Secondary Structures upon Ribosome Binding to mRNA during

THE JOURNAL OF BIOLOGICAL CHEMISTRY Q 1993 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 268, No.35, Issue of December 15, pp. 26522-26530.1993 Printed in U.S.A.

The Influenceof 5’-Secondary Structures upon RibosomeBinding to mRNA during Translationin Yeast* (Received for publication, June 11,1993)

Francis A. SaglioccoS, MariaR. Vega Lasoj, Delin Zhull, Mick F. Tuitell, John E. G. McCarthyj, and Alistair J. P. BrownSII From the $Department of Molecular & Cell Bwlogy, University of Aberdeen, Marischal College, Aberdeen AB9 lAS, United Kingdom, the §Departmentof Gene Expression, Gesselkchuft fur Bwtechrwlogische Forschung mbH, Mascheroder Weg I , D3300 Braunschweig, Germany, and the llBwlogical Laboratory, University of Kent, Canterbury, Kent CT2 7NJ, United Kingdom

The influence of 6‘-secondary structure formation can strongly influence the frequency of initiation at themajor and 6‘-leader length upon mRNA translation in yeast coding region (6), and thisprovides the basis for the translahas been analyzed using a closely related set of cut tional regulation of the GCN4 and CPAl mRNAs in yeast (7, mRNAs (Vega Laso, M. R., Zhu, D., Sagliocco, F. A., 8). (iv) Secondary structure formation near the 5’-end of a Brown, A. J. P., Tuite, M. F., and McCarthy, J. E. G. eukaryotic mRNA can have negative or positive effects upon (1993) J. Biol. Chem. 268,6463-6462). A cut mRNA translation initiation (9). The inhibitory effects ofmRNA with a relatively shortunstructured 6’-leader (22 secondary structures that involve the 5’-leader sequence have a cut bases) had a ribosome loading about half that of mRNA with an unstructured5’-leader of 77 bases. The been confirmed in yeast using the CYCl (IO), HIS4 (4), and introduction of 6’-secondary structures at various po- PYKl mRNAs (ll),and, more recently, we have confirmed sitions throughout the 6”leader of the cut mRNA in- this using a closely related set of cat mRNAs transcribed from hibited translation initiation, the degree of inhibition a modular yeast expression system (12). These inhibitory effects are influenced by the position of the secondary strucbeing largely dependentuponthethermodynamic stability of the structure. Each mRNA carrying a 5’- tures in the5’4eader (12, 13). Not surprisingly, therefore, the secondary structure had a biphasic polysome distribu-5’-leader sequences of most yeast mRNAs appear to be relation, indicating that the mRNA molecules were distrib- tively free of secondary structures (14). However, it has been uted between untranslated and well translated subpop- shown that some secondary structures located 3‘ to the initiulations. This suggests that once S‘-secondary struc- ation codon can stimulate translation initiation, depending tures are unwound, they reform slowly relative to the upon their stabilityand position with respect to the initiation rate of translation initiation in yeast. Untranslated codon (15). (v) The length of the 5”leader can also influence mRNA accumulated in 43 S preinitiation complexes, translation initiation rates. Shortening the 5’-leader below a even when therewere only 5 bases between the 5’-cap critical length has been shown to decrease the translatability and the base of the hairpin. The data are consistent of the PGKl mRNA in yeast (16) and of synthetic mRNAs in with the scanning hypothesis (Kozak, M. (1989)J. Cell. vitro (17, 18). Biol. 108, 229-241) and suggest that 40 S ribosomal Most cellular mRNAs are thought to initiate translation subunits bind tomRNA early in the scanning process, via a scanningmechanism in eukaryotic cells (for reviews, see probably beforemRNA unwinding has taken place. Refs. 19-22), and it is therefore generally assumed that the structuralfeatures described above modulate initiation by influencing one or more of the steps in the scanningprocess. Messenger RNA translation is an important step in the One of the first steps in the initiation of most mRNAs in regulation of the expression of many eukaryotic genes. In mammalian cells is the recognition of the 5’-cap and the most cases, the rate at which an mRNA istranslated is interaction of eIF-4F (a complex of eIF-4A, eIF-4E and p220) determined primarily at the level of initiation. At least five with the 5’-cap (reviewed in Ref. 23). This is thought to be structural features of an mRNA can influence the efficiency followedby theATP-dependent unwinding of secondary of translational initiation. (i) The presence of a methylated structures in the 5’-leader. The factors eIF-4A and eIF-4B 5’-cap stimulates the translation of most eukaryotic mRNAs appear to be involved (24, 25), but other factors, and possibly (I),although the extent towhich the translation of a cellular the small ribosomal subunit itself, may contribute to mRNA mRNA is stimulated by a 5’-cap seems to depend to some unwinding (21, 26). At some point in this process, the small extent upon the structure of its B’-leader sequence (2, 3). (ii) ribosomal subunit (with associated factors) binds tothe The sequence context of the translationinitiation codon mRNA and scans to the initiation codon. Several aspects of influences the efficiency with which it is recognized by the this scheme remain unresolved. In particular, it is not clear when the small ribosomal subunit joins the initiation complex initiation complex, although to a lesser extent in the yeast Saccharomyces cereuisiue (4) than in vertebrate cells (5). (iii) and whether or not mRNA unwinding precedes the binding The existence of upstream AUGs in the5’-leader of an mRNA of the small ribosomal subunit (12, 23, 27). We have previously described the constructionof a modular * This work was funded by Commission of the European Commu- reporter system based on the cat gene from the bacterial nities Grant SCI*0193-C(SMA). The costs of publication of this transposon Tn9 for in vitro and in uiuo analyses of gene article were defrayed in part by the payment of page charges. This expression in yeast (12). Using this system, we compared the article must therefore be hereby marked “advertisement” in accordtranslation of a closely related set of cat mRNAs in yeast and ance with 18 U.S.C. Section 1734 solely to indicate this fact. 11 To whom correspondence should be addressed. Tel.: 224-273183/ in two cell-free systems, confirming the inhibitory effect of 5”secondary structures upon translation and also demon46; Fax: 224-273144.

26522

Initiation of Translation in

S. cerevisiae

26523

strating themodulating influence of their position (12). Here we describe a detailed analysis of the effects of 5’-secondary structure formation upon the translation of these cat mRNAs in vivo and compare the translation of cat mRNAs carrying non-ideal5”leaders with endogenous yeast mRNAs carrying structurally comparable 5”leaders. Our data are consistent with the idea that the small ribosomal subunit binds the mRNA early in the initiation process before the unwinding of 5’-secondary structures has takenplace. They also suggest that once unwound, 5”secondary structures are held in an unfolded state, probably by bound ribosomal subunits and/or initiation factors.

pared as described above. Cells wereresuspended in lysis buffer B (50 mMKC1, 5 mM MgCl,, 5 mM @-mercaptoethanol,25 mM Tris.HC1, pH 7.2) containing 0.5 mM phenylmethylsulfonyl fluoride, 1 mM Ntosyl-L-phenylalanine chloromethyl ketone, 1 mM benzamidine, and 200 pg/ml cycloheximide using 0.1 ml of solution/100 ml of culture volume. Cells were then sheared with glass beads as described above, the homogenate was centrifuged a t 5,000 X g for 5 min, Triton X-100 was added to the supernatant to 1%final concentration, and the sample was centrifuged at 10,000 X g for 10 min a t 4 “C. The supernatant was loaded onto a 15-30% (w/w) sucrose gradient prepared in lysis buffer B and centrifuged a t 22,000 r.p.m. for 14 h at 4 “C using a Beckman SW41 rotor (60,000 X g). Gradients were treated as described above for polysome gradients except that they were divided into twelve fractions.

MATERIALS AND METHODS

RESULTS

Strains and Media-The S. cerevisiue strain BWG-7a (MATa, Ieu2-3,leu2-112, his4-519, del-100, u r d ” ) was used throughout. All experiments were performed oncultures grown inYPD(2% glucose, 2%bacteriological peptone, 1% yeast extract). cat mRNA Structure-The construction of the cat genes, the mapping of the 5‘-end formed in vivo on each cat mRNA, and the experimental confirmation of the predicted 5’-secondary structures have been described previously (12). For this study, the FOLD Program (28) on the SERC computer a t Daresbury was used to predict the secondary structures formed by the 5”region of each cat mRNA and thethermal stabilityof these structures.The whole 5”leader and 150 bases of the coding region were analyzed, since the secondary structures formed at the5’-end of some cat mRNAs may extend into the coding region (for example, the cot152 mRNA) (12). Computer analysis suggested that numerous small secondary structures might form in the 5”region of the cat126 mRNA. Yet the 5’-leader of the cat126 mRNA has been shown experimentally to form no significant secondary stuctures (12), and,therefore, the thermodynamic stability of the short B’-structures predicted by the computer to form at the 5’-end of the cat126 mRNA (AG = -29.2 kcal.mo1”) was used as a baseline for the structures formed by other mRNAs (see Fig. 1).The FOLD program correctly predicted the structure of the stable 5’secondary structures that have been shown to exist experimentally (12). RNA Isolation and Northern Analysis-mRNA levels were measured by Northern analysis relative to the actin mRNA to correct for differences in RNA yield between samples (29). Total yeast RNA was prepared (30), denatured and electrophoresed on agarose gels containing formaldehyde (31). The RNA was then transferred to Hybond-N membranes (321, uv fixed, and probed with gel-purified DNA fragments radiolabeled with [32P]dCTP by random priming (33). Hybridizations were always performed under conditions of probe excess as described previously (29). Accurate quantification of hybridization signals was achieved directly from the Northern filters using the AMBIS 2D-Radioimaging System (Labhgic). Probes were then stripped from the filters, andthe filters were subjected to autoradiography and reprobed up to a maximum of four times (29). Polysome Analysis-Sucrose densitygradient centrifugation of yeast polysomes was performed essentially according to the procedures described by Tzamarias and coworkers (34). A 400-ml yeast culture grown in a 2-liter flask was harvested in mid-exponential growth phase when the A m was 0.5 and resuspended in 2.5 ml of lysis buffer A (100 mM NaCl, 30 mM MgCl,, 10 mM Tris.HC1, pH 7.5) containing 100 pg/ml cycloheximide. Cells were sheared by vortexing continuously with 4-g glass beads (0.4-mm diameter) for 5 min a t 4 “C, the homogenate was centrifuged a t 10,000 X g for 10 min a t 4 “C, and 1.5 mlof the supernatantwas layered onto a 36-ml10-45% (w/w) sucrose gradient prepared in lysis buffer A without cycloheximide. The gradient was centrifuged at 25,000 r.p.m. for 170 min at 4 “C using a Beckman SW28 rotor (average 90,000 X g), after which the gradient was drawn through a spectrophotometer to monitor the Am profile (see Fig. 3), anddivided into 15 X 2.5-ml fractions. 300 pl of 10% (w/v) SDS and 300 pl of 1 M Tris.HC1, pH 7.5, were added to each fraction andthen RNA prepared by three rounds of phenol:CHC13 extraction and one extraction with CHC13 alone. The final aqueous phase was diluted 3-fold with ddHnO, sodium acetate was added to a finalconcentration of0.2 M, and the RNA was precipitated with 2 volumes of ethanol. The RNA was reprecipitated with ethanol, resuspended in 40 pl of ddHzO, and 4 pl was subjected to Northern analysis (see Fig. 3). For extended polysome gradient analysis, yeast cultures were pre-

The Degree of Inhibition of Translation by a 5’-Secondary Structure Dependsupon the Predicted Thermodynamic Stability of theStructure-In this study, we exploited our new modular invitrolin vivo system for the analysis of gene expression in S. cerevisiae (12). In this system, the transcription of the cat gene (derived from the bacterial transposon Tn9) is driven by the TEFl promoter and terminated by CYC1 transcriptionaltermination sequences (Fig. 1). The convenient restriction sites located throughout the expression cassette were used to insert new 5”leaders or 3’-trailers and, in some cases, to insert new sequences at the 5’-end of the cat coding region (Fig. 1). Allof the constructs have been sequenced, their transcriptional start sites mapped, and computer predictions of 5’-secondary structure formation tested experimentally (12). Furthermore, the translatability of these cat mRNAs was compared in vivo in yeast and in vitro using yeast andrabbit reticulocyte cell-free translation systems (12). The translation in vivo of the various cat mRNAs shownin Fig. 1 were analyzed in more depth. The CAT activity generated by each construct in yeast was dividedby the abundance of its mRNA (measured by Northern blotting relative to the actin mRNA as an internal loading control) (12) and plotted against the thermodynamic stability of any 5’-secondary structures formed (Fig. 2). The cat126 mRNA was excluded from this analysis, because it carries a 5’-leader that is significantly shorter that the other mRNAs (Fig. l), and this results in its being translated inefficiently relative to the cat127 control (see below). It is clear that 5”secondary structures inhibit the translation of the cat mRNA (as shown previously) (12) and that the degree of inhibition is dependent primarily upon the thermodynamic stability of the 5‘-secondary structure. The cat128, cat154, and cat157 mRNAs carry a secondary structure at their 5‘-end, which is created by the annealing of sequences in the 5‘-leader with those in the 3’trailer.’ Interestingly, these mRNAs do not conform closely to the pattern observed forthose mRNAs with local secondary structures that are formed at the5’-end of the mRNA (compare asterisks and boxes in Fig. 2). Nevertheless, these interactions between the 5’-leader and the 3’-trailer do inhibit translation (Fig. 2). The inhibitory effects of such interactions upon translation have been described previously for a PGK1 mRNA carrying modified leader and trailer sequences (16). The Polysome Distribution of cat mRNAs Carrying B’-Secondary Structures Is Biphusic-The translation of the cat mRNAswas analyzed further by studying their ribosome loading in vivo using the procedure illustrated in Fig.3. Briefly, yeast transformants carrying the cat construct of interest were harvested at thesame point during the exponential growth phase ( A m = 0.5). Polysome gradients were then

’F. A. Sagliocco, D. Zhu, M. R. Vega Laso, J. E. G. McCarthy, M. F. Tuite, andA. J. P. Brown, manuscript submitted.

Initiation of Translation in

26524

S. cerevisiae

A

N

C

B/Bg

Nd St

E X

BNr

cat mRNA

B

0‘

’ 0

I

10

5’

-0.0 -12 -231

20

30

0

I

40

50

80

Secondary Structure I-kcal/moll

FIG.2. The degree of inhibition of mRNA translation by a 5’-aecondary structure in yeast is dependent largely upon its thermodynamic stability. The translatability of each cat mRNA was estimated by dividing the in uiuo cat activity by the mRNA level for each construct (12). The cat126 mRNA was not included because of ita short 5”leader (see text). Asterisks represent cat mRNAs with secondary structures formed by sequences at the5’-end of the mRNA, whereas boxes represent cat mRNAs with 5’-secondary structures formed via interactions with 3’-sequences.

-22.0

-382 -55.1

-19 -153

FIG.1. Summary of the cat constructs used in this study. Panel A, cartoon of the cat gene carried on YCp50 (adapted from Ref. 12). The arrow represents the position and direction of the transcript. The cat mRNAs differ in length due to their different 5’-leaders, but the cat127 mRNA is 835 nucleotides long. B, EarnHI; B/Bg,EarnHI/ EglII fusion; C, ChI; E, EagI; N , NotI; Nd, NdeI; Nr, NruI; St,StuI; X , XhoI. Panel E , the sequence and structure of each 5”leader has been described before (12) (see Fig. 3 for cartoons of the structures). The thermodynamic stability of each structure was determined as described under “Materials and Methods” and is expressed in kcal. mol”. The secondary structure formed at the5’-end of each mRNA is represented in cartoon form with thin lines representing the 5’- or 3”untranslated region and thick lines representing the coding region. The cat144 mRNA controls for a change made to the 5’-end of the cat127 coding region, and thecat147 mRNA controls for a change in the cat127 5’-leader, both of these changes being required to create the stem loop in cat152 (12). prepared by sucrose density gradient centrifugation, the gradients were fractionated, Northernanalysis was performed on RNA prepared from each fraction, and the signals were quantified by 2D-Radioimaging. Filters were probed for the cat mRNA and for the actin mRNA, which controlled for slight differences in the polysome profile between gradients. Despite the care taken to ensure that cells were harvested at the same phase of growth, differences between gradients were occasionally observed, probably due to minor differences in the physiological status of the cells and/or to small fluctuations in the density distributionin the sucrose gradient. However, the ribosome loadings on the actin mRNA control were similar for all gradients (Fig. 4B). The cat127 mRNA, which carries an unstructured 5”leader (from H s P 2 6 ) , was heavilyloaded with ribosomes, the profile peaking with the actin mRNA at about 10 ribosomes/mRNA (Fig. 4). The polysome distribution of the cat126 mRNA peaked at about five ribosomes/mRNA. These data correlate well with the relative translatabilities of these mRNAs as measured by the in vivo CAT activity generated by each; the

cat126 generated 56% of the activity from cat127 (12). The short 5”leader on the cat126 mRNA (22 bases) probably accounts for the reduced rate of translation, since previous work has demonstrated that the PGKl mRNA is translated less efficiently if the length of its 5”leader is decreased to below about 27 bases (36). In fact, van den Heuvel et al. (36) demonstrated that a deletion in the 5”leader of the PGKl mRNA, which reduces its length to 21 bases, was translated with about half of the efficiency of PGKl mRNAs with 5’leaders longer than 27 bases. For each c a t mRNA that carries a 5‘-secondary structure, a large proportion of the molecules accumulated in fractions near the top of the polysome gradients indicating that these mRNAs are poorly translated (Fig. 4). This was consistent with the relatively low amounts of CAT polypeptide translated from these mRNAs in vivo (12). However, the proportion of an mRNA that accumulated at the top of a gradient fluctuated between experiments. For this reason, we have not attempted to quantify the translatability of individual mRNAs on the basis of their ribosome loading. The fluctuation was possibly due to some extent to the sensitivity of mRNA translation to the physiological status of the yeast cell. Nevertheless, the accumulation of a large proportion of each mRNA with a 5’-secondary structure nearthe top of the polysome gradient was entirely reproducible (e.g. the catl30, cat143, cat147,and cat152 mRNAs) (Fig. 4).This was also the case for the cat128 mRNAwhose 5”leader and3”trailer interact to form a stable secondary structure. The short5”leader on the cat126 mRNA and thesecondary structures formed at the5’-ends of the cat128, catl30, cat143, cat147, and cat152 mRNAs inhibited the translation of these mRNAs. However, their effects upon the ribosome loading of the cat mRNA differed significantly (Fig. 4). The polysome distribution of the cat126 mRNA showed a single peak with a reduced number of ribosomes compared with the control mRNAs (actin andcat127). This indicates that cat126 mRNA molecules werebeing translated as homogeneous a population. In contrast, the polysome distributions of all of the mRNAs with 5”secondary structures showed two peaks. For each of these cat mRNAs, a large proportion of the molecules were poorly translated, accumulating near the top of the gradient. In addition, a small but significant proportion of the molecules was as heavily loaded with ribosomes as the cat127 mRNA with an unstructured 5”leader (Fig. 4). This observation,

Initiation of Translation in S. cerevisiae

26525

A

*

1 .oo 260

0.75

0.50

0.25

FIG. 3. The polysome method used to determine the ribosome loading on each cat mRNA in vivo. A , the absorbance profile at 260 nm following sucrose density gradient centrifugation was monitored to determine the size distribution and positions of the polysomes on the cat226 polysome gradient. Sedimentation is from right to left. B, RNA prepared from gradientfractions was subjected to Northern analysis and probed for experimental ( C A T )and control ( A C T ) mRNAs. C, signals on the Northern filters were quantified by 2DRadioimaging. Numbers above the graph show the positions of polysomes containing a given number of ribosomes.

Ac 0

R

CAT w

C

1

3

2

4

5

6

% of Total Radioactivity 25 10

8

7

9

8 4 3

2

10

1112

13

1415

1

20 15 10

-

50

i

2

3

4

5

I3

7

8

9

1 0 1112 1 3 1 4 1 5

Polysome Fraction

which was entirely reproducible, wasnot dependent upon the position of the secondary structure with respect to the 5’leader. Primer extension analysis (12) revealed no difference between the 5’-ends of cat143 mRNA molecules present in polysome fractions compared with monosome fractions (not shown). Therefore, for each cat sequence carrying a 5’-secondary structure, themRNA molecules could be divisible into poorly translated andwell translated populations. Cat mRNAs Carrying 5’-Secondary Structures Accumulate in 43 S Initiation Complexes-The experiments described above showed that the major proportion of the cat mRNAs with 5’-secondary structures accumulated near the top of the polysome gradients (Fig. 4). The poorly translated subset of the cat128, cat143, cat147, and cat152 mRNAsappeared to be accumulating in a single peak near the top of each of the gradients. The only exception to this pattern was the cat130 mRNA, which consistently showed a double peak near the top of the gradient (Fig. 4). The behavior of this mRNA, which appears to be translated into twoCAT proteins of different lengths in cell-free translation systems possibly through initiation at anadditional site upstream of the stem loop, is currently being investigated.2 In an effort to determine the extent to which poorlytrans-

* D.Zhu, S. Christodoulou, and M. F. Tuite, unpublished results.

CAT mRNA

-+

ACT mRNA

lated cat143 mRNA molecules were interacting with the translational apparatus, their distribution was analyzedonextended polysome gradients (see “Materials and Methods”). The absorbance profiles (at 600 nm) and thevisualization of the ribosomal RNAs present in each fraction of these extended gradients, which allow one to define more accurately the complex(es) to which the mRNA is bound (by ethidium bromide staining of Northern gels), demonstrated that 40,60, and 80 S complexes were clearly resolved onthese gradients (not shown). Consistent with its distribution on the “normal” polysome gradients (Fig. 4)) the cat143 mRNA accumulated in a single peak on extended gradients in a position corresponding to the 43 S initiation complex(Fig. 5). Also,no unbound cat130 mRNA was observed at the top of extended polysome gradients; all of the “untranslated” cat130 mRNA was present in preinitiation complexes? Under similar conditions, the cat127 mRNA (which has an unstructured 5’leader) showed a vastly lower level of accumulation in 43 S initiation complexes (note thedifferent scales in Fig. 5). The cat126 and cat143 mRNAs Are Translated Less Efficiently in Vivo Than Natural Yeast mRNAs with Analogous 5’-Leaders”As described above, the cat126 mRNA is translated with approximately 50% of the efficiency of the cat127 F. A. Sagliocco and A. J. P. Brown, unpublished results.

26526

Initiation of Translation in S. cerevisiue A

1

t;

127

FIG.4. The effect of6'-secondary structure formation upon the ribosome loading of the cat mRNA in vivo. A, Northern analysis of each cat mRNA in polysome gradient fractions. Fraction numbers are along the top. I, the bottom of each gradient; 15, the top of each gradient,sedimentation being from right to left. Numbers down the left side indicate the cat mRNA oneach gradient (see Fig. 1). Cartoons down the right side illustrate the secondary structure formed at the5'-end of each mRNA (see Fig. 1).B, the distributions of the cat and ACT1 mRNA on each gradient (see "Materials and Methods" and Fig. 3). The number at the top left of each graph represents the cat mRNA being analyzed. Solid line, cat; dotted line, ACTI.

'

2

3 4

5

6

7

8

9

101112131415


CAT 2 127

26527

ACT

3

4

5

6

7

8

9

1

02

3

4

5

6

'

7

8

9

1

0

5'

143 5'

B

60

CAT mRNA (cpm x 1000)

127

ACT mRNA (cpm x 1000)

80s

80s

60

TOP

403

40t 1 30

20

10 o1

t2 - 3 - 4 " 6 4 8 0 7

8

9

1

0

1

1

1

2

Polysome Fraction

ACT mRNA (cpm x 1000)

CAT mRNA (cpm x 1000)

143

900

809

803

403

Top

700 -

20 ~~

800

800

16

-

600 400

-

300

-

200

-

100

-

0

10

6

" 1

'

2

"

3

"

4

'

6

6

7

8

9

1

0

1

1

.

1

!. 2

O 0

Polysome Fraction

Fk. 5. The accumulation of the cat143 mRNA in 43 S complexes. A, Northern analysis of the cat127 and cat143 mRNAs in fractions from extended polysome gradients. B, quantified signals from the Northern blots shown in A showing the positions at which 80 S ribosome and 60 and 40 S subunits sediment on these extended gradients (see "Materials and Methods"). Sedimentation is from right to left. Dots, cat; crosses, ACTl.

mRNA, and this is probably due to its relatively short 5'leader (22 bases). Therefore, we compared the translation of the cat126 mRNA with the HIS3 mRNA,which has an unstructured 5"leader (14) of an equivalent length (23 bases) (37) by reprobing a cat126 polysome gradient for HIS3 (Fig. 6). The HIS3 and cat126 mRNAshave coding regions of equivalent lengths (219 and 222 codons, respectively) (12,37),

and both have poor codon biases for yeast (codon bias index = 0.01 and -0.02, respectively) (38-40). The ribosome loading on the HIS3 mRNA was onlyslightly greater than thecat126 mRNA under identical conditions and, therefore, the natural HIS3 mRNA appears to be translated at similar rate to the artificial cat126 mRNA (Fig. 6 ) . The 5'-end of the PMAl mRNA forms a relatively strong

26528


% 01 Total Radioactivity

26

-

20

-

16

-

10

-

101 8 4 3

2

6 O

+ 1

2

” 3

” 4

”

6

”

6

“

?

’

~ 8

’ 9

10

11

1 2 1 3 1 4 1 6

Polysome Fraction

-CAT128 mRNA

+HIS3

% of Total Radioactivity I*]

X of Total Radioactivity 1.1

461

mRNA

10

8 4 3 1

2

h

40

I \

1

2

3

4

6

8

7

8

9

i

2o

1 ’€I

1 0 1 1 1 2 1 3 1 4 1 6

Polysome Fraction

-

CAT143 mRNA

than in a vertebrate system (12) (comparewith Ref. 13). These data were consistent with previous studies performed on the yeast HIS4 (41, CYCl (lo), and P Y K l genes (11) and on various mammalian mRNAs (6,42, 43). In this study, further analysis of the cat mRNAs revealed that thedegree of translational inhibition by local secondary structures formed within the 5’-region of an mRNAwas primarily dependent upon the predicted thermodynamic stability of these structures (Fig. 2). 5’-secondary structures are probably unwound before translation can proceed. Hence, the data presented in Fig. 2 are consistent with the idea that the intracellular equilibrium betweenthe folded and unfolded states is strongly influenced by the stability of the secondary structure (Fig. 7). These data would also be consistent with modelsinwhichmRNAunfoldingoccurssimply by the breathing of secondary structures followed by the advance of the translational apparatus(reviewed in Ref. 9). Equally, the data would fit models invoking mRNAunwinding factors such as eIF4A and eIF-4B (23, 24), possibly acting in concert with the 40 S ribosomal subunit (9), since their ability to denature an mRNA secondary structure might be related to the stability of the structure. The position of a secondary structure with respect to the 5’-cap and the initiation codon did affect the translation of the cat mRNA (12) but to alesser extent than thestability of the structure (Fig. 2). Nevertheless, the significance of such position effects has been emphasized by further analyses of cat and luciferasemRNAs (13) and by detailed studies of

1

Jb

2

l b

+PMA1 mRNA

FIG. 6. Translation of cat mRNAs compared with natural yeast mRNAs carrying analogous 5‘-leaders. Northern blots used to analyze the polysome distribution of the cat126 and cat143 mRNAs (Fig. 4) were stripped and reprobed for the HIS3 and PMAl mRNAs, respectively. Signals were quantified by 2D-Radioimaging as described under “Materials and Methods.”

secondary structure with a predicted thermodynamic stability of -54 kcal. mol-’ (41). This secondary structure is roughly analogous to thestructure in the 5’-leader of the cat143 mRNA; both structures lie very closeto the5‘-cap, and both mRNAs are translated poorly in rabbit reticulocyte lysates (12, 41). For this reason, we compared the ribosome loading of these mRNAs in vivo by reprobing the cat143 polysome gradient for PMAl (Fig. 6). Interestingly, the PMAl mRNA is heavily loaded with ribosomes under conditions when the cat143 mRNA is very poorly translated. Therefore, the 5’secondary structure on the PMAl mRNA apparently does not significantly inhibit its translation in uioo. DISCUSSION

In a previous study, we constructed a modular cat reporter system to facilitate the analysis of gene expression in yeast (12) (Fig. 1). A series of 5’-leader sequences wereintroduced into thecat reporter gene, and the translationof the resultant mRNAs was compared in uiuo and in vitro by measuring the amount ofCAT protein (or activity) generated by each mRNA. In this way, we confirmed that 5’-secondary Structures inhibit translation in yeast and that the translational apparatus in yeast is moresensitive to 5‘-secondarystructures

V

3

1-

t

AUC

Translation FIG. 7. A model that attempts to account for the observed influence of S’-secondary structures upon the translation of the cat mRNAs. According to this model, an early stepin the translation of cat mRNAs with 5‘-secondary structures appearsto be the binding near or at the5’-end of the mRNA by the 40 S ribosomal subunit (plus associated factors (step 1)). At some stage, this is probably followed by the unfolding of the 5”secondary structure (step 2), the frequency of this event depending upon the thermodynamic stability of the secondary structure. This, in turn,is followed by 40 S scanning and the subsequent formation of the 80 S complex at the initiation codon (step 3), following which translational elongation ensues (step 4). At some point during steps 3 or 4, further 40 S ribosomal subunits probably bind. The double arrows suggest that the steps are probably reversible, but the different sizes of the arrows are indicative of the apparentbias in each “equilibrium” (see text).

Initiation of Translation in genes that suppress the inhibitory effects of 5“secondary structures upon HIS4 mRNA translation in yeast (26). One such gene (SSL2) only suppresses the inhibitory effects of a secondary structure that lies close to the 5’-cap of the HIS4 mRNA but does not overcome an equivalent secondary structure placed closer to theinitiation codon. The distribution of the cat143 mRNA on extendedpolysome gradients revealed that untranslated molecules accumulate in 43 S preinitiation complexes in vivo (Fig. 5). This accumulation was not an artifact induced by the preincubation of cells with translation inhibitors, since we did not add cycloheximide to cultures before harvesting. The equilibrium between free and 40 S-bound cat143 mRNA was heavilybiased toward the bound form (Fig. 7). Since the 5’-cap is only 5 bases from the base of the hairpin in the 5’-leader of the cat143 mRNA, this result would appear to be inconsistent with the data of Kozak (6) who showedthat thesmall ribosomal subunit would not complex with an mRNA that had a secondary structure located 12 bases from the 5’-cap. However, Kozak‘sstudy was performed using a wheat germ cell-free system in the presence of sparcomycin, whereas our analysis was performed in yeast cells, and, therefore, several explanations of this apparent discrepancy are possible. For example, the binding of the small subunit to the 5’-end of the mRNA might not have been observed in the wheat germ system because of the altered concentrations of key factors in uitro, the presence of sparcomycin, ora fundamentaldifference in the behavior of wheat and yeast translational apparatuses. Two significant conclusions can be drawn from the observed accumulation of untranslated cat143 mRNA in 43 S preinitiation complexes (Fig. 5). First, it would appear that in yeast, secondary structures located close to the 5’-cap apparently do not inhibit translation by preventing the binding of the small ribosomal subunit (9). Instead, such structures seem to block 40 S subunit scanning to the initiation codon. Second, this observation strongly suggests that during translation initiation, the binding of the 40 S ribosomal subunit to the cat143 mRNA precedes the unfolding of the 5‘-secondary structure. Obviously,this does not favor a model of translation initiation based upon the unfolding of the 5’-secondary structures prior to 40 S subunit binding (27). All of the cat mRNAs with 5”secondary structures that we analyzed showed biphasic distributions across polysome gradients (cat128, catl30,cat143, cat147, and cat152) (Fig. 4). For each of these mRNAs, the major proportion of the mRNA molecules sedimented in the monosomal region of the polysome gradient, but a small proportion of the molecules was heavily loadedwith ribosomes. This was reproducible and was not due to differences intheirtranscriptional start sites. Therefore, each of these mRNA sequences was divisible into a subset of poorly translated mRNA molecules, and a subset of mRNA molecules that were apparently translated as efficiently as anmRNA with an unstructured 5’-leader (cat127). In some cases,these cat mRNAs differ only in their 5’-leader sequences (for example, cat127 and cat143) (Fig. l), and, therefore, only the translationalinitiationrates on these mRNAs will have been affected by the sequence changes. Hence, the clear implication of the biphasic polysome distribution is that the once a 5”secondary structure has been unfolded, it remains in an unfolded state for a significant period, allowing it to be translated efficiently. Depending upon the position of the secondary structure, once opened, its refolding could besterically hindered either by scanning 40 S ribosomal subunits, by elongating ribosomes, by specific initiation factors (for example, eIF-4A and/or eIF-B), or by a combination of these. The biphasic polysome distribution of

S. cerevisiue

26529

these mRNAs also implies that once a 5”secondary structure has been unfolded, the intracellular equilibrium is shifted rapidly toward translation (Fig. 7). If a secondary structure refolded rapidly in vivo (relative to a translational initiation event) a monophasic polysomal distribution with a relatively low ribosome loading would havebeen expected. This was not observed for any of our cat mRNAs carrying 5“secondary structures. Our data can be interpreted on the basis of the model presented in Fig. 7. According to this model, the 40 S ribosomal subunit associates with a cat mRNA with a 5’-secondary structure at an early stage in translation initiation. This seems to occur beforethe unfolding of the 5‘-secondary structure, the frequency of mRNA unfolding being linked in an as yet undetermined way to the thermodynamic stability of the secondary structure. mRNA unfolding would then allow 40 S scanning to proceed with the subsequent formation of the 80 S complex at the initiation codon. Followingthis step, translational elongation could then begin. At some point during the latter steps, further initiation events probably proceed since the 5’-secondary structures refold relatively slowly. By definition, each of these steps can be represented in terms of a thermodynamic equilibrium (Fig. 7). Our data suggest that in yeast, the equilibrium at step 1 is strongly biased toward 40 S-bound mRNA. Probably due to the relative stability of the structures we have analyzed, the equilibrium at step 2 is biased toward the folded (and hence untranslated) form of the complex. However, once melted, the equilibria at steps 3 and 4 favor the efficient translation of the mRNA (yielding the biphasic polysomal distribution) and limit the refolding of the mRNA. The data are consistent with the scanning hypothesis (19) but could also beexplained by an alternative mechanism of unfolding. It is possible to interpret the data presented in terms of a model in which mRNA unwinding precedes the formation of a productive 43 S complex if it is assumed that the 43 S complexes that accumulate on mRNAs with 5’-secondary structures in vivo are incapable of forming an 80 S complex. It is conceivable that translation initiation might be unable to proceed further if a 40 S subunit binds an mRNA with a strong 5’-secondary structure before this structure has been unwound. In thiscase, the 40 S ribosomal subunit would have to disassociate from the folded mRNA,and mRNA unwinding would have to take place before the 40 S ribosomal subunit could reassociate on the unstructured 5‘-leader to form a productive initiation complex. We must assume thatthe effectively %on-productive” binding observed in vivo is (i) not artifactual, and (ii)at the 5’-end of the mRNA. Models of translation initiation based on 40 S ribosomal subunit binding occurring before or after mRNA unwinding are not mutually exclusive. It is possible that some mRNA unwinding might take place beforethe 40 S ribosomal subunit binds, the remainder occurring afterwards. This might depend upon the thermodynamic stability and possibly the position of the 5’-secondary structure with respect to the 5’-cap and initiation codon. In an attempt to confirm our observations made using an artificial mRNA in yeast, we analyzed the polysomal distribution of a natural yeast mRNA, which can form a stable secondary structure at its5’-end (AG = -54 kcal. mol”); the PMAl mRNA is translated very inefficiently in a rabbit reticulocyte cell-free system (41) in a similar fashion to the cat143, which carries a 5’-secondary structure roughly equivalent to thePMAl mRNA (12). It would therefore be expected that the PMAl mRNAwouldshow a biphasic polysomal distribution like the cat143 mRNA(Fig. 4). However, the

26530


PMAl mRNA is very heavily loaded with ribosomes in vivo (Fig. 6), suggesting that its 5"secondary structure, which blocks translation in vitro,does not do so in vivo. The cat126 mRNA is translated at about 50% of the rate of the cat127 mRNA (Fig. 4) (12). This is entirely consistent with previous studies on the translation of the yeast PGKl mRNA (36) and synthetic cat mRNAs in vitro (17,18),which showed that translation is adversely affected oncethe length of the 5'-leader is reduced below a critical length (about 30 bases), probably due to the inability of additional 40 S subunits to "stack" onshort 5"leaders once an 80 S complex has formed at the initiation codon (18).The monophasic polysomal distribution of the cat126 mRNA is consistent with this idea (Fig. 4). The HZ83 and cat126 mRNAs, which have 5'leaders and coding regionsof similar lengths as well as similar codon biases (37), are translated at approximately the same rate (Fig. 6). However, TCMl mRNAs without a 5'-leader (i.e. where the initiation codon abutsthe 5'-cap) can be translated efficiently inyeast (44). Hence, someyeast mRNAs appear to have evolvedmechanisms of overcoming the inhibitory effects of short 5"leaders (e.g. the TCMl mRNA) (44) or 5'-secondary structures (e.g. the PMAl mRNA) (Fig. 6.). In some cases, these mechanisms might involve the use of mRNA-specific translation factors, whereas other mRNAs might exploit alternative mechanisms of translational initiation. A precedent for this has been set by the human BiP mRNA, a cellular mRNA that can initiate translationthrough an internal ribosome-binding mechanism (35). Acknowledgments-We are grateful to Dr. Ian Stansfield for advice on the extended polysome gradients. We also thank Etienne Capieaux and Andre Goffeau for providing a probe for the PMAlmRNA. REFERENCES 1. Shatkin, A. J. (1976)CeU 9,645-653 2. Lindquist. S., and Petersen, R. (1991)Enzyme 44, 147-166 3. Gerstel, B., Tuite, M. F., and McCarthy, J. E. G. (1992)Mol. Microbiol. 6, 2339-2348

4. Cigan, A. M., Pabich, E. K., and Donahue, T. F. (1988)Mol. CeU. BWL 8, 2964-2975 ~.~~ 5. Kozak, M. (1987)Nucleic Acids Res. 16,8125-8148 6. Kozak, M. (1989)Mol. CeU. Biol. 9,5134-5142 7. Hmnebuseh, A. G. (1990)Trends Bwchem. Scr. 16,148-152 8. WtTyer, M., Feller, A., Messenguy, F., and Pierard,A. (1987)CeU 49,805"

813

9. Kozak, M. (1991)J. Biol. Chem. 266,19867-19870 10. Baim, S. B., and Sherman, F. (1988)Mol. CeU. Biol. 8, 1591-1601 11. Bettany, A. J. E., Moore, P. A., Cafferkey, R., Bell, L. D., Goodey, A. R., Carter, B. L. A., and Brown, A. J. P. (1989)Yeast 6,187-198 12. Vega Laso, M. R., Zhu, D., Sagliocco,,F. A., Brown, A. J. P., Tuite M. F., and McCarthy, J. E. G. (1993)J. Bwl Chem. 268,6453-6462 13. Oliveira, C. C., van den Heuvel, J. J., and McCarthy, J. E. G. (1993)Mol Microbiol., 9,521-532 14. Cigan, A. M., and Donahue, T. F. (1987)Gene 69,l-18 15. Kozak, M. (1990)Proc. Natl. Acad. Sci. U.S.A. 87,8301-8305 16. van den Heuvel, J. J., Planta R. J., and Raue, H. A. (1990)Yeast 6,473482 17. KO-&, M. (1991)Gene Expression 1,111-115 18. Kozak, M. (1991)Gene Expression 1, 117-125 19. Kozak, M. (1989)J. CeU Biol. 108,229-241 20. Hershey J. W. B. (1991)Annu Rev. Bwchem. 60,717-755 21. Kozak. M. (1992)Crit. Rev. Biochem. MOL Biol. 27.385-402 22. Yoon, H., &d Donahue, T. F. (1992)Mol. Microbioi 6, 1413-1419 23. Thach, R. E. (1992)Cell 68,177-180 24. Jaramillo, M., Dever, T. E., Merrick, W. C., and Sonenberg, N. (1991)Mol. CeU. Biol. 11,5992-5997 25. Blum, S., Schmid, S. R., Pause, A., Buser, P., Linder, P., Sonenberg, N., and Trachsel, G. (1992)P m .Natl. Acad. Sci. U.S.A. 89, 7664-7668 26. Gulyas, K. D., and Donahue, T. F. (1992)Cell 69,1031-1042 -27. . . Sonenbere. ..""~. . N. (1991) Trends Genet. 9.105-106 28. Zuker, M.:& ' d ~Stieaei,P.