tRNA Structure
Secondary article
Eric Westhof, Institute of Molecular and Cellular Biology, Strasbourg, France
Pascal Auffinger, Institute of Molecular and Cellular Biology, Strasbourg, France
Article Contents
. Introduction
. Cloverleaf Structure
Transfer ribonucleic acid (tRNA) molecules that participate in the elongation step of
protein synthesis on the ribosome have a conserved secondary structure, known as the
cloverleaf, and fold into a common three-dimensional architecture.
. Universally Invariant and Semi-invariant Bases
Introduction
. Conservations of Bases and Loop Conformations
Single-stranded RNA (ribonucleic acid) molecules are
ubiquitous in cells and play many biological roles.
However, three main types of RNA dominate molecular
biology: the messenger RNAs (mRNAs), the transfer
RNAs (tRNAs) and the ribosomal RNAs (rRNAs). They
all participate in the biosynthesis of proteins. The mRNAs
contain the sequence of nucleotide triplets or codons which
will be read by tRNAs during the translation process on the
ribosomes of which the rRNAs constitute central components. In mRNAs, the sequence alone is the primary
determinant of biological function. However, for tRNAs
and rRNAs, the folding in space of the polynucleotide
RNA chain in a native tertiary structure, forms the basis of
biological activity.
The existence of some type of ‘adaptor’ RNA molecules
acting as go-betweens between the RNA world and the
protein universe was predicted by Francis Crick as early as
1955, long before transfer RNA molecules were characterized biochemically. Because there are 61 different codons in
the genetic code, one would expect about the same number
of different tRNAs, distributed among the 20 different
isoacceptor families corresponding to the 20 different
amino acids. However, if the wobble hypothesis of Crick
(which relaxes the Watson–Crick complementarity and
considers not only G.U but also I.U, I.C, or I.A pairings
involving inosine (I), an oxidized form of adenosine) is
followed, only 31 different tRNAs are necessary to decode
all codons (32 with the additional initiator tRNA) (Crick,
1966). The recent sequencing of whole eubacterial genomes
has revealed that the number of tRNA genes varies
between 33 (in Mycoplasma genitalium, a parasitic
bacterium with a minimal genome) and 88 (in the Grampositive bacterium Bacillus subtilis) with most being
between 44 and 46. It should be added that the copy
number of tRNA genes can be as high as four and that
some tRNA variants present nucleotide changes outside
the anticodon. Thus, in Escherichia coli the 84 tRNA genes
code 45 tRNA species with only 41 tRNAs having different
anticodons.
As with every biological macromolecule, the molecular
evolution of tRNAs is intimately coupled to the structural
constraints imposed by the nature of the polymer and its
. Summary
. The Two Domains of the Three-dimensional
Architecture
. Unusual Hydrogen Bonds Maintain the Tertiary
Structure
. Unusual tRNAs
functions. Each tRNA molecule has to evolve under two
opposing constraints. On the one hand it needs a threedimensional architecture that allows it to fit precisely in the
ribosome-binding sites for promoting protein synthesis,
but on the other hand it needs to contain enough molecular
diversity to guarantee specific recognition with a unique
cognate tRNA synthetase. Recall that tRNA synthetase
aminoacetylates the tRNA 3’-end adenosine with the
amino acid specific to that tRNA. Indeed, a tRNA carries
an anticodon triplet complementary to a given codon and
should consequently be charged by the synthetase solely
with the amino acid specified by the anticodon–codon
interactions. The signatures of those opposing biological
constraints are apparent in the network of invariant
residues and interactions that maintain a common
architecture on which enough molecular diversity can be
coded for specific recognition with the tRNA synthetases.
These structural constraints are clearly seen in the
invariance in the length of some helices of the secondary
structure and in the constant presence of some residues at
definite positions. However, tRNAs constitute a set of
molecules with which biological evolution has tinkered
enormously. Although plant chloroplast and some mitochondrial tRNAs present high sequence homologies to
eubacterial tRNAs, other mitochondrial tRNAs display a
palette of unusual structural features. tRNAs responsible
for the cotranslational insertion of the 21st amino acid,
selenocysteine, also contain odd features.
Nature has further exploited the sequence and structural
diversity compatible with tRNA folding to adapt tRNA
structure to functions unrelated to protein synthesis. For
example, a special tRNA charged with a glutamic acid is
used during chlorophyll synthesis in plants, while another
tRNA charged with a glycine is necessary for peptidoglycan synthesis in bacteria. Retroviruses, like VIH, require a
tRNA (in human cells, a tRNALys) as a primer for the
replication of their genomic RNA. Several plant viruses
contain at their 3’ ends, a tRNA-like structure, the integrity
of which is necessary for viral replication.
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
1
tRNA Structure
Cloverleaf Structure
The secondary structure of an RNA molecule is generally
defined by the set of contiguous Watson–Crick pairs
(including the G.U wobble pairs) forming helices between
segments of the single-stranded RNA. For a given family
of RNA sequences, the ensemble of the so-defined helices is
conserved as well as their relative arrangement. This is why
secondary structures can be established by sequence
comparisons. The basic hypothesis is, indeed, that partly
divergent, but nevertheless functionally and historically
related, molecules from various biological origins fold into
similar secondary and tertiary structures. Helices can occur
at homologous positions in the secondary structures of two
different sequences if compensatory changes occur on both
strands so as to maintain complementarity between the
bases in the Watson–Crick sense. For example, when
Holley sequenced the first tRNA in 1965, he suggested
several possible secondary structures (Holley et al., 1965).
However, when a second tRNA was sequenced, the only
secondary structure common to both sequences was the
one in the form of a planar cloverleaf (Figure 1a). All known
cytoplasmic elongator tRNA sequences (of lengths varying
between 74 and 98 nucleotides) can be folded in the
cloverleaf structure. The cloverleaf structure consists of
four helices and three hairpin loops. The four helices are
called: the acceptor (AA) helix (or stem) because it will
carry the amino acid once charged by the cognate
aminoacyl synthetase; the dihydrouridine (D) hairpin
because it often contains the modified base dihydrouracil;
the anticodon (AC) hairpin because its apical loop presents
the anticodon triplet; and the thymine (T) hairpin because
its loop often contains the thymine base unusual in RNAs.
Universally Invariant and Semiinvariant Bases
Alignments of tRNA sequences, equivalent to inserting
tRNA sequences on the cloverleaf template, reveal
positions that are always occupied by a single type of base
(‘invariant’) and positions that are occupied either only by
purines or by pyrimidines (‘semi-invariant’). About 22
nucleotides belong to these categories. As is usual with
biological sequences, those conservations are never absolute and, for clarity, we will focus the following description
on cytoplasmic elongator tRNAs, for which the numbering
of yeast tRNAPhe constitutes the reference system, since it
was the first tRNA structure solved by X-ray crystallography. Figure 1b shows the distribution of the four
nucleotides found in the 932 cytoplasmic tRNA genes
listed in the 1998 release of the tRNA database (Sprinzl
et al., 1998). In the following, R stands for a purine (A or
G), Y for a pyrimidine (U or C) and N for any of the four
nucleotides; standard Watson–Crick secondary base pairs
2
will be separated by a dash (e.g. U12–A23); non-Watson–
Crick pairs by a dot (e.g. G10.U25 or the trans U8.A14);
and the third base of triples will be separated by three dots
from the interacting secondary base pair on the side of the
contacting base (e.g. G45...G10.U25).
tRNA synthetases attach the cognate amino acid (i.e. the
one dictated by the codon–anticodon pairings) on either
the 3’-OH or the 2’-OH group of the terminal adenosine.
The first residue of all tRNAs, which starts the first base
pair of the acceptor helix, carries a 5’-phosphate group. It
originates in the biosynthesis of tRNAs. After being
transcribed from tRNA genes, the tRNA transcripts are
maturated at the 5’-end by the ribonucleoprotein enzyme
RNAaseP, which clips off the 5’-end segment of the tRNA
gene. The catalysis is effected by the RNA moiety of the
RNAaseP, a process which leaves a 3’ OH and a 5’
phosphate. The maturation of tRNAs is very complex and
requires a great number of enzymes since tRNAs are
characterized by the presence of several modified nucleotides.
The lengths of the helices and loops are either conserved
or vary between defined limits. The distribution of the four
Watson–Crick pairs within helices is not uniform. For
example, the acceptor helix starts in three-quarters of
sequences with a G1–C72 pair and ends very rarely with a
C7–G66 base pair. The D helix also frequently starts with a
R10–Y25 followed by a Y11–R24 base pair and ends in
half of the sequences with a C13–G22 pair. The conservation is very strong for the last base pair of the thymine helix
which is almost always G53–C61. Only the central base
pairs of the acceptor helix as well as the very central one in
the anticodon helix present an almost equal distribution of
the four Watson–Crick combinations. Interestingly, those
two sets of base pairs constitute determinants for recognition by aminoacyl synthetases.
The acceptor helix has seven base pairs, except in
tRNAHis where an eighth residue at the 5’ end is added
posttranscriptionally. The acceptor helix possesses a 3’
dangling strand with four nucleotides, the sequence of
which is -RCCA-3’OH (R is most often A and half as
frequently G; that position is unfrequently occupied by a
pyrimidine). Position 73 is called ‘discriminator’ owing to
its role in synthetase selection of tRNAs. In RNA helices,
interstrand stacking occurs in 5’-YR-3’ steps. The last base
pair of the acceptor helix, often a R1–Y72 base pair, is thus
probably stabilized by interstrand stacking with R73. The
first four base pairs of the acceptor helix are important
discriminant elements for synthetase recognition.
The dihydrouridine helix generally possesses four base
pairs, while its hairpin loop is more variable with 7–11
residues. The variation occurs in two regions, called a and
b, situated on either side of two invariant guanine residues,
G18 and G19 (Figure 1a). The dihydrouridine residues
occupy either or both of those positions. In the loop, the
first (14) and last (21) residues are always a purine, mainly
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
tRNA Structure
5’ P
0
Dihydrouridine hairpin (D hairpin)
1
72
2
71
3
70
4
69
5
68
6
67
7
66
8
α
17
16 15
73 C
Acceptor stem (AA helix)
Thymine hairpin (T hairpin)
60 59
65 64 63 62 61
58
49 50 51 52 53
56
57
9
14
13 12 11 10
17a
54 55
48
18
22 23 24 25
19
20 20a 20b
A 3’ OH
C
44
26
21
β
Anticodon hairpin
(AC hairpin)
27
43
28
42
29
41
30
40
31
39
e21
Variable loop
e22
e23
e24
e11
e25
e12
e26
e27
e13
e5
e14
e15
e4
e16
e17
e3
e1 e2
37
33
(a)
46
38
32
34
45
47
35
36
Anticodon triplet
5’ P
3’ OH
75
70
60
65
16 15
17
10
17a
50
55
18
47
25
19
20
21
46
e21
45
20a 20b
e27
e11
Single strands
A G
e5
30
40
T C
e17
Helix
A G
T C
T C
A G
34
35
e1
36
(b)
Figure 1 Nomenclature and base distributions in elongator transfer RNAs. (a) The accepted nomenclature of transfer RNA molecules (Sprinzl et al., 1998).
Variable positions are present in the dihydrouridine and variable loops. The variable loop itself forms a hairpin when long enough. Residue 0 occurs only in
histidinyl-tRNAs. Straight lines indicate secondary base pairing and broken lines unusual base pairings at the beginning or end of a helix. (b) The distribution
of the four common bases at corresponding positions along the sequence in 932 sequences of elongator tRNA genes (Auffinger and Westhof, 1998). In
single strands, the adenine region always starts at 2 908 from the vertical. For the variable positions which are not always occupied, the proportion of
sequences where they are occupied can be evaluated starting from the outer ring. Thus, position 17 is present in less than half of the sequences and
positions 45 and 46 in more than half of the sequences. In helices, the colour codes for paired residues are arranged so as to follow Watson–Crick pairings;
the 5’ strand has a thin outer circle and the 3’ strand a thick outer circle.
an adenine. The first three base pairs of the D helix
constitute identity elements for aminoacyl synthetases.
The anticodon hairpin has five base pairs in the helix and
seven residues in the loop. The second residue of the
anticodon loop is always a uridine (U33); it precedes the
three bases of the anticodon, the first position of which is
called the wobble base (34). Following the anticodon
triplet, a highly modified nucleotide is generally present
(position 37). Except for the first two residues (32 and 33),
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
3
tRNA Structure
the residues of the anticodon loop are important identity
elements for synthetases.
After the anticodon helix, there is a variable region
which may vary between 4 and 21 residues. According to
the size of the variable loop, tRNAs belong to class I or II.
Class I tRNAs have 4–5 nucleotides in the variable loop
and class II tRNAs from 10 to 24. In class II tRNAs, the
variable loop is long enough to form a fifth helix. Only
tRNAs specific for the amino acids leucine and serine (with
tyrosine in eubacteria and organelles only) belong to class
II.
Finally, the thymine hairpin has, like the anticodon
hairpin, five base pairs and seven residues in the loop.
However, the base conservation in the thymine loop is very
different from that in the anticodon loop. The last base pair
of the thymine helix, G53–C61, is highly conserved. In the
thymine loop, the first three residues are rather well
conserved and modified to -TCC- where C stands for
pseudouridine, a modified base in which the link between
the base and the sugar is C(1’)-C5 instead of C(1’)-N1. A
pseudouridine is generally present at the second position of
the loop in tRNAs but, since pseudouracils represent more
than 50% of the modified bases in tRNAs, many more
modified positions exist in the cloverleaf (Auffinger and
Westhof, 1998).
The Two Domains of the Threedimensional Architecture
The folding in three dimensions of the cloverleaf secondary
structure posed a tantalizing problem for many years. In a
remarkable paper of 1968, Levitt proposed a threedimensional model of tRNA (Levitt, 1969). Although the
overall architecture of the model turned out to be wrong,
that article laid the foundations for our thinking about the
relations between biological evolution and biomolecular
structure and function. Since the first correct tracing of a
tRNA chain in crystals of yeast tRNAPhe in 1974 (Suddath
et al., 1974), only one other crystal structure of a free tRNA
has been solved (yeast tRNAAsp in 1980) (Moras et al.,
1980). In addition, several structures of tRNAs complexed
with their cognate aminoacyl synthetases now exist. An
examination of the three-dimensional folding revealed by
the crystal structures allows a structural basis for all of the
base conservations observed to be suggested.
The four helices of the cloverleaf stack on each other
coaxially and two by two, forming two main arms or
domains (Figure 2a). Thus, the acceptor helix and the
thymine hairpin form the acceptor arm and the dihydrouridine together with the anticodon hairpins form the
anticodon arm. The two coaxial and contiguous stacks
make an angle of about 908 between them, giving to the
overall architecture of tRNAs the appearance of the letter
L or the Greek G. At the two ends, which are about 7.5–
4
8.0 nm apart, are the anticodon triplet and the -CCA 3’terminus, extremities which, when the tRNA is bound to
the acceptor site of the ribosome (A site), should,
respectively, bind to the codon triplet of the mRNA and
bring the attached amino acid into the peptidyl reaction
centre for reacting with the peptide chain carried by the
preceding tRNA present in the peptidyl site (P site). The
interfaces between the two coaxial helices are different in
each domain. The tRNA chain is continuous between the 3’
end of the thymine helix and the beginning of the acceptor
helix, while the last residue of the 5’ strand of the acceptor
helix adopts the C2’-endo sugar pucker which facilitates
branching off the helix. By contrast, at the interface
between the dihydrouridine and the anticodon helices,
there is usually a non-Watson–Crick base pair linking
residues 26 and 44 (e.g. an imino G.A pair) with a twist
angle with the last base pair of the dihydrouridine stem of
around 458, much larger than the standard 338 present
between base pairs of RNA helices.
The L-shaped architecture is locked in place by two main
structural features. First, the single-stranded regions
linking the two domains (residues 8–9 between the
acceptor and the dihydrouridine helices, as well as the
variable region between the anticodon and the thymine
helices) adopt conformations such that their residues form
base triples in the deep groove of the dihydrouridine
(Figure 2b). Secondly, those base triples position the D and
T loops so that extremely precise tertiary interactions can
occur between them. Fluorescence and ultraviolet melting
experiments have shown that formation of the base triples
in the deep groove of the D helix is the rate-limiting step in
the tertiary folding of tRNAs.
The stereochemistry of the nucleotides is rather uniform
in helices (ribose puckers in the C3’-endo domain, bases in
anti orientation with respect to the sugar, helical phosphate
torsion angles) and the folding of the chain is accomplished
in single-stranded segments by altering the ribose pucker to
the C2’-endo domain or by rotating the torsion angles at
the phosphorus atoms away from the helical conformations (Sundaralingam, 1973).
Unusual Hydrogen Bonds Maintain the
Tertiary Structure
Residues not participating in helices defining the secondary
structure, i.e. residues of loops and single-stranded regions,
form tertiary interactions either by interacting between
themselves or by interacting with base pairs within a helix
(leading to base triples). The four natural bases can interact
via hydrogen bonding in many different ways. There are 27
pairs with two standard hydrogen bonds. The recent
crystal structures of other RNAs expand further the
recognition modes, since they reveal pairings with only
one interbase hydrogen bond or with a water molecule
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
tRNA Structure
Figure 2 Two-dimensional and three-dimensional representation of the tertiary structure of elongator transfer RNAs. (a) Two-dimensional representation
of the tertiary structure of tRNAs, as proposed by Kim (1978), which emphasizes the two main domains and the tertiary contacts linking them. Only the
secondary structure can be represented in a plane without crossing lines (in other words, mathematically, a secondary structure is equivalent to a planar
graph). The representation of a three-dimensional structure and of the underlying tertiary contacts can be drawn in a plane, but with several line crossings.
Such schematic drawings are, however, useful for quick assessment and comparisons of tertiary contacts. The Kim representation shows clearly the two
domains, the contacts between the T and D loops and the tertiary base pairs and triples between the single-stranded segments and the D hairpin. The
contacts represented correspond to those of yeast tRNAAsp (Westhof et al., 1985). (b) Stereoview of a schematized representation of the tertiary structure of
yeast tRNAAsp. The sugar–phosphate backbone is drawn as a ribbon and the base pairs as rods. The colour code is the same as in Figure 2a. Notice the
characteristic deep and shallow grooves of an RNA helix in the acceptor and thymine helices, respectively.
bridging the two bases. The fundamental property of the
Watson–Crick pairs (the isosteric geometry of all four
possible pairs G–C, C–G, A–U, U–A involving complementary bases) is thus even more remarkable. In order to
discuss the tertiary interactions in tRNAs, it is necessary to
explain two terms. First, the nucleotides in a pair may
approach with their sugars on the same sides of the H
bonds or on either side of them. Following the chemical
literature, in the first case, the pairing is said to be cis and, in
the second, trans (often called ambiguously ‘reverse’).
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
5
tRNA Structure
Secondly, purine bases can interact via their edge carrying
the imidazole ring (involving ring nitrogen N7) instead of
only their pyrimidine edge, as in Watson–Crick pairings.
Such pairings are called Hoogsteen pairs, after their
discovery by Hoogsteen in 1956 in cocrystals of derivatives
of U and A. Figure 3a shows examples of a variety of
pairings. Thus, elongator tRNAs contain two Hoogsteen
U.A pairs, both with a trans orientation of the sugars. The
first trans Hoogsteen pair occurs between the invariant U
at position 8, the first residue after the acceptor helix, and
the invariant A at position 14, the first residue of the D
loop. The U8.A14 pair, thus, links the two main domains
of tRNAs. The second trans Hoogsteen pair (T54.A58)
occurs between two invariant residues within the T loop,
the T at position 54 and the A at position 58. There is also a
third trans base pair, between the semi-invariant residues
R15 and Y48, but involving the Watson–Crick sites. The
trans Watson–Crick R15.Y48 base pair, sometimes called
after Levitt who first noticed the covariation, stacks with
the trans Hoogsteen U8.A14 pair. They are both central
for the maintenance of the L-shaped structure of tRNAs
(Figure 3a).
Class I tRNAs contain three base triples, all in the deep
groove of the D helix. Of the four base pairs of the D helix,
only the 11–24 pair does not contact a third base. Both
single-stranded regions linking the two domains participate in the network of triple interactions. Residue 9
interacts with base 23 of base pair 12–23 (12–23...9), while
residues 45 and 46 of the variable loop interact, respectively, with base pair 10–25 (45...10–25) and base pair 13–
22 (13–22...46). The type of the hydrogen bonding sites
(Watson–Crick or Hoogsteen) is related to the local
orientation of the chains. Thus, cis Watson–Crick pairs
(as in usual double helices) lead to antiparallel strands
while, with equivalent stereochemistry of the nucleotides,
trans Watson–Crick pairs lead to a local parallel orientation of the strands (e.g. R15.Y48). The reverse is true for
Hoogsteen pairs between a pyrimidine and a purine (e.g.
U8.A14) and for Hoogsteen pairs involving the Watson–
Crick of one purine with the Hoogsteen sites of another
purine (e.g. G22...A46 in yeast tRNAAsp). However,
purine–purine pairs involving only the Hoogsteen sites
are locally antiparallel when cis and parallel when trans
(e.g. A9...A23 in yeast tRNAPhe). The local orientations of
the chains cannot be altered without altering profoundly
the overall topology or flipping some bases in the syn
orientation with respect to the sugar (a conformation
extremely rare in crystal structures of RNA fragments).
These structural constraints impose definite covariations
between residues involved in triple interactions.
The set of triples in the deep groove of the D helix is such
that residue 9 interdigitates between residues 45 and 46 of
the variable link and the contacts are made alternatively to
the 5’ and 3’ strands of the D helix. Again, this topology
dictates the relative strand orientations: residue R45
interacts with G10 and is parallel to it, but R46 is
antiparallel to residue R22 to which it binds, while the
inbetween residue R9 is parallel to its binding residue 23.
Because the D and anticodon helices stack coaxially with a
right-handed twist at the interface, the 3’-dangling variable
region faces the deep groove and is antiparallel to the 3’
strand of the D helix, which leads to the 5’ strand of the
anticodon helix. On the other hand, the junctions between
the two domains occur at the internucleotide linkages of
residues 7–8 and 48 so that residues 7 and 49 could almost
be linked as in a continuous helix. In short, the two singlestranded segments linking the two domains run antiparallel to each other, facing the deep groove of the D helix.
Figures 2 and 3 correspond to the molecular structure of
the yeast tRNAAsp as determined by X-ray crystallography
(Westhof et al., 1985). Figure 4 illustrates the sequence
variability in 33 available tRNAAsp genes from other
organisms. Invariant residues are clearly seen, e.g. U8,
A14, and G18 or G10. Covariations typical of secondary
structure are displayed by residues 11 and 24 or 12 and 23,
the two central base pairs of the D helix. The flanking base
pairs present biases, G10–Y25 and Y13–R22, with some
unusual oppositions (for example Y13.Y22). Correlations
between the third base and the secondary pair forming
triples are not easy to detect. For example (not apparent on
Figure 4), any of the four bases is found with the very
frequent G10–C25, although a G at position 45 is by far the
most frequent situation. In contrast, position 9, either A or
G, covaries with base pair 12–23 so that A9...U12–A23
and G9...G12–C23 are most frequently observed,
although either A9 (and almost G9) can be found together
with any of the four Watson–Crick base pairs (see Table 1).
Even the residues involved in triples appear to vary. For
example, in the structure of E. coli tRNAGln complexed
with its cognate synthetase (Rould et al., 1989), it is residue
45 which interacts with the 13–22 base pair, and not residue
46 which is pointing on the exterior of the molecule. In that
structure, the 13–22 base pair is an unusual A.A pair (see
below). Such adaptability blurs the covariation tables
(Klug et al., 1974; Gautheret et al., 1995).
Several isolated additional H bonds contribute to the
integrity of the tertiary structure. They involve the O2’
Figure 3 Atomic representations of the tertiary contacts in elongator transfer RNAs. (a) Tertiary interactions between bases as observed in yeast tRNAAsp.
From top to bottom: the two trans Hoogsteen pairs T54.A58 and U8.A14; the trans Watson-Crick pair A15.U48; the cis Watson–Crick G.A pair (also called
imino G.A pair); the standard cis Watson–Crick G19–C56 base pair; the two unusual bifurcated pairs G18.C55 and C32.C38. (b) The tertiary triple
contacts present in yeast tRNAAsp. The colour code is the same as in Figure 2a. In green, the G45...G10.U25 triple in which the amino N2 group of G45 Hbonds to the Hoogsteen sites of G10 (N7 and O6). In orange, the U12–A23...A9 triple which includes a trans symmetric Hoogsteen A.A pair. In red, the
C13.G22...A46 triple. Notice how, in G10.U25 and C13.G22, the pyrimidine base protrudes into the deep groove.
6
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
tRNA Structure
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
7
tRNA Structure
D stem
D loop
1 1 1 1 1 1 1 1 1
7 8 9 0 1 2 3 4 5 6 7 7
Organisme
A
Methanococcus Jan. … G U G G U G U A G C C C
…G U G G U G U A G C U C
Methanococ. Vani.
Methanotherm. Fer. … G U G G U G U A G U - Methanococ. Voltae … G U G G U G U A G C U C
Mycoplasma Capric. … A U A G C G A A G U U …C A U GGU G U A G U Mycoplasma Gen.
Mycoplasma Mycoid. … A U A G C G A A C G U U
Mycoplasma Pneumo.… A U G G U G U A G U - Acholeplasma Laid. … G U G G U G U A G G - …G U A G U G A A G U U Spiroplasma Melif.
…G U G G A G C A G U U U
Streptomyces Liv.
Staphlylococ. Aure. … G U A G U G U A G C - Staphlylococ. Aure. … G U A G U G U A G C - …U U G G A G C A G U - Lactobac. Bulg.
…G U A G U U C A G U U Bacillus Subtilis
…G U A G U G U A G U - Bacillus Sp. PS3
…G U A G U U C A G U C E. Coli
…G U A G U U C A G C U Haemophilus Influ.
…G U A G U U C A G C U Haemophilus Influ.
…G U A G U U C A G C U Haemophilus Influ.
…G U A G U U C A A U U Synechocystis Sp.
…U U A GU A U A G U - Phytophthora Par.
Saccharomyces Cer. … A U A G U U U A A U - Saccharomyces Cer. … A U A G U U U A A U - Schizosaccha. Pom. … U U A G U A U A G G - …G U A G U A U A G U - Glycine Max.
Caenorhabdi. Eleg. … G U A G U A U A G U - Drosophila Melano. … A U A G U A U A G U - …U U A GU A U A G U - Chicken
…U U A GU A U A G U - Chicken
…U U A GU A U A G U - Mouse
…U U A GU A U A G U - Rat
…U U A GU A U A G U - Rat
…N U R G U R Y A R Y - Consensus
1 1 2 2 2
8 9 0 0 0
A B
GG C C U
GG C C U
GG C U GG C C U
GG U U GA U - GG U U GG U U GG U U GG U U GG A G GG U U GG U U GG U C U
GG U U GG U U GG U U GG U U GG U U GG U U GG U U GG U U GG U C GG U - GG U - GG U A GG U G GG U U GG U G GG U G GG U G GG U G GG U G GG Y N -
D stem
Var. loop
2 2 2 2 2 2 2
1 2 3 4 5 6 7
A
A
A
A
A
A
A
A
A
A
U
A
A
A
A
A
A
A
A
A
A
A
A
C
A
A
A
A
A
A
A
A
A
A
U
U
U
U
U
A
U
A
A
U
G
A
A
U
G
A
G
G
G
G
G
G
G
A
G
G
G
G
G
G
G
G
G
R
C
C
C
C
C
G
C
C
C
C
C
C
C
C
A
C
A
A
A
A
A
U
A
G
U
U
U
U
U
U
U
U
U
Y
A
A
A
A
G
A
G
A
A
A
U
A
A
U
A
A
A
A
A
A
G
A
A
A
A
A
A
A
A
A
A
A
A
A
U
U
U
U
C
U
C
U
U
U
C
C
C
C
U
U
U
U
U
U
C
U
U
A
C
U
U
U
U
U
U
U
U
Y
A
A
G
A
G
A
G
A
G
G
G
G
G
G
G
G
A
A
A
A
A
A
G
U
A
U
C
C
C
C
C
C
C
R
C
C
C
C
C
U
C
U
C
C
C
C
C
C
C
C
C
C
C
C
C
C
G
G
C
C
C
C
C
C
C
C
C
C
4 4 4 4 4 4 4
3 4 5 6 7 8 9
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
…C
…C
…G
…G
…G
…G
…G
…G
…G
…G
…G
…G
U
U
C
U
A
G
A
G
A
A
A
A
A
A
A
A
G
G
G
G
A
U
A
A
C
U
A
A
A
A
A
A
A
R
GA
GA
GA
GA
GA
GU
GA
GU
GA
GA
GG
GA
GA
GA
GG
GA
GG
GU
GG
GG
AG
GA
GA
GA
AG
GA
GA
GA
GA
GA
GA
GA
GA
GR
U
U
U
U
C
U
U
U
U
U
U
U
U
U
-
C
C
C
C
C
U
C
U
C
C
C
C
C
C
C
C
C
C
C
C
U
C
U
U
C
C
C
C
C
C
C
C
C
Y
U
U
U
U
A
G
A
G
G
G
G
G
G
G
G
G
G
G
G
G
G
C
C
C
C
C
C
C
C
C
C
C
C
N
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
Figure 4 Sequence comparison of the 33 available sequences of aspartic acid specific tRNAs. The disposition and colour codes emphasize the structural
alignment. Identical colour codes emphasize observed covariations. The sequence of the yeast tRNAAsp, illustrated in Figures 2 and 3, is boxed. The
consensus sequence, which reflects the most frequent base at each position, is shown at the bottom. Because of the small number of sequences, this
comparison gives only a glimpse of the possible base variations.
Table 1 Table of covariations between the secondary base
pair 12–23 in the D helix and residue 9 in class I tRNAs (analysis is made on 745 sequences of class I elongator tRNAs)
Residue 9
Base pair 12–23
A
U
G
C
A–U
21
–
–
4
U–A
412
–
34
–
G–C
39
4
104
–
C–G
23
2
64
13
Others
9
1
10
1
hydroxyl group of the ribose or the anionic phosphate
oxygen atoms to each other or to a base atom. Thus, the N1
atom of the invariant base A21 binds to the O2’ hydroxyl of
residue U8; the O2’(C55) forms an H bond to the N7 of
G57; the O2’(A58) gives a proton to an anionic oxygen of
phosphate 60. The amino nitrogen N2 of guanine often
binds either to O2’ hydroxyl or to the intraring oxygen O4’
(e.g. N2(G57) to O4’(G19) and O2’(G18)). The anionic
phosphate oxygen atoms of the 5’ phosphate of residue 60
receive each a H bond, one from the hydroxyl of residue 58
8
and the other from the amino group of the invariant C61.
Only some of those additional contacts are observed in all
crystalline forms of tRNAs and a comparison between
different crystal structures reveals a subtle diversity of
slight rearrangements and alternative contacts, often
including distorted hydrogen bond geometries and the
use of the weaker C-H...O/N hydrogen bondings (see
Figure 5).
Conservations of Bases and Loop
Conformations
As mentioned above, both anticodon and thymine loops
always contain seven residues. However, the patterns of
conservation are rather different and reflect their respective
functions. In the anticodon loop, the only strictly invariant
residue is U33, while in the thymine loop almost five
residues are highly conserved, -T54CCRA58-. The anticodon loop should bind precisely the mRNA triplet
complementary to its anticodon bases to form a regular
RNA helix, since recognition is mainly of the Watson–
Crick type. In contrast, structurally the T loop binds
intramolecularly to the D loop. The conformation of the T
loop might also be important for RNAaseP binding to
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
tRNA Structure
Figure 5 Two examples of triples implicating a sheared R.R pair. In the structure of the class I yeast tRNAGln complexed with its cognate synthetase (Rould
et al., 1989), the sheared A13.A22 forms a trans Watson–Crick contact with A46 (both strands are parallel). Notice the contact between the C2-H group of
A13 and the N7 of A22 marked by a lightly dotted line. Notice also in both triples, the H bond between the hydroxyl O2’ atom of the ribose of R13 with the
amino N6 of A22. In the structure of the class II yeast tRNASer (Biou et al., 1994), the sheared G13.A22 pair forms a trans Watson–Crick/Hoogsteen contact
with G9 (both strands are antiparallel). Notice the H bond between the N1(G9) and an anionic phosphate oxygen of A22.
tRNAs during maturation and for recognition by protein
elongation factors during protein synthesis.
The contact between the D and T loops occurs via the
invariant -GG- residues of the D loop and the quasiinvariant -CC- residues of the T loop. The interactions
connect the first G residue with the C residue and the
second G with the C residue. The G18.C55 contact is
locally parallel with a trans orientation of the H bonds
which are bifurcated and occur between the O4(C) and
N1(G)/N2(G). The G19–56 interaction is a classical cis
Watson–Crick pair, antiparallel following a rotation
about its 5’ phosphate. At the same time, the -GG- segment
interdigitates with the R57A58 segment so that R57 of the
T loop intercalates between G18 and G19 and G18 of the D
loop intercalates between R57 and A58.
The only strictly invariant residue in the anticodon loop
is situated where the chain reversal occurs. It is accomplished by a U turn identical to that present in the T loop. It
is the phosphate group 3’ of U33 (or C55 in the T loop),
which rotates about the P-O3’ torsion angle. The U turn is
stabilized by contacts with residue n 1 2, A35 (or G57 in
the T loop). The pyrimidine ring of the U turn stacks over
the 5’ phosphate of residue n 1 2, the O2’ hydroxyl of the
pyrimidine interacts with the N7 of the purine at n 1 2, and
the N3 imino proton of the uracil H bonds to an anionic
phosphate oxygen (the Rp oxygen) of the 3’ phosphate of
residue n 1 2. Position 37 is occupied invariably by a
purine (mainly A) and is often heavily modified. It
stabilizes the codon–anticodon triplet interaction by
stacking on the very discriminating base pair formed
between the first base of the codon and the third base of the
anticodon. The structural and functional basis for the
chemical modifications on residue 37 are difficult to
establish: stabilization by 3’-end dangling purines (as at
the discriminator 73 position), blocking of the Watson–
Crick sites preventing pairing with the preceding residue on
the mRNA. Correlations have also been found between the
chemical nature of the modifications and the hydrophobicity/hydrophilicity of the coded amino acids. The first and
last bases of the anticodon loop have restricted variations,
leading in more than 60% of sequences to a one H bond
Y32.A38 base pair between O2(Y32) and N6(A38).
Unusual tRNAs
In the preceding sections, the class I tRNAs, the best
known structurally, have been extensively described. Only
one class II tRNA (or long arm tRNA), the E. coli
tRNASer, is known crystallographically in a complex with
its cognate aminoacyl tRNA synthetase (Biou et al., 1994).
In class II tRNAs, residue 45 base-pairs with a residue in
the variable helix and cannot form a triple with base pair
10–25. Also, residue 9 does not interact with 12–23 but
with base pair 13–22, which is often a purine–purine
G13.A22 pair, or A13.A22 (see Table 2). The R13.R22
base pair belongs to a special type of base pair, the sheared
type where the Hoogsteen sites of the A interact with the
N3 and N2 atoms of the G (in sheared A.A pair, there is
only one bona fide H bond between N6(A) and N3(A)). In
the triple 9...13–22, the Hoogsteen sites of residue 9
interact with the Watson–Crick sites of residue 13, still
available since residue 22 interacts with N3 and N2 on the
shallow groove side. The triple A9...A13.A22 is also
observed, which implies a slight displacement of A9 so that
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
9
tRNA Structure
Table 2 Table of covariations between the secondary base
pair 13–22 in the D helix and residue 9 in class II tRNAs (analysis is made on 187 sequences of class II elongator tRNAs)
Residue 9
Base pair 12–23
A
U
G
C
A•A
32
–
8
3
G•A
–
4
115
–
G•U
–
–
11
–
Others
7
1
4
2
its Hoogsteen sites (N7 and N6) interact with the Watson–
Crick sites (N6 and N1) of A13.
The tRNASec is a tRNASer able to read, in an
appropriate context, UGA stop codons for incorporation
of selenocysteine and constitutes a variant of class II
tRNAs. In archaea and eukaryotes, the acceptor helix is
9 bp long and the T helix is only 4 bp long (9/4 model). On
the other hand, in prokaryotes, the acceptor helix is 8 bp
long with a 5-bp T helix (8/5 model). Thus, for all
phylogenetic groups, the length of the acceptor arm is
13 bp. The D helix is also changed and displays 6 bp. The T
loop is shorter with the invariant -GG- residues centrally
disposed.
Initiator tRNAs (Schmitt et al., 1998) also display some
sequence peculiarities linked to their biological function:
they must bind directly into the ribosomal P site and not in
the A site where elongator tRNAs are channelled to. Thus,
the first base pair of the acceptor helix is never Watson–
Crick in prokaryotes, which favours recruitment of
initiator factor IF-2. Prokaryotic initiator tRNAs also
present an interesting reversal at the level of the 11–24 base
pair where it is R11–Y24, instead of the Y11–R24 as in
elongators. The T loop, normal in prokaryotes, is different
in eukaryotes where positions 54 and 60 are occupied by
adenines. The usual trans Hoogsteen pair T54.A58 is thus
replaced by the similar trans A54.A58 pair between the
Watson–Crick sites of A54 and the Hoogsteen sites of A58.
Eubacterial and eukaryotic initiator tRNAs have an odd
distribution of G–C pairs at the beginning of the acceptor
helix (alternating G–C/C–G) and at the end of the
anticodon helix (a run of three G–C pairs). Eukaryotic
initiator tRNAs have other unusual features, like the
unique sugar modification on residue 64, a phosphoribosyl
group on the 2’ hydroxyl, a modification which hinders
binding of the elongation factor EF-1a and, thus, access to
the A site.
The structural diversity of mitochondrial tRNAs is
enormous. In some of them the T hairpin is missing and
replaced by a single-stranded segment, while in others it is
the D hairpin. In the remaining stems and loops, there is a
10
systematic loss of base invariance or semi-invariance
compared to cytoplasmic tRNAs. In the absence of a
reference crystal structure, it is difficult to discuss the
structural aspects of such truncated and functional
tRNAs.
Summary
tRNA constitutes a paradigm for RNA structure and
folding. It possesses a well-defined consensus secondary
structure, the cloverleaf, which folds into a common threedimensional architecture with a characteristic L shape. The
three-dimensional fold can be divided into two domains,
similarly built of two coaxial helical stems. The threedimensional architecture is kept in place by loop–loop
tertiary contacts between the thymine and dihydrouridine
loops, together with tertiary contacts involving the singlestranded junctions linking the helical domains and the deep
groove of one helix, the dihydrouridine helix. The latter
contacts involve non-Watson–Crick pairings and formation of base triples.
tRNA structure displays several elements of the RNA
folding logic: coaxial stacking of contiguous helices,
formation of base triples in a RNA helix deep groove,
loop–loop interactions with Watson–Crick and nonWatson–Crick pairs, the U-turn motif for hairpin formation, and the extensive use of additional hydrogen bonding
involving the ribose hydroxyl O2’, the anionic phosphate
oxygen atoms, and some polar C-H groups. The modular
view of RNA architecture, based on a hierarchical
assembly of recurrent RNA motifs, first glimpsed at in
the tRNA structure, has been corroborated by the recent
crystal structures of larger RNA domains and ribozymes.
The increasing number of crystal structures has revealed
an amazing and subtle variability in precise atomic
contacts. However, the microheterogeneities in the specific
atomic contacts between residues important for the
stability of the global tertiary fold maintain identical
overall topological arrangements. These structural constraints lead to base conservations or base covariations in
sequence comparisons and alignments. Comparisons
between RNA molecules show that topologically and
functionally distinct molecules share quasi-identical threedimensional motifs which display clear signatures in
sequence conservation and variability.
References
Auffinger P and Westhof E (1998) Location and distribution of modified
nucleotides in tRNA. In: Grosjean H and Benne R (eds) Modification
and Editing of RNA, pp. 569–576. Washington, DC: American Society
for Microbiology.
Biou V, Yaremchuk A, Tukalo M and Cusack S (1994) The 2.9 Å crystal
structure of T. thermophilus seryl-tRNA synthetase complexed with
tRNASer. Science 263: 1404–1436.
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
tRNA Structure
Crick HFC (1966) Codon–anticodon pairing: the wobble hypothesis.
Journal of Molecular Biology 19: 548–555.
Gautheret D, Damberger SH and Gutell RR (1995) Identification of
base-triples in RNA using comparative sequence analysis. Journal of
Molecular Biology 248: 27–43.
Holley RW, Apgar J, Everett GA et al. (1965) Structure of a ribonucleic
acid. Science 147: 1462–1465.
Kim S-H (1978) The three dimensional structure of transfer RNA and its
functional implications. Advances in Enzymology 46: 279–315.
Klug A, Ladner J and Robertus JD (1974) The structural geometry of cocordinated base changes in transfer RNA. Nature 89: 511–516.
Levitt M (1969) Detailed molecular model for transfer ribonucleic acid.
Nature 224: 759–763.
Moras D, Comarmond MB, Fisher J et al. (1980) Crystal structure of
yeast tRNAAsp. Nature 288: 669–673.
Rould MA, Perona JJ, Söll D and Steitz TA (1989) Structure of E. coli
glutaminyl-tRNA synthetase complexed with tRNA(Gln) and ATP at
2.8 Å resolution. Science 246: 1135–1142.
Schmitt E, Panvert M, Blanquet S and Mechulam Y (1998) Crystal
transformylase complexed with the
structure of methionyl-tRNAMet
f
. EMBO Journal 17: 6819–6826.
initiator formyl-methionyl-tRNAMet
f
Sprinzl M, Horn C, Brown M, Loudovitch A and Steinberg S (1998)
Compilation of tRNA sequences and sequences of tRNA genes.
Nucleic Acids Research 26: 148–153.
Suddath FL, Quigley GJ, McPherson A et al. (1974) Three-dimensional
structure of yeast phenylalanine transfer RNA at 3.0 angstroms
resolution. Nature 248: 20–24.
Sundaralingam M (1973) The concept of a conformationally ‘rigid’
nucleotide and its significance in polynucleotide conformational
analysis. Jerusalem Symposia of Quantum Chemistry and Biochemistry
5: 417–456.
Westhof E, Dumas P and Moras D (1985) Crystallographic refinement
of yeast aspartic acid transfer RNA. Journal of Molecular Biology 184:
119–145.
Further Reading
Grosjean H and Benne R (eds) (1998) Modification and Editing of RNA.
Washington, DC: American Society for Microbiology.
Quigley GJ and Rich A (1976) Structural domains of transfer RNA.
Science 194: 796–806.
Rich A and RajBhandary UL (1976) Transfer RNA: molecular
structure, sequence, and properties. Annual Review of Biochemistry
45: 805–860.
Saenger W (1984) Principles of Nucleic Acid Structure. New York:
Springer-Verlag.
Söll D and RajBhandary UL (eds) (1995) tRNA Structure, Biosynthesis,
and Function. Washington, DC: American Sociey for Microbiology.
ENCYCLOPEDIA OF LIFE SCIENCES / & 2001 Nature Publishing Group / www.els.net
11