 1996 Oxford University Press
Nucleic Acids Research, 1996, Vol. 24, No. 4
573–581
Base and sugar requirements for RNA cleavage of
essential nucleoside residues in internal loop B of
the hairpin ribozyme: implications for secondary
structure
Sabine Schmidt, Leonid Beigelman1, Alexander Karpeisky1, Nassim Usman1,
Ulrik S. Sørensen+ and Michael J. Gait*
Medical Research Council, Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK and
1Department of Chemistry and Biochemistry, Ribozyme Pharmaceuticals Inc., 2950 Wilderness Place, Boulder,
CO 80301, USA
Received November 22, 1995; Revised and Accepted January 8, 1996
ABSTRACT
The hairpin ribozyme is a small self-cleaving RNA that
can be engineered for RNA cleavage in trans and has
potential as a therapeutic agent. We have used a
chemical synthesis approach to study the requirements of hairpin RNA cleavage for sugar and base
moieties in residues of internal loop B, an essential
region in one of the two ribozyme domains. Individual
nucleosides were substituted by either a 2′-deoxynucleoside, an abasic residue, or a C3-spacer (propyl
linker) and the abilities of the modified ribozymes to
cleave an RNA substrate were studied in comparison
with the wild-type ribozyme. From these results,
together with previous studies, we propose a new
model for the potential secondary structure of internal
loop B of the hairpin ribozyme.
INTRODUCTION
The hairpin ribozyme is a region of the negative strand of the
satellite RNA from tobacco ringspot virus that undergoes
self-cleavage (1–3). The hairpin (reviewed in 4) is one of several
small RNAs, which include the hammerhead and hepatitis delta
virus RNAs, that can be engineered for cleavage in trans (5,6). In
such cases, a substrate RNA strand is cleaved following
incubation with a catalytic RNA (ribozyme) strand in the
presence of divalent metal ions such as magnesium. The power
of such ribozymes is being harnessed for gene therapy applications through intracellular, vector-driven RNA production (for
example, see 7) as well as for exogenous delivery as potential
therapeutics (8,9).
Much effort has centered on understanding the structures and
mechanisms of action of small RNA ribozymes in order that this
will lead to the design of more effective RNA-cleaving thera-
* To
peutic agents (6,10). Particular success has come from crystallographic and other structural studies of the hammerhead ribozyme
where knowledge of the overall folding is now leading to the first
proposals for the structural basis of the cleavage mechanism
(11,12). By contrast, much less is known about the conformation
and mechanism of cleavage of the hairpin ribozyme and there are
no data available yet as to its three-dimensional structure.
Mutagenesis and in vitro selection studies have revealed that
the hairpin ribozyme consists of four regions of Watson–Crick
duplex RNA and two internal loops (A and B) (Fig. 1) (13–18).
All of the essential nucleotides, the identity of which are required
for catalytic cleavage, are located in the two internal loop regions
(16,18). It has recently become clear that the hairpin ribozyme
can be divided into two domains, the first containing the substrate
strand (internal loop A and helixes 1 and 2) and the second
consisting of internal loop B and helixes 3 and 4. These domains
appear to be hinged at the junction of helixes 2 and 3 (between
residues 14 and 15) such that the ribozyme can bend and the two
internal loop regions can approach and presumably interact with
each other during catalytic cleavage. Such conclusions have been
reached primarily from studies where the two domains are
constrained by linkers of variable length between residue –5 on
the substrate strand and residue 50 on the ribozyme strand (19,20)
and also by linking the two domains in a completely different
‘reverse’ configuration (21). Very recently it has been shown that
the two domains can be completely separated into independent
sections and ribozyme activity reconsitituted by addition of one
to the other (22). Although a 104 increase in KM was found, there
was remarkably little effect on kcat. The domains may play
somewhat different roles, however. Whereas internal loop A may
be conformationally flexibile (23), loop B appears to undergo an
initial folding process which is independent of binding of
ribozyme to substrate (24).
Although the existence of the four Watson–Crick helix regions
is now well established, the secondary structures of the two
whom correspondence should be addressed
+Present
address: Department of Chemistry, Odense University, Campusvej 55, 5230 Odense M, Denmark
574
Nucleic Acids Research, 1996, Vol. 24, No. 4
Figure 1. Secondary structure of the hairpin ribozyme shown in trans-cleaving
3-stranded mode (34). The arrow denotes the site of cleavage. The four helixes
(H-1–H-4) are marked together with the two internal loops. Positions of the
residues of loop B that are substituted by analogues in this study are highlighted
in yellow. A dashed line marks the region of extension of helix 4 in the
3-stranded ribozyme which replaces a UGG loop in the natural ribozyme.
internal loops (intra-loop contacts) and their orientations with
respect to one another (inter-loop contacts) during the process of
catalytic cleavage are less clear. This has impeded progress
towards molecular modelling of the hairpin ribozyme, since there
have been insufficient constraints available. Various approaches
have been taken to try to determine some of these inter- and
intra-loop contacts and to get information about the overall
conformation. One method has involved the measurement of
accessibility of the ribozyme to chemical modifying reagents
(25). By this procedure, non-canonical base-pairs were proposed
between G21:A43, A22:U42 and A23:A40. Another informative
finding was that a stable photo-crosslink could be formed
between G21 and U41 without significant detriment to the
catalytic activity, suggesting reasonably close proximity of these
two residues (26). Sequence similarity was also noticed between
this region of loop B and UV sensitive domains found in the loop
E of eukaryotic 5S RNA and in the conserved C domain of
viroids. On the basis of these data, Butcher and Burke have
proposed a partial secondary structure model of loop B (25).
An alternative procedure involves the substitution of individual
nucleotide residues in an RNA with nucleotide analogues
carrying specific base, sugar or phosphate modifications (27–29).
Such a procedure has been widely used in the study of the
hammerhead ribozyme (reviewed in 6,30,31). An early finding
was that a hairpin ribozyme where G+1 in the substrate strand was
substituted by inosine lost all detectable cleavage activity but no
such loss occurred with 2-aminopurine substitution (32). This
suggested that the 2-amino group of G+1 was crucial for ribozyme
activity and that it may be involved directly in the catalytic
mechanism. More recently, all the essential purine residues in
loops A and B of a 3-stranded ribozyme (Fig. 1) have been studied
with regard to their functional group necessity for ribozyme
activity (33). None of these changes was as drastic as loss of the
G+1 2-amino group. In loop B, substantial effects were observed
for loss of amino groups from G21, A22, A23 or A38 but not for
A40 or A43, making hydrogen-bonding contacts to these residues
less likely and providing no corroboration for two of the
non-canonical base-pairs suggested by chemical modification
mapping. We therefore proposed a slightly different secondary
structure model of the lower section of loop B (33).
There has been only one study of the sugar requirements for
catalytic cleavage of the ribozyme strand of the hairpin (34). For
ribozyme strand A (Fig. 1), replacement of individual ribonucleoside residues at A10, G11, A24 or C25 by either 2′-deoxy- or
2′-O-methyl-nucleosides were found to be deleterious to ribozyme cleavage. For ribozyme strand B, no dramatic reductions in
cleavage rate were noticed for global deoxynucleotide substitution (replacement of either all A, G, C or U in separate
transcripts), but complete deoxy-substitition caused total loss of
cleavage ability.
In order to obtain further information about the sugar and base
requirements of loop B of the hairpin ribozyme for catalytic
cleavage, we have now studied the effects of single substitution
of most of the nucleotide residues in loop B by either a
2′-deoxynucleotide, an abasic residue, or a C3-spacer (propyl
linker), concentrating on ribozyme strand B where data has been
particularly sparse. The results, when taken together with
previous data, have allowed us to draw some important conclusions concerning the potential secondary structure of loop B of
the hairpin ribozyme.
MATERIALS AND METHODS
Preparation of oligoribonucleotides
Oligoribonucleotides were synthesized on a 1 µmol scale using
2′-O-t-butyldimethylsilylnucleoside-3′-O-(2-cyanoethyl-N,N-diisopropyl) phosphoroamidite monomers having phenoxyacetyl
amino group protection for A and G and benzoyl protection for
C (Glen Research via Cambio). 2′-deoxy-3′-O-(2-cyanoethylN,N-diisopropyl)phosphoramidites and the C3-spacer phosphoramidite, used as modifications, were also obtained from Glen
Research. 1-Deoxy-D-ribofuranose phosphoramidite was synthesised from D-Ribose as described recently (35,36). The
syntheses of the modified and unmodified oligoribonucleotides
were undertaken using RNA synthesis procedures as previously
described (33,37,38) but with some minor improvements as very
recently described (39). In some syntheses, the standard tetrazole
activator was replaced by the more powerful 5-ethylthio-1Htetrazole (40,41). Minor changes for deprotection and purification were also introduced. Thus, oligoribonucleotides were
deprotected by suspension of the controlled pore glass in
methanolic ammonia overnight, decanting and evaporation of the
resultant solution to dryness. Treatment with triethylamine
trihydrofluoride at room temperature overnight (42) or triethylamine trihydrofluoride/DMF (3:1) at 55C for 1.5 h (41) was
carried out to remove silyl groups, followed by desalting via
Sephadex NAP-10 (Pharmacia) filtration. Oligonucleotides were
purified by strong anion exchange chromatography on a semipreparative NucleoPac PA-100 column (Dionex) (40) using
Buffer A: 10 mM and buffer B: 400 mM sodium perchlorate,
20 mM Tris–HCl (pH 6.8), 25% formamide, flow rate 3 ml min–1
575
Nucleic Acids
Acids Research,
Research,1994,
1996,Vol.
Vol.22,
24,No.
No.14
Nucleic
575
with gradients of 15–45%B over 20 min (39). Desalting was
achieved via extensive dialysis against water. The purity of
oligoribonucleotides was assayed by 5′-32P-labelling (43) followed by electrophoresis on a 20% polyacrylamide gel (PAGE)
in the presence of 7 M urea.
Characterization of modified oligonucleotides
To verify the presence of the modified residues, 32P-end-labelled
oligoribonucleotides were subjected to partial alkaline hydrolysis. Each sample (∼20 pmol) was dissolved in 15 µl 1 M sodium
carbonate/sodium bicarbonate solution (pH 10) and incubated at
80C over 5 min. After cooling on ice, 1 µl 2 M HCl solution was
added and the mixture was kept at room temperature for 2 h.
Finally, 0.5 µl 4 M sodium hydroxide solution was added to the
reaction mixture and a 2 µl aliquot of each sample was subjected
to electrophoresis on a 20% denaturing polyacrylamide gel for 2
h at 30 W followed by autoradiography. The presence of a
2′-deoxynucleotide or a C3-spacer was indicated by a gap at the
modified position within the sequence ladder. By contrast in the
sequence ladder for oligoribonucleotides containing an abasic
residue, a band of enhanced intensity was seen at the site of
modification, presumably due to better accessibility of the abasic
residue to alkali.
Determination of ribozyme steady state parameters
A stock solution equimolar in ribozyme strand A and ribozyme
strand B (50 nM) was prepared as well as a 2 µM stock solution
of 32P-labelled substrate strand, each in 40 mM Tris–HCl (pH
7.5) and incubated separately at 90C for 1 min followed by
cooling to 37C for 15 min. The ribozyme stock solution was
adjusted to 10 mM MgCl2 and incubated at 37C for 15 min. The
cleavage reactions were carried out at 37C by adding appropriate volumes of substrate strand solution to a reaction mixture
containing the ribozyme strands followed by brief vortexing.
Final concentrations were 2–20 nM ribozyme strands and 20–300
nM substrate strand (for unmodified ribozyme or faster cleaving
modified ribozymes) or 5–30 nM ribozyme strands and 20–300
nM substrate strand (slower cleaving modified ribozymes) in a
final volume of 100 µl, 10 mM MgCl2, 40 mM Tris–HCl (pH 7.5).
Aliquots (10 µl) of the reaction mixture were removed at six
suitable time intervals and the reactions quenched by addition to
20 µl of stop mix (7 M urea, 50 mM EDTA, 0.04% xylene cyanol,
0.04% bromophenol blue). The samples were analysed by
polyacrylamide gel electrophoresis on 20% denaturing gels, the
resultant gels dried and subjected to scanning using a PhosphorImager (Molecular Dynamics). The data were processed using the
programme Image Quant (Molecular Dynamics) and quantitated
by use of the Geltrak programme (44; Smith, J. and Singh, M.,
manuscript submitted).
Kinetic parameters were determined from non-linear regression fitting of the data to the Michaelis–Menten equation.
Thus:
v k cat [S]
[E]
K M [S]
where v = initial rate, [S] = substrate concentration and [E] =
total enzyme concentration. The values of kcat and KM are shown
in Table 1.
Figure 2. Structures of (a) a wild-type ribonucleoside, (b) a 2′-deoxynucleoside, (c) an abasic residue and (d) a C-3 spacer (propyl linker).
Magnesium ion titrations
The effect of increasing magnesium ion concentration on the kinetic
parameters of the cleavage reaction was determined using a substrate
concentration of 200 nM and an enzyme concentration of 10 nM (for
fast-cleaving ribozymes) or 25 nM (for slow-cleaving ribozymes) at
37C and with magnesium chloride concentrations ranging from 2
to 150 mM. The Tris concentration was 40 mM. Rate constants for
cleavage were calculated by measuring the initial cleavage rates at
varying magnesium ion concentrations (Table 2). The apparent
magnesium dissociation constant K Mg2 was calculated graphically
from the magnesium ion concentration at which the observed
cleavage rate was half-maximal.
RESULTS
Assembly of modified hairpin ribozymes and
measurement of kinetic parameters
A three-stranded ribozyme was chosen for the nucleotide
substitution studies (Fig. 1). This is identical to that used in our
previous purine nucleotide substitution studies (33) and as also
utilised in investigations of the effect of 2′-deoxynucleotide
substitution in ribozyme strand A (34). In this way the new results
could be compared directly with previous studies in consideration
of loop B structure. We decided to concentrate in particular on the
nucleotides in loop B located on ribozyme strand B. There was no
previous information as to the requirements of individual
2′-hydroxyl groups in this strand. In addition, although we had
previously obtained functional group data on the three A residues
(A38, A40 and A43) (33), very little information was available as
to the base requirements of the five pyrimidine residues present
in this strand.
As in our previous study (33), we chose to prepare a fully
chemically synthetic hairpin ribozyme, which allows for the
ready site-specific substitution of modified nucleotides. Each
analogue could then be inserted into the required oligoribonucleotide strand merely by the replacement in stepwise solid-phase
synthesis of an unmodified nucleoside phosphoramidite derivative by the appropriately modified nucleoside phosphoramidite.
576
Nucleic Acids Research, 1996, Vol. 24, No. 4
Each of the eight nucleotides in ribozyme strand B of loop B was
replaced by either a 2′-deoxynucleotide, an abasic residue
(1-deoxy-D-ribofuranose) or a C3-spacer (propyl linker) (Fig. 2).
The use of these three analogues tests the requirements respectively of the 2′-hydroxyl group, the heterocyclic base, and the
sugar and base together. Partly as a control, we also prepared the
same analogues in ribozyme strand A at positions A20 and G21 on
the opposite side of loop B, since 2′-deoxynucleotide substitution
data was already available for these sites (34).
Hairpin ribozymes were assembled from three strands of
synthetic oligoribonucleotide, two strands being unmodified and
the third containing a single modified residue at the appropriate
position (Fig. 1). Kinetic parameters for the modified and
unmodified ribozymes were obtained from determination of the
initial rate of reaction at concentrations of substrate around the
KM (Table 1). When modification resulted in slow cleavage, the
concentration of enzyme was increased but was always at least
five times lower than the substrate concentration. The reaction
velocities, once normalised to a fixed enzyme concentration,
followed a conventional Michaelis–Menten curve. For each
analogue, six individual cleavage reactions were set up under
different substrate/ribozyme concentration conditions with six
time points per reaction (see Materials and Methods). Thus for the
30 analogue substitutions studied there were a total of 1080
separate measurements of the amounts of cleaved and uncleaved
substrate.
Table 1. Kinetic data for cleavage of the hairpin ribozyme containing analogues in loop B
Position
RzA A20
RzA G21
RzB U37
RzB A38
RzB U39
RzB A40
RzB U41
RzB U42
RzB A43
RzB C44
anm,
Substitution
KM
kcat
kcat/KM
µM
min–1
(rel)
none
0.02
(±0.01)
0.23
(±0.01)
1.0
2′-deoxy
0.02
(±0.01)
0.45
(±0.03)
1.96
abasic
0.06
(±0.01)
0.0094
(±0.0006)
0.014
propyl
0.010
(±0.001)
0.0022
(±0.0001)
0.019
2′-deoxy
0.03
(±0.01)
0.11
(±0.01)
0.32
abasic
0.05
(±0.02)
0.0012
(±0.0002)
0.002
propyl
nma
2′-deoxy
0.15
(±0.07)
0.13
(±0.03)
0.076
abasic
0.015
(±0.003)
0.0196
(±0.0008)
0.11
propyl
0.17
(±0.06)
0.12
(±0.02)
0.061
2′-deoxy
0.06
(±0.01)
0.015
(±0.001)
0.022
abasic
0.07
(±0.04)
0.0010
(±0.0002)
0.001
propyl
0.010
(±0.001)
0.0012
(±0.0001)
0.01
2′-deoxy
0.08
(±0.01)
0.35
(±0.02)
0.38
abasic
0.04
(±0.02)
0.40
(±0.06)
0.87
propyl
0.028
(±0.004)
0.66
(±0.02)
2.05
2′-deoxy
0.03
(±0.01)
0.45
(±0.05)
1.3
abasic
0.07
(±0.01)
0.0017
(±0.0001)
0.002
(<0.0002)
propyl
0.010
(±0.001)
0.0015
(±0.0001)
0.013
2′-deoxy
0.04
(±0.01)
0.0239
(±0.0020)
0.052
abasic
0.03
(±0.01)
0.0010
(±0.0001)
0.003
propyl
0.010
(±0.005)
0.0006
(±0.0001)
0.005
2′-deoxy
0.011
(±0.005)
0.74
(±0.04)
5.85
abasic
0.04
(±0.02)
0.0034
(±0.0005)
0.007
propyl
0.025
(±0.004)
0.0025
(±0.0001)
0.009
2′-deoxy
0.025
(±0.005)
0.57
(±0.02)
1.98
abasic
0.04
(±0.01)
0.0013
(±0.0001)
0.003
propyl
0.049
(±0.005)
0.0011
(±0.0001)
0.002
2′-deoxy
0.03
(±0.02)
0.41
(±0.02)
1.19
abasic
0.04
(±0.01)
0.17
(±0.01)
0.37
propyl
0.10
(±0.05)
0.021
(±0.004)
0.018
not measureable (the cleavage rate was too slow to determine KM).
577
Nucleic Acids
Acids Research,
Research,1994,
1996,Vol.
Vol.22,
24,No.
No.14
Nucleic
577
Table 2. Magnesium ion dependence of cleavage of hairpin ribozymes
containing analogues
Position
RzA A20
RzA G21
RzB U37
RzB A38
Figure 3. Relative cleavage efficiencies (kcat/KM) of hairpin ribozymes
containing analogues in loop B (d = 2′-deoxy, a = abasic, p = propyl linker).
Those highlighted in magenta reflect wild-type or close to wild-type cleavage.
Undetectable cleavage was found for propyl linker substitution at G21 (yellow).
For each analogue, the relative catalytic cleavage efficiency
(kcat/KM) (Table 1 and Fig. 3) was calculated, from which the
change in apparent binding energy (G ]app) resultant from the
modification may be obtained (45). In general, the cleavage
efficiencies can be considered in three groups: (i) those close to
wild-type or better (kcat/KM rel. 0.3–6.0, G ]app 3.0 to –5.0 kJ.
Mol–1), (ii) those only modestly affected (kcat/KM rel. 0.05–0.3,
G ]app 8.0–3.0 kJ. Mol–1), and (iii) those whose cleavage
efficiency is drastically affected (kcat/KM rel. <0.05, G ]app >8.0
kJ. Mol–1). Apart from the three analogues at U37 and the 2′-deoxy
analogue at U41, all other analogues fell into either the first or third
category. In this third category, it is very likely that hydrogen
bonding is affected as a result of the modification, whereas in the
first category it is not likely that hydrogen bonding is affected.
We also carried out experiments to study the magnesium
dependence of cleavage for the unmodified and modified ribozymes. Values for the apparent K Mg2 and the fold increase in
cleavage rate between 5 and 150 mM Mg2+ are shown in Table 2.
RzB U39
RzB A40
RzB U41
RzB U42
RzB A43
RzB C44
anm,
Increase in k between
5 and 150 mM Mg2+
x-fold
KMg2+app
mM
none
5
10
2′-deoxy
6
15
abasic
5
15
propyl
13
35
2′-deoxy
23
25
abasic
8
35
propyl
nma
nma
2′-deoxy
24
15
abasic
4
10
propyl
12
30
7
40
2′-deoxy
abasic
29
35
propyl
5
25
17
25
2′-deoxy
abasic
4
20
propyl
3
10
2′-deoxy
5
10
abasic
15
35
propyl
20
35
2′-deoxy
8
40
abasic
6
15
propyl
13
40
2′-deoxy
13
15
5
25
abasic
Ribozyme strand A substitution results
It has previously been shown that A20 can be substituted by either
C, U or G without significant loss of cleavage ability (16). Also
2′-deoxynucleotide substitution was previously shown to be
non-deleterious [relative kcat/KM 0.3 determined under single
turnover conditions (34)]. Our results for deoxy-substitution at
A20 (Table 1) show wild-type cleavage (note that a factor of 2 is
within experimental error). Abasic and propyl substitution each
resulted in a 50–70-fold drop in catalytic efficiency. This suggests
that the base may play a structural role in either providing
stacking potential, or possibly in providing some non-specific
hydrogen bonding. By contrast, the sugar does not appear to have
a significant hydrogen-bonding or conformational role, since the
catalytic efficiency of the propyl analogue is not further decreased
compared with the abasic residue.
Substitution
propyl
16
10
2′-deoxy
16
15
abasic
5
10
propyl
4
15
21
25
2′-deoxy
abasic
4
10
propyl
5
25
not measurable (the cleavage rate was too slow to determine K
).
Mg 2app
The identity of G21 is essential for cleavage, since replacement
by either A, C or U has been shown to be harmful (16). As
expected therefore, substitution of G21 by an abasic residue
resulted in a 500-fold drop in cleavage efficiency. Clearly this
base is involved in important hydrogen bonding and possibly in
stacking interactions. Deoxy-substitution had an insignificant
effect and the value of kcat/KM rel. was identical (0.3) to that
previously reported (34). By contrast, a dramatic loss of all
cleavage activity to below detectable levels was found for the
propyl linker substitution. This suggests that the sugar of G21
plays a vital conformational or other structural role in addition to
578
Nucleic Acids Research, 1996, Vol. 24, No. 4
the role of the base. No other residue that we have tested in loop
B has shown this dramatic behaviour upon propyl substitution
and this demonstrates the crucial role this residue plays in hairpin
ribozyme cleavage. None of the analogues in positions A20 and
G21 showed a significant effect of magnesium ion (Table 2).
Ribozyme strand B substitution results
U37 is not amongst those nucleotides previously identified as
being essential for hairpin cleavage. Our results (Fig. 3 and Table
1) show a 10-fold reduction in cleavage efficiency for the abasic
analogue, reflected entirely in kcat reduction. Thirteen to sixteenfold reductions in cleavage efficiency were found for both the
2′-deoxynucleoside and the propyl linker, each primarily due to
increases in KM. These modest though significant effects show
that both the base and sugar have minor structural roles to play in
attaining an active conformation, yet are unlikely to play a
significant part in catalysis.
Substitution of A38 by 2′-deoxy or propyl resulted in 50- and
100-fold reductions in cleavage respectively, the majority of these
effects being seen in kcat. Abasic residue substitution led to a
1000-fold decrease in kcat/KM, the only position where abasic
subsititution was more harmful than propyl substitution and
which also showed the largest increase in cleavage rate (6-fold
above the unmodified hairpin) when magnesium ion concentration was increased (Table 2). Both base and sugar therefore must
play important roles. Since the correct identity of the base at A38
is essential also (15,16) and both purine and N7-deaza substitution are also harmful (33), this base is likely to be involved in
hydrogen-bonding contacts.
U39 is the only position we have found in loop B where none
of the three substitution mutants had a significant effect on
catalytic cleavage. It is clear from the tolerance we have found to
both abasic and propyl linker substitution that U39 merely acts as
a spacer and plays no significant role either in folding or catalysis.
The results at A40 show that, whereas there is no effect of
deoxy-substitution, substitution by an abasic residue or propyl
linker resulted in 100-fold or more reductions in cleavage
efficiency. Thus the presence of a base here is required but, when
taken together with other data (see below), do not indicate
specific base-pair formation required for catalytic activity.
The identities of U41 and U42 are amongst those said to be
essential for cleavage based on the fact that no mutations at these
sites were found in selection experiments (15). Our results show
that deoxy-substitution of U41 results in a 20-fold reduction in
cleavage efficiency. Similar substitution at U42 results in a small
but significant increase, due to improvements in both kcat and KM.
This is the only position that we have found that is actually
improved by analogue substitution and may reflect a preference
of the sugar for the 2′-endo configuration. By contrast, both
abasic and propyl substitution resulted in >100-fold drops in
catalytic efficiency for both U positions. Thus it is clear that these
two residues play important structural roles and do not act merely
as spacers.
Both abasic and propyl substitution at A43 causes
300–500-fold loss of activity and it is clear that this residue must
play an important role. However, the 2′-hydroxyl group is not
involved since deoxy-substitution was without effect. Surprisingly, neither abasic nor propyl mutants showed any significant
effect of magnesium ion concentration on cleavage rate, in
contrast to what has been previously found for N7-deazaA43
where K Mg2app was seen to increase 5-fold (33). A similar result
was found also for all other analogues tested in strand B (Table
2). An insignificant effect of abasic substitution at C44 was found
and deoxy-substitution was also tolerated. By contrast, propyl
linker substitution caused a 50-fold decrease in catalytic efficiency. These results demonstrate that the base at this position is
not required for hairpin catalytic cleavage.
DISCUSSION
Analogue substitution in the hairpin ribozyme
To determine the requirements for both sugar and base in loop 2
of the hairpin ribozyme we chose to use three analogues, namely
a 2′-deoxynucleoside, an abasic residue, and a C3-spacer (propyl
linker). Single 2′-deoxynucleoside substitution has been used
previously to test the requirements for 2′-hydroxyl groups in the
case of the hammerhead ribozyme (46,47) and also more recently
in parts of the hairpin ribozyme (34).
The synthesis and use of an abasic residue was recently
described as a replacement for stem–loop II of the hammerhead
ribozyme (35). It was further used as a replacement for two of the
conserved U residues in the catalytic core of the hammerhead
ribozyme and in the case of U4 replacement there was hardly any
effect observed in hammerhead ribozyme cleavage (36). This was
interesting because the nucleotide at the cleavage site of the
hammerhead (C17) was found in the crystal structure to make a
stacking interaction with the exocyclic O-2 of U4 (3 Å
aromatic-nπ-interaction) (12). This demonstrates that abasic
chemical substitution information can be helpful in assessing
whether base interactions seen in the ground state (crystal
structure) are also likely to be present in the transition state
(cleavage).
For those residues where a heterocylic base has little or no
involvement in attaining a catalytically active structure, substitution by a C3-spacer (propyl linker) tests the need for the
2′-hydroxyl group and/or sugar conformation in maintaining a
particular sugar–phosphate backbone geometry. Tolerance to
propyl substitution shows that a residue acts merely as a structural
separator between the adjacent nucleotides, as has been found in
HIV-1 for two pyrimidine residues in the bulge of the transactivation response region (TAR) RNA for recognition by the tat
protein (48) and of a U residue in the ‘bubble’ structure of the
rev-responsive element (RRE) RNA for recognition by the rev
protein (38). The number and types of atoms separating
sequential phosphates is identical for both a ribose and a
C3-spacer.
More precise functional group mapping can be effected with
base analogues. For the hairpin ribozyme, this has been effectively carried out for essential purine residues (32,33). By contrast,
although one useful pyrimidine analogue (2-pyrimidinone-1-βD-riboside) which is an exocyclic deletion mutant of both uridine
and cytosine has recently been prepared as an amidite suitable for
oligoribonucleotides synthesis (49), a range of pyrimidine
analogues that satisfactorily tests hydrogen-bonding potential is
not yet available as amidite derivatives.
Requirements for sugars and bases in loop B
In strand A of the hairpin ribozyme, the identities of residues G21,
A22, A23, A24 and C25 were previously thought to be essential
(16,18). Our new data for A20 and G21 are consistent with this.
579
Nucleic Acids
Acids Research,
Research,1994,
1996,Vol.
Vol.22,
24,No.
No.14
Nucleic
On the three A residues, each of the exocyclic amino groups and
two of the N7-positions (A22 and A23) are required for efficient
cleavage (33). For G21, 100-fold reductions in catalytic efficiency
were observed for inosine (loss of exocyclic amino group) or
N7-deaza-substitution (33).
Our data concur also with previous results showing that neither
hydroxyl group is required at residues A20 and G21 (34). The
same authors have also found that deoxy-substitution of either
A22 or A23 had no significant effect but very large
(300–1000-fold) reductions in catalytic efficiency were noted for
both 2′-deoxy and 2′-O-methyl substitution at positions A24 or
C25 (34). Activities of the A24 analogues could be rescued by
increasing magnesium ion concentration suggesting an important
contact of the hydroxyl group to magnesium in the transition state
of the reaction.
For strand B, U37 was thought previously not be essential,
based on the wild-type activity of a U37 deletion mutant (16).
However, this isolate also contained simultaneous G36C and
U38G mutations. Three other cases of near wild-type cleavage
were reported where U37 had been changed to A or C, but G36 was
also mutated in these constructs (16). Other mutants of U37 were
also double with either A24 (16,18) or A43 (18) and both these
double mutants were inactive. Thus no single point mutation of
U37 has been reported. U37 was also not included in in vitro
selection experiments (15,17). Our new results are consistent
with a minor, yet significant role of both base and sugar.
The important role of the base of A38 in catalysis (15,16) is
confirmed by our results and in addition there is a possible contact
involving the 2′-hydroxyl group. By contrast, neither base nor
sugar of U39 plays a significant role. This agrees with and extends
the previous observations that U39C is a general up mutant (16)
and that both C and A were selected as active alternatives in in
vitro selection experiments (15).
The role of A40 has been difficult to assess up to now. Single
point mutations at this site that maintained activity were not
selected in vitro, but a double mutant U39C, A40C was active at
a reduced level (15). Substitution of A40 by U or G has been
reported to result in a reduced (3%) activity in each case (18) but
no significant reductions in cleavage were noted for either purine
substitution or N7-deazaA substitution (33). Our results confirm
that specific base-pair formation is unlikely but that the base
nevertheless plays an important structural role. The only tolerated
substitutions at positions 41 and 42 are U41C and U42C having 25
and 3% cleavage activities respectively (18). U41 is also the site
shown to be capable of intra-strand UV cross-linking to G21 (26).
Our results now show strong evidence for the participation of
these residues in base-pairing.
Another residue which has been somewhat difficult to assess is
A43. This residue is not invariant in in vitro selection, A43G being
found regularly, preferably associated with the U39C up mutation
(15). However A43U was inactive (18). Purine substitution had
no significant effect which is indicative of no strong intra-loop
base-pairing. By contrast, N7-deazaA substitution was extremely
harmful (33). Since the K Mg2app of the N7-deazaA mutant was
increased 5-fold, it was thought previously that the N7-position
might be a magnesium binding site in the ground state. Our new
data are more consistent with the N7-position being involved in
a hydrogen-bonding interaction of some sort. Note that the
N7-position is preserved by G substitution and by purine
substitution but not by substitution with other bases. C44 had been
579
Figure 4. Model for the secondary structure of loop B of the hairpin ribozyme
showing the three proposed non-canonical base-pairs (dashed) and their
possible types. The arrow denotes the UV cross-linkable residues and the box
(green) shows the homology with other known UV-sensitive RNA regions (26).
Two other possible non-canonical pairs are shown (dotted), but there is
insufficient evidence for these to be formally proposed. Residues U39 and C44
which are not likely to be involved in base-pairing are shown in magenta.
proposed to be an invariant nucleotide based on the fact that no
mutations were found at this site during in vitro selection (16).
However, no C44 mutagenesis data has been reported either. Our
results show clearly for the first time that the base at this position
plays no significant role in hairpin cleavage.
Implications for the secondary structure of loop B
Two models have been suggested recently for the secondary
structure of sections of loop B. Butcher and Burke have proposed
three non-canonical base-pairs (G21:A43, A22:U42 and A23:A40)
based on homology with other UV sensitive RNA domains and
with interpretation of accessibility of the region to chemical
reagents. Our previous results showing the tolerance of A40 and
A43 to removal of the exocyclic amino group (33) were not
consistent with this model. We proposed instead that Watson–
Crick pairing might take place in the lower part of loop B between
residues A24:U37 and non-canonical pairing between A23:A38
and either A22 with U39 or U41. However, our new data is not
consistent with a A22:U39 pair. Nor is the relatively small
(10-fold) effect of abasic site substitution at U37 consistent with
a strong Watson–Crick pair involving U37. Thus this model too
is unsatisfactory.
Two clear results from our data are that neither U39 nor C44 is
likely to be involved in base-pairing interactions. Preclusion of
these two residues in possible base-pairing schemes limits the
remaining possibilities. Having considered all the data from our
new results together with the previous studies, we now propose
a new model of loop B secondary structure (Fig. 4). In this model,
the major elements of UV-sensitive RNA domains are preserved,
namely (i) a reverse-Hoogsteen pair between U41 and A22,
580
Nucleic Acids Research, 1996, Vol. 24, No. 4
located within the sequence GAA, (ii) the same relative locations
of U41 to G21, the cross-linked residues, and (iii) a bulged base
below U41. In addition to A22:U41, two other non-canonical base
pairs, G21:U42 and A24:A38, are also proposed.
This model takes account of the tolerance of U41 substitution
by C (18) in that a reverse-Hoogsteen A:C pair has very similar
geometry to that of an A:U pair. That U42 can also be substituted
by C, albeit at much reduced efficiency, may be because the
resultant Watson–Crick or reverse-Watson–Crick pair (G21:C42)
would be stronger (and perhaps less flexible) than the wild-type
wobble or reverse wobble G21:U42 pair. The third non-canonical
base-pair (A24:A38) is suggested mainly by the substitution data
at these sites indicating the potential for hydrogen bonding.
However, the tolerance to N7-deaza substitution of A24 (33)
would limit the pairing schemes either to N1-amino symmetrical
or to N1-aminoA24:N7-aminoA38. The two bases on both strands
below could also form non-canonical base pairs (C25:U37 and
A26:G36) but insufficient data is available to make predictions at
these sites.
Another aspect of this model is that it predicts two small
‘bubble’ regions where a single nucleoside is sited opposite two
nucleosides. One bubble includes A20, the residue which may be
substituted by any of the other three nucleosides, and C44, which
we have shown to make no base contacts required for cleavage.
A43 is also located in this bubble. The N7-position of the base is
clearly important (33) and in principle it could be involved in
some unusual cross-strand or intra-domain contact. The second
bubble includes on one side the spacer residue, U39, and the
residue A40 which has tolerance for purine and N7-deaza-substitution (33) and which is not highly inactivated when substituted
by other nucleosides (18). Opposite these residues is A23 which
is the only residue of the six suggested to be within bubbles that
is both completely invariant and which cannot be substituted by
either purine or N7-deazaA. Thus cross-strand pairing of this
residue cannot be ruled out, but if so, it is unlikely to be with either
U39 or A40 as outlined above. A rearrangement of the lower stem
would be necessary juxtaposing A23 with A38. However, this
would put A24 and C25 now in positions to Watson–Crick pair
with the other strand, which has been shown to be unlikely (16).
How does this model fit with the chemical accessibility results
previously reported (25)? Relatively good accessibility of
residues A23, A38, A40 and A43 was noted to dimethyl sulphate
(N1-modification). Three of these residues are predicted to be
within bubbles. The fourth would be accessible at N1 if paired
with A24 in one of the two likely schemes using its N7 and amino
groups. Diethylpyrocarbonate (N7-modification) sensitivity was
generally low, although A20, A22, A23, A38 and A40 were more
accessible than other positions. Again three of these residues are
within bubbles. Interestingly A43 was not sensitive to the reagent
even in the absence of substrate. This is consistent with the
involvement of A43 N7 in an important or unusual hydrogen
bonding contact, although this seems unlikely to be with A20.
Modification of G21 by either kethoxal (N1, N2 modification) or
by a bulky nickel reagent (N7-modification) was suppressed at
G21 following magnesium addition, suggesting a high degree of
structure at this site. Thus G21 is likely to be involved in
cross-strand interactions. Finally residues U37, U39, U41 and U42
were found to be accessible to a carbodiimide reagent
(N3-modification) with U41 and U42 being noticeably protected
in the presence of magnesium ion and U39 being the most
accessible with or without magnesium and/or substrate. Thus,
there is reasonable consistency of this accessibility data with our
model. However, it should also be borne in mind that non-canonical pairs across a large RNA structure such as loop B would still
allow the RNA to be relatively flexible and thus some chemical
reagents would be expected to have greater accessibility to
residues in non-canonical pairing than in regions of tight
Watson–Crick pairing.
Butcher and Burke also observed that a large change in
chemical accessibility of loop B took place when magnesium ion
was added but there were only smaller effects when substrate was
added (25). Thus in the presence of magnesium, the ribozyme
must adopt a defined structure independent of substrate binding.
Magnesium ion then also plays a second and distinct role in
achieving a catalytically proficient complex (24). A caveat of the
chemical accessibility data as well as our own chemical
substitution data is that interpretation is complicated by the
possibility of inter-domain interactions as well as intra-strand
hydrogen bonding. We have interpreted the substitution data of
loop B primarily in terms of intra-loop secondary structure,
because it seems likely that the UV-sensitivity of internal loops is
a consequence of the attainment of a particular RNA structural
element. However, it is possible that some of the drastic
reductions in cleavage rate observed for certain modified residues
might be due instead to loss of important inter-domain contacts.
It is hard from our data to distinguish such residues with any
certainty, but the most likely candidates are G21, A23, A40 and A43
(N7). In addition, we have shown previously that the number of
critical functional groups in purine residues in the non-substrate
strand of loop A is too large to be rationalised merely by
intra-strand hydrogen-bonding alone (33). However, alternative
techniques will be needed to define unequivocally inter-domain
interactions present during hairpin ribozyme cleavage.
The secondary structure model of loop B is the best fit possible
taking into account all the data available. Other pairing schemes
can be reconciled only with subsets of the data. There are few
other nucleoside analogues available which can test this model
further, although one possibility would be the replacement of U41
by 2-pyrimidinone-1-β-D-riboside (49). If this residue is involved
in reverse-Hoogsteen pairing, substitution by this analogue
should be tolerated. Confirmation of loop B structure is perhaps
best obtained via NMR or crystallography, which will hopefully
be forthcoming in the future. Meanwhile, our model should be
useful towards the molecular modelling of the hairpin ribozyme.
In addition, the complete data set of 2′-deoxy-substitution in loop
B will aid ongoing experiments in this laboratory aimed at
defining inter-domain distances via cross-linking through
2′-tethered disulphide formation (50). Further, abasic and propyl
substitution data should prove valuable for future design of
hairpin ribozymes as potential therapeutic agents.
ACKNOWLEDGEMENTS
We thank Richard Grenfell, Jan Fogg and Terry Smith for
assembly and purification of oligoribonucleotides and Mohinder
Singh for technical assistance. We acknowledge grateful support
to DAAD (Germany) for a fellowship grant to Sabine Schmidt.
REFERENCES
1 Feldstein, P.A., Buzayan, J.M. and Bruening, G. (1989) Gene, 82, 53–61.
2 Hampel, A. and Tritz, R. (1989) Biochemistry, 28, 4929–4933.
581
Nucleic Acids
Acids Research,
Research,1994,
1996,Vol.
Vol.22,
24,No.
No.14
Nucleic
3 Haseloff, J. and Gerlach, W.L. (1989) Gene, 82, 43–52.
4 Burke, J.M. (1994) In Eckstein, F. and Lilley, D.M.J. (ed.), Nucleic Acids
and Molecular Biology. Springer-Verlag, Berlin, pp. 105–118.
5 Symons, R.H. (1994) Curr. Opin. Struct. Biol., 4, 322–330.
6 Tuschl, T., Thomson, J.B. and Eckstein, F. (1995) Curr. Opin. Struct. Biol.,
5, 296–302.
7 Yu, M., Leavitt, M.C., Maruyama, M., Young, D., Ho, A.D. and
Wong-Staal, F. (1995) Proc. Natl. Acad. Sci. USA, 92, 699–703.
8 Marschall, P., Thomson, J.B. and Eckstein, F. (1994) Cell. Mol. Neurobiol.,
14, 523–538.
9 Christoffersen, R.E. and Marr, J.J. (1995) J. Med. Chem., 38, 2023–2037.
10 Tuschl, T., Gohlke, C., Jovin, T.M., Westhof, E. and Eckstein, F. (1994)
Science, 266, 785–789.
11 Pley, H.W., Flaherty, K.M. and Mckay, D.B. (1994) Nature, 372, 68–74.
12 Scott, W.G., Finch, J.T. and Klug, A. (1995) Cell, 81, 991–1002.
13 Hampel, A., Tritz, R., Hicks, M. and Cruz, P. (1990) Nucleic Acids Res.,
18, 299–304.
14 Chowrira, B.M. and Burke, J.M. (1991) Biochemistry, 30, 8518–8522.
15 Berzal-Herranz, A., Joseph, S. and Burke, J.M. (1992) Genes Dev., 6,
129–134.
16 Berzal-Herranz, A., Joseph, S., Chowrira, B.M., Butcher, S.E. and Burke,
J.M. (1993) EMBO J., 12, 2567–2574.
17 Joseph, S. and Burke, S. (1993) J. Biol. Chem., 268, 24515–24518.
18 Anderson, P., Monforte, J., Tritz, R., Nesbitt, S., Hearst, J. and Hampel, A.
(1994) Nucleic Acids Res., 22, 1096–1100.
19 Feldstein, P.A. and Bruening, G. (1993) Nucleic Acids Res., 21,
1991–1998.
20 Komatsu, Y., Koizumi, M., Nakamura, H. and Ohtsuka, E. (1994) J. Amer.
Chem. Soc., 116, 3692–3696.
21 Komatsu, Y., Kanzaki, I., Koizumi, M. and Ohtsuka, E. (1995) J. Mol.
Biol., 252, 296–304.
22 Butcher, S.E., Heckman, J.E. and Burke, J.M. (1995) J. Biol. Chem., 270,
29648–29652.
23 Vitorino Dos Santos, D., Fourrey, J.-L. and Favre, A. (1993) Biochem.
Biophys. Res. Comm., 190, 377–385.
24 Chowrira, B.M., Berzal-Herranz, A. and Burke, J.M. (1993) Biochemistry,
32, 1088–1095.
25 Butcher, S.E. and Burke, J.M. (1994) J. Mol. Biol., 244, 52–63.
26 Butcher, S.E. and Burke, J.M. (1994) Biochemistry, 33, 992–999.
27 Doudna, J.A., Szostak, J.W., Rich, A. and Usman, N. (1990) J. Org.
Chem., 55, 5547–5549.
28 Talbot, S.J., Goodman, S., Bates, S.R.E., Fishwick, C.W.G. and Stockley,
P.G. (1990) Nucleic Acids Res., 18, 3521–3528.
581
29 Usman, N. and Cedergren, R. (1992) Trends Biochem. Sci., 17, 334–339.
30 Grasby, J.A. and Gait, M.J. (1994) Biochimie, 76, 1223–1234.
31 Grasby, J.A., Singh, M., Karn, J. and Gait, M.J. (1995) Nucleosides
Nucleotides, 14, 1129–1132.
32 Chowrira, B.M. and Burke, J.M. (1991) Nature, 354, 320–322.
33 Grasby, J., Mersmann, K., Singh, M. and Gait, M.J. (1995) Biochemistry,
34, 4068–4076.
34 Chowrira, B.M., Berzal-Herranz, A., Keller, C.F. and Burke, J.M. (1993) J.
Biol. Chem., 268, 19458–19462.
35 Beigelman, L., Karpeisky, A. and Usman, N. (1994) Biorg. Med. Chem.
Lett., 4, 1715–1720.
36 Beigelman, L., Karpeisky, A., Matulic-Adamic, J., Gonzalez, C. and
Usman, N. (1995) Nucleosides Nucleotides, 14, 907–910.
37 Gait, M.J., Pritchard, C.E. and Slim, G. (1991) In Eckstein, F. (ed.),
Oligonucleotides and Analogues: A Practical Approach. Oxford
University Press, Oxford, UK, pp. 25–48.
38 Pritchard, C.E., Grasby, J.A., Hamy, F., Zacharek, A.M., Singh, M., Karn,
J. and Gait, M.J. (1994) Nucleic Acids Res., 22, 2592–2600.
39 Schmidt, S., Grenfell, R.L., Fogg, J., Smith, T.V., Grasby, J.A., Mersmann,
K. and Gait, M.J. (1996) In Epton, R. (ed.), Innovation and Perspectives in
Solid Phase Synthesis. Mayflower Scientific Ltd, Birmingham, UK, in
press.
40 Sproat, B., Colonna, F., Mullah, B., Tsou, D., Andrus, A., Hampel, A. and
Vinayak, R. (1995) Nucleosides Nucleotides, 14, 255–273.
41 Wincott, F., DiRenzo, A., Shaffer, C., Grimm, S., Tracz, D., Workman, C.,
Sweedler, D., Gonzalez, C., Scaringe, S. and Usman, N. (1995) Nucleic
Acids Res., 23, 2677–2684.
42 Gasparutto, D., Livache, T., Bazin, H., Duplaa, A.-M., Guy, A., Khorlin,
A., Molko, D., Roget, A. and Teoule, R. (1992) Nucleic Acids Res., 19,
5159–5166.
43 Slim, G. and Gait, M.J. (1991) Nucleic Acids Res., 19, 1183–1188.
44 Smith, J.M. and Thomas, D.J. (1990) Comp. Appl. Biosci., 6, 93–99.
45 Fersht, A.R. (1988) Biochemistry, 27, 1577–1580.
46 Perreault, J.-P., Wu, T., Cousineau, B., Ogilvie, K.K. and Cedergren, R.
(1990) Nature, 344, 565–567.
47 Perreault, J.-P., Labuda, D., Usman, N., Yang, J.-H. and Cedergren, R.
(1991) Biochemistry, 30, 4020–4025.
48 Sumner-Smith, M., Roy, S., Barnett, R., Reid, L.S., Kuperman, R., Delling,
U. and Sonnenberg, N. (1991) J. Virol., 65, 5196–5201.
49 Adams, C.J., Murray, J.B., Arnold, J.R.P. and Stockley, P.G. (1994)
Tetrahedron Lett, 35, 1597–1600.
50 Sigurdsson, S., Tuschl, T. and Eckstein, F. (1995) RNA, 1, 575–583.