Supporting Information

Zhou et al. 10.1073/pnas.0801313105 SI Text Falcon tubes for 15 min to clear the bulk of the cellular debris, ESI and MALDI MS Analysis of the Human eIF3 Subunits. After LC-MS and the supernatant was decanted and incubated with gentle separation, the effluent from monolithic column was passed mixing for 1 hour at 6°C with IgG-Fast Flow Sepharose resin (GE through a nanospray ionization interface into a QSTAR Elite Healthcare), using Ϸ400 ␮l of 50% resin suspension (prewashed mass spectrometer (Applied Biosystems). The eluent was split so with the same IPP150 buffer) per 40 ml of lysate supernatant. that 1 ␮l/min was directed to a 50-␮m ID PicoTip emitter (New The resin was collected in a Polyprep column (Bio-Rad) using Objective) maintained at 2,500 V. The first and second declus- manually applied pressure from a disposable syringe to maintain tering potentials were 30 and 15 V, respectively. high flow rates. The resin was washed three times with 20 ml of To identify the proteins, LC separation was repeated and the TEV cleavage buffer [10 mM Tris-HCl (pH 8), 150 mM NaCl, effluent from the monolithic column connected to a Probot 1.0 mM DTT) containing no detergent, resuspended in 3 ml of sample fraction system (Dionex) for spotting onto a MALDI the same buffer, and incubated with 600 units of AcTEV plate for digestion and subsequent sequence analysis. Auto- protease (Invitrogen) at room temperature for 1 h. After TEV mated spotting was carried out at 10 s intervals onto a MALDI cleavage, the solution was collected and the resin washed with plate prespotted with 0.5 ␮l of trypsin (0.16 ␮g/␮l) (Promega). another 2 ml of the same buffer. Analysis of the eluted proteins The dry sample spots were then overlaid with 0.5 ␮lof25mM by SDS/PAGE and proteomics experiments revealed the pres- ammonium bicarbonate, digested for 10 min in a humidifier, ence of all five of the ‘‘core’’ yeast eIF3 subunits (a, b, c, g, and vacuum-dried and overlaid with ␣-cyano-4-hydroxycinnamic i) plus eIF5 (Fig. S5). The yeast equivalent of eIF3j (Hcr1) was acid matrix before analysis by automated MALDI-TOF/TOF- not observed. Other weaker bands were often present but varied MS/MS (Applied Biosystems 4700). Data-dependent peak se- in intensity between different preparations. The combined elu- lection of the three most abundant MS ions was used for CID. ate and resin washings were concentrated to Ϸ200 ␮l and then Ar was used as the collision gas. Resulting data were analyzed by purified/buffer-exchanged by size-exclusion chromatography on GPS Explorer (Applied Biosystems), which involved a MAS- a Sephadex 200 column (GE Healthcare) in 200 mM ammonium COT (Matrix Science) database search using all entries in the acetate (pH 7). The complex eluted close to the void volume of National Center for Biotechnology Information (NCBI) data- the column (Ϸ600 MDa), indicating a larger apparent mass than base to determine candidate peptides. Mass tolerance for the the expected for a globular protein of 411 kDa, the combined precursor ion was set to 100 ppm; mass tolerance for the mass of the core yeast eIF3 subunits plus eIF5 (data not shown). fragment ions was 0.3 Da. Manual validation was performed for each MS/MS spectrum. Expression and Analysis of eIF3 Subcomplexes Formed in Insect Cells. Baculoviruses expressing N-terminally Flag-tagged eIF3 sub- Preparation of the Yeast eIF3. Yeast cultures were grown in YPD units e, f, h, k, and l and untagged subunits e and k were medium [10 g/liter yeast extract, 20 g/liter Bacto-Peptone (both generated from the appropriate pcDNA5 vectors (3) essentially from BD Bioscience); 20 g/liter glucose (Sigma)] in batches of as described (4). HA-tagged eIF3b was made by excising the 500 ml of culture per 2-liter flask, inoculated with 2.5 ml of HA-3b coding region from pAMV-pA-HA (3) and inserting it overnight culture, and incubated at 30°C and 220 rpm for 22 h. into pFASTBAC1 (Invitrogen). The eIF3m sequence (GenBank (N.B. At this point, the yeast culture was determined by OD600 accession no. BC019103.2) was amplified by PCR and inserted to have passed the diauxic shift—the point where the glucose in into the NdeI-XhoI sites of pET28c to generate His-6-tagged the medium is exhausted and cells switch to ethanol as a carbon eIF3m, which was subsequently moved into pFASTBAC1. For source—but not to have reached the stationary phase. In this Flag-eIF3m, the PCR product was directly inserted into FLAG- state, the mRNA levels of proteins involved in translational FASTBAC1 (4). processes are lower than before the diauxic shift (1), but the total Sf9 insect cells were coinfected with various combinations of recoverable eIF3 is still higher due to the increased total cell baculoviruses designed to express human eIF3 subunits and density.) Cell pellets from a total of 12 liters were combined, analyzed essentially as described by Fraser et al. (4). Cells were flash-frozen, and stored at Ϫ80°C until processed. Cells were incubated for 72 h, lysed, and subjected to immunoprecipitation lysed by mechanically from partially defrosted cell pellets using with antibody-bound beads (Sigma) against the Flag or HA a bead-beater in the IPP150 buffer described by Gould et al. (2) epitope. The beads were washed, and bound proteins were eluted with added protease inhibitors (Pierce). The total lysate, plus with SDS buffer and subjected to SDS/PAGE, followed by further washings of the glass beads, was spun at 2,000 ϫ g in immunoblotting or staining with Coomassie blue.

1. DeRisi JL, Iyer VR, Brown PO (1997) Exploring the metabolic and genetic control of gene 4. Fraser CS, et al. (2004) The j-subunit of human translation initiation factor eIF3 is expression on a genomic scale. Science 278:680–686. required for the stable binding of eIF3 and its subcomplexes to 40 S ribosomal subunits 2. Gould KL, Ren L, Feoktistova AS, Jennings JL, Link AJ (2004) Tandem affinity purifica- in vitro. J Biol Chem 279:8946–56. tion and identification of protein complex components. Methods 33:239–244. 3. Zhang L, Pan X, Hershey JW (2007) Individual overexpression of five subunits of human translation initiation factor eIF3 promotes malignant transformation of immortal fibroblast cells. J Biol Chem 282:5790–800.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 1of13 Identified Subcomplexes Connectivity Map Inferences

0 k-l-e-d c a-b-g-i h-f-m j 1 k-l-e-d c b-g-i h-f-m I. Fidelity of Subcomplexes II. Minimal Connectivity 2 e-d c b-g-i h-f-m 3 k-l-e-d c h-f-m Pairs Modules Building on Connecting 4 k-l-e-d c h-f 5 Modules Modules k-l-e-d c h a b 6 k-l-e-d c 7 c a-b-g-i b g i 8 e a-b-g-i a b i 9 g a b k-l-e-d c g-i i g g 10 l-e-d c h-f-m Module A c 11 l-e-d c h-f b i 12 l-e-d c h A + 30+31 13 l-e-d c 14 k-l-e c k l i 15 k-l-e-d i l e d a b a 16 l-e-d k l e g g b 17 e-d k e d Module B e A + 8 e 18 b-g-i l c 19 g-i d f 20 b-g h f m m 21 h-f-m f k l e d 22 f-h h f h B + 13+14 23 h---m m c 24 h m f-m Module C 25 k-l-e 26 k-l k l e d 27 l-e c f 28 b---i h 29 a-b B + C + 12 m 30 c b---i yeast 31 c b-g 32 g-i

Fig. S1. Here, we illustrate the process of building a connectivity map for the eIF3 complex, using only the observed subcomplexes from Table 1 and the data from the yeast eIF3 complex (presented in a column-aligned format at the left). The connectivity map is inferred through application of two hypotheses: (i). The ‘‘Fidelity of Subcomplexes,’’ which states that the intersubunit connections present in the observed subcomplexes are interactions retained from the original full complex. (ii) ‘‘Minimal Connectivity,’’ which states that the full complex contains the smallest set of connections between subunits required to explain those present in multiple subcomplexes. The first hypothesis allows us to identify connected pairs of subunits (indicated by red bars), and we define a pairwise- connected module as a set of pairs with common subunits. Minimal connectivity allows us to build up and connect modules from the subcomplexes and modules denoted below each complex. We use a colored bar and circle to indicate an interaction between a subunit and one or more subunits enclosed within the circle. The resulting ‘‘minimal map’’ contains the same overall connectivity as the model in Fig. 5.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 2of13 Fig. S2. MS/MS spectra of in-solution-generated subcomplexes were used to confirm their compositions. (a) eIF3l:k dimer. (b) eIF3i:g dimer. (c) eIF3k:l:e. (d) eIF3b:g:i. (e) eIF3k:l:e:d. (f) eIF3k:l:e:d:c. (g) eIF3k:l:e:d:c:h:f and eIF3k:l:e:d:c:h:f:m.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 3of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 4of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 5of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 6of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 7of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 8of13 Fig. S2. Continued.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 9of13 Fig. S3. Immunoprecipitation of baculovirus-expressed eIF3 subunits. (a) eIF3 binary complexes with eIF3b. Recombinant baculoviruses expressing HA-tagged eIF3b (3b) and/or recombinant viruses expressing a Flag-tagged e-, f-, h-, k-, l- or m-subunit (3x) of eIF3 were used to coinfect Sf9 cells as described in Experimental Methods. The combinations of viruses used are indicated at the top of each row, with 3x identified along the left side of the figure. Cells were lysed, and then aliquots were subjected to immunoprecipitation with an anti-HA resin or an anti-Flag resin, followed by SDS/PAGE and immunoblotting with anti-HA (␣-HA) antibodies to detect the 3b subunit or with anti-Flag (␣-Flag) antibodies to detect the second eIF3 subunit (3x). Regions of the immunoblot corresponding to HA-eIF3b and the various Flag-eIF3x subunits were excised and photograghed. The asterisk identifies the region where the coimmunoprecipitated protein is expected to migrate. In the cases of analyses with eIF3f and eIF3h, the HA-eIF3b protein when expressed alone bound weakly to the anti-Flag resin; however, in the presence of F-3f or F-3h, its abundance was substantially increased, indicating a binary complex. In the case of eIF3k, both HA-3b and F-3k alone bound to both resins, and coexpression did not enhance either intensities, indicating an absence of a stable 3b–3k interaction. (b and c) eIF3 trimeric subcomplex. (b) eIF3fhm. Combinations of recombinant baculoviruses encoding Flag-tagged eIF3h (F-3h) or eIF3m (F-3m), untagged eIF3f (3f) or His6-tagged eIF3m (H-3m) were expressed in Sf9 cells and analyzed as described in a. The upper rows identify which eIF3 subunit is expressed, and the probe for each gel or immunoblot is shown on the left. Lane 5 contains four times less virus expressing His6-eIF3m than lanes 2 or 4. The absence of a detectable protein band in lanes 2, 3, and 6 indicates that eIF3f and His6-eIF3m do not bind to anti-Flag beads in the absence of Flag-eIF3h. (c). eIF3ekl. Combinations of recombinant baculoviruses were used to express Flag-tagged eIF3l or eIF3k (F-3l; F-3k) and untagged eIF3e (3e) and eIF3k (3k) in Sf9 cells as described in Experimental Methods. The viruses used for each infection are identified at the top of each row. Infected cells were lysed and subjected to immunoprecipitation with anti-Flag resin and SDS/PAGE. The gels were analyzed by staining with Coomassie blue (stain) to detect F-eIF3l and untagged eIF3k, and by immunoblotting with anti-eIF3 antiserum (␣-3e) to detect untagged eIF3e and with anti-Flag antibodies (␣-Flag) to detect Flag-eIF3k. Regions corresponding to the proteins identified on the left were excised and photographed. The absence of a protein band in lanes 2 and 3 indicates that eIF3e and eIF3k do not bind to anti-Flag beads in the absence of Flag-eIF3l.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 10 of 13 Fig. S4. Comparison of MS spectra of eIF3 (a), the HCV IRES-eIF3 binary complex (b), tandem mass spectrum of the HCV IRES-eIF3 binary complex (c), and insert (d) mass spectrum of HCV IRES RNA with its mass measured as 107,953 Ϯ 66 Da (calculated mass is 107,018 Da; see sequence below). Binding of IRES to eIF3 results in a shift in m/z value. Comparison of the free and IRES-bound eIF3 shows that eIF3 is stabilized by binding to IRES. There is also a deterioration in the resolution of charge states within all spectra shown here because of the necessity of using the binding buffer and magnesium adduct formation on the exposed RNA surface. The dissociation of eIF3i subunit (purple box) and l:k dimer (blue box) were unaffected by IRES binding, whereas the peaks corresponding to i:g and f:h dimer (yellow box and expansion) greatly decrease upon IRES binding. Spectra of IRES–eIF3 binary complex and the eIF3 control were acquired under identicalMS conditions, capillary voltage of 1,250 V, and declustering potentials of 60 V and 15 V. (c) Tandem MS of the isolation of ions at 13,800 m/z with an acceleration voltage of 180 V gives rise to marginally resolved charge state peaks of the IRES–eIF3 complex, allowing its mass to be measured as 907 kDa and releasing the individual subunit eIF3k. Inset (d) MS spectrum of HCV-IRES RNA alone acquired in 150 mM AmAc on the QSTAR XL mass spectrometer using positive ion mode with declustering potentials of 100 V and 15 V. Sequence of HCV-IRES RNA used in this study: UCCCCUGUGAGGAACUACUGUCUUCACG- CAGAAAGCGUCUAGCCAUGGCGUUAGUAUGAGUGUCGUGCAGCCUCCAGGACCCCCCCUCCCGGGAGAGCCAUAGUGGUCUGCGGAACCGGUGAGUACACCGGAA- UUGCCAGGACGACCGGGUCCUUUCUUGGAUAAACCCGCUCAAUGCCUGGAGAUUUGGGCGUGCCCCCGCAAGACUGCUAGCCGAGUAGUGUUGGGUCGCG- AAAGGCCUUGUGGUACUGCCUGAUAGGGUGCUUGCGAGUGCCCCGGGAGGUCUCGUAGACCGUGCACCAUGAGCACGAAUCCUAAACCUCAAAGAAAAA

Zhou et al. www.pnas.org/cgi/content/short/0801313105 11 of 13 Fig. S5. SDS/PAGE analyses of TAP-tagged eIF3 purification from 11-liter yeast cell culture. (A) Prt1 (eIF3b) tagged purification, using IgG affinity only: lane 1, MW marker (Seeblue ϩ 2; Invitrogen); lane 2, cell lysis supernatant; lane 3, IgG resin unbound fraction; lane 4, IgG resin combined washes; lane 5, post-TEV protease elution; lane 6, post-TEV further washings. (B) Nip1 (eIF3c) tagged IgG then calmodulin affinity purification: lane 1, EGTA elution from calmodulin; lane 2, combined calmodulin resin washes; lane 3, calmodulin resin unbound fraction; lane 4, post-EGTA elution calmodulin resin resuspension; lane 5, post-TEV elution IgG resin resuspension; lane 6, MW marker.

Zhou et al. www.pnas.org/cgi/content/short/0801313105 12 of 13 Table S1. Masses of human eIF3 subunits measured by capillary HPLC-MS analysis Subunit Calculated Measured name mass, Da mass, Da eIF3a 166,569 Not resolved eIF3b 92,492 92,561 Ϯ 43.5 eIF3c 105,344 106,855 Ϯ 40 eIF3d 63,973 64,046.7 Ϯ 1.4 eIF3e 52,221 52,133.4 Ϯ 0.2 eIF3f 37,564 37,475.6 Ϯ 0.2 eIF3g 35,611 35,481.1 Ϯ 0.4 eIF3h 39,930 39,842.6 Ϯ 0.4 eIF3i 36,502 36,503.2 Ϯ 0.4 eIF3j 29,062 28,974.2 Ϯ 0.3 eIF3k 25,060 24,971.1 Ϯ 0.2 eIF3l 66,727 66,640.2 Ϯ 0.5 eIF3m 42,503 42,414.7 Ϯ 0.2

Zhou et al. www.pnas.org/cgi/content/short/0801313105 13 of 13