ABSTRACT
Ricin is a cytotoxic plant protein that inactivates ribosomes by hydrolyzing the
N-glycosidic bond at position A4324 in eukaryotic 28S rRNA. Recent studies
showed that a four-nucleotide loop, GAGA, can function as a minimum substrate for ricin (the
first adenosine corresponds to the site of depurination). We previously
clarified the solution structure of this loop by NMR spectroscopy [Orita
et al.
(1993)
Nucleic Acids Res.
21, 5670-5678]. To elucidate further details of the structural basis for recognition of its substrate by ricin, we studied
the properties of a synthetic dodecanucleotide, r1C2U3C4A5G6dA7G8A9U10G11A12G
(6dA12mer), which forms an RNA hairpin structure with a GdAGA loop and in which the site
of depurination is changed from adenosine to 2
'
-deoxyadenosine. The N-glycosidase activity against the GdAGA loop of the A-chain of ricin was 26 times higher than that against the GAGA
loop. NMR studies indicated that the overall structure of the GdAGA loop was
similar to that of the GAGA loop with the exception of the sugar puckers of 6dA
and 7G. Therefore, it appears that the 2
'
-hydroxyl group of adenosine at the depurination site (6A) does not
participate in the recognition by ricin of the substrate. Since the 2
'
-hydroxyl group can potentially destabilize the developing positive charge
of the putative transition state intermediate, an oxycarbonium ion, the
electronic effect may explain, at least in part, the faster rate of
depurination of the GdAGA loop compared to that of GAGA loop. We also show that
the amino group of 7G is essential for substrate recognition by the ricin A-chain.
Ricin is a cytotoxic heterodimeric glycoprotein with an A-chain of 267 amino acids that is linked by a disulfide bond to a B-chain of 262 residues (
1
). This toxin is isolated from the seeds of
Ricinus communis
and inactivates mammalian ribosomes. The B-chain is a lectin that recognizes galactose-containing receptors on the surface of sensitive eukaryotic cells (
2
). Binding to the surface of eukaryotic cells results in the reduction of a
disulfide bond, and the A-chain is translocated to the cytosol (
3
). The A-chain catalyzes the hydrolysis of a single N-glycosidic bond of the adenosine at position 4324 among a total of
7000 nucleotides (nt) in 28S rRNA (
4
,
5
). Depurination of A4324 then results in inactivation of the ribosome,
apparently by disrupting the structure of the binding site for elongation
factors (
6
,
7
). The sequence of the ricin-substrate domain in 28S rRNA is highly conserved in many eukaryotes and it
has been demonstrated that the GAGA tetraloop is essential for any ricin-sensitive substrate (the first adenosine of the GAGA tetraloop corresponds
to A4324) (
8
,
9
).
Many approaches have been taken in attempts to define the nature of the
interaction between ricin and ribosomes. Site-directed mutagenesis of the ricin A-chain (
10
-
12
) and X-ray analysis of substrate analogs in the active site of the ricin A-chain (
13
,
14
) revealed that Tyr80, Val81, Tyr123, Glu177 and Arg180 are critical for the
interaction between the ricin A-chain and the substrate, and Monzingo and Robertus proposed a putative
depurination mechanism from these observations (
12
,
13
). In this mechanism, the target adenosine is sandwiched between Tyr80 and
Tyr123, and protonation of the adenosine by Arg180 and Val81 facilitates the
cleavage of the bond between N9 and C1'; Glu177 stabilizes the transition-state oxycarbonium ion. Ren
et al.
also studied the crystal structure of [alpha]-momorcharin, an N-glycosidase whose amino acid sequence is 34% homologous to
that of the ricin A-chain, and they proposed another mechanism for the action of the N-glycosidase whereby the glycosidic bond is weakened by the strained
conformation of the furanose (C1'-exo conformation) that is the results of binding to the active site
of the enzyme rather than by the protonation of the base by Arg180 (
15
).
In the previous report we described a short substrate for ricin, namely, a dodecaribonucleotide r-1C2U3C4A
5G6A7G8A
9U10- G11A12G (r12mer; Fig.
1
a), that has the potential to form a hairpin structure with a stem of 4 bp and a
GAGA loop. We determined the solution structure by NMR spectroscopy and
restrained molecular-dynamic calculations (
16
). The stem region exists as an A-form duplex. In the loop region, 5G and 8A form an unusual G:A base pair
(the hydrogen bonds are formed between the amino proton of 5G and the N7 of 8A,
and between the amino proton of 8A and N3 of 5G). The bases of 6A, 7G and 8A
form a continuous stack while 5G is isolated from them. That is, the
phosphodiester backbone has a turn between 5G and 6A. We proposed that this
turn might help ricin to gain access to 6A, which is the only site of
depurination in the entire structure (
16
). Heus and Pardi reported the three-dimensional solution structures of the GAAA and GCAA loops that displayed
unusual thermodynamic stability (
17
). Our study by NMR of the GAGA loop indicated that the overall structure of the
GAGA loop is similar to those of the GAAA and GCAA loops that are not
recognized by ricin (
16
). Therefore, we suggested that, in addition to the adenosine at the
depurination site, the neighboring guanosine on the 3' side (7G) might also play a role in the recognition by ricin (
16
).
The oligonucleotides were synthesized by the phosphoramidite method on an
automated DNA synthesizer (model 394B; Applied Biosystems). The deprotection
and purification of the oligonucleotides were done as described previously (
16
).
UV spectra were recorded on a model UV-2100 spectrophotometer (Shimadzu). The buffer for thermodynamic studies
was 0.1 M NaCl, 10 mM phosphate buffer (pH 7.0) and was the same as the buffer
used for NMR studies.
NMR spectra were recorded with an ALPHA-500 spectrometer (500 MHz for
1
H, 202 MHz for
31
P and 125 MHz for
13
C; JEOL). Imino proton spectra were collected in H
2
O-D
2
O (4:1, v/v) and non-exchangeable proton spectra were collected in 99.98% D
2
O after H-D exchange. Exchangeable proton spectra were obtained with a 1-1 pulse sequence for H
2
O signal suppression (
18
). NOE experiments were performed at 5oC to reduce the rate of exchange with water protons.
All two-dimensional NMR spectra were recorded in the phase-sensitive mode by the TPPI-State method (
19
) at 15 and 30oC. NOESY spectra were recorded with 2048 points in t2 and 512 points in t1
at different mixing times (60, 120 and 250 ms). The DQF-COSY spectrum was obtained from data sets of the same size. HOHAHA spectra
were obtained from data sets of the same size with an MLEV-17 spin locking pulse sequence (90, 120 and 150 ms mixing times). A
natural-abundance
13
C-
1
H HSQC spectrum was collected using the sensitivity-enhanced field gradient pulse sequence (
20
). The data sets were recorded with 1024 points in t2 and 256 points in t1.
13
C decoupling was achieved by an MPF-8 (multi pulse decoupling with phase and frequency switching) decoupling
sequence (
21
). A proton-detected
31
P-
1
H heteronuclear COSY experiment (
22
,
23
) was performed in the inverse mode. The data sets were recorded with 1024
points in t2 and 128 points in t1.
The N-glycosidase activity of the ricin A-chain against the GdAGA loop (6dA12mer), the GAGA loop (r12mer) and
the GAIA loop (7I12mer) was examined in a similar way (
8
,
9
). The RNA oligoribonucleotides (3 pmol), containing small amounts
32
P-5'-end-labeled RNAs, were incubated with the ricin A-chain (3.6 [mu]g; Wako Pure Chemical Co.) in 18 [mu]l of reaction solution (4 mM EDTA, 4 mM
MgCl
2
, 17 mM DTT at pH 7.4) at 37oC. Aliquots (4 [mu]l) were removed at appropriate times and diluted with TE buffer (46 [mu]l). The RNA solution was extracted with phenol and chloroform (1:1)
and precipitated with ethanol after addition of carrier tRNA (5 [mu]g). The RNA was dissolved in H
2
O (5 [mu]l) and a solution (25 [mu]l) of acetic acid (2.8 M) and aniline (1 M) was added that cleaves RNA
at the site of depurination by the ricin A-chain. After incubation at 40oC for 10 min the aniline-treated RNA was precipitated with ethanol that contained 0.3 M
NaOAc, dissolved in 8 [mu]l of gel-loading solution (50 mM EDTA, 9 M urea, 0.1% xylene cyanol, 0.1%
bromophenol blue) and subjected to electrophoresis in a 20% polyacrylamide gel
that contained 7 M urea. The radioactivity of the bands on the gel was determined with a Bioimaging
analyzer (BAS 2000; Fuji Film). The cleavage activity was indicated by the rate
of formation of the aniline-cleaved product.
Recent studies have indicated that Watson-Crick and non-Watson-Crick base pairs contribute to the stability of RNA folding. In addition to a wobble-type G:U base pair, a G:A base pair has also been
identified in DNA and RNA by X-ray crystallographic and NMR analyses (
24
-
26
), and Santa Lucia
et al.
showed that neighboring G:A mismatches are unusually stable (
27
). Since the 6dA12mer (Fig.
1
b) seems able to form a self-complementary duplex that contains a GdAGA/GdAGA internal loop, we first
analyzed the dependence on concentration of the melting transition temperature
(
T
m
) of the oligomer. The profile of UV absorption at 260 nm versus temperature at
different concentrations (from 3.88 to 71.5 [mu]M) revealed the same
T
m
value (50.0oC) in each case and, furthermore, the
T
m
determined from variations in
1
H NMR chemical shifts of base protons was also ~50oC at 2 mM. Since the oligomer gave the same
T
m
value at concentrations from 3.88 [mu]M to 2.0 mM, we concluded that the 6dA12mer adopts a unimolecular hairpin
stem-loop structure in solution, at the concentrations used for NMR analysis.
This result is similar to that observed for the r12mer (
16
).
The N-glycosidase activity of ricin A-chain with the GdAGA loop (6dA12mer), GAGA loop (r12mer) and GAIA
loop (7I12mer) as substrates was measured as described in Materials and
Methods. We repeated each assay three times. Figure
2
B shows that the half-lives for the complete depurination of the 6dA12mer and r12mer were 18 min and 7.9 h, respectively. Thus, the N-glycosidase activity of the ricin A-chain against the GdAGA loop was 26 times higher than that
against the GAGA loop. Therefore, it seemed that the 2'-hydroxyl group did not participate in the interaction between the
ricin A-chain and the substrate RNA.
Figure
3
A shows the NMR spectrum of the 6dA12mer in the low-field region. There are four hydrogen-bonded imino proton signals in the region between 11.5 and 15.0
p.p.m., and one unpaired imino proton at 10.52 p.p.m. These protons were
assigned by nuclear Overhauser effect (NOE) experiments, as shown in Figure
3
B, though the imino proton of 7G was not detected even at 0oC. The dependence on temperature of the imino proton signals (Fig.
3
A) revealed that the signal of the imino proton of 5G was lost at a higher
temperature, at which signals for the stem region were also lost. These results
indicate that the NH of 7G exchanges rapidly with water protons while the NH of
5G exchanges relatively slowly. Because the imino proton of 5G is thought to be
buried more deeply inside the loop, the rate of exchange of the imino proton of
5G with a water proton should be low. The NMR studies of the GAGA (
16
), GAAA and GCAA (
17
) loops also indicated that the imino proton of the first guanosine is exchanged
relatively slowly with water protons. When the imino proton of 5G was
irradiated, the NOE on H8 of 8A was observed (Fig.
3
Bf). This NOE suggests the existence of an usual G:A base pair with hydrogen
bonds formed between the amino proton of 5G and the N7 of 8A, and between the
amino proton of 8A and N3 of 5G as shown in Figure
1
d. This anti-anti shared-type G:A base pair has been observed in several RNAs (
26
) and NMR studies showed that it also exists in GAGA (r12mer) (
16
), GAAA and GCAA (
17
) loops.
In order to obtain detailed information about the conformation of the oligomer,
we assigned the non-exchangeable proton resonances by two-dimensional NOESY, DQF-COSY, HOHAHA and heteronuclear
31
P-
1
H COSY spectroscopy, using previously established method (
28
,
29
). The results of assignments are shown in Table
1
. We also calculated the difference in chemical shifts between the 6dA12mer and
r12mer.
Table 1
Figure
4
shows the expanded NOESY spectrum from the aromatic to the H1' region. The H8/H6-H1' cross-peaks can be traced sequentially along 1C-5G and 6dA-12G, even though the NOE between H1' of 8A and H6 of 9U was very weak.
However, there are no cross-peaks between H1' of 5G and H8 of 6dA. We assumed that, in the loop region, the
bases of 6A, 7G and 8A form a continuous stack while 5G is isolated from them.
H1' shows an unusual upfield shift (>1 p.p.m. from the normal position). The
ring-current effects of aromatic rings are assumed to be the cause of this
shift. Therefore, we confirmed the assignment of H1' of 9U by a natural-abundance
13
C-
1
H HSQC because, as a rule, a
13
C nucleus does not show as much of a ring-current shift as a
1
H nucleus (data not shown). This upfield shift of H1' of 9U was thought to come from the ring current effect of the adenine
ring of 8A, which formed an unusual G:A base pair, as described in the NMR
studies of r12mer (
16
). In the stem region, H6 and H8s NOEs on 5'-neighboring H2' were stronger than those to their own H2' (data not shown). This result indicates that the
stem region of the oligomer forms an A-type helix. The observation of the pyrimidine H5 cross-peaks with 5'-neighboring H2' and H3' also supports the proposed A-form duplex conformations (
29
).
Since chemical shifts of phosphorus signals provide information about torsion
angles of [zeta](O3'-P) and [alpha](P-O5') (
30
), the backbone conformation of the oligomer was examined by
31
P-NMR (Fig.
5
). The phosphorus signals were assigned from a heteronuclear
31
P-
1
H COSY spectrum (Fig.
5
a), although the connectivity between H3' of 8A and P of 9U was not detected because of overlapping with the
solvent signal. When an oligonucleotide forms a normal, right-handed double helix, which has a smooth sugar-phosphate backbone,
31
P signals are observed in the region of -3.0 to -4.5 p.p.m. Moreover, it is well known that a change in
conformation about O3'-P and P-O5' causes a downfield shift of the phosphorus signal (
30
). In the GdAGA loop (6dA12mer), all
31
P signals except for those from the phosphorus atoms of 5Gp6dA and 8Ap9U were observed in the region of -3.0 to -4.5 p.p.m. This result means that the backbones around these
phosphorus atoms were in the normal
gauche-gauche
conformation. The phosphorus atoms of 5Gp6dA and 8Ap9U has slight downfield shifts (-2.03 and -2.42 p.p.m., respectively), but it is likely that the [xi]-[alpha] conformations of the 5Gp6dA and 8Ap9U phosphorus atoms were also
gauche-gauche
because the signal from a phosphorus atom whose [xi]-[alpha] conformation is
gauche-trans
is normally observed in the region of -1 p.p.m. Detailed assumptions cannot be made from these shifts because
several conformational effects might be responsible for such shifts.
Figure
6
shows schematic representations of the structures of the 6dA12mer and the
r12mer.
Monzingo
et al.
and Ren
et al.
proposed a mechanism for the depurination by the ricin A-chain from the results of site-directed mutagenesis and of X-ray analysis (
13
,
15
). It is well-known that the depurination of a purine nucleoside can also occur by acid-catalysis (
33
). Moreover, because the transition state in such a reaction resembles an
oxycarbonium ion, the acid depurination is thought to be similar to their N-glycosidase mechanism. Our kinetic data indicate that the N-glycosidase activity against the GdAGA loop by the ricin A-chain was 26 times higher than that of the GAGA loop. From the
standpoint of electronegativity, 2'-hydroxyl group is expected to destabilize the positive charge of
the oxycarbonium ion, the transition state intermediate in both reactions.
Therefore, it seems reasonable that the electronic effect of 2'-oxygen plays a role, at least in part, in the destabilization of
the putative oxycarbonium ion and that a potential anchimeric assistance by 2'-OH is negligible. The reaction with the r12mer was too slow for
kinetic analysis and it is, thus, difficult to define the role of each
functional group. Accordingly, the 6dA12mer should be a good model substrate
for studying the mechanism of recognition of the substrate by the ricin A-chain if 6dA is changed, for example, to deoxynebularine or N6-methyl dA.
Our data can be summarized as follows. (i) The overall structures of the
6dA12mer and the r12mer are almost the same. (ii) There are, however, two minor
differences between the 6dA12mer and the r12mer; the sugar puckers of 6dA and
7G and the dispersion of the chemical shifts of the phosphorus signals from the loop region. (iii)
Because the N-glycosidase activity against the GdAGA loop of the ricin A-chain was 26 times higher than against the type GAGA loop, it seems
that the 2'-hydroxyl group of adenosine (6A) at the site of depurination does
not participate in the recognition by the ricin A-chain of its substrate. (iv) Finally, from the glycosidase assay with the
7I12mer, we also showed that the amino group of 7G in the GAGA loop is critical
for recognition by ricin A-chain.
REFERENCES
Return