LETTER TO THE EDITOR
Transcript selection and the recruitment of mRNA
decay factors for NMD in Saccharomyces cerevisiae
MICHAEL R. CULBERTSON and ERIC NEENO-ECKWALL
Laboratories of Genetics and Molecular Biology, University of Wisconsin, Madison, Wisconsin 53706, USA
In Saccharomyces cerevisiae, nonsense-mediated mRNA decay (NMD) requires Upf1p, Upf2p, and Upf3p to accelerate the
decay rate of two unique classes of transcripts: (1) nonsense mRNAs that arise through errors in gene expression, and (2)
naturally occurring transcripts that lack coding errors but have built-in features that target them for accelerated decay (error-
free mRNAs). NMD can trigger decay during any round of translation and can target Cbc-bound or eIF-4E-bound transcripts.
Extremely low concentrations of the Upf proteins relative to the total pool of transcripts make it difficult to understand how
nonsense transcripts are selectively recruited. To stimulate debate, we propose two alternative mechanisms for selecting
nonsense transcripts for NMD and for assembling components of the surveillance complex, one for the first (pioneer) round
of translation, called ‘‘nuclear marking,’’ and the other for subsequent rounds, called ‘‘reverse assembly.’’ The model is designed
to accommodate (1) the low abundance of NMD factors, (2) the role of nucleocytoplasmic shuttling proteins in NMD, (3) the
independent and nonobligate order of assembly of two different subcomplexes of NMD factors, and (4) the ability of NMD to
simultaneously reduce or eliminate the synthesis of truncated proteins produced by nonsense transcripts while down-regulating
but not completely eliminating functional proteins produced from error-free NMD-sensitive transcripts.
Keywords: translation; RNA transcript degradation; RNA binding proteins; messenger ribonucleoprotein; nucleocytoplasmic
NMD in yeast and mammals
A great deal of attention has been paid to the question of
where the steps in nonsense-mediated mRNA decay (NMD)
take place: in the nucleus, during export to the cytoplasm, or
in the cytoplasm. In mammals, there is good evidence that
nonsense transcripts are degraded by NMD during the first
round of translation, called the pioneer round, when they are
bound to the CBC20/CBC80 (Cbc) cap binding complex and
the exon junction complex (EJC), both of which associate
with transcripts in the nucleus. In contrast, nonsense tran-
scripts undergoing bulk translation in the cytoplasm that are
bound to the cytoplasmic cap binding protein eIF-4E are
reportedly immune to NMD or, at the very least, are not
targeted by a mechanism directly related to the EJC (Ishigaki
et al. 2001). These results suggest that the major mechanism
for NMD in mammals occurs during the first round of
translation, which could take place concomitantly with
mRNA export from the nucleus.
The ideas presented below were initially prompted by a
recent report in RNA (Kuperwasser et al. 2004) claiming
important differences in the mechanism of NMD between
mammals and the yeast Saccharomyces cerevisiae. The
authors presented evidence that nonsense transcripts are
insensitive to NMD when restricted to the nucleus, due to
a block in nuclear export. In addition, analysis of a non-
sense reporter transcript indicated that it was insensitive to
NMD when bound to the nuclear cap binding complex
(Cbc1p/Cbc2p). They concluded that nonsense transcripts
are not degraded directly in the nucleus and that there is no
special pioneer round of translation in yeast. NMD must
therefore occur during bulk translation of eIF-4E-bound
transcripts in the cytoplasm.
These conclusions were quickly challenged (Gao et al.
2005). Examining six NMD-sensitive transcripts, those
authors found that all of them were susceptible to NMD
regardless of whether they were bound to Cbc- or eIF-4E
cap binding complexes. This evidence supports the exis-
tence of a special pioneer round of translation. Further-
more, the results are compatible with other evidence
showing that NMD can occur during any round of transla-
tion (Maderazo et al. 2003; Keeling et al. 2004). The finding
that the Upf proteins associate with polyribosomes and are
not restricted to 80S monosomes (Atkin et al. 1997) is
Reprint requests to: Michael R. Culbertson, Laboratories of Genetics
and Molecular Biology, University of Wisconsin, Madison, WI 53706,
USA; e-mail: email@example.com; fax (608) 262-4570.
Article published online ahead of print. Article and publication date are
RNA (2005), 11:1333–1339. Published by Cold Spring Harbor Laboratory Press. Copyright ª 2005 RNA Society.
another indication that NMD is not limited to Cbc-bound
transcripts. Considering all of these observations, the most
reasonable premise is that yeast lacks a Upf-dependent
mechanism for degrading nonsense transcripts trapped in
the nucleus but retains a special pioneer round of transla-
tion. Nonsense transcripts can be degraded during the
pioneer round or during bulk cytoplasmic translation.
Recruitment of the factors required for NMD in yeast
An unanswered question remains, however, as to the
mechanism by which appropriate transcripts are selectively
recruited for NMD. Recruitment for the pioneer round could
depend on Upf3p (Shirley et al. 2002) or Hrp1p (Gonzalez et
al. 2000). Both proteins shuttle between the nucleus and the
cytoplasm. Upf3p has been studied extensively, and there is
no doubt that it is required for NMD. Hrp1p binds to AU-
rich RNAs, is essential for growth, and performs functions
outside of NMD. A role in NMD was suggested by the finding
that it interacts with Upf1p. Also, a temperature-sensitive
hrp1 allele stabilized a nonsense reporter transcript in a tem-
perature-shift experiment, but evidence of similar effects on
other transcripts has not been reported. As proposed in an
earlier model (Culbertson 1999; Culbertson and Leeds 2003),
shuttling proteins like Upf3p or Hrp1p could mark Cbc-
bound transcripts in the nucleus for potential decay during
pioneer translation by setting the stage
for the recruitment of other factors
needed for NMD. However, this does
not readily explain how the same factors
are recruited to trigger NMD during sub-
sequent rounds of translation in the cyto-
To stimulatedebate about
recruitment might work for an NMD
pathway that can occur during any
round of translation, we propose two
mechanisms: ‘‘nuclear marking’’ for the
pioneer round of translation, and an
alternate mechanism called ‘‘reverse
assembly’’ for subsequent rounds of
translation. Figure 1 illustrates both
mechanisms. The cartoon depicts Cbc-
bound mRNPs exiting the nucleus to
engage in a pioneer round of transla-
tion. This aspect of the model deserves
further explanation, because the current
claim is that a pioneer ribosome can
translate either Cbc-bound or eIF-4E-
bound transcripts. This could be true,
but there is a caveat to consider.
eIF4E-bound PGK1 mRNA is targeted
by NMD in a cbc1-D strain, which is
surprisingly viable (Gao et al. 2005). Due
to the absence of Cbc1p and because
NMD requires translation, it was argued that pioneer transla-
circumstance. Furthermore, in wild-type strains, intron-con-
taining pre-mRNAs are bound to either Cbc or eIF-4E. Since
nuclear splicing produces mature mRNAs with both kinds of
cap binding complexes before any translation is possible, it
was argued that some mRNAs that undergo pioneer transla-
tion must have eIF-4E bound to the cap. The percentage of
transcripts falling into this category is substantial. Although
only 264 of ?5800 genes contain introns, they are concen-
trated in the most highly expressed genes. Moreover, 50% of
all mRNAs are derived from pre-mRNAs (Li et al. 1999).
However, pre-mRNAs (by virtue of intronic stop
codons) must be exported to the cytoplasm to be degraded
by NMD. From that location, they are no longer precursors
to mature mRNA because the splicing machinery is in the
nucleus. If some of the pre-mRNAs escape NMD during
pioneer translation, eIF-4E could replace Cbc before the
next round of translation, and that could be why Gao
et al. (2005) detected eIF-4E-bound pre-mRNAs. The only
way their model for pioneer translation of eIF-4E-bound
pre-mRNAs holds up is if some eIF-4E is nuclear. Because
it is only 24.3 kDa in size, eIF-4E could import by passive
diffusion (Pante and Aebi 1995), but this has not been
tested. When cells over-express eIF-4E at 100-fold above
normal (1.42 3 106molecules per cell), some nuclear accu-
FIGURE 1. Nuclear marking/reverse assembly model for yeast NMD during pioneer
translation and subsequent rounds of translation. According to the model, the two sub-
complexes of NMD factors assemble sequentially but in a nonobligate order. Nuclear
marking: The Upf3p-Upf2p subcomplex assembles prior to the pioneer round of transla-
tion followed by translation, and then assembly of the Upf1p/RF subcomplex. Reverse
assembly: The order of assembly of the subcomplexes is reversed during subsequent
rounds of translation. Regardless of the order of assembly, a structurally and functionally
identical surveillance complex is formed, leading to translation termination, decapping,
Xrn1p-mediated 50decay, and exosome-mediated 30decay.
1334RNA, Vol. 11, No. 9
Culbertson and Neeno-Eckwall
mulation occurs (Lang et al. 1994; Ghaemmaghami et al.
2003), but mislocalization at this level of expression is
common. Lang et al. (1994) also claimed that eIF-4E can
be detected in the nucleus when expressed at the normal
level, but the data was not shown. This point is so critical to
the conclusion that eIF-4E-bound transcripts undergo pio-
neer translation that we should not incorporate this into the
dogma of NMD until better evidence for nuclear eIF-4E is
As illustrated in Figure 1, nuclear marking posits that
shuttling proteins function in NMD by exporting from the
nucleus in association with every mRNP (mRNA/protein)
complex. During or following export, the pioneer ribosome
displaces the transcript-bound marker proteins, after which
they shuttle back into the nucleus for another round of
mRNP-associated export. However, when translation ter-
minates prematurely, the marker proteins remain asso-
ciated with the transcript, followed by assembly of a
‘‘surveillance’’ complex. The complete NMD-competent
complex, which contains the factors required for translation
termination and NMD (eRF1, eRF3, Upf1p, Upf2p, Upf3p),
triggers termination of translation, decapping, and 50
decay-mediated by the Xrn1p exonuclease. There is also a
link between NMD and exosomal 30decay mediated by an
interaction between Upf1p and Ski7p (Mitchell and Toller-
vey 2003; Takahashi et al. 2003). Thus, nonsense transcripts
appear to be degraded from both ends.
Full assembly of the surveillance complex is most likely
preceded by the prior assembly of two subcomplexes (Atkin
et al. 1997). One contains Upf2p and Upf3p, and the other
contains Upf1p, eRF1, and eRF3 (Upf1p/RF). Staged assem-
bly of the two subcomplexes is supported by two observa-
tions. (1) Upf2p fails to associate with translating ribosomes
in the absence of Upf3p. (2) Upf1p associates with translat-
ing ribosomes in the absence of Upf2p and Upf3p. These and
other findings suggest that the two subcomplexes are
recruited separately and assembled independently. NMD
only occurs when both subcomplexes are present.
We propose that the Upf3p/Upf2p subcomplex assembles
on Cbc-bound transcripts before the pioneer round of trans-
lation commences. If the pioneer ribosome encounters a
premature stop codon, the Upf1p/RF subcomplex is
recruited, and the surveillance complex is assembled. During
bulk translation of eIF-4E-bound transcripts, the subcom-
plexes assemble in reverse order. The structure and function
of the fully assembled surveillance complex could be identical
regardless of the order of assembly of the subcomplexes.
Limitations caused by the low abundance
of Upf proteins
The primary impetus for suggesting reversible assembly of
the subcomplexes is that it helps explain how nonabundant
Upf proteins, one of which shuttles between the nucleus
and the cytoplasm as part of its function in NMD, can
selectively form surveillance complexes on nonsense
mRNAs during both pioneer translation and subsequent
rounds of translation. In mammals the Upf proteins are
present at tens of thousands to millions of copies per cell
(Maquat and Serin 2001), but in S. cerevisiae the Upf
proteins are present at very low intracellular concentrations
(Table 1). The estimates for protein abundance vary (Atkin
et al. 1997; Li et al. 1999; Maderazo et al. 2000; Ghaemma-
ghami et al. 2003), but they are consistently much lower
than cellular transcripts
(?200,000/cell), and release factors (variable but as high
as 79,000/cell), and they consistently point to Upf3p as the
least abundant of the three proteins. A major challenge is to
understand how a small number of Upf proteins can locate
and assemble on nonsense transcripts in the total pool of
transcripts. We argue that prior to the pioneer round of
translation, nuclear marking simplifies the identification of
minority populations of nonsense transcripts in the rela-
tively small pool of exporting transcripts compared to the
larger pool of cytoplasmic transcripts.
To illustrate the limits imposed by low Upf protein
abundance, let’s assume that the cytoplasmic mRNA pool
has an average half-life of 20 min, such that nascent tran-
scripts must export from the nucleus at a rate of 375/min
(15,000/2/20) to replace degraded cytoplasmic transcripts
and to maintain a constant steady-state bulk mRNA level.
This and the requirement for translation in NMD imply
that the minimum required abundance of nuclear marking
proteins is dictated by the combined rates of shuttling and
translation. Protein import from the cytoplasm to the
nucleus through nuclear pores takes 1 msec (Weis 2002).
In contrast, estimates of the transit time for elongation
indicate that translation of the average yeast ORF requires
about 10 sec. Assuming that the combined rate of shuttling
and translation is 10 sec, then one marker protein molecule
could shuttle six times per minute. Thus, the minimum
abundance required for a marker protein to export with
every mRNP is ?60 molecules/cell (375/6). This is slightly
below the lowest estimate for Upf3p abundance of 80 mole-
TABLE 1. Estimates for haploid intracellular concentrations of yeast
a(Maderazo et al. 2000). Numbers indicate molecules/cell. The
estimates were derived by quantitative Western blotting of proteins
detected by a common monoclonal antibody to an epitope tag.
b(Ghaemmaghami et al. 2003). Numbers indicate molecules/cell.
The estimates of abundance for proteins listed in the GFP protein
localization database are typically much higher than estimates
derived by other methods. The actual concentrations of the Upf
proteins are likely to be somewhere in between the estimates
shown in the table for the two methods.
Transcript selection for NMD
cules per cell and considerably less than the highest estimate
(Table 1). It therefore seems plausible that shuttling proteins
could mark every mRNP, leaving the nucleus even when they
are present at low intracellular concentrations.
Despite the apparent feasibility of nuclear marking based
on the above arguments, it would only be fair to point out a
caveat of the model regarding Upf3p that needs to be
addressed. Conclusions grounded in firm evidence include
the following: (1) The import mechanism for Upf3p has
been established—it binds to b-importin Srp1p and is
excluded from the nucleus in an srp1 mutant (Shirley et
al. 2002). (2) Mutations in UPF3 have been identified that
block export, cause nuclear accumulation, and confer an
Nmd?phenotype. (3) Upf3p associates with 80S particles
and polyribosomes (Atkin et al. 1997). However, one
assumption of nuclear marking has not been resolved at
the molecular level. Efforts by us and by others to detect
Upf3p bound to mRNA have failed. Upf3p might associate
with mRNPs indirectly by binding to another mRNP pro-
tein, but besides Upf2p, no other interacting proteins have
surfaced in global two-hybrid screens. These difficulties
could indicate one of two possibilities: Either the nuclear
marking model is wrong or the interaction of Upf3p with
mRNPs is too transient or too weak to purify in immuno-
For subsequent rounds of translation in the cytoplasm,
the low concentrations of the Upf proteins impose severe
restrictions on any mechanism for efficient selection of
NMD substrates among the 15,000 cytoplasmic mRNAs.
Compared to the Upf proteins, most translation factors
are highly abundant, including the release factors eRF1
and eRF3. Complexes between the release factors and
Upf1p can only form with a small fraction of the release
factor pool per unit time, suggesting that the Upf1p/RF
complex may be restricted to aberrant termination events.
Although the rapid cycling of factors must be an essential
feature of shuttling proteins that function in NMD, the
need for rapid cycling is even more critical in subsequent
rounds of translation in the cytoplasm. The Upf proteins
cannot be sequestered on any one transcript for very long.
There is not enough to go around.
Upf protein localization is compatible with assembly
on Cbc- or eIF-4E-bound transcripts
Upf3p shuttles between the nucleus and cytoplasm and
accumulates in the nucleus when overexpressed or when
export is blocked, but most of the protein is cytoplasmic in
wild-type cells (Shirley et al. 1998, 2002). With the use of
sensitive techniques such as confocal microscopy, Upf1p
appears cytoplasmic and cannot be detected in the nucleus
under any conditions (Atkin et al. 1995). Because the local-
ization of Upf2p has never been reported, we determined
the distribution of Upf2-GFP (Fig. 2). The fluorescent sig-
nal was predominantly cytoplasmic even when overex-
pressed. A putative nuclear localization signal sequence
(NLS) that potentially directs Upf2p import into the
nucleus failed to direct a cytoplasmic GFP reporter to the
nucleus. In contrast, three similar NLSs located in Upf3p
direct nuclear import through the b-importin Srp1p (Shir-
ley et al. 1998). These results support a cytoplasmic location
Upf1p was recently shown to interact with nuclear pore
proteins on the cytoplasmic side of nuclear pores (Nazar-
enus et al. 2005). From this location, Upf1p could be poised
to associate with Cbc-bound-mRNPs carrying prebound
Upf3p. The mRNP protein Hrp1p is normally localized
around the nucleus on both sides of the nuclear envelope
but is not distributed throughout the cytoplasm, suggesting
that whatever role it might play in NMD may be restricted
to the translation of nascent transcripts emerging from
nuclear pores. It is worth noting, however, that Hrp1p
spreads throughout the cytoplasm during heat-shock and
could influence NMD in a more general way when cells are
subjected to stress (Henry et al. 2003).
Recruitment of factors for NMD following
the first round of translation
All three Upf proteins are required when NMD occurs
during subsequent rounds of translation following the pio-
neer round (Maderazo et al. 2003; Keeling et al. 2004). The
rapid nucleocytoplasmic shuttling of marking proteins pro-
FIGURE 2. Localization of GFP-Upf2p. GFP-Upf2p was localized
by visualizing green fluorescent protein fusions in live cells by
expressing hybrid fusions from plasmids introduced into strain
NEY4 (MATa ura3-52 trp1-D1 leu2-2 tyr7-1 can1-100 his3-D200
upf2D1<HIS3). Cells were treated with DAPI to visualize nuclear
DNA. The plasmids used were (A) pNE22 (GFP-UPF2, CEN4,
TRP1); (B) pNE25 (GFP-UPF2, 2 mm origin of replication, TRP1);
(C) pNE85-UPF2 (CUP1-GFP-UPF2, CEN4, TRP1); (D) pNE85-
UPF2(19–47) (amino acids 19–47 including the putative nuclear loca-
lization signal motif in Upf2p). Cells were grown in medium contain-
ing 25 mM Cu++without tryptophan.
1336 RNA, Vol. 11, No. 9
Culbertson and Neeno-Eckwall
vides a rational way to recruit RNA substrates and assemble
the surveillance complex in preparation for the pioneer
round of translation, but what happens when transcripts
that are targeted for NMD escape first-round decay? The
two possibilities are: (1) NMD factors are retained on the
transcript, or (2) they are displaced by the pioneer ribo-
some and must be re-assembled. We favor displacement
and reassembly. These proteins cannot be sequestered dur-
ing multiple rounds of translation from the pools of free
Upf proteins needed to find and recruit new NMD-sensitive
transcripts among the constantly emerging nascent tran-
scripts. Rapid cycling between successive rounds of transla-
tion would be required. NMD-sensitive transcripts must be
re-recruited and Upf proteins re-assembled between each
round of translation. The arguments that make nuclear
marking an attractive model for recruitment in preparation
for the pioneer round no longer apply in subsequent rounds,
because recruitment and assembly take place among the
larger pool of cytoplasmic transcripts. A different mecha-
nism seems likely.
We propose that in subsequent rounds of translation, the
Upf1p/RF subcomplex is the most likely one to be recruited
first when a premature termination codon is encountered
and recognized as an aberrant termination codon. Recent
evidence shows that aberrant termination occurs when the
stop codon is followed by an imposter sequence posing as
an improperly configured 30-UTR (Amrani et al. 2004).
This sequence is most likely functionally equivalent to the
degenerate AU-rich downstream element (DSE) shown to
be required for NMD (Zhang et al. 1995).
In the cytoplasm where the subcomplexes must find the
substrates for NMD in the larger pool of total transcripts, a
reverse order of assembly makes sense, but not out of
necessity. It is favored by probability; Upf1p/RF is 5–20
times more abundant than Upf3p/Upf2p (Table 1). After
the Upf1p/RF subcomplex is recruited, the Upf2p/Upf3p
subcomplex associates through the binding domains known
to tether all three proteins together in the surveillance
complex (He et al. 1997). The Upf3p/Upf2p subcomplex
forms its own association with RF3 (Gonzalez et al. 2001;
Wang et al. 2001). As mentioned earlier, we don’t envision
any necessity for the fully assembled surveillance complex
to differ in structure or function between the pioneer round
and subsequent rounds of translation.
Error-containing versus error-free targets of NMD
Transcripts that contain coding errors are best eliminated
during the pioneer round of translation to prevent any
truncated protein product from being made. However,
wild-type transcripts have been identified where the turn-
over rate is controlled by the Upf genes as part of the
normal repertoire of gene expression (Lelivelt and Culbert-
son 1999; He et al. 2003). These transcripts, which code for
important functional proteins, are targeted for NMD not
because they contain coding errors, but for other reasons.
The Upf proteins down-regulate their abundance and
reduce the amount of protein product made, but without
altogether eliminating their translation.
The difference between transcripts targeted to eliminate
a protein product versus those targeted to reduce but not
eliminate a protein product is not immediately obvious
because for both of them, steady-state RNA transcript levels
are not reduced by NMD to zero. Most nonsense transcripts
are reduced fourfold to 10-fold. For example, met8 and
pgk1 nonsense transcripts are reduced fourfold and 10-
fold, respectively (Gao et al. 2005). In addition, 82% of
529 error-free transcripts that are targeted by NMD are
reduced by threefold or less, but 18% are reduced much
more—as high as 11-fold (Lelivelt and Culbertson 1999).
We believe the reason that low levels of transcripts are
always detected is that for all transcripts there is a protected
class—those nascent transcripts that have not yet been
exported from the nucleus and have therefore not been
exposed to NMD through translation. Since the balance
between rates of transcription and decay vary for each
transcript, the steady-state percentage of protected tran-
scripts will vary. Because of this, it is reasonable to assume
that RNA surveillance, while not eliminating all detectable
mRNA, may very well limit the synthesis of truncated
proteins to the one made by necessity during pioneer
translation to test whether the transcript can be translated
full-length. This argues for very efficient RNA surveillance
during the first round of translation, a level of efficiency
that could be achieved by nuclear marking. In the case of
error-free transcripts, the example below illustrates one way
transcripts can be targeted by NMD while maintaining
production of a functional protein product.
Error-free transcripts are most likely substrates for NMD
during all rounds of translation. One of these, the wild-type
SPT10 transcript, is three times more abundant and three-
fold more stable when NMD is inactivated, and is targeted
for NMD by ‘‘leaky scanning’’ (Welch and Jacobson 1999).
SPT10 is NMD-sensitive because ribosomes fail to initiate
efficiently at the first AUG codon due to a suboptimal
context. As a result, ribosomes frequently bypass the initi-
ation codon and scan to a second out-of-frame AUG
(Fig. 3). When out-of-frame initiation occurs, a stop
codon in this alternate frame triggers NMD. SPT10 tran-
scripts escape NMD when translation initiates at the normal
AUG, but they are repeatedly exposed to the possibility of
NMD, which can be triggered or bypassed in each succes-
sive round of translation. The beauty of leaky scanning is
that decay rates can be reduced by a mechanism that per-
mits synthesis of a necessary protein product. Transcripts
targeted by leaky scanning most likely decay with composite
rates that depend on the frequency of ribosomal scan-
ning past the first AUG and the relative efficiencies of
Upf protein recruitment during pioneer versus bulk trans-
Transcript selection for NMD
Why should there be two mechanisms for transcript selec-
tion and complex assembly in NMD? We think the question
can be rephrased—why should Upf3p shuttle between the
nucleus and the cytoplasm if there is only one mechanism?
To accommodate shuttling, we propose that the pioneer
round of translation is distinguished from subsequent
rounds by the order of the steps leading to NMD. For
NMD to occur during the pioneer round, the Upf3p/Upf2p
subcomplex is preloaded on mRNPs by nuclear marking, the
pioneer round commences, and the presence of a premature
stop codon leads to termination of translation concomitant
with assembly of the Upf1p/RF complex. Nuclear marking in
preparation for the pioneer round of translation could
enhance the probability of NMD in the pioneer round.
In subsequent rounds of translation, neither subcomplex
is associated with a nonsense transcript until translation
comes to a premature halt, at which time the Upf1p/RF
subcomplex is most often the first to associate by virtue of
being more abundant. During bulk translation of eIF-4E-
bound transcripts, reverse assembly enhances the probability
of NMD. The efficiencies of NMD during pioneer translation
and subsequent rounds of translation may not be the same.
Sufficient similarities exist between yeast and mamma-
lian NMD to suppose that divergent pathways in these
organisms represent alternative evolutionary outcomes
derived from a common ancestral pathway (Culbertson
and Leeds 2003). In yeast, NMD targets both Cbc-bound
and eIF-4E-bound transcripts. Two alternatives may be
available to assemble the full surveil-
lance complex from two subcomplexes.
By accounting for differences between
rounds of translation, the combination
of nuclear marking and reverse assem-
bly explains how the majority of non-
sense transcripts might be preselected
by one or more shuttling proteins in
preparation for degradation during the
pioneer round of translation, while pro-
viding an alternate mechanism using
the same factors to degrade transcripts
that are targeted for NMD but escape
We thank Dr. Jon Warner for sharing
unpublished data and ideas, Dr. Gerry
their ideas. Support was from the Univer-
sity of Wisconsin College of Agricultural
and Life Sciences, the University of Wis-
consin Medical School, and NIH grant
GM65172 (M.R.C.). This is Laboratory of Genetics paper no.
Received May 17, 2005; accepted June 16, 2005.
Amrani, N., Ganesan, R., Kervestin, S., Mangus, D., Ghosh, S., and
Jacobson, A. 2004. A faux 30-UTR promotes aberrant termination
and triggers nonsense-mediated mRNA decay. Nature 432: 112–
Atkin, A., Altamura, N., Leeds, P., and Culbertson, M. 1995. The
majority of yeast UPF1 co-localizes with polyribosomes in the
cytoplasm. Mol. Cell Biol. 6: 611–625.
Atkin, A., Schenkman, L., Eastham, M., Dahlseid, J., Lelivelt, M., and
Culbertson, M. 1997. Relationship between yeast polyribosomes
and Upf proteins required for nonsense-mediated mRNA decay.
J. Biol. Chem. 272: 22163–22172.
Culbertson, M. 1999. RNA surveillance. Unforeseen consequences for
gene expression, inherited genetic disorders and cancer. Trends
Genet. 15: 74–80.
Culbertson, M. and Leeds, P. 2003. Looking at mRNA decay pathways
through the window of molecular evolution. Curr. Opin. Gen. Dev.
Gao, Q., Das, B., Sherman, F., and Maquat, L. 2005. Cap-binding
protein 1-mediated and eukaryotic translation initiation factor
4E-mediated pioneer rounds of translation in yeast. Proc. Natl.
Acad. Sci. 102: 4258–4263.
Ghaemmaghami, S., Huh, W., Bower, K., Howson, R., Belle, A.,
Dephoure, N., O’Shea, E., and Weissman, J. 2003. Global analysis
of protein expression in yeast. Nature 425: 737–741.
Gonzalez, C., Ruiz-Echevarria, M., Vasudevan, S., Henry, M., and
Peltz, S. 2000. The yeast hnRNP-like protein Hrp1/Nab4 marks a
transcript for nonsense-mediated mRNA decay. Mol. Cell 5: 489–
FIGURE 3. Translation and decay of NMD-sensitive transcripts targeted by leaky scan-
ning. The figure depicts two mechanisms for the assembly of the surveillance complex,
‘‘nuclear marking’’ during the pioneer round of translation, and ‘‘reverse assembly’’ for
subsequent rounds of translation.
1338 RNA, Vol. 11, No. 9
Culbertson and Neeno-Eckwall
Gonzalez, C.I., Bhattacharya, A., Wang, W., and Peltz, S.W. 2001.
Nonsense-mediated mRNA decay in Saccharomyces cerevisiae.
Gene 274: 15–25.
He, F., Brown, A., and Jacobson, A. 1997. Upf1p, Nmd2p, and Upf3p
are interacting components of the yeast nonsense-mediated mRNA
decay pathway. Mol. Cell. Biol. 17: 1580–1594.
He, F., Li, X., Spatrick, P., Casillo, R., Dong, S., and Jacobson, A. 2003.
Genome-wide analysis of mRNAs regulated by the nonsense-
mediated and 50to 30mRNA decay pathways in yeast. Mol. Cell
Henry, M., Mandel, D., Routson, V., and Henry, P. 2003. The yeast
hnRNP-like protein Hrp1/Nab4 accumulates in the cytoplasm
after hyperosmotic stress: A novel Fps1-dependent response. Mol.
Biol. Cell 14: 3929–3941.
Ishigaki, Y., Li, X., Serin, G., and Maquat, L.E. 2001. Evidence for a
pioneer round of mRNA translation: mRNAs subject to nonsense-
mediated decay in mammalian cells are bound by CBP80 and
CPB20. Cell 106: 607–617.
Keeling, K., Lanier, J., Du, M., Salas-Marco, J., Gao, L., Kaenjak-
Anageletti, A., and Bedwell, D. 2004. Leaky termination at prema-
ture stop codons antagonizes nonsense mediated mRNA decay in
S. cerevisiae. RNA 10: 691–703.
Kuperwasser, N., Brogna, S., Dower, K., and Rosbash, M. 2004. Non-
sense-mediated decay does not occur within the yeast nucleus.
RNA 10: 1907–1915.
Lang, V., Zanchin, I., Lunsdorf, H., Tuite, M., and McCarthy, J. 1994.
Initiation factor eIF4E of Saccharomyces cerevisiae. J. Biol. Chem.
Lelivelt, M. and Culbertson, M. 1999. Yeast Upf proteins required
for RNA surveillance affect global expression of the yeast tran-
scriptome. Mol. Cell. Biol. 19: 6710–6719.
Li, B., Nierras, C., and Warner, J. 1999. Transcriptional elements
involved in the repression of ribosomal protein synthesis. Mol.
Cell. Biol. 19: 5393–5404.
Maderazo, A., He, F., Mangus, D., and Jacobson, A. 2000. Upf1p
control of nonsense mRNA translation is regulated by Nmd2p
and Upf3p. Mol. Cell. Biol. 20: 4591–4603.
Maderazo, A., Belk, J., He, F., and Jacobson, A. 2003. Nonsense-
containing mRNAs that accumulate in the absence of a functional
nonsense-mediated mRNA decay pathway are destabilized rapidly
upon its restitution. Mol. Cell. Biol. 23: 842–851.
Maquat, L. and Serin, G. 2001. Nonsense-mediated mRNA decay:
Insights into mechanism from the cellular abundance of human
Upf1, Upf2, Upf3, and Upf3X proteins. Cold Spring Harbor Symp.
Quant. Biol. LXVI: 313–320.
Mitchell, P. and Tollervey, D. 2003. An NMD pathway in yeast involv-
ing accelerated deadenylation and exosome-mediated 30fi50degra-
dation. Mol. Cell 11: 1405–1413.
Nazarenus, T., Cedarberg, R., Bell, R., Cheatle, J., Forch, A., Haifley,
A., Hou, A., Wanja Kebaara, B., Shields, C., Stoysich, K., et al.
2005. Upf1p, a highly conserved protein required for nonsense-
mediated mRNA decay, interacts with the nuclear pore proteins
Nup100p and Nup116. Gene 345: 199–212.
Pante, N. and Aebi, U. 1995. Toward a molecular understanding of the
structure and function of the nuclear pore complex. Int. Rev. Cytol.
Shirley, R., Lelivelt, M., Schenkman, L., Dahlseid, J., and Culbertson,
M. 1998. A factor required for nonsense-mediated mRNA decay in
yeast is exported from the nucleus to the cytoplasm by a nuclear
export signal sequence. J. Cell Sci. 111: 3129–3143.
Shirley, R., Ford, A., Richards, R., Albertini, M., and Culbertson, M.
2002. Nuclear import of Upf3p is mediated by importin-a/b and
export to the cytoplasm is required for a functional nonsense-
mediated mRNA decay pathway in yeast. Genetics 161: 1465–
Takahashi, S., Araki, Y., Sakuno, T., and Katada, T. 2003. Interaction
between Ski7p and Upf1p is required for nonsense-mediated 30-to-50
mRNA decay in yeast. EMBO J. 22: 3951–3959.
Wang, W., Czaplinski, K., Rao, Y., and Peltz, S.W. 2001. The role of
Upf proteins in modulating the translation read-through of non-
sense-containing transcripts. EMBO J. 20: 880–890.
Warner, J.R. 1999. The economics of ribosome biosynthesis in yeast.
Trends Biochem. Sci. 24: 437–440.
Weis, K. 2002. Nucleocytoplasmic transport: Cargo trafficking across
the border. Curr. Opin. Cell Biol. 14: 328–335.
Welch, E. and Jacobson, A. 1999. An internal open reading frame
triggers nonsense-mediated decay of the yeast SPT10 mRNA.
EMBO J. 18: 6134–6145.
Zhang, S., Ruiz-Echevarria, M., Quan, Y., and Peltz, S. 1995.
Identification and characterization of a sequence motif involved
in nonsense-mediated mRNA decay. Mol. Cell. Biol. 15: 2231–
Transcript selection for NMD