ABSTRACT
A
trans
-acting system has been designed in order to explore the three-dimensional structure of the antigenomic HDV ribozyme. In this
system, the substrate (SANT) is associated by base-pairing to the catalytic RNA (RzANT) forming helix H1. RzANT is able to
cleave specifically the RNA substrate as well as a deoxysubstrate analogue
containing a single ribocytidine at the cleavage site (position -1). This demonstrates that such deoxysubstrate analogues are valuable
tools for structural studies of this ribozyme domain. They form however weak
complexes with RzANT which is due in part to their ability to fold as stable
hairpins unlike the RNA substrate. Using a set of full deoxy or of mixed deoxy-ribo substrate analogues site-specific substituted with the photoaffinity probe deoxy-4-thiouridine, ds
4
U, at a defined position, we were able to determine a number of long range
contacts between the substrate and the ribozyme core. In particular, crosslinks
between substrate position -1 and position -2 with residues C15, G19 and C67, thought to be involved in the
ribozyme catalytic site, were detected. A three dimensional model of the
antigenomic ribozyme system, derived from the structure proposed by Tanner
et al.
[
Current Biol.
(1994) 4, 488-498] for the genomic system was constructed. Apart from residue deletion
or insertion, only minor accommodations were needed to account for all
photocrosslinks but one which is attributed to an alternative hybridization of
the substrate with the ribozyme. This study therefore further supports the
structure proposed by Tanner
et al
. for the pseudoknot model.
Hepatitis delta virus (HDV) is an infectious human pathogen that requires the
coat proteins of hepatitis B virus for encapsulation. Both the genomic and
antigenomic HDV RNAs probably replicate by a RNA-dependent rolling-circle mechanism that generates linear multimers (
1
). These multimers exhibit self-cleaving activity which requires the presence of divalent cations and
generates a 2',3'-cyclic phosphate and a free 5'-hydroxyl at the cleavage site (
2
-
5
).
Under
in vitro
conditions, optimal catalytic activity is obtained with 85 nucleotide (nt) long
genomic or antigenomic ribozymes, i.e. with only one nucleotide 5' to the cleavage site (
6
). The corresponding sequences cannot be folded in the hammerhead or hairpin
forms and therefore represent a distinct self-cleaving motif (
7
). At least five different secondary structures have been proposed (
2
,
7
-
10
) but the pseudo-knot model of Perotta and Been, which is common to both the genomic and
antigenomic ribozymes, best fits the experimental data (
7
,
13
). It contains four base-paired regions (H1-H4) and the cleavage site is positioned at the 5'-end of the 84 nt ribozyme sequence. H3 and H4 are
stems with hairpin structures and, together with H1 and H2, form a pseudo-knot. Hairpin IV can be reduced in size without complete loss of catalytic
activity (
11
-
13
) and residues involved in the catalytic site are thought to be located within
the loop of hairpin III and in the sequence connecting hairpin IV to helix II (
11
-
12
,
14
). Recently, a three dimensional model of the genomic HDV ribozyme based on the
pseudo-knot secondary structure has been proposed (
15
). In this genomic model two conserved residues of loop III and a residue of
junction II/IV essential for cleavage activity are positioned in close
proximity to the scissile phosphodiester bond (for review see ref.
16
).
In order to check this model we have attempted to identify couples of residues
which are distant in the primary sequence but maintained in close contact
within the folded structure of the ribozyme. This can best be achieved using an
intrinsic photoaffinity methodology based on the site-specific insertion of thionucleotides. The corresponding thiobases can be
selectively photoexcited with light of wavelength >310 nm and then form mixed
photoadducts with neighbour residues. In particular, 4-thiouracil can substitute U residues in RNA (T in DNA) with minimal
structural perturbation and is able, upon 365 nm irradiation, to yield stable
photoadducts with any of the current nucleosides (
17
-
19
). So far, this methodology has been successfully applied to the analysis of the
conformation of various RNA in solution including the hairpin or hammerhead
ribozyme domains (
20
-
21
).
In the present work we have taken advantage of the capability of the HDV
ribozyme to undergo
trans
-cleavage reactions. Two forms of
trans
reaction have been generated (
10
,
22
). In the first form the self-cleaving molecule is interrupted at H4 and the two sequences reassociate
via interactions within H4 and H2. In the alternative form, the sequence is
separated between H1 and H2 and the two sequences reassociate by base-pairing within H1 to generate the cleavable structure. Our present studies
were conducted with this latter form of the antigenomic ribozyme since (i) it
exhibits higher catalytic activity in
trans
, (ii) the minimal substrate size is only 8 nt long (Fig.
1
) and, accordingly, substrate or substrate analogues containing the deoxy-4-thiouridine probe (ds
4
U) inserted at a selected position can readily be obtained by chemical
synthesis. Here we show for the first time that a deoxy-substrate analogue containing a single ribocytidine at the cleavage site
(position -1) can be enzymatically cleaved. This substrate analogue exhibits much
lower binding affinity toward the ribozyme than its parent substrate which is
due in part to its ability to fold as a hairpin. Using a series of full deoxy
or mixed ds
4
U containing substrate analogues, we have then determined a number of long-range contacts within the hybrid substrate analogue-antigenomic ribozyme
complexes. Overall, our data are consistent with a three dimensional folding of
the antigenomic ribozyme derived from the one proposed by Tanner
et al.
(
15
) for the genomic ribozyme.
The
trans
-cleaving ribozyme (RzANT) was synthesized using a large scale
transcription method (
24
) for <100 nt long RNAs. We used the sequence that Perrotta
et al
. proposed (
14
), including the non-HDV derived nucleotides. The transcription buffer contained 40 mM Tris-HCl pH 8.0, 15 mM MgCl
2
, 5 mM DTT and 1 mM spermidine (
23
). The template was a double-stranded DNA containing a T7 RNA polymerase promoter formed by the
hybridization of two chemically synthesized (GENSET, France) and purified DNAs.
The final concentrations of the bacteriophage T7 RNA polymerase and the DNA
template were 5.6 U/ml and 200 nM, respectively. The 2 ml reaction mixture was
incubated 4 h at 37oC. After a PAGE purification, we obtained 0.21 mg of unlabelled active RNA.
The ribozyme was dephosphorylated using calf intestinal phosphatase
(Pharmacia), and the mixture treated
with 10 U Proteinase K in the presence of 2% SDS. After a phenol extraction, the
ribozyme was precipitated with ethanol and then phosphorylated with T4
polynucleotide kinase and [[gamma]-
32
P]ATP at 37oC for 30 min. The reaction was terminated by adding an equal volume of
formamide containing xylene cyanol and purified by gel electrophoresis in 10%
acrylamide. The ribozyme was located by autoradiography, cut from the gel,
eluted overnight in NaCl (0.15 M) and precipitated with ethanol. The
corresponding 15mer RNA substrate SANT was synthesized using the same method.
Full deoxy or mixed substrate analogues were synthesized by GENSET (France)
using the phosphoramidite method. Analogues containing a single ds
4
U residue at a selected position were synthesized similarly, following the
protocol which uses 4-
S
-pivaloyloxymethyl-4-thiodeoxyuridine phosphoramidite (
17
). After deprotection the oligomers were purified by 20% PAGE in the presence of
8 M urea. Fully thiolated oligomers were obtained by an additional affinity
electrophoresis step (
20
,
25
). These thiolated oligomers were referenced in Table
1
. The substrates used in the cleavage reaction are either oligoribonucleotides
SANT, 5'-GGUUCGGUCGGCAU-3'; SANT', 5'-UUGGUUCGGUCGGCAU-3'; or a mixed DNA-RNA analogue S', 5'-ccaaaCgggtcggt-3'. In native gel electrophoresis experiments C, 5'-UGCCCGUCUGUUGU-3', was used as a control.
The incubation buffer was usually 10 mM MgCl
2
and 50 mM Tris-HCl (pH 8.0) and
trans
-cleavage reactions were initiated by adding the 5'-end-labelled substrate and incubating at 37 or 55oC. Aliquots were removed at various time intervals
and reactions were stopped by adding an equal volume of 50 mM EDTA, 0.1% xylene
cyanol and 80% formamide. Samples were electrophoretically separated on a
denaturating 20% acrylamide gel and autoradiographed. The bands corresponding
to full-length precursors and 5' cleaved products were quantified by counting Cerenkov emission of
the excised gel slices in a Beckman LS 6000IC scintillation counter. The
percentage of cleavage was determined by taking the ratio of product to total
radioactivity for each time point.
5'-
32
P-labelled thiolated substrate analogues were dissolved at defined
concentrations in the cleavage buffer. The solution (5 [mu]l) was then introduced in a siliconized glass capillary and placed at 2 cm
from the exit slit of a Bausch and Lomb monochromator equipped with a HBO 200W
super pressure mercury lamp and a Schott WG 345 nm filter. The sample was
irradiated for 45 min (100 kJ/m
2
) i.e. to completion of the reaction. The crosslinked and uncrosslinked
oligomers were then separated by 15% PAGE.
The same general procedure was applied to ribozyme-substrate analogue complexes. Each complex was formed by incubating 100
nM 5'-
32
P-labelled ribozyme with 2 [mu]M cold substrate analogue in the cleavage buffer at 4oC. Irradiation was performed at 15oC and the crosslinked forms of the ribozyme were separated
from uncrosslinked ribozyme by 12% PAGE. The crosslinked species were detected
by autoradiography quantified using a phosphorimager (Molecular Dynamics),
eluted and precipitated with ethanol.
Limited alkaline hydrolysis of end-labelled intact or crosslinked species was carried out in 50 mM carbonate
buffer pH 9.2 in the presence of tRNA (0.5 mg/ml). Each sample (5 [mu]l) was heated at 100oC for 1 min, cooled and loaded onto a sequencing gel (10% PAGE) at 60 W
for 1.5 h. The correspondence between the individual bands of the ladder
obtained with the intact ribozyme and its sequence was established using
limited T1 RNase digestion.
5'-
32
P-labelled substrate analogues and control oligonucleotides pre-incubated in the cleavage buffer for 15 min at 55oC were analysed by a non-denaturating 15% PAGE in the presence of 10 mM Mg
2+
, 50 mM Tris-HCl, pH 8. Samples were run for 15 h at 11 W at various temperatures (see
Results).
The three-dimensional modeling was based on the structure proposed by Tanner
et al
. (
15
) for the genomic ribozyme. Coordinates were kindly provided by E. Westhof.
Structural elements were manipulated on a PS 390 using SYBYL software package
(Tripos, St Louis). At various steps, the results of the manipulations were
energy minimized to satisfy stereochemical constraints. Coordinates are
available from the authors by e-mail to laugaa@ccr.jussieu.fr.
The sequences of the 15mer RNA substrate (SANT) and of the corresponding
trans
-acting 81mer ribozyme (RzANT) are shown in Figure
1
. The cleavage reaction was performed at a constant initial substrate
concentration (0.1 [mu]M) and at three different ribozyme to substrate ratios,
r
. The half-reaction times at 55oC were found to be respectively 0.9, 3 and 120 min at
r
values 10, 1 and 0.1, respectively. The cleavage reaction is less efficient at
37oC since the half-reaction time was close to 5 min at
r
= 10. A turn-over was observed only at 55oC and in our conditions (
r
= 0.1) 3.5 molecules of SANT were cleaved per molecule of RzANT, after 1 h
incubation. These data are in line with those previously reported by Perrotta
and Been (
14
). Under the conditions used here, RzANT was far more active than its genomic
counterpart in the
trans
-cleavage reaction (data not shown). Therefore, one should expect RzANT to
be a more valuable tool for structural analysis of the active form of the HDV
catalytic RNA.
In order to examine whether a deoxy substrate analogue can be catalytically
cleaved we have synthesized a DNA oligomer S' which contains a ribocytidine at the cleavage site. S' is cleaved in the presence of RzANT at both 37 and 55oC (Fig.
2
) generating a 5'-
32
P-labelled product that co-migrates with the 5' fragment (7mer) obtained upon alkaline hydrolysis of S'. As expected, the
trans
-cleavage reaction is abolished when 50 mM EDTA is added to the ribozyme
prior to S' addition. Perrotta and Been (
14
) have shown that a deoxycytidine 5' to the cleavage site blocked the reaction. Taken together, these data
demonstrate for the first time that the presence of a 2'-hydroxyl group i.e. a ribonucleotide at position -1 of the substrate is strictly required for the catalytic
cleavage by RzANT as previously established in the case of hairpin and
hammerhead ribozymes (
27
-
26
). The cleavage of S' is highly sensitive to temperature; the substrate is apparently not
stably associated with the RzANT at 55oC, which dramatically reduces the rate of the reaction. Consistent with
such a view, at 55 or 37oC no competitive inhibition of SANT cleavage in the presence of large
excess of a deoxy substrate analogue ds (up to 0.1 mM) was detected. Under the
same conditions the 10mer product of SANT cleavage behaved as an inhibitor
(data not shown). It can thus be concluded that S' is able to bind to the correct site on RzANT albeit with a low affinity.
In order to unravel long-range contacts within the ribozyme domain (Fig.
1
) we first examined the behaviour of a series of photoactivable deoxysubstrate
analogues and then extended this study to a few selected mixed ribo-deoxy oligomers. The full deoxy analogues used here are derived from the
12mer analogue ds which is expected to form a complete hybrid helix H1 when
combined with RzANT. Each dsi contains a single deoxy-4-thiouridine residue, ds
4
U, at a position occupied either by U (-2, +4, +10) or by C (-1, +5, +8) in ds (Table
1
). The substituted positions were selected in order to explore extensively the
long range interactions within the dsi/RzANT complex and to minimally affect
the stability of the helix I which is common to all the proposed secondary
structures. Indeed, residues at positions -2, -1, +8 and +10 are not involved in base pairs (Fig.
1
). Substitutions at position +4 and +5 will introduce either a ds
4
U-A and or wobble ds
4
U-G base pair in H1, respectively, with unchanged (
18
) or a decreased stability of this helix. In order to establish appropriate
conditions for the
trans
-photocrosslinking experiments various concentrations of the ds-2 analogue, ranging from 100 nM to 2 [mu]M, were mixed with 100 nM 5'-
32
P-labelled RzANT in the cleavage buffer at 15oC. The mixtures were irradiated with 365 nm light at 15oC. Each sample was then run on a denaturing electrophoresis gel
to separate the retarded crosslinked species from the unreacted RzANT. The
fraction of RzANT crosslinked to ds-2 increased in parallel with ds-2 concentration reaching 8% at 2 [mu]M. While the same crosslink patterns were obtained at 15 and 25oC, the overall crosslinking yield appears sensitive to the
temperature at which the irradiation step is performed. It decreases with an
increasing temperature and for ds-2 at temperatures >37oC no crosslinks could be detected. Further experiments were thus
performed at 15oC in the presence of 2 [mu]M dsi since under these conditions the control thiolated 12mer dc, which is unable to anneal to RzANT, yielded no detectable
photocrosslink (Table
1
).
Table 1
As shown in Table
1
, some dsi oligomers (ds10, ds-1, ds-2) yielded two to three crosslinked species (Fig.
3
A) while others (ds4, ds5, ds8) generated no products. Some crosslinks were
formed in low yield however, <= 1%, and could possibly have resulted from side photoreactions not involving
the ds
4
U probe. To check this point, the unsubstituted deoxysubstrate ds was irradiated
together with 5'-
32
P-labelled RzANT and no crosslink could be detected under these conditions
(Table
1
). Hence the covalent bridges analysed below are necessarily formed between ds
4
U and a ribozyme residue. In order to determine the nature of this residue in
crosslinked species each of them was isolated, repurified by denaturing PAGE if
necessary and sequenced using limited alkaline hydrolysis. A control ladder was
obtained starting from 5'-
32
P-labelled RzANT and scaled using limited RNases digestions. When RzANT was
crosslinked at position Y, random cuts 5' to the crosslink generated a partial ladder that paralleled the control
one, while all the cuts 3' to the crosslink resulted in bands with retarded migration due to the
attached substrate analogue. This led to a gap in the alkali ladder thus
allowing an unambiguous identification of Y (
20
-
21
) (Fig.
3
B).
A full comparison of the behaviour of the RNA and DNA substrate analogues in the
presence of RzANT should take into account their eventual structured forms
which, for short oligonucleotides, can be either hairpins or dimers (
28
). Thus, we first compared by native gel electrophoresis (pH 8, 10 mM Mg
2+
) the migration rates of SANT' and of a control C, unable to fold as a hairpin or to give rise to self-associated forms. The gel was run at 37oC and irrespective of the applied concentrations (5 nM to 1 [mu]M) SANT' and C yielded a single spot that migrates at
closely similar rates indicating that SANT' is devoid of any kind of secondary structure.
We then compared, by the same procedure, the behaviour of ds-2 with that of a control 12mer dc of similar nucleotide composition (Table
1
). Again, dc was designed in order to minimize its ability to fold as a hairpin
or to self-associate. Figure
4
A illustrates the data obtained at 55oC. ds-2 yielded a minor spot comigrating with dc and a major faster
migrating spot. The ratio of counts in the two bands was found to be constant,
11 +- 2%, irrespective of the applied ds-2 concentration. The data therefore strongly suggest that at 55oC, ds-2 is in equilibrium between an unfolded form and a
major fast migrating hairpin form. The existence of this hairpin form was
confirmed by the photoaffinity methodology (
28
). 5'-
32
P-labelled ds-2 was irradiated to completion of the reaction with 365 nm light at
15oC in the cleavage buffer and the irradiated mixtures subjected to
denaturing PAGE analysis. At ds-2 concentrations <= 1 [mu]M only two species were detected on the gel (Fig.
4
B). One of them behaves as a single-stranded 12mer co-migrating with irradiated dc or non-irradiated ds-2 (not shown) and is expected to occur from unproductive
ds
4
U photolysis. The major fast migrating band corresponds to a crosslinked hairpin since: (i) it adopts a compact structure in
denaturing conditions; (ii) its yield of formation, 70 +- 3%, remains constant whatever the initial ds-2 concentration <1 [mu]M. When ds-2, 1 [mu]M, was irradiated at different temperatures the
yield of formation of this crosslinked hairpin remained practically constant,
up to 55oC (data not shown), again indicating that ds-2 is predominantly in the hairpin form at this temperature.
Irradiation of 5 or 10 [mu]M solution of ds-2 at 15oC generated new strongly retarded species on the gel (Fig.
4
B) that formed in competition with the crosslinked hairpin (Fig.
4
C). These species certainly correspond to crosslinked ds-2 dimers as shown by their migration behaviour and their concentration
dependent formation. Thus ds-2 is able to self-associate as a duplex form under cleavage conditions.
The data of Figure
4
indicate that ds-2 either folds as a hairpin or self-associates to yield imperfect dimeric forms. As the hairpin form
(unlike the dimers) is stable up to 55oC, it should largely predominate over unfolded ds-2 at lower temperatures. At 15oC it can thus be considered that the dimeric and hairpin forms
are in equilibrium with an apparent dissociation constant,
K
ap
(~10 [mu]M), derived from the data of Figure
4
C according to Tanaka
et al.
(
28
). The structures proposed for the hairpin (H, H') and the dimeric forms (D, D') of ds-2 are shown in Figure
5
. A single crosslinked hairpin species can be detected (Fig.
4
B) and we favor form H as its stem is far more stable that the corresponding
stem in H'. Both dimeric forms D and D' contain several mismatched base-pairs of type g-t, g-g, g-a that are known to slightly
destabilize a duplex (
29
). The presence in D of three mismatched pairs and two couples of alternative g-c, c-g pairs, nevertheless, favors this form. As the photochemical data
(Fig.
4
B) indicate the formation of multiple crosslinked dimers, it cannot be excluded
that D' may be present in low relative amounts. In contrast to the behaviour of
the dsi family, both SANT' and s''-2, which contains a ribonucleotide sequence
potentially involved in the hairpin H, revealed unable to adopt such a
structure. They behave as unfolded oligomers retaining the potential to form
dimeric structures at 15oC.
Recently, three-dimensional models have been proposed for the genomic pseudoknot (
15
) and both antigenomic pseudoknot and axehead (
32
) ribozymes. A basic difference between the pseudoknot and axehead lies in the
Watson-Crick base-pairing involving the ribozyme 3'-end sequence which belongs to helix II in the former
while it forms a helix with residues 5' to position -2 in the latter (Fig.
1
). The models of Branch and Polaskova (
32
) were designed after co-mutation data interpreted in terms of base-pairing or proximity between couples of residues, though no upper
limit was set to define proximity. They claimed that their data were largely
compatible with the model of Tanner
et al
. (
15
) but it should however be noticed that, in this model, distance measurements
between C1' atoms yielded values ranging from 20 to 33 Å for four out of their eight sets of nearby residues. As neither
coordinates, nor details concerning software and building methodology were
given, the models of Branch and Polaskova (
32
) are inadequate for structural purpose and precludes any comparison with our
photocrosslinking data. It is questionable whether the
trans
-acting ribozyme constructions described here could adopt the axehead
structure. None of the substrate analogues (cleavable or not) used here have a
5'-end sequence able to form a Watson-Crick minihelix with the 5'-GGAG sequence at the 3'-end of the ribozyme as required
for the axehead. It is also noteworthy that position -1 yielded a crosslink with position 31, in agreement with the pseudoknot
model, but none with positions 70-72 which would have supported the axehead. With regards to our system,
the pseudoknot secondary structure appears to be the more likely.
It cannot be excluded, however, that alternative hybridization occurs. For
example, the crosslink between ds10 and G21 is best explained by the structure
depicted in Figure
1
B, whereas those between s''-2, ds-1 and G31 can be accounted for by both. The lack of
reactivity of ds4 and ds5 is expected since, in helix H1, the ds
4
U probe form standard (ds4) or wobble base pair (ds5) and the minihelix geometry
and flexibility disfavor inter-strand photocrosslink formation as observed in a number of model systems
(unpublished data). The long range crosslinks obtained with ds
4
U at position -1 or -2 of the substrate indicate a tertiary folding of the pseudoknot
structure placing C67, G19 and C15 in close proximity to the cleavage site. The
antigenomic and genomic ribozymes can adopt similar pseudoknotted secondary
structures (
23
). Some residues, located in the same secondary patterns, are strictly conserved
in both ribozymes: G31, G32 and G33 in junction I/IV; G66, C67, A69 and A70 in
junction II/IV; U14-C20 in loop III (antigenomic numbering). Mutational studies performed on
the genomic ribozyme have shown that substitution of any of these residues
strongly decreases the cleavage efficiency and that the residue corresponding
to C67 cannot be changed to any other nucleotide. On this basis, a three-dimensional model has been proposed for the genomic ribozyme by Tanner
et al
. (
15
). In this model the residue of the cleavage site is within 12 Å of the residues corresponding to C67 and C15. Given these similarities,
both ribozymes are expected to have very closely related three-dimensional folding. We therefore derived a model for the antigenomic
ribozyme from the structure proposed for the genomic one, assuming that the
conserved aforementioned residues have similar spatial arrangement.
Apart from the sequence, the main differences between both ribozymes result from
either deletion of non-conserved residues in loop III and junction I/IV or insertion of a residue
at the 3'-end of junction II/IV. Compared to the genomic model, the
antigenomic construction (Fig.
6
) is as follows, (i) Loop III is one nucleotide shorter, which results from the
deletion of the dispensable U residue at the 3'-end of the genomic loop (
33
). The 5' side of the loop which contains catalytically important residues is left
unchanged whereas, on the 3' side, G19 is now 4 Å closer to the cleavage site. (ii) Junction I/IV is three
nucleotides long (GGG) instead of six (GGGCAA). The gap formed by the deletion
of the CAA trinucleotide is now filled by G32 and G33 which adopt an unwinded A-type stacking between helix I and helix IV. The relative spatial
disposition of helix I with respect to helix IV remains almost unchanged. It
should be noticed that residues G32 and G33 are stacked between helices I and
IV whereas they are tilted and point outwards in the genomic ribozyme. (iii) In
order to accommodate the additional guanine at the 3'-end of junction II/IV, helix II has been slightly tilted (~15o) with respect to helix III via a +20o rotation about the [epsilon] angle of C10, at the junction between both
helices. Therefore, the remainder of junction II/IV which contains residues important for cleavage is
left unchanged. Starting from its 5'-end, the conformational path of the junction can be seen as a right handed
helical extension of helix IV, the first three residues yielding close contacts
between their sugars and helix/loop III. Their bases point towards junction
I/IV and helix I. There is a turn 3' to the third (variable) residue, induced by steric repulsion with helix
I which induces the last residues to wrap around helix III in a right helical
way, therefore insuring continuity with helix II. Such a way of incorporating
the extra G is devoid of effect on the relative positions of the residues
important for cleavage and induces only minor effects on the structure of
junction I/II of
cis
-cleaving ribozymes. This is consistent with data showing that a genomic
ribozyme in which junction II/IV has been replaced by the antigenomic one is
still active (Lescure, unpublished results).
While cleavage experiments were performed at 37 or 55oC, most of our crosslinking data were obtained at 15oC. In a few cases ds-2 and s''-2, we have shown however that the crosslink
pattern was largely temperature independent (up to 55oC for s''-2). Formation of a photoadduct between ds
4
U and an acceptor residue implies that the two reactants come in close
proximity, <4.5 Å, within the complex (
18
).
Photocrosslink with s
'''
-2 and s
''
-2
. Both s'''-2 and s''-2 form an A-type RNA-RNA helix I. In the
model the residue at position -1 is closely surrounded by residues G1, G31-G33 (junction I/IV) and residues U14-C15 (loop III). The strand 5' to C-1 (dc-1) is constrained to go out either on
loop III side, at the back of the molecule or on junction I/IV side at the
front of the molecule. The presence of a C3'-endo ribocytidine in s'''-2 orients its 5'-phosphate on the front of
the molecule. The residue at position -2 slips in between G33 on the one side and C67, G66 on the other side and
the remainder of the 5' strand lies along the major groove of helix IV. These steric constraints
bring residue -2 in close proximity to C67. The carbon-sulfur bond of ds
4
U is parallel to the C5-C6 double bond of C67 and at a distance of 3.8 Å which nicely accounts for the observed crosslink (Fig.
6
, top). Furthermore, none of the other residues is liable to form a
photocrosslink. Interestingly, the Rz/s'''-2 complex becomes inactive upon formation of the -2 to C67 crosslink which could well be due to
the alteration of the chemical structure and spatial disposition of C67. This
is in line with the statement discussed above that C67 is a key residue of the antigenomic catalytic site. The non-cleavable substrate analogue s''-2 differs from s'''-2 only by the presence of a
deoxy residue at the cleavage site. It however yields a distinct crosslink
pattern which involves C15 and G19 in loop III, at the back of the molecule,
whether in Mg
2+
or high Na
+
containing media. Its behaviour is necessary due to the presence of a deoxy
sugar at position -1: C'2-endo preferred conformation, greater flexibility and absence
of 2'-hydroxyl. The model shows that, when the residue of the cleavage
site is a deoxyribonucleotide with a C2' endo conformation its 5'-phosphate is in a less favorable configuration to position
the 5'-end of the substrate at the front of the molecule. The substrate 5'-end is easily built running on loop III side, at the
back of the molecule (Fig.
6
, bottom). Residue -2 now points towards residues C15 and G19. Both photoreacting bonds are
sterically accessible to residue -2 and are at a distance of 6-7 Å. A slight rotation (+20o) of helix and loop III was performed to bring these
bonds in a more favorable orientation. The rotation axis was perpendicular to
helix III average bases plane, and passed through the center of helix III
atomic coordinates. The distance between ds
4
U-2 and G19 photoreacting bonds average to ~4 Å with an antiparallel orientation which accounts for the
photocrosslink detected. The distance between ds
4
U-2 and C15 is larger (8 Å) but a distance as low as 4 Å can be reached without steric hindrance, at the expense of
that between G19. Thermal motions of the substrate 5'-end and of loop III may well account for such behaviour. Therefore,
both photocrosslinks are in agreement with the proposed three-dimensional structure (Fig.
6
).
Photocrosslinking with ds-2 and ds-1.
ds-2 is made of deoxyribonucleotides and contains only one residue 5' to the cleavage site. A deoxy substrate is not expected to induce structural
modifications of the ribozyme since it has been shown by NMR that DNA-RNA hybrids retain most of the structural features of A-type RNA helices (
34
). Furthermore, the two crystal structures of the hammerhead ribozyme are almost
identical though one of them has been performed with a deoxy substrate (
35
-
36
). A photocrosslink is obtained with G19, as in the case of s''-2. Besides, the crosslink with U23 may well be explained on
the basis of thermal motions. Indeed, the lack of any nucleotide 5' to residue -2 is likely to increase its mobility, hence it can reach U23 which
is on the same side of the ribozyme and accessible at the top of helix III.
Furthermore, U23, located 5' to a sharp turn, is expected to experience a large degree of fraying.
Unlike to s''-2, no photocrosslink is observed with C15 which might arise
from the larger movement of residue -2 between G19 and U23. The ds-1 to C15 photocrosslink is well accounted for by the structure. The
photocrosslink observed with U23 can be explained by the mobility of the short
5'-end of the substrate, including G1, allowing a back folding which
brings residue -1 in the vicinity of U23.
ds10 photocrosslinking with C7, G21 and U23.
Residue 10 is the penultimate residue of the 3'-end of the substrate. This residue located on a dangling strand is
expected to have a large degree of freedom. The occurrence of photocrosslinks
with C7 and U23 which are both sterically accessible is thus fully compatible
with the three-dimensional structure of the ribozyme. Besides, residue G21, located at
the junction between helix III and loop III, is much less accessible and too
far from residue 10 to yield a photocrosslink. This adduct might be explained
by an alternative co-folding of the ribozyme and the substrate as shown in Figure
1
B.
The herein proposed
model for the antigenomic HDV ribozyme accounts for the above photocrosslinking
data. Apart from residue deletion or insertion, the only minor
difference with the previously described model of the genomic HDV ribozyme (
15
), resides in a slight rotation of helix and loop III with respect to helix II
and helix I. It is noteworthy that, for the first time, the model of Tanner
et al
. (
15
) as well as the pseudoknot secondary structure
receives a strong experimental support due to the possibility to obtain
information on inter-residue distances.
We thank Kyle Tanner and Ilya Pashev for constructive comments on the
manuscript, Prof. Marc Vasseur for his constant interest in this work and
GENSET (France) for its kind gift of synthetic oligonucleotides. This work was
partly sponsored by grants from ANRS and Interface Chimie-Biologie (CNRS) to A.F. C.B. was supported by a fellowship from Conseil Général d'Ile de France, GENSET France and Institut de
Formation Supérieure Biomédicale.
REFERENCES
Return