produced in K1 and xrs6-Ku80. Joining of Sma/Sal yields 51%
of ‘accurate’ junctions (Fig. 4C, I) which indicates that xrs6 has
a strong capability of filling 5′-PSS. The fact that the accuracy of
Sma/Sal joining is even higher than the one obtained for blunt
end ligation (39%, Fig. 4A, III) indicates the possibility that
the polymerase responsible for 5′-PSS fill-in might be able to
support subsequent blunt end ligation by transiently stabilizing
the joining intermediate. In spite of this strong polymerase
activity in xrs6 which also conserves the 5′-PSS sequences of
the other abutting terminus configurations (Fig. 4C, III and IV),
not a single ‘accurate’ junction is formed from substrates
containing 3′-PSS (Table 2). Instead, all 3′-PSS sequences are
lost entirely leading to blunt end ligation between the trimmed
3′-PSS and the partner blunt end (Fig. 4C, IIb) or the filled 5′-PSS
(IIIb and IVb). The fact that the ability to conserve entire 3′-PSS
sequences is restored to almost wild-type levels in xrs6-Ku80
demonstrates that the ‘fill-in’ joining pathway is strictly
dependent on Ku.
20%), while in K1 and xrs6-Ku80, the same homology patches
are not used at all (IIc) or with significantly lower frequency
(IVc; both 6%). This suggests that NHEJ in the absence of Ku
is assisted by base pairing interactions which is confirmed by
results obtained in vivo in mammalian cells (41) and yeast
(37,38). The fact that the use of terminus-flanking micro-
homology patches leads to the generation of small deletions is
consistent with the idea that the fidelity of NHEJ is reduced in
Ku-deficient cells (37–39).
Stability of DNA ends is decreased in xrs6
To investigate whether the accuracy of end joining is related to
the stability of DNA ends, we have calculated the total number
of each type of end (5′-PSS, 3′-PSS and blunt) from the 1024
junctions analyzed, taking into account that each junction
originates from the interaction of two ends (Table 4). In K1-
and xrs6-Ku80, more 5′-PSS (79%; 76%) and blunt ends
(81%; 75%) participate in the formation of ‘accurate’ junctions
than 3′-PSS (49%; 31%). This bias is even more pronounced in
xrs6 (31 and 35% versus 13%), and remarkably, the 13% of
‘accurate’ junctions stem exclusively from accurate ligation of
cohesive 3′-PSS (78% of all ligation junctions are ‘accurate’)
while not a single 3′-PSS participated in the formation of an
‘accurate’ NHEJ junction.
The stability of DNA ends in xrs6 was further analyzed in
the fraction of ‘inaccurate’ junctions (Table 4) where end
stability is expressed as percentage of ends that remained intact
(–0 nt; for 5′- and 3′-PSS this means sequence conservation by
fill-in DNA synthesis), or have lost 1–4 nt, or have suffered
from deletions reaching into the flanking duplex region
(‘beyond blunt’). In xrs6, the fraction of junctions in which at
least one end is degraded ‘beyond blunt’ is strongly elevated
(64 versus 18% and 22% in K1 and xrs6-Ku80, respectively).
Although the extent of duplex degradation is generally only
modest (xrs6: 20%, –1 nt; 32%, –2 to 14 nt; 12%, larger. K1:
7%, –1 nt; 7%, –2 to 9 nt; 4%, larger. xrs6-Ku80: 9%, –1 nt;
6%, –2 to 8 nt; 8%, larger) opposed to in vivo studies (39,41),
this result indicates that degradation of DNA ends is minimized
in the presence of Ku.
Furthermore, we found that blunt ends are the most stable
ends in K1 and xrs6-Ku80 (–0, both 14%), followed by 5′-PSS
(7%; 11%) and finally 3′-PSS (4%; 2%). In xrs6, blunt ends
and 5′-PSS exhibit similar stability (35 and 32%) indicating
that there is a strong tendency for fill-in of 5′-PSS to generate
blunt ends that may subsequently undergo blunt end ligation.
By contrast, the stability of 3′-PSS in xrs6 is zero with most
ends being degraded to the blunt end position (–4 nt, 30%) or
beyond (48%). The few cases in which 3 nt (–1, 0.4%) and 1 nt
(–3, 8%), respectively, of the 3′-PSS sequences are conserved
Besides blunt end joining, another type of NHEJ that creates
small deletions with microhomologies at the breakpoints is
preferentially realized in xrs6 (Fig. 4C, IIc, 48% and IVc,

Nucleic Acids Research, 2000, Vol. 28, No. 13 2593
Table 4. Stability of ends
Terminus
5′-PSS
CHO
297
3′-PSS
CHO
271
Blunt ends
configurations
Σ ends
xrs6
288
36
Ku80
260
153
76%
48
xrs6
276
36
Ku80
262
80
CHO
132
107
81%
72
xrs6
130
46
Ku80
132
99
Σ ends in acc. jcts
235
130
79%
50
31%
92
49%
46
13%
46
31%
48
35%
72
75%
72
lig.
acc. lig.
88%
247
59%
196
18%
198
69%
92%
212
72%
63
78%
225
78%
230
0%
83%
214
19%
182
69%
97%
60
39%
58
86%
60
NHEJ
acc. NHEJ
Σ ends in inacc. jcts
77%
62
42%
141
62%
25
31%
84
62%
33
240
87%
21%
24%
51%
17%
65%
25%
Loss of term. nucl.
–0
22/7%
6/2%
8/3%
7/2%
8/3%
11/4%
92/32%
12/4%
17/6%
11/4%
15/5%
51/18%
28/11%
9/3%
10/4%
13/5%
24/9%
19/7%
30/11%
45/16%
0
6/2%
18/14%
45/35%
19/14%
–1
0
7/3%
–
–
–
–2
3/1%
1*/0.4%
23*/8%
84/30%
132/48%
19/7%
24/9%
74/28%
53/20%
–
–
–
–3
5/2%
–
–
–
–4
11/4%
7/3%
–
–
–
Beyond blunt
7/5%
39/30%
14/11%
The total number of ends (Σ), calculated individually for each terminal structure (5′-PSS, 3′-PSS and blunt), was derived from the total of 1024 junctions taking
into account that two ends have participated in the formation of each junction. Given are the absolute numbers of ends (Σ) involved in ligation (lig.) and NHEJ
and the distribution of the corresponding ‘accurate’ junctions (acc. jcts, acc. lig., acc. NHEJ). The fraction of ‘inaccurate’ junctions (inacc. jcts) comprises the
numbers of ends involved in both ‘inaccurate’ ligation and NHEJ junctions. For the evaluation of the stability of a particular type of end only ‘inaccurate’
junctions are considered. Stability is reflected by the absolute number and percentages of intact ends (–0). In addition, the values for loss of 1, 2, 3 and 4 nt (coincides with
the blunt end position) from 5′- and 3′-PSS are given [asterisks in the xrs6/3′-PSS column indicate loss from 3′-PSS derived from junctions with a 3 bp (1 case)
and 1 bp microhomology (23 cases) at the breakpoint]. Deletions affecting the flanking duplex are summarized in the ‘beyond blunt’ fraction.
are without exception derived from junctions containing a
microhomology of 3 or 1 bp, respectively, at their breakpoints.
This indicates that the conservation of parts of 3′-PSS
sequences can be realized only in junctions that provide base
match possibilities between the 3′-PSS and the partner
terminus. In ‘genuine’ blunt end junctions (no homology at
breakpoint), however, 3′-PSS are always degraded to the blunt
position or beyond.
In conclusion, our results show that xrs6 cells completely
lack the ability to perform DNA fill-in synthesis on 3′-PSS
and, furthermore, are hardly able to form ‘overlap’ junctions.
This indicates that Ku, on the one hand, facilitates the fill-in of
3′-PSS in abutting terminus configurations, the crucial step in
the ‘fill-in’ joining pathway (14), and, on the other hand,
enhances the formation of mismatched overlaps between anti-
parallel PSS, the crucial step in the ‘overlap’ pathway (18,19).
Together these data provide strong evidence that Ku acts as an
alignment factor in NHEJ.
is required for efficient and accurate joining of complementary
and non-complementary ends.
In vivo studies in Ku70/80-deficient yeast strains (37,38)
indicated that Ku might be of crucial importance for the
efficiency and fidelity of NHEJ, presumably by protecting
DNA ends from degradation and thus enhancing the accuracy
of NHEJ. Data obtained from transfection of linear DNA
substrates in the Ku80-deficient xrs6 cell line, however,
remained controversial with regard to this aspect because some
authors observed decreased levels of efficiency and increased
levels of end degradation (39) while others did not and
therefore suggested that Ku is not required for efficient and
‘accurate’ NHEJ (40,41). The cause of these different results is
probably the transfection procedure itself, because terminal
degradation of the substrate molecules is not necessarily a
result of the NHEJ reaction but may occur in the nucleus and/
or during passage of the substrate through the cytoplasm. Thus,
only a fraction of the substrate molecules undergoing NHEJ
would maintain their original terminal structures, which in turn
would significantly influence the spectra of NHEJ products
observed. This problem is minimized in vitro because joining-
active extract components gain access to the substrates directly
upon incubation so that possibly observed terminal degradation
is indeed the result of the NHEJ process itself.
DISCUSSION
Using DSB repair competent and deficient CHO cell lines, we
have established an in vitro system that facilitates studying the
functions of DSB repair proteins, such as the Ku70/80
heterodimer described here, in the mammalian NHEJ process.
Our data provide direct evidence that the Ku70/80 heterodimer
In summary we have shown that the efficiency and accuracy
of ligation as well as NHEJ is significantly decreased in the

2594 Nucleic Acids Research, 2000, Vol. 28, No. 13
absence of Ku and that accuracy is related to the type of ends
being joined as reflected by the differential stability of the
different DNA ends (blunt ≈ 5′-PSS >> 3′-PSS) in xrs6. In this
context, it is remarkable that ligation of cohesive 3′-PSS is
very efficient and accurate (3′-PSS > 5′-PSS > blunt ends) in
the absence of Ku while NHEJ of non-homologous terminus
configurations containing 3′-PSS is not (accuracy of zero in all
cases: 3′/3′ = bl./3′ = 5′/3′ < 5′/5′ < bl./5′). This apparent
paradox may be explained by the properties of the proteins
involved in ligation and NHEJ. Ligation is expected to require
only XRCC4 and DNA ligase IV which is known to be essential
for NHEJ in vivo (51,52). In another in vitro system containing
ligase IV, highest ligation efficiency was observed for 3′-PSS
> 5′-PSS > blunt ends (49) which is consistent with our results.
A substrate preference of ligase IV for 3′-PSS would explain
why ligation of cohesive 3′-PSS can occur efficiently and
accurately in xrs6 and is not significantly stimulated by Ku in
K1- and xrs6-Ku80. On the other hand, the efficiency and accu-
racy of cohesive 5′-PSS and blunt end ligation is considerably
increased in K1- and xrs6-Ku80, which indicates that Ku may
enhance the efficiency of DNA ligase IV on less preferred
DNA ends and is consistent with results from in vitro studies
using purified Ku and ligase IV (58).
As seen by the strong tendency in xrs6 for fill-in of 5′-PSS to
generate blunt ends and degradation of 3′-PSS to the blunt
position, blunt end joining is one of two alternative NHEJ
pathways in the absence of Ku. The other Ku-independent
pathway involves the formation of short deletions at sites of
microhomology. This becomes most evident in the fractions of
junctions displaying a 4 bp microhomology at their break-
points (see Fig. 4C, II and IV) which is significantly elevated
in xrs6 (6.3 versus 0.3% in K1 and 0.9% in xrs6-Ku80). The
underlying reaction is reminiscent of mechanisms designated
as direct-repeat end joining (62,63) or microhomology-driven
SSA (45,64,65) both of which use small homology patches
such as direct repeats. Basically, such a mechanism can proceed
by exonuclease-mediated trimming or helicase-mediated
duplex unwinding of the ends to generate extended 3′-PSS that
can undergo either SSA at the exposed microhomology
patches or duplex invasion to create a D-loop. In this context,
it is interesting to note that increased fractions of deletions with
breakpoint homologies have been found in vivo in xrs6 cells
(41) and Ku-deficient yeast strains (37,38), and in vitro in Ku-
devoid protein fractions partially purified from Xenopus egg
extracts (45) and calf thymus (66). Together with the data
presented here, this suggests that an error-prone pathway that
creates deletions is dominant in the absence of Ku.
In contrast to ligation, NHEJ is expected to involve other
factors in addition to XRCC4 and ligase IV, such as DNA
polymerase(s) and exonuclease(s). As postulated in the ‘fill-in’
joining pathway, ‘accurate’ NHEJ of 3′-PSS in abutting
terminus configurations occurs by DNA fill-in synthesis
primed at the 3′-OH group of the partner terminus (14). This
pathway is entirely blocked in xrs6, since not a single ‘accurate’
junction is formed from terminus configurations containing 3′-PSS.
This result strongly suggests that Ku may act as an alignment
factor by transiently holding abutting ends together and thus
allowing the polymerase to prime DNA synthesis at the partner
terminus and to bridge the gap. Previously, DNA polymerases
were suggested to take the part of an alignment factor due to
their ability to perform DNA synthesis on discontinuous
templates (56,57,59). This possibility, however, appears to be
unlikely because DNA polymerase β, which was implicated in
NHEJ recently (60), is not able to perform this type of reaction
under physiological conditions (61). Furthermore, xrs6 cells
which are expected to contain similar DNA polymerase levels
as K1 cells display efficient polymerase activity on 5′-PSS but
are unable to fill 3′-PSS. This suggests that DNA polymerases
alone are not able to exert alignment function.
In addition to facilitating discontinuous DNA synthesis in
abutting terminus configurations, Ku may also stabilize short
mismatched overlaps formed between insufficiently stable
single complementary base pairs of anti-parallel PSS and thus
permits fill-in synthesis, nick ligation and other repair reactions
to occur on these intermediates. Together these results show
that Ku fulfills all previously proposed criteria of an alignment
factor in NHEJ (14,19). The previous finding of elevated
frequency of imprecise NHEJ in cells with mutated Ku (37–39)
has led to the argument that Ku protects DNA ends from
degradation. However, in the light of Ku acting as an alignment
factor, we rather support the model of Ramsden and Gellert
(58) which suggests that processing of ends and end joining are
competitive processes, and Ku acts to increase the frequency of
accurate end joining by stabilizing crucial intermediates thus
increasing the rate of accurate NHEJ.
In this context, it is interesting that the absence of Ku
appears to affect mainly the formation of circles while the
formation of linear multimers is much less strongly influenced.
This intriguing phenomenon could be in part explained by the
fact that Ku promotes DNA looping in vitro by self-association
which would lead to the preferential formation of circles (67).
On the other hand, Baumann and West showed that the formation
of linear multimers is also strictly dependent on Ku in extracts
derived from human lymphoblastoid cell lines (49). Interestingly,
these authors neither obtained any circles in their extracts nor
did they retain multimer formation activity upon addition of
several inhibitors of NHEJ. These results indicate a crucial
difference between their extracts and ours, which is due to the
different preparation procedures rather than the different cell
lines used.
In the light of Labhart’s results showing that Xenopus egg
extracts retain a strong multimer formation activity even after
immuno-depletion of Ku while circle formation is almost
completely abolished (50)—which is fully consistent with our
own results obtained in xrs6—it is likely that dimer formation in
the absence of Ku is the result of a secondary Ku-independent
NHEJ activity. This NHEJ activity, which appears to be absent
in the extracts of Baumann and West (49), is not only error-
prone but also promotes preferentially the formation of linear
multimers. In the presence of Ku, both circles and multimers
would be formed at about similar levels by accurate NHEJ. In
addition to that, part of the multimer fraction could arise by the
secondary error-prone NHEJ pathway. In the absence of Ku,
this pathway can become fully active and promote the formation
of linear multimers (but no circles) so that the wrong impression
could arise as if Ku-dependent NHEJ formed only circles.
Why does the secondary Ku-independent NHEJ pathway
form mainly linear multimers and hardly any circles? Interestingly,
the linear dimers formed in the Ku-depleted Xenopus egg
extract are mainly arranged in head-to-head (H/H) and tail-to-tail
(T/T) but rarely in head-to-tail (H/T) orientation, although
random joining of linear molecules is expected to create twice

Nucleic Acids Research, 2000, Vol. 28, No. 13 2595
other NHEJ proteins and identify the factors involved in Ku-
independent NHEJ.
ACKNOWLEDGEMENTS
We thank Drs M. F. Rajewsky, J. Thomale, C. Schunck and the
members of AGI for stimulating discussions and laboratory
space. This work was funded by a grant (98.053.1/2) to P.P. by
the Wilhelm Sander Stiftung für Krebsforschung. P.P. is
holder of a fellowship of the Heisenberg program of the
Deutsche Forschungsgemeinschaft.
REFERENCES
1. Shinohara,A. and Ogawa,T. (1995) Trends Biochem. Sci., 20, 387–391.
2. Kanaar,R., Hoeijmakers,J.H.J. and van Gent,D.C. (1998) Trends Cell Biol.,
8, 483–489.
3. Haber,J.E. (1999) Nature, 398, 665–667.
4. Critchlow,S.E. and Jackson,S.P. (1998) Trends Biochem. Sci., 23, 394–398.
Figure 5. Features of the Ku-dependent and -independent pathway of NHEJ.
In the Ku-dependent pathway, the Ku-heterodimers (grey circles) bind to the
ends of a broken DNA duplex (black lines) and serve as alignment factors to
mediate, together with DNA-PKCS, XRCC4 and ligase IV, accurate NHEJ that
forms both circles and linear dimers. The Rad50/Mre11/NBS1 nuclease
complex (in yeast NBS1 is Xrs2) is probably also involved in this pathway (4).
In the error-prone Ku-independent pathway, microhomology patches (black
boxes) are used for SSA that forms mainly dimers containing deletions. The
factors involved in this pathway are presently unknown but the Rad52 protein,
one of the key players in homologous recombination, might be a possible
candidate because it binds, like Ku, to DNA ends and promotes SSA (3).
5. Thacker,J. (1999) Biochimie, 81, 77–85.
6. Pellicer,A., Robins,D., Wold,B., Sweet,R., Jackson,J., Lowy,I., Roberts,J.M.,
Sim,G.K. and Silverstein,S. (1980) Science, 209, 1414–1422.
7. Perucho,M., Hanahan,D. and Wigler,M. (1980) Cell, 22, 309–317.
8. Orr-Weaver,T.L. and Szostak,J.W. (1983) Microbiol. Mol. Biol. Rev., 63,
349–404.
9. Wilson,J.H., Berget,P.B. and Pipas,J.M. (1982) Mol. Cell. Biol., 2, 1258–1269.
10. Roth,D.B, Porter,T.N. and Wilson,J.H. (1985) Mol. Cell. Biol., 5, 2599–2607.
11. Roth,D.B. and Wilson,J.H. (1986) Mol. Cell. Biol., 6, 4295–4304.
12. Roth,D.B. and Wilson,J.H. (1988) In Kucherlapati,R. and Smith,G. (eds),
Genetic Recombination. Am. Soc. Microbiol., Washington, DC, pp. 621–653.
13. Pfeiffer,P. and Vielmetter,W. (1988) Nucleic Acids Res., 16, 907–923.
14. Thode,S., Schäfer,A., Pfeiffer,P. and Vielmetter,W. (1990) Cell, 60, 921–928.
15. Goedecke,W., Pfeiffer,P. and Vielmetter,W. (1994) Nucleic Acids Res.,
22, 2094–2101.
as many H/T as either H/H or T/T molecules (50). This is
reminiscent of a mechanism observed previously in human cell
extracts and purified protein fractions thereof (65,68,69). Since
the fraction of multimers formed in our extracts are composed
mainly of H/H and T/T products whose junctions did not arise by
ligation of the corresponding complementary ends (E.Feldmann,
W.Goedecke and P.Pfeiffer, unpublished result), micro-
homology-driven SSA (45,64,65) appears to be suited best to
account for multimer formation in our extracts. As pointed out
previously by Young et al. (65), the palindromic nature of the
plasmid polylinker used for substrate preparation exhibits
more extended regions of complete homology in H/H and T/T
orientation than in H/T orientation, which leads to pairing in
the region of homology and loss of a certain number of nucleo-
tides. A prediction of this model is that microhomology-driven
SSA of substrate molecules in H/T orientation would be less
efficient, and the precise number of base pairs lost from the
junction would be dependent on the position of small
fortuitous patches of homology. This is consistent with the
elevated fractions in xrs6 of junctions (all of which are H/T
because they are derived from cloned circles) containing 4 bp
microhomologies at their breakpoints and would explain why,
in the absence of Ku, the fraction of H/T products (circles)
decreases while the fraction of multimers (mainly H/H and T/T)
remains largely unchanged. Further experiments are required
to investigate this interesting phenomenon.
16. Kramer,K.M., Brock,J.A., Bloom,K., Moore,J.K. and Haber,J.E. (1994)
Mol. Cell. Biol., 14, 1293–1301.
17. Mézard,C. and Nicolas,A. (1994) Mol. Cell. Biol., 14, 1278–1292.
18. Pfeiffer,P., Thode,S., Hancke,J., Keohavong,P. and Thilly,W.G. (1994)
Mutagenesis, 9, 527–535.
19. Pfeiffer,P., Thode,S., Hancke,J. and Vielmetter,W. (1994) Mol. Cell. Biol.,
14, 888–895.
20. Bøe,S.-O., Sodroski,J., Helland,D.E. and Farnet,C.M. (1995)
Biochem. Biophys. Res. Commun., 215, 987–993.
21. Daza,P., Reichenberger,S., Göttlich,B., Hagmann,M., Feldmann,E. and
Pfeiffer,P. (1996) Biol. Chem., 377, 775–786.
22. Jeggo,P.A. (1990) Mutat. Res., 239, 1–16.
23. Jackson,S.P. and Jeggo,P.A. (1995) Trends Biol. Sci., 20, 412–415.
24. Li,Z., Otevrel,T., Gao,Y., Cheng,H.-L., Seed,B., Stamato,T.D.,
Taccioli,G.E. and Alt,F.W. (1996) Cell, 83, 1079–1089.
25. Critchlow,S.E., Bowater,R. and Jackson,S.P. (1997) Curr. Biol., 7, 588–598.
26. Grawunder,U., Wilm,M., Wu,X., Kulesza,P., Wilson,T.E., Mann,M. and
Lieber,M.R. (1997) Nature, 388, 492–495.
27. Taccioli,G.E., Gottlieb,T.M., Blunt,T., Priestly,A., Demengeot,J.,
Mizuta,R., Lehmann,A.R., Alt,F.W., Jackson,S.P. and Jeggo,P.A. (1994)
Science, 265, 1442–1445.
28. Blunt,T., Finnie,N.J., Taccioli,G.E., Smith,G.C., Demengeot,J.,
Gottlieb,T.M., Mizuta,R., Varghese,A.J., Alt,F.W., Jeggo,P.A. and
Jackson,S.P. (1995) Cell, 80, 813–823.
29. Gu,Y., Jin,S., Gao,Y., Weaver,D.T. and Alt,F.W. (1997) Proc. Natl Acad.
Sci. USA, 94, 8076–8081.
30. Featherstone,C. and Jackson,S.P. (1999) Mutat. Res., 434, 3–15.
31. Roth,D.B., Lindahl,T. and Gellert,M. (1995) Curr. Biol., 5, 496–499.
32. Jeggo,P.A. and Kemp,L.M. (1983) Mutat. Res., 112, 313–327.
33. Chen,F., Peterson,S.R., Story,M.D. and Chen,D.J. (1996) Mutat. Res.,
362, 9–19.
34. Errami,A., Smider,V., Rathmell,W.K., He,D.M., Hendrickson,E.A.,
Zdzienicka,M.Z. and Chu,G. (1996) Mol. Cell. Biol., 16, 1519–1526.
35. Pergola,F., Zdzienicka,M.Z. and Lieber,M.R. (1993) Mol. Cell. Biol., 13,
3464–3471.
In conclusion, the results presented here suggest that Ku
enhances the efficiency of ‘accurate’ NHEJ by functioning as
an alignment factor. In the absence of Ku, NHEJ is not
completely disabled but considerably less efficient and error-
prone in that secondary mechanisms based on blunt end joining
and microhomology-driven SSA are used preferentially (Fig. 5).
Further studies in cell-free extracts from other DSB repair
deficient cell lines will help to elucidate the functions of the
36. Taccioli,G.E., Rathburn,G., Oltz,E., Stamato,T., Jeggo,P.A. and Alt,F.W.
(1993) Science, 260, 207–210.

2596 Nucleic Acids Research, 2000, Vol. 28, No. 13
37. Boulton,S.J. and Jackson,S.P. (1996) EMBO J., 15, 5093–5103.
38. Boulton,S.J. and Jackson,S.P. (1996) Nucleic Acids Res., 24, 4639–4648.
39. Liang,F. and Jasin,M. (1996) J. Biol. Chem., 271, 14405–14411.
40. Han,J.-O., Stehen,S.B. and Roth,D.B. (1997) Mol. Cell. Biol., 17, 2226–2234.
41. Kabotyanski,E.B., Gomelsky,L., Han,J.-O., Stamato,T.D. and Roth,D.B.
(1998) Nucleic Acids Res., 26, 5333–5342.
42. Singleton,B.K., Priestley,A., Steingrimsdottir,H., Gell,D., Blunt,T.,
Jackson,S.P., Lehman,A.R. and Jeggo,P.A. (1997) Mol. Cell. Biol., 17,
1264–1273.
43. Whitmore,G.F., Varghese,A.J. and Gulyas,S. (1989) Int. J. Radiat. Biol.,
56, 657–665.
44. Stamato,T.D., Weinstein,R., Giaccia,A. and Mackenzie,L. (1983) Somat.
Cell Mol. Genet., 9, 165–173.
45. Göttlich,B., Reichenberger,S., Feldmann,E. and Pfeiffer,P. (1998) Eur. J.
Biochem., 258, 387–395.
46. Engelbergs,J. (1998) PhD Thesis. University of Essen, Essen, Germany.
47. Getts,R.C. and Stamato,T.D. (1994) J. Biol. Chem., 269, 15981–15984.
48. Rathmell,W.K. and Chu,G. (1994) Mol. Cell. Biol., 14, 4741–4748.
49. Baumann,P. and West,S.C. (1998) Proc. Natl Acad. Sci. USA, 95,
14066–14070.
54. Clark,J.M. (1988) Nucleic Acids Res., 16, 9677–9688.
55. Clark,J.M. (1991) Gene, 104, 75–80.
56. King,J.S., Fairley,C.F. and Morgan,W.F. (1994) J. Biol. Chem., 269,
13061–13064.
57. King,J.S., Fairley,C.F. and Morgan,W.F. (1996) J. Biol. Chem., 271,
20450–20457.
58. Ramsden,D.A. and Gellert,M. (1998) EMBO J., 17, 609–614.
59. Islas,A.L., Fairley,C.F. and Morgan,W.F. (1998) Nucleic Acids Res., 26,
3729–3738.
60. Wilson,T.E. and Lieber,M.R. (1999) J. Biol. Chem., 274, 23599–23609.
61. Reichenberger,S. and Pfeiffer,P. (1998) Eur. J. Biochem., 251, 81–90.
62. Thacker,J., Chalk,J., Ganesh,A. and North,P. (1992) Nucleic Acids Res.,
20, 6183–6188.
63. Thacker,J. (1999) Acad. Sci. Paris, 322, 103–108.
64. Lehman,C.W., Trautman,J.K. and Carroll,D. (1994) Nucleic Acids Res.,
22, 434–442.
65. Nicolás,A.L., Munz,P.L. and Young,C.S.H. (1995) Nucleic Acids Res.,
23, 1036–1043.
66. Mason,R.M., Thacker,J. and Fairman,M.P. (1996) Nucleic Acids Res., 24,
4946–4953.
50. Labhart,P. (1999) Mol. Cell. Biol., 19, 2585–2593.
67. Cary,R.B., Peterson,S.R., Wang,J., Bear,D.G., Bradbury,E.M. and
Chen,D.J. (1997) Proc. Natl Acad. Sci. USA, 94, 4267–4272.
68. Derbyshire,M.K., Epstein,L.H., Young,C.S.H., Munz,P.L. and Fishel,R.
(1994) Mol. Cell. Biol., 14, 156–196.
51. Wilson,T.E., Grawunder,U. and Lieber,M.R. (1997) Nature, 388, 495–498.
52. Frank,K.M., Sekiguchi,J.M., Seidl,K., Swat,W., Cheng,H.-L.,
Davidson,L., Kangaloo,L. and Alt,F.W. (1998) Nature, 396, 173–177.
53. Peterson,S.R., Kurimasa,A., Oshimura,M., Dynan,W.S., Bradbury,E.M.
and Chen,D.J. (1995) Proc. Natl Acad. Sci. USA, 92, 3171–3174.
69. Nicolas,A.L. and Young,C.S.H. (1994) Mol. Cell. Biol., 14, 170–180.