(Received for publication, January 31, 1997)
From the Department of Microbiology, Biomedical Center, Uppsala University, Box 581 S-751 23 Uppsala, Sweden, § Department of Microbiology, Uppsala Genetic Center, SLU (Swedish University of Agricultural Sciences), Box 7025, Genetikvägen 5, S-75007 Uppsala, Sweden, and ¶ UPR 9002 du CNRS, Institut de Biologie Moléculaire et Cellulaire, 15 rue R. Descartes, Strasbourg cedex, France
The replication frequency of plasmid R1 is
controlled by an antisense RNA (CopA) that binds to its target site
(CopT) in the leader region of repA mRNA and inhibits
the synthesis of the replication initiator protein RepA. Previous
studies on CopA-CopT pairing in vitro revealed the
existence of a primary loop-loop interaction (kissing complex) that is
subsequently converted to an almost irreversible duplex. However, the
structure of more stable binding intermediates that lead to the
formation of a complete duplex was speculative. Here, we investigated
the interaction between CopA and CopT by using Pb(II)-induced
cleavages. The kissing complex was studied using a truncated antisense
RNA (CopI) that is unable to form a full duplex with CopT. Furthermore,
RNase III, which is known to process the CopA-CopT complex in
vivo, was used to detect the existence of a full duplex.
Our data indicate that the formation of a full CopA-CopT duplex
appears to be a very slow process in vitro. Unexpectedly,
we found that the loop-loop interaction persists in the predominant
CopA-CopT complex and is stabilized by intermolecular base pairing
involving the 5-proximal 30 nucleotides of CopA and the complementary
region of CopT. This almost irreversible complex suffices to inhibit
ribosome binding at the tap ribosome binding site and may
be the inhibitory complex in vivo.
Natural antisense RNAs control a variety of biological functions,
mainly in prokaryotic accessory genetic elements such as phages,
transposons, and plasmids (1). Most antisense RNAs found regulate the
plasmid copy number. The efficiency of antisense RNA control depends on
the rate constant for the binding between antisense and target RNA and
on the intracellular concentration of the inhibitor RNA. The binding
pathways were extensively studied in ColE1 (2-4) and R1 (5-7)
plasmids. In both systems, the binding process is seen as a series of
reactions leading to a stable duplex. The initial, rate-limiting step
involves the formation of a transient loop-loop interaction (kissing
complex) that is followed by a subsequent nucleation step via the
single-stranded 5 segment of the antisense RNA. This interaction is
believed to result in the formation of a double-stranded RNA along the
entire length of the target site. However, the full duplex may not be
required for inhibition since the transient kissing complex appears to be sufficient for control in ColE1 (8), R1 (9), and IncB (10)
plasmids.
Plasmid R1 belongs to the IncFII group whose initiation frequency is
controlled by an antisense RNA, CopA. CopA inhibits RepA synthesis by
binding to the leader region of the repA mRNA (CopT) 80 nucleotides upstream of the repA translational start codon (11). RepA synthesis requires the translation of a short leader peptide
(Tap) encoded immediately downstream of the CopA target site,
overlapping the repA reading frame by two nucleotides (12). This implies that CopA inhibits repA translation indirectly
by preventing tap translation in vivo (13).
Extensive mutational analyses revealed that the structures of both the
antisense and the target RNAs are critical for obtaining an optimal
binding rate and in vivo control (1). However, although the
role of loop-loop interactions in the binding pathway of the antisense RNA and its target has been characterized (7-9), the structure(s) of
more stable binding intermediate(s) that lead to a complete duplex
remained speculative. In the present study, the interaction between
CopA and CopT was followed by using Pb(II)-induced cleavages. The
loop-loop interaction was also probed using a truncated antisense RNA
(CopI) that is unable to form a full duplex with CopT (6). Furthermore,
RNase III that processes the CopA-CopT complex (14) was used to
distinguish between antisense-target RNA complexes differing in the
extent of duplex formation. Our results indicate that the stable
CopA-CopT complex is not fully base-paired in the region of
complementarity. Remarkably, an extended loop-loop interaction persists
in the CopA-CopT complex, maintaining the structures of the major CopA
and CopT stem loops. This kissing complex is additionally stabilized by
an intermolecular helix involving the 5-proximal 30 nucleotides of the
antisense RNA and its counterpart in CopT. The significance of this
CopA-CopT complex for the regulatory mechanism is discussed.
CopT, CopA, and CopI were
synthesized with T7 RNA polymerase from polymerase chain
reaction-generated DNA fragments as described by Hjalt and Wagner (15).
Transcription of CopT yields a run-off product of 302 nucleotides that
carries two G residues at its 5 end instead of the GU sequence of the
wild type repA mRNA. The CopA RNA contains a 5
-terminal
G instead of an A residue. 5
-End labeling of dephosphorylated RNA was
performed with T4 polynucleotide kinase and [
-32P]ATP
(16). Labeled RNAs were purified by polyacrylamide-urea gel
electrophoresis, eluted, and precipitated twice with ethanol. Before
use, the different RNAs were dissolved in RNase-free water and
renatured by incubation at 90 °C for 2 min followed by slow cooling
at room temperature in TMN buffer (20 mM Tris acetate, pH
7.5, 10 mM magnesium acetate, 100 mM sodium
acetate).
Experimental conditions were
adapted from Hjalt and Wagner (15). Complex formation was carried out
at 37 °C for 5 or 15 min in TMN buffer with 5-end-labeled CopT
(3 × 10
8 M) and a 5-fold excess of
unlabeled CopA or CopI (1.5 × 10
7 M) or
with 5
-end-labeled CopA or CopI (4 × 10
8
M) and a 5-fold excess of unlabeled CopT (2 × 10
7 M). Full duplexes between 5
-end-labeled
CopT and CopA or CopI were formed by incubation at 90 °C for 2 min
followed by slow cooling to 37 °C in TMN buffer. Reaction mixtures
(10 µl) contained end-labeled RNA (20,000 cpm) either free or in
complex, carrier tRNA (2.5 µg), and lead(II) acetate at a final
concentration of 8 or 16 mM in TMN buffer. Incubation was
done at 37 °C for 5 min, and the reactions were stopped by the
addition of 5 µl of 0.1 M EDTA followed by ethanol
precipitation. Pellets were dissolved in 10 µl of stop buffer (93%
formamide, 30 mM EDTA, 0.05% xylene cyanol, 0.05%
bromphenol blue, 0.5% SDS). Before electrophoresis on 15%
polyacrylamide-urea 8 M gels, the samples were heated at 90 °C for 3 min.
5-End-labeled CopT was incubated with
a 5-fold excess of CopA or CopI (10
7 M) in 10 µl of TMN buffer containing 1 mM dithiothreitol at
37 °C for 5, 15, or 30 min. Full RNA duplexes were formed as
described above. Cleavage of free or complexed CopT was at 37 °C
from 25 s to 10 min with 0.002 µg of RNase III or from 1 to 10 min with 0.25 µg of RNase III (a gift from D. Court). All reactions
contained 10 µg of carrier tRNA. Reactions were stopped by phenol
extraction followed by RNA precipitation. Pellets were dissolved in 10 µl of stop buffer and analyzed as described above.
The primary binding intermediate involving the loop-loop
interaction was studied by using the truncated antisense RNA, CopI, which is unable to promote a full duplex with CopT (6). In both free
RNAs, the major cleavages occur in single-stranded regions, mainly in
the hairpin loops and the interhelical regions (Fig. 1).
The cleavage positions are shown schematically on the secondary structures of the RNAs (Fig. 2). The cleavages in loop
II of CopI (but also CopA) and in loop II of CopT extend into the
helical regions, which is probably due to the presence of several
bulged nucleotides (15). Under native conditions (kissing complex), CopI binding induces significant changes in the Pb(II) cleavage pattern
of CopT (Fig. 1A, lanes 5 and 6).
Strong protection is observed on the 3
side of the upper stem II
at
U117 to A123 and to a lesser extent at
U102 to C107. Furthermore, increased reactivity
induced by CopI binding occurs at U109 and
G112. The Pb(II) sensitivity of the non-complementary
regions is not altered, indicating that CopI binding induces structural
changes restricted to the upper part of stem loop II
of CopT (Fig.
1A). When full duplexes are formed, complete resistance to
Pb(II) hydrolysis is observed in the region encompassing
U94 to C138 (Fig. 1A, lanes
8 and 9). Therefore, Pb(II) is a sensitive probe that
allows discrimination between a full duplex and the kissing complex.
Conversely, binding of CopT induces enhanced Pb(II) cleavages at
A29 and, to a lesser extent, at C28,
A30, and A31 in CopI (Fig. 1A). Most
of the nucleotides located in loop II of CopI are still cleaved by
Pb(II) upon CopT binding, indicating that the upper part of helix II
participates in the formation of the intermolecular helix that
characterizes the kissing complex. Significant protection on both sides
of the upper helix of CopI is observed, especially on the 5
side of
helix II at U33 to A39 (Fig.
1B).
Pb(II) Probing Shows That a Full Duplex Is Not the Major Product of CopA-CopT Binding
A CopA-CopT complex in which either CopA or
CopT was labeled at the 5 end was formed under native conditions at
37 °C and subjected to Pb(II) probing. Under these conditions,
stable complexes were formed within 5 min at 37 °C (data not shown).
Unexpectedly, CopA binding induces the same reactivity changes in stem
loop II
of CopT as CopI (cf. lanes 5 and 6 with
lanes 11 and 12 in Fig. 1A). In
particular, the bulged cytosine 101 and part of the external loop II
of CopT remain accessible to Pb(II). Additional protections occur in
the 140-162 region (Fig. 2). No major structural rearrangement of CopT
RNA occurs in the non-complementary region since the Pb(II) sensitivity
remained unchanged (Fig. 1A). In contrast, the formation of
a full CopA-CopT duplex entails complete protection of the CopA binding
site from nucleotide 81 to 162 of CopT (Fig. 1A, lanes
14 and 15). Conversely, CopT binding induces the same
reactivity changes in stem loop II of CopI and CopA (see Fig. 1,
B and C). This is clearly illustrated by a strong
enhancement at A60 and weaker increase at C59,
A61, and A62 in CopA. In addition, the
5
-proximal 30 nucleotides of CopA are protected from Pb(II) in the
presence of CopT (Fig. 2).
Altogether, these results suggest that a full duplex is not the major
product of the CopA-CopT complex in vitro. The extended loop-loop contact, as found in the CopI-CopT complex, persists in the
CopA-CopT complex, but this complex is additionally stabilized by the
formation of an intermolecular helix that involves the 5-proximal 30 nucleotides of CopA and the complementary region of CopT.
RNase III was
shown to process the CopA-CopT complex in vivo (9).
Therefore, we reinvestigated RNase III-dependent cleavages under conditions of full duplex formation and native CopA-CopT complex
formation to define the extent of duplex formation. The cleavage
positions were mapped using 5-end-labeled CopT RNA (Fig. 3A). The cleavage sites are indicated on a
schematic drawing (Fig. 3B). When CopA or CopI was heated
and slowly cooled in the presence of labeled CopT, allowing the
formation of full RNA duplexes, the same concentration of RNase III was
able to fully digest both duplexes. The strongest cleavages were
located at A93, U94, U99, and
U100 (Fig. 3A). Conversely, with the native
CopA-CopT complex, RNase III produced two major cleavages in CopT at
U149 and C154 (Fig. 3A). This result
is in agreement with the primary RNase III cleavage identified at
C154 in vivo and in vitro (14).
However, weak cleavages corresponding to the presence of the full
duplex were also found in CopT bound to CopA, especially after a
prolonged time of incubation (at least up to 15 min, Fig.
1A). Under the same conditions, the CopI-CopT complex formed
at 37 °C was completely resistant to RNase III hydrolysis (data not
shown). Therefore, these experiments indicate that the full duplex is
not the major product of CopA-CopT complex in vitro and that
its formation appears to be a slow process in vitro.
Previous studies suggested that the binding of CopA to CopT
proceeds in at least two experimentally distinguishable steps: kissing
and hybridization duplex formation (5-7). The first, rate-limiting
step is the formation of a transient loop-loop interaction (a so-called
kissing complex) that was supposed to permit the subsequent formation
of a complete RNA duplex. The present structural probing performed on
the kissing complex shows that most of the nucleotides in loop II of
the truncated antisense RNA CopI, as well as in the complementary loop,
are accessible to Pb(II). However, since point mutations within the
CGCC sequence in loop II of CopA strongly affect the pairing rate and
result in copy number-up phenotypes (5, 11, 17), these nucleotides are
inferred to be part of the primary kissing complex (Fig.
4B). This intermediate, not detectable by the
present probing approach, rapidly rearranges to an extended kissing
complex in which the 5 side of the upper part of stem II interacts in
an interstrand pairing with CopT (Fig. 4C). This is further
supported by RNase V1 cuts in the upper stem II of CopI and increased
accessibility of the 3
side of loop II toward RNase T2 (6). Notably,
the formation of the extended kissing complex induces a strong Pb(II)
cleavage in the loop II of the antisense RNA. This cleavage may be due
to the presence of a magnesium binding site. In line with this view, Gregorian and Crothers (18) showed that the uptake of at least two
magnesium ions stabilized the loop-loop interaction in the RNAI-RNAII
stem loop complex of plasmid ColE1.
The Pb(II) probing experiment also demonstrates that the extended
kissing complex persists in the CopA-CopT complex. Furthermore, extensive protection in the 5 segment of CopA and the complementary region of CopT supports the formation of 30-34 interstrand base pairs.
We also showed that RNase III, which processes the CopA-CopT complex
in vivo (9), cleaves the native CopA-CopT complex and the
full duplex, respectively, in completely different ways. The major
RNase III cleavage site in CopT bound to CopA under native conditions
is centered within the 30-base pair intermolecular helix (Fig.
3B), which is consistent with known site preferences of
RNase III (19). The extended kissing complex CopT-CopI that is not
processed in vivo (9) is also resistant to RNase III hydrolysis in vitro. The RNase III experiments indicate that
only a minor fraction of the CopA-CopT complex proceeds to a full
duplex in vitro. This suggests that the asymmetric extended
kissing complex induces a structural bend that brings together the two
complementary single-stranded segments neighboring the hairpins II and
II
to facilitate their rapid pairing (Fig. 4D). Since
previous studies showed that stable CopA-CopT complexes were formed
very rapidly (5), the identification of a stable and not fully
base-paired CopA-CopT complex is somewhat unexpected. The present work
indicates that this long lived intermediate (Fig. 4D)
predominates for at least 15 min under conditions in which kissing
complex formation occurs within a few seconds and that conversion to a
full duplex is very slow, as it was proposed in the ColE1 system
(4).
Similar conclusions for the binding pathway of the antisense RNA
involved in translational control of repA in IncB plasmids have been reported (10). The antisense RNAI primarily inhibits the
formation of a pseudoknot structure that activates repA
translation (20). Both CopA and RNAI contain stable hairpins with
identical hexanucleotide sequences within the loops. However, RNAI has
only a short 5 tail (9 nucleotides) and lacks a 5
stem loop
structure. Enzymatic probing indicated that RNAI and the target RNAII
of plasmid pMU720 do not rapidly form a full duplex in vitro
but rather form an extended kissing complex stabilized by additional base pairs involving the 5
tail of RNAI. In ColE1, NMR spectroscopy experiments showed that a kissing complex between two complementary stem loops of RNAI-RNAII involves base pairing of all loop residues (21). In this system, the stems of both hairpins are rather short (6 base pairs) and lack bulges. Comparing the different structures of the
RNAI-RNAII stem loop complex (ColE1) with that of CopI (or CopA)-CopT
(R1) and RNAI-RNAII (pMU720), it is tempting to speculate that bulges
near a recognition loop permit an induced rearrangement to obtain a
transient but stable extended kissing complex. Removal of the bulged
residues in stem II of CopA and stem II
of CopT strongly decreases
both binding rate and kissing complex stability (22).
The same major RNase III cleavage site was identified in
vitro and in vivo at C154 in CopT RNA (9).
Thus, the majority of the CopA-CopT complexes that inhibit
repA expression in vivo may not be fully
base-paired. The extent of complete pairing in vivo may be
underestimated due to RNase III processing followed by very rapid decay
of the RNAs. Nevertheless, we propose that the formation of the long
lived CopA-CopT complex detected in vitro is sufficient to
account for the observed translational control. We have recently shown
that this complex prevents initiation complex formation at the
tap ribosome binding site (23). Even the extended kissing
complex (CopI-CopT) was able to transiently interfere with ribosome
binding that correlates well with in vivo results (9). In
agreement with previous observations (24), the present data also show that the binding of CopI or CopA does not induce major conformational changes in the non-complementary sequences of CopT, i.e. in
the tap and repA ribosome binding site regions.
That Pb(II) is very sensitive to subtle conformational changes suggests
that the structure of the kissing complex itself, rather than an
induced RNA structure change downstream, prevents ribosome binding at
tap. In the wild type situation, the presence of the 5
segment of CopA stabilizes the transient kissing complex, providing
approximately three helical turns of double-stranded RNA immediately 5
of the tap Shine-Dalgarno sequence, which is sufficient to
completely and irreversibly block tap translation.
The finding that complete pairing between an antisense and a target RNA is slow and can be unnecessary for inhibition may illustrate a rule rather than an exception. Structurally complex recognition stem loops of antisense and target RNAs appear to be intrinsically difficult to unfold during complex formation, and hence many biological control systems may have evolved to employ binding intermediates active in the inhibitory step. An advantage of such a mechanism is that inhibition kinetics can be faster than pairing kinetics (e.g. see Ref. 25) since the involved stem loops appear optimized for binding rate (26, 27), whereas unfolding rates may be difficult to optimize (unless unwinding activities are present). We suggest that such considerations may be relevant for the design of efficient antisense RNA inhibitors.
We thank Eric Westhof and Hervé Moine for stimulating discussions and Steve Lodmell for critical reading of the manuscript.