ABSTRACT
We analyzed the structure and replication of the mitochondrial (mt) circular DNA
plasmid mp1 (1309 bp) from the higher plant
Chenopodium album
(L.). Two dimensional gel electrophoresis (2DE) revealed the existence of
oligomers of up to a decamer in addition to the prevailing monomeric form. The
migration behavior of cut replication intermediates during 2DE was consistent
with a rolling circle (RC) type of replication. We detected entirely single-stranded (ss) plasmid copies hybridizing only with one of the two DNA
strands. This result indicates the occurence of an asymmetric RC replication
mechanism. mp1 has, with respect to its replication, some unique features
compared with bacterial RC plasmids. We identified and localized a strand-specific nicking site (origin of RC replication) on the plasmid by primer
extension studies. Nicks in the plasmid were found to occur at any one of six
nucleotides (TAAG/GG) around position 735 of the leading strand. This sequence
shows no homology to origin motifs from known bacterial RC replicons. mp1 is
the first described RC plasmid in a higher plant.
The mitochondria of numerous groups of eukaryotic organisms, such as fungi and
plants, harbor several extrachromosomal elements in addition to the genomic DNA
(reviewed in
1
). By far the largest number and strongest diversity among these mitochondrial
(mt) plasmids have been described for higher plants (reviewed in
2
,
3
). These plasmids were classified into three categories: circular DNA plasmids,
linear DNA plasmids and RNA plasmids. The circular DNA plasmids are very small
and lack homology to known genes. Their origin remains a matter of debate. A
few mt plasmids were reported to share homology with sequences in the nucleus (
4
) or with parts of the chloroplast genome (
5
). Plasmids can be lost without phenotypic consequences to the plant, possibly
with one exception, a 2.3 kb DNA molecule from maize was reported to bear a
tRNA gene (
6
). Almost nothing is known so far about replication of these molecules.
Most of our knowledge about replication of circular plasmids was obtained from
bacteria (
7
,
8
). Two modes of replication have been described for these molecules. According
to the characteristic structures of replication intermediates, these modes were
conventionally named [theta] and [sigma]. During the [theta] mode, which is used by most of the plasmids, the sites
of priming of leading and lagging strand synthesis are located close to one
another within the origin of replication (
7
-
11
). Elongation of DNA synthesis can proceed either unidirectionally or
bidirectionally to dimers of the replicon.
In the case of the [sigma] or rolling circle (RC) mode of replication, priming events for
replication of the two strands are unlinked, occuring at different origins
(reviewed in
12
-
16
). In the first step, the plasmid-encoded nicking/closing protein introduces a strand- and site-specific nick in the so-called double-stranded (ds) replication origin (
dso
) (
17
-
19
). The free 3'-OH end generated is then utilized as a primer for leading strand
replication. Usually, after one round of replication the nicking/closing enzyme
terminates strand displacement at its recognition sequence. Two full-sized products, a ds and a single-stranded (ss) circular molecule, are generated. However, the
production of long linear plasmid concatemers is well known from phage [lambda] replication (
20
) and has also been described for bacterial plasmids (
21
-
24
). In the latter case, [sigma]-type replication was found to be recombination dependent. The
second RC replication step is synthesis of the lagging strand. It is initiated
via oligonucleotide priming in a different plasmid region, the ss origin (
sso
) and is discontinous.
To date, plasmid replication in plant mitochondria has only been described for
two circular DNA molecules from
Vicia faba
(
25
) and one DNA circle from
Chenopodium album
(
26
-
28
). Electron microscopic analyses of linearized replicative intermediates of the
V.faba
plasmids indicated that replication originates at a specific origin and
proceeds in a unidirectional manner around the molecules via [theta]-shaped intermediates. More recently, we reported on the unusual
migratory behavior of the circular 1.3 kb mt plasmid mp1 from
C.album
(see map in Fig.
1
) during pulsed-field gel electrophoresis, showing additional linear molecules and signals retained in the well (
26
). This pattern was very similar to that observed for several mt plasmids from
fungi, for which an RC mechanism of replication was proposed (
29
,
30
). During further EM studies we detected [sigma]-shaped molecules of mp1 and other subgenomic circles (
27
). The structure of these molecules suggested that they could indeed represent
intermediates of an RC type of replication. An RC type of replication was also
indicated by the observation of ss copies of mp1. We have localized a
dso
around position 730 of the mp1 sequence (Fig.
1
) (
28
).
Mitochondria were isolated from suspension culture C.9.1. of
C.album
. Conditions of cultivation have been described previously (
34
). Cells were usually harvested 6 days after transfer into new medium during the
logarithmic growth phase. Mitochondria were isolated and lysed as described
recently (
27
,
28
). Total mtDNA, including the plasmid mp1 (1309 bp; Fig.
1
; EMBL accession no. X58911) was purified by RNase digestion, phenol/chloroform
extraction and ethanol precipitation (
35
).
The mt plasmid mp1 was cut with
Bam
HI, ligated into the
Bam
HI site of vector pGEM3zf(+) (Promega, Madison, WI) and cloned in
Escherichia coli
cells (
28
). Transformants were grown at 37oC for 3-5 h in LB medium supplemented with ampicillin (50 [mu]g/ml). The cells were harvested and recombinant plasmid DNA was
isolated according to a standard protocol (
35
).
This method was performed according to Brewer and Fangman (
36
-
38
) and used for replication studies with plasmid mp1. For this purpose, ~3 [mu]g mtDNA was cut with restriction endonucleases. To digest mp1, enzymes were selected which had only one cutting site
in the plasmid (see map of the plasmid in Fig.
1
). Restriction enzymes were purchased from Amersham-Buchler (Braunschweig, Germany). Cut and uncut samples were separated for 25 h at 1 V/cm in the first dimension in 0.4% agarose gels in 1* Tris-borate-EDTA (TBE) electrophoresis buffer without ethidium
bromide in a large electrophoresis chamber (model HRH; IBI, New Haven, CT). The
lanes were cut out (in the absence of UV light) and stained by ethidium
bromide. Separation in the second dimension was done in 1.5% agarose gels in 1* TBE with 0.3 [mu]g/ml ethidium bromide at 5 V/cm for 4 h at 90o orientation to the first dimension. All electrophoresis steps
were performed at 4oC.
After electrophoresis, the DNA was blotted by alkaline transfer to Zeta Probe GT
membranes according to the instructions of the supplier (BioRad, Richmond, VA).
The cloned plasmid mp1 (cut out of the vector) was used as a probe for
hybridization. Radioactive labeling of the plasmid DNA was performed with the
Rediprime kit and 1.85 MBq [[alpha]-
32
P]dCTP, provided by DuPont (Bad Homburg, Germany). For identification of ssDNA
forms of mp1, we prepared ss-specific RNA probes of the plasmid integrated in vector pGEM3zf(+) using
the MAXIscript
in vitro
transcription kit (Ambion Inc., Austin, TX). Filters were hybridized overnight in 6-8 ml 7% SDS, 250 mM NaH
2
PO
4
, pH 7.2, at 65oC in hybridization tubes from Schott (Mainz, Germany) and then washed under stringent conditions according to standard protocols (
35
). Quantification of hybridization signals was done with a GS-363 phosphorimager (BioRad).
For primer extension studies the sequence-specific primers 1 (5'-GCCATCTAAAACGAGCGACG-3'), 2 (5'-CCTTGTAAACATCCCCCCGA-3') and 3 (5'-GGGAGCACAACCGAGTAGCG-3') were 5'-end-labeled using the Ready-To-Go T4 polynucleotide kinase kit (Pharmacia Biotech, Uppsala, Sweden) and [[gamma]-
32
P]ATP (0.37 MBq; DuPont). Asymmetric PCR reactions with one of the primers were
performed in a 50 [mu]l volume including 0.2 mM dNTPs, 2 mM MgCl
2
in 1* PCR buffer in the presence of 0.5 [mu]g total mtDNA (harvested 1 or 6 days after transfer into new medium),
0.5 [mu]g open circular and [sigma]-like mp1 molecules (electroeluted from the respective zones of
an agarose gel as depicted in Fig.
2
c) or 0.5 [mu]g cloned mp1 in vector pGEM3zf(+). Thermostable Goldstar DNA polymerase was
purchased from Eurogentec (Seraing, Belgium). Cycling was done at 95oC for 30 s, 58oC for 30 s and 72oC for 1 min for 40 cycles. Extension products were resolved in
denaturing polyacrylamide gels (4.5 and 6% polyacrylamide, 7 M urea) in 1* TBE buffer. As a size marker, a 5'-end-labeled 10 bp ladder was used (Gibco BRL, MD).
Additionally, a sequencing reaction was done by the method of 3-dNTP internal label cycle sequencing according to the instructions of the
manufacturer (Amersham-Buchler) using [[alpha]-
35
S]dATP (1.85 MBq; DuPont) and primer 1. Template DNA was plasmid mp1, cloned in
the vector pGEM3zf(+). After electrophoresis the DNA in the gel was fixed by
incubation in 5% acetic acid. The gels were dried on a glass plate and exposed
to X-ray films (Amersham-Buchler).
In the last few years 2DE of DNA molecules has been developed into a powerful
tool for the detection of replication intermediates and for determining the
replication type (
36
-
47
). This technique takes advantage of the fact that DNA molecules are separated
according to their molecular mass in the first dimension and that a non-linear DNA molecule does not migrate at the same rate as a linear molecule
of equal mass in the second dimension, i.e. migration is additionally dependent
on the structure (
36
,
46
). Replicative DNA forms can be unequivocally distinguished from recombination
intermediates. Therefore, it should be possible to determine whether the [sigma]-like structures of plasmid mp1 recently observed by EM (
27
,
28
) represent replication or recombination intermediates.
In a first experiment, uncut mtDNA was separated in two dimensions as described
above, blotted by denaturing transfer and then hybridized with a leading strand-specific radioactively labeled RNA probe obtained by
in vitro
transcription of mp1 (Fig.
2
a). After exposure, the filter was stripped and reprobed with a lagging strand-specific RNA probe (Fig.
2
b). The patterns of hybridization signals obtained were completely identical
with probes for both strands, except a faint spot in the lower part of the gel.
The signals are explained schematically (Fig.
2
c) according to Brewer and Fangman (
36
-
38
). The strongest signals were always located at the position of the open
circular, linear and supercoiled forms of the monomer as well as at a curve
representing linear molecules starting from 1.3 up to 10-12 kb, which should represent oligomeric plasmid forms. At the position
of linear multimers, signals appeared over a strong background. This smear
stops exactly at the position of the monomer, i.e. there was no breakage of the
monomers during preparation. Open circular forms up to a 5mer could be
observed. Moreover, we detected a curve between the linear and circular
molecules which extended past the linear dimer. This signal most probably
represents plasmid molecules with a growing tail of up to 2-3 contour lengths of the corresponding circle, since this arc looks very
similar to that obtained from rolling circles on analysis of
in vitro
(
43
,
44
) and
in vivo
(
45
) replication in other systems. Therefore, this curve could represent the [sigma]-like mp1 molecules observed by EM studies (
27
,
28
). The observation that an arc only originates from the open circular monomer
spot suggests that monomeric forms are the predominant templates for plasmid replication. Bubble-like structures as known for [theta] replication were not found. Such molecules would form an arc
between the open circular forms of the monomer and dimer (
36
). Furthermore, in the lower part of the 2DE gel a faint spot appeared which
migrated in the second dimension faster than the supercoiled monomer.
Hybridization showed only signals with the leading strand-specific plasmid probe (Fig.
2
a) and not for the lagging strand (Fig.
2
b). Position and hybridization behavior are in agreement with the ss nature of
this molecule (
12
-
16
). Hence, this spot represents the ss circular form, more precisely the leading
strand, of the mp1 monomer. In addition to the ss monomer, a much weaker signal
appeared on an imaginary line of ss molecules which should represent the ss
circular dimer of mp1 (data not shown). Quantitative analysis of the plasmid
hybridization signals in Figure
2
a revealed the following distribution: 48% linear molecules, 42% circles, 6% [sigma]-like replication intermediates and 4% ssDNA molecules. In the total
fraction of plasmid DNA, monomers comprised only ~43% of the sequences.
In further experiments mtDNA was digested with restriction endonucleases that
linearize the circular plasmid mp1. The digested DNA samples were separated in
2DE gels, transferred to Nylon membranes and hybridized with the plasmid mp1
DNA as a probe (Fig.
3
a-c). The patterns of hybridization signals revealed different types of DNA
molecules, including replicative forms. The interpretation of these patterns
according to Brewer and Fangman (
36
-
38
) is depicted in Figure
3
d. A very strong signal was observed at the position of linearized monomers. A
much weaker but clearly visible spot occurred at the position of linearized
dimers. Other linear molecules were expected to migrate on a straight line
between these two spots. All hybridization signals above the line of linear
molecules represent forms of mp1 other than linear molecules. A continuous arc
of growing Y-shaped replication intermediates expanding from the linear monomer to
dimer was observed. A complete arc was seen only with DNAs digested by
Hin
dIII (Fig.
3
a) and
Bam
HI (data not shown). In the case of all other enzymes used (
Acc
I,
Bgl
I,
Cfr
I,
Fok
I,
Kpn
I,
Mlu
I,
Pst
I,
Pvu
II,
Sca
I and
Sma
I), this arc of simple Y molecules ended before reaching the dimer (shown for
Pst
I and
Bgl
I digests in Fig.
3
b and c). The patterns are not compatible with those obtained from intermediates
of [theta]-type replication, which would migrate in a much higher position (
36
-
38
). We never detected intermediates which could result from digestion of
molecules with replication bubbles (compare with the scheme in Fig.
3
d). The observed arcs of simple Y structures are best explained by [sigma]-like intermediates of replication initiating at a position near the
cutting sites of
Bam
HI and
Hin
dIII, but distant from the sites of those enzymes which did not lead to patterns
with complete arcs. In addition, we found weak signals at the position of so-called double Ys. The presence of such structures could be attributed
either to recombination intermediates (
36
-
38
,
46
,
47
), to fragments with replication bubbles at both ends cut by the enzyme (
36
-
38
), to circles with a tail exceeding the contour length of the circle (designated
as extended `E' arcs by Han and Stachow;
42
) which were not cut (e.g. because of stretches of ssDNA at the cutting site or
entirely ss tails respectively) and to circles with two tails, as detected in
EM analyses (
27
). Circles with two tails could be generated, for example, by a second
initiation of replication at the origin before completion of the first round of
replication or by the simultanous initiation of replication at more than one
origin (
28
).
The existence of ss mp1 copies (see above) of only one of the DNA double strands
is not compatible with a [theta] type of replication. Such intermediates should occur, however, during
RC replication, where the introduction of a site-specific nick in only one strand of the supercoiled template DNA
characterizes the initial event of replication (
12
-
16
). We identified such an origin of RC replication of mp1 by mapping short ds
fragments produced by cleaving the tails of [sigma]-like replication intermediates with restriction endonucleases (
28
). This is not a very precise approach, but showed that the origin is located
around position 730 of the physical map of mp1 (cf. Fig.
1
).
We have presented here several lines of evidence for replication of the
mitochondrial plasmid mp1 according to a rolling circle mechanism. This
includes detection of uncut [sigma]-shaped and cut Y-shaped replication intermediates by 2DE, observation of ss
copies of only one strand of the plasmid and identification of a nicking site
in the leading strand, a characteristic feature of RC replication. The
observation of a strand-specific nicking site and of circular ss forms of the same mp1 strand
support the idea of the activity of a nicking/closing enzyme (
17
-
19
) in the mitochondria of
C.album
. The localization of the
dso
on mp1 around nucleotide 735 by primer extension is in good agreement with
recent mapping of the origin around position 730 by a less precise approach (
28
) and is also compatible with the 2DE data. When mtDNA was cut by restriction
endonucleases which linearized mp1 and separated electrophoretically by 2DE, we
could detect continous arcs of hybridization signals only for
Bam
HI and
Hin
dIII. These cutting sites are clustered together around positions 750-800, i.e. not far from the
dso
. In the case of all other enzymes the arcs stopped before reaching the dimer,
indicating the absence of an origin close to these endonuclease recognition
sites. The circular DNA of mp1 is the first RC plasmid detected in a higher
plant. A model of the replication cycle of mp1 based on the data outlined here
and in recent reports (
27
,
28
) is depicted in Figure
5
. This plasmid shows some unique features in comparison with bacterial RC
plasmids.
(i) The replication of RC plasmids has been studied very extensively in bacteria
(
12
-
16
). These replicons replicate in a similar manner to a mechanism described for
ssDNA phages and accumulate ss plasmid copies during this process. In an early
study on
Bacillus subtilis
and
Staphylococcus aureus
it was shown that the ss plasmid DNA exists as a circular molecule of the same
size as the parental monomer and corresponds to only one of the two DNA strands
(
51
). It represented ~20% of the total hybridization signal. The
in vivo
occurence of one of the two possible ssDNA circles of mp1 in mtDNA preparations
from
C.album
is a strong indication for RC replication, the only known process producing ss
copies of ds molecules. In the case of asymmetric RC replication, ss copies of
only one of the DNA strands are to be expected. This is by definition the
leading strand of replication (
12
-
16
). The percentage of single-stranded plasmid mp1 copies (~4% of the hybridization signal) is lower than normally observed for
the bacterial RC plasmids described above. Notable exceptions, for example, are
plasmid pUB110 (
S.aureus
) and pBC16 (
Bacillus cereus
), which generated amounts of ssDNA in the same range as mp1 (
51
).
(ii) The organization of bacterial RC plasmids is highly conserved (
12
-
16
). The
dso
region is placed immediately upstream of a gene which encodes a nicking/closing
enzyme involved in the initiation and termination of leading strand synthesis.
This replication initiator protein binds to and introduces a strand- and site-specific nick in the leading strand of supercoiled DNA, providing a
free 3-OH end for elongation. The function of this protein can be substituted,
however, by nucleases which create random nicks in both plasmid strands,
leading to recombination-dependent replication (
21
-
24
). Like most of the described circular plasmids in plant mitochondria, mp1 does
not bear genetic information necessary for the function of the organelle (
52
). The sequence of mp1 contains small putative ORFs (Fig.
1
). Database alignments exhibited insignificant sequence homology of ORF2 to DNA
and RNA polymerase genes. However, this ORF would not be large enough to encode
a complete polymerase. No homolog of a gene encoding a complete replication
protein, including the nicking/closing enzyme on plasmid mp1, was found. Even
shorter conserved motifs, including tyrosine or serine residues, which are part
of the active center of the latter proteins, could not be detected. These facts
do not rule out the possibility that such a replication protein is encoded in
the nucleus or in chromosomal mtDNA.
(iii) DNA sequences proximal to the
dso
of mp1 (TAAGGG) show no homology to consensus motifs of bacterial
dso
s. A cluster of G residues (GGG) at the nicking site was not found in any RC
origin of bacterial plasmids or phages. An AT-rich sequence of 5-8 nt upstream of the nicking site, which is common to many RC
systems (
12
-
16
), is absent in the case of mp1. In comparison with bacterial RC plasmids, which
have a relatively low GC content (
16
), mp1 contains ~47.5% GC. Moreover, unlike the situation in bacteria, the nicking site of
mp1 is not represented by a single nucleotide. Our data from primer extension
experiments demonstrated that there is a nicking region of ~5 nt. The most prominent extension product occurs at position 734, which
indicates the most common nicking site to be TAAG/GG (behind position 735).
Degradation of DNA nicked at only one position would be an alternative
explanation for the observed multiple bands. However, since the cleavage sites
of restriction endonucleases were exactly determined by single bands (cf. Fig.
4
, lanes c), degradation seems not to occur under the applied conditions. In
previous mapping studies we obtained data suggesting the existence of an
additional, less often used origin on mp1 located between positions 510 and 560
(
28
). Interestingly, the
dso
sequence TAAGGG is also found at position 540. During the present study the
introduction of nicks at this position could not be observed, which is likely
due to its rare usage. In addition, we have found no evidence for termination
of plasmid mp1 replication at a second
dso
with subsequent re-initiation, which would result in greater size diversity of ss and ds
plasmid copies.
(iv) Many of the [sigma]-like mtDNA molecules of
C.album
were found to have tails several times longer than the circumference of the
corresponding circle, suggesting the synthesis of concatemeric replication
products (
27
,
28
), as known from classical phage [lambda] replication (
20
). The products of replication of several classes of RC plasmids from Gram-positive and Gram-negative bacteria are monomers (
12
-
16
), whereas long linear concatemers are produced during recombination-dependent RC replication (
21
-
24
). In contrast to the situation with mp1, where replication is initiated at one
or two distinct origins, recombination-dependent replication in bacteria is mostly not initiated at specific
origins (reviewed in
22
,
24
). In the case of mp1, monomers represented ~50% of the total plasmid DNA. mp1 also exists in linear and circular
oligomeric forms which may be products of recombination events and/or represent
concatemeric products of RC replication (Fig.
5
, pathway b).
In conclusion, our data revealed new features of an RC plasmid. This is the
first report of an organellar plasmid in plants replicating via an RC
mechanism. Its high copy number could make mp1 an interesting and promising
model system for further studies of the replication and structural organization
of chromosomal mtDNA in higher plants, because [sigma]-like molecules and entirely ss circles were also found for the
chromosomal mtDNA in
C.album
(
28
). This study may also provide clues for the explanation of a common phenomenon
of plant mitochondria, the occurence of a heterogenous population of linear
molecules (
26
,
27
), which could also arise by a rolling circle mechanism of replication in these
organelles.
We thank Ken Kreuzer and Karyn Belanger (Duke University, NC) for helpful
discussions and Brent Nielsen (Auburn University, AL) for his support and
critical reading of the manuscript. This work was supported by grants from the
BMBF, Bonn, and the Fond der Chemischen Industrie, Frankfurt, to T.B.
REFERENCES
Return