ABSTRACT
A small RNA derived from the decoding region of
Escherichia coli
16S rRNA can bind to antibiotics of aminoglycosides (neomycin and paromomycin)
that act on the small ribosomal subunit [Purohit,P. and Stern,S. (1994)
Nature
, 370, 659-662]. In the present study, the P-site subdomain was removed from this decoding region RNA to construct a
27mer RNA (designated as ASR-27), which includes the A-site-related region (positions 1402-1412 and 1488-1497) of 16S
rRNA. Footprint experiments with dimethyl sulfate as a chemical probe indicated
that the ASR-27 RNA can interact with the neomycin family in the same manner as the
decoding region RNA. A mutagenesis analysis of the ASR-27 RNA revealed that paromomycin binding of ASR-27 involves the C1407[middot]G1494 and C1409[middot]G1491 base pairs, and the internal loop comprising
A1408 and the nucleotides in positions 1492-1493, located between the two C[middot]G base pairs. In addition, a G or U in position 1495, and base
pairing between positions 1405 and 1496 are also involved. These structural
features were found in a viral RNA element, the Rev-binding site of human immunodeficiency virus type-1, which may explain why neomycin can bind to this viral RNA.
Increasing evidence has indicated that ribosomal RNAs play crucial and positive
roles in ribosome functions (
1
,
2
). This already has been established for mRNA selection, which occurs through
the base-pairing interaction between the 3'-end region of 16S rRNA and the Shine-Dalgarno sequence on the mRNA (
3
,
4
). The peptidyl transferase activity of the large ribosomal subunit has been
suggested to be an intrinsic function of 23S rRNA, on the basis of several
lines of study (
5
-
7
), although conclusive evidence has yet to be provided.
While the peptide transfer occurs on the large subunit, the decoding function of
the ribosome is attributed to the small subunit (
1
,
2
). Some groups of antibiotics impair the decoding function by acting on this
subunit (
8
). Resistance to these antibiotics is conferred by methylation or mutation at
specific sites in 16S-like rRNAs (
8
). Furthermore, it has been shown that characteristic sets of the 16S-rRNA bases from
Escherichia coli
are protected by these antibiotics from the attack of chemical probes (
9
,
10
). These observations led to suggestions that 16S(-like) rRNA is involved in ribosomal decoding, and that the antibiotics act
by interfering with this function of 16S(-like) rRNA (
2
,
9
).
The interactions of
E.coli
16S rRNA with mRNA and tRNA have been investigated by cross-linking (
11
-
14
) and chemical protection studies (
15
,
16
). These studies have defined the decoding region in 16S rRNA, which largely
overlaps two phylogenetically conserved sequences, positions 1390-1407 and
1492-1506 (according to the numbering scheme of
E.coli
16S rRNA), of 16S-like rRNA (
17
,
18
). Furthermore, observations that the sites of antibiotic binding are located in
or near the decoding region thus defined (
9
,
10
) also point to the involvement of this region in decoding (
1
,
2
).
P. Purohit and S. Stern showed by a chemical protection study that a small RNA
derived from the decoding region of
E.coli
16S rRNA, or the decoding region RNA, can interact with mRNA, tRNA and
antibiotics of the neomycin family (neomycin and paromomycin) in a manner
similar to that of the small subunit (
19
). This implies that the binding of the small subunit to its ligands may be due
to the intrinsic property of the RNA moiety (
19
). Since the neomycin family protects the A-site bases, rather than the P-site bases, from chemical probes (
9
), in this study we removed the P-site subdomain from the decoding region RNA, and thus constructed a 27-residue A-site-related RNA (designated as ASR-27). Chemical protection experiments showed that
this ASR-27 RNA still can bind to the neomycin family, in the same manner as the
decoding region RNA. Therefore, 24 variants of the ASR-27 RNA were examined for interactions with paromomycin, in order to
identify the nucleotide residues required for binding to the drug.
Neomycin, paromomycin and hygromycin B were purchased from Sigma. The neomycin
was a mixture of neomycin B (85%) and C (15%). Tetracycline, streptomycin and
kanamycin were from Wako pure chemical industries, Ltd (Osaka, Japan). Dimethyl
sulfate was from Nacalai Tesque Inc. (Kyoto, Japan), and diethyl pyrocarbonate
was from Sigma. Aniline (aniline-point test grade) was purchased from Wako.
T7 RNA polymerase was purified from an overproducing strain, kindly provided by
Dr W. Studier (Stonybrook, New York), according to the method described (
20
). Oligodeoxyribonucleotides were chemically synthesized with an Expedite 9600
DNA synthesizer (Perseptive Biosystems). The ASR-27 RNA and its variants were obtained by run-off transcription with T7 RNA polymerase. Separation from the minor
T7 transcripts was carried out by 16% denaturing polyacrylamide gel
electrophoresis. The RNA was eluted from the gel with buffer A (0.5 M ammonium
acetate, 0.1% sodium dodecyl sulfate and 0.1 mM EDTA), and was then
precipitated with ethanol. Sequencing of the RNA was performed as described (
21
).
The RNA was labeled at the 3'-end with [5'-
32
P]cytidine 3',5'-bis(phosphate) (111 TBq/mmol, Dupont/NEN research products)
using T4 RNA ligase (Takara Shuzo Co. Ltd, Kyoto, Japan). The labeled RNA was
subjected to 16% denaturing polyacrylamide gel electrophoresis, and was
recovered from the gel as described above.
The labeled sample of RNA was dissolved in 80 mM HEPES-KOH (pH 7.85), and was incubated at 65oC for 3 min followed by gradual cooling to room temperature. Binding
reaction mixtures, containing this renatured RNA, antibiotics, 80 mM HEPES-KOH (pH 7.85) and 50 mM ammonium chloride, were incubated at 37oC for 15 min, and then at 0oC for 60 min. An aliquot (20 [mu]l) of the mixture, containing ~1 [mu]M of the RNA, was added to 1.5 [mu]l dimethyl sulfate/ethanol (1:5, v/v), and was
incubated at 0oC for 30 min. Another aliquot was added to 2 [mu]l diethyl pyrocarbonate, and was incubated at 0oC for 2 h. For investigating the effects of magnesium ion, the
labeled RNA was dissolved in 80 mM HEPES-KOH (pH 7.85) containing MgCl
2
(0-10 mM), and the binding reaction mixtures were supplemented with the
corresponding concentrations of MgCl
2
. For the reactions of the RNA with diethyl pyrocarbonate in the absence of
drugs, a 20 [mu]l reaction mixture, containing the renatured RNA (~1 [mu]M), 80 mM HEPES-KOH (pH 7.85) and 50 mM ammonium chloride, was added to 2 [mu]l diethyl pyrocarbonate, and was incubated at 0oC for 2 h, 37oC for 30 min or 90oC for 5 min. After the chemical
modifications, the RNA was treated with aniline, and the RNA fragments thus
produced were separated on 20% denaturing polyacrylamide gels. These chemical
reactions were performed according to the standard procedure as described (
21
,
22
), except that the lyophilization steps were replaced by a phenol/chloroform extraction and an ethanol precipitation. The reactivity toward dimethyl sulfate
or diethyl pyrocarbonate was estimated on the basis of the band intensity on
the autoradiogram, which was measured with a Bio-Imaging Analyzer BAS2000 (Fuji Photo Film Co., Ltd, Tokyo).
The ASR-27 RNA involves the
E.coli
16S-rRNA sequences, from positions 1404-1412 and 1488-1497 (Fig.
1
). The nucleotide sequences at positions 1409-1412 and 1488-1491 form a stem
structure in the context of the ribosome (
23
). Therefore, in the ASR-27 RNA, these sequences are linked with the stable UUCG tetraloop present
at the distal end of the 3'-penultimate stem of 16S rRNA (
19
). Since it has been suggested that the C1404[middot]G1497 and G1405[middot]C1496 base pairs are formed in 16S rRNA (
24
), complementary GG and CC sequences are added to the 5'- and 3'-ends, respectively, of the ASR-27 RNA, in order to clamp the top of ASR-27.
The ASR-27 RNA was investigated for interactions with hygromycin B, kanamycin, neomycin, paromomycin, streptomycin and tetracycline. The concentrations of these drugs were 5-10-fold higher than those used in a previous protection study of 16S
rRNA (
9
).
Both neomycin (50 [mu]M) and paromomycin (50 [mu]M) were found to remarkably decrease the reactivities of the N7
positions of G1405, G1491 and G1494 toward dimethyl sulfate (Fig.
2
). On the other hand, both hygromycin B (500 [mu]M) and kanamycin (500 [mu]M) appreciably decreased the reactivity of G1494, with slight
protections at G1405 and G1491 (Fig.
2
). It should be noted here that the protection of G1494 is characteristic of the
binding of these four antibiotics to the small subunit (
9
). In contrast, no significant change in the reactivity of the N7 of any guanine
base was detected for streptomycin (50 [mu]M) or tetracycline (500 [mu]M) (Fig.
2
), which interact with other sequences than that corresponding to ASR-27 within 16S rRNA (
9
). An increase in the concentration of streptomycin (to 500 [mu]M) still did not change the reactivity of any purine at N7 (data not shown).
The aforementioned results showed that the binding of antibiotics to the
decoding region RNA is hardly affected by the removal of the P-site bases. Therefore, we focused on the ASR-27 RNA in order to determine which nucleotide residues are required
for antibiotic binding. Twenty-four variants of ASR-27 were examined for interactions with paromomycin of 50 [mu]M in terms of the protection from dimethyl sulfate (Fig.
5
), because the neomycin was a mixture of neomycin B and C. Figure
6
summarizes the levels of protection at positions 1405, 1491 and 1494, relative
to the corresponding protection levels in the `parental' ASR-27 RNA, which consists of the wild-type sequence.
Figure
Figure
Furthermore, it was determined whether the number of `spacer' bases (positions
1492 and 1493) between G1491 and G1494 is important for ASR-27 binding to paromomycin. The protection levels were found to be the same
for the parental ASR-27 and its variant with three adenosine residues in place of A1492 and
A1493 (Fig.
5
, lanes 47-48; Fig.
6
). On the other hand, the substitution of these two adenosine residues with a
single adenosine largely reduced the protection levels, especially at position
1491 (Fig.
5
, lanes 49-50; Fig.
6
).
From the aforementioned mutagenesis study, the paromomycin-binding motif of ASR-27 was identified (Fig.
7
A). This motif was searched for in the antibiotic-binding region of a group I intron, in the Rev-binding element (RBE) from HIV-1 and in the hammerhead ribozyme; the functions of these RNAs
have been reported to be inhibited by neomycin (
29
-
33
).
Figure
Among these RNA sequences, that of the RBE (positions 47-51 and 67-73) appears
to be identical to the paromomycin-binding motif, except for the replacements of A1408 by G50, and of the
1405[middot]1496 base pair by the non-canonical G47[middot]A73 base pair with a bulged U72 (
34
) (Fig.
7
A and B). The effect of the A1408G substitution is relatively small, both for
paromomycin binding (described above) and neomycin binding (data not shown).
Therefore, it was necessary to determine if the 1405[middot]1496 base pairing is compatible with a G[middot]A base pair with a U bulge. This replacement was introduced into
ASR-27, together with that of C1404[middot]G1497 by stem IID (residues 45-46 and 74-75) of the RBE (Fig.
7
C), because the stacking with stem IID may be important for residues G47 and A73
to form a non-canonical base pair with each other (
34
). These replacements do not affect the ASR-27 binding to either neomycin (data not shown) or paromomycin (Fig.
5
, lanes 51-52; Fig.
6
). Therefore, we suggest that the paromomycin-binding motif may underlie the neomycin binding of RBE.
Aminoglycoside antibiotics have a polycationic character; neomycin B has six
amino groups, five of which are protonated at pH 6-8 (
35
). Therefore, it is likely that aminoglycosides bind to RNA through
electrostatic interactions between their amino groups and the phosphate groups
of the RNA backbone. This is probably the case for the neomycin binding of the
recently isolated RNA aptamers, which share a hairpin structure featuring a
widely opened major groove, rather than a particular nucleotide sequence (
36
,
37
). On the other hand, for the ASR-27 RNA, we observed that the base substitutions affect paromomycin
binding, and thus identified a nucleotide set required for antibiotic binding.
Furthermore, we showed that the drug binding of ASR-27 does not require, and is not affected by, the presence of magnesium
ion. This result is apparently inconsistent with the observation that magnesium
ion at a physiological concentration inhibits binding of the decoding region
RNA to the immobilized neomycin on the agarose support (
37
). However, this disagreement may be due to the covalent linkage of neomycin to
agarose in the previous study (
37
). Thus, we conclude that paromomycin binding to ASR-27 is not simply due to ionic interactions between the RNA backbone and
the amino groups of the drug; it has a sequence-specific character.
The paromomycin-binding motif is thought to be involved in neomycin binding as well, for
three reasons. First, the chemical structure of paromomycin is identical to
that of neomycin, except for the replacement of the 6'-amino group by a hydroxyl group. Secondly, these two
aminoglycosides protect the same set of bases in the ASR-27 RNA and 16S-rRNA (
9
). Finally, 16S rRNA does not appear to discriminate between these
aminoglycosides; no mutation in 16S rRNA has been reported to confer resistance
to only one of the two. On the other hand, although hygromycin B and kanamycin
can bind to ASR-27, these antibiotics are thought to recognize different structures of 16S
rRNA (
28
,
38
).
Two natural, single-base mutations in 16S-like rRNA that confer resistance to paromomycin have been reported (
27
,
28
). These mutations occur in positions 1409 and 1491, and disrupt the base
pairing between these positions. This resistant phenotype correlates with the
weak binding of paromomycin to the mutant 16S-like RNAs (
39
). Consistent with these observations, the G1491U substitution in the ASR-27 RNA, which disrupts the 1409[middot]1491 base pair, was found to significantly inhibit the interaction
with paromomycin (Fig.
6
). The replacement of the C1409[middot]G1491 base pair by an A[middot]U or G[middot]C base pair also reduced the affinity of ASR-27 for the drug, but to a smaller extent than
G1491U. On the other hand, the cytoplasmic 16S-like RNAs from eukaryotes have G in place of A1408, and these ribosomes
are less sensitive to paromomycin than the prokaryotic ribosomes (
40
-
42
). For ASR-27, A1408G reduced the protection levels, but to an appreciably smaller
extent than the `resistant mutation', G1491U. Thus, the result with ASR-27 correlates well with the
in vivo
observations of the drug sensitivities of 16S-like rRNAs, although the effect of the U1495C substitution may be an
exception, because it largely reduced the affinity of ASR-27 for paromomycin, but did not affect the sensitivity of
Tetrahymena
17S rRNA to this drug (
28
).
The paromomycin-binding motif involves some bases of 16S rRNA that are associated with the
decoding function. The adenosine residues comprising the internal loop (A1408,
A1492 and A1493) have been reported to be protected from a chemical reagent by
the tRNA bound to the A site of the ribosome (
15
,
16
). The phosphodiester bond between A1493 and G1494 is susceptible to the
cleavage by colicin E3, which inactivates the ribosome (
43
,
44
). Furthermore, a mutation that confers a dominant lethal phenotype occurs in
position 1407 (
45
), while frameshift-suppressor mutations occur in the base-paired 1409 and 1491 positions (
46
,
47
). These observations imply important roles for these residues in the A-site functions of 16S rRNA. Our identification of the paromomycin-binding motif will provide a structural basis for delineating the
molecular mechanism by which the antibiotics impair ribosomal decoding.
The neomycin family has been reported to inhibit the functions of various RNAs
besides 16S rRNA, including group I introns (
30
,
31
), the RBE of HIV-1 RNA (
32
) and the hammerhead ribozyme (
33
). The structural features underlying the antibiotic binding of these RNA
molecules have not yet been unraveled. On the other hand, novel neomycin-binding RNAs, which were recently isolated by
in vitro
selections, have a common hairpin structure featuring a widely opened major
groove (
36
,
37
). This structural feature may be extrapolated to the paromomycin-binding motif with an asymmetric internal loop, because this type of internal
loop is accessible in the major groove (
26
).
Our discovery of a nucleotide sequence related to the paromomycin-binding motif in the RBE of HIV-1 suggests that this motif may underlie the RBE binding to neomycin.
Indeed, a recent report showed that the substitution of G71 with A reduced the
ability of the RBE to bind neomycin (
48
). The interactions of a 66-residue RNA, involving the RBE, with neomycin have been investigated by
chemical probing (
32
). The set of protected guanine bases of this RNA partly overlaps, but does not
correspond to that of the protected ASR-27 bases, probably because the probed position of the guanine base is
different between these analyses. Although binding of the Rev protein to the
RBE is inhibited by neomycin, rather than paromomycin (
32
), our result suggests that the RBE can accommodate paromomycin. Therefore, the
6'-amino group of neomycin, which is absent in paromomycin, may make
additional interactions with the RBE that are important for the inhibitory
activity of this drug.
On the other hand, the RBE interacts with the arginine-rich domain of the Rev protein (
49
,
50
). By
in vitro
genetic studies, the structural features of the RBE that are important for Rev
binding have been elucidated (
34
,
51
). The two C[middot]G base pairs (positions 49[middot]70 and 51[middot]67) are conserved in all the selected aptamers to Rev.
Interestingly, these base pairs are involved in the paromomycin-binding motif. Furthermore, a computerized analysis has revealed that
accessibility in the major groove of the RBE is required for the interaction
with the arginine-rich peptide derived from Rev (
52
). Thus, some important features are partly overlapped between the Rev binding
and the drug binding of the RBE, raising the possibility that neomycin may
mimic the Rev-arginine-rich peptide in its interaction with the RBE. Delineation of the
structural basis for the RBE binding to neomycin will be helpful for the design
of aminoglycosides that more effectively target the viral RNA.
This work is supported by a Grant-in-Aid for Scientific Research on Priority Areas (No. 04272103) from
the Ministry of Education, Science and Culture of Japan.


C1409 and G1491.
Base pairing between positions 1409 and 1491 is conserved in prokaryotic 16S
rRNAs (
18
), and single mutations that disrupt this base pair have been reported to confer
resistance to paromomycin (
27
,
28
). Therefore, the substitution of G1491 with U (denoted by G1491U) was
introduced into the ASR-27 RNA. G1491U significantly reduced the protection levels at positions
1405 and 1494, and thus weakened the interaction with paromomycin (Fig.
5
, lanes 3-4; Fig.
6
). The replacement of C1409[middot]G1491 with G[middot]C or A[middot]U also weakened the drug binding, but to a slightly
smaller extent than G1491U (Fig.
5
, lanes 5-8; Fig.
6
).
G1405 and C1496.
G1405 has been shown to be base paired with C1496 in 16S rRNA (
24
), and its protection from a chemical reagent was observed for the ASR-27 RNA, as described above. Single mutation G1405C appreciably reduced the
protection levels at positions 1491 and 1494 (Fig.
5
, lanes 9-10; Fig.
6
), while C1496G reduced the protection levels at positions 1405 and 1491
significantly, and at position 1494 slightly (Fig.
5
, lanes 9-12; Fig.
6
). On the other hand, the reversal of this G1405[middot]C1496 base pair, to C[middot]G, had no effect on the paromomycin binding (Fig.
5
, lanes 13-14; Fig.
6
). These observations indicate that the base pairing between these positions,
rather than the base type, is important for paromomycin binding. Thus, although
protected by paromomycin, G1405 may not be involved in a direct interaction
with it.
U1406 and U1495.
These bases are universally conserved in 16S-like rRNAs (
18
), and it is possible that a `short wobble' base pair may be formed between
these uridine residues. The paromomycin binding was not affected by either of
U1406A and U1406C, which introduce a Watson-Crick base pair and a U[middot]C mismatch, respectively, in place of the putative U1406[middot]U1495 pair (Fig.
5
, lanes 15-18; Fig.
6
). In contrast, for position 1495, the replacement of U by either A or C largely
reduced the affinity for the drug, while that by G did not affect the drug
binding (Fig.
5
, lanes 21-26; Fig.
6
). In order to determine if this is due to the replacement of the putative U1406[middot]U1495 pair by another `wobble' base pair, a U[middot]G pair, substitution U1495G was introduced together with U1406G,
resulting in no effect on the protection levels (Fig.
5
, lanes 19-20; Fig.
6
). It was concluded therefore that the paromomycin binding of ASR-27 requires either of U and G in position 1495, but allows any of U, C, A
and G in position 1406, while base pairing between residues 1406 and 1495 is
not involved.
C1407 and G1494.
These bases are also universally conserved in 16S-like rRNAs (
18
), and protection of G1494 was observed for ASR-27, as described above. Single base changes (C1407G and G1494C) that each
abolish the potential C1407[middot]G1494 pairing were introduced (Fig.
5
, lanes 27-28 and 31-32; Fig.
6
). Furthermore, this putative C1407[middot]G1494 pair was substituted with U[middot]G, G[middot]C and U[middot]A (Fig.
5
, lanes 29-30 and 33-36; Fig.
6
). All of these substitutions impaired paromomycin binding more severely than
the substitutions described above, indicating that the nucleosides in positions
1407 and 1494 are stringently required to be C and G, respectively. It is yet
to be clarified whether C1407 and G1494 form a Watson-Crick base pair in the complex of ASR-27 with paromomycin.
The A-rich internal loop.
The internal loop of the ASR-27 RNA comprises A1408, A1492 and A1493. The latter two are universally
conserved in 16S-like rRNAs, while A1408 is replaced by G in the cytoplasmic ribosomes from
eukaryotes (
18
). The substitutions of A1408 with C, G and U all weakened the binding to
paromomycin; the effect of A1408G is relatively small, while the effect of
A1408U is largest among those of these substitutions (Fig.
5
, lanes 37-42; Fig.
6
). In contrast, the double substitution of A1492 and A1493 with CC was found to
have no effect on paromomycin binding (Fig.
5
, lanes 43-44; Fig.
6
). In order to determine if the effect of A1408U is due to the possible base
pairing of U1408 with either A1492 or A1493, A1408U was introduced, together
with the double substitution A1492C/A1493C. This triple substitution still
affected the interaction with the antibiotic, to essentially the same extent as
the single substitution of A1408U (Fig.
5
, lanes 45-46; Fig.
6
), showing that the effect of A1408U is due to the base substitution itself.

REFERENCES
Return

