Paralogous translation factors target distinct mRNAs to differentially regulate tolerance to oxidative stress in yeast

Abstract Translation initiation factor 4G (eIF4G) is an integral component of the eIF4F complex which is key to translation initiation for most eukaryotic mRNAs. Many eIF4G isoforms have been described in diverse eukaryotic organisms but we currently have a poor understanding of their functional roles and whether they regulate translation in an mRNA specific manner. The yeast Saccharomyces cerevisiae expresses two eIF4G isoforms, eIF4G1 and eIF4G2, that have previously been considered as functionally redundant with any phenotypic differences arising due to alteration in eIF4G expression levels. Using homogenic strains that express eIF4G1 or eIF4G2 as the sole eIF4G isoforms at comparable expression levels to total eIF4G, we show that eIF4G1 is specifically required to mediate the translational response to oxidative stress. eIF4G1 binds the mRNA cap and remains associated with actively translating ribosomes during oxidative stress conditions and we use quantitative proteomics to show that eIF4G1 promotes oxidative stress-specific proteome changes. eIF4G1, but not eIF4G2, binds the Slf1 LARP protein which appears to mediate the eIF4G1-dependent translational response to oxidative stress. We show similar isoform specific roles for eIF4G in human cells suggesting convergent evolution of multiple eIF4G isoforms offers significant advantages especially where translation must continue under stress conditions.


INTRODUCTION
Transla tion initia tion is the process by which a ribosome is recruited to a messenger RN A (mRN A) transcript to begin protein synthesis.For initiation to occur, the mRNA must first be bound by a series of initiation factors which pr epar e the transcript and promote recruitment of the ribosome.These stages are crucial, often rate-limiting steps in translation and are subject to e xtensi v e regulation, particularly in eukaryotes ( 1 ).Translational regulation in response to stress conditions is especially important as cells must be able to make the specific proteins r equir ed to survi v e environmental challenges, whilst not wasting energy making any unnecessary proteins ( 2 ).The control of protein synthesis in response to external stimuli is ther efor e thought to play an important role in the overall regulation of gene expression in eukaryotic cells and some estimates have highlighted a dominant role for translational controls over transcriptional controls ( 3 , 4 ).
For most eukaryotic translation initiation e v ents, the mRNA to be translated must be selected in a process where specific factors interact with the mRN A 5 ca p or 3 pol y(A) tail ( 5 ).The e ukaryotic i nitiation f actor (eIF)4E interacts with the cap structur e ( 6 ), wher eas the poly(A) binding protein (Pab) binds to the poly(A) tail ( 7 ).eIF4G acts as a large scaffold protein which is capable of interacting with both eIF4E and Pab ( 8 ).eIF4G also interacts with RNA helicases such as eIF4A, which act to unwind the transcript allowing recruitment of the 43S pre-initiation complex (PIC) and subsequent scanning of the 5 UTR for a start codon ( 9 ).The 43S PIC contains the small ribosomal subunit bound by the initiation factors eIF1, eIF1A, eIF3, and the eIF2 Met-tRN A i met GTP ternary complex (TC) ( 1 ).Criticall y, eIF4G mak es k ey contacts with eIF3 (in mammals) or eIF5 / eIF1 (in yeast) to recruit the 43S PIC to the mRNA (10)(11)(12).The 43S complex then scans the mRNA in a 5 to 3 direction until it identifies an AUG start codon in an appropriate sequence context.Start codon recognition by the TC triggers GTP hydrolysis on eIF2 followed by dissociation of eIF2 •GDP.Ejection of the majority of the eIFs after start codon recognition and recruitment of the large ribosomal subunit (60S) has been thought to complete the initiation process ( 1 , 5 ).It is generally unclear when and where during scanning and start codon recognition the interactions among eIFs, mRNA, and ribosomal subunits are established and broken ( 13 ).For instance, following joining of the 60S subunit, it appears that some eIFs may remain transiently associated with ribosomes through the early elongation and termination phases ( 14 ).Both eIF4G and eIF4A are specifically enriched on ribosomes engaged in the translation of short, reinitiation-permissi v e uORFs ( 15 ).
Throughout eukaryotes the mRNA selection process forms a focal point for the regulation of translation initiation.Specific proteins are capable of interacting with the core translation initiation factors to modify their activity.For instance, eIF4E binding proteins (4E-BPs) competiti v ely r epr ess eIF4G binding to eIF4E ( 16 ).These proteins target mRNAs either globally or more specifically to reduce the le v el of translation initiation, often in response to cellular signalling pathways ( 17 ).In addition, many of the mRNA selection factors are present as multiple isoforms, which via subtle sequence differences allow fine-tuning of the mRNA selection process.Extreme examples are eIF4E in Drosophila , where eight isoforms of the eIF4E gene are pr esent and ar e thought to dir ect tissue and de v elopmental specificity to mRNA translation ( 18 ).There are also eight isoforms of PAB in Arabidopsis ( 19 ) and at least ten eIF4A genes have been identified in Tobacco ( 20 ).The roles of many of these different proteins are often quite poorly understood.
For the key scaffold molecule eIF4G, there are at least two isoforms in many eukaryotes, with evidence of differential roles for eIF4G domain-containing proteins in humans, flies , and plants , amongst others (21)(22)(23).Humans have three eIF4G genes, eIF4G1, eIF4G2 (DAP5) and eIF4G3.eIF4G1 and eIF4G3 are the main eIF4G isoforms, with most homology to the yeast homologues ( 24 , 25 ), whereas DAP5 is only similar to the C-terminal part of the other proteins, and lacks the eIF4E binding domain ( 26 ).In Sacchar om y ces cer evisiae ther e ar e two isoforms of eIF4G encoded by TIF4631 and TIF4632 , r eferr ed to as eIF4G1 and eIF4G2 respecti v ely.Similar to human eIF4G, yeast eIF4G is a large protein w hich directl y interacts with Pab, eIF4E, and eIF4A (27)(28)(29).It also contains three RNA binding domains which are important for growth and translation ( 30 ).Sequence alignment between the two eIF4Gs in yeast shows conservation mostly towards the C-terminus and in the protein binding domains, with variability towards the N-terminal end and the RNA binding domains (Figure 1 A).Although the yeast isoforms are similar, they are only ∼50% identical, leaving ample variability that might contribute to isoform-specific functions.
The presence of multiple eIF4G isoforms in di v erse eukaryotes is suggesti v e that these isof orms ma y have distinct functions that have been independently selected for at various stages of ev olution.Ho wever, there is little functional evidence for isoform specificity in regulating translation for the different eIF4G isoforms in Saccharomyces cerevisiae .The two S. cerevisiae eIF4G genes ar ose fr om a whole genome duplication e v ent (WGD), w hich largel y defines the post-WGD clade of yeast species ( 31 , 32 ).After this ancestral genome duplication e v ent, one copy of the duplicated genes was lost for most genes, but some duplicated genes wer e pr eserv ed.S .cerevisiae has maintained 551 duplicated genes.Intriguingly, many of these duplicated genes are involved in the process of protein synthesis and the two genes have diverged very little.
Yeast deletion mutants lacking individual eIF4G isoforms are viable, although loss of eIF4G1 has been reported to cause slo w gro wth and impair translation initiation ( 33 , 34 ).These phenotypes are thought to deri v e from the lowered le v els of total eIF4G, since eIF4G1 is more highly expressed than eIF4G2.Later studies have used homogenic yeast strains to avoid the problems associated with the slow growth of eIF4G1 mutants.These are strains which express a single eIF4G isoform at similar le v els to the total amount of eIF4G present in a wild-type strain and no differences in growth kinetics or translational activity were found in these strains expressing single eIF4G isoforms ( 34 ).These results support little or no isoform specificity between the two eIF4G genes.
In this current study we hav e re-e xamined this pre vailing model, considering whether the yeast eIF4G isoforms ar e differ entially r equir ed across a range of conditions, focussing specifically on tolerance to oxidati v e stress.In contrast to previous data, we show that eIF4G1, but not eIF4G2, promotes tolerance to oxidati v e stress, highlighting significant differences in the way these isoforms work to regulate gene expression via translational control.

Growth and stress conditions
Yeast strains were grown at 30 • C with shaking at 180 rpm in SCD complete media (0.67% w / v yeast nitrogen base without amino acids, 2% w / v glucose) supplemented with appropriate amino acids and bases (ForMedium, UK).For oxidati v e stress e xperiments, H 2 O 2 was added to e xponential phase cells (OD 600 0.6-1.0) to a final concentration of 0.4 mM.Stress sensitivity was determined by growing cells to exponential phase in SCD media and spotting serially diluted cultures ( A 600 = 1.0, 0.1, 0.01, 0.001) onto SCD agar plates containing various concentrations of hydrogen peroxide or sulpholmeturon methyl (SMM).For growth in the absence of glucose, SCGE contained 3% w / v glycerol and 1% v / v ethanol instead of glucose.
Two A 260 units of lysate were layered onto 15-50% sucrose gradients pr epar ed in 12 ml thin-walled open polyallomer tubes (Seton Scientific) and separated by ultracentrifugation in an SW41 Ti rotor (2.5 h at 278 000 × g , 4 • C).
Profiles were generated by continuous A 254 recording using a UA-6 UV / Vis detector and chart recorder (Teledyne ISCO).Profile images were analysed using ImageJ (version 1.52q) ( 39 ).Fractions (14 fractions, 0.8-1 ml) were collected manually for western blotting starting from the top of the gr adient.Protein was extr acted fr om sucr ose gr adient fr actions by precipitation in 20% trichloroacetic acid (TCA) overnight at -20 • C. Precipitated protein was pelleted by centrifugation (15 min at 20 000 × g , 4 • C) and washed twice with ice-cold acetone.

Label-free mass spectrometry analysis
Cell extracts were prepared from four biological replicates essentially as described in ( 43 ).Protein samples were prepared for MS by adding equal volumes of sample and protein loading buffer (2 × NuPAGE LDS sample buffer [Invitro gen], 715 mM 2-merca ptoethanol) and incubating for 5 min at 95 • C. Samples were briefly run on NuPAGE 4-12% Bis-Tris gels (Invitrogen) and then excised from gels.Samples were dehydrated using acetonitrile and centrifuged under vacuum.Dried gel pieces were reduced with 10 mM DTT and alkylated with 55 mM iodoacetamide, then twice washed alternately with 25 mM ammonium bicarbonate and acetonitrile.Gel pieces were dried by vacuum centrifugation and samples digested using trypsin overnight at 37 • C.
Liquid chromato gra phy was carried out using an Ul-tiMate 3000 Rapid Separation Binary System (Thermo Fisher Scientific).Peptides were concentrated using an AC-QUITY UPLC M-Class Symmetry C18 Trap Column (180 m inner diameter, 20 mm length [Waters]) and then separated using an ACQUITY UPLC M-Class Peptide BEH C18 Column (75 m inner diameter, 250 mm length, 1.7 m particle size [Waters]).A gradient starting with 99% Buffer A (0.1% formic acid in water) and 1% Buffer B (0.1% formic acid in acetonitrile) and increasing to 75% Buffer A and 25% Buffer B was used to separate the peptides over 45 min at a flow rate of 200 nl / min.Label-free tandem MS was performed using an Orbitrap Elite Hybrid Ion Trap-Orbitrap Mass Spectrometer (Thermo Fisher Scientific).Peptides were selected for fragmentation and MS2 analysis automatically by data-dependent analysis.
Raw MS data were processed using MaxQuant version 1.6.17.0 ( 44 ).A peptide mass tolerance of 20 ppm was used for the first search, 4.5 ppm for the main search, and 0.5 Da for the MS / MS fragment ions.The 'match between runs' feature was used.The peak list was searched against the Uniprot Sacchar om y ces cer evisiae da tabase (accessed 10 February 2017) using the built-in Andromeda search engine ( 45 ).Data from the 'pr oteinGr oups' file produced by MaxQuant were imported into R (RStudio 2021.09.2 + 382).One biological replicate from the '4G2h Unstressed' condition was identified as an outlier and omitted from the analysis.Proteins identified as potential contaminants, re v erse proteins, and only identified by site were filtered out.Proteins without a gene name were filtered out.Proteins with only one razor and unique peptide were filtered out.Normalisation was performed using the limma package ( 46 ).The normalizeQuantiles function was used to normalise the LFQ intensities by the mean of each quantile yielding the same empirical distribution for each sample.All protein intensities ('LFQ.intensity'columns) were log 2transformed and filtered by > 2 v alid v alues in a condition using Perseus (version 1.6.15.0) ( 47 ).The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE ( 48 ) partner repository with the dataset identifier PXD041682.

Bioinformatic analyses
GO analysis was performed using the clusterProfiler package (version 3.18.1)( 49 ).The compareCluster function was used to compare GO term enrichments between proteins significantly different in pairwise comparisons (parameters: organism database = S. cerevisiae (org.Sc.sgd.db),P -value cutoff = 0.05, function = 'enrichGO').Following overrepresentation analysis, the simplify function was used to remo ve o verla pping GO terms.GOSlim anal ysis using PAN-THER (Version 17.0) was used to generate simplified GO term enrichment categories between the same lists ( 50 , 51 ).

eIF4G1 is r equir ed f or o xidative str ess toler ance
The presence of two di v erged eIF4G genes in the yeast Sacchar om y ces cer evisiae provides a fle xib le system for studying the role of par alogous tr anslation factor genes.Previous work has suggested that the underlying rationale for maintaining two eIF4G genes relates to the dosage of eIF4G.Such a hypothesis is difficult to link with the high degree of di v ergence in protein sequence between eIF4G1 and eIF4G2, and the fact that over 85% of genes have been lost after the genome duplication e v ent in S .cerevisiae ( 52 ).Previous work on the sub-functionalization of duplicated genes in yeast has posited tha t dif ferential functionality is often associated with stress response pathways ( 53 , 54 ).Therefore, to explore whether there are any functional differences in the r equir ement for the duplicated eIF4G genes, the stress sensitivity of different eIF4G mutants was examined.
Str ess conditions wer e selected that ar e known to cause a rapid a ttenua tion of protein synthesis including oxida ti v e stress, amino acid starvation and glucose depletion (55)(56)(57).Strains were grown to exponential phase and serially diluted before spotting onto plates containing hydrogen peroxide to cause oxidati v e stress, sulpholmeturon methyl (SMM) to cause an amino acid starv ation (isoleucine / v aline starv ation) or plates containing glycerol / ethanol instead of glucose (SCGE).We initially used deletion strains where either TIF4631 encoding eIF4G1 (denoted 4G1 ) or TIF4632 encoding eIF4G2 (denoted 4G2 ) was deleted.The 4G1 strain was strongly sensiti v e to oxidati v e stress and amino starvation compared with the wild-type and 4G2 strains (Figure 1 B).No differences in sensitivity to growth on SCGE were observed.
Gi v en pre vious reports that phenotypes observed in the 4G1 strain might arise due to lowering the eIF4G dose rather than through any isoform-specific effects, it was important to measure the cellular concentrations of eIF4G1 and eIF4G2 in the deletion strains ( 34 ).Using isoformspecific antibodies, we found that a similar concentration of eIF4G1 was present in both the 4G2 and wild-type strains (Figure 1 C).In contrast, the cellular concentration of eIF4G2 increased a pproximatel y two-fold in the 4G1 strain presumably due to the previously reported compensatory transcriptional increase in TIF4632 expression observed in 4G1 strains ( 34 ).Thus, even though eIF4G2 concentrations are significantly increased in the 4G1 strain, it is not sufficient to promote oxidati v e stress tolerance, suggesting tha t dif fer ences in str ess tolerance may arise due to functional differences between the two proteins.
We next examined the stress sensitivity of homogenic strains which are strains where the coding region of eIF4G1 is replaced with the coding region of eIF4G2 in one strain (4G2h), and in another strain the coding region of eIF4G2 is replaced with eIF4G1 (4G1h) ( 34 ).The homogenic eIF4G1 (4G1h) strain expressed eIF4G1 at comparable levels to eIF4G1 present in the wild-type strain (Figure 1 C).The cellular concentrations of eIF4G2 were again significantly increased in the 4G2h strain w hich onl y contains eIF4G2 and lacks any eIF4G1.Analysis of stress sensitivity for the 4G1h and 4G2h strains re v ealed that neither strain was affected in growth on SMM or SCGE relati v e to the wild type-strain, whereas the 4G2h homogenic strain was sensiti v e to oxidati v e stress (Figure 1 B).It is unclear why the 4G1 strain is sensiti v e to SMM, whereas the 4G2h strain is resistant.It may be related to the different amino acid auxotrophies present in the deletion (BY4741 -met15 ) and homogenic (BY4742 -Lys2 ) strains since SMM is an inhibitor of acetolactate synthase (ALS), a branched-chain amino acid biosynthetic enzyme ( 58 ).Howe v er, the consistent sensitivity of both the 4G1 and 4G2h strains to hydr ogen per oxide indica tes tha t eIF4G1 is r equir ed for oxidant tolerance, and the rest of this study has focused on the role of eIF4G in oxidati v e stress tolerance.
Comparison of growth rates for the various strains indica ted tha t the 4G1 strain is slow growing, whereas the homogenic strains grow at wild-type rates (Figure 1 D).The fact that the 4G1 strain is slow growing compared with the 4G2h strain presumably reflects a threshold affect in the r equir ement for eIF4G2 cellular concentrations to support the growth of a strain lacking eIF4G1.Subsequent studies have ther efor e focussed on the homogenic eIF4G strains to avoid any complications that might arise due to differences in growth rate.

eIF4G1 is r equir ed to r emodel the proteome in response to o xidative str ess
We reasoned that the differential requirement for eIF4G1 to promote tolerance to oxidati v e stress might arise due to dif ferences in transla tion and pr otein pr oduction dri v en by the presence of the different eIF4G isoforms.To examine any oxidati v e stress induced proteome changes, whole cell extracts were prepared and proteome changes analysed using a label-free quantitati v e mass spectrometry (LC-MS) approach comparing total cell extract proteomes ( 59 , 60 ).For these experiments, the wild-type and homogenic strains were grown to exponential phase befor e tr eatment with 0.4 mM H 2 O 2 for 1 h.
The differences in protein abundance between the homogenic strains and oxidati v e stress conditions were first assessed using volcano plots where the log 2 (fold change) and -log 10 ( P value) for different pairwise comparisons were plotted, and proteins that were significantly different across the comparison highlighted (Figure 2 ).Following one hour of oxidati v e str ess tr eatment, ther e wer e significant changes to all the strains' proteomes.For the wild-type and 4G1h strains there was a ske w towar ds higher protein abundance under oxidati v e str ess conditions (Figur e 2 A, B).Howe v er, for the 4G2h strain higher protein le v els were seen pr e-str ess and fewer proteins increased in le v el after oxidati v e stress conditions (Figure 2 C).These differences are also highlighted by the lists of significantly upregulated proteins for each strain following oxidati v e str ess, wher e the wild-type strain had 51 proteins which were significantly more abundant after stress, the 4G1h strain had 118, but the 4G2h strain only had 22.This indicates that the 4G2h response to stress may be more muted in comparison to the wild-type or 4G1h strains, potentially explaining the observed differences in sensitivity to H 2 O 2 .The deficiencies in the 4G2h oxidati v e str ess r esponse wer e further emphasised by the gene ontolo gy (GO) anal ysis.For instance, enrichment of the biological process category 'Cellular response to oxidati v e stress' and the molecular function category 'Oxidoreductase activity' were only found in the wild-type and 4G1h strain samples after H 2 O 2 treatment, but not in the 4G2h H 2 O 2 condition (Figure 2 A-C).
To further compare the oxidati v e stress proteomes of the homogenic strains, each of the strains' proteomes following oxidati v e str ess wer e subject to pairwise comparisons.Strikingl y, w hile the wild-type and 4G1h strains showed very similar proteomes follo wing o xidati v e stress, the 4G2h strain had significantly lower abundances for many proteins when compared to both the wild-type and 4G1h strains (133 and 127 respecti v ely) (Figure 2 D-F).Conversely, there were only nine proteins with a significantly higher abundance in the 4G2h strain follo wing o xidati v e stress compared with the wild-type strain follo wing o xidati v e str ess (Figur e 2 F).These data illustrate that while the WT and 4G1h strains both express proteins to similar levels of abundance, a substantial number of proteins ar e pr esent at lower le v els in the 4G2h strain under the same condition.Gene ontology analysis again re v ealed an enrichment for the molecular function category 'Oxidoreductase activity' in the wild-type and 4G1h hydrogen peroxide proteomes compared with the 4G2h hydrogen peroxide proteome (Figure 2 D-F).

The translational response to oxidative stress is different in strains containing eIF4G1 or eIF4G2 alone
Since the presence of different eIF4G isoforms caused alterations in the proteome and eIF4G is an essential translation factor ( 33 ), we next addressed whether eIF4G1 or eIF4G2 affect the translational response to oxidati v e stress.Translational activity was analysed by examining the sedimentation of ribosomes within a sucrose density gradient during non-stress conditions and in response to hydrogen peroxide.During unstressed growth conditions, similar polysome profiles were observed in the wild-type, 4G1h and 4G2h strains (Figure 3 A), similar to previous studies ( 34 ).Most ribosomes are in the polysome fractions and so are acti v ely engaged in translating mRNA.Ther efor e, the translational activity, defined as the polysome to monosome (P / M) ratio was high and comparable in all three strains (Figure 3 B).
It is well known that stress conditions such as oxidati v e stress cause a marked inhibition of translation initiation promoted by Gcn2-dependent phosphorylation of eIF2 ␣ ( 56 ).On polysome gradients, such an inhibition of transla tion initia tion is observed as an increase in the 80S monosome peak along with reduced le v els of acti v ely translating polysomes.This is because initiation is the rate limiting step in translation and continued elongation causes a characteristic 'run-off' of polysomes resulting in the increased 80S peak ( 55 , 61 ).Accordingly, in keeping with previous studies ( 56 ), we found that in the wild-type strain after a 15 minutes exposure to 0.4 mM H 2 O 2 significant polysome run-off was evident (Figure 3 A).This result is characteristic of an inhibition of translation initiation and can be measured from the decreased polysome to monosome (P / M) ratio (Figure 3 B).A very similar scale of polysome run-off and decrease in the P / M ratio was also observed in the 4G1h strain.Interestingly howe v er, the P / M ratio observ ed in the 4G2h strain in response to hydrogen peroxide treatment was significantly higher than for the wild-type ( P < 0.01) and 4G1h (* P < 0.05) strains (Figure 3 A and B).This result suggests that the inhibition of translation initiation caused by oxidati v e stress may not be so pronounced for the 4G2h strain and the mechanism underlying this finding is further explor ed in Figur e 4 .We further verified this result by examining translational activity in the 4G1 and 4G2 strains exposed to increasing concentrations of hydrogen peroxide.A comparable decrease in the P / M ratio was observed in response to hydrogen peroxide treatment in the wild-type and 4G2 strains (Supplementary Figure S1).In contrast, a less dramatic decrease in the P / M ratio was observed in the 4G1 strain supporting the idea that eIF4G1 is r equir ed for the translational inhibition caused by oxidati v e stress (Supplementary Figure S1).
Exposure to oxidati v e stress is known to inhibit transla tion initia tion to cause polysome run-of f via the Gcn2dependent phosphorylation of eIF2 ␣ ( 56 ).Gi v en the difference in P / M ratios observed between the 4G1h and 4G2h strains after exposure to oxidative stress, we examined whether there are any differences in hydr ogen per oxideinduced phosphorylation of eIF2 ␣.This analysis re v ealed similar increases in eIF2 ␣ phosphorylation in response to hydr ogen per oxide treatment in the wild-type, 4G1h and 4G2h strains (Figure 3 C).This indicates that the difference in 80S monosome accumulation observed in the 4G2h strain does not arise due to any differences in Gcn2dependent phosphorylation of eIF2 ␣.

The 4G2h strain is defective in translation initiation following o xidative str ess
Oxidati v e stress is known to inhibit protein synthesis at multiple le v els including inhibition a t the initia tion and elongation phases of translation ( 56 ).The decreased 80S monosome peak observed in response to oxidati v e stress in the 4G2h strain compared with the 4G1h strain might therefore arise due to a more se v ere elongation b lock pre v enting polysome run-off and the accumulation of free 80S ribosomes, or alternati v ely, it may be due to an initiation block pre v enting 80S ribosomes from being assembled on mR-NAs.The ribosomes that are detected in the 80S monosome peak on standard sucrose density gradients r epr esent both mRNA-fr ee, and ther efor e translationally inacti v e 80S ribosomes, or 80S ribosomes that are bound to mRNAs.The 80S ribosomes that are bound to mRNAs are more stable than free 80S ribosomes and high magnesium concentrations can be used in sucrose density gradients to differentially dissociate free 80S ribosomes into their constituent 40S and 60S ribosomal subunits (Figure 4 A) ( 62 ).
As anticipated, we observed dissociation of the 80S monosome peaks into their constituent 40S and 60S ribosomal subunits for all strains analysed using high magnesium sucrose density gradients (Figure 4 B).Quantification of the 80S monosome peaks remaining in the wild-type, 4G1h and 4G2h strains indicated that similar proportions of mRNA bound 80S ribosomes are present in all three strains during non-stress conditions (Figure 4 C).In contrast, similar quantitation of mRNA bound 80S ribosomes after hydrogen peroxide treatment re v ealed a difference for the 4G2h str ain.In contr ast to the WT and 4G1h str ains where the proportion of the 80S ribosomal peak made up of mRNA bound ribosomes increases after stress, for the 4G2h strain there was no increase in the proportion of mRNA-bound versus free 80S ribosomes after oxidati v e str ess (Figur e 4 C).This difference may explain the reduced height of the 80S monosome peak in the 4G2h strain after stress (Figure 3 A  and B).This suggests that despite eIF2 being phosphorylated to a similar extent in response to oxidati v e stress, fe wer ne w initiation e v ents occur in the 4G2h strain during oxidati v e stress conditions.

eIF4G1 remains associated with actively translating polysomes following oxidative stress
The above data suggest that in the presence of eIF4G1, transla tion initia tion persists a t some small le v el to produce mRNA associated 80S ribosomes under oxidati v e stress conditions.To explore whether eIF4G1 remains associated with translated mRNAs after stress, the distribution of both eIF4G1 and eIF4G2 in sucrose density gradients was examined.Analysis of the wild-type strain grown under non-stressed conditions re v ealed that eIF4G1 and eIF4G2 show a similar pattern of distribution with enrichment in the 40S and 80S fractions along with association with ribosomes in the heavy acti v el y translating pol ysome fractions (Figure 5 A).The association of eIF4G1 across a polysome gradient was only subtly altered following oxidati v e stress conditions.In contrast, a significant proportion of eIF4G2 was shifted from the heavy polysome fractions to the 80S monosome fraction following exposure to hydrogen peroxide (Figure 5 A).This means that a lower proportion of eIF4G2 is associated with polysomes compared with eIF4G1 during oxidati v e stress conditions.The continued association of eIF4G1 with acti v ely translating ribosomes may explain the role of eIF4G1 in promoting translation during oxidati v e stress conditions.

No differences in eIF4E association or mRNA-cap binding are detected for eIF4G1 or eIF4G2 following oxidative stress conditions
One possibility for the transla tion tha t persists after oxidati v e stress, is that eIF4G1 is part of a mechanism that favours the selection of str ess r esponsi v e mRNAs for translation.A major function of eIF4Gs in translation is that together with eIF4E they interact with the mRN A ca p.So a key question is whether the eIF4E binding and cap interaction are maintained for eIF4G1 and eIF4G2 after stress.We used two different approaches to examine these interactions.We first compared the interaction between eIF4E and eIF4G during normal and oxidati v e stress conditions.For this experiment we used an eIF4E-TAP strain and performed TAP affinity purifications and Western blot-ting to examine the interaction between eIF4E and eIF4G1 or eIF4G2.As expected, eIF4E immunoprecipitated a fraction of both eIF4G1 and eIF4G2 (Figure 5 B).Howe v er, no differ ences wer e observed in the interactions between eIF4E and eIF4G1 or eIF4G2 during normal or oxidati v e stress conditions.
As a second approach, we used Cap affinity chromatography to assess the ability of eIF4G to bind the mRN A ca p structure with eIF4E.Both eIF4G1 and eIF4G2 were found (UB) and bound (IP) fractions were analysed by western blotting and membranes probed with antibodies for eIF4E, eIF4G1 and eIF4G2.Quantification is shown for the normalised ratio of eIF4G: eIF4E from triplicate experiments.No significant differences were found using one-way ANOVA.( C ) A whole cell extracts were prepared from the wild-type strain either unstressed or treated with 0.4 mM H 2 O 2 for 15 minutes and incubated with m 7 GTP-sepharose.Total (T), unbound (UB), and bound (B) fractions were analysed by Western blotting.Membranes were probed with antibodies for eIF4G1, eIF4G2 and eIF4E.Quantification is shown for bound factions as a percentage of totals.No significant differences were found using one-way ANOVA.
to interact with the mRN A ca p although the le v els bound across these experiments were quite variable.This analysis re v ealed that the proportion of eIF4G1, eIF4G2 or eIF4E bound to the cap structure did not change following oxidati v e str ess compar ed to the unstr essed condition (Figur e 5 C).Differences in eIF4E or mRN A ca p binding do not ther efor e appear to explain the different requirements for eIF4G1 or eIF4G2 during oxidati v e stress conditions.The Slf1 la-related protein (LARP) is required for the eIF4G1-mediated response to oxidative stress Taken together, our data so far suggest that the continued association of eIF4G1 with acti v ely translating mR-NAs may mediate the translational regulation of mRNAs that ar e r equir ed for str ess adaptation.This raises the question as to how these mRNAs are translated during oxidati v e stress conditions that cause a global inhibition of translation via increased eIF2 ␣ phosphorylation.A large number of RNA-binding proteins (RBPs) are known to interact with translation factors and / or the ribosome to either promote or inhibit mRNA translation during different growth and stress conditions ( 63 ).One key example is the yeast La related protein Slf1, which is a key activator of translation during oxidati v e stress conditions and has also been shown to remain associated with polysomes during ROS exposure ( 64 ).Additionally, similar to the increased P:M ratio observed in the 4G2h strain in response to oxidati v e stress (Figure 3 A), the P:M ratio is also increased in an slf1 mutant compared with a wild-type strain following exposure to the same concentration of hydr ogen per oxide ( 64 ).We ther efor e tested the interaction between eIF4G and Slf1, and whether Slf1 is r equir ed to promote eIF4G1-dependent oxidant tolerance.
Previous studies found that Slf1 immunoprecipitated a fraction of key closed-loop proteins consistent with the idea that Slf1 remains associated with acti v ely translating mR-NAs during oxidati v e stress conditions ( 64 ).This included eIF4G1, but the interaction with eIF4G2 was not examined.We ther efor e r epeated this immunopr ecipitation using Slf1-TAP and the resulting western blot was probed using antibodies specific for eIF4G1 and eIF4G2.Slf1 was found to precipitate a fraction of eIF4G1 during normal and oxidati v e stress conditions as expected (Figure 6 A).In contrast we could not detect any interaction between Slf1 and eIF4G2.
The previous study found that RNase I treatment diminished the co-immunoprecipitation of Slf1 and eIFG1 which was taken as evidence that this interaction might be mRNA mediated ( 64 ).We first confirmed that the interaction between Slf1 and eIF4G1 is decreased by treatment with RNase I (Figure 6 A).We reasoned that eIF4G1 and Slf1 maybe in close proximity on mRNAs and so we used formaldehyde cross-linking to stabilise protein and RNA interactions.This increased the amount of eIF4G1 that was immunoprecipitated by Slf1, but again we could not detect any precipitation of eIF4G2 (Figure 6 A).Formaldehyde cross-linking was also found to pre v ent the decreased eIF4G1-Slf1 interaction observed in response to RNase I tr eatment (Figur e 6 A).Taken together these da ta indica te that not only do Slf1 and eIF4G1 appear to interact in an RNA-dependent manner, but they must do so in close proximity such that their RNA-dependent interaction can be stabilized by cross-linking.
Having established a physical interaction between eIF4G1 and Slf1, we ne xt e xamined combined eIF4G and slf1 mutant strains in terms of sensitivity to oxidati v e stress.A strain lacking SLF1 was found to show similar sensitivity to oxidati v e str ess compar ed with the 4G2h strain (Figur e 6 B).Loss of Slf1 resulted in oxidati v e stress sensiti vity in the 4G1h strain indicating that Slf1 is r equir ed to mediate eIF4G1-dependent oxidant tolerance.Surprisingly, loss of Slf1 was found to rescue the oxidati v e stress sensiti vity of the 4G2h strain.The reasons for this rescue are unclear but may indicate that Slf1 plays an antagonistic role in preventing eIF4G2-mediated translation of mRNAs r equir ed during oxidati v e stress conditions.
As a final test of the overlapping r equir ement for eIF4G1 and Slf1 in the translational response to oxidati v e stress, we compared Slf1 mRNA targets with our proteomic analysis in homogenic strains exposed to hydrogen peroxide.RN A-imm une precipitation sequencing (RIP-Seq) has previously been used to identify mRNAs bound by Slf1 ( 64 ).We mapped these targets onto the volcano plot comparing the 4G1h and 4G2h proteomes exposed to oxidative str ess (Figur e 6 C, purple highlighted cir cles).This analysis re v ealed that most of the Slf1 target mRNAs encode proteins that are increased in abundance in the 4G1h strain compared with the 4G2h strain.

Isof orm specific r equir ements f or eIF4G during o xidative stress conditions in human cells
The two eIF4G genes in S. cerevisiae almost certainly arose as a result of a genome duplication e v ent in the Sacchar om y ces lineage ( 65 ), and these genes have likely been maintained to provide specific functional properties.Multiple eIF4G isoforms are also prevalent across other eukaryotic organisms, either by virtue of multiple genes or alternati v e splicing.To assess whether differences in eIF4G function ar e r ela ted to oxida ti v e stress tolerance in other eukaryotic systems, eIF4G1 and eIF4G3 expression was specifically decreased in HeLa cells using siRNAs.qRT-PCR was used to confirm the degree of specificity to the knockdowns.eIF4G3 siRN A specificall y lowered eIF4G3 mRN A levels without affecting eIF4G1 (Figure 7 A).Whilst eIF4G1 siRN A treatment did strongl y reduce eIF4G1 mRN A levels, it also decreased eIF4G3 expression by a pproximatel y 40%.Immunofluor escence microscop y was performed using eIF4G1 or eIF4G3 antibodies to examine eIF4G protein expression.The analysis confirmed that eIF4G1 and eIF4G3 siRNAs depleted cellular protein concentrations of eIF4G1 and eIF4G3, respecti v ely (Figure 7 B).
Treatment of HeLa cells with hydrogen peroxide causes a dose-dependent decrease in cell viability (Figure 7 C).This effect on cell viability was exacerbated following eIF4G1 or eIF4G3 knockdown.Intriguingly, at the very highest hydrogen peroxide concentration tested, knockdown of eIF4G3 resulted in strong sensitivity to oxidative stress compared with the eIF4G1 knockdown suggesting isoform-specific requirements for the human eIF4Gs in promoting oxidati v e stress tolerance.We confirmed that the depletion of eIF4G did not affect the basal or induced le v el of eIF2 ␣ phosphorylation in response to peroxide treatment of human cells suggesting that global translation is similarly regulated (Figure 7 D).We used a puromycin incorporation assay to determine how the knockdowns and changes in eIF2 ␣ phosphorylation le v els compare with translational inhibition (Figure 7 E).Interestingly, we found tha t transla tion is inhibited follo wing knockdo wn of eIF4G3 which may suggest that eIF4G3 ma y pla y a mor e pr edominant role in translation initiation during unstressed conditions than eIF4G1, similar to the more dominant role for yeast eIF4G1 compared with yeast eIF4G2.Following oxidati v e stress, translation was similarly inhibited under control conditions or follow-ing eIF4G1 or eIF4G3 knockdown (Figure 7 E).Overall, these data suggest that in an analogous manner to the roles of yeast eIF4G1 and eIF4G2 in facilitating tolerance to oxidati v e stress, the human eIF4G genes might similarly differ in their capacity to specify the translational activity of specific mRNAs r equir ed for adaptation to stress.

DISCUSSION
Man y eIF4G isof orms have been described both at the le v el of gene duplication and alternati v e splicing in di v erse eukaryotic organisms.For instance, Xenopus, Arabidopsis, Drosophila and humans all have two or more eIF4G genes, and in extreme scenarios some plants such as Soya bean contain up to eight eIF4G and eIFiso4G genes ( 66 ).This evolutionary r eoccurr ence of multiple distinct forms of eIF4G raises the possibility that they may have isoformspecific functions.Additionally, differences in splicing isoforms, e xpression le v els, and tissue specific expression further contribute to the idea that different eIF4G isoforms provide an important regulatory step in translation ( 67 , 68 ).In this current study we found that the two isoforms of fulllength eIF4G present in yeast and human cells are differentially r equir ed to media te tolerance to oxida ti v e stress.Understanding the functional specificity of eIF4G isoforms is important since potential disease-causing variants of eIF4G have been linked with Parkinson's disease and defects in eIF4G are associated with cancer and ineffecti v e cancer treatments (69)(70)(71).
Yeast strains lacking eIF4G1 and human cells with reduced expression of eIF4G3 were found to be sensiti v e to hydr ogen per oxide suggesting isoform-specificity in the translational response to oxidati v e stress conditions.In response to external stimuli such as ROS, global translation initiation is normally reduced, matched with complementary mechanisms that allow certain mRNAs to escape global r epr ession ( 2 ).One key mechanism that inhibits transla tion initia tion in eukaryotic cells is through phosphorylation of the alpha subunit of translation factor eIF2 ( 72 ).We found that eIF2 ␣ was similarly phosphorylated in response to hydrogen peroxide in y east str ains containing eIF4G1 or eIF4G2 as the sole eIF4G present or in human cells with reduced expression of eIF4G1 or eIF4G3.This indica tes tha t the regula tion of transla tion initia tion via eIF2 ␣ phosphoryla tion is unaf fected by the presence of different eIF4G isoforms.It is unclear how the mRN As w hich show increased translational activity during oxidati v e stress conditions are resistant to the translation inhibition mediated via eIF2 ␣ phosphorylation and reduced ternary complex le v els.Our da ta indica te a conv ergent e volutionaril y ad vantage to eIF4G isoforms in promoting oxidati v e stress tolerance, suggesting that eIF4G contributes to the continued translation of key mRNAs during conditions that globally inhibit translation initiation.
Pre vious studies hav e shown that the proteome is significantly altered following exposure to ROS, independent of any changes in transcription ( 56 , 73 ).While the majority of protein synthesis is inhibited follo wing o xidati v e stress, it is vital for the survival of the cell that the proteins involved in the response to oxidati v e stress continue to be made ( 56 , 73 ).The mechanism by which specific mRNAs are translated in response to stress is not yet fully elucidated but our quantitati v e proteomics da ta indica te tha t yeast eIF4G1 accounts for some of this oxidati v e stress dependent translation.Polysome profiling performed under high magnesium concentrations showed that mRNA-bound ribosomes incr ease compar ed with fr ee ribosomes in eIF4G1, but not eIF4G2 containing strains, in response to oxidati v e stress.This altered pattern of free and mRNA-bound ribosomes in the homogenic strains was also mirrored by the patterns of eIF4G1 and eIF4G2 association across polysome gradients.eIF4G2 shifts from the heavily translated region on polysome profiles to be more associated with inacti v e ribosomes following oxidati v e stress conditions.Conversely, eIF4G1 remained associated with heavy polysomes during oxidati v e stress suggesting that new initiation and contin-ued translation specifically r equir es eF4G1 function during oxidati v e stress conditions.
Further analysis aimed at re v ealing possib le mechanistic reasons for the differences in eIF4G1 and eIF4G2 function using immunoprecipitation of the eIF4E cap binding protein did not re v eal any alterations in its association with eIF4G1 or eIF4G2 follo wing o xidati v e stress conditions.Similarl y, ca p affinity chromato gra phy showed that different eIF4F complexes associated with eIF4G1 or eIF4G2 do not appear to be recruited to mRNAs following oxidati v e stress to promote translation in an isoform-specific manner.One possibility to explain the different functions of eIF4G1 and eIF4G2 is that they interact with mRNAs via different RNA binding proteins.One key difference observed in eIF4G isoform protein interactions was the specific interaction of eIF4G1 with Slf1.Slf1 is a member of the conserved La motif containing RNA-binding (LARP) proteins, and has been shown to have an important role in the oxidati v e stress response by binding to mRNAs specifically encoding oxidati v e str ess r esponse proteins ( 64 ).Slf1 r emains associated with ribosomes following oxidati v e stress conditions and has ther efor e been proposed to mediate the translational response to oxidati v e stress via mRNA-specific translational control ( 64 ).We found that eIF4G1, but not eIF4G2, interacts with Slf1 in both an RNA-dependent and independent manner, suggesting that yeast eIF4G isoforms have different protein binding partners.This is perhaps not surprising, since although eIF4G1 and eIF4G2 are similar proteins, they also have large regions with low homology including their N-terminal regions which are not very well conserved ( 30 ).
Combined with the finding that eIF4G1 can facilitate the oxidati v e str ess r esponse, one possibility is that Slf1 plays a role in mediating the translation of eIF4G1-bound mR-NAs.RBPs such as LARPs have increasingly been recognized as playing key roles in providing specificity to specialized ribosomes that play regulatory functions in the selecti v e translation of subsets of mRNAs ( 74 ).Slf1 is an atypical LARP that has a conserved La-motif (LaM) RNA binding domain but lacks any RN A reco gnition motif (RRM) typical of the LARP family ( 75 ).Slf1 has been shown to bind small ribosomal subunits via a novel binding motif within the N-terminal region of Slf1 ( 64 ).Slf1 may ther efor e act as an adapter protein between specific oxidati v e stress related mRNAs via its LaM domain and ribosomes via its ribosome binding domain.Our data suggests these mRNAs are only accessible to eIF4G1 rather than eIF4G2.It is also interesting that the simultaneous loss of both eF4G1 and Slf1 promotes oxidant tolerance suggesting that eIF4G2 is capable of driving oxidative stress-specific translation under certain conditions.One possibility is that Slf1 acts to preclude eIF4G2 mRNA binding during oxidati v e stress conditions.More studies will be r equir ed to determine the functional r equir ement for eIF4G2 and whether it is similarly r equir ed for mRNA-specific translation during particular growth or stress conditions.

DA T A A V AILABILITY
Proteomics data are available via ProteomeXchange with identifier PXD041682.

Figure 1 .
Figure 1.eIF4G1 is r equir ed for oxidati v e stress tolerance.( A ) Schematic of protein sequence alignment of eIF4G1 and eIF4G2 where yellow sections (scored 3) are conserved residues and purple sections (scored 0) are not similar.Alignment and similarity scores from ClustalOmega ( 76 ).( B ) The wildtype (BY4741) and isogenic eIF4G deletion strains (4G1 , 4G2 ) or wild-type (BY4742) and isogenic homogenic strains (4G1h, 4G2h) were grown to exponential phase and the A 600 adjusted to 1, 0.1, 0.01 or 0.001 before spotting onto the indicated plates.This included glucose (SCD), glycerol / ethanol, H 2 O 2 or sulpholmeturon methyl (SMM) plates.( C ) Western blot analysis of the same strains probed with antibodies that recognise eIF4G1 or eIF4G2 normalized to Zwf1 (glucose 6-phophate dehydrogenase) as a loading control.Data shown are the means of 3 independent biological repeat experiments ± SD.Significance is shown using a one-way ANOVA test.( D ) Doubling times for the same strains grown on SCD media.Data shown are the means of three independent biological repeat experiments ± SD.Significance is shown using a one-way ANOVA test.** P < 0.01, *** P < 0.001, **** P < 0.0001.

Figure 2 .
Figure 2. eIF4G1 is r equir ed to r emodel the proteome in response to oxidati v e str ess.( A-C ) Volcano plots ar e shown comparing the values for log 2 (fold change) and -log 10 ( p v alue) v alues for proteins identified in the wild-type and homogenic strains in the presence or absence of oxidati v e stress conditions.( D-F ) P airwise comparisons ar e shown for the indicated strains in the presence of hydr ogen per oxide.Pr oteins w hich are significantl y increased ( P < 0.05) are coloured blue and proteins which are significantly decreased are coloured yellow ( P < 0.05) and are listed in Supplementary Table S1.GOSlim enrichments are shown for significantly different proteins in each comparison.

Figure 3 .
Figure 3.The translational response to oxidati v e stress is different in strains containing eIF4G1 or eIF4G2 alone.( A ) Representati v e polyribosome traces are shown for the wild-type and homogenic strains before or after treatment with 0.4 mM H 2 O 2 for 15 min.The peaks that contain the small ribosomal subunit (40S), the large ribosomal subunit (60S), and both subunits (80S) are indicated along with the polysome peaks generated by 2, 3, 4, 5, etc., 80S ribosomes on a single mRNA.( B ) Quantification of polysome:monosome ratios is shown from triplicate traces determined by the ratio between the area under the monosome to the polysome peaks.Error bars denote standard deviation and significance is shown comparing unstressed or stressed ratios using a one way ANOVA test.( C ) Western blot analysis of eIF2 ␣ (Sui2) and eIF2 ␣-P using the same strains and conditions as for panels A and B. Quantification is shown from triplicate experiments comparing phosphorylation in the presence or absence of hydr ogen per oxide for each strain using one-way ANOVA.* P < 0.05, ** P < 0.01.

Figure 4 .
Figure 4.The proportion of mRNA-bound versus free 80S ribosomes increases in eIF4G1 containing strains but not in eIF4G2-containing strains strain follo wing o xidati v e stress conditions.( A ) Schematic showing the effects of high magnesium concentrations in polysome buffers on either free or mRNA bound 80S ribosomes.Free 80S subunits dissociate to their 40S and 60S constituent parts whereas mRNA bound 80S subunits are stabilised by their mRNA interaction and remain as 80S monosomes.( B ) Representati v e polyribosome traces are shown in the presence or absence of high magnesium for the wild-type and homogenic strains before or after treatment with 0.4 mM H 2 O 2 for 15 minutes.Green diamonds (40S), red ovals (60S) and blue triangles (80S) denote ribosomes and subunits.( C ) Quantification of the area of the 40S and 60S peaks (free) and the 80S and polysomes (bound) as a proportion of the total trace, where the free fraction represents ribosomal subunits separated by magnesium and the bound fraction r epr esents ribosomes stabilised by mRNA.Significance is shown using a one-way ANOVA test for the bound fraction; * P < 0.05, *** P < 0.001.

Figure 5 .
Figure 5. eIF4G1 is associated with acti v ely translating mRNAs following oxidati v e stress conditions.( A ) Polyribosome profiles are shown for wild-type strain before or after hydrogen peroxide treatment for 15 min .Sucrose density gradients were fractionated and proteins visualised by Western blotting using antibodies specific for eIF4G1 or eIF4G2.The intensities of eIF4G1 and eIF4G2 bands were quantified and are shown as a percentage of the total cumulati v e intensity across each polyribosome trace.Quantification is shown from triplicate experiments (pairwise t -test, ** P < 0.001) ( B ) eIF4E-TAP was immunoprecipitated from an exponential phase whole cell extract either unstressed or treated with 0.4 mM H 2 O 2 for 15 min.The total (T), unbound (UB) and bound (IP) fractions were analysed by western blotting and membranes probed with antibodies for eIF4E, eIF4G1 and eIF4G2.Quantification is shown for the normalised ratio of eIF4G: eIF4E from triplicate experiments.No significant differences were found using one-way ANOVA.( C ) A whole cell extracts were prepared from the wild-type strain either unstressed or treated with 0.4 mM H 2 O 2 for 15 minutes and incubated with m 7 GTP-sepharose.Total (T), unbound (UB), and bound (B) fractions were analysed by Western blotting.Membranes were probed with antibodies for eIF4G1, eIF4G2 and eIF4E.Quantification is shown for bound factions as a percentage of totals.No significant differences were found using one-way ANOVA.

Figure 6 .
Figure 6.Slf1 is r equir ed for the eIF4G1-mediated response to oxidati v e stress.( A ) Slf1-TAP was imm unoprecipitated from an exponential phase w hole cell extract either unstressed or treated with 0.4 mM H 2 O 2 for 15 minutes , formaldehyde , or RNase I.The bound (IP) fractions were analysed by Western blotting and membranes probed with antibodies for Slf1, eIF4G1 and eIF4G2.( B ) The wild-type (BY4742), homogenic strains (4G1h, 4G2h), and strains deleted for SLF1 ( slf1 , 4G1h slf1 , 4G2h slf1 ) were grown to exponential phase and the A 600 adjusted to 1, 0.1, 0.01 or 0.001 before spotting onto the indica ted pla tes.( C ) Volcano plot of the log 2 fold change and -log 10 P value of the pairwise comparisons between the homogenic strains in oxidati v e stress conditions with Slf1 mRNA targets highlighted in purple.Box and whisker plots show the average fold changes for all proteins and Slf1 mRNA targets.

Figure 7 .
Figure 7. Isoform specific r equir ements for eIF4G in human cells during oxidati v e stress conditions.( A ) qRT-PCR analyses are shown to confirm the degree of specificity for eIF4G1 and eIF4G3 knockdowns.eIF4G3 siRN A specificall y lowers eIF4G3 mRNA le v els without affecting eIF4G1, whereas, eIF4G1 siRN A treatment strongl y reduces eIF4G1 mRN A le v els and decreases eIF4G3 e xpression by a pproximatel y 40%.Statistical anal ysis was carried out using one-way ANOVA.( B ) The indicated HeLa cells were left untreated or treated with 1 mM hydrogen peroxide for one hour.eIF4G1 and eIF4G3 were visualized by immunofluorescence staining using specific antibodies (green).Mitochondria were visualized using MitoTracker Red and merged images are shown.( C ) HeLa cells were treated with the indicated concentrations of hydrogen peroxide and viability measured using crystal violet.Statistical analysis was carried out using one-way ANOVA to compare the viability of the control and knock-down cells.( D ) HeLa cells were left untreated or treated with 1 mM hydr ogen per oxide for one hour.Protein extracts were immunoblotted for phosphorylated eIF2 ␣ (p-eIF2 ␣) or eIF2 ␣.Quantification is shown from triplicate experiments comparing the fold change (FC) increase in phosphorylation in response to hydrogen peroxide exposure.( E ) Protein synthesis was analysed using puromycin labelling under the same conditions as shown in panel D. The incorporation of puromycin into newly synthesised protein was assessed by immunoblotting.Quantification is shown from triplicate experiments normalised against the intensity of Coomassie blue staining presented as fold change (FC) in puromycin labelling.