doi:10.1006/jmbi.2001.4801 available online at http://www.idealibrary.com on
J. Mol. Biol. (2001) 310, 987±999
A Sequence Required for ÿ1 Ribosomal Frameshifting
Located Four Kilobases Downstream of the
Frameshift Site
Cynthia P. Paul, Jennifer K. Barry, S. P. Dinesh-Kumar
VeÂronique Brault and W. Allen Miller*
Plant Pathology Department
351 Bessey Hall, Iowa State
University, Ames
IA 50011, USA
Programmed ribosomal frameshifting allows one mRNA to encode regulate expression of, multiple open reading frames (ORFs). The polymerase
encoded by ORF 2 of Barley yellow dwarf virus (BYDV) is expressed via
minus one (ÿ1) frameshifting from the overlapping ORF 1. Previously,
this appeared to be mediated by a 116 nt RNA sequence that contains
canonical ÿ1 frameshift signals including a shifty heptanucleotide followed by a highly structured region. However, unlike known ÿ1 frameshift signals, the reporter system required the zero frame stop codon and
did not require a consensus shifty site for expression of the ÿ1 ORF. In
contrast, full-length viral RNA required a functional shifty site for frameshifting in wheat germ extract, while the stop codon was not required.
Increasing translation initiation ef®ciency by addition of a 50 cap on the
naturally uncapped viral RNA, decreased the frameshift rate. Unlike any
other known RNA, a region four kilobases downstream of the frameshift
site was required for frameshifting. This included an essential 55 base
tract followed by a 179 base tract that contributed to full frameshifting.
The effects of most mutations on frameshifting correlated with the ability
of viral RNA to replicate in oat protoplasts, indicating that the wheat
germ extract accurately re¯ected control of BYDV RNA translation in the
infected cell. However, the overall frameshift rate appeared to be higher
in infected cells, based on immunodetection of viral proteins. These ®ndings show that use of short recoding sequences out of context in reporter
constructs may overlook distant signals. Most importantly, the remarkably long-distance interaction reported here suggests the presence of a
novel structure that can facilitate ribosomal frameshifting.
# 2001 Academic Press
0
*Corresponding author
Keywords: Luteovirus; recoding; reporter genes; 3 untranslated region;
plant virus translation
Introduction
Ribosomal frameshifting is a recoding event that
provides an economical means of storing and
expressing genetic information.1,2 One nucleotide
sequence tract encodes two overlapping open reading frames (ORFs) and provides signals for control
of their translation. This compression of information is especially advantageous for viruses in
which nucleotide sequence length must be mini-
Present addresses: C. P. Paul, Department of Biological Chemistry, Medical Science I, University of Michigan, 1301
E. Catherine, Ann Arbor, MI 48109-0606, USA; J. K, Barry, Pioneer Hi-Bred International, Department of
Biotechnology Research, Johnston IA, 50141-1004, USA; S. P. Dinesh-Kumar, Molecular, Cellular & Developmental
Biology Department, Yale University, OML 451, P.O. Box 208104, New Haven, CT 06520-8104, USA; V. Brault, INRA,
Unite de Recherche en Biologie des Interactions Virus-Vecteur, 28, rue de Herrlisheim, BP 507, 68001 Colmar Cedex,
France.
Abbreviations used: BYDV, barley yellow dwarf virus; GUS, b-glucuronidase; HTLV-2, human T-cell leukemia
virus type II; ORF, open reading frame; PPT, polypyrimidine tract; RCNMV, red clover necrotic mosaic virus; TE,
cap-independent translation element; UTR, untranslated region.
E-mail address of the corresponding author: wamiller@iastate.edu
0022-2836/01/050987±13 $35.00/0
# 2001 Academic Press
988
mized. In minus one (ÿ1) frameshifting, ribosomes
shift one nucleotide in the 50 direction during translation of an ORF, then resume translation of an
overlapping reading frame. Viruses in the Retroviridae, Nidovirales, astrovirus, Luteoviridae, dianthovirus, sobemovirus, Totiviridae, and mushroom
bacilliform
virus
groups
employ
ÿ1
frameshifting.2 ± 4 In most cases the ÿ1 reading
frame encodes the active site of the viral polymerase. In eukaryotes, ÿ1 frameshifting is not known
to occur in non-viral genes, although some candidate frameshift sequences exist in cellular genes.5
Thus, the ÿ1 frameshift mechanism is a potential
target for antiviral agents.6 Furthermore, understanding the mechanism of ÿ1 frameshifting can
shed light on fundamental mechanisms of mRNAribosome interactions.
The prevailing simultaneous slippage model for
ÿ1 frameshifting purports that the tRNAs, basepaired to the shifty heptanucleotide in the zero
frame, and in the ribosomal A and P sites, simultaneously slip back one base on the mRNA.2,7
Translation then resumes in the new, ÿ1 reading
frame. Two cis-acting signals in the mRNA are
necessary.1,2,8 One is the heptanucleotide at the
shift site, which usually ®ts the consensus: X XXY
YYZ, where X is any base, Y is A or U, and Z is
any base but G, and the spaces separate codons in
the initial (zero) frame.7 This sequence allows at
least two of the three anticodon bases to re-pair to
the mRNA in the ÿ1 frame after the shift.
The second signal is a highly structured region,
usually a pseudoknot, that begins ®ve or six
nucleotides downstream from the shifty site.8 ± 10 In
some cases, a stem-loop11,12 will suf®ce. In gene 10
of bacteriophage T7, a sequence in the 30 untranslated region (UTR), 200 nt downstream of the frameshift site, is necessary.13 The structured region
favors the chances of ÿ1 frameshift by inducing
elongating ribosomes to pause on the shifty
site.14,15 However, the structure may have a more
speci®c role than just inducing pausing, because
some highly stable structures do not facilitate
frameshifting16 ± 18 even if they can cause
pausing.14,15 The potential role(s) of protein factors
in ÿ1 frameshifting is controversial,19,20 but the
prevailing theory is that the downstream element
both pauses the ribosome and speci®cally interacts
with it in a way that is not yet clear, to facilitate
the ÿ1 frameshift.
Early studies showed that the shifty site and the
adjacent pseudoknot (or stem-loop) were necessary
and suf®cient for frameshifting at a level similar to
that of the full-length virus sequence context.7,21
Thus, frameshifting is now studied mostly using
reporter genes linked to the frameshift element, so
that ÿ1 frameshifting is needed for reporter gene
expression. However, discrepancies have been
observed between in vitro and in vivo frameshift
assays.22,23 Thus, understanding of the mechanism
and control of frameshifting by a replicating virus
is incomplete. Here, we show that out-of-context
frameshift studies can lack important signals, and
Frameshifting Promoted by the 3 0 UTR
Figure 1. (a) Genome organization of full-length infectious BYDV transcript, PAV6. Molecular masses of proteins encoded by viral ORFs (numbered boxes) are
indicated in kilodaltons (K). Nucleotide positions of
selected restriction enzyme sites are indicated below
viral genomic RNA (bold line). Shaded box: 100 nt 30
cap-independent translation element (TE). (b) Viral
insert in GUS expression vectors. The BYDV insert,
bases 1131 to 1246, includes the overlapping region of
ORFs 1 and 2, with the shifty heptanucleotide (black
background). Bases in italics are vector-derived. Underlined bases are paired in the bulged stem-loop
structure.26 Amino acid sequences of the zero frame
(top) and ÿ1 frame (bottom) encoded by the 30 end of
ORF 1 and 50 end of ORF 2, respectively, are indicated
below the nucleotide sequence. Lower-case amino acid
residues are translated only in the constructs with the C
substitution in the ORF 1 stop codon (bold UAG) that
converts it to a serine codon (UCG). Amino acid residues in italics comprise the beginning of the GUS coding region.
that new, additional sequences besides the canonical shifty site and adjacent structured region are
needed for frameshifting on a full-length, plant
viral genomic RNA.
In vitro24 studies indicated that the polymerase
gene (ORF 2) of Barley yellow dwarf virus (BYDV,
PAV serotype) is translated via ÿ1 ribosomal
frameshifting of ribosomes from ORF 1 in the
short, 13 nt region of overlap (Figure 1). The frameshift site (bases 1152-1158) consists of a canonical
shifty sequence, G GGU UUU, followed by a
region beginning 6 nt downstream that is predicted
to form either a large bulged stem-loop or two
smaller kissing stem-loops.25 Only the bulged
stem-loop is phylogenetically conserved.26 A virus
with a related polymerase gene, Red clover necrotic
mosaic virus (RCNMV), was shown to require a
similar bulged stem-loop for ÿ1 frameshifting.12
The BYDV frameshift sequence, when placed in
front of a b-glucuronidase (GUS) reporter gene
gave about 1 % apparent frameshifting,25 which is
at the low end of the range of known viral frameshift ef®ciencies (1-33 %). The zero frame ORF
(ORF 1) stop codon, which is located immediately
989
Frameshifting Promoted by the 3 0 UTR
30 of the shifty heptanucleotide, was required for
b-glucuronidase (GUS) expression. This differs
from other ÿ1 frameshifting viruses where a stop
is not necessary,7,18 as the frameshift site is often
far upstream of the zero frame stop codon.4 Previous studies, in which we mapped a cap-independent translation element to a 30 portion of the
BYDV genome, indicated that an even more distal
sequence in the 30 untranslated region (UTR) may
be necessary for frameshifting.27 Here, we verify
that observation and report a new frameshift
sequence element that is located thousands of
bases downstream of the frameshift site.
to that of RCNMV (GGAUUUU) had little effect
(pM2SH18, Table 1). This is consistent with work
by Kim & Lommel,12 who showed that the BYDV
shifty site had no effect on RCNMV frameshifting.
Surprisingly, constructs with major mutations in
the shifty heptanucleotide still allowed wild-type
GUS expression. This includes one construct that
completely disrupts codon-anticodon base-pairing
by both the A and P site tRNAs after the frameshift
(pM7SH27, shifty sequence changed to G CUG
CAU, Table 1). Of the 17 different mutant heptanucleotides tested (Table 1 and data not shown),
none had a signi®cant effect on GUS expression,
with the exception of UUUUUUU, which caused a
slight increase (pM4SH33, Table 1). Thus, it is clear
that a canonical shifty heptanucleotide is not
required for expression of GUS from the ÿ1 frame
in this context.
Carrot is not a host of BYDV, so we tested
frameshifting in protoplasts from oat, a BYDV
host. These experiments utilized the maize Adh1
promoter and ®rst intron because it expresses at
much higher levels in monocotyledoneae than
the CaMV 35S promoter used in carrot cells
(Table 2; Callis et al.28). A more appropriate inframe positive control plasmid, pADHS(0)UCAG,
was used. It fuses the upstream portion of ORF
2 to GUS exactly as would occur in the ÿ1 frameshift (Figure 1(b)), whereas the previous inframe control, pS(0)UCG (and pADHS(0)UCG),
fuses the 33 codons in the zero frame, downstream of the ORF 1 stop codon, with GUS.
When pADHS(0)UCG and pADHS(0)UCAG were
compared, the latter gave about 40 % greater
GUS activity (Table 2, Experiment 2). This, in
turn, gave a lower calculated frameshift
ef®ciency of less than 1 %. The ability of
constructs with the Adh1 promoter, and any
stop codon at the end of ORF 1, to facilitate
apparent ÿ1 frameshifting was evident in
oat cells (pADHS(ÿ1)UAG, pADHS(ÿ1)UAA,
pADHS(ÿ1)UGA) (Table 2). We then tested the
Results
Out of their natural context, canonical BYDV
frameshift signals do not play a role in ÿ1
GUS expression in reporter constructs
Previously, we investigated frameshifting by
BYDV sequences using reporter plasmids with the
GUS ORF out of frame from the start codon, so
that a ÿ1 ribosomal frameshift was required for
GUS expression in carrot cells.25 The sequence
between the start codon and GUS ORF contained
116 bases of viral sequence, including the shifty
heptanucleotide and downstream secondary structure (Figure 1(b)). Using that construct, apparent
frameshifting ef®ciency was measured at about
1 %. Replacement of the ORF 1 UAG stop codon
with a UCG sense codon abolished apparent
frameshifting.25 To further examine the dependence on a zero frame stop codon, we tested other
stop codons. Replacement of the ORF 1 UAG with
UAA also supported wild-type levels of GUS
expression in carrot cells (pM9SH1, Table 1), while
UGA gave about 50 % as much GUS activity
(pM9SH2). Thus, in this assay, any stop codon is
suf®cient for GUS expression from the ÿ1 frame.
To test for other differences from consensus ÿ1
frameshift signals, we investigated the role of the
shifty heptanucleotide (GGGUUUU). Mutating it
Table 1. GUS activity in carrot cells
Ept. 1
a
shift site/stop codon
GGGUUUUUCG
GGGUUUUUAG (wt)
GGAUUUUUAG
UUUUUUUUAG
GGGGCAUUAG
GCUGCAUUAG
GGGUUUUUAA
GGGUUUUUGA
GGGUUUUUCG
GGAUUUUUCG
a
b
c
b
plasmid
c
GUS
pS(0)UCG (inframe control) 12,158 204
pS(ÿ1)UAG
203.5 30
pM2SH18
170.3 1.9
pM4SH33
299.2 5.8
pM6SH8
pM7SH27
pM9SH1
pM9SH2
pS(ÿ1)UCG
4.0 1.2
pM8SH14
Expt. 2
Expt. 3
%FS
GUS
%FS
GUS
%FS
100
1.7
1.4
2.5
10,943 10
170.9 4.4
254.8 9.4
100
1.6
2.3
13,796
360
100
2.6
182.0 5.3
167.4 8.7
147.2 6.8
96.5 5.4
2.4 0.5
1.1 0.1
1.7
1.5
1.4
0.9
0
0
183.8
321.9
328.0
151.1
1.3
2.3
2.4
1.1
0
Shifty site in italics, stop codon in bold, mutations underlined.
Details of constructs are in Materials and Methods and Figure 1(b).
GUS assays are described in Materials and Methods and as described.25
990
Frameshifting Promoted by the 3 0 UTR
Table 2. GUS expression in oat cells
Experiment 1
Plasmid
pS(0)UCG
pS(ÿ1)UAG
pS(ÿ1)UCG
pADHGUS
pADHS(0)UCG
pADHS(0)UCAG
pADHS(ÿ1)UAG
pADHS(ÿ1)UAA
pADHS(ÿ1)UGA
pADHS(ÿ1)UCG
pADHS(ÿ1)UGG
pADHS(ÿ1)CGG
Experiment 2
GUS
%FS versus UCG
1236 204
3.7 6.6
ÿ1.9 1.5
20,058 315
18,507 2212
100
0.3
ÿ0.2
322.4 56.3
1.7
8.0 4.4
0
100
roles of other sense codons besides the previously tested UCG (serine) codon. Codons for
tryptophan (pADHS(ÿ1)UGG) and arginine
(pADHS(ÿ1)CGG),
as
well
as
serine
(pADHS(ÿ1)UCG) in place of the ORF 1 stop
codon abolished GUS activity (Table 2). Thus,
there is a requirement for a stop codon for
GUS expression in the ÿ1 frame in the reporter
constructs in both carrot and oat. The lack of a
requirement for a shifty heptanucleotide and the
requirement for a stop codon indicate either that
BYDV uses a frameshift mechanism fundamentally different from that of other viruses, or that
the constructs do not re¯ect viral translation
accurately.
Frameshifting is more efficient on replicating
viral RNA than on reporter genes
To understand frameshifting in the natural viral
infection, oat protoplasts were inoculated with
inoculum: no RNA
dilution: 1 0.5 0.25
BYDV-PAV RNA
1
0.5
0.25
99K
39K
apparent % frameshift:
20
12
6
Figure 2. Western blot detection of proteins containing 39 kDa protein antigen in uninfected (no RNA) and
PAV6-infected protoplasts. Total cell lysate 24 hours
after inoculation was separated by polyacrylamide gel
electrophoresis, blotted onto nylon membrane, probed
with antiserum prepared against 39 kDa protein, and
detected using the ECL luminescent system as described
in Materials and Methods. Different dilutions of the
same cell lysate were loaded and the calculated frameshift rates based on signal intensity of 39 kDa and
99 kDa bands are shown.
GUS
5780 1706
6176 473
10,044 1859
16.6 3.2
54.5 2.3
85.7 43.0
ÿ0.9 4.3
ÿ0.2 1.2
ÿ2.6 3.5
%FS versus UCG
100
0.3
0.9
1.4
0
0
0
%FS versus UCAG
100
0.2
0.5
0.9
0
0
0
infectious viral RNA, and the products of ORF 1
(39 kDa) and ORFs 1 ‡ 2 (99 kDa) were detected
by immunoblotting using anti-39 kDa antiserum
(Figure 2). The ratio of 99 kDa to 39 kDa protein
varied from 6-20 % among several different
measurements and experiments, and depended on
the amount loaded onto the gel (Figure 2). The
ability to quantify the luminescent signal from the
Western blots was very limited, owing to the
narrow dynamic range of the detection system.
Regardless of the exact frameshift rate, it appears
that frameshifting on full-length, replicating viral
RNA in plant cells is more ef®cient than in reporter
constructs (Tables 1 and 2) containing only 116
bases of viral sequence.
The 39K ORF stop codon is not required for
frameshifting on full-length transcripts in vitro,
but is required for infectivity in oat protoplasts
The apparent discrepancies in frameshifting
between the reporter constructs and the replicating
virus prompted us to investigate the sequence
requirements for frameshifting within the natural
context of full-length viral RNA. Mutations in the
ORF 1 stop codon, corresponding to those made in
the GUS reporter constructs, were introduced into
full-length BYDV genomic clone PAV6 and translated in wheat germ extracts. Both the 39 kDa and
99 kDa proteins were translated from full-length
transcripts harboring any of the three ORF 1 stop
codons (Figure 3(a), PAV6, PAVORF1UAA,
PAVORF1UGA). Replacement of the ORF 1 stop
codon with a sense codon increased the size of the
ORF by 44 codons, resulting in a predicted 43 kDa
protein product. Both 43 kDa and 99 kDa proteins
were translated from mutants with the UCG, CGG,
or UGG codons in place of the UAG stop codon,
with no signi®cant change in frameshift ef®ciency
(Figure 3(a), lanes PAVORF1UCG, PAVORF1CGG,
PAVORF1UGG). Thus, in contrast to the GUS
reporter results, the stop codon of the zero frame
ORF was not necessary for translation of the
99 kDa fusion protein from full-length viral RNA
in the wheat germ system.
991
Frameshifting Promoted by the 3 0 UTR
without a stop codon (preventing translation of the
polymerase) as was observed with the GUS reporter constructs, or (more likely) that the 4 kDa
extension to ORF 1 in the absence of the stop
codon was lethal to the function of the ORF 1
protein.
The shifty heptanucleotide is required for
frameshifting in full-length viral transcripts
Figure 3. Translation and replication of mutant fulllength viral genomic RNAs. (a) Autoradiograph of
wheat germ translation products of PAV6 mutants with
the ORF1 stop codon (UAG) replaced by the three or
four bases indicated at the end of the transcript name.
Mobilities of wild-type (39 kDa) or extended by the
absence of stop codon (43 kDa) products of ORF 1 and
the 99 kDa frameshift product of ORFs 1 ‡ 2 are indicated at the right. Percentage frameshift is calculated as
in Materials and Methods. ``100`` indicates equivalent of
100 % frameshifting, due to fusion of ORFs 1 and 2 by
insertion of C in the ORF1 stop codon. Each of the
RNAs was inoculated in oat cells, and total viral coat
protein accumulation measured by ELISA with antibody
against BYDV virions. Lane BMV shows translation products of all four BMV RNAs with mobilities at left. The
39 kDa protein from ORF 1 reproducibly migrates as a
46 kDa protein.24 (b) Wheat germ translation products
of wild-type (PAV6) RNA and mutant form (SHSIL)
with shifty site GGGUUUU changed to CGGCUUC.
To test for biological effects of the mutations in
the ORF 1 stop codon, the ability of the mutant
full-length transcripts to infect protoplasts was
determined. Frameshifting is necessary for
expression of the polymerase and thus for replication. Replication was assayed by ELISA detection
of virions (Figure 3(a)). (Production of coat protein
for virions requires subgenomic RNA synthesis via
RNA replication.) Virions generated by replication
of transcripts with mutant UGA or UAA termination codons accumulated to wild-type levels in
protoplasts (Figure 3(a), ELISA). However, no virions accumulated when the ORF 1 stop codon was
changed to any of the three sense codons, or to
UCAG (Figure 3(a)). These replication results were
con®rmed by Northern blot detection of viral RNA
accumulation (data not shown). Thus, only those
transcripts with a functional ORF 1 stop codon
were able to replicate in oat protoplasts. We conclude that either no frameshifting occurred in vivo
We next re-examined the role of the shifty heptanucleotide, this time in the full-length context. A
``translationally silent'' mutant was constructed by
replacing the wild-type shifty heptanucleotide
sequence G GGU UUU with C GGC UUC in fulllength transcripts (mutant SHSIL). This mutant
should completely disrupt slippage of the tRNAs,
by preventing re-pairing in the ÿ1 frame, but the
amino acid sequence of the frameshift product
would be the same as in the wild-type protein
(Gly-Phe) if frameshifting did occur. This mutant
did not direct frameshifting in wheat germ extract
(Figure 3(b)). Furthermore, this mutant RNA failed
to replicate in oat protoplasts, as indicated by
Northern blot hybridization of viral RNA accumulated in the cells (Figure 4). Lack of replication
must be due to changes in RNA sequence because
no amino acid residues would have changed if
frameshifting took place. Thus, in contrast to
results with reporter constructs, the shifty heptanucleotide is required for frameshifting within the
natural context of full-length BYDV RNA both
in vitro and, by inference, in vivo.
Figure 4. Replication of selected BYDV mutants. Transcripts from SmaI-linearized plasmids were electroporated into oat protoplasts, and total RNA extracted 24
hours post-infection and analyzed by Northern blot as
in Materials and Methods, using a probe complementary
to the 30 -terminal 1000 nt of the BYDV genome. Equal
loading of total RNA in each lane was veri®ed by ethidium bromide staining.
992
Frameshifting Promoted by the 3 0 UTR
Presence of a 50 cap decreases frameshifting
Figure 5. Effects of 30 truncation and capping on
frameshifting by BYDV RNA in wheat germ extracts. (a)
Products of transcripts from pPAV6 that had been linearized with the indicated restriction enzymes (see map,
Figure 1(a)). (b) Products of capped or uncapped fulllength (SmaI-linearized) transcripts from pPAV6 or
pPAV3788-4515,G4922C, which translates and replicates like wild-type PAV6 RNA.39 % fs indicates percentage frameshifting. rel. % fs indicates relative
frameshifting normalized to the wild-type percentage
frameshifting on PAV6 RNA.
0
The 3 end of the viral genome contains
elements required for frameshifting
The discrepancies between the reporter construct
and full-length viral RNA may be due to sequences
in the full-length RNA that are absent from the
reporter constructs. To test this, deletions were
made in full-length viral RNA and translated
in vitro. When full-length PAV6 transcripts were
translated in wheat germ extracts, ÿ1 ribosomal
frameshifting occurred at rates from 1-4 %
(Figures 3, 5 and 6). This varied between wheat
germ batches and with potassium concentration.24
Thus the level of frameshifting in wheat germ
extract may not quantitatively re¯ect that in
infected cells. However, wheat is a host of BYDV,
so the wheat germ extract should respond to
BYDV RNA translation signals.
Surprisingly, frameshifting was reduced greatly
on viral genomic transcripts that were truncated in
the 30 untranslated region (Figure 5(a)). When the
30 cap-independent translation element (30 TE, nt
4810-4918) was deleted, all translation of uncapped
mRNA dropped by at least 30-fold (Figure 5(a),
ScaI4514 digestion). For the smaller 30 truncations,
the amount of 39 kDa protein remained fairly constant, while the amount of 99 kDa frameshift product decreased as the size of the 30 truncation
increased. These data suggest that a sequence
located about 4 kb downstream of the frameshift
site is required for frameshifting.
The lack of frameshifting in wheat germ extracts
when PAV6 transcripts were truncated at the
PstI5010 site indicated that at least part of the
required 30 element lies 30 to base 5010. To map the
50 boundary of the 30 frameshift element, a series of
internal deletion mutants between the shifty site
and the 30 UTR was constructed. Many of these
deletions included the 30 TE, and thus required a 50
cap on the mRNA for ef®cient translation. Therefore, we ®rst investigated the effect of a 50 cap on
frameshifting, by comparing frameshift rates of
capped and uncapped transcripts that still harbor
the 30 TE. Capped transcripts frameshifted less
than half as ef®ciently as their uncapped counterparts (Figure 5(b)). In 15 different experiments,
capped transcripts frameshifted 26(10) % as ef®ciently as their uncapped counterparts. Mutant
3788-4515,G4922C, in which the ORF6 start
codon was altered, frameshifts about the same as
PAV6 with or without a cap. This also shows that
intact ORF 6 is not necessary for frameshifting. In
both cases, the reduction in frameshift percentage
was a consequence of the increase in 39 kDa protein due to presence of the 50 cap. The amount of
99 kDa protein remained constant in capped and
uncapped transcripts. Thus, increasing the
initiation rate decreased the percentage of ribosomes that frameshift.
Mapping the distant 30 UTR sequence required
for frameshifting
One of the largest deletions that permitted
frameshifting was construct PAV1745-4919
(Figure 6(a)). This transcript lacks the entire 30
TE,29 therefore capped transcript was used. Importantly, this result demonstrates that the elements
required for cap-independent translation and frameshifting in vitro do not overlap. The greater than
threefold increase in frameshift rate over wild-type
may result from moving the downstream element
closer to the frameshift site. The large deletion,
PAV1684-4837 frameshifted ef®ciently, but
PAV1684-5112 frameshifted poorly (Figure 6(a)).
Thus, the 30 UTR frameshift element has a 50 boundary located somewhere between bases 4920 and
5112, and a 30 boundary upstream of the PvuI5322
site. To localize the 30 boundary, smaller deletions
within this 400 nt region were tested. Transcript
PAV5280-5426 frameshifted at wild-type levels
(Figure 6(b)). This localizes the 30 end of the
sequence needed for full-frameshifting to nucleotide 5279.
A 29 base polypyrimidine tract (PPT) lies
between bases 5016 and 5046 (Figure 6(e)). PPTs
have been implicated in RNA-protein interactions
that regulate many functions, including cap-independent translation30,31 and RNA replication.32
However, deletion of the PPT in transcript
PAV5016-5045 had little effect on frameshifting
(Figure 6(d)). In contrast, an adjacent deletion in
993
Frameshifting Promoted by the 3 0 UTR
Figure 6. Effect of deletions in distant downstream regions of the BYDV genome on ribosomal frameshifting. (a)(d). Wheat germ translation products of transcripts from SmaI-linearized plasmids containing the indicated deletions.
Transcripts in (a), (b) and (d) are capped. Uncapped PAV5046-5100 gave no detectable frameshifting (not shown).
(e) Downstream sequence required for ribosomal frameshifting.
mutant PAV5046-5100 virtually abolished frameshifting (Figure 6(d)). Thus, the PPT is not required
for frameshifting, but a sequence between bases
5046 and 5100 is essential.
Transcript PAV5019-5279 did not frameshift,
but PAV5118-5279, PAV5186-5279 (data not
shown), PAV5158-5202, PAV5180-5205, and
PAV5183-5205 all frameshifted at 30-50 % of
wild-type levels (Figure 6(c)). Together, all the
deletion data indicate the presence of an essential
distant frameshift core element with a 50 boundary
at nucleotide 5046 and a 30 boundary between
nucleotides 5100 and 5117, and a distant frameshift
enhancer element located between nucleotides
5118 and 5279. The sequence between nucleotides
5183 and 5205 is certain to play a role in the enhancer activity.
smallest deletion that abolished frameshifting,
PAV5046-5100, did not replicate (Figure 4). All
other mutants that did not frameshift in wheat
germ extract also did not replicate (data not
shown). Transcript PAV5183-5205, which frameshifted about 50 % of wild-type levels in wheat
germ extracts (Figure 6(c)), replicated at a reduced
level in protoplasts, accumulating about 50 % as
much viral RNA. Similar mutants, PAV5180-5205
and PAV5158-5202, frameshifted (Figure 6(c))
and replicated (data not shown) about half as well
as wild-type RNA. These results show a correlation
between frameshift level and replication, which
indicates that BYDV tolerates a reduced frameshift
rate (PAV5158-5202 and PAV5183-5205), but
once it is below a certain level, as in PAVSHSIL or
PAV5046-5100, the virus is dead.
The 30 frameshift element is required for viral
replication in plant cells
Discussion
To verify the biological relevance of the in vitro
translation results, we tested the ability of selected
30 deletion mutants to replicate in oat protoplasts.
The PPT deletion mutant, PAV5016-5045, which
frameshifted at wild-type levels in wheat germ
extracts, replicated as ef®ciently as wild-type
(PAV6) RNA (Figure 4). Thus, PPT is not necessary
for RNA replication and must have little effect on
frameshifting in vivo. The transcript containing the
Out-of-context reporter constructs may not
represent biologically relevant
translation events
We conclude that the translation properties of
full-length genomic RNA most accurately re¯ect
events in infected cells, on the basis of several
observations. First, the requirement for a stop
codon and the tolerance of complete disruption
of the shifty site suggest that the ÿ1 frame ORF
994
is expressed in the reporter assay by a mechanism different from standard ÿ1 frameshifting. It
may be a stop codon-mediated event like those
observed in Escherichia coli33 and at low levels
(0.3 %) in potato virus M RNA.34 However, both
systems required a run of at least four identical
bases, which was absent from our mutants that
disrupted the shifty heptamer but still gave GUS
activity (Table 1). The shiftiness may be aided
by the very short zero frame ORF used in the
reporter assays, in which the ribosome may still
be paused at the initiation site and not fully
locked into frame.35 Secondly, the reporter constructs frameshifted at a much lower rate than
the apparent 6-20 % that occurs in infected cells
(Figure 2). However, the immunodetection of
viral proteins in cells also may not accurately
re¯ect frameshift rate: the luminescent system is
dif®cult to quantify (Figure 2); the amount of
99 kDa fusion protein protein may be underestimated, because the antibody was raised against
the 39 kDa protein and the 99 kDa protein may
alter some 39 kDa epitopes; and the stability of
the two proteins may differ. Despite these
caveats, the immunodetection data suggest that
frameshifting probably occurs at a higher rate in
a natural infection than was measured with the
reporter constructs.
In vitro translation of full-length viral RNA
correlates with in vivo replication
The third observation that supports validity of
the full-length viral RNA in vitro translation experiments is that the effects of mutations correlate
well with their effect on replication in oat cells.
Like the reporter assays, the wheat germ extract
gave a low frameshift rate, but this varies depending on wheat germ batch, potassium concentration,24 mRNA concentration,36 and presence
of a cap (Figure 5(b)). For these reasons, we cannot
estimate the actual in vivo frameshift rate from
in vitro systems. However, wheat germ extracts
respond to other cis-acting translational control
sequences on BYDV RNA similarly in vitro and
in vivo.27,29,37 Indeed, mutations that knocked out
frameshifting, knocked out replication, and those
that reduced frameshifting only partially, such as
5183-5205, reduced replication only partially
(Figure 4; data not shown). Mutations that completely disrupted post-shift base-pairing between the
tRNAs and mRNA had no effect on reporter gene
expression (Tables 1 and 2), but ablated frameshifting in the full-length context in wheat germ
(Figure 3(b)) and prevented replication despite the
lack of any changes in predicted amino acid
sequence (SHSIL, Figure 4). Therefore, it is highly
likely that this mutation is acting by blocking frameshifting, and that frameshifting occurs by the
simultaneous slippage mechanism.38
The only mutants that did not affect frameshifting but prevented replication were those that converted the ORF 1 stop codon into a sense codon
Frameshifting Promoted by the 3 0 UTR
(Figure 3(a)). It is likely that the 4 kDa C-terminal
extension to the 39 kDa protein that results from
this mutation renders the protein non-functional.
The same mutations also had no effect on frameshifting but knocked out replication of the closely
related RCNMV.12 Moreover, all mutations that we
have introduced into the 39 kDa protein have been
lethal to replication.39
Efficient translation initiation conditions
reduce frameshift efficiency
An important trend is that conditions that favor
a high initiation rate correlate with reduced frameshifting. Addition of a 50 cap, which increased the
amount of 39 kDa protein synthesis, had no effect
on 99 kDa protein production, thus reducing the
frameshift rate (Figure 5(b)). At higher potassium
levels, 39 kDa protein synthesis decreased while
the
99 kDa
frameshift
product
remained
unchanged, giving a frameshift rate as high as
7 %24 (C.P.P., unpublished results). Thus, frameshifting may have been low in our in vitro assays
because we used conditions optimal for cap-independent translation initiation.27 Lucchesi et al36
reported that use of high mRNA concentrations
gave higher frameshift rates for cocksfoot mottle
virus RNA in wheat germ extract. These observations differ from the human T-cell leukemia
virus type II (HTLV-2) sequence, in which capping
enhanced frameshifting in reticulocyte lysates,40
but agrees with detailed observations by Lopinski
et al.,15 who also found more ef®cient frameshifting
in conditions of reduced overall translation.
In all of the above cases, with the exception of
HTLV-2, a higher number of ribosomes would be
expected on the mRNA in the conditions that give
reduced frameshifting. We and others15 speculate
that the heavy load of ribosomes will tend to keep
the structure adjacent to the frameshift site melted.
This would reduce its ability to pause ribosomes at
the shifty site. With fewer ribosomes translating
the RNA, there would be more time for the structure to re-fold and induce pausing of a higher percentage of ribosomes. Another possibility is that
only a certain absolute number (rather than proportion) of ribosomes is allowed to shift under all
conditions. Given that the 30 and 50 ends of mRNA
interact in the translation initiation complex, it is
possible that the downstream frameshift element
associates with (or modi®es) a ribosome at
initiation, and the ribosome remains associated (or
modi®ed) during elongation until it reaches the
frameshift site. Only one elongating ribosome per
mRNA could have such an association (modi®cation), which would explain the decreased frameshift rate with increased ribosomes per mRNA.
Are distant frameshift elements present in
other viral genomes?
The sometimes misleading nature of the reporter
gene assay using out-of-context viral frameshift
Frameshifting Promoted by the 3 0 UTR
sequences suggests that some frameshift studies
may be incomplete. A good correlation between
frameshifting in reporter assays and in the replicating viral genome has been shown for only a few
viruses.21 Discrepancies between out-of-context
reporter assays, and in vitro translation of intact
viral genes has been observed with HTLV23 and
HIV RNA.22 Using only out-of-context reporter
assays, Wilson et al.41 reported that the shifty heptanucleotide alone was suf®cient for HIV frameshifting, but Parkin et al.42 found that the adjacent
stem-loop structure was also necessary when intact
gag/pol mRNA was translated. Bases immediately
upstream of the shifty site enhanced HTLV-2
frameshifting.23
Immunoprecipitation of HIV gag and gag/pol
proteins expressed from a non-replicating construct
revealed an apparent frameshift rate of only
0.82 %.42 This was much lower than that observed
in rabbit reticulocyte lysates, but may have been
fraught with the inaccuracies of immunodetection
discussed by the authors.42 For many frameshifting
viruses, such as the large and complex Coronaviridae, full-length infectious clones are lacking, and
studies focus only on regions around the frameshift
site. Because of the requirement for a pseudoknot
in most cases, Farabaugh2 proposed that those
RNAs that appear to employ only a stem-loop
structure (e.g. HIV) may also need more distant
elements for complete frameshift activity. Thus, for
a variety of reasons in certain cases it may be
worth the effort to examine the potential roles of
distant viral sequences on frameshifting in vitro
and in vivo.
A long-range ribosomal frameshift signal
Alignment of the sequence with those of other
BYDV strains reveals conserved primary and
possible secondary structures, that may effect
frameshifting (Figure 7). A perfectly conserved
PuUCUGUG sequence occupies a loop in a stable
stem-loop in the essential 5046-5100 region. This
sequence has the potential to base-pair to a ®vebase bulge on the 30 side of the large bulged stemloop adjacent to the frameshift site (Figure 1(b), nt
1235-1240). Such base-pairing across kilobases
would be reminiscent of the kissing between stemloops in the 30 and 50 UTRs that facilitates capindependent translation of BYDV RNA.43 Downstream of the essential region, deletion of just 23 nt
(5183-5205) reduced frameshifting by 50 % and disrupts a different potential stem-loop (Figure 7).
Future mutagenesis will determine the roles, if
any, of these structures.
The requirement for the sequence 4 kb downstream from the frameshift site indicates a new
type of structure capable of facilitating frameshifting. Downstream elements of other frameshifting
RNAs have been limited to those beginning within
6 nt of the frameshift site, including certain
pseudoknots,10,18 less common stem-loops,12,42 or a
branched structure.44 The event reported here may
995
Figure 7. Conserved primary and secondary structures in the downstream frameshift element. Sequences
were aligned with using CLUSTAL (Lasergene Megalign
program). Except for MAV, all isolates are PAV serotype
of BYDV. Accession numbers: PAV6, X07653; Ill,
AF235167; Pur, D11032; JPN, D85783; P129, AF218798;
MAV, D11028. Sequences 13T, FHV1, FHV2 are published only by Chaloub et al.51 Secondary structures
were predicted using STAR52 and MFOLD.53 Predicted
stem-loops that were found consistently with both programs, and which are supported by phylogenetic covariation are indicated by converging arrows over
shaded (paired) bases.
roughly resemble the gene 10 frameshift signal of
bacteriophage T7. It requires a sequence in the 30
UTR 200 nt downstream of the shifty site.13 Its
mechanism of action is unknown. It is intriguing
that no distant element was reported for frameshifting of RCNMV RNA which has a bulged
stem-loop adjacent to the frameshift site12 resembling the structure predicted in BYDV RNA.26
Those authors employed reticulocyte lysates that
are shiftier than wheat germ.12
Recoding mediated by a sequence located kilobases downstream occurs during translation of
selenoproteins. UGA codons accept a novel selenocysteyl-tRNA in mRNAs containing a speci®c
bulged stem-loop (SECIS element) in the
30 UTR.45,46 Proteins that recognize the SECIS
element and selenocysteyl-tRNA carry out the process by an as yet unresolved mechanism.47 Because
the downstream BYDV element facilitates frameshifting rather than readthrough, the mechanistic
996
details would be different from the SECIS element,
but some similarities may exist in the long-distance
communication processes.
The long-distance interaction reiterates the
theme for translational control of BYDV in which
downstream sequences affect translation far
upstream.48 The 30 TE facilitates cap-independent
translation initiation 4 kb upstream,43 and a
sequence in ORF 5 facilitates read-through of the
coat protein gene stop codon 750 nt upstream.37
This reveals a complex series of interactions within
the RNA that regulate translation and possibly
replication.
Materials and Methods
Reporter plasmids
Sequences of all constructs were con®rmed by DNA
sequencing on an ABI377 automated sequencer. PCR
was performed with high-®delity Hot Tub or Vent polymerases. Mutations in the shifty heptanucleotide and
stop codon were introduced into the reporter construct
in which GUS was under the dicot CaMV 35S promoter,
pS(ÿ1)UAG,25 by two-step PCR mutagenesis.49 The primer containing the mutated sequence was used in the
®rst round with one ¯anking primer. The upstream
¯anking primer containing the NcoI site (includes start
codon) and the downstream ¯anking primer containing
a HindIII site were used for the second round of
PCR. The product was cloned into NcoI-HindIII-cut
pS(ÿ1)UAG.
To test for GUS expression in oat cells under a monocot promoter, pS(0)UCG, pS(ÿ1)UCG, pS(ÿ1)UAG,
pS(ÿ1)UAA, and pS(ÿ1)UGA were digested with NcoI
and ApaI and the fragments containing the BYDV
sequence were ligated into the large fragment of
pADHGUS(ÿAUG)50 that had been digested with the
same enzymes to make pADHS(0)UCG, pADHS(ÿ1)
UCG,
pADHS(ÿ1)UAG,
pADHS(ÿ1)UAA,
and
pADHS(ÿ1)UGA. pADHS(0)UCAG, pADHS(ÿ1)UGG,
and pADHS(ÿ1)CGG were made using two-step PCR
mutagenesis with pS (ÿ1)UAG as the template. The PCR
products were digested with NcoI and ApaI, and ligated
into pADHGUS(ÿAUG) that had been digested with the
same enzymes.
Full-length BYDV genomic clones
The stop codon mutations were introduced by PCR49
into the full-length infectious BYDV clone, pPAV6, to
make pPAVORF1UAA, pPAVORF1UGA, pPAVORF1UCG, pPAVORF1UCAG, pPAVORF1CGG, and pPAVORF1UGG. The appropriate mutagenic primer paired
with the universal forward and reverse ¯anking primers
ampli®ed a template consisting of the ClaI1044-HindIII1591
genomic fragment in pGem. The mutant PCR product
was digested with ClaI and HindIII, and cloned into the
same sites in pPAV2 (consisting of the 50 end of the
BYDV genome through the HindIII1591 site24). The resulting intermediate plasmids were digested with NotI and
HindIII, and cloned into similarly-cut pPAV6.24
pSHSIL, which is a full-length BYDV clone with the
shifty heptanucleotide changed to CGGCUUC, was con-
Frameshifting Promoted by the 3 0 UTR
structed by PCR mutagenesis of pPAV6. The upstream
primer contained the shifty heptanucleotide mutations as
well as an Eco72I site and the downstream primer contained the HindIII1591 site. The Eco72I-HindIII-cut PCR
product was inserted into the large fragment of pPAV6
after digestion with the same enzymes. The deletion
mutants are named for the bases deleted from the genome. pPAV4838-5190 is pPAV24 and pPAV37884515AUC is pPAV30 as described.39 pPAV1741-4837
was made by the digestion of pPAV6 with BamHI, followed by religation. This construct was then digested
with HindIII, the single-stranded ends ®lled in with
Klenow fragment and religated, to make pPAV17414837UAG. This created an in-frame stop codon at position 1596 that results in a truncated frameshift product.
Deletion mutants in the region around the BclI5190 site
were made by controlled exonuclease III digestion essentially as described in the Promega Erase-a-base technical
manual. pPAV6 was grown in the dam ÿ E. coli strain,
GM33. After digestion with BclI, the DNA was digested
with exonuclease III on ice and aliquots were removed at
30 second intervals. The DNA was digested with S1
nuclease for 30 minutes at room temperature. The S1
nuclease was inactivated by the addition of S1 stop buffer (0.3 M Tris-base, 0.05 M EDTA) followed by heating
the samples at 70 C for ten minutes. The samples were
then ethanol-precipitated and resuspended in TE. Singlestranded ends were ®lled in using Klenow fragment and
then ligated. Deletion mutants pPAV5180-5205,
pPAV5158-5202, and pPAV5183-5205 were recovered
after transformation of the DH5a strain of E. coli. A similar approach was used to make pPAV1684-4837 and
pPAV1684-5112. The DNA was treated as described
above for the deletions around the BclI site, except that
these plasmids were made by exonuclease III digestion
of pPAV1741-4837UAG that had been cleaved at the
BlnI1684 site and ®lled in with a-phosphorothioate dNTPs
using Klenow fragment. The DNA was then digested
with BamHI prior the treatment with exonuclease III at
30 C. This strategy resulted in a nested set of deletions
extending 30 to base 1684.
Deletions in pPAV5016-5045 and pPAV5280-5426
were created by two-step PCR mutagenesis in which
the mutagenic primer contained sequences on either
side of the region to be deleted. The ¯anking primers
contained unique restriction sites in pPAV6, BsaMI
and SmaI. After digestion of the PCR products and
pPAV6 with these enzymes, the PCR products were
inserted into pPAV6.
In vitro transcription
Transcription templates were prepared by run-off
transcription of pPAV6 or mutant derivatives cleaved
with SmaI for full-length transcripts, or transcription
from plasmids cleaved with PvuI, BclI, PstI, or ScaI to
make truncated transcripts (Figure 5(a)). The 30 overhangs created by cleavage with Pst I and PvuI were ®lled
in with Klenow fragment prior to transcription. RNA
was transcribed using phage T7 RNA polymerase with
one of the following kits: Megascript (Ambion), Ampliscribe (Epicentre Technologies), or Ribomax (Promega),
according to the manufacturer's speci®cations. To make
capped transcripts, the concentration of GTP in the reaction was reduced to one-®fth that of the other nucleoside
triphosphates and cap analog (7mG(50 )ppp(50 )G, New
England Biolabs) was included in the reaction mix at a
ratio of 4:1 (cap analog:GTP).
997
Frameshifting Promoted by the 3 0 UTR
Protoplasts and immunoblotting
Protoplasts were prepared from suspension cells, electroporated, and assayed for GUS activity as described
for carrot25 and oat50 cells. The 39 kDa and 99 kDa proteins were detected on immunoblots of plant extracts
from infected protoplasts using polyclonal antibodies
raised against 39 kDa protein that was overexpressed in
E. coli. Antibodies were raised in rabbits to the BYDV
39 kDa protein that was expressed in pET11-d in BL21
cells (Novagen). The diluted 39 kDa antiserum was
cross-absorbed with acetone-precipitated plant proteins
before incubation with membranes for immunoblots.
Viral transcripts were introduced into oat protoplasts by
electroporation as for reporter assays. The electroporation voltage was 450 V in early experiments, but later
decreased to 300 V. Protoplasts were harvested at 24
hours post-infection and stored at ÿ80 C. Just before
loading onto the gel, an equal volume of gel loading buffer was added and the protoplasts were heated in a boiling waterbath for four minutes. The lysed protoplast
extracts were then separated by SDS-PAGE (10 % (w/v)
polyacrylamide) and transferred to Hybond-ECL membrane (Amersham) by electroblotting in buffer containing
50 mM Tris, 380 mM glycine, 0.1 % (w/v) SDS, and 20 %
(v/v) methanol at 30 V overnight with cooling. Viral
proteins were detected using the Amersham ECL kit.
The blot was blocked in 5 % (w/v) powdered milk in
TBS-T(20 mM Tris-base, 137 mM NaCl, 0.1 % (v/v)
Tween-20, pH 7.6) at 4 C overnight. The blocked blot
was probed with the 39 kDa antibody for one hour and
washed with TBS-T. The secondary antibody was peroxidase-linked donkey anti-rabbit antibodies (Amersham,
NA.934) at a dilution of 1:1000 and incubation was for
one hour. The proteins were visualized using the Amersham ECL chemiluminescent detection system to expose
Hyper®lm-ECL (Amersham). The 39 kDa and 99 kDa
proteins were quanti®ed by scanning on a Hoeffer densitometer. Percentage frameshifting was calculated as the
area under the 99 kDa protein peak divided by the sum
of the areas under the 99 kDa and 39 kDa peaks and
then multiplied times 100.
In vitro translation
Transcripts were translated in wheat germ S30 extracts
(Promega) according to the manufacturer's instructions
in buffer containing 134 mM potassium acetate. Equimolar amounts of all transcripts were used. The products
were separated by denaturing SDS-PAGE (10 % polyacrylamide). The protein products were quanti®ed using
a phosphorimager and Imagequant software (Molecular
Dynamics). To correct for the number of methionine residues in each protein (28 in the 99 kDa, ten in the 39 kDa
protein), and the percentage frameshift was calculated
as ((99 kDa counts/28)/(99 kDa counts/28 ‡ 39 kDa
counts/10)) 100. All constructs were tested two or
three times. In most cases, standard error was <10 %
where PAV6 frameshift ˆ 100 %.
Acknowledgments
The authors thank Holly Lephardt for preparing
39 kDa protein antigen and antiserum, and Randy Beckett for technical assistance. The work was funded by
USDA National Research Initiative grant no. 95-37303-
1813, and NIH fellowship no. 1 F32 AI10445-01 (to
J.K.B.). This is paper 19230 of the Iowa Agriculture and
Home Economics Experiment Station, project 6542.
References
1. Gesteland, R. F. & Atkins, J. F. (1996). Recoding:
dynamic reprogramming of translation. Annu. Rev.
Biochem. 65, 741-768.
2. Farabaugh, P. J. (1997). Programmed Alternative Reading of the Genetic Code, R.G. Landes Co, Austin, TX.
3. Brierley, I. (1995). Ribosomal frameshifting on viral
RNAs. J. Gen. Virol. 76, 1885-1892.
4. Baranov, P. V., Gurvich, O. L., Fayet, O., Prere,
M. F., Miller, W. A. & Gesteland, R. F. et al. (2001).
RECODE: a database of frameshifting, bypassing
and codon rede®nition utilized for gene expression.
Nucl. Acids Res. 29, 264-267.
5. Hammell, A. B., Taylor, R. C., Peltz, S. W. &
Dinman, J. D. (1999). Identi®cation of putative programmed ÿ1 ribosomal frameshift signals in large
DNA databases. Genome Res. 9, 417-427.
6. Dinman, J. D., Ruiz-Echevarria, M. J., Czaplinski, K.
& Peltz, S. W. (1997). Peptidyl-transferase inhibitors
have antiviral properties by altering programmed ÿ1
ribosomal frameshifting ef®ciencies: development of
model systems. Proc. Natl Acad. Sci. USA, 94, 66066611.
7. Jacks, T., Madhani, H. D., Masiarz, F. R. & Varmus,
H. E. (1988). Signals for ribosomal frameshifting in
the Rous Sarcoma virus gag pol region. Cell, 55, 447458.
8. Brierley, I., Digard, P. & Inglis, S. C. (1989). Characterization of an ef®cient coronavirus ribosomal frameshifting signal: requirement for an RNA
pseudoknot. Cell, 57, 537-547.
9. Alam, S. L., Atkins, J. F. & Gesteland, R. F. (1999).
Programmed ribosomal frameshifting: much ado
about knotting!. Proc. Natl Acad. Sci. USA, 96, 1417714179.
10. Giedroc, D. P., Theimer, C. A. & Nixon, P. L. (2000).
Structure, stability and function of RNA pseudoknots involved in stimulating ribosomal frameshifting. J. Mol. Biol. 298, 167-185.
11. Bidou, L., Stahl, G., Grima, B., Liu, H., Cassan, M. &
Rousset, J. P. (1997). In vivo HIV-1 frameshifting
ef®ciency is directly related to the stability of the
stem-loop stimulatory signal. RNA, 3, 1153-1158.
12. Kim, K. H. & Lommel, S. A. (1998). Sequence
element required for ef®cient ÿ1 ribosomal frameshifting in red clover necrotic mosaic dianthovirus.
Virology, 250, 50-59.
13. Condron, B. G., Gesteland, R. F. & Atkins, J. F.
(1991). An analysis of sequences stimulatingframeshifting in the decoding of Gene-10 of bacteriophage-T7. Nucl. Acids Res. 19, 5607-5612.
14. Somogyi, P., Jenner, A. J., Brierley, I. & Inglis, S. C.
(1993). Ribosomal pausing during translation of an
RNA pseudoknot. Mol. Cell. Biol. 13, 6931-6940.
15. Lopinski, J. D., Dinman, J. D. & Bruenn, J. A. (2000).
Kinetics of ribosomal pausing during programmed
ÿ1 translational frameshifting. Mol. Cell. Biol. 20,
1095-1103.
16. Chen, X., Chamorro, M., Lee, S. I., Shen, L. X.,
Hines, J. V., Tinoco, I., Jr & Varmus, H. E. (1995).
Structural and functional studies of retroviral RNA
pseudoknots involved in ribosomal frameshifting:
nucleotides at the junction of the two stems are
998
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
important for ef®cient ribosomal frameshifting.
EMBO J. 14, 842-852.
Napthine, S., Liphardt, J., Bloys, A., Routledge, S. &
Brierley, I. (1999). The role of RNA pseudoknot stem
1 length in the promotion of ef®cient ÿ1 ribosomal
frameshifting. J. Mol. Biol. 288, 305-320.
Kim, Y. G., Su, L., Maas, S., O'Neill, A. & Rich, A.
(1999). Speci®c mutations in a viral RNA pseudoknot drastically change ribosomal frameshifting ef®ciency. Proc. Natl Acad. Sci. USA, 96, 14234-14239.
Stahl, G., Bidou, L., Hatin, I., Namy, O., Rousset,
J. P. & Farabaugh, P. (2000). The case against the
involvement of the NMD proteins in programmed
frameshifting. RNA, 6, 1687-1688.
Dinman, J., Ruiz-Echevarria, M., Wang, W. & Peltz,
S. (2000). The case for the involvement of the Upf3p
in programmed ÿ1 ribosomal frameshifting. RNA, 6,
1685-1686.
Dinman, J. D. & Wickner, R. B. (1992). Ribosomal
frameshifting ef®ciency and gag/gag-pol ratio are
critical for yeast M(1) double-stranded RNA virus
propagation. J. Virol. 66, 3669-3676.
Cassan, M., Delaunay, N., Vaquero, C. & Rousset,
J. P. (1994). Translational frameshifting at the gagpol junction of human immunode®ciency virus type
1 is not increased in infected T-lymphoid cells.
J. Virol. 68, 1501-1508.
Kollmus, H., Hongiman, A., Panet, A. & Hauser, H.
(1994). The sequences of and distance between two
cis-acting signals determine the ef®ciency of ribosomal frameshifting in human immunode®ciency virus
type 1 and human T-cell leukemia virus type II
in vivo. J. Virol. 68, 6087-6091.
Di, R., Dinesh-Kumar, S. P. & Miller, W. A. (1993).
Translational frameshifting by barley yellow dwarf
virus RNA (PAV serotype) in Escherichia coli and in
eukaryotic cell-free extracts. Mol. Plant-Microbe Interact. 6, 444-452.
Brault, V. & Miller, W. A. (1992). Translational
frameshifting mediated by a viral sequence in plant
cells. Proc. Natl Acad. Sci. USA, 89, 2262-2266.
Miller, W. A., Dinesh-Kumar, S. P. & Paul, C. P.
(1995). Luteovirus gene expression. Critic. Rev. Plant
Sci. 14, 179-211.
Wang, S. & Miller, W. A. (1995). A sequence located
4.5 to 5 kilobases from the 50 end of the barley
yellow dwarf virus (PAV) genome strongly stimulates translation of uncapped mRNA. J. Biol. Chem.
270, 13446-13452.
Callis, J., Fromm, M. & Walbot, V. (1987). Introns
increase gene expression in cultured maize cells.
Genes Dev. 1, 1183-1200.
Wang, S. P., Browning, K. S. & Miller, W. A. (1997).
A viral sequence in the 30 untranslated region
mimics a 50 cap in facilitating translation of
uncapped mRNA. EMBO J. 16, 4107-4116.
Ito, T. & Lai, M. M. (1999). An internal polypyrimidine-tract-binding protein-binding site in the hepatitis C virus RNA attenuates translation, which is
relieved by the 30 -untranslated sequence. Virology,
254, 288-296.
Witherell, G. W., Schultz-Witherell, C. S. &
Wimmer, E. (1995). Cis-acting elements of the
encephalomyocarditis virus internal ribosomal entry
site. Virology, 214, 660-663.
Li, H. P., Huang, P., Park, S. & Lai, M. M. (1999).
Polypyrimidine tract-binding protein binds to the
leader RNA of mouse hepatitis virus and serves as a
regulator of viral transcription. J. Virol. 73, 772-777.
Frameshifting Promoted by the 3 0 UTR
33. Weiss, R. B., Dunn, D. M., Atkins, J. F. & Gesteland,
R. F. (1990). Ribosomal frameshifting from ÿ2 to
‡50 Nucleotides. Prog. Nucl. Acids Res. Mol. Biol. 39,
159-183.
34. Gramstat, A., PruÈfer, D. & Rohde, W. (1994). The
nucleic acid-binding zinc ®nger protein of potato
virus M is translated by internal initiation as well as
by ribosomal frameshifting involving a shifty stop
codon and a novel mechanism of P-site slippage.
Nucl. Acids Res. 22, 3911-3917.
35. Wolin, S. L. & Walter, P. (1988). Ribosome pausing
and stacking during translation of a eukaryotic
mRNA. EMBO J. 7, 3559-3569.
36. Lucchesi, J., Makelainen, K., Merits, A., Tamm, T. &
Makinen, K. (2000). Regulation of ÿ1 ribosomal
frameshifting directed by cocksfoot mottle sobemovirus genome. Eur. J. Biochem. 267, 3523-3529.
37. Brown, C. M., Dinesh-Kumar, S. P. & Miller, W. A.
(1996). Local and distant sequences are required for
ef®cient read-through of the barley yellow dwarf
virus-PAV coat protein gene stop codon. J. Virol. 70,
5884-5892.
38. Jacks, T., Power, M. D., Masiarz, F. R., Luciw, P. A.,
Barr, P. J. & Varmus, H. E. (1988). Characterization
of ribosomal frameshifting in HIV-1 gag-pol
expression. Nature, 331, 280-283.
39. Mohan, B. R., Dinesh-Kumar, S. P. & Miller, W. A.
(1995). Genes and cis-acting sequences involved in
replication of barley yellow dwarf virus-PAV RNA.
Virology, 212, 186-195.
40. Honigman, A., Falk, H., Mador, N., Rosental, T. &
Panet, A. (1995). Translation ef®ciency of the human
T-cell leukemia virus (HTLV-2) gag gene modulates
the frequency of ribosomal frameshifting. Virology,
208, 312-318.
41. Wilson, W., Braddock, M., Adams, S. E., Rathjen,
P. D., Kingsman, S. M. & Kingsman, A. J. (1988).
HIV expression strategies: ribosomal frameshifting is
directed by a short sequence in both mammalian
and yeast systems. Cell, 55, 1159-1169.
42. Parkin, N. T., Chamorro, M. & Varmus, H. E.
(1992). Human immunode®ciency virus type-1
gag-pol frameshifting is dependent on downstream
messenger RNA secondary structure - demonstration by expression in vivo. J. Virol. 66, 5147-5151.
43. Guo, L., Allen, E. & Miller, W. A. (2001). Basepairing between untranslated regions facilitates
translation of uncapped, nonpolyadenylated viral
RNA. Mol. Cell. 7, 1103-1109.
44. Rettberg, C. C., Prere, M. F., Gesteland, R. F.,
Atkins, J. F. & Fayet, O. (1999). A three-way junction
and constituent stem-loops as the stimulator for programmed ÿ1 frameshifting in bacterial insertion
sequence IS911. J. Mol. Biol. 286, 1365-1378.
45. Berry, M. J., Banu, L., Chen, Y., Mandel, S. J.,
Kieffer, J. D., Harney, J. W. & Larsen, P. R. (1991).
Recognition of UGA as a selenocysteine codon in
type-I deiodinase requires sequences in the 30
untranslated region. Nature, 353, 273-276.
46. Low, S. C. & Berry, M. J. (1996). Knowing when not
to stop: selenocysteine incorporation in eukaryotes.
Trends Biochem Sci. 21, 203-208.
47. Atkins, J. F. & Gesteland, R. F. (2000). The twenty®rst amino acid. Nature, 407, 463-465.
48. Miller, W. A., Brown, C. M. & Wang, S. (1997). New
punctuation for the genetic code: luteovirus gene
expression. Semin. Virol. 8, 3-13.
49. Landt, O., Grunert, H. P. & Hahn, U. (1990). A general method for rapid site-directed mutagenesis
Frameshifting Promoted by the 3 0 UTR
using the polymerase chain reaction. Gene, 96, 125128.
50. Dinesh-Kumar, S. P. & Miller, W. A. (1993). Control
of start codon choice on a plant viral RNA encoding
overlapping genes. Plant Cell, 5, 679-692.
51. Chaloub, B. A., Kelly, L., Robaglia, C. & Lapierre,
H. D. (1994). Sequence variability in the genome-30 terminal region for 10 geographically distinct PAVlike isolates of barley yellow dwarf virus: analysis of
the ORF6 variation. Arch. Virol. 139, 403-416.
999
52. Gultyaev, A. P., Van Batenburg, F. H. D. & Pleij,
C. W. A. (1995). The computer simulation of RNA
folding pathways using a genetic algorithm. J. Mol.
Biol. 250, 37-51.
53. Zuker, M., Mathews, D. H. & Turner, D. H. (1999).
Algorithms and thermodynamics for RNA secondary structure prediction: a practical guide. In RNA
Biochemistry and Biotechnology (Barciszewski, J. &
Clark, B. F. C., eds), pp. 11-43, Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Edited by D. E. Draper
(Received 20 February 2001; received in revised form 17 May 2001; accepted 18 May 2001)