Structural integrity of the PCI domain of eIF3a/TIF32 is required for mRNA recruitment to the 43S pre-initiation complexes

Transfer of genetic information from genes into proteins is mediated by messenger RNA (mRNA) that must be first recruited to ribosomal pre-initiation complexes (PICs) by a mechanism that is still poorly understood. Recent studies showed that besides eIF4F and poly(A)-binding protein, eIF3 also plays a critical role in this process, yet the molecular mechanism of its action is unknown. We showed previously that the PCI domain of the eIF3c/NIP1 subunit of yeast eIF3 is involved in RNA binding. To assess the role of the second PCI domain of eIF3 present in eIF3a/TIF32, we performed its mutational analysis and identified a 10-Ala-substitution (Box37) that severely reduces amounts of model mRNA in the 43–48S PICs in vivo as the major, if not the only, detectable defect. Crystal structure analysis of the a/TIF32-PCI domain at 2.65-Å resolution showed that it is required for integrity of the eIF3 core and, similarly to the c/NIP1-PCI, is capable of RNA binding. The putative RNA-binding surface defined by positively charged areas contains two Box37 residues, R363 and K364. Their substitutions with alanines severely impair the mRNA recruitment step in vivo suggesting that a/TIF32-PCI represents one of the key domains ensuring stable and efficient mRNA delivery to the PICs.

Crystallization of the a/TIF32 276-494 domain and crystal structure determination X-ray diffraction images were collected at 100 K at beamline 14.1 (BESSY, Berlin, Germany; (3)) equipped with a MAR Mosaic 225mm CCD detector (Norderstedt, Germany). The oscillation images were indexed, integrated, and merged using the XDS package (4,5) to the final resolution of 2.65 Å for the native and to 2.79, 2.81 and 2.97 Å for Se-Met derivative crystal (peak, inflection and remote datasets, respectively). The crystal structure of a/TIF32 was solved by means of SAD using the Se-Met dataset at the peak wavelength in SHARP/autoSHARP (6). Within autoSHARP the heavy atom search was performed by SHELXD (7) and resulted in localizing four heavy atom positions that were further refined using SHARP followed by density modification in Solomon (8) and automatic model building in Arp/wARP (9). The resulting protein model comprising 172 amino acids has been refined using torsion angle dynamics in CNS (10) and manually rebuild and verified in Coot (11) against Simulated Annealing (SA) omit maps. The model comprising 217 residues (from 6 to 223, including three methionines) and belonging to one a/TIF32 monomer has been refined using PHENIX (12) at resolution of 2.65 Å to R and R free factors of 29.63% and 33.85%, respectively. Unusually high R-factors in concert with the solvent content of 72% and one additional (the fourth) Se peak found during the heavy atom search suggested the presence of additional protein molecule or its fragment in the asymmetric unit. The presence of the second full length a/TIF32 molecule in the asymmetric unit was unlikely due to the moderate resolution and resulting low solvent content of 42%. Performed MS analysis could not confirm any proteolytic digestion of protein samples obtained from dissolved crystals (data not shown). Difference electron density maps calculated with PHENIX (2mFo-DFc and mFo-DFc contoured at 1 and 3 sigma respectively) did not indicate any missing protein fragments as a few small separated blobs of density could not be interpreted as even a small part of a polypeptide chain. However the difference electron density maps (3mFo-2DFc, mFo-DFc) calculated with CNS revealed the presence of most probably some polypeptide fragments in the solvent channels, although the maps were highly diffused and not interpretable (no secondary structure elements could be recognized). In order to test if fragments of mostly α-helical a/TIF32 monomer could be the source of diffused electron density maps observed in solvent channels, we decided to carry out several molecular replacement (MR) searches with short a/TIF32 fragments as search models using PHASER (13) and keeping the already refined model as the fixed partial solution. About 650 short polypeptide fragments differing in length (25, 30, 35, 40, 55, 60 aa) have been generated with an offset of two and five amino acids covering the complete a/TIF32 monomer structure. The individual MR searches resulted in several well scoring solutions (TFZ scores between 9 to 11) which, when displayed simultaneously, formed an ensemble of overlapping fragments building a fragment of the second a/TIF32 molecule covering the amino acids range from 102 to 205. In order to localize the missing N-terminal fragment of the second a/TIF32 molecule the search was repeated, this time using the refined a/TIF32 monomer and formerly found 103 amino acid long fragment of the second a/TIF32 molecule as the fixed partial model. Well scoring solutions could only be identified after increasing the value of RMSD from 0.35 Å (used for the first runs) to 0.5 Å. The necessity of increasing RMSD implicated higher level of positional disorder of the missing N-terminal part of the second a/TIF32 molecule in comparison to already localized 103 residues long fragment of it. Solutions with the highest TFZ scores (8 to 11) formed an ensemble comprising residues 4 to 65 of a/TIF32 monomer. Based on these results, two a/TIF32 fragments comprising residues 4 to 65 and 102 to 205, respectively have been subjected to additional molecular replacement search and were successful only when using RMSD of 0.9 Å and 0.6 Å for fragment one and two, respectively. The presence of the second a/TIF32 molecule was in addition confirmed by calculating the self-rotation function using GLRF (14) program giving one clear solution at 10 sigma level. The self-rotation axis corresponds to the rotation between two a/TIF32 monomers present in the asymmetric unit (the difference in kappa angle is 7 degrees). Refinement of the structure, comprising complete a/TIF32 monomer and two additional a/TIF32 fragments -residues 8 to 64 and 104 to 205, resulted in decrease of R and R free factors to 24.85 % and 29.56 %, respectively. Due to high level of disorder of the second a/TIF32 molecule (average B-factor of 165 Å 2 ) reference model restrains generated from the complete a/TIF32 molecule, as implemented in PHENIX, have been used during the refinement. As a consequence no conformational differences can be observed between the two a/TIF32 molecules occupying the asymmetric unit of which only the complete a/TIF32 molecule has been refined independently.

RNA synthesis
A template for DAD4 RNA synthesis was prepared by PCR amplification from yeast genomic DNA using the following primers: GAAATTAATACGACTCACTATAAGCAGATAGGGAGGAAAAGAAGTGAGTTTA and ATGCGTATATAGAAAATTGGTGAATTAAA(T) 20 . In the forward primer the sequence of the T7-promoter was included. A stretch of 20 T was added to the 5' end of the reverse primer to mimic the presence of a poly-A in the resulting RNA. The PCR product was precipitated with ethanol. To remove potential RNAse contaminations, Proteinase K was added and subsequently denatured by heating to 95°C. 1 -1.5 mg DNA template was employed in an in vitro transcription approach containing T7 Polymerase and each 40 mM rNTPs in 1× HT buffer (30 mM HEPES pH 8.0, 25 mM MgCl 2 , 10 mM DTT, 2 mM spermidine, 0.01% Triton X-100). After incubation at 37 °C for 3 h, the transcript was ethanol precipitated. The resulting pellet was dissolved in RNase-free water.
Other biochemical methods β-galactosidase assays were conducted as described previously (15). Polysome profile analysis, 2% HCHO cross-linking, WCE preparation and fractionation of extracts for analysis of pre-initiation complexes were carried out as described by (16).
Analysis of the 48S PICs was done as described by(1)with the following exceptions. Total RNA was isolated from 0.5 ml of gradient fractions by hot-phenol extraction, and resuspended in 26 µl of diethyl pyrocarbonate (DEPC)-treated H 2 O. Isolated total RNA was treated with 0.7 µl of DNaseI(NEB)in the total volume of 30 µl. 3 µl of RNA were subjected to reverse transcription with SuperScript III reverse transcriptase (Invitrogen) in the total volume of 20 µl. Aliquots of cDNA were diluted 3-fold or 12-fold for measuring mRNA or 18S rRNA levels, respectively (this way the cDNA was diluted 20 or 80-fold in total, respectively, in comparison to non-diluted RNA). qPCR amplifications were performed on 2 µl of diluted cDNA in 10-µl reaction mixtures prepared with the Brilliant II SYBR green qPCR Master Mix (Stratagene) and primers for RPL41A mRNA (0.3 µM), DAD4 mRNA (0.3 µM), SME1 mRNA (0.3 µM) or 18S rRNA (0.4 µM) using the Mx3000P system (Stratagene). For each round of qPCR, each fraction was measured in triplicates together with no-RT control. The experiment with each strain was performed at least three times for RPL41A mRNA and two times for DAD4 and SME1 mRNAs with similar results.

TABLES, FIGURES AND FIGURE LEGENDS
Supplementary Figure S1.Solubility test of different fragments of recombinant a/TIF32. Cell lysate was clarified by centrifugation, resulting in the soluble protein (S) and insoluble protein in the pellet (P). In each case, the total cell lysate prior to centrifugation is loaded on the gel (T). The best soluble fragment was a/TIF32 276-494 which is indicated by a black arrow.
Supplementary Figure S2. Multiple sequence alignment of eIF3a/TIF32 from different organisms. Sequence alignment was done using ClustalW (17). Espript (18) was used for graphical presentation of the results. The crystallized fragment is marked by a box; Box37 is indicated by a grey bar. Green bars and orange arrows represent helices and strands, respectively.
Supplementary Figure S3. The tif32-Box37 substitution eliminates association of three reporter mRNAs with 43S PICs in vivo. (A -C) The isogenic rpl11b∆ strains carrying either wt or mutant a/TIF32 were heat shocked at 36˚C for 4 hours and processed for mRNA binding analysis as described in Figure 6. The amounts of 18S rRNA and (A) RPL41A, (B) DAD4, and (C) SME1 mRNAs were measured by realtime quantitative PCR (qPCR). (D) The relative amounts + SDs of all three mRNAs in the tif32-Box37 mutant versus wt in 18S rRNA containing fractions were calculated.