Mammalian mitochondrial endonuclease G
Digestion of R-loops and localization in intermembrane space
Takashi Ohsato
1
, Naotada Ishihara
2
, Tsuyoshi Muta
1
, Shuyo Umeda
1
, Shogo Ikeda
3
, Katsuyoshi Mihara
2
,
Naotaka Hamasaki
1
and Dongchon Kang
1
1
Department of Clinical Chemistry and Laboratory Medicine, Kyushu University Graduate School of Medical Sciences, Fukuoka,
Japan;
2
Department of Molecular Biology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan;
3
Department of Biochemistry, Faculty of Science, Okayama University of Science, Japan
Mammalian mitochondria contain strong nuclease activity.
Endonuclease G (endoG), which predominantly resides in
mitochondria, accounts for a large part of this nuclease
activity. It has been proposed to act as an RNase H-like
nuclease on RNAÆDNA hybrids (R-loops) in the D-loop

with template strands around conserved sequence blocks
(CSBs) during transcription, forming R-loops consisting of
two DNA strands and one RNA strand, and serve as
primers for mtDNA replication (Fig. 1A). EndoG cleaves
the RNA of a linear RNAÆDNA duplex preferentially in the
CSB region [3], raising the possibility that endoG can
generate RNA primers for mtDNA replication [3]. How-
ever, endoG is not a specific RNase. On the other hand,
RNase MRP
1
, which is also thought to provide RNA
primers by cleaving the RNA of R-loops, is a specific
RNase. In addition, the endogenous RNAÆDNA hybrid
is formed in supercoiled mtDNA and should be a
triple-stranded R-loop [4]. The cleavage of the RNA of
triple-stranded R-loops by endoG has never been shown,
while RNase MRP has been shown to cleave the RNA of
triple-stranded R-loops preferentially at the CSBs [5].
Furthermore, NUC1, which is a yeast homolog of endoG
and is also found in mitochondria, is not essential for
mtDNA replication in yeast, as disruption of the gene leads
to no obvious derangement of the metabolism of mtDNA
[6]. Thus the role of endoG in mtDNA replication is still
ambiguous.
EndoG has recently been reported to be an apoptotic
nuclease [7]. It translocates to the nucleus upon apoptotic
stimulus and extensively degrades nuclear DNA, suggesting
that, in mitochondria, it has the potential to fully digest
mtDNA. This raises another issue of how mtDNA escapes
extensive digestion by endoG under steady-state conditions.

M
guanidine, 0.5
M
NaCl, 50 m
M
Tris/HCl, pH 8.0, and 1 m
M
dithiothreitol. The solubilized protein was bound to Ni
2+
-
chelating Sepharose resin (Amersham Pharmacia Biotech).
The resin was washed with buffer consisting of 20 m
M
Tris/
HCl, pH 7.5, 0.5
M
NaCl, 5 m
M
imidazole, 0.1% Triton
X-100, 2 m
M
2-mercaptoethanol, 10% glycerol, and 6
M
urea. The denatured His-endoG was renatured on the resin
by sequentially reducing the urea in the buffer from 6
M
to
0
M
, in 10 steps. The renatured His-endoG was finally eluted

polymerase instead of mtRNA polymerase, the human
mitochondrial D-loop region lacking authentic promoters
for the light and heavy strands was inserted downstream of
the SP6 promoter in pGEMhmD. R-loops were reconsti-
tuted as described previously [8]. Briefly, a reaction mixture
containing 5 n
M
pGEMhmD, 50 m
M
KCl, 20 m
M
Tris/
HCl, pH 8.0, 10 m
M
MgCl
2
,1m
M
dithiothreitol, 0.1 m
M
NTPs, and 0.2 UÆlL
)1
SP6 RNA polymerase was incubated
for 30 min at 37 °C essentially as described by Lee &
Clayton [9]. To remove NTPs, the reaction mixture was
applied to a gel-filtration spin column. The R-loops in the
eluate were ethanol-precipitated, dried, and resolubilzed in
distilled water. R-loop resolution with RNase H was
performed in 20 lL buffer containing 0.1 pmol R-loops,
20 m

indicated concentration of endoG for 10 min at 37 °C. The
reaction was stopped by the addition of 1 lgÆmL
)1
prote-
inase K and 0.5% SDS, and then incubated for another
10 min. R-loops were analyzed by 0.7% agarose gel
electrophoresis in buffer consisting of 89 m
M
Tris base,
89 m
M
boric acid, and 2 m
M
EDTA.
Determination of cleavage sites
After 1 lg R-loops or pGEMhmD had been treated with
50 ng endoG for 10 min, DNA was ethanol-precipitated,
washed with 70% ethanol, dried, and solubilized in distilled
water. To determine the 5¢ ends of cleavage sites, one cycle
of primer extension reactions was performed using
5¢-fluorescein isothiocyanate (FITC)-labeled primers [FD7
(FITC-ctacgttcaatattacaggcg) and FpGEM (FITC-
ctttatgcttccggctcgtatg) for the heavy and light strands,
respectively]. DNA was denatured for 5 min at 95 °C, the
primer was annealed for 0.5 min at 55 °C, and then an
extension reaction was performed for 1 min at 72 °Cusing
LA Taq polymerase (Takara, Seta, Japan). For sequence
ladders, 25 cycles of primer extension reactions were
similarly performed, but using a Thermo Sequence Core
Sequencing kit (Amersham Pharmacia Biotech) and 0.5 lg

ively. Antisera against rat Tim23, Tim43, and Tom20 were
as previously described [10,11].
Rat liver mitochondria were prepared as previously
described [10]. For preparation of soluble and particulate
fractions, intact rat mitochondria (1 mg proteinÆmL
)1
)
were disrupted by sonication in low-salt isotonic buffer
(10 m
M
Hepes/KOH, pH 7.4, 0.22
M
sucrose, and 0.07
M
mannitol), with 0.1
M
Na
2
CO
3
added before sonication in
some experiments. Next, 2.0
M
NaCl and 1.0% Triton
X-100 were added, and the samples were centrifuged at
100 000 g for 30 min and separated into pellets and
supernatants. The pellets were suspended in the same
volume of buffer, and each fraction was solubilized with an
equal volume of SDS sample buffer. The outer membranes
were disrupted by hypotonic treatment (10 m

R-loop mimics the endogenous RNAÆDNA hybrid better
thandoesalinearRNAÆDNA duplex. R-loops were
linearized at concentrations at which endoG converted
ordinary supercoiled plasmids into open circular forms
(Fig. 1B, lanes 2 and 3 in upper and lower panels). At higher
concentrations, R-loops were eventually degraded into
small pieces. These results suggest that endoG directly
produced double-stranded breaks in the R-loop, instead of
introducing nicks. Subsequently, the linearized DNA was
further digested. Consistent with this, the linear form
appeared as early as 1 min after the addition of endoG
(Fig. 2, lower panel). At this time point, the ordinary
supercoiled plasmid was hardly converted into a relaxed
form at all (Fig. 2, upper panel). It is noteworthy that a
closed circular form was never observed during endoG
treatment of R-loops over the time course of the study
(Fig. 2. lower panel). If RNA was first cleaved with
RNase H activity of endoG, the R-loops would have
reverted to closed circular plasmids as they were with
RNase H (Fig. 2, lane 2 in lower panel). This suggests that
cleavage of DNA is a primary event.
Fig. 2. Time course of endoG digestion. Supercoiled pGEMhmD
plasmids (upper) or R-loops (lower) were incubated with 50 ng endoG
for the indicated periods (lanes 4–10). Supercoiled pGEMhmD plas-
mids were incubated without endoG (lanes 2 and 3 in upper panel).
R-loops (lane 1 in lower panel) were treated with RNase H (lane 2 in
lower panel).
Fig. 3. Cleavage sites of R-loops. (A) Diagram of pGEMhmD. (B)
R-loops after cleavage by 50 ng endoG (lane 2), and after further
cleavage with ScaI(lane3)orNaeI(lane4).

and 8) of template DNA, respectively. Signals marked with * may be nonspecific artifacts caused by stalling of the extension reaction. (B) Light
strand. The reactions were performed as in (A), except the primer FpGEM was used. Except for the sequence lanes (lanes 6–9), each sample
corresponds to 1.0 lg template DNA (lanes 1–5).
5768 T. Ohsato et al.(Eur. J. Biochem. 269) Ó FEBS 2002
Because these signals were observed even in nondigested
plasmids (Fig. 4A, lanes 1 and 2), the DNA polymerase
used in the primer extension reaction may tend to pause
around the sites, and the signals may be technical artifacts.
We also saw several signals for the light strand when
nondigested R-loops were used (Fig. 4B, lane 3). Because
such signals were not observed when normal pGEMhmD
was used (Fig. 4B, lane 2), primer extension reactions
would frequently and artificially terminate only when using
R-loops. Although the exact reason for the termination is
not clear, considering that R-loops are fairly stable to heat
[9] and the template light strand was originally hybridized
with RNA, one possibility is that RNAÆDNA hybrids that
remain even after denaturation by heat may block the
extension reaction.
Submitochondrial localization of endoG
Next we examined the submitochondrial localization of
endoG. When rat liver mitochondria were disrupted by
sonication in low-salt isotonic buffer, endoG was mainly
recovered from a particulate fraction (Fig. 5A, lanes 7 and
8). On the other hand, it was released into the soluble
fraction in the presence of 2.0
M
NaCl, 0.1
M
Na

digested. This indicates that the protease has access to the
outside of the inner membranes but not the matrix side.
Under these conditions, both cytochrome c and endoG
were digested (Fig. 5C, lane 3). Both mHSP70 and Tim44
were completely digested with proteinase K when mito-
chondria were solubilized with 0.5% Triton X-100 (results
not shown). Thus, endoG and mtDNA are localized to
different compartments, which may protect mtDNA from
extensive digestion by endoG.
DISCUSSION
EndoG is essentially a nonspecific nuclease [1,2], which is
evident in vivo after apoptotic stimulation, when endoG
Fig. 5. Submitochondrial localization of endoG. (A) Intact rat mitochondria (lane 9) were disrupted by sonication in low-salt isotonic buffer (lanes 7
and 8). Then 2.0
M
NaCl (lanes 5 and 6) and 1.0% Triton X-100 (lanes 1 and 2) were added. In some preparations, 0.1
M
Na
2
CO
3
was added before
sonication and there was no further treatment (lanes 3 and 4). The samples were centrifuged and separated into pellets (P) and supernatants (S).
Samples corresponding to 20 lg mitochondrial protein were applied to each lane. EndoG was detected by Western blotting. (B) Intact mito-
chondria were incubated without proteinase K (lane 1) or with proteinase K (lane 2) for 1 h at 4 °C. Outer membranes were disrupted by hypotonic
treatment. The mitoplasts were then incubated without proteinase K (lane 3) or with proteinase K (lane 4) in the low-salt isotonic buffer for 1 h at
4 °C. Tom20, endoG, cytochrome c, Tim44, and mHSP70 were detected by Western blotting. (C) Intact mitochondria were treated with digitonin
for 10 min in the low-salt isotonic buffer at 4 °C followed by proteinase K digestion (lanes 1–3). mHSP70, Tim44, Tim23, endoG, and cytochrome c
were detected by Western blotting.
Ó FEBS 2002 Endonuclease G in mitochondria (Eur. J. Biochem. 269) 5769

regions [14]. These structures block transcription and
replication. For instance, it has been reported that
expansion of the GAA triplet repeat of the frataxin
gene results in triplex structures, thereby causing tran-
scription to be paused [15]. R-loop formation can occur
in a nonspecific manner during transcription [16] and can
serve as a primer for DNA synthesis [17], although the
nascent RNA molecule is normally displaced from the
DNA template strand shortly after synthesis. Accumula-
tion of R-loops is hazardous to cells, because of
induction of unregulated replication [18]. In addition,
guanine-rich RNA can participate in very stable three-
stranded structures by establishing both Watson–Crick
and Hoogsteen-type hydrogen bonds [19], the formation
of which inhibits transcription. Given that endoG is
normally regulated to exist in nuclei at a very low level,
it may have a role in survival in normal states
by eliminating kinked DNA formed in guanine-rich
regions.
EndoG has been proposed to provide primers for
mtDNA replication by cleaving the RNA moiety of
RNAÆDNA hybrids formed at the origin of replication [3].
However, based on our observations, endoG would not
act as an RNase in these conditions, but would instead
preferentially degrade mtDNA in an R-loop. Further-
more, we have shown that endoG exists in the intermem-
brane space (Fig. 5B,C), whereas mtDNA replication
should take place in the matrix. On the other hand,
RNase MRP has been shown to cleave RNA of triple-
stranded R-loops preferentially at the CSBs [5], and is

mitochondrial DNA is resolved by RecG, a Holliday junction-
specific helicase. Biochem. Biophys. Res. Commun. 255,1–5.
9. Lee, D.Y. & Clayton, D.A. (1996) Properties of primer RNA–
DNA hybrid at the mouse mitochondrial DNA leading origin of
replication. J. Biol. Chem. 271, 24262–24269.
10. Ishihara, N. & Mihara, K. (1998) Identification of the protein
import components of the rat mitochondrial inner membrane,
rTIM17, rTIM23, and rTIM44. J. Biochem. (Tokyo) 123,722–
732.
11. Iwahashi, J., Yamazaki, S., Komiya, T., Nomura, N., Nishikawa,
S., Endo, T. & Mihara, K. (1997) Analysis of the functional
domain of the rat liver mitochondrial import receptor Tom20.
J. Biol. Chem. 272, 18467–18472.
12. Ruiz-Carrillo, A. & Renaud, J. (1987) Endonuclease G: a (dG) n X
(dC) n-specific DNase from higher eukaryotes. EMBO J. 6, 401–407.
13. Ikeda, S. & Ozaki, K. (1997) Action of mitochondrial
endonuclease G on DNA damaged by
L
-ascorbic acid, peplomy-
cin, and cis-diamminedichloroplatinum (II). Biochem. Biophys.
Res. Commun. 235, 291–294.
14. Gilbert, D.E. & Feigon, J. (1999) Multistranded DNA structures.
CurR Opinion Sruct. Biol. 9, 305–314.
15. LeProust, E.M., Pearson, C.E., Sinden, R.R. & Gao, X. (2000)
Unexpected formation of parallel duplex in GAA and TTC
trinucleotide repeats of Friedreich’s ataxia. J. Mol. Biol. 302,
1063–1080.
16. Masse, E. & Drolet, M. (1999) Escherichia coli DNA topoisome-
rase I inhibits R-loop formation by relaxing transcription-induced
negative supercoiling. J. Biol. Chem. 274, 16659–16664.