role of an upstream open reading frame in the translation of

© 1992 Oxford University Press Nucleic Acids Research, Vol. 20, No. 15 3851-3857

Role of an upstream open reading frame in the translationof polycistronic mRNAs in plant cells

Johannes Futterer+ and Thomas Hohn*Friedrich Miescher-lnstitute, PO Box 2543, CH-4002 Basel, Switzerland

Received May 28, 1992; Revised and Accepted July 13, 1992

ABSTRACT

The influence of an upstream small open reading frame(URF) on the translation of two consecutive codingregions on an eukaryotic mRNA was studied. The ciseffects of leader length, URF length, the sequences ofthe URF and neighboring regions, and the trans effectsof the Cauliflower mosaic virus transactivator (TAV)were analyzed. Translation efficiency of the immediatedownstream open reading frame (ORF) decreased withincreasing URF length. Short URFs did not drasticallyinhibit translation of immediate downstream ORFs butsupported far downstream translation in the presenceof TAV. In the latter case, the optimal URF length was30 codons.

INTRODUCTION

On the vast majority of eukaryotic mRNAs, ribosomes select thetranslation initiation site by a scanning mechanism (1). Scanningusually commences at the capped 5' end. As a consequence mostprotein-coding regions (open reading frames, ORFs) oneukaryotic mRNA species begin with the 5' proximal ATGcodon. However, 5 to 10% of eukaryotic mRNAs contain oneor more upstream ORFs (URFs) in the 5' 'untranslated' leadersequence (2). URFs are usually short and therefore also referredto as short ORFs (sORFs). The presence of URFs influences thetranslation of the downstream ORF(s) because—as current modelssuggest—initiation factors dissociate from the ribosome upon eachinitiation event (e.g. at the URF's start codon) and are no longeravailable for further downstream translation (3). Therefore, URFsoften inhibit downstream translation in proportion to the efficiencyof their own translation (4,5). Translation of an URF and thusits inhibitory potential is reduced when the sequence context ofits start codon is suboptimal (4,5), or when the start codon islocated close (< 10 nt) to the cap site (6,7). In a few cases, thecoding sequence of an URF was itself found to influence the effecton downstream translation (8,9).

It has been suggested that after translation of an URF, scanningcan be resumed and subsequently initiation competence isregained, probably by binding a new set of initiation factors. Sincethis process is time consuming, reinitiation efficiency dependson the distance between the upstream termination site and the

downstream initiation site (10). It has further been suggested thaton leader sequences with multiple URFs, the translation of thefirst URF enables ribosomes to pass further downstream URFsin an initiation incompetent state and only become competent intime to translate the main ORF (10). A good example for thismodel is the GCN4 mRNA (11,12), where regulated translationof the main mRNA coding region depends on the presence ofat least two URFs in the leader sequence, the first being requiredto overcome a translational block by the second (13). WithGCN4-derived mRNAs, it was shown that the quality of an URFor its surrounding sequence influences the efficiency ofdownstream translation initiation (14,15).

Another system described recently in which an URF enhancesthe translation of a far downstream ORF is the translation fromthe polycistronic 35S RNA of cauliflower mosaic virus (CaMV).This requires transact!vation by a CaMV-encoded protein (TAV).Transactivation is weak when the first ORF on an artificialpolycistronic mRNA is long, but it is strong when this long ORFis preceded by a short URF (16). In this system, the translationof the far downstream reporter ORF is linked to translation ofthe upstream ORF in a way that is explainable by a reinitiationmechanism (16).

We have studied the expression of such polycistronic mRNAsin transfected Orychophragmus violaceus protoplasts in order toanalyze the influence of URFs on the translation of an ORFimmediately downstream and one further downstream, andspecifically addressed the question of how the length of the URFinfluences the translation of the downstream reporter ORFs.

MATERIALS AND METHODS

Plasmid construction

The construction of the basic plasmids pNRF4-GusCat andpURF4GusCat has been described before (as 'pGC4.NS' and'pGC4.OS' in ref. 16). In brief, these plasmids contain adicistronic reporter gene cassette consisting of an upstream GUSand a downstream CAT ORF. An additional short ORF betweenGUS and CAT is opened by an ATG codon that overlaps theGUS stop codon. We have shown previously (16) that thisadditional ORF does not influence the translation efficienciesdescribed. Transcription is controlled by the CaMV 35S promoter

* To whom correspondence should be addressed+ Present address: Institute for Plant Sciences, Swiss Federal Institute of Technology Zurich, Universit2tsstrasse 2, CH-8092 ZOrich, Switzerland

Downloaded from https://academic.oup.com/nar/article-abstract/20/15/3851/2376578by gueston 18 February 2018

3852 Nucleic Acids Research, Vol. 20, No. 15

and the CaMV polyadenylation signal (Fig. 1). Only the regionbetween the transcription start and the GUS ORF (the leadersequence; L in Fig. 1) was modified between different constructs.In pURF4-GusCat, this leader contains a short ORF (URF forupstream reading frame) of 4 codons. In pNRF4 • GusCat, theATG codon of the URF was mutated to AAG while the rest ofthe sequence is identical in the two plasmids. We use the term'NRF' (no reading frame) and also give the length of the NRFregion in base triplets to stress the close resemblance of plasmidsfrom the pURF and the pNRF series.

The ATG codon of the URF and the corresponding AAG tripletof the NRF are contained within a small Xhol-Sall fragment(Fig. 2). By making use of the compatibility of these restrictionsites, this sequence module could be reiterated, giving rise toa series of plasmids with different leader lengths (Fig. 2). AnyURF opened by the start codon in one of the modules terminatesat a TAA codon immediately downstream of the Sail site. Bycombination of a first ATG containing module with an increasingnumber of ATG less ones the pURF series was obtained (Fig.2);each additional module adds six codons to the original 4 codonURF. For example, replacement of the Sall—EcoRl fragmentof pURF4 • GusCat by the Xhol -EcoKI fragment of pNRF4 • GusCat yielded pURFlO- GusCat (Fig. 2). By varying the positionof the first ATG containing module in the series of modules,URFs starting at different distances from the cap-site wereobtained (Figs. 2 and 5) and by combination of more than oneATG containing modules, URFs with additional internal ATGcodons could be constructed (Figs. 2 and 6).

Plasmid pURFl in which the last three codons of the URFare deleted and also the sequence derivatives shown in Fig. 6were constructed by inserting synthetic oligonucleotides intosuitable restriction sites. The plasmid pNRF10-Cat was generatedby deletion of the GUS ORF containing BamHl fragment frompNRF4-GusCat.

All plasmids were characterized by restriction endonucleasedigestions or by sequencing of double-stranded DNA usingstandard methods.

Protoplast transfections and reporter gene assays

Protoplasts from a cell suspension culture of O. violaceus weretransfected by electroporation (17). After overnight incubation,a protein extract was prepared and GUS- and CAT activitymeasured as previously described (16). All plasmids were testedat least four times, most of them more than ten times. Since theabsolute expression levels varied between different protoplastbatches, expression levels were normalized with respect topNRFlO-GusCat for GUS expression and to pNRFlOCat forCAT expression.

RNA analysisTotal RNA was isolated from transfected protoplasts afterovernight incubation and subjected to an RNAse A/Tl protectionassay as described (18) because RNA levels were too low fordecisive Northern experiments. Since we have shown previously(16) that the reporter proteins are translated from a polycistronicmRNA, only RNA quantification was performed here. To obtainprotected fragments of the same size for pN(U)RF • GusCat andpN(U)RFCat constructs an RNA antisense probe wassynthesized that contains homologies to the last 132 nt of the GUSORF and to the first 258 nt of the CAT ORF, but not to theintercistronic region present in pN(U)RF-GusCat. This was

5000

4000

1000

2000

KpnlSphI

PstI

BamHIXbal

EcoRINcol

3000

Figure 1. The circular map shows the basic setup of the plasmids used, consistingof the PUC18 vector, the 35S promoter, a variable leader (L) with a facultativeURF (small box), the GUS and CAT ORFs and the CaMV polyadenylation signal.The relevant restriction sites and the transcript produced in transfected plant cellsare shown.

derived from another dicistronic expression unit in which a partof the CaMV 35S leader sequence separates the GUS and CATORFs (18). As internal standard, a plasmid encoding thetransactivator (19) or an inactive mutant (M.deTapia,unpublished) was cotransfected and RNA from these plasmidswas detected by an antisense probe covering a common part ofactive and inactive transactivator.

RESULTSInfluence of URF length on downstream translationFollowing construction of the basic plasmid pNRF4-GusCat(Fig. 1), protoplasts were transfected with this plasmid to yielda polycistronic mRNA containing a /3-glucuronidase (GUS) anda chloramphenicol-acetyltransferase (CAT) ORF. Transcriptionis under the control of CaMV signals. The region preceding theGUS ORF on the mRNA (the leader) lacks ATG codons andhas none of the 'non-ATG' start codons that have been shownto allow translation initiation in plant protoplasts with efficienciesgreater than 1 % (20). The short ORF between the GUS and CATcoding regions has no effect on expression of GUS and CAT(16). The GUS ORF is translated efficiently from pNRF4 • GusCatmRNA but the CAT ORF is not translated at all (Fig. 3a).

A derivative of pNRF4 • GusCat was constructed by a singlepoint mutation in the 5' leader sequence, creating an Ncol


Nucleic Acids Research. Vol. 20, No. 15 3853

pNRF4eGusCat ACAGGGTACCCGGGCCTUGAGAAAACC AAG OAA G T C I GAC TAA GGATCCGGGGGAAAAG ATG (gus)

CTC- QAC TAA - . --

r

I I X hol/EcoRI RESTRICTION EcoRI/Sal l

LIGATION ,"

X S

pURF10.GusCat

+ / - -

1

Figure 2. The leader sequences of pNRF4.GusCat and pURF4,GusCat are shown on the top and symbolic representations of the total transcribed sequences in the line below. The part of the sequences located between the Xhol (X) and the Son ( S ) sites and used for reiteration is boxed and the ATGcontainig module shaded: for th~s case the coding region is symbolized by an arrow. The positions of the &mHI (B) and EtnRI (E) sites used for manipulation and analysis are indicated. Next, the manipulations (restriction, fragment isolation, ligation) yielding pURFIO.GusCAT from pNRF4.GusCat and pURF4.GusCat are shown. At the bottom examples of transcribed sequences of two of the other plasmids are shown, which were obtained by reiteration of the XhollSnfl box.

restriction site and an ATG start codon that opens a 4codon URF in plasmid pURF4.GusCat (Fig. 2). By reiteration of the small XhoI -Sun fragment within the leader and by the recombination of fragments from the pNRF- and pURF-series. it was possible (without introducing new sequence elements) to construct a variety of plasmids which encode mRNAs with leader sequences containing either URFs of different lengths or corresponding non- coding regions (NRFs). URFs could be inserted at different positions and could also contain additional internal ATG codons.

Since RNA steady state levels in transfected protoplasts were too low to be detected by Northern experiments, total RNA synthesized from the different plasmids was determined by RNAse protection assays. RNA levels were found to be similar for all plasmids (examples are shown in Fig. 4). Differences in reporter gene expression can, therefore, be attributed to differences in translation efficiency.

In the pNRF series (Fig. 3a), GUS expression was not significantly altered by the mere elongation of the leader sequence upstream of the GUS ORF in the range from 19 (4 triplets from the NRF region plus 15 surrounding ones) to 67 triplets. Further

elongation (1 15 and 21 1 triplets) led to a reduction of GUS expression and a leader length of 403 triplets completely abolished GUS expression, possibly because of increased secondary structure due to the reiterated complementary restriction sites. GUS expression obtained with pNRF 10. GusCat was set to 100% and used as an internal standard to normalize expression rates obtained in different experiments.

As expected, the introduction of an URF led to a significant reduction in GUS expression (Fig. 3b). The extent of the reduction was correlated to the length of the URF: the shortest possible URF, derived from pURF4. GusCat by deleting the last three codons (yielding pURFl .GusCat) and consisting only of a start codon, reduced GUS expression by a factor of two while longer URFs reduced it still further (Fig. 3b).

The position of a short URF (4 codons) relative to the cap-site did not influence its inhibitory effect (Fig. 5). With a 34codon URF, elongation of the upstream leader led to a slight reduction in expression of the downstream reporter ORFs (Fig. 5). Again, this might be due to the increased secondary structure caused by the reiterated complementary restriction sites.



140% QUS activity % CAT activity M 1 M

NRF insertion QUS without TW

0U8 with W

CAT with W

30

25

20

100 200 300length of leader NRF (triplets)

4 0 0

i QUS activity

URF insertion

% CAT ictlvlty

QUS without TW

aus with w

CAT without T *

CAT with TJV

30

10

100 200 300length of leader ORF (codons)

400

Figure 3. Reporter ORF expression from polycistronic transcripts lacking orcontaining an URF: A: Plasmids of the pNRF4+n-GusCat series (inset) withn=0,6(l2,24,30,48,96,l32 and 384 were transfected into O. vioiaceus protoplastseither alone or together with a plasmid expressing the CaMV TAV gene (19).Curves show the GUS and CAT reporter activities measured after overnightincubation in dependence to the number inserted NRF triplets. Values presentedare the means of at least four independent experiments and are normalized asdescribed in the text. Error bars have been omitted for clarity. The variabilitybetween different experiments is about 20% of the respective value. Note thatthe scales for GUS activity (left) and CAT activity (right) are different. B: AsA, but plasmids pURF 1 • GusCat and pURF4 + n • GusCat with n values as abovehave been used.

Translation initiation downstream of the URF is reinitiationTo determine whether the GUS ORF in pURF constructs isreached by ribosomes due to leaky scanning, we comparedplasmids pURF 10 GusCat and pURF10(2)- GusCat (Fig. 6a).The latter contains two ATG containing modules and therefore,the URF has an additional internal in-frame ATG codon identicalin context to the first. If only a fraction of ribosomes initiatetranslation at the first ATG, a similar fraction of the remainingribosomes should initiate at the URF's internal ATG. The twoupstream ATG's are so closely spaced that ribosome occlusioncan be excluded for steric reasons. Therefore, if GUS expressiondepended on those ribosomes that missed the first ATG codonQeaky scanning) an attenuating effect of the additional internalATG codon would be expected. However, this was not observed(Fig. 6a), indicating that the GUS ORF is translated byreinitiating ribosomes. A similar conclusion can be drawn fromthe results with plasmid pURF10(2) • TGG • GusCat in which theURF's stop codon is mutated to TGG. The elongation of the URF

5 1 7

5 0 6

3 9 6 -

344—|

2 9 8 ^

rr;

£

221 —

1 5 4 ^

> * •

- C A T* V

i

* #

• m9 *# t" t

1!

I — TAV

- G U S

Figure 4. RNA analysis. RNAse A/Tl protection experiments analyzingrepresentative RNA mixtures harvested from transfected protoplasts after 18 hincubation are shown. The antisense probes used were a mixture of a GUS-CATprobe, which covers the first 258 nt of the CAT ORF and the last 132 nt of theGUS ORF, and a transactivator (TAV) probe (columns 1 to 7). Extracts werefrom non-transfected cells (column 1) and those transfected with pNRFlOCat(columns 3,4), pURF28• GusCat (columns 5,6) and pNRF 196 GusCat (column7), either with the active transactivator plasmid (columns 2,4,6,7) or with theinactive transactivator control plasmid (columns 3,5). The locations of the protectedbands are indicated on the right.

across the two potential GUS start codons produces completeinhibition of GUS expression (Fig. 6a); this would not beexpected for a leaky scanning mechanism.

We conclude that all ribosomes initiate translation at the URFand that termination of URF translation must occur to allow GUSexpression. Ribosomes emerging from a longer URF are lesscompetent to reinitiate than those that emerge from a shorterURF. Reinitiation competence thus reflects a state of the ribosomewhich slowly changes during the translation process, and is notdefined by only the first initiation or termination event.

The influence of URF length on transactivated polycistronictranslationIn our system, translation of the CAT reporter ORF downstreamof the GUS ORF requires the presence of a CaMV ORF VItranslation product (TAV) which acts as a translation


Nucleic Acids Research, Vol. 20, No. 15 3855

--

.. GAGAAAACCAAGGAAGTCy 'GAQAAAACC ATG Q A A G T ~ G A Q A A A ACC AAG G A A GTC1,GAC TAA .. ---- - . --

Example: y.2, x=5

G U S

Reporter act ivi ty 80 --

I

1 i I

3 6 Leader Lengl

G U S O G U S ; TAV CAT ; TAV

Increment

Figure 5. The effect of leader length upstream of the URF on GUS and CAT expression. The GUSICAT dicistronic plasmid series with URFs positioned downstream 0 to 6 NRF modules was used and the relevant portion of this series is shown at the top of the figure together with a graphic representation of one example (y =2, x=5). The effect was measured for two different URF lengths. GUS expression is shown both in the absence and the presence of transactivator (TAV), but CAT expression is only measurable in ~ t s presence.

transactivator (19,2 1). Transactivation acts on a scanning related mechanism and is greatly enhanced by the presence of an URF (16). To estimate the efficiency of the transactivated translation in the system described here, we constructed the reference plasmid pNRF10-Cat by deleting the GUS ORF from pNRF 10. GusCat. Plasmid pNRF 10. Cat produces a monocistronic CAT mRNA with a leader sequence similar to that of the GUS reference plasmid pNRF 10. GusCat; therefore. the CAT ORF and the GUS ORF should be translated with similar efficiencies. RNA levels produced by the two plasmids were similar (Fig. 4), and thus CAT expression from the constructs directly reflects the translation efficiency. CAT expression from pNRF4Cat was normalized to 100%.

The level of transactivated CAT expression from pURF.GusCat plasmids was found to vary with the length of the URF (Fig. 3b). However. in contrast to the inhibiting effect of the URF on GUS expression. the stimulating effect on CAT expression was not linearly related to URF length. The strongest stimulation (25% of the monocistronic construct) was observed with URFs of 28- and 34 amino acids. Longer and shorter URFs had weaker effects. Those long URFs that completely inhibited

GUS expression also abrogated transactivated CAT expression (Fig. 3b).

As described previously for similar constructs (16), CAT expression from pURF-plasmids, including pURF lO(2) TGG . GusCat, required that translation of the preceding (GUS) ORF terminated before or at the CAT ATG codon (results not shown) indicating that CAT is translated by reinitiating ribosomes. Since URF translation in pURFIO(2)TGG.GusCat terminates downstream of the proper GUS ORF start codon and a functional GUS protein is therefore not produced (Fig. 6a), we must assume that ribosomes eventually thread into the GIJS reading phase producing a truncated GUS protein during their passage from the URF's stop codon to the CAT ORF. This process may require several reinitiation events until ribosomes enter the GUS reading phase.

The influence of TAV on GUS expression is less pronounced than on far downstream CAT expression, but the TAV-induced increment of GUS expression nevertheless roughly parallels the increase in CAT expression both in amount and URF-length dependence (Fig. 3b). The precise value of this increment is difficult to evaluate since TAV not only increases the reinitiation



CCAISGAAGTCGGAAGTCGGAAAACCEXPRESSION

COS CAT

pURTIO'GuaCat

pOKF10(2)GuaCat

pURF1O( 2 )TAG-GuaCat

pUHF1O<2)TGA.Gu«Cat

pCBFI 0 (2) TGG. CusCst

pOHFI O(TAA) 2-GusCat

pURFIO(tat)• GuaCat

pUBTIO(tgg)-GuaCat

pUITI0(o9Q)-GuaCat

AAGGAAGTCGACTAA£

AJSGAAGTCGACTAA£

AJBGAAGTCGACTAfiG

MSGAAGTCCACTtaC

AffiGAAGTCCACTOGC

AAGGAAGTCGACTAArAACTVI

AASGAAGTCGACTAA.TOTCTA

AAGQAACTCGACTACTCatG

AAtf2UU7TCGACTA£CCIS>IG

40 48

41 45

48 59

45 56

0 0

40 4S

44 SO

45 48

43 47

0 12

0 13

0 12

0 11

0 8

0 12

0 12

0 12

0 11

B)

c

pUBF4-GuaCat

pURFS(CGO)•GuaCat

pOSF5(AOA)-GuaCat

pORFA-GuaCat

pURTA1 • GuaCat

pURJVII'-GuaCat

EXPRESSIONCOS CAT

AIS

Alfi

ias

ATS

GAA

GAA

GAA

TOT

COT

GAT

-

COG

GAG

ana

coo

GTC

GTC

GTC

asasTTT

GAC

GAC

OAC

GAG

GAG

AAA

I M

lAi

I M

CTC

CTC

GTC

GAC

GAC

GAC

IAA

IM,

IAA

48 50

34 35

44 45

52 54

0 8

0 11

0 8

0 9

0 10

0 11

Figure 6. Dicistronic plasmids with modified URFs. A: URFs with two ATGcodons and with modifications at the stop codon. The part of the leader sequencethat was kept constant is shown on the top with the box signalling the part thatwas varied. The varied sequences are shown in boxes below. Start and stop codonsare underlined and mutated nucleotides are italized. GUS and CAT activities inthe absence (—) and the presence (+) of transactivator are given. B: URFs withdifferent coding sequences. The sequence of the URF region of pURF4 • GusCatand its derivatives is shown with the respective expression data as in A. URFA has the same sequence as the first sORF in the CaMV 35S RNA leader sequencewhile URF VH' presents the first codons of the CaMV ORF VD.

efficiency but also has some negative effects on expression. Thisis apparent from the slight reduction of GUS expression fromplasmids of the pNRF • GusCat series in the presence of TAV(Fig. 3b).

Influence of sequences around the URF termination codonon downstream translationFor the constructs described here, the nature of the URF's stopcodon is not important for reinitiation efficiency since all threestop codons led to similar expression levels[pURFl 0(2) TGA-GusCat and pURF 10(2) TAG GusCatcompared with pURF10(2)GusCat, which has a TAA stopcodon; Fig. 6a]. Furthermore, a construct containing twoconsecutive stop codons was equally active [Fig. 6a;pURFlOCTAAVGusCat vs. pURF10(TAT)-GusCat].

A survey of eukaryotic translation stop signals has shown thatin invertebrates CGG and CGT codons very rarely directly followa functional stop codon (22). However, a CGG codon is presentimmediately downstream of the GCN4 URF4 in a region whichinfluences the reinitiation properties in the yeast cells negatively(14,15). Therefore, we tested the influence of a CGG codondownstream of the URF's stop codon compared to constructscontaining a TGG codon [Fig. 6a; pURF10(TGG)-GusCat vs.pURFlO(CGG)-GusCat]. Reinitiation efficiency for bothconstructs was comparable and similar to that of all the others.

Modulation of the sequence upstream of the URF's stop codonalso did not significantly alter the reinitiation efficiencies.

Different URF sequences derived from the CaMV genome hadeffects similar to die artificial ones described above (Fig. 6b).

In order to see whether the length of the URF or the actualtime required for its translation is the critical parameter, weintroduced two different alternative arginine codons into the URFassuming that rare codons are decoded slower. However, an URFcontaining the rare CGG codon (5 % of the arginine codons indicot plants according to ref. 23), was only slightly moreinhibitory to GUS expression and allowed only a slightly highertransactivated CAT expression than an URF containing thefrequently used AGA codon [Fig. 6b; pURF5(AGA) • GusCat vs.pURF5(CGG)-GusCat]. A more systemic study would have tobe performed to verify this effect.

DISCUSSION

We have analyzed the influence of an URF on the translationof two consecutive coding regions on polycistronic mRNAs inplant protoplasts. The translation of a GUS ORF 18 nucleotidesdownstream of the URF was inhibited by the URF and the degreeof inhibition correlated with URF length. The shortest possibleURF, consisting of a start codon immediately followed by a stopcodon, inhibited GUS translation to about 50%, and a 100-codonURF to almost undetectable levels. It is noteworthy that theinhibition by longer URFs (> 50 codons) was found to vary morebetween different batches of protoplasts than that by shorterURFs, suggesting that the degree of the translational effect ofan URF depends on the general condition of the cell.

In the constructs described, the GUS ORF is most likelytranslated by ribosomes that have also translated the URF, i.e.by reinitiation. The alternative explanation of leaky scanning canbe excluded since an overlap between the URF and the GUS ORFabolished GUS expression, and the number of ATG codonswithin the URF did not influence GUS expression. Bothobservations are incompatible with a leaky scanning mechanism.

Processes interpreted as efficient reinitiation of translation havebeen observed with a variety of mRNAs (e.g. 10,11,16,24,25).However, it is not yet clear how reinitiation occurs. Translationinitiation requires a variety of initiation factors. The currentmodels assume that at least some of these initiation factors bindto 40S ribosomes before ribosomal subunits associate with themRNA. During the initiation event some factors dissociate whilstthe ternary complex eIF2GTPtRNAf*a is utilized and thenreleased as eIF2-GDP (reviewed in refs. 3, 26). This model,which is based mainly on in vitro reconstitution experiments,appears to predict that an RNA-bound ribosome can initiatetranslation only once. To explain reinitiation, it has beensuggested that ribosomes scanning from the URF's terminationcodon to a downstream start codon can bind a new set of initiationfactors if this scanning distance is long enough (10). Formammalian cells a distance of 79 nt was sufficient (10), whilstin yeast—especially under starvation conditions (11)—, and inplant cells (16), longer distances were required.

In the system described in this paper, the distance between theURF and the GUS ORF is short and invariable, and thereinitiation efficiency is inversely proportional to the length ofthe URF. This may be explained in two ways:

Firstly, increased URF length could cause more ribosomes tocompletely disengage from the RNA upon translation termination.The remaining ones would then regain initiation competenceduring scanning the short distance between the URF and the GUSORF. This scenario, however, is in conflict with our finding that


Nucleic Acids Research, Vol. 20, No. 15 3857

even after translation of the 600-codon-long GUS ORF, asignificant number of ribosomes continue to scan and reinitiateat a further downstream CAT ORF when the intercistronic regionis long (16).

Secondly, reinitiation competence might not be lostimmediately at the first initiation event but during translation ofthe URF as was suggested by Kozak (10). Ribosomes emergingfrom the translation of a short URF would then be either in astate of intrinsic initiation competence, because they still containnecessary initiation factors, anaVor in a state with an exceptionallyhigh affinity for initiation factors. The required factors could thusbe recruited faster by these ribosomes than by ribosomesemerging from a long URF. One could also speculate thateIF2 • GDP might still be loosely associated with the ribosomeafter the translation of a short URF and might be recycled toeIF2GTPtRNA|^a directly in this complex.

The transition between initiation competence and incompetenceoccurring during protein synthesis may involve a variety of well-defined intermediates. One such intermediate structure, whichis present when translation terminates after about 30 codons,would be a putative target for the CaMV TAV protein. TAVcould induce the stabilization of reinitiation competence. TAVaction allows multiple reinitiation events in the constructsdescribed here: first, at the GUS ORF, second, most likely ata short ORF between GUS and CAT, which does not influenceCAT expression (16; results for pURF constructs not shown),and third, at the CAT ORF.

It is interesting to note that TAV also causes an increase inGUS translation comparable to that of CAT translation. Thisindicates that TAV acts on ribosomes that would otherwise nottranslate the GUS ORF, i.e. that have lost or are in the processof loosing their TAV-independent initiation competence. Apositive influence of TAV on GUS expression has not beenrecorded so far (16), probably because the increment in somecases is small and because, in the previously studied examples,the GUS ORF was preceded by parts of the CaMV 35S RNAleader sequence containing multiple sORFs and other elementspossibly influencing translation. This made the regulation of GUStranslation much more complex than in the simple systemdescribed here.

In a variety of experimental systems, the nature of the URFor its surrounding sequences influenced its effect on downstreamtranslation (8,9,14,15). We have not found any parameters otherthan URF length (or possibly the time required to translate theURF) which influence the type of reinitiation mechanismdescribed here. Small alterations of sequences around the stopcodon which created or destroyed features similar to those thatseem to be important in the GCN4 system had no detectableinfluence. This may be due to differences between yeast andplants, but could also indicate that the requirements for reinitiationin the system described here are different to the cases wherereinitiation depended on a long intercistronic distance. We assumethat such parameters nevertheless exist. In particular, theinteraction of the translation machinery with the CaMVtransactivator TAV can also occur on types of mRNA other thanthose predictable from the ones presented here; e.g. CaMV ORFVII, the first longer ORF on the CaMV 35S RNA, may playa particular role in the translation of downstream ORFs:translation of a reporter ORF preceded only by the CaMV ORFVII is very efficiently transactivated by TAV despite the fact thatORF VQ is long (100 codons) and effectively inhibits downstreamtranslation in the absence of TAV (19). In this latter respect, ORF

Vn acts rather like the longer URFs used in the present studythat prohibited transactivation. A specific feature of ORF VTItranslation, such as pausing or premature termination, might beresponsible for this effect. It is noteworthy that the ORF VIIanalogue of the related figwort mosaic virus (FMV) was foundto be required in cis for transactivation of downstream translationby the FMV TAV protein (27).

ACKNOWLEDGEMENTS

We highly acknowledge the expert technical help of HannySchmid-Grob and Matthias Miiller, the construction of some ofthe plasmids by Werner Ruppitsch and the critical reading of theMS by Pat King, Tamas Kiss and Simon-John Morley.

REFERENCES1. Kozak, M. (1989) Mol. Cell. Bid. 9, 5073-5080.2. Kozak, M. (1987) Nucl Acids Res. 15, 8125-8131.3. Merrick, W.C. (1990) In Hill, W.E. et. al., (eds.), The Ribosome: Structure,

Function and Evolution. American Society for Microbiology,Washington.D.C, pp. 292-298.

4. Kozak, M. (1984) Nucl. Acids Res. 12, 3873-3893.5. Kozak, M. (1986) Cell 44, 283-292.6. Sedman, S.A., Gelembiuk, G.W., and Metz, J.E. (1990) J. Virol. 64,

453-457.7. Kozak, M. (1991) Gene Expression 1, 111-115.8. Schleiss, M.R., Degnin, C.R., and Geballe, A.P. (1991) J. Viol. 65,

6782-6789.9. Werner, M.A., Feller, A., Messenguy, F., and Pierard, A. (1987) CW/49,

805-813.10. Kozak, M. (1987) Mol. Celt. Bid. 7, 3438-3445.11. Abastado, J-P., Miller, P.F., Jackson, B.M., and Hinnebusch, A.G. (1991)

Mol. Cell. Biol. 11 486-496.12. Abastado, J-P., Miller, P.F., and Hinnebusch, A.G. (1991) The New Biologist

3, 511-524.13. Hinnebusch, A.G. (1988) 77G 4, 169-174.14. Miller, P.F., and Hinnebusch, A.G. (1989) Genes & Development 3,

1217-1225.15. Williams, N.P., Mailer, P.P., and Hinnebusch, A.G. (1988) MoL Cell. Biol.

8, 3827-3836.16. Ffltterer, J., and Hohn, T. (1991) EMBO J. 10, 3887-3896.17. Futterer, J., Gordon, K., Pfeiffer, P., Sanfacon, H., Pisan, B., Bonneville,

J-M., and Hohn, T. (1988) Virus Genes 3, 45-55.18. Futterer, J., Gordon, K., Sanfacon, H., Bonneville, J.M., and Hohn, T.

(1990) EMBO J. 9, 1697-1707.19. Bonneville, J-M., Futterer, J., Sanfacon, H., and Hohn, T. (1989) Cell 59,

1135-1143.20. Gordon, K., Futterer, J., and Hohn, T. (1992) Plant J., in press.21. Gowda, S., Wu, F.C., Scholthof, H.B., and Shepherd, R.J. (1989) Proc.

Natl. Acad. Sd. USA 86, 9203-9207.22. Cavener, D.R., and Ray, S.C. (1991) Nucl. Acids Res. 19, 3185-3192.23. Murray, E.E., Lotzer, J., and Eberel, M. (1989) Nucl. Adds Res. 17,

477-493.24. Levine, F., Yee, J.K., and Friedmann, T. (1991) Gene 108, 167-174.25. Peabody, D.S., and Berg, P. (1986) Mol. Cell. Biol. 6, 2695-2703.26. Pain, V.M. (1986) Biochem. J. 235, 625-637.27. Gowda, S., Scholthof, H.B., Wu, F.C., and Shepherd, R.J. (1991) Virology

185, 867-871.


role of an upstream open reading frame in the translation of

Documents