Reconfiguration of yeast 40S ribosomal subunitdomains by the translation initiationmultifactor complexRobert J. C. Gilbert*†, Yulya Gordiyenko‡, Tobias von der Haar‡, Andreas F.-P. Sonnen*, Gregor Hofmann*,Maria Nardelli‡, David I. Stuart*†, and John E. G. McCarthy‡§

*Division of Structural Biology, Wellcome Trust Centre for Human Genetics, University of Oxford, Roosevelt Drive, Oxford OX3 7BN, United Kingdom;†Oxford Centre for Molecular Sciences, Central Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QH, United Kingdom;and ‡Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester M1 7DN, United Kingdom

Edited by Fred Sherman, University of Rochester Medical Center, Rochester, NY, and approved February 7, 2007 (received for review August 10, 2006)

In the process of protein synthesis, the small (40S) subunit of theeukaryotic ribosome is recruited to the capped 5� end of the mRNA,from which point it scans along the 5� untranslated region in searchof a start codon. However, the 40S subunit alone is not capable offunctional association with cellular mRNA species; it has to beprepared for the recruitment and scanning steps by interactionswith a group of eukaryotic initiation factors (eIFs). In buddingyeast, an important subset of these factors (1, 2, 3, and 5) can forma multifactor complex (MFC). Here, we describe cryo-EM recon-structions of the 40S subunit, of the MFC, and of 40S complexeswith MFC factors plus eIF1A. These studies reveal the positioningof the core MFC on the 40S subunit, and show how eIF-bindinginduces mobility in the head and platform and reconfigures thehead–platform–body relationship. This is expected to increase theaccessibility of the mRNA channel, thus enabling the 40S subunit toconvert to a recruitment-competent state.

posttranscriptional gene expression � protein synthesis � ribosome structure

Elucidation of the mechanisms underlying ribosome function andprotein synthesis remains one of the major challenges of mo-

lecular biology. Recent progress in structural analysis of bacterialribosomes has provided insight into the likely modes of action ofcore functional centers, including those for decoding and peptidyltransferase, and the tRNA-binding sites (1). Analogous core struc-tural features are clearly shared by the eukaryotic counterpart, butthere is much less structural and mechanistic information availablethat is specific to the eukaryotic ribosome. This limits our under-standing of the process of translation initiation, the step wheremajor differences between the prokaryotic and eukaryotic systemsare evident (2).

The small ribosomal subunit in both prokaryotes (30S) andeukaryotes (40S) is responsible for controlling base pairing betweenthe tRNA anticodon and each mRNA codon during proteinsynthesis. However, unlike its prokaryotic (30S) counterpart, theeukaryotic 40S subunit does not locate directly to the position of themRNA AUG codon where protein synthesis begins. Instead, re-cruitment onto cellular mRNAs generally occurs via the capped 5�end. Because the AUG start codon can be located many hundredsof nucleotides downstream, the 40S subunit then has to translocateto reach the initiation site (3) [see supporting information (SI) Fig.5]. During this processive, sequence-independent scanning phase,the 40S subunit manifests some characteristics that appear to besimilar to those of a molecular motor (2, 4).

The eukaryotic 40S subunit alone is incapable of stable recruit-ment onto the capped 5� ends of cellular mRNA molecules. Its rolein translation initiation depends on a large number of eukaryoticinitiation factors [the eIFs (5)]. According to the current classifi-cation, 11 distinct eIFs (including eIF2B, a guanine nucleotideexchange factor) are involved in (steady-state) translation initiation.There has been considerable progress in recent years in under-standing the functions of the individual eIFs, although much

remains to be learned about their contributions to the mechanismand control of the initiation pathway in vivo. Binding of the ternarycomplex (comprising Met-tRNAi

Met, eIF2, and GTP) to the 40Ssubunit is stabilized by eIF1A and eIF3 (6). Because eIF3 can bindto both eIF2 and the 4G component of the cap-binding complexeIF4F, it promotes recruitment of mRNA to 40S (5). Geneticexperiments in yeast have indicated that eIF1, eIF2, and eIF5influence start codon selection (7), whereas in vitro biochemicalexperiments have shown that eIF1 and eIF1A play roles in scanningand formation of the 48S complex, which comprises 40S, the eIFs,and mRNA (8, 9). A growing body of evidence indicates that, atleast in budding yeast, eIF1, eIF2, eIF3, and eIF5 may bind to the40S subunit as a preformed multifactor complex (MFC) (10). Thusthe MFC components, together with eIF1A, play a key role in40S-mRNA recruitment, scanning of the 5� untranslated region,and start codon recognition (6, 10–12) (see SI Fig. 5).

In this article, we address the question how the ribosomal 40Ssubunit becomes rendered competent for translation initiation asthe result of association with the eIFs. We have applied cryoelec-tron microscopy to this problem, with the intent of gaining insightinto any changes in 40S conformation that might explain itsconversion to an initiation-competent state. The results indicatethat eIF-dependent reconfiguration of the 40S subunit domainsfacilitates efficient access of cellular mRNAs to the mRNA channel,thus promoting the process of translation initiation.

ResultsTo perform the structural analyses described in this article, wegenerated highly purified preparations of 40S subunits and of eIFs(Fig. 1A). These components were used to prepare 40S–eIFpreinitiation complexes (see Fig. 1 B–D and Table 1). Like 40Ssubunits, the 40S–MFC complex (hereafter referred to as the 43Scomplex) ran as a single peak during analytical ultracentrifugation(Fig. 1C), consistent with low heterogeneity in the preparation. Thecryo-EM reconstruction of the yeast 40S subunit at 19-Å resolution

Author contributions: R.J.C.G. and Y.G. contributed equally to this work; D.I.S. and J.E.G.M.designed research; R.J.C.G., Y.G., T.v.d.H., A.F.-P.S., G.H., and M.N. performed research;T.v.d.H. and M.N. contributed new reagents/analytic tools; R.J.C.G. and J.E.G.M. analyzeddata; and R.J.C.G. and J.E.G.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Abbreviations: eIF, eukaryotic initiation factor; IRES, internal ribosome entry site; MFC,multifactor complex.

Data deposition: The reconstructions reported in this paper have been deposited in theMacromolecular Structure Database, http://www.ebi.ac.uk/msd (accession nos. EMD-1327–EMD-1332).

§To whom correspondence should be addressed. E-mail: john.mccarthy@manchester.ac.uk.

This article contains supporting information online at www.pnas.org/cgi/content/full/0606880104/DC1.

© 2007 by The National Academy of Sciences of the USA

5788–5793 � PNAS � April 3, 2007 � vol. 104 � no. 14 www.pnas.org�cgi�doi�10.1073�pnas.0606880104

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0

(Fig. 2A, SI Fig. 6, and Table 1; see Materials and Methods fordefinition of resolution) shows good agreement with the smallsubunit within the previously published map of the yeast 80Sribosome filtered to a comparable resolution (14) (SI Fig. 6) andwith the atomic model of the yeast subunit generated by Frank andcolleagues (13) (Fig. 2 A and B and SI Fig. 6). Modest changes inthe relative arrangement of the head, body, and platform arehowever apparent (Fig. 2 A and B), as expected when comparingthe small subunit with the 80S ribosome (14). Of the four connec-tions between the head, platform, and body domains of the subunit(Fig. 2B), the strongest is between the head and the platform/bodyjunction just above rpS0A and close to the fitted position of RNAhelices 1/2 and 28. The latter form the covalent head/body link. Thenoncovalent links observed include the interaction between RNA

helices 18 and 34 and a weaker interaction between rpS5 and rpS14,which form the mRNA exit channel. The most obvious differencein the arrangement of the head, body, and platform in the isolatedsubunit compared with the subunit within the 80S ribosome is araising of the head (by �10 Å) above the body/platform, resting onthe noncovalent interactions just discussed (SI Fig. 6B). Fig. 2Cshows the 40S reconstruction with its head and platform/bodysurfaces rendered in different colors to assist in the interpretationof the data described below.

A reconstruction (14-Å resolution; see SI Materials and Methodsand Table 1) was also obtained for the complex of eIF2, eIF3, andeIF5 (Fig. 2D), which is equivalent to the MFC lacking eIF1. ThisMFC subcomplex was chosen because we observed that eIF1 (by farthe smallest component of the MFC) was substoichiometric in ourstandard MFC preparations, as was previously noted for MFCcomplexes isolated directly from yeast extracts (10) (Fig. 1A). Thereconstruction (Fig. 2D) confirms that these eIFs form a stable corecomplex, which comprises two ellipsoidal lobes of density orientedperpendicularly to one another and assists in our interpretation ofthe 43S reconstruction.

A number of analytical methods were used to follow the assemblyof complexes of 40S with different combinations of eIFs. Oneapproach was to include [35S]methionyl-tRNAi, so that complexformation could be followed via native-gel electrophoresis (16)(Fig. 1D). For example, in the presence of eIF1, eIF1A, [35S]Met-tRNAi.eIF2.GMP-PNP and eIF5, the addition of eIF3 led toformation of a band with shifted mobility corresponding to the 43S

S04

CF

M-S0411A523kD

eIFs

1758362

47.5

32.525

16.5

6.5

tS

M

35S-Met-tRNA

35S-Met-Puro

35S-MetmRNA++

GDPnP++++

43S(-eIF3)

35SMet-tRNAi+ +

+++-

Time(min)

40S-MFC

eIF1

eIF2?

eIF3i

eIF5

eIF1A

090

40S-MFC

40S

8

76

54

3

210

0 20 40 60 80 100Time (min)

53e

ditpe

pid

oru

P-teM-

S

+eIF5B

-eIF5BD

O452

*

*

+ +++++++++++-

++++

60SeIF5BeIF3

eIF1,1A,2,540S

--+

++

++- -

- -

43S(+eIF3)

80S

A B C

D E F

Fig. 1. Preparation and hydrodynamic behavior of 40S complexes (see SI Materials and Methods for full details). (A) eIFs, 40S, and 40S-MFC (43S) complex loadedonto an SDS/PAGE gradient gel. A few of the eIF proteins show very similar electrophoretic behavior to 40S proteins and thus generate multiple bands (markedby asterisks) in the 43S lane. The 43S complex loaded here was isolated as fractions (see vertical dotted lines in B) from a sucrose-density gradient. (B) A 254-nmabsorbance elution profile from a sucrose-density gradient; the presence of the eIFs in the 43S sucrose-gradient peak was routinely confirmed by means ofWestern blotting using the antibodies defined in SI Materials and Methods. (C) The 40S (gray, circles) and the 43S complex (red, diamonds) were subjected tosedimentation velocity analytical ultracentrifugation, yielding apparent sedimentation coefficient distributions. These were fitted with Gaussian equations(lines), giving sedimentation coefficients of 36.4S for the small subunit and 41.5S for the preinitiation complex. Corrected for finite solute and solvent effectson the basis that the small subunit value is 40S, the preinitiation complex value becomes 45.6S. (D) Assembly of complexes between 40S and MFC eIFs.PhosphorImages of native gels loaded with 40S–eIF complexes incorporating [35S]Met-tRNAi and GMP-PNP. The addition of eIF3 leads to a shift in the ribosomecomplex band (left-hand side). The further addition of eIF5B and the 60S subunit led to a further shift, associated with the formation of 80S (right-hand side).(E) A typical experiment following the kinetics of methionyl-puromycin synthesis by 80S complexes prepared by using the components shown in A plus 60S, eIF5B,methionyl-tRNAi, and GTP (instead of GMP-PNP). A PhosphorImage of samples from a typical experiment that were allowed to run on cation exchange TLC isshown. (F) Time course of formation of the 35S-labeled methionyl-puromycin product, taking the samples featured in E plotted against time. The y axis iscalibrated in relative signal intensity. Also plotted are data from an experiment in which eIF5B was omitted.

Table 1. Complexes studied by cryoEM in this work

Complexname Components

Resolution,Å

Particlesimaged

Mw,kDa

40S 40S 19 6,267 1,400MFC Met-tRNAi.eIF2.GMP-PNP,

eIF3, eIF514 19,946 556

43S Met-tRNAi.eIF2.GMP-PNP,eIF3, eIF1, eIF1A, 40S

20 31,756 1,941

Molecular masses of the component eIFs are (in kDa): eIF1, 12.3; eIF1A, 17.4;eIF2, 124.2; eIF3, 361.9; and eIF5, 45.2.

Gilbert et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 5789

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0

complex (see left-hand side of Fig. 1D). eIF3 is by far the largestsingle component of the MFC and in mammals has been shown tostabilize the 40S.Met-tRNAi

Met.eIF2.GTP complex (5). Here, wesee that eIF1:eIF1A:[35S]Met-tRNAi.eIF2.GMP-PNP:eIF5, as wellas the complex of these eIFs together with eIF3, are both stable. Infurther experiments, we investigated whether the 43S complex wascompetent to build the 80S complex by adding eIF5B and the 60Ssubunit. Previous work has shown that the addition of an excess ofGMP-PNP can stabilize 80S formation, most likely by inhibiting therelease of eIF5B that is normally triggered by GTP hydrolysis (17);likewise, we observed conversion of the 43S complex into 80S (Fig.1D, right- hand side).

We also investigated the functionality of the 43S complex formedby mixing the components shown in Fig. 1A. To do this, wesubstituted GTP for GMP-PNP in the 43S complex, added theremaining components required for assembly of the 80S complex,

i.e., eIF5B and 60S, and tested the assembled complex for its abilityto catalyze methionyl-puromycin formation (Fig. 1 E and F). Theresults show that the addition of 60S and eIF5B allowed us to‘‘chase’’ the 43S complex into an active 80S complex, thus confirm-ing that the components used in this study are functional. Althoughthere are no established ‘‘benchmark’’ figures for specific methio-nyl-puromycin synthesis rates catalyzed by such a complex fromyeast, this is a widely used indicator of functionality.

The cryo-EM reconstruction of the 43S complex, at 20-Å reso-lution, indicates that major conformational changes occur in the40S subunit on binding of the MFC (Fig. 3A). Multiple and entirelyindependent reconstructions of the 43S complex yielded verysimilar maps in which one region of the structure (identifiable as thesubunit head) was not joined by discernible density to the rest of thestructure. Because the single covalent link formed by helices 1/2 and28 between the head and the body cannot break, this apparentdetachment presumably derives from mobility in the head, such thatits link to the body is averaged out, and its form is smeared. Toinvestigate whether there were discrete subpopulations with betterdefined structures, we subclassified the 43S images (see SI Materialsand Methods) and were able to obtain three separate lower-resolution reconstructions (30 Å) in which the interactions betweenthe head and the rest of the structure were observable. Thesereconstructions (subclasses I–III, Fig. 3 B–D) contain images inwhich the very distinctive shape of the head is easily discernible.Subclass I derives from 6,864 of 31,756 images in the whole data set,subclass II from 8,269 of 31,756 images, and subclass III from 9,467of 31,756 images; these three subclasses therefore account for 77%of the data. In three of the four reconstructions shown in Fig. 3, thestructures of the body and platform are well defined and consistent;however, in subclass III, the head is particularly well ordered at theexpense of some blurring of the body and platform.

We tested our classification by looking, for some typical views ofeach of the three subclasses, at averages of the contributingprojections and also the variances within the averages (SI Fig. 7).The clear features in the averages, together with the low level ofvariance, show that the subclasses shown in Fig. 3 are genuine; thequality of these data seems equivalent to those found in othersimilar analyses (see, e.g., ref. 18). The relatively low resolution ofthe subclass reconstructions suggests that there is residual variabil-ity within them but markedly less than in the data set as a whole(SI Fig. 7D). We also performed a 3D variance and covarianceanalysis, using the recently developed methodology of Penczek andcolleagues (18–20) (see SI Materials and Methods, Fig. 3, and SI Fig.8). This indicated that the maximum variance in the reconstructionof the whole data set was associated with the head region and thejoin between the body and platform regions (SI Fig. 8 A and B) andthat the variance was reduced by subclassification (SI Fig. 8 C–E).Subclass III exhibited significantly more variance than subclasses Iand II, which may derive from its consisting of more particles and/orfrom preferential alignment of the head at the expense of the restof the structure. The features of the covariance (correlation) mapsfor significant variance peaks indicated that the high variancefeatures have only local correlations. These results indicate that thehead occupies a broad continuum of positions, which we havesampled by subclassification (Fig. 3E). The majority of the particles(represented by subclasses I and II) show a rotation within an arcof �25° that includes the locked conformation observed in theisolated 40S subunit, but a minority (�9,500; subclass III) displayrotation of up to 45°. The rotation observed is about the same axisas (although of greater magnitude than) that for the recentlydiscovered movement within prokaryotic ribosomes (15).

Careful analysis of the reconstruction of the 43S complex shownin Fig. 3A reveals further details about the interaction of the smallsubunit and MFC. The body and platform can be identified by theirpositions and characteristic shapes. The body has a similar confor-mation to that found in the isolated 40S, but aligning the head, body,and platform density of the 40S to that of the 43S does not account

Fig. 2. Cryo-EM density maps of the 40S subunit and MFC. (A) The 40Ssubunit reconstruction (6,267 particles; resolution 19 Å) contoured at 1.5� inthree orientations, with the atomic coordinates of the yeast 40S subunit (14)fitted: head rRNA, blue; platform, green; body, red; and helix 44, cyan. Thesmall subunit proteins are shown in magenta. Three orthogonal views aboutaxes as marked are shown, and major structural landmarks are labeled. (B)Close-up, sectioned views of the head–body/platform connections at a con-tour level of 1.5� in the same orientations and with the same color scheme asin A (thumbnail images indicate the directions of view). (C) Views of the 40Ssubunit reconstruction after computational segmentation, with the densityfor head colored blue and the platform and body together colored yellow.(Scale bar, 50 Å.) (D) Orthogonal views of the MFC reconstruction (19,446particles, 14-Å resolution). The scale is identical to that of A and C. (Scale bar,50 Å.)

5790 � www.pnas.org�cgi�doi�10.1073�pnas.0606880104 Gilbert et al.

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0

for all of the 43S density (SI Fig. 9B). By elimination, this densitycomprises the bound MFC. The volumes of the regions of densityidentified with the head, body, platform, and MFC in the 43Scomplex account well for their known masses and display highcorrelation coefficients and low R factors when fitted to equivalent

portions of the isolated subunit reconstruction or to the recon-struction of the MFC alone (see Table 1 and SI Materials andMethods). In Fig. 4A, we show the body, platform, and MFC regionsof the 43S reconstruction, colored as in Fig. 2 C and D. We alsoshow a superposition of the MFC density from the 43S reconstruc-tion and the isolated MFC structure (Fig. 4B). Their shapes are verysimilar, allowing for the loss of eIF5, the incorporation of eIF1 andeIF1A, the difference in resolution of the two maps (see SI Materialsand Methods), and changes consequent on binding the smallsubunit. The MFC lies between the body and platform, extending

Fig. 3. Reconstructions of the 43S complex. (A) The main figure (in yellow/blue) depicts orthogonal views of the 43S reconstruction at 20-Å resolution,incorporating 27,101 of 31,756 images. The head (h), body (b), and platform(p) are labeled. The head is colored blue, and the rest of the reconstruction isin yellow. The left-hand image is of the reconstruction in gray mesh with thebootstrap variance map computed from its data set displayed at 1.7� (green)and 1.77� (red) (the maximum peak in each case was 2�). (B–D) Equivalentorthogonal views to those shown in A for three subclass reconstructions of the43S, displaying the variable position adopted by the head. B was calculatedfrom 6,864 particles, C from 8,269 particles, and D from 9,467 particles. In eachcase, the resolution of the map is 30 Å. See SI Fig. 7 for representative classaverages and 2D variance maps for these subclasses of 43S particles. See SI Fig.8 for 3D variance analysis of these data; the variance map for the firstorientation displayed is shown on the left of each figure, as in A. (E) Orthog-onal views of the head regions of the reconstructions shown in B–D super-posed (B, green; C, blue; and D, red). On the right (upper thumbnail) is theatomic structure of the head of the ribosome from the T. thermophilus crystalstructure (21) (red) superimposed with reference to its body on that of the E.coli crystal structure (15) to demonstrate crystallographic evidence for themodes of head movement observed between the 40S and 43S structures. Thelower three thumbnails show the 40S head aligned on its body to the 43S foreach view. The thumbnails are 50% the size of the main images.

Fig. 4. Analysis of the 43S complex. (A) Orthogonal views of the platform/body region of the 20-Å 43S reconstruction (as in Fig. 3A) with the platformand body colored as in 2C and the MFC colored magenta, oriented such thatthe body is in the same orientation in each image as in Fig. 2C. Thumbnailsshow the 40S platform and body aligned by reference to the body of the 43Sfor each view. Thumbnails are 50% of the main image size. (B) Equivalentviews to A in which the 43S density corresponding to the reordered smallsubunit is colored gray (using the head from 3B for the head and the 20-Åstructure (A) for the body/platform) fitted with the atomic models for thehead, platform, and body (13), colored blue, green, and red. The MFC densityis shown as a mesh surface, colored magenta. Thumbnails of the MFC withinthe 43S (semitransparent surface) superimposed on the isolated MFC recon-struction (magenta mesh) are shown in each orientation. Asterisk, helix 18.Thumbnails are 75% of the main image size. (C) Schematic diagram of theconformational changes observed here in the 40S subunit (Left, inactive) onbinding the eIFs that form the MFC (Center). The head is colored blue and theplatform/body is yellow in each case, as in Figs. 2C and 3, with arrowsindicating the domain movements after eIF binding that release the subunithead. (Center) The head displays movement along with the covalent linkbetween it and the body. (Right) The small subunit has reverted to its usualstructure within the elongating ribosome with the binding of the large 60Ssubunit. h, head; p, platform; b, body; mfc, multifactor complex.

Gilbert et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 5791

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0

from the front to the back of the subunit (Fig. 4 A and B and SI Fig.10) and appears to act as a wedge, opening up the mRNA channelbetween the head, body, and platform of the subunit and leading tosubunit reorganization. In releasing the head from its stabilizingnoncovalent interactions with the body and platform, the MFC hasinduced a rotation of the platform around toward the solvent faceof the body and a spring-like movement of helix 18, which is veryclearly identifiable in the body density (Fig. 4B and SI Fig. 10, wherea stereoview is provided). However, it is likely that the conforma-tional change involving the platform–body interface cannot bemodeled in a simplistic fashion because it results from cumulativeeffects of many small changes (see below).

The underlying connectivity of the subunit domains is consonantwith the proposed conformational changes (13) (SI Fig. 11C). In SIFig. 11, we show the interactions among the four small subunitdomains of the Thermus thermophilus 30S subunit (1,21; PDB IDCode 1FJF). The backbones of these domains, on the basis of whichwe have colored both the 30S atomic structure and the yeast modelused in Figs. 2 and 4, all converge at the physical join of head, body,platform, and helix 44 in the small subunit (SI Fig. 11). The platformtherefore has a single covalent link to the body, with the rRNApassing from body to platform at nucleotide �537 and returningfrom platform to body at nucleotide �860, with the regionscontaining nucleotides 537 and 860 being close in space (SI Fig. 11).The platform otherwise engages in a clasping interaction with thebody, with helix 21 wrapped around onto the solvent face of thesmall subunit. Given the conformational changes already observedfor the 30S, it is entirely conceivable that this clasping could beloose, permitting rotation in the platform–body region.

In SI Fig. 12, we show how the underlying architecture of thedensity in our reconstruction of the 43S matches the underlyingrRNA structure as defined crystallographically and in SI Fig. 11. Ahigh-density tube connects the body and platform regions of thesmall subunit within the 43S complex, and, with helix 18 aligned toits visible density (Fig. 4B), the yeast model of the subunit rRNAaligns such that the covalent connection between the central(platform) and 5� (body) domains align with this tube. Thus, thedensity distribution in our 43S reconstruction matches the crystal-lographically defined rRNA structure. The conformational changesin the platform/body between the 40S and 43S structures requiresmuch further detailed characterization at higher resolution, but inaddition to the rearrangement of electrostatic bonds (e.g., thoseinvolving helix 21) may also involve morphing of the platform/bodystructure as a whole. Such morphing is a well documented aspectof ribosome structure/function (1, 15, 22). In support of this, weperformed an analysis of the solvent content of the small subunit,using a crystal structure of the prokaryotic 30S subunit (21), andfollowing the work of Voss and coworkers who showed the largesubunit to be riddled with solvent channels in a manner reminiscentof a sponge (23). Using this approach, we showed the presence ofa number of large solvent cavities and channels within the subunitthat would allow conformational changes by solvent extrusion andthat the solvent content of the small subunit is similar to that of thelarge (SI Fig. 13). Interestingly, the regions that are most implicatedin conformational rearrangements by our study contain the mostextensive regions of solvent. Thus, the rearrangements could verylikely be achieved without the breakage of a large number of(noncovalent) bonds but by flexing around solvent voids. Suchremodeling is a known factor in ribosome function, e.g., the gatingof the mRNA exit channel (24) and the movement of the L7/L12stalk (25).

DiscussionComparison of our results with previous studies is informative. Anearlier 48-Å reconstruction of a negatively stained eIF3-containingcomplex of 40S obtained from rabbit reticulocyte lysate (26)concluded that eIF3 is located on the solvent face in a similarposition to that identified here and that, in the absence of the other

MFC eIFs, it causes no more than a modest conformationalrearrangement of the rabbit 40S subunit. This conclusion is con-sistent with the findings reported here, because we observe that thewhole MFC, including the other eIFs, is responsible for opening upthe head/body/platform junction. A more recent study (27) re-ported reconstructions of human eIF3 both in isolation and incomplex with the hepatitis C virus internal ribosome entry site(IRES). The eIF3-IRES structure was then docked in silico onto apreviously determined structure of a mammalian ribosome incomplex with the IRES alone, and this fit was compared with anegative stain study of the 40S– complex. The results are consistentwith our data in that they correspond to a different (most likelylater) stage in the initiation pathway at which the conformationalchanges associated with recruitment have been completed (indeed,the yeast 40S subunit is incapable of mRNA recruitment in thepresence of eIF3 alone).

Unfortunately this earlier work on the mammalian eIF3 is at toolow a resolution to allow a detailed structural comparison with ourresults showing the yeast eIF3 in the context of the MFC. For thesame reason, it is also not possible for us to compare our MFCreconstruction meaningfully with the isolated human eIF3; inaddition, human eIF3 is more than twice the mass of the eIF3 ofbudding yeast; 807 kDa compared with 362 kDa. Similarly, becauseIRES elements bypass the scanning stage in initiation [and indeedIRES-mediated translation initiation requires only a subset, ornone, of the eIFs involved in cap-dependent initiation (27–29)], thestudy of 80S human ribosomes bound with the cricket paralysis virus(CrPV) IRES by cryo-EM (28) generated a reconstruction that islikely to be equivalent to a state of the 40S subunit normallyachieved downstream of mRNA insertion into the channel. Evenmore pointedly, the ribosome/CrPV IRES interaction involved an80S ribosome because of the so-far unique characteristics of theCrPV IRES (28). Overall, the previous studies of 40S-IRESinteractions (28, 29) suggest that there are at least two different setsof interactions that can clamp the 5� end of an IRES to the 40Ssubunit. There may therefore be distinct structural strategies forbinding each type of IRES in the correct position. Moreover, theseearlier reconstructions do not indicate how the 40S subunit solvesthe topological problem of inserting the mRNA into the ribosomechannel, most likely because they represent a step in the IRES-mRNA recruitment process that occurs after loading into themRNA channel.

Our work suggests that, for capped message, the topologicalproblem of mRNA loading is solved in a general way by majorconformational changes that have a significant impact on therelationship between the head and body of the 40S subunit (Fig.4C). This rearrangement will necessarily remodel the channelthrough the 40S subunit identified as the conduit for mRNA (13,29, 30). In the isolated 40S subunit, this channel is relatively narrow,such that it is hard to envisage diffusion processes threading a 5�capped mRNA through it, and there is no effective bindingmechanism to facilitate vectorial, 5�-specific entrance of the mRNAinto it (compare Figs. 2A, 3 B–D, and 4B), consistent with the 40Ssubunit not being recruited on to the 5� end of capped mRNA inthe absence of the eIFs. In this article, we have not characterizeddirectly the (in vivo) functional consequences of the conformationalchange but have managed to capture a structural snapshot thatsuggests a model in which the levering open of the noncovalenthead–body-platform contacts by the MFC renders the mRNAchannel accessible to introduction of the 5� nucleotides of themRNA mediated by the cap-binding complex eIF4F. This confor-mational change is much greater that those seen in prokaryotes,where initiation does not involve the loading of mRNA associatedwith a large cap-binding complex through the narrow mRNAentrance channel (21), and it seems that eukaryotes have evolveda translation machinery (eIFs and ribosome) that promotes generalguided access to the channel for any capped mRNA through largeconformational changes in the 40S subunit.

5792 � www.pnas.org�cgi�doi�10.1073�pnas.0606880104 Gilbert et al.

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0

Materials and MethodsFurther details are provided in SI Materials and Methods.

Sample Preparation. The 43S complexes were prepared by mixing 50pmol of 40S with 150 pmol each of eIF2, eIF3, eIF5; 500 pmol ofeIF1; and 5 nmol of eIF1A in a reaction volume of 250 �l. We thenadded 150 pmol of Met-tRNAi

Met, ATP and GMP-PNP (finalconcentrations of 1 mM and 0.4 mM, respectively). Reactions wereincubated at 26°C for 10 min, loaded on 10–30% gradients ofsucrose in buffer A and centrifuged at 39,000 rpm for 4 h (SW 40Tirotor). Peak fractions were dialyzed in buffer A for 2 h at 4°C andconcentrated to OD260 � 1.5–2.5.

Assaying Formation of 40S-eIF Complexes and Methionyl-Puromycin.The native gel assay for the formation of ribosomal complexes (RC)and the methionyl-puromycin assay were performed as described(17), except that purified 40S and 60S fractions, rather than 80Sfractions, were used. The various components (including themRNA 5�-GGAA[UC]7UAUG[CU]10C-3�) were brought togetherin 38 mM Hepes-KOH (pH 7.4), 135 mM KOAc, 3.25 mMMg(OAc)2, and 4% glycerol. GTP or GMP-PNP were used, asappropriate. Incubations of 25 �l were run on a 4% acrylamide gel(37.5:1 acrylamide:bisacrylamide) in THEM buffer [66 mMHepes/34 mM Tris base/2.5 mM MgCl2/0.1 mM EDTA (final pH7.5)] for 6 h at 8 W. Formation of the 80S complex was achieved byalso including eIF5B and 60S (17). In the puromycin assay, allcomponents except puromycin were initially preincubated at 28°Cfor 10 min. The reactions were then started through the addition ofpuromycin. Two-microliter aliquots were withdrawn at differenttime points and quenched in 0.5 �l of 3 M NaOAc (pH 4.6).Samples were spotted onto Polygram IONEX-25 SA-Na cationexchange TLC plates (Macherey–Nagel, Duren, Germany). Theplates were developed in 2M NH4OAc, pH5.2 plus 10%acetonitrile.

Cryoelectron Microscopy. Images of plunge-frozen microscope gridswere captured by using a CM200 field emission gun cryoelectronmicroscope operating under low dose conditions. The micrographswere scanned by using a UMAX PowerLook 3000 scanner on an8.322-�m raster. Images were excised by using XIMDISP (31) and

SPIDER (32), or EMAN (33). The CTF of each micrograph wasdetermined, and the data deriving from it was corrected by phase-flipping. For all samples, reconstructions were computed ab initioby using IMAGIC (34), with refinement in EMAN (33) andSPIDER (32). Reconstructions were computed for both the wholedata set and for three subclasses, as described in SI Materials andMethods. Variance and covariance analyses were carried out byusing the methods described by Penczek and colleagues (19). SI Fig.14 shows the isotropic distribution of the image orientations usedin the 43S reconstruction. SI Fig. 15 shows determination of thehand of the 43S reconstruction, which also serves as a validation ofits correctness.

Fitting and Analysis of Atomic Structures. Atomic models were firstfitted manually and then automatically by using the program URO(35). Analysis of the solvent volume within the small subunit wascarried out according to the method of Voss and colleagues (23).

Fitting of Densities. Maps, or fragments thereof, were fitted to eachother by using the program GAP (24, 36), refining relative rotationsand translations by the method of steepest descent. The targetfunctions were real space correlation coefficients and R-factors.

Accession Numbers. The reconstructions reported in this manuscripthave been deposited in the Macromolecular Structure Database(http://www.ebi.ac.uk/msd) with accession numbers EMD-1327–EMD-1332.

We thank John Hershey (Davis, CA), Alan Hinnebusch [National Institutesof Health (NIH), Bethesda, MD], Tom Dever (NIH, Bethesda, MD), andGraham Pavitt (University of Manchester) for providing expression plas-mids. Robert Esnouf, Richard Brimacombe, Peter Rosenthal, and RichardHenderson provided valuable advice. Images used in this analysis werecaptured on CM200 microscopes at the Division of Structural Biology,University of Oxford and at the European Molecular Biology Laboratory(EMBL), Heidelberg, Germany. We are grateful to the EMBL, and inparticular to Kenneth Goldie. This research was supported by grants fromthe Wellcome Trust and the Biotechnology and Biological Sciences Re-search Council (U.K.). R.J.C.G. is a Royal Society University ResearchFellow and D.I.S. is an Medical Research Council Research Professor.J.E.G.M. is a Wolfson–Royal Society Research Fellow and a member ofboth the School of Chemical Engineering and Analytical Science and theFaculty of Life Sciences at the University of Manchester.

1. Ramakrishnan V (2002) Cell 108:557–572.2. McCarthy JEG (1998) Microbiol Mol Biol Rev 62:1492–1553.3. Kozak M (2002) Gene 299:1–34.4. Spirin AS (2002) FEBS Lett 514:2–10.5. Hershey JWB, Merrick WC (2000) in Translational Control of Gene Expression (eds

Sonenberg N, Hershey JWB, Mathews MB (Cold Spring Harbor Lab Press, ColdSpring Harbor, NY), pp 33–88.

6. Chaudhuri J, Si K, Maitra U (1997) J Biol Chem 272:7883–7891.7. Donahue T (2000) in Translational Control of Gene Expression (eds Sonenberg N,

Hershey JWB, Mathews MB (Cold Spring Harbor Lab Press, Cold Spring Harbor,NY), pp 487–502.

8. Pestova TV, Borukhov SI, Hellen CU (1998) Nature 394:854–859.9. Singh CR, He H, Ii M, Yamamoto Y, Asano K (2004) J Biol Chem 279:31910–

31920.10. Asano K, Clayton J, Shalev A, Hinnebusch AG (2000) Genes Dev 14:2534–2546.11. Phan L, Schoenfeld LW, Valasek L, Nielsen KH, Hinnebusch AG (2001) EMBO

J 20:2954–2965.12. He H, von der Haar T, Singh CR, Ii M, Li B, Hinnebusch AG, McCarthy JEG,

Asano K (2003) Mol Cell Biol 23:5431–5445.13. Spahn CM, Beckmann R, Eswar N, Penczek PA, Sali A, Blobel G, Frank J (2001)

Cell 107:373–386.14. Spahn CM, Gomez-Lorenzo MG, Grassucci RA, Jorgensen R, Andersen GR,

Beckmann R, Penczek PA, Ballesta JP, Frank J EMBO J 23:1008–1019.15. Schuwirth BS, Borovinskaya MA, Hau CW, Zhang W, Vila-Sanjurjo A, Holton

JM, Cate JHD (2005) Science 310:827–834.16. Lorsch J, Herschlag D (1999) EMBO J 18:6705–6717.17. Shin B-S, Maag D, Roll-Mecak A, Arefin MS, Burley SK, Lorsch JR, Dever TE

(2002) Cell 111:1015–1025.18. Grob P, Cruse MJ, Unouye C, Peris M, Penczek PA Tjian R, Nogales E (2006)

Structure (London) 14:511–520.

19. Penczek PA, Yang C, Frank J, Spahn CMT (2006) J Struct Biol 154:168–183.20. Penczek PA, Frank J, Spahn CMT (2006) J Struct Biol 154:184–194.21. Wimberly BT, Brodersen DE, Clemons WM, Morgan-Warren RJ, Carter AP,

Vonrhein C, Hartsch T, Ramakrishnan V (2001) Nature 407:327–339.22. Ogle JM, Murphy FV, Tarry MJ, Ramakrishnan V (2002) Cell 111:721–732.23. Voss NR, Gerstein M, Steitz TA, Moore PB (2006) J Mol Biol 360:893–906.24. Gilbert RJC, Fucini P, Connell S, Nierhaus KH, Robsinson CV, Dobson CM,

Stuart DI (2004) Mol Cell 14:57–66.25. Agrawal RK, Heagle AB, Penczek PA, Rassucci RA, Frank J (1999) Nat Struct

Biol 6:643–647.26. Srivastava S, Verschoor A, Frank J (1992) J Mol Biol 226:301–304.27. Siridechadilok B, Fraser CS, Hall RJ, Doudna JA, Nogales E (2005) Science

310:1513–1515.28. Spahn CMT, Jan E, Mulder A, Grassucci RA, Sarnow P, Frank J (2004) Cell

118:465–475.29. Spahn CM, Kieft JS, Grassucci RA, Penczek PA, Zhou K, Doudna JA, Frank J

(2001) Science 291:1959–1962.30. Dube P, Bacher G, Stark H, Mueller F, Zemlin F, van Heel M, Brimacombe R

(1998) J Mol Biol 279:403–421.31. Crowther RA, Henderson R, Smith RM (1996) J Struct Biol 116:9.32. Frank J, Radermacher M, Penczek P, Zhu J, Li Y, Ladjadj M, Leith A (1996) J

Struct Biol 116:190–199.33. Ludtke SJ, Baldwin PR, Chiu W (1999) J Struct Biol 128:82.34. van Heel M, Harauz G, Orlova EV, Schmidt R, Schatz M (1996) J Struct Biol

116:17–24.35. Navaza J, Lepault J, Rey FA, Alvarez-Rua C, Borge J (2002) Acta Crystallogr D

58:1820–1825.36. Grimes JM, Burroughs JN, Gouet P, Diprose JM, Malby R, Zientara S, Mertens

PP, Stuart DI (1998) Nature 395:470–478.

Gilbert et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 5793

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 2,

202

0