lessons not learned fromtranslation

Upload: plastioid4079

Post on 03-Apr-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/28/2019 Lessons Not Learned Fromtranslation

    1/10

    Lessons (not) learned from mistakes about translation

    Marilyn Kozak

    Department of Biochemistry, Robert Wood Johns on Medical School, 675 Hoes Lane, Piscataway, NJ 08854, USA

    Received 9 July 2007; received in revised form 16 August 2007; accepted 27 August 2007

    Received by A.J. van Wijnen

    Available online 5 September 2007

    Abstract

    Some popular ideas about translational regulation in eukaryotes have been recognized recently as mistakes. One example is the rejection of a

    long-standing idea about involvement of S6 kinase in translation of ribosomal proteins. Unfortunately, new proposals about how S6 kinase might

    regulate translation are based on evidence that is no better than the old. Recent findings have also forced rejection of some popular ideas about the

    function of sequences at the 3 end of viral mRNAs and rejection of some ideas about internal ribosome entry sequences (IRESs). One long-held

    belief was that tissue-specific translation via an IRES underlies the neurotropism of poliovirus and the attenuation of Sabin vaccine strains. Older

    experiments that appeared to support this belief and recent experiments that refute it are discussed. The hypothesis that dyskeratosis congenita is

    caused by a defect in IRES-mediated translation is probably another mistaken idea. The supporting evidence, such as it is, comes from a mouse

    model of the disease and is contradicted by studies carried out with cells from affected patients. The growing use of IRESs as tools to study other

    questions about translation is discussed and lamented. The inefficient function of IRESs (if they are IRESs) promotes misunderstandings. I explain

    again why it is not valid to invoke a special mechanism of initiation based on the finding that edeine (at very low concentrations) does not inhibit

    the translation of a putative IRES from cricket paralysis virus. I explain why new assays, devised to rule out splicing in tests with dicistronic

    vectors, are not valid and why experiments with IRESs are not a good way to investigate the mechanism whereby microRNAs inhibit translation.

    2007 Elsevier B.V. All rights reserved.

    Keywords: Translational control; S6 kinase; IRES; Edeine, Poliovirus translation; 14-3-3; dyskeratosis congenita; microRNA; 3 UTR

    Science, my lad, is made up of mistakes, but they are

    mistakes which it is useful to make because they lead little

    by little to the truth.

    Jules Verne

    1. Introduction

    It is not useful, however, to keep making the same mistakes. Toprevent repetition, it helps to know how an erroneous idea initially

    took hold and how it was eventually recognized as wrong. With

    that as the goal, this review discusses some once-popular ideas

    about translational regulation that recently were found to be mis-

    taken. The flaws that led to these misunderstandings still pervade

    the literature, suggesting the lessons have not yet taken hold.

    The first topic (Section 2) concerns the regulatory effects of a

    well-known kinase (S6K) thattargets ribosomal protein S6 (rpS6).An old hypothesis linking translation of ribosomal proteins to

    phosphorylation of rpS6 persisted for decades, although it was

    based only on correlative evidence. Recent studies undertaken to

    identify new targets for S6K again make the mistake of relying

    more on correlations than on meaningful functional tests.

    Section 3 is about unusual mechanisms of translation thought

    to be mediated by sequences near the 3 end of viralRNAs. Recent

    evidence disproves a model for initiation of translation involving

    the tRNA-like sequence at the 3 end of turnip yellow mosaic virus

    (TYMV) RNA. An idea about regulation of translation via a

    protein that binds the 3 end of rotavirus mRNAs has also been

    Gene 403 (2007) 194203

    www.elsevier.com/locate/gene

    Abbreviations: CrPV, cricket paralysis virus; DC, dyskeratosis congenita;

    eIF, eukaryotic initiation factor; Fluc, firefly luciferase; HCV, hepatitis C virus;

    HIV, human immunodeficiency virus; IRES, internal ribosome entry sequence;

    mTOR, (mammalian) target of rapamycin; NSP3, nonstructural protein 3; PABP,

    poly(A) binding protein; PDCD4, programmed cell death protein 4; PTB,

    polypyrimidine tract binding protein; Rluc, Renilla luciferase; rpS6, ribosomal

    protein S6; S6K, rpS6 kinase; TOP, 5-terminal oligo-pyrimidine tract; TYMV,

    turnip yellow mosaic virus; XIAP, X-linked inhibitor of apoptosis. Tel.: +1 732 235 5355; fax: +1 732 235 5356.

    E-mail address: [email protected].

    0378-1119/$ - see front matter 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.gene.2007.08.017

    mailto:[email protected]://dx.doi.org/10.1016/j.gene.2007.08.017http://dx.doi.org/10.1016/j.gene.2007.08.017mailto:[email protected]
  • 7/28/2019 Lessons Not Learned Fromtranslation

    2/10

    retested and rejected. Even before these corrections came out last

    year, deficiencies in the original experiments were obvious

    (Kozak, 2004), albeit not obvious enough to keep the erroneous

    ideas from entering some textbooks (Flint et al., 2004).

    Section 4 concerns mRNA sequences, called IRES elements,

    that purportedly mediate internal entry of ribosomes. Advocates

    of the internal initiation hypothesis are not yet willing to acknowl-edge many of the experimental flaws pointed out by others

    (Kozak, 2003, 2005), but advocates do admit that some long-held

    ideas about poliovirus translation are incorrect. Recent experi-

    ments revealed, for example, that poliovirus is not translated

    better in neuronal cells than in other tissues, and Sabin vaccine

    strains are not attenuated specifically in neuronal cells. Other

    misunderstandings involve IRESs that purportedly exist in

    cellular mRNAs. This idea was invoked to explain the

    pathological effects of mutations in the dyskerin (DKC1) gene,

    but closer study of cells from affected patients suggests an

    alternative mechanism unrelated to IRESs. A postulated con-

    nection between IRES-mediated translation and a protein called14-3-3, which is frequently lost in tumors, is another case where

    the evidence falls far short of what is claimed.

    The growing use of putative IRESs as tools to test other ideas

    about translation (Section 5) makes it important to think clearly

    about this subject. The discussion herein of admitted and

    possible misunderstandings might help.

    2. How does S6K regulate translation (if it does)?

    2.1. An old hypothesis about translation of mRNAs that encode

    ribosomal proteins is wrong

    It was long believed that, in response to mitogens, phosphor-ylation of rpS6 by the eponymous kinase stimulates translation of

    mRNAs thatbeara 5-terminal oligo-pyrimidine tract (TOP). This

    hypothesis is no longer tenable, inasmuch as translation of TOP

    mRNAs was perfectly normal in knock-in mice whose rpS6 gene

    contains serine-to-alanine substitutions at all five sites normally

    phosphorylated by S6K (Ruvinsky et al., 2005). The overall rate

    of translation in liver cells from these mice was actually higher

    than normal (2.5-fold). In a complementary experiment, S6K-

    knockoutmice weretested; and again, translation of TOPmRNAs

    was not impaired (Pende et al., 2004).

    While nearly everyone now admits that the hypothesis linking

    S6K and rpS6 to translation of TOP mRNAs was wrong, no onetalks about how the misunderstanding became entrenched. The

    answer is that the evidence never went beyond showing a

    correlation: when resting cells were stimulated to resume

    growing, mRNAs were recruited onto polysomes and rpS6

    became phosphorylated (Gressnerand Wool, 1974;Thomas et al.,

    1980); and the recruitment onto polysomes depended on 5 TOP

    sequences (Levy et al., 1991; Schwab et al., 1999). No one

    attempted to define the mechanism. Small differences in activity

    between phosphorylated and nonphosphorylated ribosomes were

    occasionally reported (Duncan and McConkey, 1982), but

    functional effects of phosphorylation were not seen consistently

    (Mastropaolo and Henshaw, 1981). People simply found new

    ways to demonstrate a correlation between S6K and translation of

    TOP mRNAs (Jefferies et al., 1997; Terada et al., 1994). The

    complexity of mammalian signaling pathways makes it hard to

    interpret those findings, however. The rapamycin-sensitive kinase

    (mTOR) that activates S6K induces elevated production of all

    components of the translational machinery (e.g. transcription and

    processing of rRNAs increases; Hannan et al., 2003) and many

    other growth-related processes are stimulated (Wullschleger et al.,2006). The bottom line is that, although mTOR somehow brings

    aboutan increasein translation of TOP (and many other) mRNAs,

    S6K and phosphorylation of rpS6 are not involved.

    2.2. New ideas about how S6 kinase regulates translation are

    based on flimsy evidence

    The new ideas focus on translation factors. Holz et al. (2005)

    postulate that phosphorylation of eukaryotic initiation factor

    (eIF)4B by S6K is regulated by binding of S6K to eIF3. The

    evidence boils down to showing that S6K associates and

    dissociates from eIF3 in a way that correlates with activation ofS6K by mTOR. No hard evidence supports the hypothesis that

    release of activated S6K from eIF3 puts the kinase in position

    to act on rpS6 and eIF4B, and some evidence contradicts that

    idea: the small fraction of eIF3 found to be associated with S6K

    was not associated with ribosomes.

    The functional consequence of eIF4B phosphorylation is also

    dubious. Phosphorylation was shown to increase the amount of

    eIF4B that co-immunoprecipitates with eIF3 (Holz et al., 2005,

    Fig. 4B; Shahbazian et al., 2006, Fig. 6), but that is not a test of

    function. Raught et al. (2004) say a mutation in eIF4B that

    prevents phosphorylation by S6K abolishes the activity of the

    factor, but the activity they monitored was remarkably artificial:

    over-expression ofwild typeeIF4B was found to inhibittranslationof a co-transfected reporter gene, while over-expression of

    nonphosphorylatable eIF4B did not inhibit translation.

    The other newly identified substrate for S6K is programmed

    cell death protein 4 (PDCD4), which has been described as an

    inhibitor of eIF4A. There is compelling evidence that phos-

    phorylation by S6K promotes degradation of PDCD4 (Dorrello

    et al., 2006) but no convincing evidence that PDCD4 regulates

    translation. PDCD4 does bind to eIF4A, but over-expression of

    PDCD4 caused only a two-fold reduction in translation of a

    reporter gene (Yang et al., 2003). These marginal findings

    demonstrated by using a synthetic mRNA and over-expressing

    PDCD4 do not justify the assertion that PDCD4 is a noveltumor suppressor that inhibitstranslation of proteins that are

    essential for neoplastic transformation (LaRonde-LeBlanc

    et al., 2007).

    3. The 3 end of some viral RNAs does not perform a

    special function in initiation of translation

    3.1. TYMV

    The hypothesis proposed by Barends et al. (2003) and

    refuted by Matsuda and Dreher (2007) is that a tRNA-like

    sequence at the 3 end of TYMV mRNA, which can be charged

    with valine, reaches back to the 5 end of the mRNA and

    195M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    3/10

    functions instead of Met-tRNAi to initiate translation from the

    second AUG codon, producing a 206 kDa polyprotein. In

    principle, there is no need for such a fancy mechanism to explain

    translation of the polyprotein. The first AUG in TYMV mRNA,

    which initiates translation of a 69 kDa movement protein, is in a

    weak context which should allow leaky scanning; and there is

    indeed evidence that some ribosomes gain access to the secondAUG via leaky scanning (Matsuda and Dreher, 2006).

    As evidence for the reachback hypothesis, Barends et al.

    (2003) claimed that [3H]valine was donated from the 3 tRNA-

    like element to polypeptide(s) synthesized in a cell-free

    translation system programmed with TYMV mRNA. The

    main product labeled by [3H]valine was said to be a fragment

    of the TYMV polyprotein, although its molecular weight was

    different (19 kDa) and the identification was not confirmed by

    immunological techniques. Thus, the evidence was weak to

    begin with. A followup test in which the incorporation into

    proteins of viral-linked radiolabeled valine was suppressed in

    the presence of free unlabeled valine (Matsuda and Dreher,2007, Fig. 4) makes it unlikely that valine really is donated

    directly from TYMV RNA into viral proteins.

    As a second line of evidence, Barends et al. (2003) examined

    [35S]methionine-labeled proteins directed by TYMV mRNA

    which had been pre-treated with RNase H to remove the 3

    tRNA-like element. The authors say the truncated mRNA

    specifically lost the ability to synthesize the polyprotein; but the

    evidence was not convincing because (i) the 206 kDa protein

    was difficult to distinguish from other bands in the gel, and (ii)

    there was no proof that RNase H cut only at the intended sites in

    the 3 UTR. When the point was retested by Matsuda and

    Dreher (2007) using antisera to identify the polypeptides and

    avoiding the use of RNase, deletion of the 3 tRNA-likestructure had no effect on translation of viral proteins in vitro.

    In formulating the reachback model, Barends et al. (2003)

    ignored their own finding that oxidation of the 3 cytosine in

    TYMV mRNA, which obviously precludes aminoacylation,

    caused no reduction in synthesis of viral proteins in vitro. Thus,

    the accumulated evidence strongly contradicts the reachback

    mechanism.

    3.2. Rotavirus

    The closed-loop model postulates that contact between poly

    (A) binding protein (PABP) and eIF4G augments translation.This caused people to wonder how non-polyadenylated

    rotavirus mRNAs get translated. One idea was that a

    nonstructural viral protein (NSP3) might substitute for PABP,

    based on the finding that NSP3 binds to both eIF4G and a

    sequence near the 3 end of rotavirus mRNAs (Piron et al.,

    1998). Vende et al. (2000) argued that this binding has functional

    consequences; i.e. they say circularization of rotavirus mRNAs

    by NSP3 is necessary for efficient translation. Their experiments

    employed unnatural systems, however; e.g. translation of

    uncapped mRNAs in vitro, and translation of chimeric (rather

    than authentic viral) mRNAs in a cell line that expresses NSP3

    constitutively. Other deficiencies in these experiments were

    explained in an earlier review (Kozak, 2004).

    The notable new finding derived from straightforward

    measurements of viral protein synthesis in infected monkey

    cells is that siRNA-mediated knockdown of NSP3 had no

    detrimental effect on production of other rotavirus proteins.

    Thus, Montero et al. (2006) concluded that NSP3 does not

    augment translation of rotavirus mRNAs.

    Not yet settled is the question of whether NSP3 inhibits hostprotein synthesis (perhaps by binding to eIF4G). Montero et al.

    (2006) noted that synthesis of host proteins persisted, to some

    extent, when NSP3 was knocked down.1 Piron et al. (1998)

    emphasized a temporal correlation between formation of

    eIF4G-NSP3 complexes and disruption of eIF4G-PABP com-

    plexes (detected by immunoprecipitation), but there was a

    temporal discrepancy in that the reduction in host protein

    synthesis preceded the disruption of eIF4G-PABP complexes.

    Host translation was markedly reduced by 3 h post-infection,

    while the reduction in eIF4G-PABP complexes did not begin

    until 4.5 h and was not complete until 6 h.

    If the old hypotheses about rotavirus translation are wrong(Piron et al., 1998; Vende et al., 2000), what mechanisms operate

    instead? My guesses are that (i) the mechanism of initiation of

    translation is the same for host and viral mRNAs. The fact that

    rotavirus mRNAs lack a poly(A) tail is irrelevant because PABP

    does not really augment translation of other mRNAs. [This

    problematic point is discussed in other reviews (Kozak, 2004,

    2006).] (ii) eIF4G is required for translation of both host and

    viral mRNAs; i.e. the normal cap-dependent mechanism

    operates. The finding that knock-down of eIF4G caused no

    reduction in translation of viral or host mRNAs (Montero et al.,

    2006, Fig. 7) simply means that eIF4G is not limiting; the

    residual amount of eIF4G (10%) is sufficient to support protein

    synthesis. (iii) The eventual decline in host protein synthesiscould be explained by competition from the massive amount of

    viral mRNAs produced in infected cells.

    4. Misunderstandings about IRESs

    4.1. Revised thinking about poliovirus neurotropism

    The ability of wild-type but not vaccine strains of poliovirus

    to cause paralytic disease, and the finding that key mutations

    map to the 5 UTR in Sabin vaccine strains, made it reasonable

    to think that poliovirus neurotropism might be determined by

    cell-type specific translation.2

    Many experiments were inter-preted as evidence for this idea, which now appears to be wrong.

    Recent experiments rule out tissue-specific translation as an

    1 An experiment thatused firefly luciferase (Fluc) as a stand-infor host mRNAs

    is suspect, however, because the cells were transfected withthe Fluc plasmid after

    the onset of infection. The strong inhibition of Fluc expression, except when

    NSP3 was knocked down, might reflect a block at the level of mRNA processing

    or transport which would not apply to the pool of cellular mRNAs that existed

    prior to infection.2 The 5 UTR of picornaviruses is involved also in RNA replication and

    packaging (Borman et al., 1994; Hunziker et al., 2007; Johansen and Morrow,

    2000), and therefore a defect at the level of translation is not the only way toexplain the effects of the mutations.

    196 M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    4/10

    explanation for why poliovirus replicates predominately in the

    brain and spinal cord.

    The new experiments involve testing the ability of the

    poliovirus IRES to support translation in mice, using an

    adenovirus vector to express a dicistronic mRNA in which a

    wild-type or mutated IRES precedes the 3 Renilla luciferase

    (Rluc) reporter gene (Kauder and Racaniello, 2004).3 Thisapproach ensures that other effects of the mutations (e.g. on

    RNA replication) are not mistaken for effects on translation. The

    most important finding was that the poliovirus 5 UTR supports

    translation, not only in the brain and spinal cord, but in many

    other organs. The authors conclude also that a translational

    defect associated with the Sabin type 3 IRES was observed in all

    organs. (That conclusion requires qualification. With dicis-

    tronic constructs containing the Sabin-3 mutation, production of

    Rluc was reduced 3-to 5-fold compared to the wild-type IRES;

    but the absence of RNA analyses makes it hard to know if the

    primary effect of the mutation was on translation or mRNA

    accumulation. Either way, there clearly was no organ-specificeffect.)

    If it is not true that neurovirulent (wild type) poliovirus is

    translated better in neuronalcells than in other tissues,and not true

    thatSabin strains are attenuated specifically in neuronal cells, how

    did those mistaken ideascome to be believed?Re-reading some of

    the early papers is informative. One study (Agol et al., 1989)

    admits to screening three neuronal cell lines and then proceeding

    with the line that best discriminated between neurovirulent and

    attenuated strains. (An addendum mentions that not all

    neurovirulent strains grow as well, and not all attenuated strains

    grow as poorly, as the ones used in the published experiments).

    Another study says we sought a cell line in which replication of

    attenuated poliovirus would be impaired (La Monica andRacaniello, 1989; the chosen line was SY5Y). As years passed,

    people forgot about this biased screening as they clung to the

    conclusion that there is a correlation between mutations in the

    IRES and restricted growth in neuronal cells.

    What generated the mistaken belief that a deficiency at the

    level of translation underlies the purportedly restricted growth of

    vaccine strains in neuronal cells? The aforementioned study with

    SY5Y cells in which the Sabin-3-like virus replicated poorly

    (ten-fold less efficiently than wild-type poliovirus; La Monica

    and Racaniello, 1989) showed only a two-fold reduction in

    synthesis of viral proteins; this was cited as evidence of impaired

    translation of the attenuated virus. In later studies, thesupposedly impaired translation in neuronal cells was blamed

    on impaired binding of polypyrimidine tract binding protein

    (PTB), although binding of PTB to the IRES was barely

    detectable even with wild-type poliovirus (Guest et al., 2004;

    Gutierrez et al., 1997). Other evidence was obtained using in

    vitro translation systems, which can be manipulated to show

    almost anything. In one study, Haller et al. (1996) tested the

    effects of adding an extract from neuronal cells to a reticulocyte

    translation system programmed with RNA from laboratory-

    constructed mutant viruses (revertants of X472) that purportedly

    mimic Sabin-1. Their main finding was that a crude extract from

    NGP neuronal cells, which stimulated translation of wild-typepoliovirus, fai led to stim ulate translation of the X472

    revertants.4 If Kauder and Racaniello (2004) are correct, all

    these hints of tissue-specific translational defects are artifacts

    worth talking about only as warnings about similar experiments

    with other putative IRESs.

    4.2. Unjustified claims about cellular IRESs and disease

    Experiments with a mouse model of dyskeratosis congenita

    (DC) gave rise to the idea that translation is impaired by

    mutations in the DKC1 gene, which encodes an enzyme

    (dyskerin) that modifies uridine residues in rRNA. The claimthatDkc1m cells are specifically impaired in translation of IRESs

    (Yoonet al., 2006) requires strongevidence that thethree affected

    mRNAs Bcl-xL, p27Kip1 and X-linked inhibitor of apoptosis

    (XIAP) are translated by internal initiation, but that key point

    was not proven. Earlier evidence for IRES activity in XIAP and

    p27Kip1 mRNAs did not hold up; re-analysis uncovered cryptic

    promoters and splicing (Liu et al., 2005; Van Eden et al., 2004a).

    Ignoring those problems, Yoon et al. (2006) argue that their claim

    is supported by the finding that XIAP expression is stimulated by

    irradiationin normalbut notDkc1m mutantcells; i.e. they argue

    (unconvincingly) that stimulation by irradiation is proof of

    IRES-dependent translation.

    Inasmuch as Bcl-xL had not previously been identified as anIRES, it was crucial to test that point, e.g. by RNA transfection

    experiments with dicistronic (5-Rluc-Fluc-3) transcripts. This

    crucial test had no controls, however (Yoon et al., 2006, Fig. 2G).

    The Bcl-xL IRES was compared only to itself(i.e. the Fluc/Rluc

    ratio in normal unirradiated cells was set at 1.0 without revealing

    how the Fluc yield from the Bcl-xL dicistronic construct

    compares to a monocistronic mRNA, or a dicistronic mRNA

    lacking an IRES, or a dicistronic mRNA containing a proven

    IRES, if such there is). The argument used here has a circular

    quality: the slightly reduced translation of Bcl-xL dicistronic

    constructs in Dkc1m cells (compared to normal cells) is cited as

    evidence that Bcl-xL is an IRES, and reduced translation of Bcl-xL is then cited as evidence that Dkc1m cells are impaired in

    translation of IRESs.

    In short, the studies with mouse Dkc1m cells are unconvinc-

    ing; and the underlying idea of a ribosomal defect in Dkc1m

    3 The claim that the poliovirus sequence indeed functions as an IRES in these

    constructs was not proven rigorously. The proffered Northern blot is not

    sufficient to rule out a small amount of monocistronic mRNA, perhaps produced

    via a splice-donor sequence in the adenovirus portion of the vector and a splice-

    acceptor sequence in the poliovirus IRES. The negative control (5NC-X472)

    was not included in the Northern blot, and therefore it is possible that the

    diminished yield of Rluc from that construct reflects diminished production ofmRNA rather than inactivation of the IRES.

    4 The conclusion from these convoluted in vitro experiments was contra-

    dicted by a straightforward experiment wherein viral protein synthesis was

    monitored in NGP cells infected with Sabin-1 or wild-type poliovirus: there

    was no difference between the two viruses. Haller et al. (1996) say this

    unexpected result increases the value of experiments using the X472 revertants

    because these viruses have neuronal cell-specific defects in translation that are

    even more pronounced than those of the Sabin type 1 attenuated strain.

    Readers might disagree about the value of substituting laboratory constructs forthe real thing.

    197M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    5/10

    cells is contradicted by studies with human cells. Instead of a

    ribosomal deficiency, cells derived from DC patients were

    found to be deficient in telomerase RNA (Wong and Collins,

    2006).

    Another recent study postulates a defect in internal initiation

    as a step in tumorigenesis (Wilker et al., 2007). The hypothesis is

    that a protein called 14-3-3, which is frequently lost in tumors,normally stimulates IRES-mediated translation of the cyclin-

    dependent kinase PITSLRE. The authors assumed, without

    retesting, that PITSLRE is translated via an IRES. [Although

    early experiments suggestive of internal initiation looked

    convincing (Cornelis et al., 2000), the apparent IRES activity

    was much lower in a follow up study (Tinton et al., 2005). Thus,

    this essentialpoint requires verification.] In support of their claim

    that IRES-dependent translation is impaired in cellslacking 14-3-

    3, Wilker et al. (2007) show experiments carried out only with

    the dubious p27Kip1 IRES (Liu et al., 2005) and an even more

    dubious IRES from human immunodeficiency virus (HIV).5 The

    bottom line is that it might be true that the aberrant phenotype of14-3-3-depleted cells results fromabsence of a short form (p58)

    of PITSLRE, but p58 is probably translated from a spliced

    mRNA (Xiang et al., 1994) rather than via internal initiation.

    4.3. False deductions about the frequency of internal initiation

    Faulty reasoning underlies the prediction that a sizeable

    proportion of cellular mRNAs, perhaps as much as 35%is

    likely to contain an IRES (Holcik and Sonenberg, 2005). That

    number is based on the finding that 35% of cellular mRNAs

    remain polysome-associated in poliovirus-infected cells, wherein

    eIF4G undergoes cleavage (Johannes et al., 1999). But the belief

    that there is a link between cleavage of eIF4G and inhibition ofcap-dependent protein synthesis is not well-founded: it is based on

    correlative evidence from infected HeLa cells; but in some other

    cell lines, host protein synthesis persists at normal levels even

    after nearly-complete cleavage of eIF4G (Yanagiya et al., 2005;

    5 hr time point).6 In one system where experimentally-induced

    cleavage of eIF4G did reduce protein synthesis, the residual

    translation was traced to localization of certain mRNAs in the

    endoplasmic reticulum rather than to a special cap-independent

    mechanism of translation (Lerner and Nicchitta, 2006).

    The disconcerting absence of shared or similar sequences

    among putative IRESs (Baird et al., 2006) is sometimes excused

    by postulating that IRESs might instead have a common

    secondary structure. Baird et al. (2007) undertook a search for

    new IRESs based on that idea. But the search started from a

    flawed premise the hypothetical folded structure proposed for

    the XIAP IRES turned out not to be important for function

    and the search uncovered no evidence for a common structure

    among cellular IRESs. The absence of structural criteria thus

    allows almost anything to be called an IRES (e.g. Dorokhovet al., 2002; Terenin et al., 2005).

    Re-examination of many putative IRESs revealed cryptic

    promoters or splice sites which undermine use of the dicistronic

    test for internal initiation (Bert et al., 2006; Elango et al., 2006;

    Kozak, 2005; Wang et al., 2005). This has taught editors to

    scrutinize manuscripts more carefully, but the scrutiny still is not

    careful enough. In some cases where authors say cryptic promoters

    were ruled out, the test was not performed properly.7 The examples

    in Table 1 reveal other ways of failing to prove while claiming to

    prove that dicistronic vectors function as intended. The simplest

    way for editors to demand more stringent criteria would be to ban

    use of the popular pRF vector. The finding that detection of IRESactivity (even with poliovirus RNA) depends on the arrangement

    of reporter genes in this vector (Hennecke et al., 2001) isa warning

    that should not be ignored. The simplest explanation is that the

    5 Rluc gene harbors a splice-donor sequence (verified by Van

    Eden et al., 2004a) and the putative IRES actually functions by

    contributing a splice-acceptor sequence.

    There is growing use oftwo new indirect assays which do not

    really rule out splicing. Van Eden et al. (2004a) say IRES activity

    can be verified and splicing ruled out by demonstrating that

    siRNA against the 5 Rluc cistron causes equivalent reduction of

    Fluc expression from the 3 cistron. But the fact that Fluc was

    resistant to inhibition by siRNARluc in one case where Fluc was

    translated from a spliced mRNA (Van Eden et al., 2004a, Fig. 1E)does not necessarily mean that splicing can be ruled out in every

    case where Fluc expression is inhibited by siRNARluc. The timing

    of events (splicing vs. cleavage by siRNA) might vary from case

    to case, and therefore the effect of siRNARluc on Fluc cannot be

    predicted. The second assay does not involve siRNA; it simply

    requires quantifying the mRNA derived from each cistron. Holcik

    et al. (2005) say splicing can be ruled out if there is no

    diminishment in the amount of RNA from the 5 cistron, reasoning

    that production of a spliced transcript that preserves the 3 cistron

    requires eliminating the 5 cistron. Although this subtraction

    assay indeed detected the efficient splicing of Rluc/Fluc

    transcripts caused by the putative XIAP IRES, the subtractionassay is not likely to detect a small amount of splicing which

    would be sufficient to give the appearance of weak IRES activity.

    In short, a positive result using one of these indirect assays might

    indicate the occurrence of splicing, but a negative result does not

    rule it out.

    Advocates invoke indirect evidence when attempting to

    prove the occurrence of internal initiation, but they overlook the

    implications of indirect evidence when it works against them.

    5 The putative IRES from HIV barely functioned when tested in vitro by

    Brasey et al. (2003). It was more active when dicistronic constructs were tested

    in vivo, but there were no RNA analyses to rule out splicing, and the IRES

    activity inexplicably depended on using Fluc as the reporter. Wilker et al.

    (2007) describe their own results with the HIV IRES in a way that hides the

    inefficiency: the Fluc/Rluc ratio is simply set at 1.0 in control cells prior to

    onset of mitosis.6 This might be because eIF4G is normally present in excess (as suggested by

    the knockdown experiment in Montero et al., 2006), or because eIF4G cleavage

    fragments can still support cap-dependent translation (Ali et al., 2001), or

    because the observed cleavage of eIF4G occurs during cell lysis rather than in

    vivo. Invoking a second form of eIF4G (eIF4GII) is not a convincing

    explanation for the continued translation, given the low abundance of thatisoform.

    7 Instead of deleting the SV40 promoter from the dicistronic pRF vector, the

    putative IRES was transposed to a promoterless monocistronic vector (Petz et al.,

    2007; Yang et al., 2006). This is not valid because enhancer elements in the

    dicistronic construct might be needed to activate a cryptic promoter (Bert et al.,2006).

    198 M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    6/10

    I refer to the repeated discovery of splicing factors that stimulate

    one or another putative IRES (Bedard et al., 2007; Bushell et al.,

    2006; Holcik et al., 2003; Lin et al., 2007). This might be

    grounds for suspecting that the putative IRESs really work via

    production of spliced transcripts.

    5. Misunderstood tools invite more misunderstandings

    5.1. Some viral IRESs (if they are IRESs) are too weak to serve

    as controls

    There is growing use of putative IRESs from cricket

    paralysis virus (CrPV) and hepatitis C virus (HCV) as tools to

    evaluate other candidate IRESs (Li et al., 2006) and to probe

    other ideas about translation (Colon-Ramos et al., 2006;

    Gorgoni et al., 2005; Humphreys et al., 2005; Petersen et al.,

    2006). This is unwise, given the poor function and uncertain

    mechanisms of these viral RNAs.8

    The CrPV IRES works only 2% as efficiently as a normalcapped mRNA (Humphreys et al., 2005). The HCV IRES also

    supports only low-level translation of the 3 cistron when

    dicistronic mRNAs are tested in vitro (e.g. Miyakawa et al.,

    2006, Fig. 4B; Van Eden et al., 2004b, Fig. 2B). When tested by

    RNA transfection, dicistronic mRNA containing the HCV IRES

    supported only threefold better translation of the 3 cistron than a

    control with no IRES (Van Eden et al., 2004b). [The HCV IRES

    appears to work better when tested by DNA transfection, but that

    can be attributed to cryptic promoter activity (Dumas et al.,

    2003).] In many studies, the poor performance of the HCV or

    CrPV IRES is concealed by setting the yield at 1.0 without

    reference to any other mRNA (e.g. Berlanga et al., 2006; Coller

    and Parker, 2005; Colon-Ramos et al., 2006; Gorgoni et al.,2005; Holz et al., 2005; Lancaster et al., 2006; Moerke et al.,

    2007; Petersen et al., 2006; Yoon et al., 2006).

    5.2. Edeine does not inhibit polypeptide elongation

    Experiments in which edeine is used to distinguish between

    standard and special initiation mechanisms often are misinter-

    preted. This is an important issue in evaluating claims about

    CrPV. The undisputed finding is that translation of CrPV is

    inhibited by edeine at a concentration of 2 M (Matsuda and

    Dreher, 2007, Fig. 2B; Wilson et al., 2000, Fig. 3K). When

    people want to argue that a special edeine-resistant mechanismof initiation operates, however, they dismiss that result by

    claiming a high concentration of edeine (e.g. 2 M) inhibits the

    elongation step. That simply is not true. Edeine is used

    routinely at a concentration of 2 or 5 M to inhibit initiation and

    thus allow study of the elongation step (Fang et al., 2004;

    Horsburgh et al., 1996; Mothes et al., 1997; Somogyi et al.,

    1993). In a classic experiment undertaken to test the point

    directly, Hunt (1974) showed that 2 M edeine inhibits

    initiation and not elongation of globin polypeptide chains.The unexpected finding that CrPV-mediated translation is

    not inhibited by lower concentrations of edeine (e.g. 0.5 M)

    might be an artifact caused by the low efficiency of the putative

    IRES: translation is probably limited, not by the availability of

    ribosomes, but by the small amount of functional (broken)

    mRNA. Not until the entire pool of ribosomes is inactivated

    (at 2 M edeine) would translation of CrPV be affected.

    Lancaster et al. (2006) defend the idea that edeine inhibits

    elongation by citing studies by Szer and Kurylo-Borowska

    (1970) and Carrasco et al. (1974). The first of these papers reports

    that edeine inhibits elongation in a bacterial system, which is not

    relevant to how the drug works in eukaryotes. The second papertested effects of edeine using poly(U) as the template, which is

    not relevant to how the drug works with natural mRNAs.

    5.3. Indirect tests and blind use of reagents contribute to

    misunderstandings about microRNAs and IRESs

    A precision tool should not be used like a hammer. It is

    worrisome to see reports wherein rapidly-sedimenting com-

    plexes are judged to be polysomes based on their disappear-

    ance from cells exposed for a very long time (3 h) to a very high

    concentration of puromycin (1 mM) (Vasudevan and Steitz,

    2007). This exceeds the conditions routinely used to dissociate

    polysomes [e.g. incubation of HeLa cells with 200 Mpuromycin for 45 min (Katze et al., 1986)], and even those

    conditions are excessive. Nottrott et al. (2006) found nearly

    complete release of mRNAs from polysomes after cells were

    incubated with 5 M puromycin for only 20 min.

    In another attempt to study the mechanism of inhibition by

    microRNAs, Thermann and Hentze (2007) used a crude

    translation system from Drosophila embryos which they say

    works only if the reporter mRNA is pre-incubated in the extract

    for three hours. They say nothing about what happens during

    that long interval. The experiments can be faulted on many

    other grounds, e.g. the rapidly-sedimenting pseudo-poly-

    somes, purportedly induced by microRNA, were seen evenunder conditions where translation was not repressed. But it

    seems trivial to fuss about such problems when the experiment

    begins with a three-hour black box step.

    IRESs are used as tools in many studies with microRNAs.

    The reasoning is that, if IRES-mediated translation is not

    inhibited, microRNAs must inhibit via the cap. The test is

    indirect, at best; but it loses all meaning when the putative IRES

    barely works [the CrPV IRES was only 2% as efficient as

    capped mRNA in the study by Humphreys et al. (2005)] and

    when the answer varies depending on the transfection technique

    (Lytle et al., 2007).

    IRES activity itself is sometimes claimed based only on

    indirect tests; e.g. demonstrating that translation is cap-

    8 Advocates claim to have proven that the CrPV IRES forms a functional

    complex with ribosomes; i.e. they claim the complexes (assembled without

    initiation factors and Met-tRNAi) go on to produce peptides or peptidyl-

    puromycin (Cevallos and Sarnow, 2005, Fig. 8; Jan et al., 2003; Pestova and

    Hellen, 2005). But those experiments were not convincing because there was

    no positive control against which to judge the efficiency of the reaction with

    CrPV mRNA. Without convincing evidence of function, elegant structural

    models cannot be interpreted. The need to use yeast ribosomes and a high

    concentration of Mg

    2+

    to obtain stable complexes with CrPV RNA (Schuleret al., 2006) adds to the worry about what the structural models really mean.

    199M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    7/10

    independent (not inhibited by a cap analog; Marr et al., 2007) or

    not inhibited when a hairpin structure is appended to the 5 end

    of the mRNA. These tests are meaningless because translation

    from broken transcripts also would be unaffected by those

    manipulations. Broken transcripts that lack a cap cannot be

    translated efficiently, but they can be translated (Bendena et al.,

    1985; Kozak, 1980; Lawrence, 1980; Pelham, 1979; Thoma

    et al., 2001). Thus, low-level translation of the 3 proximal

    cistron might be accomplished, not via internal initiation, but

    via cleavage of dicistronic mRNAs.9

    Possible cleavage of mRNA is an important issue vis--vis

    interpreting experiments with both IRESs and microRNAs.Investigators often try to address this concern by showing that

    intact transcripts can be recovered following liposome-mediated

    transfection, but the recent finding thatmostof the RNA introduced

    via liposomes does not enter the cell in a usable way (Barreau

    et al., 2006) precludes simple conclusions. When the structure and

    stability of the functional mRNA pool is at issue, analysis ofbulk

    mRNA recovered from transfected (or microinjected) cells is

    meaningless. Rulingout effects on mRNA stability can be difficult

    even when translation is studied in vitro, as in a recent attempt to

    study inhibition by microRNAs using extracts from Krebs cells

    (Mathonnet et al., in press). Sucrose gradient analysis of initiation

    complexes revealed that only 10% of the input mRNA actually

    participates in translation; and therefore, the absence of an effect

    on bulk mRNA stability does not rule out the possibility that

    microRNAs inhibit by cleaving (translationally active) mRNAs.

    6. Last thoughts

    Some problems are becoming more common as investigators

    (mis)learn from one anothercopying poor experimental

    designs (e.g. dicistronic vectors); repeating erroneous inter-

    pretations (e.g. that edeine inhibits elongation); predicating new

    models on dubious old models. The mistaken idea about

    translation of TYMV (Section 3.1) was inspired by a dubious

    model about Met-tRNAi-independent translation of CrPV. The

    mistaken idea about translation of rotavirus mRNA (Section 3.2)

    was inspired by the unproven closed-loop model.

    A frequent problem is deciding the answer first and then

    constructing a test system to show what one wants to see, as in

    the selection of a neuronal cell line that showed the expected

    9 Some picornavirus IRESs contain sequences that are specifically recognized

    by RNases (Elgadi and Smiley, 1999; Serrano et al., 2007), but random attack is

    probably the more common mechanism. Rosenfeld and Racaniello (2005)

    claimed that cleavage of dicistronic mRNA was ruled out in their study because

    translation of the 3 cistron was not reduced by mutations that inactivate the

    nonsense-mediated decay pathway, but that pathway is not the only source of

    RNases in yeast. In the case of HCV, cleavage might even occur bymagnesium-catalyzed autohydrolysis (Kieft et al., 1999).

    Table 1

    Recent poorly-documented claims of IRES activity in 5 UTR sequences from cellular mRNAs

    mRNA Problems with the evidence obtained via dicistronic vectors a References

    Cryptic promoter? Splicing checked? Activity confirmed via RNA

    transfection?

    XIAP (intron)b Ruled out. Splicing was detected with

    pRF vectorcVery weak (twofold greater

    than empty vector).

    Van Eden et al. (2004a)

    NF-B repressing factor (intron) b Ruled out. siRNA against 5 Rluc also

    inhibits 3 Fluc.

    Not tested. Reboll et al. (2007)

    p27Kip1 Detected but

    deemed minimal. dsiRNA against 5 Rluc also

    inhibits 3 Fluc.

    Absence of a positive control

    precludes conclusion.eJiang et al. (2007)

    AANAT f Ruled out. Northern blot is insufficient

    to rule out splicing.

    Not tested. Kim et al. (2007)

    Cyclin T1, SET7, SIAH2, ZNF217, etc. Ruled out. Northern blot is insufficient

    to rule out splicing. gNot tested. Bushell et al. (2006)

    a The splicing-prone pRF vector (5-Rluc-Fluc-3) was used for all these studies. Section 4.3 explains why the use of siRNA, targeted to the 5 cistron, is not a valid

    assay for ruling out production of spliced (monocistronic) mRNAs from the dicistronic vector.b cDNA analyses revealed that the long 5 UTR, here tested for IRES activity, is really an intron (Jianfeng et al., 2003). With XIAP also, the 3 end of the putative

    IRES harbors a splice acceptor site that functions naturally, i.e. the IRES is an intron ( Van Eden et al., 2004a).c Holcik et al. (2005) claim that with a different dicistronic vector (gal/CAT), the XIAP sequence supports translation in the absence of splicing, judging from the

    fact that there was no reduction in the amount ofgal RNA. But this experiment had no positive control to show that the predicted reduction in gal can be detectedwhen a verified intron is introduced.d The detected activity, equal to 15% of the strong SV40 promoter, is arguably more than minimal.e The efficiency of expression from the dicistronic transcript needs to be compared to a capped monocistronic mRNA (positive control). Claims of IRES activity

    based only on comparison to a negative control are meaningless because it is too easy to devise a negative control (in this case, an antisense version of the p27Kip1 5

    UTR) that makes the sequence of interest look good. In RNA transfection experiments with a different candidate IRES, Van Eden et al. (2004b, Fig. 5C) demonstrated

    vastly different levels of expression of the 3 cistron with three different negative controls (random insertNempty vectorantisense insert). It makes no sense to use the

    worst of these as the control against which IRES activity is calculated.f Insertion of the putative IRES from arylalkylamine N-acetyltransferase (AANAT) mRNA produced only a fourfold increase in translation of Fluc relative to the

    empty pRF vector. Thus, the putative IRES is very weak, as shown also by comparing translational yields from the 3 and 5 cistrons (Flucb2% of Rluc).g Bushell et al. (2006) argue that Northern blotting, albeit insensitive, really is the best method to check RNA structure because RT-PCR fails when the wrong

    primer is used. Internal initiation was proven, they say, by demonstrating that dicistronic mRNA is polysome-associated even when translation of the first cistron is

    blocked; but the fine-print reveals the RNA analyses were done on fractions pooled from polysomal and 80S regions of the gradient, which makes the claim

    meaningless.

    200 M. Kozak / Gene 403 (2007) 194203

    http://dx.doi.org/doi:10.1126/science.1146067http://dx.doi.org/doi:10.1126/science.1146067
  • 7/28/2019 Lessons Not Learned Fromtranslation

    8/10

    (wrongly expected) discrimination against vaccine strains of

    poliovirus (Section 4.1). The problem of allowing expectations

    to dictate the outcome can be seen also in a recent study of the

    mechanism whereby translation of PRMT5 is elevated in

    lymphoid cancer cells. Pal et al. (2007) decided a priori to

    look for microRNAs targeted to PRMT5 mRNA, without having

    ruled out simpler ways to explain the up-turn in translation; e.g.although they checked for (and ruled out) an increase in the

    amount of mRNA, they did not check for a possible change in

    structure of the 5 UTR.10 Upon finding that addition of the

    candidate microRNAs to an in vitro translation system failed to

    inhibit translation of PRMT5, the authors simply changed the

    sequence of the microRNAs to make up for the lack of

    stability.

    The internal initiation story is the most egregious example of

    deciding the answer first. Unlike the poliovirus vaccine story,

    where people selected evidence that fit expectations, the

    evidence for internal initiation was invented rather than selected.

    This was not done in a fraudulent way, of course; it was done byadopting such weak criteria for recognizing IRES activity that

    any sequence of interest could be called positive.11 The

    starting belief (pre-decided answer) was that critical regulatory

    genes e.g. genes expressed during mitosis or apoptosis need

    a special mechanism for initiating translation (Holcik and

    Sonenberg, 2005). This prompted investigators to test important

    regulatory genes for IRES activity; and because of the weak

    criteria (and use of the pRF vector), almost every test worked.

    Advocates say the belief has been vindicated. Critics are still

    watching for the first scrap of real evidence.

    References

    Agol, V.I., Drozdov, S.G., Ivannikova, T.A., Kolesnikova, M.S., Korolev, M.B.,

    Tolskaya, E.A., 1989. Restricted growth of attenuated poliovirus strains in

    cultured cells of a human neuroblastoma. J. Virol. 63, 40344038.

    Ali, I.K., McKendrick, L., Morley, S.J., Jackson, R.J., 2001. Truncated initiation

    factor eIF4G lacking an eIF4E binding site can support capped mRNA

    translation. EMBO J. 20, 42334242.

    Baird, S.D., Turcotte, M., Korneluk, R.G., Holcik, M., 2006. Searching for

    IRES. RNA 12, 17551785.

    Baird, S.D., Lewis, S.M., Turcotte, M., Holcik, M., 2007. A search for

    structurally similar cellular internal ribosome entry sites. Nucleic Acids Res.

    35, 46644677.

    Barends, S.,Bink,H.H.J., vanden Worm, S.H.E., Pleij, C.W.A., Kraal, B.,2003.

    Entrapping ribosomes for viral translation: tRNA mimicry as a molecular

    Trojan horse. Cell 112, 123129.

    Barreau, C., Dutertre, S., Paillard, L., Osborne, H.B., 2006. Liposome-mediated

    RNA transfection should be used with caution. RNA 12, 17901793.

    Bedard, K.M., Daijogo, S., Semler, B.L., 2007. A nucleo-cytoplasmic SR

    protein functions in viral IRES-mediated translation initiation. EMBO J. 26,

    459467.

    Bendena, W.G., Abouhaidar, M., Mackie, G.A., 1985. Synthesis in vitro of the

    coat protein of papaya mosaic virus. Virology 140, 257268.

    Berlanga, J.J., Baass, A., Sonenberg, N., 2006. Regulation of PABP function in

    translation:characterizationof thePaip2homolog,Paip2B. RNA12, 15561568.

    Bert, A.G., Grepin, R., Vadas, M.A., Goodall, G.J., 2006. Assessing IRES

    activity in the HIF-1 and other cellular 5 UTRs. RNA 12, 10741083.

    Borman, A.M., Deliat, F.G., Kean, K.M., 1994. Sequences within the poliovirus

    internal ribosome entry segment control viral RNA synthesis. EMBO J. 13,

    31493157.

    Brasey, A., et al., 2003. The leader of HIV-1 genomic RNA harbors an IRES thatis active during the G2/M phase of the cell cycle. J. Virol. 77, 39393949.

    Bushell, M., et al., 2006. Polypyrimidine tract binding protein regulates IRES-

    mediated gene expression during apoptosis. Mol. Cell 23, 401412.

    Carrasco, L., Battaner, E., Vazquez, D., 1974. The elongation steps in protein

    synthesis by eukaryotic ribosomes: effects of antibiotics. Methods Enzymol.

    30, 282289.

    Cevallos, R.C., Sarnow, P., 2005. Factor-independent assembly of elongation-

    competent ribosomes by an IRES located in an RNA virus that infects

    penaeid shrimp. J. Virol. 79, 677683.

    Coller, J., Parker, R., 2005. General translational repression by activators of

    mRNA decapping. Cell 122, 875886.

    Colon-Ramos, D.A., et al., 2006. Direct ribosomal binding by a cellular inhibitor

    of translation. Nat. Struct. Mol. Biol. 13, 103111.

    Cornelis, S., Bruynooghe, Y., Denecker, G., Van Huffel, S., Tinton, S., Beyaert,

    R., 2000. Identification and characterization of a novel cell cycle-regulatedIRES. Mol. Cell 5, 597605.

    Dorokhov, Y.L., et al., 2002. Polypurine (A)-rich sequences promote cross-

    kingdom conservationof internal ribosome entry. Proc.Natl.Acad. Sci. U. S. A.

    99, 53015306.

    Dorrello, N.V., Peschiaroli, A., Guardavaccaro, D., Colburn, N.H., Sherman, N.E.,

    Pagano, M., 2006. S6K1-and TRCP-mediated degradation of PDCD4

    promotes protein translation and cell growth. Science 314, 467471.

    Dumas, E., et al., 2003. A promoter activity is present in the DNA sequence

    corresponding to the hepatitis C virus 5 UTR. Nucleic Acids Res. 31,

    12751281.

    Duncan, R., McConkey, E.H., 1982. Preferential utilization of phosphorylated 40S

    ribosomal subunits during initiation complex formation. Eur. J. Biochem. 123,

    535538.

    Elango, N., Li, Y., Shivshankar, P., Katz, M.S., 2006. Expression of RUNX2

    isoforms: involvement of cap-dependent and cap-independent mechanismsof translation. J. Cell. Biochem. 99, 11081121.

    Elgadi, M.M., Smiley, J.R., 1999. Picornavirus IRES elements target RNA

    cleavage events induced by the herpes simplex virus virion host shutoff

    protein. J. Virol. 73, 92229231.

    Fang, P., Spevak, C.C., Wu, C., Sachs, M.S., 2004. A nascent polypeptide domain

    that can regulate translation elongation. Proc. Natl. Acad. Sci. U. S. A. 101,

    40594064.

    Flint, S.J., Enquist, L.W., Racaniello, V.R., Skalka, A.M., 2004. Principles of

    Virology, 2nd ed. ASM Press, Washington, D.C.

    Gorgoni, B., Andrews, S., Schaller, A., Schumperli, D., Gray, N.K., Muller, B.,

    2005. The stem-loop binding protein stimulates histone translation at an

    early step in the initiation pathway. RNA 11, 10301042.

    Gressner, A.M., Wool, I.G., 1974. The phosphorylation of liver ribosomal

    proteins in vivo. J. Biol. Chem. 249, 69176925.

    Guest, S., Pilipenko, E., Sharma, K., Chumakov, K., Roos, R.P., 2004.Molecular mechanisms of attenuation of the Sabin strain of poliovirus type

    3. J. Virol. 78, 1109711107.

    Gutierrez, A.L., Denova-Ocampo, M., Racaniello, V.R., del Angel, R.M., 1997.

    Attenuating mutations in the poliovirus 5 UTR alter its interaction with

    polypyrimidine tract-binding protein. J. Virol. 71, 38263833.

    Haller, A.A., Stewart, S.R., Semler, B.L., 1996. Attenuation stem-loop lesions in

    the 5 noncoding region of poliovirus RNA: neuronal cell-specific

    translation defects. J. Virol. 70, 14671474.

    Hannan, K.M., et al., 2003. mTOR-dependent regulation of ribosomal gene

    transcription requires S6K1 andis mediated by phosphorylation of theC-terminal

    activation domain of the nucleolar transcription factor UBF. Mol. Cell. Biol. 23,

    88628877.

    Hennecke, M., et al., 2001. Composition and arrangement of genes define the

    strength of IRES-driven translation in bicistronic mRNAs. Nucleic Acids

    Res. 29, 33273334.

    10 This is a reasonable possibility given that cDNAs with different 5 UTRs

    have been described for PRMT5 (GenBank NM_006109 and NM_001039619).

    The second entry has a small upstream ORF which is predicted to reduce

    translational efficiency.11Weak criteria means there are no structural requirements, and the slightest

    increase in expression of the 3

    cistron (compared to some arbitrary referencepoint) suffices.

    201M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    9/10

    Holcik, M., Sonenberg, N., 2005. Translational control in stress and apoptosis.

    Nat. Rev., Mol. Cell Biol. 6, 318327.

    Holcik, M., Gordon, B.W., Korneluk, R.G., 2003. The IRES-mediated

    translation of anti-apoptotic protein XIAP is modulated by the hnRNP

    proteins C1 and C2. Mol. Cell. Biol. 23, 280288.

    Holcik, M., Graber, T., Lewis, S.M., Lefebvre, C.A., Lacasse, E., Baird, S.,

    2005. Spurious splicing within the XIAP 5 UTR occurs in the Rluc/Fluc but

    not the gal/CAT bicistronic reporter system. RNA 11, 16051609.Holz, M.K., Ballif, B.A., Gygi, S.P., Blenis, J., 2005. mTOR and S6K1 mediate

    assembly of the translation preinitiation complex through dynamic protein

    interchange and ordered phosphorylation events. Cell 123, 569580.

    Horsburgh, B.C., Kollmus, H., Hauser, H., Coen, D.M., 1996. Translational

    recoding induced by G-rich mRNA sequences that form unusual structures.

    Cell 86, 949959.

    Humphreys, D.T., Westman, B.J., Martin, D.I.K., Preiss, T., 2005. MicroRNAs

    control translation initiation by inhibiting eIF4E/cap and poly(A) tail

    function. Proc. Natl. Acad. Sci. U. S. A. 102, 1696116966.

    Hunt, T., 1974. The control of globin synthesis in rabbit reticulocytes. Ann. N.Y.

    Acad. Sci. 241, 223231.

    Hunziker, I.P., Cornell, C.T., Whitton, J.L., 2007. Deletions within the 5 UTR

    of coxsackievirus B3: consequences for virus translation and replication.

    Virology 360, 120128.

    Jan, E., Kinzy, T.G., Sarnow, P., 2003. Divergent tRNA-like element supportsinitiation, elongation, and termination of protein biosynthesis. Proc. Natl.

    Acad. Sci. U. S. A. 100, 1541015415.

    Jefferies, H.B.J., Fumagalli, S., Dennis, P.B., Reinhard, C., Pearson, R.B.,

    Thomas, G., 1997. Rapamycin suppresses 5 TOP mRNA translation

    through inhibition of p70s6k. EMBO J. 16, 36933704.

    Jianfeng, D., et al., 2003. Cloning of the correct full-length cDNA of NF- B

    repressing factor. Mol. Cells 16, 397401.

    Jiang, H., Coleman, J., Miskimins, R., Srinivasan, R., Miskimins, W.K., 2007.

    Cap-independent translation through the p27 5 UTR. Nucleic Acids Res.

    35, 47674778.

    Johannes, G., Carter, M.S., Eisen, M.B., Brown, P.O., Sarnow, P., 1999.

    Identificationof eukaryotic mRNAs that aretranslated at reducedcap binding

    complex eIF4F concentrations using a cDNA microarray. Proc. Natl. Acad.

    Sci. U. S. A. 96, 1311813123.

    Johansen, L.K., Morrow, C.D., 2000. The RNA encompassing the IRES in thepoliovirus 5 nontranslated region enhances the encapsidation of genomic

    RNA. Virology 273, 391399.

    Katze, M.G., DeCorato, D., Krug, R.M., 1986. Cellular RNA translation is

    blocked at both initiation and elongation after infection by influenza virus or

    adenovirus. J. Virol. 60, 10271039.

    Kauder, S.E., Racaniello, V.R., 2004. Poliovirus tropism and attenuation are

    determined after internal ribosome entry. J. Clin. Invest. 113, 17431753.

    Kieft, J.S., Zhou, K., Jubin, R., Murray, M.G., Lau, J.Y.N., Doudna, J.A., 1999.

    Thehepatitis C virus IRESadopts an ion-dependent tertiaryfold. J. Mol. Biol.

    292, 513529.

    Kim, T.-D., Woo, K.-C., Cho, S., Ha, D.-C., Jang, S.K., Kim, K.-T., 2007.

    Rhythmic control of AANAT translation by hnRNP Q in circadian melatonin

    production. Genes Dev. 21, 797810.

    Kozak, M., 1980. Binding of wheat germ ribosomes to fragmented viral mRNA.

    J. Virol. 35, 748756.Kozak, M., 2003. Alternative ways to think about mRNA sequences and

    proteins that appear to promote internal initiation of translation. Gene 318,

    123.

    Kozak, M., 2004. How strong is the case for regulation of the initiation step of

    translation by elements at the 3 end of eukaryotic mRNAs? Gene 343,

    4154.

    Kozak, M., 2005. A second look at cellular mRNA sequences said to function as

    internal ribosome entry sites. Nucleic Acids Res. 33, 65936602.

    Kozak, M., 2006. Rethinking some mechanisms invoked to explain translational

    regulation in eukaryotes. Gene 382, 111.

    La Monica, N., Racaniello, V.R., 1989. Differences in replication of attenuated

    and neurovirulent polioviruses in human neuroblastoma cell l ine SH-SY5Y.

    J. Virol. 63, 23572360.

    Lancaster, A.M., Jan, E., Sarnow, P., 2006. Initiation factor-independent

    translation mediated by the hepatitis C virus IRES. RNA 12, 894902.

    LaRonde-LeBlanc, N., Santhanam, A.N., Baker, A.R., Wlodawer, A., Colburn,

    N.H., 2007. Structural basis for inhibition of translation by the tumor

    suppressor Pdcd4. Mol. Cell. Biol. 27, 147156.

    Lawrence, C.B., 1980. Activation of an internal initiation site for protein

    synthesis during in vitro translation. Nucleic Acids Res. 8, 13071317.

    Lerner, R.S., Nicchitta, C.V., 2006. mRNA translation is compartmentalized to the

    endoplasmic reticulum following physiological inhibition of cap-dependent

    translation. RNA 12, 775789.Levy, S., Avni, D., Hariharan, N., Perry, R.P., Meyuhas, O., 1991. Oligopyr-

    imidine tract at the 5 end of mammalian ribosomal protein mRNAs is

    required for their translational control. Proc. Natl. Acad. Sci. U. S. A. 88,

    33193323.

    Li, P.W.-L., Li, J., Timmerman, S.L., Krushel, L.A., Martin, S.L., 2006. The

    dicistronic RNA from the mouse LINE-1 retrotransposon contains an IRES

    upstream of each ORF: implications for retrotransposition. Nucleic Acids

    Res. 34, 853864.

    Lin, J.-C., Hsu, M., Tarn, W.-Y., 2007. Cell stress modulates the function of

    splicing regulatory protein RBM4 in translation control. Proc. Natl. Acad.

    Sci. U. S. A. 104, 22352240.

    Liu, Z., Dong, Z., Han, B., Yang, Y., Liu, Y., Zhang, J.-T., 2005. Regulation of

    expression by promoters versus IRES in the 5-untranslated sequence of the

    human cyclin-dependent kinase inhibitor p27kip1. Nucleic Acids Res. 33,

    37633771.Lytle, J.R., Yario, T.A., Steitz, J.A., 2007. Target mRNAs are repressed as

    efficiently by microRNA-binding sites in the 5 UTR as in the 3 UTR. Proc.

    Natl. Acad. Sci. U. S. A. 104, 96679672.

    Marr, M.T., D'Alessio, J.A., Puig, O., Tjian, R., 2007. IRES-mediated

    functional coupling of transcription and translation amplifies insulin

    receptor feedback. Genes Dev. 21, 175183.

    Mastropaolo, W., Henshaw, E.C., 1981. Phosphorylation of ribosomal protein

    S6 in the Ehrlich ascites tumor cell. Lack of effect of phosphorylation upon

    ribosomal function in vitro. Biochim. Biophys. Acta 656, 246255.

    Mathonnet, G. et al., in press. MicroRNA inhibition of translation initiation in

    vitro by targeting the cap-binding complex eIF4F. Science. doi:10.1126/

    science.1146067.

    Matsuda, D., Dreher, T.W., 2006. Close spacing of AUG initiation codons

    confers dicistronic character on a eukaryotic mRNA. RNA 12, 13381349.

    Matsuda, D., Dreher, T.W., 2007. Cap-and initiator tRNA-dependent initiationof TYMV polyprotein synthesis by ribosomes: evaluation of the Trojan

    horse model for TYMV translation. RNA 13, 129137.

    Miyakawa, S., Oguro, A., Ohtsu, T., Imataka, H., Sonenberg, N., Nakamura, Y.,

    2006. RNAaptamers to mammalian IF4G inhibit cap-dependent translationby

    blocking the formation of initiation factor complexes. RNA 12, 18251834.

    Moerke, N.J., et al., 2007. Small-molecule inhibition of the interaction between

    the translation initiation factors eIF4E and eIF4G. Cell 128, 257267.

    Montero, H., Arias, C.F., Lopez, S., 2006. Rotavirus nonstructural protein NSP3

    is not required for viral protein synthesis. J. Virol. 80, 90319038.

    Mothes, W., et al., 1997. Molecular mechanism of membrane protein integration

    into the endoplasmic reticulum. Cell 89, 523533.

    Nottrott, S., Simard, M.J., Richter, J.D., 2006. Human let-7a miRNA blocks

    protein production on actively translating polyribosomes. Nat. Struct. Mol.

    Biol. 13, 11081114.

    Pal, S., Baiocchi, R.A., Byrd, J.C., Grever, M.R., Jacob, S.T., Sif, S., 2007. Lowlevels of miR-92B/96 induce PRMT5 translation and H3R8/H4R3

    methylation in mantle cell lymphoma. EMBO J. 26, 35583569.

    Pelham, H.R.B., 1979. Translation of fragmented viral RNA in vitro. FEBS Lett.

    100, 195199.

    Pende, M., et al., 2004. S6K1//S6K2/ mice exhibit perinatal lethality and

    rapamycin-sensitive 5-TOP mRNA translation and reveal a mitogen-activated

    protein kinase-dependent S6K pathway. Mol. Cell. Biol. 24, 31123124.

    Pestova, T.V., Hellen, C.U.T., 2005. Reconstitution of eukaryotic translation

    elongation in vitro following initiation by internal ribosomal entry. Methods

    36, 261269.

    Petersen, C.P., Bordeleau, M.-E., Pelletier, J., Sharp, P.A., 2006. Short RNAs

    repress translation after initiation in mammalian cells. Mol. Cell 21,

    533542.

    Petz, M., et al., 2007. The leader region of laminin B1 mRNA confers cap-

    independent translation. Nucleic Acids Res. 35, 24732482.

    202 M. Kozak / Gene 403 (2007) 194203

  • 7/28/2019 Lessons Not Learned Fromtranslation

    10/10

    Piron, M., Vende, P., Cohen, J., Poncet, D., 1998. Rotavirus RNA-binding

    protein NSP3 interacts with eIF4GI and evicts the poly(A) binding protein

    from eIF4F. EMBO J. 17, 58115821.

    Raught, B., et al., 2004. Phosphorylation of eIF4B Ser422 is modulated by S6

    kinases. EMBO J. 23, 17611769.

    Reboll, M.R., et al., 2007. NRF IRES activity is mediated by RNA binding

    protein JKTBP1 and a 14-nt RNA element. RNA 13, 13281340.

    Rosenfeld, A.B., Racaniello, V.R., 2005. Hepatitis C virus IRES-dependenttranslation in Saccharomyces cerevisiae is independent of PTB, PCBP2, and

    La protein. J. Virol. 79, 1012610137.

    Ruvinsky, I., et al., 2005. Ribosomal protein S6 phosphorylation is a determinant

    of cell size and glucose homeostasis. Genes Dev. 19, 21992211.

    Schuler, M., et al., 2006. Structure of the ribosome-bound cricket paralysis virus

    IRES RNA. Nat. Struct. Mol. Biol. 13, 10921096.

    Schwab, M.S., et al., 1999. p70S6K controls selective mRNA translation during

    oocyte maturation and early embryogenesis inXenopus laevis. Mol. Cell.Biol.

    19, 24852494.

    Serrano, P., Gomez, J., Martinez-Salas, E., 2007. Characterization of a

    cyanobacterial RNase P ribozyme recognition motif in the IRES of

    FMDV reveals a unique structural element. RNA 13, 849859.

    Shahbazian, D., et al., 2006. The mTOR/P13K and MAPK pathways converge

    on eIF4B to control its phosphorylation and activity. EMBO J. 25,

    27812791.Somogyi, P., Jenner, A.J., Brierley, I., Inglis, S.C., 1993. Ribosomal pausing

    during translation of an RNA pseudoknot. Mol. Cell. Biol. 13, 69316940.

    Szer, W., Kurylo-Borowska, Z., 1970. Effect of edeine on aminoacyl-tRNA

    binding to ribosomes and its relationship to ribosomal binding sites.

    Biochim. Biophys. Acta 224, 477486.

    Terada, N., Patel, H.R., Takase, K., Kohno, K., Nairn, A.C., Gelfand, E.W.,

    1994. Rapamycin selectively inhibits translation of mRNAs encoding

    elongation factors and ribosomal proteins. Proc. Natl. Acad. Sci. U. S. A. 91,

    1147711481.

    Terenin, I.M., et al., 2005. A cross-kingdom IRES reveals a simplified mode of

    internal ribosome entry. Mol. Cell. Biol. 25, 78797888.

    Thermann, R., Hentze, M.W., 2007. Drosophila miR2 induces pseudo-

    polysomes and inhibits translation initiation. Nature 447, 875878.

    Thoma, C., et al., 2001. Generation of stable mRNA fragments and translation of

    N-truncated proteins induced by antisense oligodeoxynucleotides. Mol. Cell8, 865872.

    Thomas, G., Siegmann, M., Kubler, A.-M., Gordon, J., Jimenez de Asua, L.,

    1980. Regulation of 40S ribosomal protein S6 phosphorylation in Swiss

    mouse 3T3 cells. Cell 19, 10151023.

    Tinton, S.A., Schepens, B., Bruynooghe, Y., Beyaert, R., Cornelis, S., 2005.

    Regulation of the cell-cycle-dependent IRES of the PITSLRE protein

    kinase: roles of Unr protein and phosphorylated eIF2. Biochem. J. 385,

    155163.

    Van Eden, M.E., Byrd, M.P., Sherrill, K.W., Lloyd, R.E., 2004a. Demonstrating

    internal ribosome entry sites in eukaryotic mRNAs using stringent RNA test

    procedures. RNA 10, 720730.

    Van Eden, M.E., Byrd, M.P., Sherrill, K.W., Lloyd, R.E., 2004b. Translation ofc-IAP1 mRNA is IRES mediated and regulated during cell stress. RNA 10,

    469481.

    Vasudevan, S., Steitz, J.A., 2007. AU-rich-element mediated upregulation of

    translation by FXR1 and Argonaute 2. Cell 128, 11051118.

    Vende, P., Piron, M., Castagne, N., Poncet, D., 2000. Efficient translation of

    rotavirus mRNA requires simultaneous interaction of NSP3 with the

    eukaryotic translation initiation factor eIF4G and the mRNA 3 end. J. Virol.

    74, 70647071.

    Wang, Z., Weaver, M., Magnuson, N.S., 2005. Cryptic promoter activity in the

    DNA sequence corresponding to the pim-1 5-UTR. Nucleic Acids Res. 33,

    22482258.

    Wilker, E.W., et al., 2007. 14-3-3 controls mitotic translation to facilitate

    cytokinesis. Nature 446, 329332.

    Wilson, J.E., Pestova, T.V., Hellen, C.U.T., Sarnow, P., 2000. Initiation of

    protein synthesis from the A site of the ribosome. Cell 102, 511520.Wong, J.M.Y., Collins, K., 2006. Telomerase RNA level limits telomere

    maintenance in X-linkeddyskeratosiscongenita. Genes Dev. 20,28482858.

    Wullschleger, S., Loewith, R., Hall, M.N., 2006. TOR signaling in growth and

    metabolism. Cell 124, 471484.

    Xiang, J., Lahti, J.M., Grenet, J., Easton, J., Kidd, V.J., 1994. Molecular cloning

    and expression of alternatively spliced PITSLRE protein kinase isoforms.

    J. Biol. Chem. 269, 1578615794.

    Yanagiya, A., Jia, Q., Ohka, S., Horie, H., Nomoto, A., 2005. Blockade of the

    poliovirus-induced cytopathic effect in neural cells by monoclonal antibody

    against poliovirus or the human poliovirus receptor. J. Virol. 79, 15231532.

    Yang, H.-S., et al., 2003. The transformation suppressor Pdcd4 is a novel eIF4A

    binding protein that inhibits t ranslation. Mol. Cell. Biol. 23, 2637.

    Yang, D.-Q., Halaby, M.-J., Zhang, Y., 2006. The identification of an IRES in

    the 5 UTR of p53 mRNA provides a novel mechanism for the regulation of

    its translation following DNA damage. Oncogene 25, 46134619.Yoon, A., et al., 2006. Impaired control of IRES-mediated translation in X-linked

    dyskeratosis congenita. Science 312, 902906.

    203M. Kozak / Gene 403 (2007) 194203