Article No. jmbi.1999.3080 available online at http://www.idealibrary.com on
J. Mol. Biol. (1999) 292, 467±483
COMMUNICATION
Singly and Bifurcated Hydrogen-bonded Base-pairs in
tRNA Anticodon Hairpins and Ribozymes
Pascal Auffinger and Eric Westhof*
ModeÂlisations et Simulations
des Acides NucleÂiques, UPR
9002, Institut de Biologie
MoleÂculaire et Cellulaire du
CNRS, 15 rue Rene Descartes
67084 Strasbourg Cedex
France
The tRNA anticodon loops always comprise seven nucleotides and is
involved in many recognition processes with proteins and RNA fragments. We have investigated the nature and the possible interactions
between the ®rst (32) and last (38) residues of the loop on the basis of
the available sequences and crystal structures. The data demonstrate the
conservation of a bifurcated hydrogen bond interaction between residues
32 and 38, located at the stem/loop junction. This interaction leads to the
formation of a non-canonical base-pair which is preserved in the known
crystal structures of tRNA/synthetase complexes. Among the tRNA and
tDNA sequences, 93 % of the 32 38 oppositions can be assigned to two
families of isosteric base-pairs, one with a large (86 %) and the other with
a much smaller (7 %) population. The remainder (7 %) of the oppositions
have been assigned to a third family due to the lack of evidence for
assigning them into the ®rst two sets. In all families, the Y32 R38 basepairs are not isosteric upon reversal (like the sheared G A or wobble
G U pairs), explaining the strong conservation of a pyrimidine at position 32. Thus, the 32 38 interaction extends the sequence signature of the
anticodon loop beyond the conserved U-turn at position 33 and the
usually modi®ed purine at position 37. A comparison with other loops
containing both a singly hydrogen-bonded base-pair and a U-turn
suggests that the 32 38 pair could be involved in the formation of a base
triple with a residue in a ribosomal RNA component. It is also observed
that two crystal structures of ribozymes (hammerhead and leadzyme)
present similar base-pairs at the cleavage site.
# 1999 Academic Press
*Corresponding author
Keywords: Nucleic acid conformation; tRNA; anticodon; leadzyme;
hammerhead
Introduction
Hydrogen bonds between bases constitute the
cement of nucleic acid structures. Among nucleic
acid base-pairs, the well-known Watson-Crick GC
and A U pairs involve three and two hydrogen
bonds, respectively. Besides these regular WatsonCrick interactions, a large array of possible basepair interactions involving two hydrogen bonds
have been enumerated and many of them have
been observed in crystal structures (Tinoco, 1993;
Dirheimer et al., 1995; Leontis & Westhof, 1999a).
Such non-Watson-Crick interactions subtend the
great architectural diversity of nucleic acids, and
especially of RNA molecules where they are recurrently observed. Further, non-Watson-Crick pairs
E-mail address of the corresponding author:
[email protected]
0022-2836/99/380467±17 $30.00/0
contribute to the structural ¯exibility of these
molecules, since they are often strongly context
dependent. A less recognized class of base-pairs
comprises those in which the two bases are linked
by a single or a bifurcated hydrogen bond.
In transfer RNAs, the anticodon loop plays a
central role in processes associated with protein
synthesis. On the one hand, it must interact speci®cally with its cognate synthetase, and on the other
hand, it must ®t in the common tRNA binding site
of the ribosome for codon recognition on the messenger RNA. Over the years, rather extensive databases at the tRNA level (546 sequences) or at the
tDNA level (2726 sequences) have been compiled
(Sprinzl et al., 1998). Several years ago, it was
noticed that the ®rst (32) and last (38) residues of
the seven-membered anticodon loop are involved
in a bifurcated hydrogen bond contact (Quigley &
# 1999 Academic Press
468
Singly and Bifurcated Hydrogen-bonded Base-pairs
Figure 1. Secondary structures of tRNA anticodon hairpins for which crystal structures are accessible in the NDB.
The 32 38 base-pairs are shown in bold.
Rich, 1976; Westhof et al., 1985). Later, multiple
molecular dynamics simulations performed with
appropriate electrostatic treatment showed a surprising stability for this 32 38 contact (Auf®nger &
Westhof, 1996, 1998a; Auf®nger et al., 1999). Still
more surprising, in the crystal structure of the
speci®c complexes between tRNAs and their cognate synthetase (Rould et al., 1991; Ruff et al., 1991;
Cusack et al., 1996a, 1998), similar contacts between
residues 32 and 38 occur despite a severe unfolding of the rest of the anticodon loop. Therefore, we
decided to investigate the available crystal structures of tRNAs and their complexes with tRNAsynthetases, in the light of their base distribution in
the tRNA sequence database (Sprinzl et al., 1998),
in order to rationalize the possible interactions
between residues 32 and 38 (Figure 1).
Results
Crystallographic structures of free tRNAs
In all tRNA crystal structures present in the
NDB database (see Table 1), the ®rst (residue 32)
and last (residue 38) bases of the anticodon loop
(Figure 1), are linked through a bifurcated hydrogen bond and form a ``pseudo'' base-pair
(Figure 2). The only exception is the trna03 structure (Brown et al., 1985) obtained at low pH in the
presence of Pb2‡ (in trna03, a Pb2‡ is intercalated
between C32 and the hypermodi®ed Y37 base,
possibly preventing the formation of a 32 38 interaction). This particular anticodon loop conformation is probably not adopted by the tRNA
molecule under physiological conditions and will
not be further discussed.
In tRNAAsp, a 32C38 and, in tRNAPhe, a
C32 A38 base-pair are present ( stands for a
pseudouridine where the C5-H group is replaced
by a N1-H group and the C10 -N1 link between the
base and the sugar is replaced by a C10 -C5 bond;
see Figure 2). The 32 and 38 bases are linked by a
bifurcated hydrogen bond established between a
carbonyl oxygen atom and an amino group, i.e.
and
O4(32) . . . N4(C38)
in
tRNAAsp
Phe
O2(C32) . . . N6(A38) in tRNA . Interestingly, the
carbonyl and amino groups as well as the glycosidic bonds of both base-pairs superimpose very
well. Thus, although a signi®cant out-of-plane
deviation is observed (Figure 2), these 32 38 pairs
can be considered as isosteric (Figure 2). It is noteworthy that the observed out-of-plane deviation
does not affect the interaction scheme of the two
bases or the inter-base C10 (32) . . . C10 (38) distance,
Ê as in Watson-Crick pairs. Yet, the
close to 11 A
deviation of these isosteric pairs from a WatsonCrick arrangement is clearly illustrated by the
Ê ) between the C10 atoms of
large distance (7 A
residue 38 in the bifurcated hydrogen bonded
32 38 pairs and in the corresponding Watson-Crick
arrangement (Figure 2). Interestingly, a U C pair
identical with that observed in yeast tRNAAsp has
been observed in the crystal structure of the alphasarcin loop of 23 S rRNA (Correll et al., 1998). A
different situation is observed in the crystal structure of yeast tRNAMet
(Basavappa & Sigler, 1991),
i
where the C32 A38 pair (Figure 3) is not isosteric
to the 32 38 pairs described above. The authors
reported that the resolution of the anticodon hairpin is very low. Hence, this interaction scheme is
probably of poor structural relevance and will not
be further discussed. This view is also supported
by recent NMR studies of two transcripts of yeast
and Escherichia coli tRNAMet
anticodon
tRNAMet
i
m
hairpins (Schweisguth & Moore, 1997) from which
it was concluded that the observed non-canonical
C32 A38 NMR base-pairing are much closer to
those observed in yeast tRNAPhe crystal structures
structhan the one observed in the yeast tRNAMet
i
ture (Schweisguth & Moore, 1997).
A recent crystal structure of the complex
between E. coli tRNACys and the translation
elongation factor EF-Tu reveals the arrangement of
a 32 A38 base-pair (pr0004; Nissen et al., 1999).
Similarly to what is observed in the crystal structure of the complex of tRNAPhe with EF-Tu
(ptr012; Nissen et al., 1995), the anticodon loop of
tRNACys adopts a conformation close to that seen
in the crystal structures of uncomplexed tRNAs
with a non-Watson-Crick 32 A38 pair. Again the
A pair is isosteric to the C pair observed in
Singly and Bifurcated Hydrogen-bonded Base-pairs
469
Figure 2. The 32 38 base-pairs observed in crystal structures of uncomplexed tRNAs (see Table 1). (a) Average
structures of the 32 C38 and C32 A38 pairs extracted from the crystal structures of uncomplexed tRNA molecules,
including also the structure of the yeast tRNAPhe/RS complex (ptr012) where the anticodon loop does not interact
with the protein. The individual structures used for the averages are drawn with small lines. (b) Superposition (left)
of the 32 C38 (yellow) and C32 A38 (red) average structures showing the isostericity of these base-pairs in the
tRNA anticodon loop structural context. For comparitive purposes, a regular A U Watson-Crick pair is also drawn.
Orthogonal view (right) showing the observed out-of-plane deviation of these pairs. In all these Figures, the glycosidic bonds of nucleotides at position 32 have been superimposed.
yeast tRNAAsp and to the C A pair observed in
yeast tRNAPhe (Figure 2). Interestingly, in both
tRNACys and tRNAPhe (trna06) structures, a water
molecule is seen in the vicinity of the N7 and N6H groups of A38. In trna09, there is a comparable
water molecule which belongs, however, to the
coordination sphere of a Mg2‡.
Hence, from the available tRNA crystal structures where the anticodon hairpin keeps its folded
structure, it can be deduced that a set of non-canonical but isosteric base-pairs involving a bifurcated
hydrogen bond between a carbonyl oxygen atom
and an amino group is recurrently present at the
junction between the stem and the loop of the
anticodon hairpin. On a thermodynamic basis,
base-pairs involving a single or bifurcated hydrogen bond should be more labile compared to those
involving two or three hydrogen bonds. However,
as described below, formation of speci®c tRNA/
synthetase complexes, which involve the splaying
out toward the exterior of ®ve (33 to 37) of the
bases of the loop, do not disrupt the direct hydrogen bonding contacts established between bases 32
and 38.
Crystallographic structures of tRNA/
synthetase complexes
The coordinates of several crystal structures of
tRNAs complexed with their cognate synthetase
are available (Table 1). Upon formation of the
tRNA/synthetase complexes, slight conformational
rearrangements of the 3238 base-pair are
observed.
32 C38 pairs. Two structures of the complex of
the wild-type yeast tRNAAsp, including all the
modi®ed bases with its cognate synthetase
(tRNAAsp/RS) are available (Table 1). In these
structures, the 32 C38 pair adopts a conformation different from that observed in the uncomplexed tRNA structures (Figure 4). The C38 amino
group forms a single hydrogen bond rather than a
bifurcated hydrogen bond with the O4(32) atom.
Further, a signi®cant out-of-plane orientation of
base C38 is observed. These arrangements appear
less stable than those formed in uncomplexed
tRNAs and are, possibly, stabilized by additional
protein/tRNA interactions.
U32 U38 and Um32 38 pairs. No structure has
been solved of an uncomplexed tRNA having a
U32 U38 pair. However, in the structures of the
transcripts of E. coli tRNAGln/RS complex (Table 1),
a U32 U38 pair is observed (Figure 4). Again, a
single hydrogen bond links the two bases. Note
that this base-pair is approximately isosteric to a
Watson-Crick pair (see the bottom drawing in
right-hand column of Figure 6). Thus, its geometry
470
Singly and Bifurcated Hydrogen-bonded Base-pairs
Table 1. Description of the crystal structures of complexed and uncomplexed tRNA molecules referenced in the NDB
(Berman et al., 1992) before the 1st August 1999
NDB codea
Free tRNAs
trna01
trna02
trna03
trna04
trna05
trna06
trna07
trna08
trna09
trna10
trna12
tRNA/EF-TU complexes
pr0004
ptr012
tRNA/RS complexes
pte001
pte002
pte003
ptr001
ptr002
ptr003
ptr004
ptr005
ptr007
ptr008
ptr009
ptr010
ptr011
ptr014
ptr015
PDB code
Ê)
Res. (A
3238
Transcript
Coord.
/
/
1TN2
6TNA
/
1TRA
2TRA
3TRA
4TRA
4TNA
1YFG
yeast
yeast
yeast
yeast
yeast
yeast
yeast
yeast
yeast
yeast
yeast
tRNAPhe
tRNAPhe
tRNAPhe
tRNAPhe
tRNAAsp
tRNAPhe
tRNAAsp
tRNAAsp
tRNAPhe
tRNAPhe
tRNAMet
i
2.5
3.0
3.0
2.7
3.0
3.0
3.0
3.0
3.0
2.5
3.0
CmA
CmA
CmA
CmA
C
CmA
C
C
CmA
CmA
C A
N
N
N
N
N
N
N
N
N
N
N
N
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
1B23
1TTT
E. coli tRNACys
yeast tRNAPhe
2.6
2.7
A
CmA
N
N
Y
Y
/
/
1QTQ
1GSG
1GTS
1GTR
1SER
1ASY
T. thermo. tRNAPro(GGG)
T. thermo. tRNAPro(CGG)
E. coli tRNAGln
E. coli tRNAGln
E. coli tRNAGln
E. coli tRNAGln
T. thermo. tRNASer
yeast tRNAPhe
T. thermo. tRNASer
yeast tRNAAsp
E. coli tRNAGln
E. coli tRNAGln
E. coli tRNAGln
T. thermo. tRNALys
T. thermo. tRNALys
3.6
3.5
2.2
2.8
2.8
2.5
2.9
2.9
2.7
3.0
2.6
2.7
3.0
2.7
2.9
UU
UA
UU
Um
UU
UU
C A
C
C A
C
UU
UU
UU
C A
C A
N
N
Y
N
Y
Y
N
N
N
N
Y
Y
Y
N
Y
N
N
Y
Nb
Y
Y
Yc
Y
N
Y
Y
Y
Y
N
N
1ASZ
1QRS
1QRT
1QRU
/
/
a
Crystal structure references are given next: (trna01, Quigley et al. (1978); trna02, trna03, Brown et al. (1985); trna04, Sussman et al.
(1978); trna05, Comarmond et al. (1986); trna06, Westhof & Sundaralingam (1986); trana07, trna08, trna09, Westhof et al. (1988);
trna10, Hingerty et al. (1978); trna12, Basavappa & Sigler (1991); pr0004, Nissen et al. (1999); ptr012, Nissen et al. (1995); pte001,
pte002, Cusack et al. (1998); pte003, Rath et al. (1998); ptr001, Rould et al. (1989), ptr002, Perona et al. (1993); ptr003, Rould et al.
(1991); ptr004, Biou et al. (1994); ptr005, Ruff et al. (1991); ptr007, Cusack et al. (1996a,b); ptr008, Cavarelli et al. (1994); ptr009, ptr010,
ptr011, Arnez & Steitz (1996); ptr014, ptr015, Cusack et al. (1996a,b).
b
Only backbone coordinates have been deposited in the NDB.
c
The coordinates for the atoms of the anticodon hairpin could not be derived from the crystal data.
is clearly distinct from that observed in the crystal
structures of uncomplexed tRNAs for C A, U A,
or U C pairs. The stabilization of such base-pairs
involves a network of water molecules, some linking one base to the other and some linking a base
to its backbone (see Figure 3 by Rould et al., 1991).
An additional contact between Asn370 and U38 is
present in these structures. The Um32 38 basepair of the E. coli tRNAGln/RS complex, which
includes all tRNA modi®ed residues, displays a
geometry identical with that observed for the
tRNA transcripts (Rould et al., 1991).
Other 32 38 pairs. In the structures of the two
complexes formed between Thermus thermophilus
tRNALys synthetase and transcripts of E. coli and
T. thermophilus tRNALys, the occurrence of a nonstandard C32 A38 base-pair has been reported
(Cusack et al., 1996a; S. Cusack, personal communication). Beside contacts made by A38 with the protein, a N3(C32) . . . N6(A38) is formed. This contact
is different from those observed in the structures of
uncomplexed tRNAPhe. Furthermore, in the struc-
tures of the two complexes of T. thermophilus
tRNAPro synthetase with two tRNAPro isoacceptors,
Ê resolution, the occurrence of a
determined at 3.5 A
non-standard U32 U38 and of a Watson-Crick
U32 A38 base-pair is mentioned (Cusack et al.,
1998; S. Cusack, personal communication).
Phylogenetic analysis
tDNA gene sequences. In order to detect any
sequence preferences for base-pairs at position
32 38, the tRNA gene database which contains up
to 2726 tDNA sequences (Sprinzl et al., 1998) was
analyzed. First, there is a very large proportion of
pyrimidines at position 32 (98 %, see Table 2) and
a signi®cant number of purines at position 38
(71 %). Secondly, sequences which could form
C G (six occurrences), G C (no occurrence) or A U
(ten occurrences) Watson-Crick pairs are very rare.
Nevertheless, there is a large proportion of
U32 A38 pairs (18 %). Note that the number of
U A pairs might be slightly lower since it has been
Singly and Bifurcated Hydrogen-bonded Base-pairs
Ê
Figure 3. The C32 A38 pair observed in the 3.0 A
(Basavappa & Sigler,
crystal structure of yeast tRNAMet
i
1991). This pair displays a unique conformation which
cannot be easily compared to the base pair arrangements shown in Figure 2. For comparison, a C G Watson-Crick pair and a C32 A38 pair extracted from yeast
tRNAPhe (see Figure 2) are also represented. The glycosidic bonds of 32 have been superimposed.
shown that U ! C editing can occur at position 32
(Beier et al., 1992).
The preferred base-pair sequence is C A
(48 %), which is followed by the T A (18 %),
T T (11 %), C C (8 %) and T C (8 %) pairs.
Most of these pairs, with the exception of the T A
pair, cannot form Watson-Crick arrangements.
Note that these proportions do not vary signi®cantly if one considers the two following subsets:
subset 1 (932 sequences of tDNA elongators) and
subset 2 (352 sequences of eukaryotic tDNA elongators) of the whole 2726 tDNA sequences.
471
tRNA sequences including modified bases. The
tRNA sequence database is approximately ®ve
times smaller (546 sequences) than the tDNA database (2726 sequences). Despite this reduced number of sequences, useful information can be
gathered due to the presence of modi®ed nucleotides (Table 3). It appears that, at least for C A or
U C pairs, only modi®ed bases compatible with
the formation of the non-canonical interaction
shown in Figure 2 are reported. For example, the
occurrence of C32 A38 pairs is close to 42 %; that
of C A pairs close to 11 % (a methylation of the
20 -hydroxyl group does not prevent the formation
of a bifurcated hydrogen-bonded pair as shown in
yeast tRNAPhe); that of m3C‡ close to 4 % (the
methylation of the N3 atom of a cytosine base is
compatible with the C A arrangement represented
Figure 2); and that of a s2C A pair close to 1 % (a
sulfur atom replacing the O2 atom of a cytosine
base could also form a S2(s2C32) . . . N6(A38) bifurcated hydrogen bond). Thus, 59 % of C A pairs are
observed at position 32 38. The majority of these
modi®ed C A pairs can adopt the conformation
shown in Figure 2. Similarly, all the modi®ed U U
pairs reported in Table 3 are compatible with the
Watson-Crick like arrangement shown in Figure 4.
Is the sequence at position 32 38 specific for a
given tRNA type? In order to detect a possible correlation between the sequence of the 32 38 pair
and either the amino acid type of the tRNA or the
phylogenetic group in which they occur, the family
of tDNAs containing a T32 T38 pair was analyzed
(Sprinzl et al., 1998). The T T pairs are found in
tDNAs coding for 14 different amino acid residues,
in eukaryotic, eubacterial, mitochondrial, and
chloroplastic tRNAs (in archaea, only one occurrence is reported). Further, no correlation between
the ``base type'' at positions 37 and 38 could be
noted (146 A37 T38 and 140 G37 T38 sequences).
As in the case of tDNAs containing a T32 T38 pair,
Figure 4. The 32 38 base-pairs observed in crystal structures of tRNAs complexed with synthetases (see Table 1).
The C base-pairs are shown on the left and U U base-pairs are shown on the right. For comparative purposes, a
regular U A Watson-Crick pair as well as the average C pair of free tRNAs shown Figure 1 are drawn. The glycosidic bonds of 32 have been superimposed.
472
Singly and Bifurcated Hydrogen-bonded Base-pairs
Table 2. Percentages of occurrence of the four A, G, C and T bases at the non-canonical 32 38 base-paired position
for all 2726 tDNAs contained in the tRNA database (Sprinzl et al., 1998), a subset of 932 elongator tDNAs, and a subset of 352 eukaryotic elongator tDNAs (percentages larger than 5 % are idicated in bold)
Base 32
A
G
A
G
C
T
Total:
R/Ya
1
0
48
18
66
69
0
0
3
3
Base 32
A
G
A
G
C
T
Total:
R/Ya
1
54
13
68
69
0
1
1
2
Base 32
A
G
A
G
C
T
Total:
R/Ya
1
58
2
89
61
0
0
0
1
a
C
Base 38
All tDNAs (2726)
T
0
0
0
1
8
1
8
11
16
13
30
All tDNA elongators (932)
C
T
0
0
12
1
7
10
18
11
30
All tDNA eukaryotic elongators (352)
C
T
14
8
22
39
2
15
17
Tot.
2
1
56
39
100
Tot.
1
68
30
100
Tot.
1
74
25
100
R/Ya
3
97
100
R/Ya
1
99
100
R/Ya
1
99
100
Indicates the percentages of purine (R) and pyrimidine (Y) bases at position 32 and 38.
no apparent correlation exists for tDNAs displaying other types of 32 38 pairs.
Model structures for the 32 38 interaction in tRNAs
C A, U A, U C, C C and U U base-pairs. From
the preceding structural and phylogenetic data, it
is possible to propose models for base-pairs located
at position 32 38 of tRNA anticodon hairpins
(Figure 5). As the vast majority of 32 38 pairs are
of the C A type, the arrangement observed in the
crystal structures of yeast tRNAPhe (see Figure 2)
seems the most plausible (family I). The same con®guration can be adopted by U A pairs which
may be linked by a bifurcated rather than by two
regular hydrogen bonds as in Watson-Crick pairs
(Figure 5). As shown before, the much rarer U C
and C C pairs could adopt conformations isosteric
to C A and U A pairs. Interestingly, the shallow
groove patterns of C A and U A pairs are similar,
as are the shallow groove patterns of U C and
C C pairs. Note that the U C pair shown in
Figure 5 is different from the U C pair observed in
the crystal structure of several RNA oligomers
(Holbrook et al., 1991; Cruse et al., 1994; Tanaka
et al., 1999) where the two bases are linked by a
O4(U) . . . N4(C) instead of an O2(U) . . . N4(C)
hydrogen bond with a water molecule bridging the
two N3 sites of both pyrimidines.
However U U pairs, which account for approximately 7 % of the total number of pairs, cannot
adopt a conformation isosteric to that observed for
C A pairs of family I. The arrangement shown in
Figure 5 is the one observed in the crystal struc-
tures of the wild-type and transcripts of E. coli
tRNAGln complexed with its cognate synthetase
(see Figure 4). It is proposed to be a representative
of family II. tRNAs with a U U pair at position
32 38 may also belong to a separate family of
tRNAs with speci®c and unknown features.
Rare 32 38 base-pairs. Besides the pairs in
families I and II shown in Figure 5, several rare
base oppositions are found at positions 32 38.
They account for 7 % of the total number of tDNA
(Table 2) and tRNA (Table 3) sequences and are
gathered in family III. Table 2 indicates that only
G C pairs are completely avoided in the 2726
sequences reported in the tRNA database (Sprinzl
et al., 1998). For several of these pairs, pairing
arrangements approximately isosteric to family I or
to family II can be suggested (Figure 6). In the following, only cis pairing types will be considered.
A A (1.1 %, see Table 2), GU (0.6 %), and G A
(0.1 %) pairs are roughly isosteric to C A pairs
(family I of Figure 5), while U G (2.7 %), C U
(1.3 %), and A U (0.4 %) pairs are approximately
isosteric to U U pairs (family II shown in Figure 5).
For the ®rst family, the A13 A22 pair, found in the
crystal structure of E. coli tRNAGln (pte003), is close
to the bifurcated C A pair. For G U pairs, no
structures close to that of the C A pair have been
observed and, thus, the G U pair shown in Figure 6
is suggested. The crystal structure of the 5 S rRNA
loop E (ur1064; Correll et al., 1997) contains a
sheared G98 A78 pair superposable to the singly
hydrogen bonded CA pair. For the second family,
the GU and A U pairs shown in Figure 6 have
473
Singly and Bifurcated Hydrogen-bonded Base-pairs
Table 3. Percentages of occurrence of the four A, G, C and U bases at the non-canonical 32 38 base-paired position
for all 546 tRNAs contained in the tRNA database (Sprinzl et al., 1998), a subset of 367 elongator tRNAs, and a subset
of 191 eukaryotic elongator tRNAs
Base 32
A
G
C
A
G
C
Cm
m3C‡
s2C
U
Um
m
Total:
R/Ya
1
0
42
11
4
1
5
1
3
68
70
0
2
2
7
0
4
5
0
16
30
Base 32
A
G
C
A
G
C
Cm
m3C‡
s2C
U
Um
m
Total:
R/Ya
35
14
5
1
7
1
2
66
68
1
1
5
2
7
0
4
1
16
32
Base 32
A
G
C
A
G
C
Cm
m3C‡
s2C
U
Um
m
Total:
R/Ya
27
23
10
2
64
64
-
5
1
1
9
1
18
36
a
Base 38
All tRNAs (546)
m5C
U
1
1
0
2
0
1
0
2
0
0
4
All tRNA elongators (367)
m5C
U
2
1
1
3
1
1
0
3
0
4
All tRNA eukaryotic elongators (191)
m5C
U
3
2
1
6
0
1
0
1
Tot.
0
2
2
1
1
7
1
0
53
12
4
1
16
2
2
0
100
Tot.
0
2
3
2
1
8
1
0
47
16
5
1
17
3
8
1
100
Tot.
1
4
2
2
2
10
0
1
40
26
10
4
3
13
1
100
R/Ya
1
99
100
R/Ya
1
100
R/Ya
99
100
Indicates the percentages of purine (R) and pyrimidine (Y) bases at position 32 and 38.
been modeled in order to be roughly isosteric to
the U U pair while interacting only through a
single hydrogen bond. The C U pair is extracted
from the same crystal structure of an RNA duplex
(ar0005; Tanaka et al., 1999). Note that NMR data
obtained for E. coli tRNAPhe revealed signals
associated to a rare A32 38 interaction (Hyde &
Reid, 1985a,b). For all these base-pairs, slight
changes in the positions of the bases may
``improve'' their isosteric character. For A G
(0.3 %), A C (0.2 %), CG (0.2 %), G C (0.04 %),
and G C (0 %) oppositions, no arrangement close
to either the C A or the U U pair shown in
Figure 5 can be proposed. Yet, for these rare basepairs, the possibility of base editing (Price & Gray,
1998) and the probable existence of a noise level in
the tRNA sequence database leading to a small
number of incorrect sequences should be taken
into account. Therefore, the sequences and model
structures of family III should be considered with
caution.
Discussion
A conserved non-canonical 32 38 base-pair in the
anticodon loop of tRNAs
The preceding structural and phylogenetic data
lead to the conclusion that particular types of pairing occur systematically between the ®rst (32) and
last (38) residues in the anticodon loop of tRNAs.
The largest family of base-pairs (family I) comprises the C A pair, where the two bases are linked
by a bifurcated hydrogen bond, and the isosteric
U A, U C and C C pairs (Figure 5). This category
represents 86 % of the total number of sequences.
A second family of base-pairs (family II) comprises
the U U pairs (7 %), which are not isosteric to the
474
Singly and Bifurcated Hydrogen-bonded Base-pairs
Figure 5. The possible single
hydrogen-bonded arrangements of
32 38 base-pairs in anticodon loop
of tRNAs organized in three
families. In family I, the structure
of the U A pair is inferred from
that of the C A pair extracted from
yeast tRNAPhe, and the structure of
the C C pair is inferred from that
of the U C pair extracted from
yeast tRNAAsp (see Figure 2). In
family II, the structure of the U U
pair is extracted from the E. coli
tRNAGln/RS complex (see Figure 4).
Family III gathers the rare oppositions. The glycosidic bonds of base
32 are similarly oriented in all the
base-pair oppositions.
pairs of family I. Finally, a third family (family III)
comprises a set of rare base-pairs (7 %) for which
few structural data are available (Figure 6), but
which could be roughly distributed among the ®rst
two families.
Importance of bifurcated hydrogen bonds
Up to now, pairs where the bases are linked by
a single or a bifurcated hydrogen-bonding interaction have encountered limited interest, probably
because of their apparent lability. Yet, such pairs
were observed in the crystal structures of yeast
tRNAPhe (Quigley & Rich, 1976) and yeast tRNAAsp
(Westhof et al., 1985) at position 32 38 (where two
chemical groups are involved) and at position
G18 55 (where three chemical groups are
involved). Furthermore, subsequent molecular
dynamics simulations have shown that they are
stable over nanosecond time scales in yeast
tRNAAsp (Auf®nger & Westhof, 1996; Auf®nger
et al., 1999). Since the resolution of the ®rst tRNA
crystal structures, the bestiary of non-canonical
base-pairs including bifurcated hydrogen bonds
has been extended and comprises, among others,
G G and G U pairs (Correll et al., 1997). Thus, the
consideration of singly or bifurcated hydrogen-
bonded pairs should improve the structures resulting from crystallographic, NMR, and modeling
re®nement processes. The fact that such pairs are
additionally stabilized by water-mediated interactions should also be taken into account (Rould
et al., 1991; Auf®nger & Westhof, 1998c).
Why a conserved pyrimidine at position 32
in tRNAs?
The non-canonical C32 A38 base-pair is characterized by a Watson-Crick-like distance between
the C10 atoms and an asymmetric disposition of
the angles at the C10 atoms (Figure 2). The latter
geometrical characteristic leads to a pronounced
non-isostericity upon reversal of the base-pair.
Thus, C A pairs are not isosteric to A C pairs
(Figure 7), U A pairs are not isosteric to A U
pairs, and U C pairs are not isosteric to C U pairs.
The C C and U U pairs shown in Figure 5 are
obviously self-isosteric. Thus, the anticodon loop
must begin with a pyrimidine at position 32 in
order to adopt its characteristic functional shape.
These structural considerations are in agreement
with phylogenetic data indicating that, in tRNAs, a
pyrimidine is present in 98 % of the sequences at
position 32 (Tables 2 and 3).
Singly and Bifurcated Hydrogen-bonded Base-pairs
475
Figure 6. Possible arrangements
of the rare 32 38 oppositions gathered in family III. Only those
arrangements which can approximately be consisdered as isosteric
to the base-pairs of families I and II
have been represented. The structure of the C A pair (family I),
extracted from yeast tRNAPhe, is
shown in blue and the structure of
the U U pair (family II), extracted
from the E. coli tRNAGln/RS complex, is shown in red (see Figure 5).
The glycosidic bonds of residues 32
are similarly oriented in all the
base-pair oppositions.
Is isostericity always required?
As demonstrated by the Watson-Crick pairs, isostericity is a central concept in nucleic acid structure. From crystal and phylogenetic data gathered
from a set of sequences of the 5 S rRNA loop E,
isostericity has also been proposed for several noncanonical base-pairs (Leontis & Westhof, 1998a,b,
1999b). In the present study, a certain number of
isosteric arrangements have been observed at position 32 38. Such isosteric pairs may result in a
similar structure of the anticodon loop for free
tRNAs. However, the occurrence of a non-negligible number of U U and related pairs (7 %) which
are not isosteric to C A pairs indicates that,
depending on the context, isosteric base-pairs may
not be systematically required. Thus, such sets of
non-isosteric base-pairs may be characteristic of
different tRNA families, for which adjustments for
interactions with proteins, other RNA motifs or
divalent ions are required. Signi®cantly, when
base-pairs are not strictly isosteric to the C A or
U U pairs, their percentage of occurrence drops
considerably (see Figure 6).
Several biochemical studies have evaluated the
effects of changing the 32 38 base-pair type. Yarus
and co-workers (1986b) have studied the suppression ef®ciency of a large number of mutants of the
476
Figure 7. Bifurcated C A pairs are not isosteric to
A C pairs as shown by the large distance separating the
glycosidic bonds. Thus, a A C pair cannot replace a
C A pair at position 32 38, indicating why a pyrimidine
at position 32 is required in tRNAs. The glycosidic
bonds of 32 have been superimposed.
amber suppressor Su7 and have shown that no
mutation at position 32 or 38 was neutral. For
example, Um and Um C pairs at position 32 38
were found to be less ef®cient than Um A or
Cm A pairs (Yarus et al., 1986a). Interestingly, the
measured suppression ef®ciency is Um '
Um C < Um A < Cm A and, thus, follows the proportion of 32 38 pairs derived from phylogenetic
analysis (Figure 5). The substitution of the Cm32
nucleotide by any purine residue reduces the suppression ef®ciency more than tenfold (Smith &
Yarus, 1989) in agreement with the fact that purines are strongly disfavored at position 32.
Mutations of base 38 show less dramatic effects
indicating that a larger mutational variability is
allowed at position 38 than at position 32 (see
Tables 2 and 3). In transcripts of E. coli tRNAGly,
with a U32 A38 pair the anticodon discriminates
the glycine codons according to the wobble rules,
but with a C A pair it loses its discriminating
power (Lustig et al., 1993). In transcripts of Mycoplasma myciodes tRNAGly, the substitution of the
wild-type C32 A38) by a U A pair (Claesson et al.,
1995) resulted in similar effects (but the transformation was the reverse of that performed in E. coli
tRNAGly). Since in both cases an isosteric substitution was performed (C A$U A, see Figure 5),
these data indicate that the nature of the contacts
between the tRNA and the ribosome may involve
the 32 38 base-pair and that such a substitution
affects the recognition pattern of the anticodon
hairpin more than its three-dimensional fold. It has
also been demonstrated that substitution of
U32 A38 by a C A pair in M. myciodes tRNAGly
results in a substantial increase in its frameshifting
ef®ciency (O'Connor, 1998). In E. coli tRNAGln transcripts, substitution of the U32 C38 opposition by
a C C or CA pair resulted in an increase of both
aminoacylation and ribosome performance, but
enhanced the latter function to a greater extent
(McClain et al., 1998). It is possible that the replace-
Singly and Bifurcated Hydrogen-bonded Base-pairs
ment of a rare U32 U38 by a non-isosteric C A
pair has as a consequence the formation of a more
``anticodon'' like hairpin. Besides the isostericity of
several of these non-canonical base-pair types, each
of these base-pairs may interact speci®cally with
proteins and ribosomal elements (see below). Isostericity, thus, is a generally valid concept indicating that it is possible to substitute, with minor
structural changes, base-pairs displaying different
shallow and deep groove recognition patterns.
However, each of these isosteric base-pairs has
speci®c functional roles, experimentally characterized by different translational ef®ciency. Furthermore,
in
some
occurrences
non-isosteric
substitutions have been observed like the U U pair
of the rare base-pairs found in Tables 2 and 3. For
example, in E. coli, a minor tRNAAla isoacceptor
differs from the other isoacceptors in possessing a
rare purine instead of a pyrimidine at position 32.
However, the structure and functional role of such
base-pairs is not yet known.
Importance of modified nucleotides at positions 32
and 38
Modi®ed nucleotides are found in many RNA
molecules and especially in tRNAs where they
account for 12 % of the total number of residues
(Grosjean & Benne, 1998). A large number of modi®ed nucleotides are located at the stem/loop junction of the anticodon hairpin in tRNAs (Auf®nger
& Westhof, 1998b). It has been proposed, in agreement with a large body of experimental evidence
and molecular dynamics simulations, that some of
these modi®cations, and especially pseudouridylation (Auf®nger & Westhof, 1998a), are required to
stabilize the three-dimensional structure of the
functionally important anticodon loop. At position
32, the stabilization mechanism involves the formation of a pseudouridine/water complex where
the water molecule mediates base to backbone
interactions, and thus helps to increase the stability
of the 3238 interaction pairs (Davis & Poulter,
1991; Arnez & Steitz, 1994; Auf®nger & Westhof,
1997, 1998a). The same nucleotide/water complexes can occur after pseudouridylation at positions 31, 38, and 39. Such modi®cations may thus
help to strengthen the structure of the tRNA anticodon hairpin in the ribosome and prevent excessive unfolding of the anticodon loop when
interacting with synthetase proteins.
Indeed, as noted above, in a few of the crystal
structures of tRNAs complexed with their cognate
synthetase, the 32 38 interaction is systematically
maintained even after partial unfolding of the
loop, and is stabilized by water mediated interactions involving modi®ed nucleotides (Roud et al.,
1989). Thus, one possible role of the 32 38 pair and
of associated modi®ed nucleotides is to limit the
``unfolding'' of the loop to the ®ve bases, 33 to 37.
The structure of base-pairs 31 39, 32 38, and the
stacking of U33 below base 32 being at least partially preserved, as shown in Figure 8(a) for the
Singly and Bifurcated Hydrogen-bonded Base-pairs
477
Figure 8. Turns in RNA motifs involving non-canonical base-pairs. (a) Views extracted from crystal structures of
yeast tRNAPhe (trna09) and tRNAAsp (trna07) as well as from the crystal structures of the complex of E. coli tRNAGln
with its cognate synthetase (pte003). For comparison, a fragment of a conventional Watson-Crick helix is drawn. The
position of the glycosidic bond, for a hypothetical 32-38 Watson-Crick base-pair in the tRNAPhe anticodon loop is
marked by a green arrow. (b) Fragments of the crystal structures of the TC loop of yeast tRNAPhe (trna09) and of
the GNRA tetraloop of the hammerhead ribozyme (uhx026). (c) Fragments of the crystal structure of the leadzyme
(ur0001) and the hammerhead ribozyme (uhx026) showing the C23 A45 and the C3 C17 base-pairs. In all these
Figures, the glycosidic bonds of 32 have been superimposed.
E. coli tRNAGln/RS complex, a subsequent ``refolding'' to the ribosomal active form, after decomplexation with the synthetase, may be facilitated.
Besides a structural role, the question of a possible
functional role of the non-canonical 32 38 base-
pair can be raised. It has been noted that in some
instances bases 32 and 38 can interact with the cognate synthetase and, thus, act as structural determinants in the recognition process. Biochemical data
fully support this view (for a review, see GiegeÂ
478
et al., 1998). Nevertheless, the largest number of
tRNA/synthetases do not use these bases as structural determinants and, thus, their structural role
seems to prevail over their direct implications in a
recognition process with a synthetase.
Is there a relationship between the 32 38
interaction and the U-turn?
The active three-dimensional structure of tRNA
anticodon loops is characterized by a U-turn
associated with the conserved base U33 and the Aform helical stack of loop residues 34 to 38 in continuity with the helical stem residues 39 to 44 (as
shown in Figure 8(a) for yeast tRNAAsp and yeast
tRNAPhe). While both residues 32 and 38 are
located in a helical track, the rotation of the glycosidic bonds on the side of residue 32 is much more
pronounced than in regular helical A-form stacks
(Figure 8(a)). Furthermore, between residues 34
and 38, the rotation angles are rather regular and
not far from the helical case. But, clearly, residue
38 has not rotated suf®ciently to form a WatsonCrick pair with residue 32. Hence, a substitution of
the non-canonical 3238 pair by a regular WatsonCrick pair would hinder the formation of the Uturn and the regularity of the loop. This explains
why no G C and very few C G pairs are reported
at these positions (Table 2). Therefore, the isosteric
C A, U A, U C, and C C pairs at position 32 38,
and possibly also U U pairs, appear as a crucial
transition element linking the stem to the loop of
the anticodon hairpin. Residues 32 and 38 should
Singly and Bifurcated Hydrogen-bonded Base-pairs
therefore be considered, along with the conserved
base U33 and likely a conserved purine at position
37, as a signature for tRNA anticodon loops
(Figure 9). The view that only speci®c base combinations occur at position 32 38 is in agreement
with the extended anticodon concept, indicating
that the coding performance of the triplet anticodon is enhanced by the appropriate loop and stem
sequence (Yarus, 1982; Yarus et al., 1986b; Yarus &
Smith, 1995).
In the seven base T-loops of tRNAs, a non-canonical two hydrogen-bonded trans Hoogsteen
T54 A58 pair precedes the U-turn. This base-pair
appears to play a similar role in T-loops as that
played by the 3238 pair in anticodon loops
(Figure 8(b)). Note that a T54 A58 pair is reported
in 78 % of tRNA gene sequences (Sprinzl et al.,
1998). Similarly, in the recent crystal structure of
the L11 binding region in LSU RNA, a trans Watson-Crick A U pair ¯anks the U-turn of a ®ve
nucleotide loop (Conn et al., 1999). GNRA tetraloops are known to form U-turns as well (Westhof
et al., 1989; Jucker & Pardi, 1995). In such motifs,
G A sheared pairs, which involve two hydrogen
bonds, are used to close the loop (Figure 8(b)).
Thus, sheared GA pairs, almost isosteric to C A
pairs (Figure 6), are used in short loops, presumably because of their intrinsic higher stability and
because they induce a sharper turn, while the
apparently more labile singly hydrogen bonded
base-pairs found in anticodon motifs are used in
longer loops.
Figure 9. tRNA anticodon loop signature showing the structure/sequence relationship at the 32 38 base-pair level.
The distribution of the four A, G, C, and U nucleotides is extracted from the set of 2726 tDNA gene sequences present in the tRNA database (Auf®nger & Westhof, 1998b; Sprinzl et al., 1998).
Singly and Bifurcated Hydrogen-bonded Base-pairs
Other examples of interfacial non-canonical basepairs
In hairpins. A recent NMR structure of the
tRNALys,3 anticodon hairpin obtained at low pH
revealed the formation of a C32 A‡38 base-pair
(Durant & Davis, 1999). The pKa of the adenine
base was estimated to be close to 6 for A38. Likewise, evidence for the formation of a non-canonical
bifurcated U C pair have been obtained by NMR
for the central hairpin of the HDV ribozyme containing a seven nucleotide loop (Kolk et al., 1997).
In this structure, the sharp turn which is associated
with the change of direction of the backbone is
shifted in the 30 direction by one nucleotide, and
characterized as a reversed U-turn. Interestingly,
an independent NMR determination of the structure of the same hepatitis delta virus hairpin proposed a different hydrogen-bonding scheme for
the U C pair of the nucleotides on the 50 and 30 ends of the loop (Lynch & Tinoco, 1998). In this latter structure, the cytosine amino group interacts
with O4 instead of O2 of U resulting in a water
inserted U C pair similar to that observed in the
crystal structure of RNA oligonucleotides
(Holbrook et al., 1991; Cruse et al., 1994; Tanaka
et al., 1999). In any case, in each example of structurally determined hairpins containing loops with
seven bases, a non-canonical base-pair is observed
at the junction between the stem and the loop.
Several examples of loops closed by non-canonical base-pairs can be found in crystal structures
(Pley et al., 1994; Scott et al., 1995; Cate et al., 1996;
Correll et al., 1998; Perbandt et al., 1998). A hairpin
closing wobble U U pair is found in the crystal
structure of the complex formed between a fragment of the U2 snRNA and a spliceosomal protein
(Price et al., 1998). Interestingly, in the crystal structure of the RNA-binding domain of the U1A
spliceosomal protein complexed with an RNA hairpin, no loop closing C A base-pair is observed
(Oubridge et al., 1994) indicating that, in this other
structural context, a possible non-canonical basepair is disrupted after complex formation.
In ribozymes. Interestingly, the crystal structure of
the leadzyme (Wedekind & McKay, 1999), a
C23 A45 pair, identical with the CA pair
observed in tRNAs, is present (Figure 8(c)). It has
been proposed, on the basis of crystallographic
data, that this non-canonical pair is involved in the
binding of a Pb2‡ in the RNA deep groove which
may be important for catalysis. In one motif of the
crystal structure of a Ba2‡ occupies this site (in
tRNA crystal structures, a related Mg2‡ binding
site is located in the deep groove of the 32 38
pair). Interestingly, the formation of a C A pair is
not mandatory for catalysis, since substitution
experiments have shown that the ribozyme is still
active after replacement of the adenine residue by
an abasic ribose (Chartrand et al., 1997).
In the hammerhead ribozyme, a U A non-canonical base-pair involving a single hydrogen bond
479
closes stem III (Pley et al., 1994; Scott et al., 1995).
This hydrogen bond, which is not a bifurcated one
as in tRNAs, is well maintained during an MD
simulation of the ribozyme (Hermann et al., 1998).
Also, and most interestingly, the base preceding
the U-turn in the hammerhead ribozyme is normally a C and it forms a wobble base-pair with the
cleavable C residue (C3 C17: see Figure 8(c)). This
base-pairs differ from the singly hydrogen-bonded
base-pair shown in Figure 5. Nevertheless, it seems
essential for the ribozyme as a substitution of this
C C pair by a G C pair results in a considerable
loss of activity (McKay, 1996).
Could the 32 38 base-pair participate in RNA/RNA
recognition in the ribosome?
As emphasized above, several biochemical studies have demonstrated that modi®cations of the
32 38 base-pair can alter both aminoacylation and
ribosome performance suggesting that this basepair is involved in direct or indirect recognition
features with their cognate synthetase and with the
ribosome (Lustig et al., 1993; Claesson et al., 1995;
Yarus & Smith, 1995; Giege et al., 1998; McClain
et al., 1998; O'Conner, 1998) or, as proposed by
Yarus and co-workers (Smith & Yarus, 1989; Yarus
& Smith, 1995), with a second tRNA inside the
ribosome. For RNA/RNA interactions within a
given organism, the Watson-Crick binding sites of
residue 38, located in the shallow groove of the
anticodon stem, are not blocked and could be
involved in the formation of a triple interaction
with a ribosomal residue (Figure 10). Thus, the
adenine base of C A and U A pairs and the cytosine base of U C and C C pairs could be involved
in Watson-Crick interactions with U or G residues,
respectively. Yet, other interaction types cannot be
excluded. For example, the formation of a trans
A38 U or C38 U pair could be proposed. This
interaction scheme presents the advantage of involving a uridine base in all four cases (Figure 10).
Further, a uridine base could also be placed in the
shallow groove of the U32 U38 pair. Interestingly,
GNRA tetraloops in which the Watson-Crick sites
of the guanine or the adenine residue of the closing
G A pair are available, are involved in RNA/RNA
recognition via the Watson-Crick sites of the A in
the shallow groove side of G C pairs in helices
(Michel & Westhof, 1990; Jaeger et al., 1994). Crystal structures have demonstrated that the G A pair
of a GNRA tetraloop can indeed form of a base triple with a helix (Pley et al., 1994).
However, RNA/RNA interactions could also
take place on the less accessible deep groove side,
especially at the border of a helix. Biochemical data
indicate that the substitution of a C A by an isosteric U A pair affects the discriminating ability of
tRNA transcripts (Lustig et al., 1993; Claesson et al.,
1995) and their frameshifting ef®ciency (O'Connor,
198). This seems only possible if recognition
phenomena involve the deep groove side which
presents different shapes. However, interaction
480
Singly and Bifurcated Hydrogen-bonded Base-pairs
Figure 10. Possible interaction of a ribosomal RNA residue with the shallow groove of the set of four isosteric
32 38 base-pairs in the largest Family. Left: Given the accessibility of the Watson-Crick sites of residue 38, formation
of Watson-Crick pairs involving either a U or a G residue can be proposed. Right: Base triples involving an uridine
residue in trans with respect to residue 38 are possible in all four cases.
patterns involving a third residue are less obvious.
Further, the deep groove side of the non-canonical
C A pair in yeast tRNAPhe ( C in yeast tRNAAsp)
has been described as forming an ion binding site
in tRNAs and in the leadzyme (Wedekind &
McKay, 1999). Several rare modi®cations were also
found to block the access to the deep groove side.
These modi®cations are m3C‡ at position 32 and
m5C at position 38. Note that m3C‡ has only been
detected at position 32 in tRNAs (Auf®nger &
Westhof, 1998b). In order to test the preceding
hypothesis, several biochemical assays can be proposed. For example, a methylation of the N1 site of
A38 or of the N3 site of C38 should block access to
the shallow groove, while a methylation of the N7
site of A38, the C5 site of C38, the N3 site of C32,
or the N3 site of U32 would at least partially hinder access to the deep groove. Interestingly, in the
recent crystal structure of a conserved ribosomal
RNA/protein domain, a U A pair adopts the conformation of the C A pair observed in yeast
tRNAPhe (Conn et al., 1999). Here, the hydroxyl
group of a distant adenine residue seems hydrogen
bonded to the uridine residue on the deep groove
Ê ). Although such a
side N3(U) . . . O20 (A) 2.6 A
hydrogen bond is not compatible with a methylation at the N3 site of the uridine base, it does
point to another potential binding site in the deep
groove. Furthermore, binding of ribosomal protein
residues cannot be excluded.
Conclusions
Biomolecular motifs with speci®c structure and
functions can be recognized by a set of sequence
characteristics constituting their molecular signatures. The knowledge of these signatures has many
implications such as (i) the ability to recognize
motifs with speci®c structure in sequence databases; and (ii) the possibility to derive a 3D model
from the sequence. Up to now, the signature of the
tRNA anticodon hairpins comprised the phylogenetically conserved U33 residue associated with a
conserved pyrimidine at position 32 and a conserved purine at position 37. A strong preference
for a purine at position 38 had also been noted
while an almost regular distribution of the four
bases is observed at the anticodon positions 35 to
36. At position 34 adenine bases are poorly represented (Grosjean et al., 1982; Auf®nger &
Westhof, 1998b). An analysis of available data led
to the re®nement of the signature of the tRNA
anticodon loop. tRNA anticodon loops are generally closed by a set of non-canonical isosteric basepairs involving a single inter-residue hydrogen
bond at position 32 38, which can be distributed in
three families. Family I (86 % of the sequences)
comprises the four C A, U A, U C, and CC pairs,
i.e. Y C/A, base-pairs (interestingly, the U A pair
is not of the Watson-Crick type). Family II (7 %)
gathers the set of U U base-pairs, not isosteric to
the preceding ones, also observed at position
32 38. Thus, sequences with U32 U38 pair may
comprise a tRNA family with distinct properties.
Family III (7 %) comprises a set of rare base oppositions for which only a limited number of approximately isosteric base-pairs could be tentatively
proposed. Besides the structural aspects, it is proposed that conserved patterns associated with the
set of the four isosteric 32 38 base-pairs of family I,
Singly and Bifurcated Hydrogen-bonded Base-pairs
could be involved in the formation of base triples
mediating tertiary interactions of the tRNA with a
ribosomal RNA residue.
Acknowledgements
The authors thank Stephen Cusack for providing coordinate sets for the tRNA anticodon hairpin of structures
pte002 and ptr014.
References
Arnez, J. G. & Steitz, T. A. (1994). Crystal structure of
unmodi®ed tRNAGln complexed with glutaminyltRNA synthetase and ATP suggests a possible role
for pseudo-uridines in stabilization of RNA structure. Biochemistry, 33, 7560-7567.
Arnez, J. G. & Steitz, T. A. (1996). Crystal structure of
three mysacylating mutants of Escherichia coli glutaminyl-tRNA synthetase complexed with tRNAGln
and ATP. Biochemistry, 35, 14725-14733.
Auf®nger, P. & Westhof, E. (1996). H-bond stability in
the tRNAAsp anticodon hairpin: 3 ns of multiple
molecular dynamics simulations. Biophys. J. 71, 940954.
Auf®nger, P. & Westhof, E. (1997). RNA hydration:
three nanoseconds of multiple molecular dynamics
simulations of the solvated tRNAAsp anticodon hairpin. J. Mol. Biol. 269, 326-341.
Auf®nger, P. & Westhof, E. (1998a). Effects of pseudouridylation on tRNA hydration and dynamics: a
theoretical approach. In Modi®cation and Editing of
RNA (Grosjean, H. & Benne, R., eds), pp. 103-112,
American Society for Microbiology, Washington,
DC.
Auf®nger, P. & Westhof, E. (1998b). Location and distribution of modi®ed nucleotides in tRNA. In Modi®cation and Editing of RNA (Grosjean, H. & Benne, R.,
eds),
pp.
569-576,
American
Society
for
Microbiology, Washington, DC.
Auf®nger, P. & Westhof, E. (1998c). RNA base pair
hydration. J. Biomol. Struct. Dynam. 16, 693-707.
Auf®nger, P., Louise-May, S. & Westhof, E. (1999). Molecular dynamics simulations of the solvated yeast
tRNAAsp. Biophys. J. 76, 50-64.
Ê crystal
Basavappa, R. & Sigler, P. B. (1991). The 3 A
structure of yeast initiator tRNA: functional implications in initiator/elongator discrimination. EMBO
J. 10, 3105-3111.
Beier, H., Lee, M. C., Sekiya, T., Kuchino, Y. &
Nishimura, S. (1992). Two nucleotides next to the
anticodon of cytoplasmic rat tRNA(Asp) are likely
generated by RNA editing. Nucl. Acids Res. 20,
2679-2683.
Berman, H. M., Olson, W. K., Beveridge, D. L.,
Westbrook, J., Gelbin, A., Demeny, T., Hsieh, S. H.
& Srinivasan, A. R. (1992). The nucleic acid database: a comprehensive relational database of threedimensional structures of nucleic acids. Biophys. J.
63, 751-759.
Biou, V., Yaremchuk, A., Tukalo, M. & Cusack, S.
Ê crystal structure of T. thermophilus
(1994). The 2.9 A
seryl-tRNA synthetase complexed with tRNASer.
Science, 263, 1404-1436.
Brown, R. S., Dewan, J. C. & Klug, A. (1985). Crystallographic and biochemical investigation of the lea-
481
d(II)-catalyzed hydrolisis of yeast phenylalanine
tRNA. Biochemistry, 24, 4785-4801.
Cate, J. H., Gooding, A. R., Podell, E., Zhou, K. H.,
Golden, B. L., Kundrot, C. E., Cech, T. R. &
Doudna, J. A. (1996). Crystal structure of a group I
ribozyme domain-Principles of RNA packing.
Science, 273, 1678-1685.
Cavarelli, J., Eriani, G., Rees, B., Ruff, M., Boeglin, M.,
Mitschler, A., Martin, F., Gangloff, J., Thierry, J.-C.
& Moras, D. (1994). The active site of yeast aspartyl-tRNA synthetase: structural and functional
aspects of the aminoacylation reaction. EMBO J. 13,
327-337.
Chartrand, P., Usman, N. & Cedergreen, R. (1997).
Effects of structural modi®cations on the activity of
the leadzyme. Biochemistry, 36, 3145-3150.
Claesson, C., Lustig, F., BoreÂn, T., Simonsson, C.,
Barciszewska, M. & Lagerkvist, U. (1995). Glycine
codon discrimination and the nucleotide in position
32 of the anticodon loop. J. Mol. Biol. 247, 191-196.
Comarmond, M. B., GiegeÂ, R., Thierry, J. C. & Moras, D.
(1986). Three-dimensional structure of yeast
tRNAAsp. I. Structure determination. Acta Crystallog.
sect. B, 42, 272-280.
Conn, G. L., Draper, D. E., Lattman, E. E. & Gittis, A. G.
(1999). Crystal structure of a conserved ribosomal
protein-RNA complex. Science, 284, 1171-1174.
Correll, C. C., Freeborn, B., Moore, P. B. & Steitz, T. A.
(1997). Metals, motifs and recognition in the crystal
structure of a 5S rRNA domain. Cell, 91, 705-712.
Correll, C. C., Munishkin, A., Chan, Y. L., Ren, Z.,
Wool, I. G. & Steitz, T. A. (1998). Crystal structure
of the ribosomal RNA domain essential for binding
elongation factors. Proc. Natl Acad. Sci. USA, 95,
13436-13441.
Cruse, W. B. T., Saludjian, P., Biala, E., Strazewski, P. &
PrangeÂ, T. (1994). Structure of a mispaired RNA
Ê resolution and implications
double helix at 1.6-A
for the prediction of RNA secondary structure. Proc.
Natl Acad. Sci. USA, 91, 4160-4164.
Cusack, S., Yaremchuk, A. & Tukalo, M. (1996a). The
crystal structure of T. thermophilus lysyl-tRNA
synthetase complexed with E. coli tRNALys and a
T. thermophilus tRNALys transcript: anticodon recognition and conformational changes upon binding of
a lysyl-adenylate analogue. EMBO J. 15, 6321-6334.
Cusack, S., Yaremchuk, A. & Tukalo, M. (1996b). The
crystal structure of the ternary complex of T. thermophilus seryl-tRNA synthetase with tRNASer and a
seryl-adenylate analogue reveals a conformational
switch in the active site. EMBO J. 15, 2834-2842.
Cusack, S., Yaremchuk, A., Krikliviy, I. & Tukalo, M.
(1998). tRNAPro anticodon recognition by Thermus
termophylus prolyl-tRNA synthetase. Structure, 6,
101-108.
Davis, D. R. & Poulter, C. D. (1991). 1H-15N NMR studies of escherichia coli tRNAPhe from hisT mutants:
a structural role for pseudouridine. Biochemistry, 30,
4223-4231.
Dirheimer, G., Keith, G., Dumas, P. & Westhof, E.
(1995). Primary, secondary, and tertiary structures
of tRNAs. In tRNA: Structure, Biosynthesis, and Function (SoÈll, D. & RajBhandary, U., eds), pp. 93-126,
American Society for Microbiology, Washington.
Durant, P. C. & Davis, D. R. (1999). Stabilization of the
anticodon stem-loop of tRNALys,3 by A‡ C basepair and by pseudouridine. J. Mol. Biol. 285, 115131.
482
GiegeÂ, R., Sissler, M. & Florentz, C. (1998). Universal
rules and idiosyncratic features in tRNA identity.
Nucl. Acids Res. 26, 5017-5035.
Grosjean, H. & Benne, R. (1998). Editors of Modi®cation
and Editing of RNA, American Society for Microbiology, Washington, DC.
Grosjean, H., Cedergreen, R. J. & McKay, W. (1982).
Structure in tRNA data. Biochimie, 64, 387-397.
Hermann, T., Auf®nger, P. & Westhof, E. (1998). Molecular dynamics investigations of the hammerhead
ribozyme RNA. Eur. J. Biophys. 27, 153-165.
Hingerty, B., Brown, R. S. & Jack, A. (1978). Further
re®nement of the structure of yeast tRNAPhe. J. Mol.
Biol. 124, 523-534.
Holbrook, S. R., Cheong, C., Tinoco, I. & Kim, S. H.
(1991). Crystal structure of an RNA double helix
incorporating a track of non-Watson-Crick base
pairs. Nature, 353, 579-581.
Hyde, E. I. & Reid, B. R. (1985a). Assignment of the
low-®eld 1H NMR spectrum of Escherichia coli
tRNAPhe using nuclear overhauser effects. Biochemistry, 24, 4307-4314.
Hyde, E. I. & Reid, B. R. (1985b). NMR studies of ion
binding to Escherichia coli tRNAPhe. Biochemistry, 24,
4315-4325.
Jaeger, L., Michel, F. & Westhof, E. (1994). Involvement
of a GNRA tetraloop in long-range RNA tertiary
interactions. J. Mol. Biol. 236, 1271-1276.
Jucker, F. M. & Pardi, A. (1995). GNRA tetraloops make
a U-turn. RNA, 1, 219-222.
Kolk, M. H., Heus, H. A. & Hilbers, C. W. (1997). The
structure of the isolated, central hairpin of the HDV
antigenomic ribozyme: novel structurl features and
similarity of the loop in the ribozyme and free in
solution. EMBO J. 16, 3685-3692.
Leontis, N. B. & Westhof, E. (1998a). The 5S rRNA loop
E: chemical probing and phylogenetic data versus
crystal structure. RNA, 4, 1134-1153.
Leontis, N. B. & Westhof, E. (1998b). A common motif
organizes the structure of multi-helix loops in 16S
and 23S ribosomal RNAs. J. Mol. Biol. 283, 571-583.
Leontis, N. B. & Westhof, E. (1999a). Conserved geometrical base pairing patterns in RNA. Quart. Rev.
Biophysics, in the press.
Leontis, N. B. & Westhof, E. (1999b). Recurrent RNA
motifs: analysis at the base pair level. In RNA Biochemistry and Biotechnology (Barciszewski, J. & Clark,
B. F. C., eds), Kluwer Academic Publishers, Boston.
Lustig, F., BoreÂn, T., Claesson, C., Simonsson, C.,
Barciszewska, M. & Lagerkvist, U. (1993). The
nucleotide at position 32 of the tRNA anticodon
loop determines ability of anticodon UCC to discriminate among glycine codons. Proc. Natl Acad.
Sci. USA, 90, 3343-3347.
Lynch, S. R. & Tinoco, I. (1998). The structure of the L3
loop from the hepatitis delta virus ribozyme: a syn
cytidine. Nucl. Acids Res. 26, 980-987.
McClain, W. H., Schneider, J., Bhattacharya, S. &
Gabriel, K. (1998). The importance of tRNA backbone-mediated interactions with synthetase for aminoacylation. Proc. Natl Acad. Sci. USA, 95, 460-465.
McKay, D. B. (1996). Structure and function of the hammerhead ribozyme: an un®nished story. RNA, 2,
395-403.
Michel, F. & Westhof, E. (1990). Modelling of the threedimensional architecture of group I catalytic introns
based on comparative sequence analysis. J. Mol.
Biol. 216, 585-610.
Singly and Bifurcated Hydrogen-bonded Base-pairs
Nissen, P., Kjeldgaard, M., Thirup, S., Polekhina, G.,
Reshetnikova, L., Clark, B. F. C. & Nyborg, J.
(1995). Crystal structure of the ternary complex of
Phe-tRNAPhe, EF-Tu and a GTP analog. Science, 270,
1464-1472.
Nissen, P., Thirup, S., Kjeldgaard, M. & Nyborg, J.
(1999). The crystal structure of Cys-tRNACys-EF-TiGDPNP reveals general and speci®c features in the
ternary complex and in tRNA. Structure, 7, 143-154.
O'Connor, M. (1998). tRNA imbalance promotes-1
frameshifting via near-cognate decoding. J. Mol.
Biol. 279, 727-736.
Oubridge, C., Nobutoshi, I., Evans, R. P., Teo, C. H. &
Ê resolNagai, K. (1994). Crystal structure at 1.92 A
ution of the RNA-binding domain of the U1A spliceosomal protein complexed with an RNA hairpin.
Nature, 372, 432-438.
Perbandt, M., Nolte, A., Lorentz, S., Erdmann, V. A. &
Betzel, C. (1998). Crystal structure of domain E of
Thermus ¯avus rRNA: a helical RNA structure
including a hairpin loop. FEBS Letters, 429, 211-215.
Perona, J. J., Rould, M. A. & Steitz, T. A. (1993). Structural basis for transfer RNA aminoacylation by
Escherichia coli glutaminyl-tRNA synthetase. Biochemistry, 32, 8758-8771.
Pley, H. M., Flaherty, K. M. & McKay, D. B. (1994).
Three-dimensional structure of a hammerhead ribozyme. Nature, 372, 68-74.
Price, D. H. & Gay, M. W. (1998). Editing of tRNA. In
Modi®cation and Editing of RNA (Grosjean, H. &
Benne, R., eds), pp. 103-112, American Society for
Microbiology, Washington, DC.
Price, S. R., Evans, P. R. & Nagai, K. (1998). Crystal
structure of the spliceosomal U2B00 -U2A0 protein
complex bound to a fragment of U2 smal nuclear
RNA. Nature, 394, 645-650.
Quigley, G. J. & Rich, A. (1976). Structural domains of
transfer RNA molecules. Science, 194, 796-806.
Quigley, G. J., Teeter, M. M. & Rich, A. (1978). Structural analysis of spermine and magnesium ion binding to yeast phenylalanine transfer RNA. Proc. Natl
Acad. Sci. USA, 75, 64-68.
Rath, V. L., Silvian, L. F., Beijer, B., Sproat, B. S. &
Steitz, T. A. (1998). How glutaminyl-tRNA synthetase selects glutamine. Structure, 6, 439-449.
Rould, M. A., Perona, J. J., SoÈll, D. & Steitz, T. A. (1989).
Structure of E. coli glutaminyl-tRNA synthetase
Ê resolcomplexed with tRNA(Gln) and ATP at 2.8 A
ution. Science, 246, 1135-1142.
Rould, M. A., Perona, J. J. & Steitz, T. A. (1991). Structural basis of anticodon loop recognition by glutaminyl-tRNA synthetase. Nature, 352, 213-218.
Ruff, M., Krishnaswamy, S., Boeglin, M., Poterszman,
A., Mitschler, A., Podjarny, A., Rees, B., Thierry,
J. C. & Moras, D. (1991). Class I aminoacyl tRNA
synthetases: crystal structure of yeast aspartyltRNA synthetase complexed with tRNAAsp. Science,
25, 1682-1689.
Schweisguth, D. C. & Moore, P. B. (1997). On the conformation of the anticodon loops of initiator and
elongator methionine tRNAs. J. Mol. Biol. 267, 505519.
Scott, W. G., Finch, J. T. & Klug, A. (1995). The crystal
structure of an all-RNA hammerhead ribozyme: a
proposed mechanism for RNA catalytic cleavage.
Cell, 81, 991-1002.
Smith, D. & Yarus, M. (1989). tRNA-tRNA interactions
within cellular ribosomes. Proc. Natl Acad. Sci. USA,
86, 4397-4401.
Singly and Bifurcated Hydrogen-bonded Base-pairs
Sprinzl, M., Horn, C., Brown, M., Loudovitch, A. &
Steinberg, S. (1998). Compilation of tRNA sequences
and sequences of tRNA genes. Nucl. Acids Res. 26,
148-153.
Sussman, J. L., Holbrook, S. R., Warrant, R. W., Church,
G. M. & Kim, S.-H. (1978). Crystal structure of
yeast phanylalanine transfer RNA. I. Crystallographic re®nement. J. Mol. Biol. 123, 607-630.
Tanaka, Y., Fujii, S., Hiroaki, H., Sakata, T., Tanaka, T.,
Uesugi, S., Tomita, K. & Kyogoku, Y. (1999). A0 form RNA double helix in the single crystal structure of r(UGAGCUUCGGCUC). Nucl. Acids Res. 27,
949-955.
Tinoco, I. (1993). Structure of base pairs involving at
least two hydrogen bonds. In The RNA World
(Gesteland, R. F. & Atkins, J. F., eds), pp. 603-607,
Cold Spring Harbor Laboratory Press, Cold Spring
Harbor, NY.
Wedekind, J. E. & McKay, D. B. (1999). Crystal structure
of a lead-dependent ribozyme revealing metal binding sites relevant to caalysis. Nature Struct. Biol. 6,
261-268.
Westhof, E. & Sundaralingam, M. (1986). Restrained
re®nement of the monoclinic form of yeast phenylalanine transfer RNA. Temperature factors and dyamics, coordinated waters, and base-pair propeller
twist angles. Biochemistry, 25, 4868-4878.
483
Westhof, E., Dumas, P. & Moras, D. (1985). Crystallographic re®nement of yeast aspartic acid transfer
RNA. J. Mol. Biol. 184, 119-145.
Westhof, E., Dumas, P. & Moras, D. (1988). Restrained
re®nement of two crystalline forms of yeast aspartic
acid and phenylalanine transfer RNA crystals. Acta
Crystallog. sect. A, 44, 112-123.
Westhof, E., Romby, P., Romaniuk, P., Ebel, J.-P.,
Ehresmann, C. & Ehresmann, B. (1989). Computer
modeling from solution data of spinach chloroplast
and of Xenopus laevis somatic and oocyte 5S rRNAs.
J. Mol. Biol. 207, 417-431.
Yarus, M. (1982). Translational ef®ciency of transfer
RNA's: uses of an extended anticodon. Science, 218,
646-652.
Yarus, M. & Smith, D. (1995). tRNA on the ribosome: a
waggle theory. In tRNA: Structure, Biosynthesis, and
Function (SoÈll, D. & RajBhandary, U., eds), pp. 443469,
American
Society
for
Microbiology,
Washington, DC.
Yarus, M., Cline, S., Raftery, L., Wier, P. & Bradley, D.
(1986a). The translational ef®ciency of tRNA is a
property of the anticodon arm. J. Biol. Chem. 261,
10496-10505.
Yarus, M., Cline, S. W., Wier, P., Breeden, L. &
Thompson, R. C. (1986b). Actions of the anticodon
arm in translation on the phenotypes of RNA
mutants. J. Mol. Biol. 192, 235-255.
Edited by D. E. Draper
(Received 7 June 1999; received in revised form 26 July 1999; accepted 30 July 1999)