TRANSLATION
Mechanisms that ensure speed and fidelity in
eukaryotic translation termination
Michael R. Lawson1
†, Laura N. Lessen2,3
†, Jinfan Wang1
, Arjun Prabhakar,1
‡, Nicholas C. Corsepius1
§,
Rachel Green3,4
*, Joseph D. Puglisi1
*
Translation termination, which liberates a nascent polypeptide from the ribosome specifically at stop
codons, must occur accurately and rapidly. We established single-molecule fluorescence assays to track
the dynamics of ribosomes and two requisite release factors (eRF1 and eRF3) throughout termination
using an in vitro–reconstituted yeast translation system. We found that the two eukaryotic release
factors bound together to recognize stop codons rapidly and elicit termination through a tightly
regulated, multistep process that resembles transfer RNA selection during translation elongation.
Because the release factors are conserved from yeast to humans, the molecular events that underlie
yeast translation termination are likely broadly fundamental to eukaryotic protein synthesis.
P
rotein synthesis concludes when a translating
ribosome encounters a stop codon
at the end of an open reading frame,
triggering recruitment of two factors to
liberate the nascent polypeptide: eukaryotic
release factor 1 (eRF1), a tRNA-shaped proteinthat
decodesthe stopcodon inthe ribosomal
aminoacyl-tRNA site (A site) and cleaves the
peptidyl-tRNA bond (1–3), and eukaryotic release
factor 3 (eRF3), a GTPase that promotes eRF1
action (4–6). After translation termination, the
ribosome, peptidyl-tRNA site (P site) tRNA, and
mRNA are released by recycling (4, 7, 8). Despite
decades of study, the order and timing of the
molecular events that drive translation termination
remain unclear because multistep
processes are difficult to assess using traditional
approaches. A cohesive understanding
of translation termination and its underlying
steps that are central to normal translation
would also support the treatment of diseases
caused by premature stop codons, which include
cystic fibrosis, muscular dystrophy, and
hereditary cancers (9). Because premature stop
codons cause 11% of all heritable human
diseases (10), stop codon readthrough therapeutics
have immense clinical potential
(9, 11).
Direct tracking of release factor dynamics
Here, we used an in vitro–reconstituted yeast
translation system (12) and single-molecule
fluorescence spectroscopy to track eukaryotic
release factor dynamics and termination directly.
We reasoned that ribosomes translating
mRNAs with very short open reading frames
would provide the simplest system for detailed
analysis of the discrete substeps of termination.
Ribosome complexes were programmed
with Met (M-Stop) or Met-Phe (M-F-Stop)
mRNAs, achieved by incubation with purified
Met-tRNAMet
i, initiation factors, elongation
factors, and tRNAs (as appropriate) and then
reacted with saturating amounts of eRF1 and
eRF3 (13). Peptide release from both M-Stop
and M-F-Stop ribosome complexes occurred at
similar rates as a longer tetrapeptide (M-F-K-KStop)–programmed
ribosome complex (Fig.
1A and fig. S1, A and B) and also matched the
rate previously characterized for tripeptideprogrammed
ribosome complexes (4, 5). To
monitor eRF1 and eRF3 binding to ribosomes
in real time, we labeled both proteins specifically
with fluorescent dyes (fig. S1, C and D)
and established that the labeled proteins exhibited
wild-type peptide release activity (Fig.
1B and fig. S1, E and F). Association of eRF1
with the ribosome was monitored by Förster
resonance energy transfer (FRET) between 60S
subunits labeled with Cy3 (FRET donor) on the
C terminus of uL18 (14) and eRF1 labeled with
Cy5 (FRET acceptor) on the N terminus. Structural
models placed these termini ~50 Å apart
when eRF1 was bound in the A site (Fig. 1C)
(3, 15). Next, Cy3-labeled ribosomal complexes
programmed with Met in the P site and either
UAA or UUC in the A site were combined
with Cy5-eRF1 and unlabeled eRF3, and FRET
was monitored at equilibrium using total internal
reflection fluorescence microscopy. eRF1-
60S FRET was only observed when a stop
codon was in the A site (Fig. 1D and fig. S1, G
and H), demonstrating the specificity of the
FRET signal for proper eRF1 association mediated
by a stop codon.
We leveraged this FRET-based binding signal
to determine the roles of eRF1 and eRF3
in translation termination. We first prepared
80S ribosomal complexes programmed on 5′biotinylated
M-Stop mRNAs with Cy5-labeled
60S subunits; these complexes were tethered
to neutravidin-coated zero-mode waveguide
(ZMW) surfaces (16, 17). Upon start of realtime
data acquisition, Cy3-eRF1, excess GTP,
and unlabeled eRF3 were added to ZMWs,
and Cy3-eRF1 and Cy5-60S fluorescence within
individual ZMWs was monitored by excitation
with a 532-nm laser (Fig. 2A).
Rapid, concentration-dependent eRF1 binding
to the ribosomal A site was detected upon delivery
of the release factors (Fig. 2, B to D, and fig.
S2, A to C). Association kinetics were fit to a
double-exponential function with a dominant
(56 to 83%) eRF1 concentration–dependent
fast phase with a pseudo–second-order rate
constant of 6.3 ± 3.9 mM−1
s−1
[95% confidence
interval (CI); Fig. 2, D and E, and fig. S2B]; a
minor (17 to 44%) slow phase, which did not
vary with eRF1 concentration, was also observed
(e.g., kobs = ~0.009 s−1
; Fig. 2E and fig.
S2B). Conversely, eRF1 bound very slowly to
these same complexes in the absence of eRF3
(e.g., kobs = ~0.008 s−1
; Fig. 2, E and F, and fig.
S2, D and E); the rate constant for this eRF3independent
binding was similar to the slow
phase observed with eRF3 and also was unaffected
by eRF1 concentration. In all cases,
eRF1-binding events were long-lived (e.g., t =
227 ± 13 s; fig. S2, B and E), and prolonged
detection was likely limited by the dye photobleaching
lifetime (fig. S2F). In the presence of
eRF3, the rapid eRF1 binding observed here
was similar to the rate of Phe-tRNAPhe
ternary
complex binding to its cognate A-site codon
under similar conditions (9.0 ± 0.4 mM−1
s−1
;
fig. S2G). These results indicate that eRF1 binding,
which would otherwise be limited by a slow
event, is rapid enough to compete with tRNAs
for A-site occupancy when assisted by eRF3.
We next tracked eRF3 dynamics directly,
independently of eRF1, to establish a baseline
understanding of its interaction with the ribosome.
We used a previously established interribosomal
subunit FRET signal to confirm 80S
complex formation (14) and monitored dyelabeled
eRF3 dynamics by fluorescent bursts
that occured upon factor binding to immobilized
ribosomes. Ribosomes, Cy3 labeled on
uL18, and Cy5 labeled on uS19 (yielding FRET
upon 80S formation) were programmed with
5′-biotinylated M-Stop mRNAs and tethered to
ZMWs. Next, Cy5-eRF3 and GTP were added to
ZMWs and illuminated with 532- and 642-nm
lasers. After an initial phase of FRET, typified
by rapid Cy5-40S photobleaching, brief bursts
of additional Cy5 signal were observed that
marked binding and dissociation of eRF3
(Fig. 3, A and B). eRF3 binding was concentration
dependent (Fig. 3C), and association
kinetics were fit to an exponential function
with a pseudo–second-order rate constant of
RESEARCH
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 1 of 6
1
Department of Structural Biology, Stanford University
School of Medicine, Stanford, CA, USA. 2
Program in
Molecular Biophysics, Johns Hopkins University School of
Medicine, Baltimore, MD, USA. 3
Department of Molecular
Biology and Genetics, Johns Hopkins University School of
Medicine, Baltimore, MD, USA. 4
Howard Hughes Medical
Institute, Johns Hopkins University School of Medicine,
Baltimore, MD, USA
*Corresponding author. Email: ragreen@jhmi.edu (R.G.);
puglisi@stanford.edu (J.D.P.)
†These authors contributed equally to this work.
‡Present address: Pacific Biosciences Inc., Menlo Park, CA, USA.
§Present address: Department of Chemistry, Fresno City College,
Fresno, CA, USA.
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
0.4 ± 0.2 mM−1
s−1
(95% CI; fig. S3, A and B).
eRF3 resided briefly on the ribosome (t =
0.15 ± 0.01 s; Fig. 3D), and the dwell times
between eRF3-binding events varied with its
concentration (fig. S3B), consistent with a bimolecular
association reaction. Inclusion of
GTP analogs, GDP, or a GTPase-deficient eRF3
mutant [H348E (4)] did not markedly affect
the association or dissociation rates of eRF3
(about twofold or less; Fig. 3D and fig. S3C)
suggesting that this binding cycle occurs independently
of the eRF3 nucleotide–bound state
or GTP hydrolysis.
Two distinct models could explain how eRF3
promotes the fast association of eRF1 with ribosomes
halted at stop codons. eRF3 may first
bind to ribosomes, triggering rearrangements
that favor subsequent association of eRF1 with
ribosomes. Alternatively, eRF3 may act as a
chaperone, directly delivering eRF1 to ribosomes
(5). To distinguish between these models,
we performed single-molecule experiments
similar to those described above but now simultaneously
tracking fluorescent eRF1 and eRF3.
We observed concurrent binding of the two
factors to M-Stop ribosomes (Fig. 4, A and B).
Although we also observed eRF1 and/or eRF3
binding individually to ribosomes in these experiments
(which was unsurprising because
the release factors can each bind alone to ribosomes
and are at subsaturating concentrations),
the likelihood of such independent
binding events occurring simultaneously was
very low (<0.1%), allowing us to rule out that
co-arrivals happen primarily by random chance
(see the materials and methods). Analysis of Cy5
(eRF1) and Cy3.5 (eRF3) fluorescence intensities,
alignedtothebeginningofapparentco-association
events (“postsynchronization”), further demonstrated
that eRF3 binding to a stop codon–
halted ribosome is transient, whereas eRF1
resides longer on the ribosome (Fig. 4C).
Omission of GTP decreased the number of observed
co-binding events by 17-fold, confirming
that eRF1, eRF3, and GTP bind the ribosome
together as a preformed ternary complex
(fig. S4A). Ternary complex association kinetics
were fit to a double-exponential function,
yielding a dominant fast-phase rate that was
dependent upon eRF1 concentration and a
pseudo–second-order rate constant of 2.6 ±
5.1 mM−1
s−1
(95% CI; Fig. 4D and fig. S4, B to D).
In contrast to the dynamics of eRF3 in absence
of eRF1 (where eRF3 lifetime had little
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 2 of 6
Fig. 1. Bulk biochemical and single-molecule studies of termination.
(A) Peptides are liberated at similar rates from ribosomes translating a variety
of model mRNAs. (B) Wild-type and labeled release factors liberate peptides
from ribosomes. Catalytically dead eRF1 (orange) is inactive. (C) Structural
modeling [Protein Data Bank (PDB) ID: 5LZT (15)] suggests that labeled eRF1
(green, Cy5 labeled at red star) would FRET with ribosomes (red, Cy3 labeled at
green star) upon binding to the A site. (D) Example of FRET observed with
Cy3-eRF1 and Cy5-60S by total internal reflection fluorescence microscopy.
Fig. 2. eRF3 promotes fast binding of
eRF1 to ribosomes halted at stop
codons. (A) Experimental setup.
(B) Assay schematic. (C) Example of fast
binding of eRF1 (Cy3, green) to M-Stop
ribosomes (Cy5, red) observed in the
presence of eRF3. (D) Binding of eRF1
to M-Stop ribosomes is fast and concentration
dependent in the presence of
eRF3. Association time distributions were
fit to a double-exponential model.
(E) Observed rates of eRF1 binding to
M-Stop ribosomes (kobs) with and without
eRF3. The plateau observed with the
eRF1/eRF3/GTP fast phase coincides
with the rate of sample mixing in ZMWs
(17). (F) Binding of eRF1 to M-Stop
ribosomes is slow and eRF1 concentration
independent without eRF3. Association
time distributions were fit to an
exponential model.
RESEARCH | RESEARCH ARTICLE
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
dependence on GTP hydrolysis and ranged
from 0.11 to 0.16 s; Fig. 3D), here, GTP accelerated
eRF3 release from ribosomes by eightfold
compared with experiments performed
with the more slowly hydrolyzed analog
GTPgS (0.3 ± 0.1 s with GTP versus 2.5 ± 0.1 s
with GTPgS; Fig. 4E and fig. S4, E and F).
Substitution of wild-type eRF3 with a GTPasedeficient
mutant similarly slowed its release
from the ribosome by fivefold (1.6 ± 0.1 s; Fig.
4E and fig. S4, E and F). Thus, eRF3 is a
chaperone that delivers eRF1 to ribosomes
halted at stop codons, and eRF3 departure
from the ribosome is partly governed by its
GTPase activity.
Real-time monitoring of translation
termination
We next sought to understand the timing
and regulation of peptidyl-tRNA ester bond
hydrolysis catalyzed by eRF1. Peptide hydrolysis
triggers rapid rearrangement of P-site
tRNA from a classical (P/P) to a hybrid (P/E)
state (18), and we hypothesized this rearrangement
could be tracked using FRET between
labeled P-site–bound tRNA and A-site–bound
eRF1 (~34 Å separation before versus ~50 Å
after rearrangement; Fig. 5A). To test this,
we tethered to ZMWs ribosomes programmed
on an M-Stop mRNA with Cy3-labeled Met-
tRNAMet
i (FRET donor) in the P site; added
catalytically inactive Cy5-eRF1 (G180A, FRET
acceptor), unlabeled eRF3, and GTP; and illuminated
with a 532-nm laser. As expected,
high FRET was observed between the classical
state tRNA and eRF1 (m = 0.63, s = 0.10;
Fig. 5B). Next, we repeated the assay but
added puromycin, a drug that cleaves the
peptidyl-tRNA bond, and indeed observed
lower-efficiency FRET between the now hybridstate
tRNA and catalytically inactive eRF1 (m =
0.53, s = 0.10; Fig. 5B). Therefore, peptidyltRNA
bond status can be deduced by monitoring
P-site tRNA conformation with respect to
eRF1 through this FRET signal.
We used this FRET signal to correlate eRF1
dynamics with peptidyl-tRNA bond hydrolysis
in real time. Ribosomes programmed on
5′-biotinylated M-F-Stop mRNAs were tethered
to ZMWs and illuminated with a 532-nm laser.
We then added a mixture of Cy3-labeled Phe-
tRNAPhe
(FRET donor), Cy5-eRF1 (FRET acceptor),
excess eRF3, elongation factors, and eIF5A
[an accessory factor that accelerates elongation
and termination (19)]. Fast tRNA binding,
denoted by high Cy3 signal, was observed
soon after factor addition and persisted until
translocation and subsequent eRF1 binding
occurred (Fig. 5, C and D, and fig. S5A). Association
of eRF1 initially resulted in a highFRET
signal (m = 0.67, s = 0.07; herein referred
to as the “pretermination state”; fig. S5, B
and C) and was followed by a lower-FRET
signal (m = 0.46, s = 0.11; “post-termination
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 3 of 6
Fig. 3. Observing eRF3 dynamics in ZMWs. (A) Assay schematic. (B) Example of eRF3 binding to
M-Stop ribosomes. (C) Binding of eRF3 to M-Stop ribosomes is concentration dependent. Association
time distributions were fit to an exponential model. (D) GTP hydrolysis by eRF3 is not required for its release
from the ribosome in the absence of eRF1.
Fig. 4. eRF3 delivers eRF1 quickly to ribosomes halted at stop codons. (A) Assay schematic.
(B) Example of simultaneous binding of eRF1 (Cy5, red) and eRF3 (Cy3.5, yellow) to M-Stop ribosomes (Cy3,
green). (C) Postsynchronization plot of fluorescence changes observed upon simultaneous binding of
eRF1 and eRF3 (dashed, black vertical line). (D) Simultaneous binding of eRF1 and eRF3 to M-Stop ribosomes
is fast and concentration dependent. Association time distributions were fit to a double-exponential model.
(E) GTP hydrolysis by eRF3 accelerates its release from the ribosome in the presence of eRF1.
RESEARCH | RESEARCH ARTICLE
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
state”; Fig. 5, C to E, and fig. S5B). Finally,
eRF1 was released from the ribosome, resulting
in restoration of high Cy3 signal (Fig. 5, C
to E). Critically, substitution of eRF1 with an
inactive mutant reduced the number of observed
high- to lower-FRET transitions by
about sevenfold (fig. S5, D and E), which
closely matched the relative extent of peptide
release observed in bulk with either wildtype
or G180A eRF1 (Fig. 1B and fig. S1, E and
F). P-site tRNA/eRF1 FRET is also specific for
stop codon recognition, because replacement
of the UAA stop codon with near-cognate UAU
completely eliminated these FRET transitions
(fig. S5D).
We used this tRNA/eRF1 FRET assay to characterize
the kinetics of peptidyl-tRNA bond
hydrolysis, focusing on ribosomal events before
and after termination. Pretermination state
lifetimes of eRF1 on the 80S ribosome fit well
to a two-step, irreversible kinetic model with
termination occurring in 2.8 s at 30°C (95%
CI: 2.6 to 3.0 s; Fig. 5F and fig. S5, F and G).
eRF1 dissociation kinetics, fit to a single-step
model by a single exponential, revealed that
eRF1 is released quickly after termination in
0.5 s (95% CI: 0.5 to 0.6 s, Fig. 5F and fig. S5, G
and H); eRF1 lifetime was unaffected by laser
power variation, showing that its lifetime is
not limited by dye photobleaching; fig. S5I).
Further support that peptidyl-tRNA bond
hydrolysis favors eRF1 release was observed
with catalytically inactive eRF1, which resided
sixfold longer on ribosomes (fig. S5J). PeptidyltRNA
bond cleavage also hindered rebinding
of eRF1 (fig. S5K), demonstrating that termination
decreases the affinity of eRF1 for ribosomes.
The long eRF1 lifetime observed in our
prior measurements (Figs. 2 and 4) is attributable
to differences in detection methods and the
omission of eIF5A from those assays, because
eIF5A does not affect eRF1 association rates but
does enhance termination and eRF1 release
rates (Fig. 5G and fig. S6) (19). Peptidyl-tRNA
bond hydrolysis rates increased with temperature
from 20 to 30°C (Fig. 5F and figs. S5G and
S7, A to C), and subsequent Eyring and Gibbs
analyses revealed that termination is regulated
by a step with a large energetic barrier (fig.
S7D); consistent with this notion, previous
structural work demonstrated that eRF1 undergoes
a large-scale conformational change
(referred to as “accommodation”) to render it
catalytically active (3, 15), and the tRNA rearrangement
tracked by this assay would also
depend upon transition of the ribosome from
a classical to a hybrid state (20). Release of
eRF1 from the ribosomal A site is also energetically
costly (Fig. 5F and fig. S7E), likely
because of extensive interactions that anchor
eRF1 to the stop codon (3). Termination proceeded
at a similar rate when the UAA stop
codon was changed to either UAG or UGA
(fig. S8), suggesting that a common mechanism
is used at all three stop codons. We
therefore observed an ordered series of events
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 4 of 6
Fig. 5. Tracking peptidyl-tRNA bond
hydrolysis. (A) Model of FRET between
eRF1 (Cy5, red star) and P-site tRNA
(Cy3, green star) before and after
termination. Left, pretermination
[modeled by 40S alignment;
PDB IDs: 5LZU and 5LZT (3, 15)]. Right,
post-termination [modeled by 40S
alignment; PDB IDs: 5LZU and 3J77
(15, 20)]. (B) FRET observed between
P-site Cy3-Met-tRNAi
and G180A
Cy5-eRF1 (black). The addition of
puromycin (red) yielded lower FRET, as
expected. (C) Assay schematic.
(D) Example of FRET observed with
Cy3-Phe (green) and Cy5-eRF1 (red).
(E) Postsynchronization plot of FRET
efficiency observed before and after
peptidyl-tRNA bond hydrolysis (dashed,
black vertical line). (F) Pre- and posttermination
state lifetimes observed
at different temperatures.
(G) Dependence of eRF1 lifetime on
eIF5A and detection methods. eRF1 lifetimes
observed using tRNA/eRF1 FRET
either represent all binding events (all)
or only those resulting in termination
(iii → iv).
RESEARCH | RESEARCH ARTICLE
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
at stop codons, with eRF1 eliciting peptidyltRNA
bond cleavage within ~3 s of ribosomal
association, followed by rapid eRF1 release.
Stop codon readthrough effectors slow
termination dynamics
We next explored the impact of cis-acting mRNA
elementsontermination. Prior workuncovered a
class of 3′ untranslated region (UTR) mRNA
sequences that promote stop codon readthrough
(21, 22), but it is unclear whether these elements
function in part by inhibiting termination. Indeed,
insertion of a six-nucleotide sequence that
promotes stop codon readthrough with 30%
efficiency (CAAUUA) into the 3′ UTR of M-FStop
mRNAs lengthened pretermination state
duration by twofold (5.9 s, 95% CI: 5.2 to 7.0 s
with CAAUUA versus 2.9 s, 95% CI: 2.7 to 3.0 s
without CAAUUA; Fig. 6A and fig. S9, A and B).
Insertion of other sequences that promote readthrough
at lower efficiencies also lengthened the
pretermination state, albeit to a lesser degree
(Fig. 6A and fig. S9, A and B). Thus, sequences
that enhance stop codon readthrough hinder
peptidyl-tRNA bond cleavage by eRF1.
The aminoglycosides paromomycin and G418
also promote stop codon readthrough (23), primarily
by stabilizing near-cognate tRNA in the
A site. Paromomycin also inhibits bacterial termination
(24), but its effects on eukaryotic termination
were not deeply characterized (25).
G418 was recently identified as a eukaryotic
termination inhibitor by bulk biochemical
studies with a mixed mammalian system
(26); however, its mode of action remains unclear.
We therefore applied our suite of assays
to determine the effects of these drugs on
eukaryotic termination. The addition of 1 mM
G418 [median effective concentration (EC50)
of ~2 mM (27)] lengthened the pre- and posttermination
state duration (each by threefold;
Fig. 6B and fig. S9, C to E). Similarly, 50 mM
paromomycin [EC50 of ~35 mM (27)] increased
the pretermination state duration (by nearly
twofold; Fig. 6C and fig. S9, C to E) and
lengthened post-termination state duration (by
fivefold; Fig. 6C and fig. S9, C to E). Titration of
paromomycin revealed concentration-dependent
effects of the drug on post-termination state
duration (Fig. 6C and fig. S9, C to E). Simultaneous
tracking of eRF1/eRF3 dynamics (as
described in Fig. 4) revealed that 50 mM paromomycin
slowed eRF1/eRF3 co-binding to the
ribosome (by more than twofold; fig. S9F). Together,
these studies demonstrate that stop
codon readthrough effectors hinder numerous
facets of termination, thus uncovering additional
nodes to target with potential therapeutics.
Discussion
Whereas prior work broadly described the
roles of eRF1 and eRF3 in regulating eukaryotic
termination (2, 4, 5, 19, 28), here, we
directly monitored the kinetics of individual
substeps to obtain a higher-resolution
view of this essential process (Fig. 6D). First,
a prebound ternary complex of eRF1, eRF3,
and GTP rapidly binds to a ribosome halted
at a stop codon (Figs. 2 and 4). eRF3 appears
to unlock eRF1 conformation to facilitate fast
ribosomal binding, because the association of
eRF1 alone is slow and governed by an eRF1
concentration–independent event (Fig. 2, D
and E). Consistent with this notion, prior
structural studies demonstrated that the
predominant conformation of eRF1 free in
solution is incompatible with ribosomal binding
(fig. S10) (3, 15, 29, 30). Next, eRF3 hydrolyzes
GTP to promote its own release (Fig. 4E),
which permits the rearrangement of eRF1 to an
active conformation (15). Accommodated eRF1
then rapidly cleaves the peptidyl-tRNA bond,
triggering ribosomal intersubunit rotation,
movement of the deacylated P-site tRNA to
a P/E hybrid state, and ejection of both eRF1
and the liberated peptide (Fig. 5). Indeed,
eRF1 release was slowed by small molecules
that hinder ribosomal rotation (G418 and
paromomycin; Fig. 6, B and C) (27), which
further shows the interdependence of these
events. Direct tracking of peptidyl-tRNA bond
hydrolysis further uncovered how small molecules
and mRNA sequences inhibit discrete
steps in termination to promote stop codon
readthrough (Fig. 6, A to C). With the critical
caveat that termination may also be influenced
by unidentified nascent chain dynamics and
other trans-acting factors, we propose that the
termination mechanisms described here are
fundamental to eukaryotic translation, because
the release factors are widely conserved from
yeast to humans (fig. S11) (1, 6).
To assess the physiological relevance of our
in vitro results, we compared them with
ribosome-profiling measurements that report
A-site occupancy using the ratio of short
(21 nucleotides: empty A site) versus long
(28 nucleotides: occupied A site) ribosomeprotected
footprints (RPFs). Prior studies
suggested that binding of eRF1 to ribosomes
is not rate limiting for termination, because
long RPFs substantially outnumber short RPFs
at stop codons (31). We confirmed this hypothesis
through direct observation of termination
substeps (Fig. 6D), which demonstrated that
release factor binding is indeed faster than
subsequent ribosomal events. The finding that
termination (~4 s) is fast relative to initiation
[~20 to 60 s (32, 33)] but somewhat slower
than elongation [0.05 to 1.4 s per codon (33)]
suggests the existence of an intricate choreography
that prevents the accumulation of
ribosomes at stop codons. Consistent with this,
ribosomal profiling in eRF1-depleted cells revealed
a marked increase in queueing of ribosomes
at stop codons (31).
Eukaryotic termination differs substantially
from the mechanisms described previously
for bacterial termination, in which the bacterial
namesake of eRF3, RF3, drives the departure of
eRF1-like factors (RF1/2) from ribosomes (34).
Instead, eukaryotic termination more closely
resembles translation elongation (5), in which
bacterial EF-Tu and eukaryotic eEF1A assist in
the selection of proper tRNAs through a tightly
regulated, multistep process. In eukaryotic termination,
eRF3 (itself an EF-Tu/eEF1A homolog)
in complex with GTP quickly delivers eRF1
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 5 of 6
Fig. 6. Termination is regulated by release factors, 3′ UTR mRNA sequence, and small molecules.
(A) Insertion of 3′ UTR mRNA sequences known to promote stop codon readthrough (“weak,” CAAAGA, 10%
efficiency; “medium,” CAAUCA, 20% efficiency; and “strong,” CAAUUA, 30% efficiency) hinder termination;
eRF1 release is unaffected. (B and C) The aminoglycosides G418 (B) and paromomycin (C) slow termination
and eRF1 release. (D) Order and timing of events in eukaryotic termination.
RESEARCH | RESEARCH ARTICLE
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
(a tRNA-shaped protein) to a stop codon in the
A site (Fig. 4, A and B), similar to the rapid
association of eEF1A/tRNA/GTP ternary complex
with a sense codon in the A site. eRF3
hydrolyzes GTP to promote its release from
the ribosome and facilitate eRF1 accommodation
(Fig. 4E) (15), just as EF-Tu and eEF1A
hydrolyze GTP to accelerate their ejection
and favor tRNA accommodation (35, 36). Thus,
the similarities between elongation and eukaryotic
termination are not limited to factor
architecture, but also include the molecular
choreography of these processes.
The fidelity of translation elongation is driven
in part by kinetic proofreading, in which EF-Tu/
eEF1A preferentially rejects noncognate tRNAs
in two sequential steps to boost overall accuracy
(35–37). Although the basis of termination
fidelity is unknown, we consider kinetic proofreading
a plausible model. eRF3 is essential
for termination fidelity, because its inclusion
boosts specificity by 2600-fold (28). Here, we
show that eRF3 conformationally unlocks and
delivers eRF1 to ribosomes (Figs. 2 and 4) and
facilitates eRF1 accommodation in an eRF3
GTPase–dependent manner (Fig. 4), thus providing
eRF3 with multiple opportunities to
favor genuine stop codons. Further study of
termination substep kinetics at cognate and
near-cognate stop codons will reveal whether
proofreading governs eukaryotic termination
fidelity.
Mutations that introduce a premature stop
codon pose a distinct challenge for therapeutic
intervention. These mutations trigger premature
termination, liberating an incomplete
polypeptide, and the defective mRNAs are further
degraded by nonsense-mediated decay
(NMD) (38). To achieve effective therapeutic
readthrough of premature stop codons, elongation,
termination, and NMD must all be
carefully tuned to avoid widespread misregulation
of gene expression while still eliciting
enough readthrough to alleviate disease. Thus,
termination and NMD inhibitors may prove
most useful as adjuvants, lengthening the kinetic
window for drug-mediated readthrough of
premature stop codons. Extension of these
single-molecule assays to monitor stop codon
readthrough and NMD will provide the quantitative
tools necessary to evaluate combination
therapies, paving the way to effective
treatments.
REFERENCES AND NOTES
1. L. Frolova et al., Nature 372, 701–703 (1994).
2. E. Z. Alkalaeva, A. V. Pisarev, L. Y. Frolova, L. L. Kisselev,
T. V. Pestova, Cell 125, 1125–1136 (2006).
3. A. Brown, S. Shao, J. Murray, R. S. Hegde, V. Ramakrishnan,
Nature 524, 493–496 (2015).
4. C. J. Shoemaker, R. Green, Proc. Natl. Acad. Sci. U.S.A. 108,
E1392–E1398 (2011).
5. D. E. Eyler, K. A. Wehner, R. Green, J. Biol. Chem. 288,
29530–29538 (2013).
6. G. Zhouravleva et al., EMBO J. 14, 4065–4072 (1995).
7. A. Heuer et al., Nat. Struct. Mol. Biol. 24, 453–460
(2017).
8. T. Becker et al., Nature 482, 501–506 (2012).
9. K. M. Keeling, X. Xue, G. Gunn, D. M. Bedwell, Annu. Rev.
Genomics Hum. Genet. 15, 371–394 (2014).
10. M. Mort, D. Ivanov, D. N. Cooper, N. A. Chuzhanova,
Hum. Mutat. 29, 1037–1047 (2008).
11. R. Bordeira-Carriço, A. P. Pêgo, M. Santos, C. Oliveira,
Trends Mol. Med. 18, 667–678 (2012).
12. M. G. Acker, S. E. Kolitz, S. F. Mitchell, J. S. Nanda, J. R. Lorsch,
Methods Enzymol. 430, 111–145 (2007).
13. Methods and materials are available as supplementary online
materials.
14. J. Wang et al., Nature 573, 605–608 (2019).
15. S. Shao et al., Cell 167, 1229–1240.e15 (2016).
16. S. Uemura et al., Nature 464, 1012–1017 (2010).
17. J. Chen et al., Proc. Natl. Acad. Sci. U.S.A. 111, 664–669
(2014).
18. J. Flis et al., Cell Rep. 25, 2676–2688.e7 (2018).
19. A. P. Schuller, C. C. Wu, T. E. Dever, A. R. Buskirk, R. Green,
Mol. Cell 66, 194–205.e5 (2017).
20. E. Svidritskiy, A. F. Brilot, C. S. Koh, N. Grigorieff,
A. A. Korostelev, Structure 22, 1210–1218 (2014).
21. O. Namy, I. Hatin, J. P. Rousset, EMBO Rep. 2, 787–793
(2001).
22. O. Namy et al., Nucleic Acids Res. 31, 2289–2296 (2003).
23. J. R. Wangen, R. Green, eLife 9, e52611 (2020).
24. E. M. Youngman, L. Cochella, J. L. Brunelle, S. He, R. Green,
Cold Spring Harb. Symp. Quant. Biol. 71, 545–549 (2006).
25. D. E. Eyler, R. Green, RNA 17, 925–932 (2011).
26. D. Susorov, S. Egri, A. A. Korostelev, RNA 26, 2044–2050 (2020).
27. I. Prokhorova et al., Proc. Natl. Acad. Sci. U.S.A. 114,
E10899–E10908 (2017).
28. G. Indrisiunaite, Dissertation, Faculty of Science and
Technology, Uppsala University (2019).
29. H. Song et al., Cell 100, 311–321 (2000).
30. Z. Cheng et al., Genes Dev. 23, 1106–1118 (2009).
31. C. C. Wu, B. Zinshteyn, K. A. Wehner, R. Green, Mol. Cell 73,
959–970.e5 (2019).
32. P. Shah, Y. Ding, M. Niemczyk, G. Kudla, J. B. Plotkin, Cell 153,
1589–1601 (2013).
33. D. Chu et al., EMBO J. 33, 21–34 (2014).
34. D. V. Freistroffer, M. Y. Pavlov, J. MacDougall, R. H. Buckingham,
M. Ehrenberg, EMBO J. 16, 4126–4133 (1997).
35. T. Pape, W. Wintermeyer, M. V. Rodnina, EMBO J. 17,
7490–7497 (1998).
36. T. Pape, W. Wintermeyer, M. Rodnina, EMBO J. 18, 3800–3807
(1999).
37. A. B. Loveland, G. Demo, A. A. Korostelev, Nature 584,
640–645 (2020).
38. F. He, A. Jacobson, Mol. Cell. Biol. 21, 1515–1530 (2001).
39. J. Yin et al., Proc. Natl. Acad. Sci. U.S.A. 102, 15815–15820
(2005).
40. E. Gutierrez et al., Mol. Cell 51, 35–45 (2013).
41. C. J. Shoemaker, D. E. Eyler, R. Green, Science 330, 369–372
(2010).
42. P. Tesina et al., EMBO J. 39, e103365 (2020).
43. J. Wang et al., Nat. Commun. 11, 5003 (2020).
44. V. A. Mitkevich et al., Nucleic Acids Res. 34, 3947–3954
(2006).
45. A. V. Kononenko et al., Nucleic Acids Res. 38, 548–558 (2010).
46. B. Ho, A. Baryshnikova, G. W. Brown, Cell Syst. 6, 192–205.e3
(2018).
47. P. Jorgensen, J. L. Nishikawa, B. J. Breitkreutz, M. Tyers,
Science 297, 395–400 (2002).
48. A. G. Johnson et al., RNA 25, 881–895 (2019).
49. O. V. Korobov, Chemical Kinetics with Mathcad and Maple
(Springer, 2011).
50. A. Petrov, R. Grosely, J. Chen, S. E. O’Leary, J. D. Puglisi,
Mol. Cell 62, 92–103 (2016).
51. R. A. Marshall, M. Dorywalska, J. D. Puglisi, Proc. Natl. Acad.
Sci. U.S.A. 105, 15364–15369 (2008).
52. C. E. Aitken, J. D. Puglisi, Nat. Struct. Mol. Biol. 17, 793–800 (2010).
53. M. F. Juette et al., Nat. Methods 13, 341–344 (2016).
54. J. E. Bronson, J. Fei, J. M. Hofman, R. L. Gonzalez Jr.,
C. H. Wiggins, Biophys. J. 97, 3196–3205 (2009).
55. J. Choi, J. D. Puglisi, Proc. Natl. Acad. Sci. U.S.A. 114,
13691–13696 (2017).
56. H. S. Steinert, J. Rinnenthal, H. Schwalbe, Biophys. J. 102,
2564–2574 (2012).
57. J. Rinnenthal, B. Klinkert, F. Narberhaus, H. Schwalbe, Nucleic
Acids Res. 38, 3834–3847 (2010).
58. C. Magis et al., Methods Mol. Biol. 1079, 117–129 (2014).
59. E. F. Pettersen et al., Protein Sci. 30, 70–82 (2021).
60. P. Saini, D. E. Eyler, R. Green, T. E. Dever, Nature 459, 118–121
(2009).
61. M. H. Park, H. L. Cooper, J. E. Folk, Proc. Natl. Acad. Sci. U.S.A.
78, 2869–2873 (1981).
ACKNOWLEDGMENTS
We thank C. Lapointe, C. Preston, and A. Gilliam Valentic for
editing assistance and members of the Puglisi and Green
laboratories for discussions and input. Funding: This work was
supported by the A.P. Giannini Foundation (postdoctoral fellowship
to M.R.L.), the Cystic Fibrosis Foundation (grant PUGLISI20GO to
J.D.P.), and the National Institute of General Medical Sciences
(grant R37GM059425 to R.G. and grants R01GM113078 and
R01GM51266 to J.D.P.). Author contributions: M.R.L., L.N.L.,
R.G., and J.D.P. conceived the project. M.R.L., L.N.L., and J.W.
generated reagents with guidance from R.G. and J.D.P. L.N.L.
performed and analyzed bulk peptide release experiments with
guidance from R.G. M.R.L. performed and analyzed all singlemolecule
experiments (except those for fig. S2G, which were
done by J.W.) with guidance from J.W., A.P., and J.D.P. N.C.C.
contributed custom single-molecule data analysis scripts.
M.R.L. and L.N.L. prepared the figures and drafted the initial
manuscript, which was revised with input from J.W., A.P., R.G.,
and J.D.P. All authors read and approved the final manuscript.
Competing interests: The authors declare no competing
interests. Data and materials availability: All data are available
in the main text or the supplementary materials. Relevant
expression constructs are available upon request.
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/373/6557/876/suppl/DC1
Materials and Methods
Figs S1 to S11
Table S1
References (39–61)
MDAR Reproducibility Checklist
View/request a protocol for this paper from Bio-protocol.
2 April 2021; accepted 7 July 2021
10.1126/science.abi7801
Lawson et al., Science 373, 876–882 (2021) 20 August 2021 6 of 6
RESEARCH | RESEARCH ARTICLE
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021
Use of think article is subject to the Terms of service
Science (ISSN ) is published by the American Association for the Advancement of Science. 1200 New York Avenue NW, Washington, DC
20005. The title Science is a registered trademark of AAAS.
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim
to original U.S. Government Works
Mechanisms that ensure speed and fidelity in eukaryotic translation termination
Michael R. LawsonLaura N. LessenJinfan WangArjun PrabhakarNicholas C. CorsepiusRachel GreenJoseph D. Puglisi
Science, 373 (6557), • DOI: 10.1126/science.abi7801
How translation stops
Protein synthesis concludes when a ribosome encounters a stop codon in a transcript, which triggers the recruitment
of highly conserved release factors to liberate the protein product. Lawson et al. used traditional biochemical methods
and single-molecule fluorescence assays to track the interplay of release factors with ribosomes and reveal the
molecular choreography of termination. They identified two distinct classes of effectors, small molecules and mRNA
sequences, that directly inhibited the release factors and promoted stop codon readthrough. These findings may
buttress ongoing efforts to treat diseases caused by premature stop codons, which cause 11% of all heritable human
diseases. —DJ
View the article online
https://www.science.org/doi/10.1126/science.abi7801
Permissions
https://www.science.org/help/reprints-and-permissions
Downloadedfromhttps://www.science.orgatMasarykUniversityonNovember14,2021