Translation regulation of specific mRNAs by RPS26 C-terminal RNA-binding tail integrates energy metabolism and AMPK-mTOR signaling

Abstract

Increasing evidence suggests that ribosome composition and modifications contribute to translation control. Whether direct mRNA binding by ribosomal proteins regulates the translation of specific mRNA and contributes to ribosome specialization has been poorly investigated. Here, we used CRISPR–Cas9 to mutate the RPS26 C-terminus (RPS26dC) predicted to bind AUG upstream nucleotides at the exit channel. RPS26 binding to positions −10 to −16 of short 5′ untranslated region (5′UTR) mRNAs exerts positive and negative effects on translation directed by Kozak and Translation Initiator of Short 5′UTR (TISU), respectively. Consistent with that, shortening the 5′UTR from 16 to 10 nt diminished Kozak and enhanced TISU-driven translation. As TISU is resistant and Kozak is sensitive to energy stress, we examined stress responses and found that the RPS26dC mutation confers resistance to glucose starvation and mTOR inhibition. Furthermore, the basal mTOR activity is reduced while AMP-activated protein kinase is activated in RPS26dC cells, mirroring energy-deprived wild-type (WT) cells. Likewise, the translatome of RPS26dC cells is correlated to glucose-starved WT cells. Our findings uncover the central roles of RPS26 C-terminal RNA binding in energy metabolism, in the translation of mRNAs bearing specific features and in the translation tolerance of TISU genes to energy stress.

INTRODUCTION

Translation initiation is a highly complex and regulated process consisting of ribosomes, regulatory proteins (eIFs) and mRNA (1,2). Protein synthesis efficiency of different mRNAs is highly variable and modulated by the availability of certain eIFs, the presence of cis-elements in the 5′ untranslated region (5′UTR) and the context of the AUG start codon (3,4). In addition, increasing evidence from budding yeast to humans suggests that ribosome heterogenicity via varying composition can potentially affect the translation efficiency (TE) of specific mRNAs (5–12). Most of the studied examples refer to the haploinsufficiency of ribosomal proteins (RPs) causing several genetic diseases, such as Diamond–Blackfan anemia, congenital asplenia and 5q− syndrome (13–16). These diseases exhibit a range of tissue-specific symptoms rather than general effects, suggesting gene-specific defects.

Selective mRNA binding by the RPs within the ribosome is another possible way for ribosome specialization and regulation of translation. There are few cases implicating the ribosome in binding to specific mRNA motifs. For instance, a few IRES elements bind directly to ribosomes (17). We reported previously the sequence-specific UV cross-linking of RPs to the Translation Initiator of Short 5′UTR (TISU) (18). Cross-linking of other ribosomal subunits to AUG flanking sequences was also reported (19). Thus far, there is little genetic evidence for the importance of specific ribosome–mRNA contacts for selective translational control.

Identification of RNA elements recognized by RPs is challenging. One possibility is to isolate ribosome–mRNA complexes from cells and identify the specific contact sites of the ribosome with the mRNA. However, since the ribosome scans all translated mRNAs, it would be difficult to identify specific recognition motifs except for ribosome pausing sites. Another possibility is to use the information of cryogenic electron microscopy (cryo-EM) studies of ribosomes bound to mRNA and to focus on the RPs situated close to the mRNA path within the ribosomal subunit. For example, RPS3, RPS10 and RPS5 contact AUG downstream nucleotides in the entry channel, while RPS26 and RPS28 are associated with AUG upstream nucleotides in the exit channel. These contacts likely stabilize the PIC, as was recently shown for the yeast RPS3 (20). Other studies in yeast also revealed that RPS26 is important for the TE of Kozak mRNAs (12). However, since in the current ribosomal structures, RPS26 RNA contacts are upstream from the critical −3 position of Kozak and are unlikely to bind it directly, the molecular basis of this effect is unclear. As most of the RP contact sites in the mRNA are beyond the AUG context, little is known about how these RNA bindings are linked to specific mRNA features and how they contribute to translational control, in particular in mammalian systems.

In this study, we used genome editing to mutate the amino acid residues of RPS26 located proximal to the −10 position and further upstream (RPS26dC). Our experiments led to the discovery of novel ribosome-specific functions and selective translational control of mRNAs bearing defined features. We found that the RPS26 C-terminus is required to facilitate cap-proximal initiation of the Kozak element by binding to −11 to −16 nt, while no requirement was seen with a Kozak initiator preceded by a longer 94 nt 5′UTR. Additionally, a sequence preference for pyrimidines in the −11 to −16 nt positions is important for the RPS26 C-terminus function. In contrast, the cap-proximal TISU activity was dramatically enhanced either in RPS26dC cells or upon the removal of the −11 to −16 nt, suggesting that RPS26 C-terminus RNA binding is inhibitory to TISU-mediated translation. The RPS26dC mutation confers resistance to glucose starvation and mTOR inhibition. Furthermore, in RPS26dC cells, the activity of the energy sensor AMP-activated protein kinase (AMPK) is induced while mTOR signaling is inhibited, akin to energy-deprived wild-type (WT) cells. Also, the translatome of RPS26dC cells resembles that of glucose-starved WT cells. Thus, RPS26 C-terminal RNA binding has important roles in the regulation of cellular energy metabolism, the translation of mRNAs bearing specific features and the translation tolerance of TISU genes to energy stress.

MATERIALS AND METHODS

Cell culture and transfections

Human embryonic kidney 293 (HEK293) cells were obtained from ATCC^® (CRL1573™) and HEK293 TAF1 temperature-sensitive (TAF1ts) cells were previously described (21). Cells were grown at 37°C in a humidified incubator with 5% CO₂ in Dulbecco’s modified Eagle’s medium (Gibco, Life Technologies, Thermo Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Gibco), 100 units/ml penicillin and 100 µg/ml streptomycin. The restrictive temperature of TAF1ts cells was 39.5°C. Cells were tested negative for mycoplasma by PCR. Cells were harvested after 4 h of glucose starvation [glucose-free media (11966025, Gibco) with 10% dialyzed FBS] for polysome profiling, 3 h for compound C treatment (171260, Sigma) or 24 h after transfection for cell viability measurement using a Cell Titer-Glo luminescent assay (G7571, Promega) after the various stresses [glucose starvation, thapsigargin (328570010, Thermo Scientific™) and Torin-1 (HY-13003, MCE)].

DNA transfections were done by the calcium phosphate method (2× HEPES buffered saline, 554Bf), by polyethylenimine 25K (Polysciences) prepared at 1 mg/ml. Lipofectamine 2000 (11668027, Invitrogen™) was used for RNA transfection. For the GFP reporter assays, sub-confluent cells in 12-well plates were co-transfected with 25 ng of the specified GFP plasmid together with 75 ng of firefly luciferase that serves as a normalizing control. Cells were harvested 24 h after transfection, luminescence activity was determined and then normalized amounts were loaded onto SDS–PAGE followed by western blot analysis using an anti-GFP antibody. For the RNA transfection, cells in 96-well plates were transfected with 50 ng of the Renilla luciferase mRNA and harvested after 7 h.

Generation of RPS26dC cells

HEK293 TAF1ts cells (21) were used to generate cells with the desired mutation. To improve the recovery of the desired RPS26 mutants, we used a co-editing selection strategy (22), where the HEK293 TAF1ts cells were transfected with the Cas9/sgRNA plasmids and ssODN donors for the repair of the ts mutation and for the generation of the RPS26 mutations. The population of cells that successfully repaired the TAF1ts mutation, restoring the WT sequence, was enriched for the desired mutation in RPS26. sgRNA was designed by the DESKGEN cloud tool. The selection of TAF1ts transfected cells was done at a temperature of 39.5°C. Colonies of cells surviving the selection, each arising from a single cell, were picked into 96-well plates and grown for further analysis. The area from the gDNA of the transfected clones containing the desirable mutations was amplified by PCR from the cell lines’ genomic DNA and sequenced by Sanger sequencing. Guide RNA and ssODN sequences and other primers used for PCR are listed in Supplementary Table S1.

Western blot and antibodies

Lysates were subjected to SDS–PAGE followed by western blot using the following antibodies: anti-RPS26 (HPA043961, Atlas Antibodies), anti-GAPDH (MAB374, EMD Millipore), anti-RPS3 (ab140688, Abcam), anti-RPL17 (sc-515904, Santa Cruz Biotechnology), anti-GFP (ab1218, Abcam), anti-HA (ab9110, Abcam), anti-4EBP (9644, Cell Signaling), anti-p-4EBP (sc-293124, Santa Cruz Biotechnology), anti-AMPK (2532, Cell Signaling), anti-p-AMPK (2535, Cell Signaling), anti-eIF2α (Antibody Verify), anti-p-eIF2α (Santa Cruz Biotechnology), anti-activating transcription factor 4 (anti-ATF4; Proteintech, 10835-1-AP), anti-puromycin (Millipore) and anti-tubulin (Sigma). Western blot quantification was done in the Li-Cor 2800 Odyssey FC instrument using the Image Studio™ Lite software.

Plasmids

The SpCas9/sgRNA expression plasmids were based on pX330-U6-Chimeric_BB-CBh-hSpCas9 as previously described (21). The guide RNA sequences were inserted into SpCas9/sgRNA expression plasmids using BsaI (R3733, NEB).

The pEGFP-N1 (Clontech) and pEGFP-N1 with 16 or 10 bp 5′UTR with TISU or sub-optimal Kozak context were constructed using the T-PCR method (23). The NUG firefly luciferase and Renilla luciferase plasmids were kindly provided by Kastura Asano (Kansas State University) and were previously described (24). HA-RPS26 WT was constructed by restriction-free cloning (25) and primers to pCruz HA™ (sc-5045, Santa Cruz Biotechnology). HA-RPS26 was used as a backbone for HA-RPS26 mutants. All the constructs were verified by Sanger sequencing.

RNA constructs were constructed using DNA templates bearing T7 promoter. The reporter RNAs were produced by PCR amplification (Kappa HiFi HotStart, Roche) of plasmids encoding for Renilla reporter using the primers listed in Supplementary Table S1. One primer was used to add a poly(A) tail of 32 nt and the second introduced both the T7 promoter and the indicated sequence manipulations. PCR products were sequenced by Sanger sequencing and cleaned using ethanol precipitation. PCR products were used for in vitro transcription reactions and DNase treatment with the RiboMAX kit (Promega) according to the manufacturer’s instructions. The transcribed RNA was cleaned by phenol–chloroform. RNA was capped using the Vaccinia capping kit (NEB) and recovered using the Direct-zol RNA miniprep kit (Zymo Research).

Analysis of global translation and mass spectrometry

Untreated and 4 h glucose-starved WT and RPS26dC were incubated with 100 μg/ml cycloheximide (CHX, Sigma) for 5 min and then washed twice with cold buffer containing 20 mM Tris (pH 8), 140 mM KCl, 5 mM MgCl₂ and 100 μg/ml CHX. The cells were collected and lyzed with 500 μl of the same buffer containing 0.5% Triton, 0.5% DOC, 1.5 mM DTT, 150 units of RNase inhibitor (Eurx) and 5 μl of protease inhibitor (Sigma). The lyzed samples were centrifuged at 12 000 × g at 4°C for 5 min. The cleared lysates were loaded onto a 10–50% sucrose gradient and centrifuged at 38 000 rpm in an SW41 rotor for 105 min at 4°C. Gradients were fractionated and the optical density at 254 nm was continuously recorded using the ISCO absorbance detector UA-6. For global protein synthesis analysis, cells were harvested after treatment with puromycin (p8833, Sigma) for 5 min and subjected to western blot with an anti-puromycin antibody.

For mass spectrometry (MS), the free and 80S fractions of WT cells were isolated and purified by trichloroacetic acid precipitation, resolved on SDS–PAGE, and the area corresponding to RPS26 was excised and subjected to MS analysis.

For global translation analysis, the collected samples were merged to create the fractions: light polysome (2–5 ribosomes) and heavy polysome (6+ ribosomes). The RNA of these fractions and the input samples were isolated using TRIzol and Direct-zol RNA miniprep kits (Zymo Research). Equivalent RNA concentrations were taken from each fraction for the library preparation for MARS-Seq (26,27).

Deep sequencing, data mapping and analysis

The RNA sequencing (RNA-seq) of the polysome profile was done on the Illumina machine in the INCPN sequencing unit at the Weizmann Institute. Polysome profile read quality control was tested using Fastqc (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Polysome profile reads were aligned to the human reference genome (hg38) using Bowtie2 (--local) (28). The number of reads per gene was counted from the uniquely aligned reads in GENCODE (GRCh38.103) genomic regions (29) using featureCounts (30). Normalization between samples was performed using the DESeq2 package in the R statistical programming language. Replicate similarities were tested using principal component analysis (PCA) and Pearson correlations and plotted using ggplot2 in R. The normalized gene read count table (DESeq2 output table) was filtered to include only genes in high expression: only genes in each sample with number of reads per gene >32 reads were used for downstream analysis. This ensured that the genes used in the downstream analyses were expressed in all tested samples. This step resulted in 7666 genes that were used for downstream analysis. The normalized counts were used to calculate TE as follows:

where Heavy′(gene i) is the normalized expression in the heavy fraction for the ith gene, Light′(gene i) is the normalized expression in the light fraction for the ith gene and Input′(gene i) is the normalized expression in the input fraction for the ith gene.

Quantification of ATP and ADP

The ATP/ADP ratio was measured in the Metabolomics unit at the Weizmann Institute (courtesy of Alexander Brandis). The LC–MS/MS instrument (LiChrosolv^®), equipped with an electrospray ion source and operated in positive ion mode, was used to analyze nucleotides. Chromatographic separation is done in UPLC HSS T3 column equipped with corresponding pre-column (Waters Corp) using a gradient as described (31).

RESULTS

Generation of RPS26 C-terminal RNA-binding mutant cell line

We envisaged that ribosome–mRNA contacts might be particularly important for initiation from short 5′UTR in which the PIC is relatively unstable (20,32). According to the ribosome–mRNA structure, RPS26 interacts with AUG flanking nucleotides upstream from the critical −3 position (33,34) (Figure 1A). In particular, K82 is close to positions −5 and −6, and the C-terminal amino acids 100–108 wrap the mRNA at nucleotides around position −10 possibly together with an adjacent segment at the C-terminus (aa 109–115), which is not visible in the structures. To examine the potential regulatory function of the RPS26 RNA-binding motif, we generated RPS26 mutants in which K82 was substituted (K82D) and the C-terminal residues 100–115 were deleted (RPS26dC). We performed a pilot experiment to test whether these RPS26 RNA-binding mutants affect the initiation of short 5′UTR mRNA. These mutants were transiently transfected into HEK293T cells with a GFP reporter gene in which the initiating AUG is in a sub-optimal Kozak context and is preceded by a short 5′UTR length of 16 nt (Supplementary Figure S1A and E). This AUG is in frame with the downstream AUG of the GFP. The translation that begins from the upstream AUG generates an ∼30 kDa protein (US AUG), whereas initiation from the downstream GFP AUG produces an ∼27 kDa protein (DS AUG). Firefly luciferase with a 5′UTR of 111 served as an internal normalizing control. The 16 nt long 5′UTR drives inefficient translation from the cap-proximal AUG as ∼50% starts from the downstream AUG due to leaky scanning (Supplementary Figure S1B, first lane). Expression of K82D and RPS26dC mutants but not the WT caused a further enhancement of leaky scanning (Supplementary Figure S1B), even though the expression level of exogenous RPS26 in these experiments was insufficient for detection by western blot. Encouraged by these results, we set out to generate cells in which the endogenous RPS26 is mutated using the CRISPR–Cas9 system with the mutation shown in Supplementary Figure S1C. While we could not obtain a homozygous K82D clone, we did get two clones bearing deletions in RPS26 C-terminus. One clone (RPS26dC1) is homozygous to the deletion of 100–115 RNA-contacting residues (Supplementary Figure S1D), while the second clone (RPS26dC2) is a complex mutant with one allele bearing the desired 100–115 deletion and the second allele has a smaller deletion that starts from 101 that includes an RNA-binding arginine residue, and four additional in-frame new residues (Supplementary Figure S1E). We, therefore, focused most of the subsequent experiments on the homozygous mutant clone. Western blot with RPS26 and GAPDH antibodies confirmed equivalent expression levels of RPS26 in the WT and RPS26dC mutant cells (Figure 1B).

Figure 1.

Generation and characterization of RPS26 C-terminal deletion mutant clones. (A) Human RPS26 [Protein Data Bank (PDB) ID: 4V6X] was structurally aligned to RPS26 from yeast 48S PIC (PDB ID: 3J81) using PyMol (shown in cartoon format). Human RPS26 is colored in turquoise, yeast RPS26 in green, mRNA backbone in orange and side chains are colored in green with blue and red. The human RPS26 C-terminus is shown in pink, R82 from yeast in red and K82 from human in purple. mRNA positions −10, −6, −5 and −4 are highlighted in black, dark gray, gray and light gray, respectively. (B) Western blot analysis of WT and RPS26dC cells using anti-RPS26 and anti-GAPDH antibodies. The graph shows the RPS26 protein levels normalized to GAPDH (N = 3). (C) WT (red) and RPS26dC (blue) cells were seeded and cell proliferation was measured using a luminescent cell viability assay at the indicated time points. Luminescence values were normalized to the values of day 1 (N ≥ 4). (D) Polysome profiling of WT and RPS26dC cells. Cell lysates of WT (red) or RPS26dC (blue) were subjected to sucrose gradient sedimentation followed by a fraction collection to obtain polysome profiles. Representative profiles are shown. The ratios of polysome to monosome (P/M) and 40S to 60S of three independent repeats are indicated. The asterisk denotes a statistically significant difference (P < 0.05). The free, 40S, 60S and 80S fractions were subjected to western blot analysis using anti-RPS26 (E), anti-RPS3 (F) and anti-RPL17 (G). Percentages of association were quantified (N = 3). Representative immunoblots are shown. Asterisks denote statistically significant differences (*P < 0.05, ***P < 0.005).

To examine the impact of this mutation on cell growth, we plated an equal number of WT and mutant cells and followed their growth for 4 days. The results revealed that RPS26dC cells grew slower and reached lower density than the WT cells (Figure 1C). We subjected WT and RPS26dC cells to polysome profiling and found a small but significant increase in the polysome/monosome ratio in the mutant cells despite their slow growth phenotype (Figure 1D). We confirm that the high-density peaks are polysomes by performing sucrose sedimentation of EDTA-treated cell lysates (Supplementary Figure S2A). Puromycin incorporation assay revealed no significant difference in protein synthesis rate (Supplementary Figure S2B). Together, these findings suggest that the reduced growth may reflect an effect on specific genes and not an overall effect on translation. Analysis of the distribution of RPS26 in the free, 40S, 60S and 80S fractions revealed a trend of reduction of the mutant RPS26 in the 40S, while the relative levels of RPS3 and RPL17, small and large RPs, respectively, remained similar compared to WT cells (Figure 1E–G).

The opposing effects of RPS26dC mutation on cap-proximal initiation of Kozak and TISU are linked to 5′UTR length and sequence

To determine the importance of the C-terminus for translation initiation driven by short 5′UTR, WT and RPS26dC cells were transfected with the 16 nt short 5′UTR GFP reporter (Figure 2A and Supplementary Figure S1E). The RPS26dC displayed an apparent enhancement of leaky scanning, confirming the requirement of the RNA binding of the C-terminus in stabilizing the PIC on short 5′UTR (Figure 2B, left). As the C-terminus interacts with the −10 position and possibly nucleotides upstream (Figure 1A), we generated a GFP reporter with a 5′UTR of 10 nt (Supplementary Figure S1E), which is expected to have fewer contacts with the RPS26 C-terminus. In the WT cells, leaky scanning is significantly enhanced upon shortening the 5′UTR from 16 to 10 nt (Figure 2B, right), suggesting that these nucleotides are important for PIC stability. In addition, with the 10 nt 5′UTR, there is no difference in the extent of leaky scanning between the WT and mutant cells (Figure 2B, graphs), consistent with the requirement of RPS26 RNA binding upstream of the −10 position for stabilization of the PIC. Similar results were obtained with these reporters in the second mutant clone, RPS26dC2 (Supplementary Figure S3A). These findings provide strong evidence that the RPS26 C-terminus is specifically required for binding upstream to position −10 to stabilize the PIC when the AUG is close to the 5′ end.

Figure 2.

RPS26 C-terminal deletion effect on start codon fidelity of different 5′UTR length and AUG contexts. (A) A schematic representation of the GFP reporter gene with an AUG in either a sub-optimal Kozak (GGAAUGU) or a TISU context (CAAGAUGGCGGC), preceded by a long or short 5′UTR and downstream in-frame AUG. US and DS (in red) denote the upstream and downstream translation initiation sites, respectively. Arrow indicates the position of the TSS. WT or RPS26dC cells were co-transfected with a GFP plasmid bearing an AUG in a strong (B) or TISU context (C) preceded by a short 5′UTR or a long 5′UTR and a strong context (D) together with a firefly luciferase as a normalizing control. Cells were harvested 24 h after transfection and normalized amounts were loaded on SDS–PAGE followed by western blot analysis using an anti-GFP antibody. US and DS denote upstream and downstream initiation sites, respectively. Representative immunoblots are shown. Quantified results are shown in the graphs in which the relative intensity of the upstream translation site is presented by orange bars and the downstream translation site by yellow bars. The overall translation of 16 or 94 nt 5′UTR in the WT cells was set to 1. Gray and black asterisks on the graphs denote a statistically significant difference in overall translation or US/DS ratio of translation directed by strong context with short 5′UTR, respectively (N ≥ 3). (E) A schematic representation of the reporter gene with a 5′UTR of 16 nt of original Renilla luciferase or mutated 5′UTR. (F) A total of five different 5′UTR reporters were transcribed and capped in vitro and the resultant mRNAs were then transfected into WT and RPS26dC cells where they underwent in vivo translation. Cells were harvested 24 h after transfection and luciferase levels were measured. Luciferase levels of the original 5′UTR in WT and RPS26dC were set to 1. Red and blue asterisks denote a statistically significant difference between original and mutated 5′UTR luciferase levels in WT and RPS26dC, respectively. Black asterisks denote a statistically significant difference in luciferase levels between WT and RPS26dC cells (N = 5). Asterisks denote statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001).

TISU is a translation initiation context that directs efficient and accurate translation from very short 5′UTR. We transfected the parental and RPS26dC cells with similar 16 and 10 nt long 5′UTR GFP reporters in which the first initiating AUG is in a TISU context (Figure 2C and Supplementary Figure S1E). Remarkably, initiation from the cap-proximal TISU is enhanced in RPS26dC compared to the WT cells, which is the exact opposite of the effect seen with the Kozak context. The shortening of the 5′UTR from 16 to 10 nt has a similar enhancing effect on initiation in the WT cells, while it had a marginal impact in RPS26dC cells. These findings clearly show that the RNA binding of −10 to −16 nt directed by the RPS26 C-terminus is inhibitory for TISU while stimulatory for Kozak. Intriguingly, in the RPS26dC2 clone, the enhancement of TISU is not observed (Supplementary Figure S3A), raising the interesting possibility that the presence of the R100 RNA-binding residue in this clone is critical for the inhibitory effect of TISU but not for the stabilizing effect of the Kozak.

We also analyzed the impact of the RPS26 C-terminus on a longer, 94 nt long 5′UTR and a Kozak initiator (Figure 2D). In this case, no difference was seen between the WT and mutant cells (Figure 2D), further highlighting the specific role of the C-terminus in cap-proximal initiation.

Considering the importance of RPS26 binding to −11 to −16 nt in short 5′UTR mRNAs, we tested whether the identity of these nucleotides is important for the initiation directed by the cap-proximal Kozak. We substituted the five nucleotides in positions −11 to −15 (Figure 2E) with five consecutive A, C, G or T and determined their effect in WT and mutant cells following RNA transfection (validation of the RNA amount and quality is shown in Supplementary Figure S3B). Compared to the original sequence, the A(s), G(s) and T(s) substitutions significantly altered translation in the WT cells, with enhancing effects of the A(s) and T(s) and repressing effect of the G(s). In RPS26dC cells, we found a significant reduction with the C(s) and T(s) but not the A(s) and G(s) at the −11 to −15 positions compared to WT cells, suggesting that the stabilizing effect of RPS26 C-terminus RNA binding involves a sequence preference toward pyrimidines.

To assess the importance of RPS26 C-terminus RNA binding for translation initiation fidelity, we used a series of luciferase reporter genes in which the initiating triplet is either AUG or near-cognate AUG such as CUG, GUG, UUG and ACG, preceded by a 5′UTR length of 111 nt (24). WT and mutant cells were transfected with these reporters, and the activities of AUG, CUG, GUG, UUG and ACG, expressed as a percentage of AUG, were determined. There was no change in the relative utilization of CUG, UUG and ACG, while GUG utilization was enhanced (Supplementary Figure S3C), suggesting a minor role of the RPS26 C-terminus in initiation fidelity.

RPS26 C-terminus mutants confer resistance to energy stress and mTOR inhibition

The translation machinery is rapidly downregulated in response to multiple stresses. In this context, the TISU genes are unique; despite their high dependence on eIF4E and mTOR (32,35–37) they are nevertheless resistant to energy stress in which AMPK is activated and mTOR is inhibited. On the other hand, cap-proximal translation directed by Kozak is sensitive to this stress (37). Considering the resemblance of the effects of RPS26dC mutation and energy stress on the translation of short 5′UTR, we examined whether the RPS26 C-terminus is linked to the energy stress response. We induced a state of energy deprivation in WT, RPS26dC and RPS26dC2 cells by a 24 h glucose starvation, and then assessed cell survival. Strikingly, both RPS26 mutant clones display significant resistance to this severe stress (Figure 3A). When mTOR was inhibited by Torin-1 (Supplementary Figure S4), the RPS26dC mutant clones were also more resistant (Figure 3B). On the other hand, they were either unaffected (RPS26dC2) or slightly more sensitive (RPS26dC) to an ER stress elicited by thapsigargin (Figure 3C and Supplementary Figure S4).

Figure 3.

The effects of stresses on RPS26dC viability and association with the ribosome. (A–C) Cell viability assay after different treatments of WT, RPS26dC and RPS26dC2 cells for 24 h. Cells were treated with 0.9 g/l glucose or without glucose (A), with 25 or 75 nM Torin-1 (B), or with 500 or 1000 nM thapsigargin (C). Luminescence values were normalized to the values of control conditions (N ≥ 3). (D) Polysome profiling of WT cells and RPS26dC cells with and without glucose. Cell lysates of WT and RPS26dC cells with and without glucose for 4 h were subjected to sucrose gradient sedimentation followed by fraction collection to obtain polysome profiles. Representative profiles are shown (WT: red and light gray, respectively; RPS26dC: blue and gray, respectively). (E) The free, 40S, 60S, 80S and two ribosomes fractions of the glucose-starved WT and RPS26dC cells were subjected to western blot analysis using anti-RPS26. Representative immunoblots are shown. Asterisks denote statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001).

Previous studies in yeast cells reported that RPS26 is released from the ribosome under several stress conditions (12,38). To determine whether RPS26 association with the ribosome in human cells is similarly affected by energy stress, we analyzed RPS26 levels in the polysome profile fractions after glucose starvation (Figure 3D and E) in WT and RPS26dC cells. The results revealed similar levels of RPS26 in the 80S and two ribosome fractions in both cells. We also used mass spectroscopy (MS) to check RPS26 in the free and 80S ribosomal fractions prepared by sucrose gradient sedimentation from normal and glucose-deprived cells. The findings revealed a marked elevation of RPS26 in the 80S fraction in the glucose-starved sample (Supplementary Figure S5A and B), consistent with the accumulation of the 80S under these conditions. These findings suggest that in mammalian cells, RPS26 remains associated with the ribosome under this stress.

Interestingly, the MS analysis revealed that the K113 of the RPS26 C-terminus is acetylated (Supplementary Figure S5A). Acetylation of this residue was confirmed in databases of post-translational modifications (39–41). To examine the potential regulatory function of this residue, we mutated it to arginine (K113R) to preserve the positive charge but prevent acetylation or to a negatively charged aspartic acid (K113D). WT, RPS26dC, K113R and K113D were transiently transfected into cells and 24 h later were subjected to energy stress for an additional 24 h, followed by cell survival analysis. As the duration of this experiment is longer than that described in Figure 3A, the cells reach higher confluency and are relatively more protected. Nevertheless, the survival of cells expressing RPS26dC mutant is significantly increased compared to cells expressing WT RPS26 (Supplementary Figure S5C), recapitulating the results with the mutant clones (Figure 3A). Likewise, the K113D but not the K113R mutant promoted a similar extent of increased tolerance to the stress (Supplementary Figure S5C). We confirmed the expression of HA-tagged WT and mutant proteins by western blot (Supplementary Figure S5C). These findings imply that the sensitivity to energy stress depends on the K113 charge, which most likely contributes to RNA binding, and this charge can be modulated by acetylation.

AMPK-dependent inhibition of mTOR in RPS26dC cells is linked to the intracellular energy status

Reduced availability of intracellular energy is known to induce the phosphorylation and activation of the AMPK, a conserved sensor of the cell energy status (42,43). Phospho-AMPK stimulates ATP production pathways and switches off anabolic pathways that consume ATP, such as cap-dependent translation by inhibiting mTOR (44–47). To elucidate the underlying basis of the reduced sensitivity of RPS26dC mutant cells to energy stress and mTOR inhibition, we monitored the phosphorylation status of eIF4EBP (4EBP), a major substrate of mTOR and an inhibitor of eIF4E and cap-dependent translation in its unphosphorylated form (48,49). The level of 4EBP phosphorylation in WT cells under normal growth conditions is relatively high and 4EBP becomes dephosphorylated upon glucose starvation, as expected (Figure 4A). In RPS26dC cells, on the other hand, the 4EBP phosphorylation is substantially lower even under normal growth conditions and is further reduced following energy stress (Figure 4A). Thus, in RPS26dC cells, the basal mTOR activity is limited and resembles cells undergoing stress. Next, we analyzed the total and the phosphorylated active form of AMPK. In WT cells, AMPK phosphorylation is induced upon energy stress (Figure 4B). However, in RPS26dC cells, the basal AMPK phosphorylation is high and is further elevated by energy starvation (Figure 4B). Notably, the total AMPK protein levels are decreased in the mutant cells, consistent with recent studies reporting reduced stability of the activated phosphorylated form of AMPK (50,51).

Figure 4.

AMPK-dependent inhibition of mTOR in RPS26 C-terminus deletion cells. (A, B) WT and RPS26dC cells were treated with 0.9 g/l or without glucose. Cells were harvested after 4 h and subjected to western blot analysis using anti-4EBP and anti-phosphorylated 4EBP (p-4EBP) (A) or anti-AMPK and anti-phosphorylated AMPK (p-AMPK) (B) and anti-tubulin antibodies (N = 3). (C) Cells were treated with or without glucose in the presence or absence of 20 µM compound C. Cells were harvested after 3 h and subjected to western blot analysis using anti-4EBP, anti-p-4EBP and anti-tubulin antibodies. Representative immunoblots are shown. Quantified results are shown in the graphs (N = 3). (D) ATP/ADP ratio per cell of WT (red) and RPS26dC (blue) cells (N = 5). The graphs represent the mean ± SE from three to five independent experiments. Asterisks denote statistically significant differences (*P < 0.05, ****P < 0.001).

To examine whether the reduced mTOR activity seen in RPS26dC cells is linked to the elevated AMPK activity in RPS26dC cells, we treated cells with the AMPK antagonist compound C and followed the 4EBP phosphorylation levels (Figure 4C). As expected, in WT cells, compound C had no significant effect on the basal 4EBP phosphorylation status, but it partially restored the reduced phosphorylation elicited by energy stress, confirming that mTOR inhibition under energy stress is AMPK dependent. In the RPS26dC mutant cells, compound C elevated the low basal activity of 4EBP phosphorylation, indicating that the reduced 4EBP phosphorylation in the mutant cells is, at least in part, mediated by the high basal activity of AMPK.

We also analyzed eIF2α phosphorylation, which is responsive to ER and oxidative stresses. We found higher basal levels of eIF2α phosphorylation in mutant cells, which were unchanged upon energy stress. Interestingly, despite the increase of eIF2α phosphorylation, the levels of ATF4 are not upregulated but in fact are diminished in the mutant cells (Supplementary Figure S6).

To elucidate the underlying basis of the constitutive high level of AMPK phosphorylation and activation in RPS26dC mutant cells, we examined whether it is linked to the energy status of the cells, which directly impacts AMPK phosphorylation. As ATP/ADP ratio is a well-established parameter of cellular energy metabolism, we used metabolomic LC–MS to measure the relative levels of ATP and ADP of an equal number of logarithmically growing WT and RPS26dC mutant cells. The results revealed an ∼3-fold reduction in the ATP/ADP ratio in the mutant relative to WT cells, indicating that RPS26dC cells are under constant energy deprivation.

Translatome analysis of RPS26dC reveals a signature of an activated energy stress

To elucidate further the regulatory role of RPS26 C-terminal RNA binding in the cellular response to energy stress, we performed genome-wide transcription and translation analyses in control and glucose-deprived cells to identify mRNAs whose translation is affected by the C-terminal deletion. For the translatome analysis, we preferred the polysome profile over Ribo-seq since the latter is limited in its ability to quantify the translation state of very short 5′UTR mRNAs. This is due to the isolation of 26–32 nt ribosome-protected fragments that exclude initiation sites preceded by an extremely short 5′ leader that generates shorter protected segments. Following polysome profiling of WT and RPS26dC cells grown under normal and energy-stressed conditions, RNA-seq was carried out on total mRNA, poorly translated mRNA (light, 2–5 polysome fractions) and well-translated mRNA (heavy, ≥6 polysome fractions). After aligning the reads to the human reference genome, >83% (average 85%) of reads were uniquely aligned. In addition, >85% (average rate per sample = 87.1%) of the reads were mapped to exons of protein-coding genes. After sample normalization, the biological replicates were correlated with R > 0.96 in all pairs (average R = 0.98). This similarity is depicted by the Pearson correlation matrix (Supplementary Figure S7A) and also by the PCA of the 1000 top variable genes (Supplementary Figure S7B). The normalized counts were used to calculate TE by calculating the ratio between the reads in the heavy + light polysomal fractions to the input.

Considering the altered response of RPS26dC mutant cells to energy stress, we compared the changes in the TE of WT cells undergoing glucose starvation to the translational differences of RPS26dC relative to WT cells (Figure 5A). We found a strong positive correlation between these two states (R = 0.435, P < 10⁻¹⁶), consistent with the idea that the basal state of RPS26dC cells partially resembles energy-stressed cells. Likewise, we examined the differences in the ratio of the heavy to light polysomes of WT cells undergoing glucose starvation and compared them to the translational changes of RPS26dC relative to WT cells and found a high correlation (R = 0.302, P < 10⁻¹⁶, Supplementary Figure S7C).

Figure 5.

Global translation analysis of RPS26dC cells revealed energy stress signature. (A) Regression analysis between log₂(RPS26dC TE/WT TE) and log₂(WT GS TE/WT TE). R = 0.435, P< 2.2 × 10⁻¹⁶. Boxplots presenting the TE distributions of TISU (B), OXPHOS pathway (C), glycolysis pathway (D), fatty acid oxidation pathway (E) and TOP-containing RP genes (F) in all samples. Asterisks denote statistically significant differences (*P < 0.05, ***P < 0.005, ****P < 0.001).

As the reporter experiments revealed that RPS26 C-terminus is inhibitory for translation initiation directed by the TISU motif with a leader >10 nt, we used these data to examine its impact on the TE of endogenous TISU genes. Our findings confirm an increase in the translation level of TISU genes in RPS26dC cells compared to WT cells under basal conditions (Figure 5B). The extent of this elevation is not as large as seen with the reporter gene, most likely because ∼50% of TISU genes have a 5′UTR length of 12 nt or less, which are expected to be unaffected (see Figure 2C).

As the energy level of RPS26dC cells is lower than that of WT cells, we analyzed the TE of major pathways associated with cellular energy metabolism. Specifically, we analyzed the TE of oxidative phosphorylation (OXPHOS), glycolysis and fatty acid oxidation genes and found perturbations in the translation of a subset of glycolysis and fatty acid oxidation pathways in the RPS26dC compared to the WT cells (Figure 5C–E) that may contribute to the severe reduction in energy availability in the mutant cells. Unexpectedly, the translation of the TOP-containing RPs was not reduced in the mutant cells (Figure 5F), despite their high sensitivity to mTOR inhibition.

It has been reported that the yeast RPS26 promotes translation initiation directed by the Kozak motif (12). We, therefore, wished to examine whether the C-terminus RNA-binding domain is involved. To this end, we determined the frequency of either full (RNNAUGG) or partial (YNNAUGG or RNNAUGH) Kozak in the upregulated or downregulated mRNA in RPS26dC compared to the parental cells. Interestingly, not only that the downregulated mRNAs were not enriched with the Kozak element, we found a small but significant increase in the frequency of the full Kozak in the upregulated mRNAs (Supplementary Figure S7D), suggesting that the RNA binding of the C-terminus is either refractory or inhibitory rather than stimulatory for the Kozak context. This finding is consistent with the reporter experiment in which the activity of a Kozak element preceded by a long 5′UTR was not affected in the mutant cells (Figure 2D).

DISCUSSION

In this study, we focused on the question of how a specific RP contributes to the regulation of translation of specific mRNAs and ribosome specialization in human cells. In the 48S mRNA cryo-EM structure, the most upstream position of the mRNA seen in the exit channel is −10 relative to the AUG, in which RPS26 residues 100–108 wrap the mRNA around the −10 position, and other unstructured C-terminal residues are expected to bind further upstream (33,34). Using a genetic approach to mutate the C-terminal mRNA-binding domain of RPS26 (RPS26dC), we provide compelling evidence for selective translational control of mRNAs bearing specific features. We found that the RPS26 C-terminus is required to facilitate cap-proximal initiation of the Kozak element by binding to −11 to −16 nt and stabilizing the PIC, while no requirement was seen with a longer 94 nt 5′UTR and a Kozak initiator. Furthermore, the identity of the −11 to −16 nt was important, with a clear preference for pyrimidines. In contrast, the binding of RPS26 C-terminus to the same position is inhibitory to the cap-proximal TISU initiator. This is evident from the dramatic rise in TISU activity either in RPS26dC cells or upon removing these nucleotides. Interestingly, downregulation of eIF1 or eIF4G1 also resulted in opposite effects of cap-proximal initiation directed by Kozak and TISU but in the reversed direction compared to RPS26 (36,37), highlighting the differential mechanistic aspects of initiation by these two elements in the context of short 5′UTR. It is, therefore, possible that RPS26 C-terminus binding to −11 to −16 interferes with the activity of eIF4G1 in these elements.

Our results suggest that RPS26dC mutant cells resemble normal cells undergoing energy stress. First, the mutant cells are partially resistant to glucose deprivation and mTOR inhibition. Second, in RPS26dC cells, the levels of the active phosphorylated AMPK are elevated while mTOR signaling is downregulated, mirroring energy-deprived WT cells. Consistent with that, the translatome of RPS26dC cells is highly similar to that of glucose-starved WT cells. These biological effects are associated with apparent changes in the translation of major energy metabolic enzymes. However, there are also differences between RPS26dC and energy-stressed WT cells, which are highlighted by the apparent differential effects of glucose starvation and the mutation on the translation of OXPHOS and TOP mRNAs. These observations are consistent with gene-specific effects of RPS26 C-terminus.

It was recently reported that eIF3a, the largest subunit of the eIF3 complex that is located close to RPS26 in the 48S complex (52,53), also controls glucose metabolism and AMPK activity via translation regulation of Rheb (54). We examined whether the regulation of energy metabolism of these two adjacent proteins is linked by analyzing the translation level of Rheb in the WT and RPS26dC mutant translatome data but found it to be unchanged, suggesting that the metabolic regulation of these two adjacent proteins is independent of each other.

TISU-containing mRNAs are highly sensitive to mTOR inhibition by either rapamycin or Torin-1 (32,35,36) but are nevertheless resistant to mTOR inhibition during energy stress or active AMPK (37). In contrast, cap-proximal Kozak is inhibited following energy stress or AMPK activation (37). However, the mechanistic basis of the continuous translation of TISU genes under energy stress is unknown and, so far, could not be attributed to any initiation factor. An attractive possibility is the direct involvement of the RPS26 C-terminal domain, which may be modified during energy stress, directly or indirectly, or be removed from the ribosome altogether, as was found in yeast (12). While our biochemical experiments do not support the dissociation of RPS26 from the ribosome nor its modification during energy stress, we cannot exclude the possibility that a small fraction of RPS26 departs from the ribosome during stress. Additional studies are required to directly link the RPS26 C-terminal domain and TISU activity under energy stress.

DATA AVAILABILITY

Data of the RNA-seq presented are available in the GEO database ID: GSE217180.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Minerva Foundation [713877]; Israel Science Foundation [843/17]; Weizmann Institute of Science (internal grants from the Estate of Albert Engleman and Estate of David Levinson). Funding for open access charge: Minerva Foundation.

Conflict of interest statement. None declared.

This paper is linked to: doi:/10.1093/nar/gkad269.