(Received for publication, December 18, 1996)
From the Département de Microbiologie et immunologie, Université de Montréal, C.P. 6128, Succursale Centre-Ville, Montréal, Québec H3C 3J7, Canada
Recent in vivo and in vitro studies have suggested an important role for DNA topoisomerases in regulating R-loop formation during transcription in Escherichia coli. In the present report we present genetic and biochemical evidence strongly suggesting that R-loop formation can occur during transcription of a portion of the rrnB operon and that it is regulated by DNA topoisomerase activity. We found that a multicopy plasmid (pBR322) carrying an heavily transcribed portion of the rrnB operon cannot be transformed in topA mutants unless RNase H is overproduced. Transcription of the 567-base pair HindIII fragment from the rrnB operon allows the extraction of large amount of R-looped plasmid DNAs from a topA mutant, in a manner that depends on the intracellular level of RNase H activity. When DNA gyrase is sufficiently active, hypernegatively supercoiled plasmid DNA is produced if the same DNA fragment is transcribed in a topA mutant. The formation of such topoisomers most likely reflect the presence of extensive R-loops since it is sensitive to the intracellular level of RNase H activity. Finally, the formation of R-looped plasmid DNAs in an in vitro transcription system using phage RNA polymerases is also detected when the 567-base pair HindIII fragment is transcribed on a negatively supercoiled DNA template.
During the process of transcription, the nascent RNA molecule is normally displaced from the DNA template strand to be translated (mRNAs) or to participate in the process of translation (stable RNAs: tRNAs and rRNAs). Indeed, early results have suggested that Escherichia coli RNA polymerase has a function allowing proper RNA displacement from the template strand (1). However, in some cases the RNA has to remain associated with the template strand to serve as a primer for DNA replication initiation. The best studied example in this regard is the replication origin of ColE1 plasmid DNA (2, 3). In addition, several observations suggest that R-loop formation can occur in a rather nonspecific manner if DNA gyrase, the enzyme responsible for the introduction of negative supercoiling into the DNA of bacteria, is present during transcription. In in vitro replication systems for oriC containing plasmid DNAs and ColE1 plasmid derivatives and where RNA polymerase and DNA gyrase are present, RNase H and E. coli topoisomerase I, an enzyme that specifically relaxes negatively supercoiled DNA, are required to maintain the specificity in the process of initiation at oriC or at the ColE1 origin of replication, respectively (4-6). This result is best explained by the introduction of negative supercoiling by DNA gyrase providing the driving force for R-loop formation at sites other than the normal origins of replication. The annealed RNAs can be used as primers to initiate replication at these sites. RNase H, an endoribonuclease degrading the RNA moiety of an RNA-DNA hybrid, eliminates these primers, whereas E. coli topoisomerase I by counteracting DNA gyrase activity, prevents their formation or destabilizes them. Recent in vitro experiments have shown that the production of hypernegatively supercoiled DNA during transcription in the presence of DNA gyrase was due to R-loop formation (7, 8). Many plasmid DNAs tested in this system were shown to become hypernegatively supercoiled during transcription in the presence of DNA gyrase but in the absence of RNase H, suggesting that extensive R-loop formation was a rather general phenomenon in the presence of DNA gyrase (7, 8). Moreover, E. coli DNA topoisomerase I was shown to efficiently abolish the formation of such supercoiled DNA during transcription, and hence R-looping as well. A model for the regulation of R-loop formation was therefore proposed in which DNA gyrase, with its supercoiling activity, favors the formation of R-loops, while DNA topoisomerase I, by relaxing negative supercoiling, inhibits the formation of such structures. In support of this model, it has been recently demonstrated that an R-loop can be a hot spot for DNA relaxation by E. coli DNA topoisomerase I (8).
To study the putative role of DNA topoisomerase I in the regulation of R-loop formation in vivo, it was necessary to construct new topA mutants. The reason is that to survive, topA mutants need to acquire compensatory mutations that most often lie in a gyr gene and as a result, lower DNA gyrase activity. Moreover, some of these strains such as the widely used DM800 with the gyrB225 mutation, has one or more additional compensatory mutation(s) (9). Since these mutations arise to compensate for the absence of DNA topoisomerase I, they will most likely mask some of the effects associated with the loss of topA gene function. Various topA mutants with a conditional compensatory phenotype were thus constructed (10). These strains have a gyrB(Ts) mutation and as a result, they can grow at 37 °C because gyrase activity is then low enough to compensate for the loss of topA. At 30 °C, they basically do not grow because gyrase regains a more wild-type level of activity that can no longer compensate for the absence of DNA topoisomerase I. Such strains are therefore cold-sensitive and are also temperature-sensitive (Ts) because of the presence of the gyrB(Ts) allele. By exposing these strains to the non-permissive cold temperature it is thus possible to study the true effects of loosing topA on cell physiology. Since RNase H overproduction allowed these mutants to grow at low temperature, it was concluded that R-loop formation was a major problem for the cell in the absence of DNA topoisomerase I. Moreover, a combination of topA and rnhA (encoding for RNase HI) mutations was found to be non-viable and the gyrB(Ts) mutation was shown to correct some phenotypes associated with an rnhA mutation at 37 °C. It was thus suggested that gyr mutations arise in topA mutants to reduce R-loop formation (10). These in vivo data are therefore in perfect agreement with the in vitro results and suggest an important role for DNA topoisomerases, and hence DNA supercoiling, in the regulation of R-loop formation. In view of the important roles of these enzymes in the regulation of DNA functions, it is thus interesting to consider the R-loop structure as a putative regulator of various genomic activities. However, this model needs to be supported by data showing that R-loop formation indeed occurs during transcription of certain genes in E. coli and that it is regulated by DNA topoisomerase activity. In the present report we present genetic and biochemical evidence that R-loop formation can occur during transcription of an rrn operon both in vivo and in vitro and that it is influenced by DNA topoisomerase activity.
E. coli strains used are listed and described in Table I. The detailed protocol used to construct the DM800 derivatives by P1vir transduction will be presented elsewhere. The RFM480 cold-sensitive strain is well described in the Introduction and in Ref. 10.
|
pNO1302 is a pBR322 derivative containing a
non-functional rrnB operon with its complete regulatory
region including P1 and P2 promoters (11; Fig. 1). It was constructed
by removing the 2451- and 53-bp1 adjacent
SalI fragments of the rrnB operon of pNO1301.
This deletion removed the 3 end of the 16 S rRNA gene, the spacer
region, and the 5
end of the 23 S rRNA gene. pEM501 was obtained by
deleting the 451-bp BstBI-StuI fragment of
pNO1302 (nucleotides 1113 to 1563 according to Ref. 12) to remove the
regulatory region of rrnB including both promoters. pEM520
was constructed by deleting the 567- and 3180-bp HindIII DNA
fragments of pNO1302, removing all the rrnB coding sequence
downstream of the first HindIII site of the rrnB
operon, and the 346-bp BamHI-HindIII fragment of
pBR322. pTrc99a, obtained from Pharmacia, is an expression vector
derived from pBR322 that carries a multiple cloning site downstream of a Shine-Dalgarno sequence and the IPTG-inducible Ptrc promoter.
This vector also carries the lac Iq gene
allowing good transcriptional repression of Ptrc, in the absence of IPTG. pMD210 is a derivative of pTrc 99a in which the Shine-Dalgarno sequence has been deleted by Bal31 nuclease
digestion from NcoI ends, as described (13). This deletion
removed nucleotides 251 to 289 of the pTrc 99a vector. pMD217 (Fig. 2)
was constructed by inserting a synthetic boxA sequence (14)
with SalI (5
) and HindIII (3
) cohesive ends
into the same sites of pMD210. This synthetic boxA sequence
was obtained by annealing to one another the two oligonucleotides with
the following nucleotide sequence: 5
-TCGACACTGCTCTTTAACAATTTA and
5
-AGCTTAAATTGTTAAAGAGCAGTG. pMD306 and pMD308 are derivatives of
pMD217 in which the 567-bp HindIII fragment from the
rrnB operon on pNO1302 was inserted into the HindIII site (nucleotides 1598 to 2165 according to Ref.
12), respectively, in its original and its reverse orientation relative to Ptrc (Fig. 2). A region from pMD306 and pMD308 including
the 5
end of the HindIII rrnB fragment, the
boxA element, and the Ptrc promoter was sequenced
and was found to be as expected (data not shown). pEM001 and pEM003
were constructed by first subcloning the
SalI-EcoRI fragment respectively from pSK760 (15)
carrying the rnhA gene and pSK762c (16) carrying a mutated
rnhA gene, into the same sites of the pBluescript KS vector
(from Stratagene). The Xba I-HincII fragment from
the respective resulting plasmid was then subcloned within the same
sites of pACYC184 to give pEM001 and pEM003. pJP459 and pJP461 were
obtained by cloning the 567-bp HindIII fragment of
rrnB into the same site of the pBluescript KS vector,
respectively, in the reverse and the physiological orientation relative
to the vector map numbering.
Media and Growth Conditions
Unless otherwise indicated,
bacteria were grown in LB media at the temperature indicated in the
table and figure legends. When needed, antibiotics were added as
follows: ampicillin at 50 µg/ml, and chloramphenicol at 30 µg/ml.
Because of the acrA13 mutation in the DM800 derivatives that
renders these cells more permeable to many antibiotics (9),
chloramphenicol was used at 10 µg/ml, for these cells.
Isopropyl--D-thiogalactoside (IPTG) and
5-bromo-4-chloro-3-indolyl
-D-glucoside were obtained
from Sigma.
The molecular biology techniques were performed essentially as described (13). CaCl2 transformations were done as described in Drolet et al. (10).
Electrophoresis in the Presence of ChloroquineOne-dimensional and two-dimensional agarose gel electrophoresis in the presence of chloroquine were performed essentially as described (7), except that 0.5 × TBE (13) was used instead of 0.5 × TPE. Chloroquine was used at the concentration indicated in the figures. After electrophoresis, agarose gels from one-dimensional electrophoresis were stained with ethidium bromide and photographed under UV light. Agarose gels from two-dimensional electrophoresis were dried and prepared for in situ Southern hybridization as described (17).
Extraction of R-looped Plasmid DNAsTo examine if plasmid DNAs extracted from various strains contained R-loops, a clear lysate protocol was used (18). Bacterial cells carrying pMD306 or pMD308 DNAs were grown at the indicated temperature to an OD600 of 0.4 at which time they were split evenly (10 ml) into two tubes. IPTG at 1 mM final was then added to one tube and both tubes were incubated for an additional 75 min. Cells were transferred in a tube half full of ice and pelleted and resuspended in 200 µl of an ice-cold solution containing 50 mM Tris (pH 8.0) and 25% sucrose. Forty µl of an ice-cold solution made of 0.25 M Tris (pH 8.0) and lysozyme at 5 mg/ml were added and the cells were incubated on ice for 5 min. After the addition of 80 µl of ice-cold 0.25 M EDTA (pH 8.0) and a 5 min incubation on ice, 320 µl of 2% Triton X-100 in water were added. The sample was incubated on ice for 10 min with periodic gentle agitation and then centrifuged for 30 min at 4 °C in a microcentrifuge. The globular pellet (containing chromosomal DNA and cell debris) was removed and SDS at 0.5% and proteinase K at 0.5 mg/ml were added to the supernatant. After a 30-min incubation at 37 °C, the sample was extracted once with phenol, once with chloroform, and the nucleic acids were precipitated with ethanol. The DNA pellet was resuspended in 100 µl of TE buffer (13) and the sample was extracted once with phenol and twice with chloroform before being precipitated with ethanol and resuspended in 40 µl of TE buffer. Two aliquots of 10 µl were treated with 250 ng of RNase A in a solution made of 10 mM Tris (pH 8.0), 20 mM MgCl2, and 100 mM NaCl. Twenty ng of E. coli RNase HI (from Robert J. Crouch, National Institutes of Health) were added to one tube and both aliquots were incubated at 37 °C for 1 h. After one phenol extraction, the samples were analyzed by agarose gel electrophoresis (1%) in 1 × TAE (13).
In Vitro Transciption ReactionsIn vitro transcription reactions for the detection of R-loop formation were performed with T3 and T7 RNA polymerases (6 units) purchased from Stratagene and Life Technologies, Inc. A 25-µl reaction mixture containing 40 mM Tris-HCl (pH 8.0), 8 mM MgCl2, 25 mM NaCl, 2 mM spermidine, 30 mM dithiothreitol, 400 µM each of NTP, and 0.33 µg of plasmid DNA with a native superhelical density, was prepared in a tube on ice. When indicated, 5 ng of RNase H and/or 1 µg of RNase A were added during transcription. After a 20-min incubation at 37 °C, the reaction was stopped by 0.2% SDS final and extracted with a mixture of phenol-chloroform. The samples were then incubated at 37 °C for 1 h in the presence of 250 ng of RNase A and, when indicated, 20 ng of RNase H. The samples were analyzed by agarose gel electrophoresis (1%) in 1 × TBE (13).
Our in vivo results strongly suggested that R-loop formation was a major problem in the absence of DNA topoisomerase I but did not indicate anything relating to the specificity, the frequency, or the consequences of R-loop formation (10). For several reasons we considered that R-loop formation might occur during transcription of the ribosomal RNA genes (rrn operons). First, since they are heavily transcribed (they contribute to more than half of the transcriptional activity in an exponentially growing culture of E. coli); they are statistically more prone to R-loop formation. Second, since rRNAs are not translated they are free to hybridize with the template strand, as opposed to mRNAs that are bound by ribosomes. Third, we observed that the growth of our cold-sensitive topA mutants was very sensitive to nutritional shift-ups unless RNase H was overproduced (10). Sensitivity of rnhA mutants to nutritional shift-ups has also been reported (19). Under nutritional shift-up conditions, the cells have to face a sudden demand for a large amount of ribosomes and therefore, for rRNAs at the very beginning. A recent report demonstrates that the seven chromosomal copies of the rrn operon confer growth advantage over cells that have two rrn operons deleted, if cells are exposed to nutritional shift-ups (20). These results suggest that it is during nutritional shift-ups that the rrn operons experiment the maximum rate of transcription and are thus more prone to R-loop formation. Fourth, we found that double nusB5-topA and nusB5-rnhA mutants are barely viable.2 One of the reported phenotype of the nusB5 mutant is a partial inhibition of rrn operons transcription elongation (21). However, these mutants are still able to make the same number of ribosomes as wild-type cells, apparently by increasing the frequency of rrn operon transcription initiation, in agreement with the feedback regulation model of ribosome synthesis (11).
We reasoned that, if R-loop formation does occur during rRNA synthesis,
increasing the total amount of cellular rrn transcription by
introducing a multicopy plasmid carrying an rrn operon, it could result in growth inhibition of topA mutants, because
of a possible depletion of the negative regulators of R-loop formation (e.g. RNase H). In the beginning we used the plasmid pNO1302
(Fig. 1), a derivative of the multicopy plasmid pBR322,
that was previously used in experiments to support the feedback
regulation model of ribosome synthesis (11). pNO1302 contains a deleted
rrnB operon with an intact promoter region but a
SalI deletion removing the 3 end of the 16 S rRNA gene, the
spacer region, and the 5
end of the 23 S rRNA gene (11). Since the
rrnB operon carried by this multicopy plasmid is
non-functional, it will not be subject to feedback regulation and, as a
result, a cell carrying this plasmid will produce much more
untranslated RNAs.
We thus transformed this pBR322 derivative in various bacterial strains. As can be seen in Table II, pNO1302 can transform all topA+ strains whether they overproduce RNase H (pEM001) or not (pEM003). However, pNO1302 could only be transformed in topA null mutants providing that RNase H was overproduced. This was true for all topA mutants tested, including the widely used DM800 strain (Table II). The fact that a pNO1302 derivative with a deletion removing the two rrnB promoters (pEM501) is able to transform topA null mutants irrespective of the cellular amount of RNase H suggests that this is due to heavy transcription from these very strong rrnB promoters. Moreover, the fact that a pNO1302 derivative carrying both rrnB promoters but in which almost all the rrnB coding sequence is deleted (pEM520) is able to transform topA mutants not overproducing RNase H, indicates that the problem is related to transcription of the rrnB operon. We also analyzed, by restriction enzyme digestions and electrophoresis, the plasmid DNA extracted from several RFM480 cells (topA::Tn10) transformed with pNO1302. Although intact pNO1302 was always found from RFM480 cells overproducing RNase H, the very few RFM480 transformants not overproducing RNase H were found to contain low-molecular weight pNO1302 derivatives, in which almost the entire rrnB operon was deleted (data not shown).
|
Our results suggested that R-loop formation was taking place during transcription of the rrnB operon on pNO1302, but did not indicate if specific rrnB sequences were involved. In addition, since the evidence we obtained supporting R-loop formation was based on growth phenotypes and therefore being rather indirect, we decided to address this question in a more direct manner by looking at the presence of R-loops on our plasmid constructs, extracted from the cold-sensitive topA mutant. The assay is based on the fact that the presence of R-loops on plasmids will alter their migration in an agarose gel, mainly due to DNA relaxation, since DNA unwinding must occur to allow RNA-DNA hybrid formation. This DNA relaxation caused by the presence of R-loops will produce conformers, not topoisomers, since the DNA linking number is unchanged. Disappearance of the conformers upon RNase H treatment will be taken as a strong evidence for the presence of R-loops.
We thus designed a system allowing the subcloning of
rrnB DNA fragments under the control of the inducible
Ptrc promoter. In this vector, the Shine-Dalgarno sequence
following the inducible promoter has been deleted, but a synthetic
boxA sequence has been added (Fig. 2,
pMD217). Because our preliminary results indicated that the NusB
protein that acts in the process of antitermination by interacting with
a boxA element (reviewed in Ref. 22) seems to have an effect
on R-loop formation,3 we considered the
possibility that a boxA sequence could be essential for the
formation of such structures. The strong -independent rrnB T1 T2 terminator sequence is also present in this
vector, in such a way that transcription from Ptrc will be
terminated whether or not a boxA sequence is present. In
addition, since transcription from Ptrc on these vectors is
tightly regulated owing to the presence of the
lacIq gene, it is therefore possible to clone
rrnB sequences and to introduce the resulting plasmid DNAs
in our topA mutants. Our work first began with the
subcloning of the 567-bp HindIII DNA fragment from the
16 S gene of the rrnB operon, previously used to
study the antitermination system of rrn operons (23, 24). This HindIII fragment was thus subcloned in both
orientations into the HindIII site of pMD217 to produce
pMD306 and pMD308, respectively, carrying this HindIII
fragment in the original (or physiological) and the reverse
orientation, relative to the IPTG-inducible Ptrc promoter
(Fig. 2). In the beginning we observed that the growth on LB plates of
our topA mutants not overproducing RNase H and carrying
pMD306 or pMD308 was inhibited upon IPTG induction and when a selective
pressure was kept for the maintenance of the plasmid DNAs
(+ampicillin). We also observed that this growth inhibition was more
severe for topA mutants carrying pMD306 as compared with the
same mutants carrying pMD308. These observations suggested to us that
R-loop formation was occurring during transcription of this
HindIII fragment, mainly in its original orientation, and
that such R-loops were interfering with the maintenance of the plasmid
DNAs.
Plasmid DNAs were extracted as described under "Experimental
Procedures," from RFM480 carrying either pMD306 or pMD308 and grown
in the presence or not of 1 mM IPTG at 35 °C. As can be seen in Fig. 3, a large amount of conformers that
disappear upon RNase H treatment (compare lane 1 with
lane 2) is only visible for pMD306 extracted from cells
grown in the presence of 1 mM IPTG (compare lane
1 with lane 5). We can also observe that the migration
of almost all plasmid DNAs is retarded suggesting that nearly 100% of
these pMD306 DNAs contains R-loops (lane 1). Basically no
conformers are observed for pMD308 extracted from cells grown in the
presence of 1 mM IPTG (compare lane 3, pMD308,
with lane 1, pMD306). These results show that under our
experimental conditions, a significant amount of R-loops is only
detected when the rrnB HindIII fragment is transcribed (IPTG
induction) in its physiological orientation (pMD306).
Early results have shown that R-loop formation could be induced by the
use of protein denaturants such as SDS and phenol, when a negatively
supercoiled DNA template was used in in vitro transcription
experiments (1). Although we considered unlikely this possibility
especially because the formation of conformers is dependent on the DNA
fragment orientation (pMD306 versus pMD308) and also because
the detrimental effect of heavily transcribed rrnB sequences
on the growth of topA mutants is corrected by RNase H
overproduction, we performed a simple experiment to address this
question. One prediction that could be made if R-loop formation was an
artifact of extraction is that RNase H overproduction in vivo should not have any effect on the formation of conformers. On
the contrary, if R-loop formation really occurs in vivo,
RNase H overproduction should abolish or decrease the production of such conformers. The results presented in Fig. 4 clearly
show that RNase H overproduction in vivo strongly decreased
the production of conformers and hence R-loop formation (compare
lane 1, RNase H overproduction with lane 3, no
RNase H overproduction). This result demonstrates that at least the
formation of a large fraction of the conformers is due to R-loop
formation in vivo. A net decrease in the formation of
conformers is also observed in an in vitro transcription
system when the same DNA fragment is transcribed in the presence of
RNase H (see below).
Transcription of the 567-bp rrnB HindIII Fragment Allows the Extraction of Hypernegatively Supercoiled Plasmid DNA from a topA Mutant When DNA Gyrase Is Sufficiently Active
Previous in
vitro studies have shown that transcription in the presence of DNA
gyrase can lead to extensive R-loop formation and the subsequent
generation of hypernegatively supercoiled plasmid DNAs upon R-loops
removal (7, 8). In one of this study, transcription of the 567-bp
rrnB HindIII fragment with phage RNA polymerases and in the
presence of DNA gyrase, was shown to generate such hypernegatively
supercoiled DNA and hence very long R-loops (8). We can therefore
predict that extensive R-loop formation during transcription of this
rrnB HindIII fragment in a topA mutant should
generate hypernegatively supercoiled DNA, if DNA gyrase is sufficiently
active. Our initial experiments failed to detect a significant amount
of such topoisomers when pMD306 DNA was extracted from our
cold-sensitive topA mutant grown at 35 °C, following IPTG
induction (data not shown). This is most likely due to the fact that
DNA gyrase, owing to the gyrB(Ts) allele, is too weak at
35 °C to promote extensive R-loop elongation. We therefore performed
an experiment in which the topA mutant cells were grown at
35 °C before being incubated at 30 °C for 75 min, to activate DNA
gyrase. Plasmid DNAs, pMD306 or pMD308, were then extracted as
described under "Experimental Procedures." The result shown in Fig.
5A suggest that extensive R-loop formation
occurred when the 567-bp HindIII fragment was transcribed
(+IPTG) in its physiological orientation, since the migration of almost
100% of the extracted pMD306 DNAs was retarded during electrophoresis
(compare lane 1, no RNase H, with lane 2, RNase
H). Significant R-loop formation also occurred when the rrnB
HindIII fragment was transcribed (+IPTG) in its reverse
orientation since the migration of a large proportion of the extracted
pMD308 DNAs was retarded during electrophoresis (compare lane
3, no RNase H, with lane 4, RNase H). However, it is
also obvious that under our experimental conditions, a larger amount of
R-loops is detected when the rrnB HindIII fragment is transcribed in its physiological orientation (compare lane
1, pMD306 with lane 3, pMD308).
We next loaded an aliquot of the samples from the experiments shown in Fig. 5A on an agarose gel with chloroquine, to detect the presence of hypernegatively supercoiled plasmid DNAs. Before loading on such gel, the DNA samples were treated with RNase H to reveal the supercoiling level after electrophoresis and ethidium bromide staining. As can be seen in Fig. 5B hypernegatively supercoiled plasmid DNA is produced following transcription of the 567-bp rrnB HindIII fragment in both orientations (lanes 1 and 2, [- -]). Despite the fact that more R-loops are detected on pMD306 than pMD308 (Fig. 5A compare lane 1 with lane 3), roughly the same amount of hypernegatively topoisomers are detected for both plasmids. This can best be explained by the fact that transcription of the rrnB HindIII fragment in the physiological orientation produce R-loops that are more stable than the ones generated in the reverse orientation (8). This stability can be manifested either in vivo or during the extraction of the plasmid DNAs and the following treatment with RNase A before electrophoresis (see "Discussion").
One prediction that can be made if the formation of hypernegatively
supercoiled plasmid DNAs in topA mutants is directly linked to R-loop formation, is that overproducing RNase H should reduce the
formation of such topoisomers. To test this hypothesis, we introduced
the plasmid pEM001 to overproduce RNase H into the cold-sensitive
topA mutant carrying pMD306. As a control, pEM003, carrying
a mutated rnhA gene, was used instead of pEM001. The cells
were grown with or without IPTG, as described for the experiment shown
in Fig. 5 and the plasmid DNAs were extracted by the alkaline lysis
protocol (13). The presence of hypernegatively supercoiled plasmid DNAs
was detected by two-dimensional agarose gel electrophoresis in the
presence of chloroquine. As can be seen in Fig. 6, a
decrease in the amount of hypernegatively supercoiled pMD306 DNA
(indicated by an arrow) is observed when RNase H is
overproduced (compare + IPTG, pEM001 with + IPTG, pEM003). Indeed,
densitometry analysis of the two-dimensional gel shown in Fig. 6,
indicates that the proportion of such topoisomers decrease by a factor
of about 3 when RNase H is overproduced (data not shown). These results
suggest that the generation of an important proportion of
hypernegatively supercoiled pMD306 DNA in the cold-sensitive
topA mutant, is linked to R-loop formation. Similar results
were also obtained when the same experiment was performed with pMD308
instead of pMD306 (data not shown). It is, however, obvious that RNase
H overproduction in vivo do not have the same impact on the
production of hypernegatively supercoiled DNA as it has in in
vitro reactions (8). This can be explained by the fact that other
factors in vivo, for example, single-stranded binding
protein and histone-like proteins, may protect hypernegatively
supercoiled DNA from relaxation.
Transcription in Vitro of the 567-bp rrnB HindIII Fragment Allows the Production of R-looped Plasmid DNA
To further support
our in vivo results and to investigate in greater detail the
process of R-loop formation, we designed an in vitro
transcription system using phage RNA polymerases. These polymerases are
highly promoter-specific and will therefore only transcribe DNA
sequences placed downstream of an appropriate promoter. We thus
subcloned the 567-bp HindIII fragment from rrnB
in both orientation into the pBluescript II KS vector in such a way
that it is flanked by promoters for phage T3 and T7 RNA polymerases. In
the pJP459 plasmid the HindIII fragment is cloned in the
reverse orientation relative to the pBluescript KS map and will thus be transcribed in this orientation by the T7 RNA polymerase and in its
physiological orientation by the T3 RNA polymerase. In the pJP461
plasmid the HindIII fragment is inverted and will therefore be transcribed in the physiological orientation by the T7 RNA polymerase and in its reverse orientation by the T3 RNA polymerase. As
can be seen in Fig. 7, a larger amount of conformers is
produced when pJP459 is transcribed by T3 RNA polymerase as compared
with transcription by T7 RNA polymerase (compare lane 1 with
lane 11), therefore when the HindIII
rrnB fragment is transcribed in its physiological
orientation. This is not an RNA polymerase effect since transcription
of pJP461 produces more conformers with T7 RNA polymerase than with T3
RNA polymerase (compare lane 6 with lane 16).
Since such conformers are either barely visible or not visible at all
after RNase H treatment during or after transcription, they are most
likely due to the presence of R-loops (lanes 3, 4, 8, 9, 13, 14, 18, and 19). All these results are therefore in good
agreement with the in vivo data showing stable R-loop formation when the rrnB HindIII fragment is transcribed in
its physiological orientation.
Interestingly, the presence of RNase A at 1 µg had almost no effect on R-loop formation, suggesting that the RNA in the R-loop is never free and that it is elongated by RNA polymerase without being displaced from the template strand (compare lane 1 with lane 2 and lane 16 with lane 17). A similar conclusion was reached from the results of in vitro transcription experiments with pJP459 and pJP461 DNAs, in the presence of DNA gyrase (8). Moreover, increasing the amount of RNase A to 2.5 µg or adding RNase T1 that has a different ribonucleotide specificity, had no effect on R-loop formation (data not shown). In addition, these results with RNase A and T1 together with the fact that RNase H during transcription strongly decreased the production of conformers (lanes 3, 8, 13, and 18), eliminate the possibility that the production of such structures is an artifact of extraction due to the use of protein denaturants. We also observed that the addition of RNase H during transcription stimulated RNA synthesis, suggesting that an R-loop can act as a road block for RNA polymerases (data not shown).
Since our in vivo results suggested that the level of negative supercoiling was an important factor for extensive R-loop formation on the rrnB HindIII fragment, we decided to look for R-loop formation during transcription of relaxed pJP459 and pJP461 DNAs. No RNase H-dependent changes in the electrophoretic mobility patterns were observed for any plasmids and polymerases used, suggesting that stable R-loops did not form on relaxed DNAs (data not shown).
In the present study we have provided genetic and biochemical evidence that R-loop formation can occur during transcription of the rrnB operon in E. coli. We first showed that heavy transcription of a portion of the rrnB operon was detrimental for the growth of all topA mutants tested unless RNase H was overproduced. We then demonstrated the formation of conformers when the 567-bp rrnB HindIII fragment was transcribed on a plasmid (pMD306) and in a topA mutant, suggesting the presence of R-loops on the plasmid DNAs. Indeed, the disappearance of such conformers upon RNase H treatment confirmed the presence of R-loops. Following these results it was important to show that these R-loops were generated during transcription in vivo and therefore, that their formation was not induced during plasmid DNAs extraction. Two sets of experiments strongly suggested that these R-loops were generated during in vivo transcription. First, we found that overproducing RNase H in vivo considerably decreased the amount of conformers produced. Second, when DNA gyrase was sufficiently active we were able to detect the presence of hypernegatively supercoiled plasmid DNAs, a predictable event if extensive R-loop formation occurs. This was indeed supported by the finding that the formation of such topoisomers was sensitive to the intracellular level of RNase H. The in vivo evidence for R-loop formation during transcription of the rrnB HindIII fragment is also supported by the results of in vitro transcription experiments: the formation of RNase H-sensitive conformers (this study) and the formation of hypernegatively supercoiled DNA in the presence of DNA gyrase and in a manner sensitive to the level of RNase H activity (8).
A larger amount of conformers, and therefore R-loops, was always detected when the rrnB HindIII fragment was transcribed, either in vivo or in vitro, in its physiological orientation as compared with its reverse orientation. Similar results were recently obtained by measuring the R-loop-dependent linking number change following the addition of DNA gyrase either during or after transcription (8). The results of such experiments allowed us to conclude that roughly the same amount of R-loops were generated in both orientations under the experimental conditions used, but that the R-loops generated in the physiological orientation were more stable than the ones generated in the reverse orientation. This conclusion regarding the stability was based on the sensitivity of the R-loops to RNase A, a ribonuclease specific to single-stranded RNA. By considering this stability factor, the absence of conformers when the rrnB HindIII fragment was transcribed in its reverse orientation (pMD308) at 35 °C may be explained by a combination of in vitro and in vivo factors. It is possible that the unstable R-loops were lost during the extraction of plasmid DNAs and/or that their RNA was degraded by RNase A before loading the samples on the agarose gels. It is predictable and it has indeed been shown that R-loops are generally less stable on relaxed DNA than on negatively supercoiled DNA (1). In this context one can imagine that the absence of conformers on pMD308 extracted from cells grown at 35 °C is due to the fact that the negative supercoiling level was too low (lower than in a wild-type cell; data not shown), whereas the presence of such conformers when the same plasmid was extracted from cells grown at 30 °C is explained by a higher negative supercoiling level which can stabilize the R-loops. Because the R-loops that are produced when the rrnB HindIII fragment is transcribed in its physiological orientation tend to be more stable, they will be less affected by the supercoiling level. This would explain why large amounts of conformers were detected on pMD306 extracted from cells grown at both temperatures (30 and 35 °C). By considering the supercoiling factor as a key element in R-loop formation one can explain why a mutation reducing the level of negative supercoiling, such as a mutation in gyrA or gyrB, will be the best mutation to compensate for the absence of DNA topoisomerase I, since such a mutation will prevent stable R-loop formation. Indeed, R-loop formation was not detected when the rrnB HindIII fragment was transcribed in its physiological orientation on a relaxed DNA template in vitro. Moreover, by exposing the cold-sensitive topA mutant to a temperature higher than 37.5 °C, no R-loops were detected on pMD306 DNA which was also found to be significantly more relaxed than DNA extracted from wild-type cells (data not shown). Therefore, as concluded from the results of in vitro experiments (8), our in vivo results allow us to suggest that DNA gyrase participates both in R-loop initiation and R-loop elongation by providing the negative superhelical density needed for these processes.
Our results also suggest that the nucleotide sequence of the DNA template can contribute to the stability of the R-loop. What sequence element is responsible for the higher stability of the R-loop generated during transcription of the HindIII rrnB fragment in its physiological orientation, is unknown for the moment. Interestingly, it has been shown that a RNA-DNA hybrid with a purine RNA is more stable than a hybrid with a pyrimidine RNA (25). In addition, recent results of in vitro experiments have shown that long polypurine stretches can favor the formation of very stable RNA-DNA hybrids if transcription occurs in the direction that makes purine-rich RNA (26, 27). We note that the RNA made when the rrnB HindIII fragment is transcribed in its physiological orientation is more purine-rich than the RNA made when it is transcribed in its reverse orientation. Whether it is significant or not remains to be determined. It is also important to mention that transcription of other DNA fragments from the rrnB operon and from other sources was recently found to have a detrimental effect on the growth of topA mutants in a manner that depends on the intracellular level of RNase H.4 We therefore believe that the major factor influencing R-loop formation in general is the supercoiling level of the DNA template but that the nucleotide sequence of the DNA template is also an important factor.
Interestingly, in all other in vitro systems where R-loop formation has been demonstrated, the primary role (probably the only one) of the transcribed RNA is to form an hybrid with the template DNA strand. Obviously, this is not the case in the present study since the role of the rrnB operon is to produce rRNAs and tRNAs. In the other described systems RNA-DNA hybrids are formed either to create RNA primers to initiate DNA replication (2, 28), or possibly to be used as a recombination intermediate in the process of immunoglobulin class switching (26, 29). Therefore, a major question is whether or not R-loop formation at rrnB is biologically significant. To be biologically meaningful, R-loop formation during transcription of the rrnB operon must occur in wild-type cells. We have found that pNO1302 DNA is not easily maintained in wild-type cells unless RNase H is overproduced. In addition, pNO1302 but not the pEM501 derivative without rrnB promoters, was found to contain R-loops when extracted from wild-type cells.4 These results suggest that even in the presence of negative regulators such as DNA topoisomerase I and RNase H, R-loop formation can occur when rrnB transcription increases considerably. This is an indication that the system regulating R-loop formation can be saturated in wild-type cells and therefore, that such structures may form in these cells. Moreover, to be biologically relevant, R-loop formation does not have to be frequent especially when one considers the possible involvement of such structures in initiation of DNA replication and DNA recombination. Interestingly, in the case of the cold-sensitive topA mutants, we repeatedly observe a 10-20-fold increase in the gene duplication rate between different rrn operons as compared with the rate in an isogenic gyrB(Ts) strain.5 Whether R-loop formation during transcription of rrn operons is directly or indirectly involved in gene duplication remains to be tested. It is worth mentioning that the absence of either type I DNA topoisomerase (Top1 or Top3) in the yeast Saccharomyces cerevisiae results in increased recombination in the rDNA multiple tandem array (30-32). The possible involvement of R-loop formation in this process has not yet been investigated.
We thank Dr. R. L. Gourse for pNO1302 plasmid, and Dr. R. J. Crouch for purified RNase HI from E. coli. We also thank Josée Prévost for technical assistance and Sonia Broccoli for careful reading of the manuscript.