Genome-wide analyses of translation can provide major contributions in our understanding of the complex interplay between virulent factors and host cells. So far, the activation of host translational control mechanisms by bacterial toxins, owing to specific recruitment of mRNAs, RNA-binding proteins (RBPs) and ncRNAs (non-coding RNAs), are far from being understood. In the present study, we characterize for the first time the changes experienced by the translational control system of host cells in response to the well-known Staphylococcus aureus α-haemolysin (AHL) under both sublytic and lytic conditions. By comparing variations occurring in the cellular transcriptome and translatome, we give evidence that global gene expression is primarily rewired at the translational level, with the contribution of the RBP ELAVL1 (HuR) in the sublytic response. These results reveal the importance of translational control during host–pathogen interaction, opening new approaches for AHL-induced diseases.
Staphylococcus aureus is a commensal pathogenic bacterium of human skin involved in different diseases such as pneumonia, infective endocarditis, skin infections, arthritis, septic shock  and squamous cell carcinoma of the skin . Additionally, methicillin-resistant S. aureus strains are the major cause of community-associated methicillin-resistant infections, making the study of its pathogenicity a worldwide priority. S. aureus secretes a wide variety of virulence factors, including the exotoxin α-haemolysin (AHL), the most studied pore-forming toxin (PFT), which induces, at certain concentrations, the formation of pores in the host plasma membrane . AHL is the primary causative agent of an extraordinary multitude of host cell responses [4–9]. The concentration of virulence factors and their interaction with host cells in vivo probably depend on the site of infection and the distance between the target cells and the bacterial foci . Even if it is not completely known, physiological concentrations of toxin seem to be primarily sublytic during in vivo infections . Therefore, the understanding of effects exerted on host cells by both sublytic and lytic doses of PFTs may be of particular relevance.
For a number of PFTs it has been shown that the target cells are able to restore the plasma membrane integrity or respond to membrane damage with mechanical survival mechanisms by using different calcium-dependent strategies [12,13]. Regulation of cell proliferation [14,15] and activation of several pathways related to stress response such as p38 and JNK (c-Jun N-terminal kinase)–MAPK (mitogen-activated protein kinase) have been reported [16–20]. At the molecular level, effects on metabolic pathways, intracellular signalling, proteasome activity, transcription [21,22] and translation [19,21,23] have been demonstrated. Besides the effects due to high toxin doses, several signalling pathways have also been found to be activated in response to low toxin concentrations , but only a few studies have investigated the global changes that occur at the transcriptome level (i.e. the set of all of the mRNAs in the cell) and none of them has examined the translatome level (i.e. the set of mRNAs associated with polysomes). Using low-throughput assays, variations in the expression of few genes have been reported [14,24–26], but these approaches underestimate and ultimately hide a more global cellular response. Although translation has been shown to be a crucial layer in the control of gene expression [27–29], very few studies tackled any genome-wide change in translational control occurring as host responses to virulent attacks .
A useful approach to reveal translational controls under different conditions is the comparison between the transcriptome and the translatome . In the context of host–pathogen interactions, this approach was previously employed to discover multiple genes translationally regulated and involved in response processes of macrophages and dendritic cells exposed to LPS (lipopolysaccharide) [31,32] or infected with hepatitis C virus . Therefore, it is of the utmost importance to investigate the genome-wide translational response to bacterial toxins to unveil how host cells react and eventually resolve the ‘crisis’ triggered by these virulence factors exploiting translational control. This is particularly crucial when considering the faster dynamics associated with translational regulation in comparison with transcriptional regulation. In fact, a sudden stress such as that generated by bacterial PTFs can be more effectively addressed by rapid changes in translation, resulting in immediate outcomes.
In the present study, we chose a host-virulent factor model comprising human cell lines as host and the staphylococcal AHL as virulent factor. By using a combination of genome-wide bioinformatics approaches and classical cellular biology, we identified and globally characterized the host cellular response to the attack of sublytic and lytic AHL at both translational and transcriptional levels. The present study represents the first attempt to specifically analyse the global impact of bacterial PFTs on the translational control of the host cells.
MATERIALS AND METHODS
Cell culture and AHL treatments
Neuroblastoma SH-SY5Y were seeded on adherent plates and maintained at 37°C under 5% CO2 in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin and 100 mg/ml of streptomycin. All AHL and control treatment were performed as follows: after seeding, cells were maintained in culture for 24–72 h in complete medium, then the medium was replaced with DMEM supplemented with 2 mM glutamine and with or without AHL. The control cells were incubated for the same time as the AHL-treated cells. If not specified otherwise, cells were grown until reaching 80% confluence before each assay.
Adenylate kinase assay
At 24 h after seeding 1×104cells/well in a 96-well plate, cells were treated with AHL at different concentrations (2 μM, 1 μM, 200 nM, 5 nM, 12 nM, 3 nM and 0.5 nM) for 4h and 24 h. The adenylate kinase (AK) release was measured using the ToxiLight® BioAssay Sample Kit (Lonza) according to the manufacturer′s instructions. A TECAN infinite M200 instrument was employed for the detection of chemiluminescence. Experiments were run in triplicate and results were assessed for statistical significance using Student's t test; P<0.05 was considered significant.
Determination of haemolytic activity
Rabbit red blood cells (RRBCs) were purchased from ‘Zootecnica Il Gabbiano’. Haemolysis after AHL addition was followed by measuring the turbidity at 650 nm in a 96-well microplate reader (UVmax, Molecular Devices) for 45 min. The percentage of haemolysis was calculated as described previously .
Protein extraction and immunoblotting
Cells seeded at 7×105 cells/well in six-well plates were lysed by RIPA buffer in the presence of phosphatase and protease inhibitors cocktails (Sigma). Equal amounts of proteins were analysed by SDS/PAGE (10/12% gel). Blotting was performed on PVDF membranes (Bio-Rad Laboratories) and blots were processed by an ECL Plus detection kit (GE Healthcare). The chemiluminescence was acquired by ChemDoc-It (Bio-Rad Laboratories) and analysed with ImageQuant Tl software (Bio-Rad Laboratories). All experiments were run at least in biological triplicates. Statistical analysis (Student's t test, P<0.05) was performed on the densitometry values. The anti-p-Akt (phosphorylation sites Ser473 and Thr308), anti-p-RPS6K (ribosomal protein S6 kinase) (phosphorylation site Thr389) anti-p-RPS6 (ribosomal protein S6) (phosphorylation sites Ser235 and Ser236) and anti-RPS6 were purchased from Cell Signaling Technology; anti-Akt, anti-4E-BP1 (eukaryotic initiation factor 4E-binding protein 1) and anti-p-4E-BP1 (phosphorylation sites Ser65 and Thr70), anti-eIF4B (eukaryotic initiation factor 4EB), anti-p-eIF4B (phosphorylation site Ser422), anti-eIF4E, anti-p-eIF4E (phosphorylation site Ser209), anti-ELAV1, anti-β-actin and anti-p53 were purchased from Santa Cruz Biotechnology. Antibodies against AHL, RPL14 (ribosomal protein L14) and CCND1 (cyclin D1) were purchased from Abcam. Secondary antibodies anti-rabbit, anti-mouse or anti-goat, conjugated with horseradish peroxidase (HRP) were purchased from Santa Cruz Biotechnology.
Expression of lytic and monomeric AHL
Recombinant wild-type AHL and recombinant AHL-H35N were produced by coupled in vitro transcription and translation (IVTT) using an Escherichia coli T7-S30 expression system for circular DNA (Promega). AHL oligomers were purified according to .
Polysomal profiling and RNA extraction
Cells were seeded at 1.5×106 cells/dish. After AHL treatment cells were incubated for 5 min with cycloheximide (10 μg/ml) at 37°C to trap the ribosomes on the mRNAs. Cells were washed with PBS complemented with cycloheximide (10 μg/ml) and scraped directly on the plate with 300 μl of lysis buffer [10 mM NaCl, 10 mM MgCl2, 10 mM Tris/HCl, pH 7.5, 1% Triton X-100, 1% sodium deoxycholate, 0.2 units/ml, DNAse I (Fermentas), RNase inhibitor (Fermentas), 1 mM DTT and 10 μg/ml cycloheximide]. Nuclei and cellular debris were removed by centrifugation for 5 min at 12000 g at 4°C. The supernatant was directly transferred on to a 15–50% linear sucrose gradient containing 100 mM NaCl, 10 mM MgCl2 and 30 mM Tris/HCl, pH 7.5, and centrifuged in a Beckmann ultracentrifuge on a swinging rotor for 100 min at 180000 g at 4°C. The collected fractions were used for RNA extraction, HuR RIP (RNA immunoprecipitation) and protein isolation (methanol/chloroform precipitation).
For microarrays, total and polysomal RNA were purified according to Tebaldi et al. . RNA quality was checked using the Agilent 2100 Bioanalyzer platform following the manufacture's guidelines. All experiments were performed in biological triplicate.
Quantitative real time-PCR
Array validation was performed using Taqman probes, whereas the validation of ELAVL1/HuR target was performed using SYBR Green. The primers used for quantitative real-time PCR (qPCR) are listed in Supplementary Table S1.
Data were analysed with Bio-Rad CFX-Manager 1.6 software. Relative quantification of target genes was determined calculating the change in cross-threshold (ΔCt) and the relative ΔΔCt after normalization with the geometric mean of three [Alu-J, 18S and β-actin (ACTB) for SYBR probe] or four [glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ACTB, peptidylprolyl isomerase A (PPIA) and MRPC19 for TaqMan probe] different housekeeping genes according to the Pfaffl method .
ELAVL1/HuR RNA immunoprecipitation from polysomes (RIP assay)
Polysomal fractions were isolated by sucrose gradient fractionation according to the above-mentioned protocol and pooled. The input (polysomal input) was considered as one-tenth of the whole volume and kept aside for RNA extraction. To perform ELAVL1/HuR immunoprecipitation, anti-HuR antibody was added to fractions containing polysomes and kept for 2 h with gentle rotation at +4°C. Dynabeads conjugated to Protein G (Life Technologies 10003) were then added and left rolling for 1 h at +4°C. Beads were washed thrice with 100 mM NaCl, 10 mM Tris/HCl and 10 mM MgCl2, pH 7.4, before RNA extraction using TRIzol reagent. RNA samples were dissolved in 20 μl of RNase-free water and retro-transcribed using SuperScript® VILO cDNA Synthesis Kit before qPCR analysis. The fold enrichment variations of ELAVL1 (HuR) targets [CCND1, TP53, RPL14 and cyclin-dependent kinase 1A (CDK1A)] upon toxin treatment were calculated using the difference between the Ct of polysome input and the Ct of ELAVL1 immunoprecipitation (ΔCt) and dividing by ΔCt for the housekeeping gene Alu-J. The level of the control without toxin addition was taken as reference and fixed to 1.
Total and polysomal RNA samples were submitted to microarray analysis, as described in . Experiments were run in biological triplicate. Differentially expressed genes (DEGs) were determined with the Bioconductor tRanslatome package  adopting a double threshold based on: (i) the magnitude of the change (absolute log2 fold change > 0.5); and (ii) the statistical significance of the change, measured with a rank product test (P-value < 0.01). Microarray data are available through the Gene Expression Omnibus database using accession numbers GSE50652 (for sublytic AHL) and GSE60543 (for lytic AHL).
Functional annotation enrichment analysis
The ToppGene resource  was used for enrichment analysis of the transcriptome and the translatome DEGs lists. The AURA 2 (Atlas of UTR Regulatory Activity 2) resource  was used for enrichment analysis of post-transcriptional regulators, based on the annotation of binding sites on UTRs of transcriptome and translatome DEGs. Enrichment was tested with the Fisher exact test. The significance of over-representation was determined at a 0.05 FDR (false discovery rate) threshold.
Cells were seeded at 7×105 cells/dish in 35-mm dishes in complete DMEM, until reaching 100% confluence. A scratch was then performed through the cell layer using a sterile micropipette tip. After washes with PBS, serum/antibiotic-free medium was added before toxin addition to the cells. The wounded area was marked with a black label and images were captured immediately after toxin addition (time 0) and at regular intervals of 3 h within a time window of 9 h. Images were acquired by using an optical microscope. The migration abilities were quantified by measuring the length of the scratched regions with ImageJ software (NIH). The experiment was performed in triplicate.
Definition of AHL sublytic doses
AHL is released by S. aureus as a 33.2 kDa monomer that oligomerizes into transmembrane pores on the host cell plasma membrane [40,41]. It causes swelling and cellular lysis at a wide range of concentrations, depending on the target cell [42,43]. Similar to most PFTs, AHL follows a multi-step mechanism: (i) release of the soluble monomers, (ii) binding to the target membrane, (iii) oligomerization in a non-lytic pre-pore, and (iv) insertion of the pore-forming domain into the lipid bilayer [43,44]. This very last event leads to osmotic imbalance and, eventually, to cell death. Interestingly, at low doses of PFTs, cellular responses have been observed to be extremely diverse . Given that pore formation is a strictly sequential and dynamic process, all of the intermediates leading to pores (i.e. monomers and non-lytic pre-pores) may be present at any time on the plasma membrane [45,46].
To separate the effect of sublytic and lytic AHL, we used as lytic AHL the purified and pre-formed lytic pores, which can be obtained according to the procedure developed in . This allowed us to study the effect of ‘lytic AHL’ without any additional intermediate on the plasma membrane (see the Materials and methods section and Supplementary Figure S1). In the case of the sublytic concentration, it is of primary importance to define it in the cell system of choice as precisely as possible. As demonstrated in model membranes , the equilibrium between monomers, pre-pores and pores on the target cell membrane depends on the lipid/toxin ratio. Therefore, prior to studying its sublytic effects on gene expression, we incubated different amounts of plasma membranes obtained from the widely used cellular model for AHL, RRBC, with 70 nM AHL (Figure 1A). As expected, high lipid/toxin ratios led to a decrease in AHL oligomers and an increase in monomeric AHL. During haemolysis, stable oligomers form on the cell membrane and are still detectable even in the absence of haemolysis (Figure 1B), suggesting that at certain sublytic doses few pores may be inserted into the membrane even in the absence of measurable lysis. The first concentration at which some oligomers form without inducing haemolysis can be considered the upper threshold of the so called sublytic dose. By comparing this evidence with the results obtained in Figure 1(A), we reasoned that a concentration window 4–6-fold lower than the sublytic upper threshold is needed to have most of the toxin in the monomeric and membrane-bound state.
Definition of AHL sublytic doses in human cell lines
Then, we measured the cytolytic effect of native AHL at different concentrations on human neuroblastoma cell lines (i.e. SH-SY5Y, Figure 1C) that have been previously employed to study cellular response to bacterial PFTs [48,49] and, more importantly, express ADAM10 (a disintegrin and metalloproteinase 10), the putative receptor for AHL [50,51]. We determined the sublytic concentration threshold after 4 and 24 h of incubation with AHL. The sublytic threshold (Figure 1C, black lines) was 50 nM. Thus, we set our ‘test sublytic’ concentration at 3 nM.
Sublytic and lytic AHL trigger different responses on transcriptome and translatome of host cells
To systematically monitor the gene expression response of host cells to sublytic and lytic AHL, we performed genome-wide multi-level analyses based on polysomal profiling , comparing the effect of lytic and native sublytic AHL at the concentration described in the previous paragraph. After exposure of SH-SY5Y to sublytic (3 nM native) and lytic (purified and pre-formed lytic pores) AHL for 2 h, polysomes were purified using sucrose gradient fractionation (Figures 2A and 2B). Then both polysomal and total RNA were extracted and both host transcriptome (total RNA, portraying the total amount of transcribed mRNAs) and translatome (polysomal RNA, portraying the transcripts associated with polysomes) in the two conditions were profiled.
Transcriptome and translatome variations (displayed in Figure 2C for sublytic treatment and Figure 2D for lytic treatment), in response to AHL are globally independent (Spearman correlation coefficient 0.19 for the sublytic treatment and 0.14 for the lytic treatment). DEGs induced by sublytic or lytic AHL treatments were determined based on fold-change magnitude and statistical significance  (see Supplementary File 1 for the complete results). After sublytic treatment, 192 DEGs were detected in the transcriptome and 920 in the translatome (Figure 2E, left panel, and Supplementary Figure S2). Comparing the two levels, 82% of DEGs (863 genes, in yellow in Figures 2C and 2E, left panel) significantly changed only at the polysomal level and 13% of DEGs (135 genes, in blue in Figures 2C and 2E, left panel) changed significantly only in the transcriptome. According to these results, cells replied to sublytic AHL treatment by chiefly changing the levels of mRNAs uploaded on to polysomes rather than inducing changes in the total level of mRNAs. Only 40 genes (4% of DEGs, in green in Figures 2C and 2E, left panel) showed significant homo-directional changes, reflecting a concordance between transcriptome and translatome variations.
Genome-wide translational reshaping after sublytic and lytic AHL treatment
Looking at the lytic treatment, we again observed a strong uncoupling between transcriptome and translatome variations. In fact, the fraction of homo-directional DEGs is only 6% (Figures 2D and 2E, right panel). Differently from the sublytic treatment, the lytic response is characterized by an almost equal number of transcriptome DEGs (638) and translatome DEGs (709).
With both lytic and sublytic treatments, more than 50% of DEGs were mainly controlled at the translation level, suggesting that translation may reply to rapid environmental stimuli before transcription.
Overall, our analysis suggests that AHL triggers in host cells an elaborate reshaping of gene expression, characterized by: (i) strong translational changes after the sublytic treatment; (ii) pronounced transcriptional changes upon the lytic treatment, and (iii) a poor overlap of transcriptome and translatome DEGs, upon both sublytic and lytic AHL treatments.
Sublytic AHL affects factors controlling translation
Genome-wide expression analysis revealed an overwhelming majority of changes at the level of translation for sublytic AHL treatment. Being the most energy consuming process in cell, translation is tightly controlled and translation initiation is commonly considered the main bottleneck of the process . Two pathways are mainly involved in controlling translation initiation: the MAPK pathway and the phosphoinositide 3-kinase (PI3K)/Akt/mammalian target of rapamycin complex 1 (mTORC1) pathway . Therefore, we checked whether the exposure to sublytic AHL induced any changes in the phosphorylation state of cellular markers for general translational control, such as the PI3K/Akt/mTORC1 downstream targets Akt, 4E-BP, RPS6K and RPS6; the MAPK/extracellular-signal-regulated kinase (ERK) downstream target eIF4E and finally eIF4B that is a common target of both pathways [53,55,56]. The cellular response upon sublytic AHL treatment was studied for three sublytic concentrations (3, 6 and 12 nM Figure 3A and Supplementary Figure S3) and at four times of exposure (5, 15, 60 and 120 min, results not shown). A significant activation of Akt (both Thr308 and Ser473 phosphorylation sites) at the lowest sublytic AHL concentration (3 nM) and after 2 h of incubation (Figure 3B) was detected, whereas no significant activation of 4E-BP was observed (Figures 3B and 3C). In addition to this, the significant increase in phosphorylation for eIF4E, RPS6 (Ser235 and Ser236) and eIF4B (Ser422) [56–58] indicates that pathways involved in translational rewiring are indeed engaged in the response to sublytic AHL. Undoubtedly, given the controversial implication of these phosphorylation events to the activation or inactivation of cap-dependent translation, additional studies are required to elucidate the precise connection between the triggering of these pathways and the positive or negative effect on translation. Nonetheless, a modulation of translational effectors after sublytic AHL treatment is indeed occurring.
Sublytic dose of AHL affects the phosphorylation state of translational factors
The results obtained with sublytic AHL treatment of SH-SY5Y were confirmed in HeLa cells (Supplementary Figure S4). Another neuroblastoma cell line (CHP-134) was almost insensitive to AHL-dependent activation of Akt (Supplementary Figure S4), owing to the lipid composition of these cell lines that is characterized by a lower amount of cholesterol and sphingomyelin (Supplementary Figure S5). In fact, the cholesterol depletion in SH-SY5Y cells with β-methyl cyclodextrin  led to the complete disappearance of the previously detected phosphorylation (Supplementary Figure S5), suggesting that membrane composition may affect the sensitivity of the cellular response to sublytic doses of AHL.
To understand whether the observed phosphorylations of translational effectors were specific to sublytic AHL treatment, we used the lytic AHL and, as a control for the sublytic exposure, a non-lytic mutant, AHL-H35N, which is able to bind to the membrane but not to oligomerize in a lytic pore [60–62]. We treated cells with 3 nM lytic AHL or with 3 nM monomeric AHL-H35N and checked for Akt (Thr308 and Ser473), eIF4E and 4E-BP phosphorylation (Figure 3D). Similarly to what observed for the sublytic AHL treatment in Figures 3(A)–3(C), the non-lytic mutant AHL-H35N induced a significant phosphorylation of Akt at both Thr308 and Ser473 phosphorylation sites and a slight increase in p-eIF4E. On the other hand, lytic AHL was completely ineffective to phosphorylate translational effectors, rather inducing a depression of 4E-BP phosphorylation, enforcing the previous idea that pores and monomers may elicit different translational responses in host cells.
Functional enrichment analysis reveals a translational-specific response toward the plasma membrane for sublytic AHL
To understand from a broader perspective the implications of the cellular reply to AHL treatments, we focused our analysis on the functional annotation of DEGs. To this end, ontological enrichment analysis was performed on the lists of transcriptome and translatome DEGs identified after sublytic and lytic AHL treatments. The comparison of the top enriched terms is shown in Figure 4(A) for annotations concerning biological processes, cellular localizations and protein domains (see Supplementary File 2 for the complete enrichment results). Enrichment results suggest that transcriptome and translatome DEGs are involved in disjoined biological processes. In accordance with the fact that the majority of variations are almost purely translational upon sublytic AHL treatment, we did not observe any significant enrichment among transcriptome DEGs (Figure 4A, first column). On the other hand, translatome DEGs triggered by sublytic AHL are consistently associated to membrane processes and components, as shown by the enrichment of terms as ‘receptor signalling pathway’, ‘plasma membrane region’, ‘7 transmembrane receptor’ and ‘cell adhesion’ (Figure 4A second column).
Uncoupling between transcriptome and translatome DEGs reflect enrichments of specific biological themes
This last result prompted us to check for any changes in the functionality of cell adhesion by using a cell motility assay, because motility could be increased following adhesion decrease. Interestingly, we found that after sublytic AHL treatment, SH-SY5Y increased their mobility (Figure 4B) in agreement with previous data demonstrating that AHL up-regulates ADAM10 enzymatic activity inducing VE (vascular endothelial) cadherin cleavage and endothelial barrier disruption . This functional analysis reinforces the evidence that, after sublytic exposure to AHL, an mRNA-specific host translational control is activated with a semantically coherent response toward the plasma membrane. Involvement of immune response is also suggested by the enrichment of terms such as ‘regulation of inflammatory response’ and ‘immunoglobulin V-set domain’.
Looking at the lytic AHL treatment, the functional scenario is considerably different. The translational response is characterized by an enrichment of genes belonging to the histone family and genes coding for proteins positioned on the mitochondrial membrane and acting as oxidoreductases (Figure 4A, fourth column). This last finding is consistent with oxidative stress associated with the lytic action of AHL [42,63]. Conversely, transcriptome DEGs did not display strong enrichments, additionally suggesting that a more selective and targeted functional response is orchestrated by translational regulation.
ELAVL1 is a post-transcriptional factor involved in response to sublytic AHL
The results described in the previous sections suggest the involvement of finely tuned translational mechanisms to guide the regulation of specific genes upon AHL injury. As outlined in the introduction, RNA-binding proteins (RBPs) and ncRNAs (non-coding RNAs) are the two main classes of trans-factors driving binding-dependent regulation of translation [64,65]. To detect which trans-factors could mediate translational specific responses of host cells to AHL, we searched for enrichments of transcripts known to be targets of RBPs and miRNAs among the populations of transcriptome and translatome DEGs. Using the experimental annotation of interactions between human UTRs and post-transcriptional trans-factors collected in the AURA2 database , we calculated the over-representation of targets of specific RBPs or miRNAs. The results obtained are shown in Figure 5(A) and are organized as a heat-map containing the set of trans-factors whose targets are enriched in transcriptome or translatome DEGs after sublytic or lytic AHL treatments. The enriched trans-factors are ordered by decreasing statistical significance (see Supplementary File 3 for the complete results). Many general factors such as AGO1 (Argonaute 1) and AGO2 have enriched targets in multiple populations of DEGs, whereas two miRNAs are specifically associated to lytic AHL translatome DEGs. One of these, miR-34, has been reported to play important roles in the onset of many tumours .
Sublytic AHL induces the recruitment of host post-transcriptional trans-acting factors
The most striking result from this analysis is that the well-known and well-studied RBP ELAVL1 (HuR), was predicted to be the RBP most selectively associated with translatome DEGs and specifically enriched in the sublytic treatment (Figure 5A, first row). The Hu/ELAV family member ELAVL1 binds to the 3′-UTR of hundreds of mRNAs, regulating transcript stability and access to polysomal complexes [67,68].
Driven by this result, we analysed by immunoblotting the cytoplasmic levels of ELAVL1 at different concentrations of sublytic AHL (Figure 5B). Only upon treatment with sublytic AHL did ELAVL1 protein increase its cytoplasmic levels without a concomitant increase in transcript expression (Supplementary File 1). In accordance with this result, the cytoplasmatic levels of ELAVL1 protein increased after treating cells with the monomeric AHL-H35N, whereas no changes occurred after the lytic treatment (Figure 5C). These results are consistent with the specific enrichment of mRNAs targeted by ELAVL1 and translationally modulated following the sublytic stimulus.
To better elucidate the possible involvement of ELAVL1 in translation rewiring, we studied the distribution of ELAVL1 protein along polysomes, finding that it was altered only upon sublytic AHL treatment. As shown in Figure 6(A), ELAVL1 is increasingly recruited on to heavy polysomes after sublytic treatment with respect to the untreated sample. As a control, we checked the distribution of the RBP QKI, showing lower target enrichment than ELAVL1 (and therefore not shown in Figure 5A). We found that QKI did not modify its localization along polysomal fractions (Figure 6A). In accordance with enrichment data in Figure 5(A), lytic AHL treatment did not change ELAVL1 protein distribution along polysomes (Figure 6A), suggesting again that ELAVL1 involvement is specific to the sublytic treatment.
ELAVL1 recruitment on heavier polysomes occurs after AHL sublytic treatment
To obtain further evidence, we selected four known ELAVL1 targets (TP53, RPL14, CDK1A and CCND1) [67,69–72] and addressed their recruitment in the translationally active compartment (polysomal fractions) with respect to the translationally silent compartment (sub-polysomal fractions, i.e. ribonucleoparticles and 80S) by using qPCR (Figure 6B, left panel). We observed a significant increase in the polysomal levels of RPL14 and TP53, a well-studied regulator of multiple cellular responses. In accordance with our results, a previous study found that TP53 is translationally enhanced by ELAVL1 in response to cell damage .
To understand whether the observed changes of ELAVL1 along polysomal fractions and of its targets were ELAVL1-specific, we performed an RIP assay on polysomal sucrose fractions and purified ELAVL1-associated transcripts from polysomes after sublytic treatment. The recruitment of ELAVL1-targets on polysomes was studied by qPCR. The ELAVL1-dependent enrichment on to polysomes of the previously analysed transcripts upon sublytic AHL exposure (Figure 6B, right panel) shows that TP53 and CCND1 transcripts are specifically regulated by ELAVL1. In particular, TP53 was uploaded on polysomes by ELAVL1, whereas CCND1 was down-regulated. RPL14 displayed an ELAVL1-dependent decrease, demonstrating that the increased translation efficiency of RPL14 observed in Figure 6(B), left panel, is not ELAVL1-specific.
To validate the previous results, we then analysed the protein levels of three ELAVL1 targets (Figure 6C). We also monitored the protein abundance after inhibition of the proteasome, because AHL is able to stimulate proteasome activation and protease up-regulation [73,74]. In fact, if post-translational degradation of proteins occurred, we could not directly compare the protein level with the polysomal analysis shown in Figure 6(C) upper and middle panel. The protein level of CCND1 shows a slight decrease after toxin addition, in agreement with data displayed in Figure 6(B), right panel, whereas RPL14 remains unchanged. TP53 protein levels significantly increased with respect to the control sample when the proteasome was inhibited. This is in accordance with the fact that the intracellular levels of proteins with short half-lives, such as TP53, are regulated by the balance between translation and proteasome degradation . These results confirm the ELAVL1-specific uploading of TP53 transcript on polysomes and its overall increased translation (Figure 6B, right panel).
In summary, we showed the activation of a complex post-transcriptional network of gene expression regulation, with the involvement of translational trans-acting factors, such as ELAVL1, in response to the sublytic attack.
Concentrations of virulent factors, such as bacterial PFTs, are most probably sublytic in vivo . To date very few studies have addressed the effects of low doses of native AHL even if several cellular responses during toxin pore-formation were described, ranging from cellular proliferation , caspase activation and oligonucleosomal DNA fragmentation [62,63] to Akt inhibition  and histone methylation . The impact of sublytic doses compared with lytic doses of toxin on translational control of host cells is still understudied, despite translation is increasingly recognized as a central player of gene expression remodelling. In the present study, we compared sublytic and lytic AHL treatment of cells to obtain information on possible translational control impairments relevant to human infections and verify the involvement of translational trans-factors, such as RBPs, as players in the host–pathogen response.
First, we characterized the effect at the translational level of both sublytic and lytic exposure of host cells to AHL, unearthing strong and translational specific reshaping of gene expression. In both cases, we found that the translational response to AHL is globally independent of the transcriptional response. Especially with sublytic AHL treatment, cells responded mainly by changing the mRNAs uploaded on to polysomes rather than inducing de novo mRNA transcription. Given that the majority of significant variations are exclusively translational, bypassing transcription, a predominant role for translational regulation seems to be implicated in mediating the cellular response to short exposure of sublytic AHL. Interestingly and in line with our results, scarce transcriptional variations have been observed as host response to pore-independent treatments of HeLa cells with AHL [25,26].
A tightly regulated control of mRNA translation is provided by phosphorylation and de-phosphorylation of components of the two pathways involved in controlling translation. At this stage, a number of critical cell signalling pathways converge. To confirm the involvement of translational controls, we checked two main signalling pathways modulating translation initiation: PI3K/Akt/mTORC1 (with Akt, 4E-BP, eIF4B, RPS6K and RPS6) and MAPK (with eIF4E and eIF4B) after sublytic and lytic AHL treatments. We found increased phosphorylation of Akt, RPS6 and eIF4B, in a sublytic specific-manner. The involvement of ERK1/2, JNK and MAPK, in addition to PI3K/Akt/mTORC1, is suggested by the observed phosphorylation of eIF4E. Even if the function of eIF4E phosphorylation remains controversial [76,77], an increase in phosphorylation of eIF4E on Ser209 was demonstrated to increase the translation of subset of tumour-promoting mRNA  and to be involved in promoting cell survival after cell stress . The effect of the lytic AHL was different from that observed for the sublytic treatment. In accordance, in previous studies [19,21], the effect on translation of lytic concentration of PFTs and the eIF2A-dependent inhibition of protein synthesis has been observed . Translation inhibition by PFTs produced by Pseudomonas entomophila was also demonstrated in Drosophila gut and correlated to 4E-BP .
The transcriptional and translational uncoupling upon AHL treatment was confirmed by functional annotation enrichment analysis, showing that the transcriptome and translatome DEGs are involved in disjoined biological processes. Many histone transcripts are translationally modulated upon lytic AHL treatment, suggesting a possible effect on chromatin remodelling, as already demonstrated in [25,26]. Remarkably, after sublytic AHL treatment, translatome DEGs are significantly associated with membrane processes and components, suggesting that membrane perturbations promote a coherent post-transcriptional response toward the same cellular compartment, with the possible involvement of translational factors (Supplementary Figure S6). Up-regulation of cell–cell adhesion and motility was also observed and experimentally validated. In line with this evidence, it has been proven that AHL plays a role in the extracellular matrix reorganization and in mediating invasiveness of S. aureus in tissues [80–82].
By enrichment analysis of experimentally known binding sites on UTRs of mRNAs, the targets of many RBPs and few miRNAs emerged to be specifically enriched only when considering translatome DEGs. Among the most important trans-factors responsible for translational control, ELAVL1 showed the highest enrichment in targets among translatome DEGs, specifically after sublytic AHL treatment. ELAVL1 is a pleiotropic protein  able to regulate many physiological processes and it is known to act on target mRNAs as a stabilizer and a translational enhancer, binding to AU-rich element (ARE)-containing mRNAs [69,84]. In addition to AMP-activated kinase family (AMPK)  and protein kinase C (PKC) , ELAVL1 shuttling from nucleus to cytoplasm is promoted by MAPK . In accordance with this latter evidence and the observed enrichment in targets among translatome DEGs, we detected an increase in ELAVL1 cytoplasmic protein levels only after sublytic AHL treatment. Since this RBP mediates stimulus-induced targeting of mRNAs from translational inactive RNPs to polysomes , we demonstrated that ELAVL1 shifts towards heavy polysomal fractions after sublytic AHL treatment. Finally, we showed that the ELAVL1 target TP53 is uploaded on to polysomes after sublytic AHL treatment. This upload is ELAVL1-dependent, as demonstrated by RIP assay. Increased polysomal upload results in increased TP53 protein synthesis, in accordance with its involvement during cellular warning in response to various cellular stresses . On the other hand, we showed that increased protein synthesis of TP53 is counterweighted by increased protein degradation, through proteasome recruitment. This result represents an intriguing example of the complex interplay between the processes triggered by the virulent factor and the counteracting response of the host at the translational level.
We performed, for the first time, a combined genome-wide analysis of total and polysomal RNA after sublytic and lytic AHL stimulation of human cells. We observed strong mRNA hijacking (i.e. selection of specific mRNAs) at the translational level after both treatments, with an almost complete bypass of transcriptional reprogramming after sublytic AHL stimulation. By functional enrichment analysis, the cellular response to sublytic AHL stimulus is compartment-specific and probably characterized by a boost for translation of many membrane-related genes and the involvement of the RBP ELAVL1 (Supplementary Figure S6). A completely different cellular response is triggered by lytic AHL, most probably mediated by alternative regulation networks.
This global approach provides new insights into the biology of an ancient and finely tuned response, representing the first demonstration that a membrane perturbing stimulus induces a translational-specific response of target cells, with recruitment of specific genes and trans-factors. Further studies will be required to understand which structural elements (cis-acting factors) on the mRNA may act in combination with trans-acting factors to finally decode a membrane-related translational regulon . In conclusion, the present work reveals the importance of translation regulation in host–pathogen interaction, opening new horizons for studying AHL-induced diseases.
Massimiliano Clamer performed the experiments and elaborated the data. Toma Tebaldi performed all of the analyses of genome-wide data. Paola Bernabò, Gabriella Viero and Marta Marchioretto performed the polysomal profiling with lytic AHL and experiments shown in Figure 6. Efrem Bertini performed the polysomal RIP assay of ELAVL1. Graziano Guella performed the NMR experiments. Graziano Guella conceived and designed the study and organized the figures. Graziano Guella, Massimiliano Clamer, Toma Tebaldi, Mauro Dalla Serra and Alessandro Quattrone wrote the paper. All authors have read and approved the final paper.
We thank Hagan Bayley and Alessandro Provenzani for the gift of lytic and AHL-H35N plasmids and of ELAVL1/HuR target primers respectively. Special thanks go to Valentina Adami [Core Facility, High Throughput Screening (HTS) CIBIO] for the technical support with Microarray hybridization, Matteo Gaglio for the skilful help with polysomal profiling and Alberto Inga for kindly giving anti-TP53 antibody and for useful comments and suggestions. This paper is dedicated to the memory of Gianfranco Menestrina.
This work was funded by Fondazione Caritro (Trento, Italy), the University of Trento (CIBIO start-up grant), and sponsored by IMMAGINA BioTECHNOLOGY.
a disintegrin and metalloproteinase 10
Atlas of UTR Regulatory Activity
cyclin-dependent kinase 1A
differentially expressed gene
Dulbecco's modified Eagle's medium
eIF4E-binding protein 1
eukaryotic initiation factor 4
c-Jun N-terminal kinase
mitogen-activated protein kinase
mammalian target of rapamycin complex
quantitative real-time PCR
ribosomal protein L14
ribosomal protein S6
rabbit red blood cell
change in cross-threshold
This paper is dedicated to the memory of Gianfranco Menestrina.
These authors contributed equally to this work.