ABSTRACT
A conserved hairpin corresponding to nt 1057-1081 of large subunit rRNA (
Escherichia coli
numbering) is part of a domain targeted by antibiotics and ribosomal protein
L11. The stem of the hairpin contains a U[middot]U juxtaposition, found as either U[middot]U or U[middot]C in virtually all rRNA sequences. This hairpin has been
synthesized and most of the aromatic and sugar protons were assigned by two-dimensional proton NMR. Distances and sugar puckers deduced from the NMR
data were combined with restrained molecular dynamics calculations to deduce
structural features of the hairpin. The two U residues are stacked in the
helix, form one NH3-O4 hydrogen bond and require an extended backbone conformation (
trans
[alpha]
and
[gamma]
) at one of the U nucleotides. The hairpin loop, UAGAAGC closed by a U-A pair, is the same size as tRNA anticodon loops, but not as well
ordered.
Many regions of rRNAs have been highly conserved in sequence or secondary
structure (
1
) and in some instances have been associated with specific ribosome functions (
2
). One such region is the domain at nt 1051-1108 of the large subunit rRNA (
Escherichia coli
numbering). A third of the 58 nt are invariant among prokaryotic and eukaryotic
sequences and the domain is the binding site for the thiostrepton family of
antibiotics that affect the GTPase activity of the ribosome (
3
). A conserved ribosomal protein, L11, specifically binds the same region (
4
,
5
). Recent work has shown that an RNA fragment duplicating the domain has a set
of tertiary interactions stabilized specifically by Mg
2+
and NH
4
+
in preference to other ions and that both thiostrepton and L11 recognize this
tertiary structure (
6
,
7
,
8
).
Within the 1051-1108 rRNA domain is a hairpin (1057-1081), shown in Figure
1
, that contains a U[middot]U or U[middot]C mismatch: 1060U is invariant, while 1078U is substituted by C
in many organisms. Tandem U[middot]U mismatches in short RNA duplexes are unexpectedly stable and form hydrogen bonded pairs (
9
,
10
), while a T[middot]T mismatch may pair as an unusual enol tautomer (
11
). Tandem U[middot]C mismatches are also hydrogen bonded in the crystal structure of an RNA
helix (
12
). Whether the ribosomal U[middot]U mismatch is paired or distorts the helix in a specific way is
potentially relevant to the function of the rRNA domain. We present here an NMR
study of this 1060U[middot]1078U mismatch conformation, as the first study of a single U[middot]U mismatch within an asymmetrical, biologically relevant context. The results clearly show that the U[middot]U mismatch is base paired and causes one of the two U
nucleotides to adopt an extended
trans
backbone conformation. Although we do not have enough information to determine
the detailed structure of the hairpin loop, it is clear that its conformation
is different from the seven base anticodon loop structure of tRNAs.
Synthetic DNA containing a T7 RNA polymerase promoter sequence, the desired A' RNA transcription sequence, and appropriate restriction sites was cloned
into pUC18 cleaved with
Eco
RI and
Sma
I endonucleases. CsCl density centrifugation was used as a final step in large
scale plasmid preparation.
Sma
I endonuclease was used to linearize the plasmid for
in vitro
transcription by T7 RNA polymerase, prepared in this laboratory. Transcription
reactions contained 50 mM Tris, pH 8.1, 30 mM MgCl
2
, 25 mM NaCl, 30 mM DTT, 40 nM template DNA, 3 mM of each NTP and ~0.4 [mu]M T7 RNA polymerase and were carried out for 12 h at 37oC. A 30 ml reaction volume produced ~13 mg purified RNA, sufficient for one NMR sample. The
transcription products were precipitated with ethanol, redissolved in ~5 ml water and dialyzed extensively against 5 mM Na
2
EDTA using dialysis membrane with a 3 kDa molecular weight cut-off. The RNA was then purified by electrophoresis on four 8% polyacrylamide gels
containing 8 M urea and measuring 3 mm thick * 12 cm wide. The desired RNA was located by its UV absorbance, excised and
eluted electrophoretically from the gels in an EluTrap (Schleicher & Schuell). The final product was precipitated by ethanol and its purity was
checked by electrophoresis on a 15% polyacrylamide gel with 8 M urea.
All water used for sample preparation was purified through a Barnstead Nanopure
ion exchange unit and further treated with Chelex (BioRad) to remove any trace
heavy metal ions. RNA was first dissolved in water and dialyzed against water
for more than 36 h to remove salts and any other low molecular weight
substances. The sample was then lyophilized and resuspended in 500 [mu]l 100 mM KCl, 4 mM sodium cacodylate, pH 6.45, and 0.4 mM EDTA. The sample was lyophilized a second time and resuspended in 450 [mu]l H
2
O plus 50 [mu]l D
2
O. This sample was used for experiments detecting exchangeable protons. It was then lyophilized, resuspended in 99.96% D
2
O, lyophilized two more times and finally redissolved in 99.996% D
2
O (Cambridge Isotope Laboratories). TMS, phopsphoric acid and
13
CH
3
I were used as references for
1
H,
31
P and
13
C respectively in all experiments.
All NMR spectra were acquired on a Varian Unity Plus 500 MHz spectrometer and processed using Felix (Biosym Technology) or Vnmr (Varian) software. A pulsed field gradient probe was used for the
ssNOESY experiments and an inverse probe for all other experiments. All
experiments were recorded in States mode (
13
). The NMR experiments are only briefly summarized here; more details are
presented elsewhere (
14
,
15
). A NOESY spectrum of exchangeable protons was taken at 10oC using a symmetrically shifted pulse sequence at 200 ms mixing time (
16
). NOESY spectra in D
2
O were recorded at 100 (30oC) and 400 ms (32 and 35oC) mixing times. The recycling time was 6 s for the 100 ms mixing time
NOESY, so as to achieve a maximum spin recovery. To distinguish primary NOEs from relayed NOEs and chemical exchange cross-peaks, a ROESY experiment was performed (
17
), with the mixing power <0.5 W and a recycling time of 5 s.
31
P-Decoupled DQF-COSY and homo-TOCSY spectra (125 ms mixing time) were also acquired. Two
heteronuclear experiments were performed: a
1
H-
13
C HSQC spectrum (
18
) and a
1
H-
31
P hetero-COSY (
19
) with corrected phase cycling.
The
T
1
relaxation times of resolved non-exchangeable proton resonances were measured using an inversion recovery
sequence in an interleaved mode. The longest
T
1
was ~7.5 s (1077AH2).
T
1
relaxation times of H1' protons were between 3 and 5 s. Although a 6 s recycling time was used
for the NOESY at 100 ms mixing time, some protons, especially adenine H2, were
not fully recovered to equilibrium.
Cross-peak volumes from the NOESY with 100 ms mixing time were integrated using
Felix software. In generating NOE distances, the molecule was assumed to be
rigid and have a uniform rotational correlation time; the cytosine H5-H6 distance (2.45 ) was used as reference. For input of restraints to
Discover, the derived distances were presented in percentage mode, by which an
error range was assigned to each distance automatically according to the signal-to-noise ratio of a peak. The error bars ranged from +-0.5 to +-2.0 Å. Errors introduced by spin diffusion at 100
ms mixing time are typically <20% and are within the conservatively estimated range of each NOE distance
generated by the Felix program. One hundred and sixty eight distance
constraints were generated from the NOE data; 20 of these involved 1060U and
1078U.
Proton-proton coupling constants were estimated from a
31
P-decoupled DQF-COSY spectrum with ~1 Hz/point digital resolution and ~4 Hz linewidths. Line-shape simulations were performed for coupling
constants under 5 Hz (
20
). In the cases where cross-peaks were very weak or missing, the coupling constants were assumed to be
<2 Hz. Torsion angles were derived from coupling constants using the generalized
Karplus equations. Some
1
H-
31
P coupling constants could be measured from the
1
H-
31
P hetero-COSY experiment, with uncertainties of +-3 Hz. Angles estimated for the corresponding [beta] and [epsilon] torsions all fell into the standard range for A-form helices.
Restrained molecular dynamics (rMD) calculations and model display were
performed with Biosym Insight II. The starting structure for rMD calculations
was an A-form helix generated by the Discover module of Biosym Insight. Distance
and torsional angle constraints were introduced as skewed biharmonic functions that have a flat-bottomed forcing potential within the estimated error limits [50 kcal/(mol
Å
2
) or 50 kcal/(mol rad
2
) for distances and torsion angles respectively]. Constraints of Watson-Crick hydrogen bonding for stem base pairs (1.8 +- 0.2 Å) and A-form ranges for [beta] (140 to 180o), [epsilon] (-150 to -180o) and [gamma] (45 to
75o) backbone torsions were introduced, except for 1060U and 1078U.
The refinement protocol was similar to that used by SantaLucia and Turner (
21
). During the refinement calculation, bond, valence, torsion angles, outplane
interactions and AMBER force field 1-4 bond interactions were turned on. Non-bonded interactions were cut off at 13.5 Å and the calculation was carried out
in vacuo
with a distance-dependent dielectric constant. The energy of the whole hairpin was minimized for 500 rounds and then equilibrated at 1000 K for 1 ps. This was followed by 10 ps restrained dynamics calculations at 1000 K simulation temperature, during which the torsional angle force constant was gradually scaled to 50 kcal/(mol rad
2
) at 5 kcal/rad
2
/ps for the angles derived from NMR experiments. Then restrained dynamics
calculations were carried out, cooling the temperature 100 K/ps until the system temperature reached 300 K. To overcome minor energy barriers, 3 ps dynamics at 300 K were
used, followed by 3000 steps of energy minimization using the steepest descent
protocol (to avoid local minima) and 5000 steps restrained minimization usinga conjugate gradient until the maximum derivative was <10
-4
.
The hairpin shown in Figure
1
, previously termed A' RNA, was designed to contain the conserved U1060[middot]U1078 mismatch and 1066-1072 hairpin loop of
E.coli
large subunit rRNA (
22
). A G-C pair was substituted for the
E.coli
A-U pair at the terminus for extra stability and convenience in transcription; this substitution is frequent among eubacterial sequences (Fig.
1
). The RNA was transcribed by T7 RNA polymerase from plasmid DNA that had been
cut with
Sma
I (CCC[up arrow]GGG) and as a result the hairpin has a C overhang at the 3'-end. The
T
m
of this RNA in 10 mM MOPS buffer, pH 7.0, was previously reported to be 52oC (
22
). Added salt caused dimerization of the hairpin, but we have since found that
dimerization is suppressed in sodium cacodylate buffer. RNA melting curves at ~2.5 and 25 [mu]M and 0.25 mM hairpin in 4 mM sodium cacodylate, pH 6.0, 100 mM NaCl and 0.2 mM Na
2
EDTA were superimposable with melting temperatures of 54oC, from which we conclude that the RNA remains monomeric at the high
concentrations required for NMR.
Proton assignments were carried out using standard procedures (
23
), starting with the identification of imino protons from a NOESY experiment
done at 10oC in H
2
O. The 1060U[middot]1078U mismatch gave two imino peaks at the high field end of the imino
region (Fig.
2
A); similar chemical shifts have been seen for the imino protons of tandem U[middot]U mismatches in other oligomers (
9
,
24
). There is a very strong NOE between these imino protons, as clearly seen in a
one-dimensional spectrum (Fig.
2
A). A weak resonance at ~13.9 p.p.m. grows more intense at lower temperatures and potentially orignates from 1065U at the base of the hairpin loop, but
an NOE to make this assignment is lacking. Cytidine amino and aromatic protons
could be deduced from the same NOESY spectrum (Fig.
2
B). Additional assignments rely on cross-peaks in the region between 4.7-5.7 p.p.m. (pyrimidine H5) and 6.5-8.5 p.p.m. (pyrimidine H6 and amino) and are listed in Table
1
.
Assignment of the loop GAAG sequence H1' and H8 depends on the assumption that an NOE between 1072C H1' and an aromatic proton involves H8 of the adjacent 1071G. The
lack of pyrimidines (with their characteristic H5-H6 cross-peaks) within this sequence and the paucity of NOEs between these
purines (no NOEs between H8 protons were observed) means that we have no
independent confirmation of the H1'-H8/H6 NOE walk within 1068-1071. Through-space assignments in loops can also be misleading.
For these reasons the Table
1
assignments of 1066-1072 should be considered provisional.
The other sugar protons in the same spin system as a given H1' were assigned from examination of a NOESY spectrum at short mixing times
(100 ms), a ROESY spectrum to distinguish primary NOEs and TOCSY and DFQ-COSY spectra, using standard methods (
23
,
25
). The TOCSY spectrum and many of the assignments are shown in Figure
4
. Most of the H5'/H5'' protons were not observed in the TOCSY spectrum and could
not be assigned.
The chemical shifts of all the assigned sugar protons are summarized in Table
1
and the coupling constants measured from a DFQ-COSY spectrum are listed in Table
2
. The latter provide an estimation of the sugar conformations. The helix
nucleotides 1058G-1065U and 1072C-1079C have J
H3'-H4'
> 8 Hz and J
H1'-H2'
< 3 Hz, which corresponds to an ~3'
endo
conformation (
26
). An important point is that the sugar conformations of the 1060U-1078U mismatch nucleotides do not differ detectably from those of the
surrounding helix. Four of the loop nucleotides have comparable J
H1'-H2'
and J
H3'-H4'
couplings, which can be an indication that the sugar is a mixture of 2' and 3'
endo
conformers (
23
). The exception is 1070A, for which J
H1'-H2'
is unusually large and J
H3'-H4'
small, suggesting a 2'
endo
conformation.
A
31
P-
1
H HETCOR spectrum of the A' RNA was run and, as expected for an RNA of this size, showed extensive
overlap of the
31
P resonances. Only about seven
31
P-H5' cross-peaks could be resolved and fewer
31
P-H3' cross-peaks. None of these were the 1060U or 1078U nucleotides of
most interest, though the adjacent 1059G-H5' and 1061G-H5' correlations were observed. P-H3' and P-H5' coupling constants that
could be measured were in ranges typical for A-form RNA (
23
). In some hairpin and internal loops there are substantial shifts of one or two
31
P resonances away from the range of chemical shifts for helical backbone
phosphates and these are usually associated with unusual P-O bond torsions (
15
,
27
,
28
,
29
).
31
P chemical shifts in A' RNA were dispersed over only ~1 p.p.m. This observation does not imply that A' RNA has no unusual P-O torsions (see below), since factors such as O-P-O bond angle may also affect
31
P chemical shift (
30
).
Table 2
U[middot]U base pairing has been suggested by unexpected thermodynamic stability and slowly exchanging imino protons in self-complementary helices with tandem U[middot]U mismatches (
9
). Pairing is also indicated in this study of a single U[middot]U mismatch in an asymmetrical sequence context. In addition, we find
that a
trans
backbone conformation is needed to accommodate the mismatch. A rationale for
this conformation is the following. The Figure
5
C U[middot]U pair has preserved the Watson-Crick C1'-C1' distance by a substantial rotation (in the
plane of the pair) of one uracil. This rotation could be accomplished by
changing the sugar pucker to 1'
exo
, though at the expense of steric problems for the 2'-OH. The
trans
conformation at [gamma] neatly solves the problem by extending the backbone and allowing the
entire sugar to rotate.
Our model can be compared with two crystal structures with similar U[middot]U conformations. For the tandem U[middot]U pairs studied by Baeyens
et al.
(
10
), one
trans
backbone conformation is associated with each pair and one of the two pairs has
a large, 45o propeller twist that permits only one NH3-O2 hydrogen bond. It was suggested that this loss of a hydrogen bond
is compensated for by a water bridging the O4 positions of uracils in
neighboring pairs on opposite strands and an intranucleotide hydrogen bond
between O2 and H2'. The two U O2-O2' distances are too large for hydrogen bonding in our model,
though the 1060U O4-1077A NH6 distance, 2.6 Å, is short enough that a water could bridge these positions in the
major groove. A U[middot]pseudouridine pair is also observed closing the anticodon loop of tRNA
Gln
bound to its cognate synthetase (
32
). In this case, the U nucleotide has a [gamma] torsion of 97o, in between (+)
gauche
and
trans
conformations, and a water bridges the NH3-O2 positions, as in Figure
5
C. The structures of U[middot]U pairs in these two crystal structures and our NMR-based model suggest that U[middot]U pairs adopt a range of conformations, varying in the
extents of propeller twist, imino proton hydrogen bonding and backbone
distortion.
The hairpin loop studied here presents an interesting contrast with the same
size tRNA anticodon loops. It has been argued that the optimum way to build a
seven base loop on the end of an A-form helix is to stack five bases on the 5'-end of the helix, which then leaves a short distance, easily
spanned by 2 nt, between the 3'-end of the helix and the 5'-end of the stacked loop nucleotides (
35
). This is the structure seen in crystallized tRNAs (
31
,
36
) and in a solution study of the tRNA
Phe
anticodon hairpin (
37
). Though we cannot make a detailed model of the A' RNA hairpin loop, the unusual 2'
endo
sugar pucker at 1070A suggests that an anticodon-like structure does not apply to it.
There are several reasons why the anticodon loop structure should not be general
for 7 nt loops. tRNA anticodon loops have a `U-turn' structure, in which a highly conserved uridine at the second
position of the loop forms a hydrogen bond between its imino proton and the 5' phosphate of position 5 (
38
). This structure is quite stable and has been found in other loop contexts (
15
,
39
). A' RNA does not have a uridine in the correct position to form this
hydrogen bond and neither does it show a shifted
31
P resonance characteristic of the
trans
P-O bond within the turn (
15
,
40
). In addition, tRNA
Phe
has two modified bases that may increase the stability of the stacked
conformation. The base pair closing the anticodon loop is a pseuodouridine-adenine pair and NMR studies have shown that the pseudouridine stabilizes
the base pair substantially compared with uridine at the same position (
41
). Position 1 of the loop is 2'-
O
-methylcytosine; based on NMR studies (
37
), it has been proposed that this methyl has a hydrophobic interaction with the
hypermodified Y base across the loop (position 6). In the absence of a U-turn and modified bases, A' RNA evidently prefers a conformation different than that of a tRNA
anticodon loop.
There is some reason to think that the stem structure is altered by formation of
tertiary structure in the intact rRNA. The 1051-1108 rRNA fragment has a set of tertiary interactions which are
dramatically stabilized by the mutation 1061U -> A (
6
). It appears that 1077A, which opposes nt 1061, is an important participant in
the tertiary structure and that pairing with 1061U competes with formation of
the tertiary structure. In support of this are the observations that 1077A is
universally conserved, while 1061 can be U, G or A and that 1061U reacts with a
single strand-specific reagent in the intact rRNA (
42
). The fact that only thermophiles have evolved the more stable 1061A structure
hints that the 1061U-1077A pairing is necessary to preserve a balance between two competing
structures (
6
). The features of the A' hairpin and stem that have been observed in this work should therefore
be taken as one of several possible conformations that may be adopted during
ribosome assembly and the ribosome cycle.
This work was supported by NIH grant GM29048.
+
Present address: National Institutes of Health, Building 30, Bethesda, MD 20892,
USA
J
H1'-H2'
J
H3'-H4'
1058G
2.9
8.5
1059G
<2.0
NA
1060U
<2.0
9.0
1061U
2.6
9.0
1062G
<2.0
8.5
1063G
<2.0
9.0
1064C
<2.0
9.0
1065U
<2.0
9.0
1066U
4.9
5.3
1067A
4.8
6.0
1068G
5.8
4.7
1069A
NA
NA
1070A
10
3.7
1071G
6.0
4.5
1072C
3.0
8.3
1073A
2.9
8.1
1074G
<2.0
9.0
1075C
<2.0
9.0
1076C
<2.0
9.0
1077A
2.7
9.0
1078U
2.0
8.5
1079C
2.6
8.5
1080C
2.8
7.7
1081C
2.8
7.7
REFERENCES
Return