SunRiSE – measuring translation elongation at single-cell resolution by means of flow cytometry

Protein synthesis is a central metabolic process that is regulated at different biochemical levels. Polypeptide elongation by ribosomes proceeds discontinuously, with pauses regulating the speed of protein synthesis (Petersen et al., 2006; Richter and Coller, 2015). The total level of engaged ribosomes and the associated elongation rate are affected by translation elongation and initiation factors, codon usage, tRNA abundance and post-transcriptional modifications, as well as by different steric impediments, such as mRNA secondary structures or microRNA (miRNA) expression (Petersen et al., 2006; Richter and Coller, 2015). In vitro translation assays and ribosome profiling have recently demonstrated that elongation speed affects co-translational protein folding and protein expression (Buhr et al., 2016; Yu et al., 2015). The relative contribution of the different environmental and endogenous factors on translation elongation in cells directly isolated from human blood or mice tissues is poorly understood and is still methodologically demanding. Approaching these questions requires a technique with high-throughput potential and that measures translation elongation rates under physiological conditions in non-abundant and non-transformed cells ex vivo. Reduced manipulation and applicability to complex mixes of cells should also be a requirement for such a method. Among the techniques available for assessing translation elongation rates, one can distinguish methods that measure an average rate within a bulk of cells (Arava et al., 2003; Brar and Weissman, 2015; Dieterich et al., 2010; Ingolia et al., 2009; Starck et al., 2004) from those that measure translation rates at the single-mRNA molecule scale and require multiple steps of molecular engineering (Halstead et al., 2015; tom Dieck et al., 2015; Wang et al., 2016; Yan et al., 2016). The first group of methods includes ribosome profiling, polysome profiling and radiolabeled methionine incorporation assays. Given the advantages and recent improvements in next-generation sequencing, ribosome profiling has become a gold standard to monitor translation and ribosome engagement (McGlincy and Ingolia, 2017). This technique allows for qualitative and quantitative measurement of translation by sequencing RNA fragments that are 28 and 29 nucleotides long and are protected from degradation through their association with ribosomes. Single-cell ribosome profiling has, however, never been performed, and seems for the moment to be out of technical reach, leaving ribosome profiling approaches with the disadvantage of measuring translation through the averaging of many cells, sometimes from very heterogeneous populations, during which individual phenotypes are binned and diluted into one unique dataset.

Puromycin (puro) is an aminonucleoside Tyr-tRNA mimetic antibiotic that enters the A site of ribosomes, and is incorporated into nascent chains. Puro incorporation induces the termination of peptide synthesis, and this covalent reaction between nascent proteins and puro is called puromycilation (Azzam and Algranati, 1973). Puromycilation detection with ad hoc antibodies and measurement by flow cytometry has proven to be an extremely versatile method to replace radiolabeled amino acid incorporation as a means to monitor protein synthesis levels (Goodman et al., 2012; Schmidt et al., 2009). Here, we propose to combine puro immuno-detection together with ‘run-off’ experiments on translating ribosomes to determine the translation elongation rate in different cell populations, including human blood cells and mouse embryonic cells. Importantly, accurate rate measurements allow the direct comparison of protein synthesis activity in different cells types and tissues, independently of the variations observed in cell size, amount of ribosomes or environmental growth conditions. This novel method called ‘SunRiSE’ (for SUnSET-based Ribosome Speed of Elongation; SUnSET refers to ‘Surface sensing of translation’) offers novel exploratory capabilities, which are illustrated here by an analysis of protein synthesis in normal and stress conditions after genetic inactivation of translation elongation factor 2 kinase (eEF2K) in mouse embryonic fibroblasts (MEFs). Surprisingly, eEF2K ablation mostly affects ribosome engagement, and most likely protein synthesis initiation at steady state, rather than ribosome processing speed. The impact of eEF2K on translation elongation is, however, revealed upon ER-stress induction, since the translation intensity and elongation rate in eEF2K-deficient cells remain unaffected by tunicamycin treatment. Interestingly, the levels of phosphorylated eEF2 (P-eEF2) are highly heterogeneous among single wild-type (WT) MEFs grown in the same culture conditions, and SunRiSE reveals that eEF2 phosphorylation at steady state does not necessarily correlate with a lower protein synthesis rate. The existence of such biochemical heterogeneity further demonstrates the importance of developing novel techniques, like SunRiSE, to measure protein synthesis at high resolution and study its impact on the control of protein expression.

By using puro as a nascent protein tag and anti-puro monoclonal antibodies for its detection, we have previously introduced puro-immunomonitoring and, in particular, flow cytometry to measure translation in living cells and organisms (Schmidt et al., 2009; Seedhom et al., 2016). Surface sensing of translation (SUnSET) allows for measuring global translation in individual cells found in mixed populations grown in vitro or ex vivo. Despite these clear advantages, quantitative SUnSET measurements are not absolute and can only be performed relative to a given reference sample, and are therefore not adapted for a direct comparison among many cellular samples of different origins and obtained at different times. A precise measure of a translation elongation rate in different populations should, however, solve this issue, by attributing an elongation rate constant to each specific cell group. Such a rate constant would allow a direct quantitative comparison of variations in protein synthesis activity among different cells present in heterogeneous populations or tissues, and growing in different environmental conditions.

In addition to puro, antibiotics interfering with ribosome assembly and function have proven to be invaluable reagents to study translation mechanisms, and in particular initiation and elongation. Harringtonine efficiently blocks the initiation of elongation during protein synthesis, and thus had been used with ribosome profiling in run-off experiments to establish the rate of elongation of mammalian ribosomes (∼330 codons/min) (Ingolia et al., 2012). We hypothesized that the rate of translation elongation in single cells could be established by measuring puro incorporation by flow cytometry after blocking translation initiation with harringtonine at different time intervals (Fig. 1). The decay in puro labeling upon harringtonine treatment is linked to the speed at which ribosomes run-off from mRNAs, and therefore directly provides an accurate measure of the global translation elongation rate (Fig. 1). We first established the kinetics at which polysomes are lost upon harringtonine treatment by following the rRNA distribution after sucrose gradient separation. Most polysomes were found to run-off from mRNAs within 8 min of translation initiation inhibition (Fig. 2A), and remained completely stabilized after treatment with the nascent chain elongation inhibitor emetine (David et al., 2013). Upon addition of puro at various times (30 s to 1 min) after harringtonine treatment, a decay in the incorporation of the nucleosidic antibiotic into neosynthetized proteins was observed by immunoblotting (Fig. 2B). Again, the loss of puro labeling upon translation initiation arrest was prevented by blocking translation elongation with emetine. As a consequence of ribosome elongation during translation initiation inhibition, loss of puro labeling was first observed within the population of small nascent proteins and, later on, in larger peptides displaying higher molecular masses (Fig. 2B,C). Taken together, these results indicate that puro incorporation decay after inhibition of the first round of translation elongation, provides an accurate measure of the processing speed of ribosomes along mRNAs, thus supporting our initial hypothesis, and calling us to explore the large potential of this method, termed ‘SUnSET-based Ribosome Speed of Elongation’ (SunRiSE), by switching to flow cytometry monitoring of puromycin incorporation.

where the k value is a constant related to the elongation rate of the ribosomes in a given cell population.

When elongation is blocked by emetine, k tends towards zero, and anti-puro staining levels at different time points remain constant and equal to puro_initial at all times points [e^0×t=1, then puro(t)=puro_initial]. k values therefore permit a direct comparison of elongation rates in situations during which many parameters fluctuate, and different intensities of protein synthesis are displayed by distinct cell groups. A comparative measurement of translation elongation rates upon HBSS starvation of mouse embryonic fibroblasts (MEFs) is shown in Fig. 3B to illustrate this point. It has been previously shown, through ribosome profiling, that a decrease in the elongation rate of polysomes can be observed in starving MEFs (Ingolia et al., 2012). We could confirm this observation by performing SunRiSE after 4 h of HBSS starvation and measuring k values, which were reduced by half compared to normally fed control cells (Fig. 3B).

We next performed multi-parametric analysis of single cells present in a complex suspension of spleen cells (Fig. 4). Mouse splenocytes were chosen as an experimental model, given their large heterogeneity, the availability of multiple cell-specific markers suitable for flow cytometry analysis, and their capacity to undergo rapid changes in protein synthesis intensity upon stimulation with bacterial lipopolysaccharides (LPS) (Lelouard et al., 2007) or a cocktail of phorbol 12-myristate 13-acetate (PMA) and ionomycin (Schmidt et al., 2009). Our experiments were designed to monitor the total levels of protein synthesis in different cell populations simultaneously, but more importantly, to define whether elongation rates are modified in different lymphocytes subsets after activation or cell cycle initiation (Fig. 4A). Flow cytometry was performed, using a gating strategy in which B cells were identified as CD4⁻ CD8⁻ CD11c⁻ CD19⁺ B220⁺ MHC-II⁺ cells, and CD4⁺ T cells by a CD4⁺ CD8⁻ CD19⁻ CD11c⁻ MHC-II⁻ B220⁻ phenotype (Fig. 4B). Phosphorylation of the ribosomal protein S6 (P-S6) (Lelouard et al., 2007) and the active cell cycle marker Ki67 (also known as MKI67) (Yu et al., 1992) were used, respectively, to evaluate metabolic activation (e.g. AKT/mTOR signaling axis) and entry in cell cycle within 16 h of stimulation (Fig. 4B,C). In B cells, overnight stimulation with PMA increased both global translation levels and speed of elongation, but did not augment the proportion of Ki67⁺ cells (Fig. 4C). In contrast, LPS treatment induced Ki67 expression and augmented the translation elongation rate. Both LPS and PMA/ionomycin treatment induced an increase in P-S6 levels in B cells, suggesting that translation upregulation is generally associated with S6 phosphorylation, but not necessarily with cell cycle entry. The translation rates (k) of cells stimulated with PMA were different from those triggered by LPS, demonstrating that SunRiSE is sensitive enough to discriminate among different levels of activation and translation regulation. As expected from previous work, showing that LPS does not affect T cell proliferation nor cytokine secretion (Zanin-Zhorov et al., 2007), CD4⁺ T cells only responded to PMA/ionomycin by increasing cycling and S6 phosphorylation (Fig. S1), while augmenting both protein synthesis intensity and associated translation elongation rate (Fig. 4C).

During mitosis, the overall level of protein synthesis has been described to be 35% lower, with some specific mRNAs showing an up to 10-fold reduction in their translation rate (Tanenbaum et al., 2015). The cell cycle induction triggered by PMA/ionomycin in CD4⁺ T cells gave us the opportunity to comparatively measure translation rates in proliferating cells based on Ki67 antigen positivity or carboxyfluorescein succinimidyl ester (CFSE) labeling over 2 days of activation (Fig. 4D; Fig. S2). This enabled us to determine the level of translation and elongation in cells that had undertaken a different number of divisions. PMA-stimulated Ki67⁻ T cells displayed a higher translation intensity and elongation rate than their Ki67⁺ counterparts, a situation also observed in the small fraction of PMA-activated B cells found to be Ki67⁺ (Fig. 4D). However, contrasting with PMA/ionomycin activation, LPS-activated Ki67⁺ B cells displayed a 4-fold higher translation intensity and elongation rate than non cycling cells (Fig. 4D).

These results also reflected the level of S6 phosphorylation found in the different populations, P-S6 being mostly associated with a low level of Ki67 in PMA-stimulated T cells but not in PMA-stimulated B cells (Fig. S1). CFSE dilution experiments revealed similar differences, with the largest number of divisions observed respectively in PMA/ionomycin-stimulated T cells and LPS-activated B cells (Fig. S2A,B).

When SunRiSE was performed together with CFSE labeling, we could also unravel major differences in protein synthesis activity within the pool of proliferating cells (Fig. S2C,D). Both control and LPS-stimulated T cells had similar puromycin incorporation profiles, with only a small cell population, characterized by a low level of CFSE dilution, that displayed measurable translation activity. Conversely, PMA/ionomycin treatment strongly increased translation initiation and elongation in most T cells, although proportionally to the level of CFSE dilution exhibited (Fig. S2C), thus giving more resolute information than achieved through Ki67 staining alone. Our results indicate that after 2 days of PMA/ionomycin stimulation, protein synthesis activity is augmented proportionally to the number of cycles achieved. In LPS-stimulated B cells, protein synthesis was strongly enhanced in all cells; this was independently of the CFSE dilution level achieved and in full agreement with the results obtained with Ki67 staining after overnight stimulation (Fig. 4). Although, PMA induced a much weaker response in B cells than in T cells, the same relationship between CFSE dilution and increased protein synthesis activity was observed (Fig. S1D). Thus, cycling and translation regulation are highly interdependent and their relationship is influenced differently by the nature and duration of the activation stimuli. Importantly, PMA/ionomycin impacts the level of translation proportionally to the number of divisions achieved by the stimulated cells, a situation different from that observed with LPS stimulation and not entirely reflected by Ki67 expression.

At the cellular level, the embryonic development program is an extremely complex, and coordinated cell migration, proliferation and differentiation, displaying different needs for protein synthesis activity, is required in order for it to be successful. SunRiSE is therefore well suited to characterize such embryonic cells on the basis of their translational activity rather than uniquely on their phenotype. Embryonic day (E)12.5 mouse embryos were obtained and fetal livers were processed (van de Pavert and Vivier, 2016) to perform flow cytometry analysis using specific markers of blood vessel endothelial cells (BVECs; CD45⁻, CD31⁺), Lyve+ sinusoidal endothelial cells (LyVECs; CD45⁻, CD31^intermediate, LyVE1⁺), hematopoietic stem cells and leukocytes (CD45+), as well as putative neuronal cells (Neuro; CD45⁻, CD24⁺) (Fig. 5A). Large variations in levels of puro incorporation and elongation rates could be observed among the different gated cells populations (Fig. 5B,C), underlining the metabolic differences already in place at this stage of development. Vascular endothelial cells displayed the lowest intensity of protein synthesis, probably indicating their high level of differentiation (Fig. 5C). Conversely, the highest translation rates were found in CD45⁺ cell populations with high differentiation potential (Fig. 5C). The relatively homogeneous rates of elongation among the different cell types (with the exception of the BVECs), suggests that translation initiation or engagement is likely to be a major post-transcriptional regulatory step used during lymphoid organ development, as has already been proposed for many other biological systems. Initiation appears, however, to be tightly coordinated with ribosome processing and translation elongation, suggesting the existence of common regulatory pathways coordinating these two key mechanistic aspects of protein synthesis. Large variations in translation activity are also likely to represent major differences in ATP production and amino acid availability in the different embryonic cells tested and could give important clues about cell fate and differentiation potential, in relation to their metabolic status.

One of the major biochemical mechanisms that regulates translation elongation is the phosphorylation of elongation factor-2 (eEF2), the main, if not unique, substrate of eEF2K (Lazarus et al., 2017; Ryazanov, 2002). eEF2 phosphorylation is inhibitory and represses translation elongation in adverse growth conditions. Cellular stress, such as starvation, or accumulation of misfolded proteins in the endoplasmic reticulum, activate eEF2K, which in turn phosphorylates eEF2 and inhibits translation elongation through what is believed to be an arrest of codon translocation from ribosomal A to P sites (De Gassart and Martinon, 2017; Richter and Coller, 2015). SunRiSE is therefore an ideal technique to quantify, in different cells, the impact on translation elongation of eEF2K-mediated phosphorylation in normal or stressful growth conditions.

We first analyzed mouse embryonic fibroblasts (MEFs) isolated from eEF2K-deficient (−/−) or WT animals (Moore et al., 2015) for their capacity to phosphorylate eEF2 in normal conditions or upon tunicamycin-induced ER stress, and how this impacted on their translation elongation rates. Immunoblotting revealed the total absence of phosphorylated eEF2 (P-eEF2) in eEF2K^−/− MEFs, and their incapacity to increase P-eEF2 upon tunicamycin treatment (Fig. 6A). Flow cytometry confirmed this result, but also demonstrated that, in steady state, P-eEF2 levels are extremely heterogeneous across individual fibroblasts (Fig. 6A), and that SunRiSE offers a mean for dissecting the role of P-eEF2 with single-cell resolution.

When protein synthesis and translation elongation rate were measured (Fig. 6B), eEF2K^−/− MEFs displayed a clear increase in global translation activity compared to WT cells. In contrast, elongation speed was found to be identical among the two cell types, suggesting that permanent ablation of eEF2K increases translation engagement, rather than elongation speed per se. This was an unexpected result, given the reported activity of eEF2 in vitro (Herbert and Proud, 2006) and the presence of a low level of P-eEF2 at steady state in WT MEFs, which should have put a cap on elongation rate. When the same experiments were conducted with cells under persistent tunicamycin-induced ER-stress (16 h, Fig. 6C), we could observe that eEF2K deletion prevented both eEF2 phosphorylation and the downregulation of protein synthesis induced by tunicamycin. Both translation initiation and elongation activities were modestly impacted in eEF2K^−/− cells, while they were considerably decreased in stressed WT cells. Thus, eEF2K and P-eEF2 control ribosome engagement, potentially through interference with translation initiation or early elongation events. Alternatively, eEF2K ablation could increase global mRNA transcription and/or stabilization by decreasing the amount of ribosome run-offs from mRNA. The levels of eIF2α phosphorylation were increased in both cell types, however, to a lower level and with faster kinetics in eEF2K^−/− MEFs (Fig. 6A), suggesting a strong interplay among the eEF2K and PERK (also known as eIF2AK3) pathways. Decreased eIF2α phosphorylation could therefore contribute to the resistance of eEF2K^−/− MEFs to tunicamycin-induced translation arrest. With this in mind, we tested the respective contribution of eIF2α and eEF2 phosphorylation by using the pharmacological ISR inhibitor (ISRIB), which prevents inhibition of eIF2B and promotes translation initiation and stress granule assembly despite eIF2α phosphorylation (Sekine et al., 2015; Sidrauski et al., 2015). As expected, WT MEFs showed a rapid decrease in their translation activity upon a short treatment with thapsigargin, which was fully rescued in the presence of ISRIB (Fig. S3). In contrast, eEF2K^−/− MEFs again displayed a modest inhibition of protein synthesis during acute ER stress, which was partially rescued by ISRIB. These results further confirm that the moderate inhibition of translation observed in eEF2K^−/− MEFs is associated with reduced levels of eIF2α phosphorylation. eEF2K activity is therefore absolutely necessary to arrest ribosomal engagement upon ER stress, potentially through a strong interplay with the eIF2α phosphorylation pathway.

We next sought to investigate the consequences of the natural variations observed in P-eEF2 levels at the single-cell level. The differences in eEF2 phosphorylation were clearly dependent on eEF2K activity (Fig. 6A,D) and would have remained undetected by traditional bulk analysis, such as ribosome profiling. During SunRiSE analysis of WT MEFs, three subsets of cells were identified by the heterogeneous display of low, intermediate and high P-eEF2 levels (Fig. 6D,E), although eEF2 levels were found to be extremely homogeneous across the entire population (Fig. S4). To our surprise, and despite the high levels of P-eEF2 found in at least one of the subsets, no differences were observed in protein synthesis intensity nor in elongation rates among the different identified MEF subsets (Fig. 6E). These results suggest that in cell culture conditions, single cells display different eEF2K activity and P-eEF2 levels, but they demonstrate a great tolerance to these variations, probably by setting up adapted molecular thresholds of activation by stress. These stochastic thresholds allow each individual cell to commensurate its protective response by matching the inhibition of translation initiation (lower levels of phosphorylation) and translation elongation (higher levels of phosphorylation) to the intensity and duration of stress exposure.

Protein synthesis is one major regulatory layer that controls the final protein expression level. The translation machinery can rapidly react to environmental and metabolic cues and fine-tune the protein production issued from a pre-existing pool of mRNAs prior to major transcription reprogramming. Although mechanistic deciphering of the translation reaction was initiated in the 1950s, quantitative, or even qualitative, measurement of its efficacy in cells and tissues encountered many experimental and analytic hurdles. Our work provides an alternative to existing methods to measure translation elongation speed through multiparametric flow cytometry data acquisition. The SunRiSE method is complementary to ribosomal profiling, since it provides quantitative information about translation activity in complex populations and at single-cell resolution, however, without giving information at the individual mRNA level. Importantly, it also permits a precise comparison of protein synthesis activity in cells from different origins. SunRiSE can also discriminate between cells displaying heterogeneous levels of actively progressing ribosomes from those with stalled ribosomes, which creates a high level of false-positive hits during ribosomal profiling (Zhang et al., 2017). This caveat highlights the need of techniques like SunRiSE that can overcome this limitation and can be used to evaluate the homogeneity of a cell population prior to analysis by ribosome profiling.

SunRiSE can be implemented with minimal laboratory equipment and is adapted to human clinical studies, since it requires only small numbers of cells isolated from blood or biopsies. Our observations suggest that eEF2 phosphorylation by eEF2K has the capacity to regulate both ribosome engagement and translation elongation, the latter being particularly obvious during ER stress induction. The heterogeneity of P-eEF2 levels in a given cell population, although it has no clear impact on translation, raises a question on the activation of eEF2K in culture conditions and whether only a subset of cells increases P-eEF2 in response to stress, and not the entire cell population, as previously assumed from bulk immunoblot-based studies (Boyce et al., 2008). In addition to being dependent on Ca^2+/calmodulin, eEF2K activity is negatively regulated by the mammalian target of rapamycin complex 1 (mTORC1). mTORC1, among other functions, controls both mRNA translation and ribosome biogenesis (Kenney et al., 2014). Importantly, the Cdc2–cyclin B complex has also been shown to regulate eEF2K in a cell cycle- and amino acid availability-dependent manner. Cdc2 is activated early in mitosis, targeting Ser359 in eEF2K and reducing eEF2 phosphorylation in mitotic cells (Smith and Proud, 2008). Inactivation of eEF2K by Cdc2 might keep eEF2 active during mitosis and permit protein synthesis to proceed, while explaining the heterogeneity observed among the single MEFs in culture. We could reveal that eEF2K controls the global level of ribosome engagement, something that was previously solely attributed to translation initiation factors. We cannot exclude that eEF2K inactivation impacts on mRNA transcription and stability, leading to an increase in the level of protein synthesis independently of translation initiation, but our results clearly underline the roles of eEF2 in controlling various steps of protein synthesis. Increased phosphorylation of eEF2 was thought to be the consequence of P-eIF2α accumulation in response to ER stress, with a direct role for eEF2K in translation inhibition (Boyce et al., 2008). Our results confirm that eEF2K is required to achieve efficient inhibition of protein synthesis during ER stress, potentially by impacting on translation initiation, through its interaction with the PERK/eIF2α pathway. Of course, one could foresee the existence of other eEF2K targets that could interfere with the initiation machinery, although recent characterization of such targets has not yielded any obvious alternative candidates other than eEF2 itself (Lazarus et al., 2017).

SunRiSE can be applied in different types of studies, whether they involve embryonic development, differentiation, activation, ex vivo characterization of cell subsets or cell cycle progression. Interestingly, different growth-promoting stimuli can have very different consequences on cell cycle entry and associated protein synthesis activity. Single-cell resolution offers a clear perspective on such different biochemical outcomes, and a combination of SunRiSE with techniques allowing the visualization of different cell cycle phases (Bajar et al., 2016) will be invaluable for our understanding of translation control during the cell cycle. In addition, to giving information on cell proliferation, the capacity of SunRiSE to be performed in bulk using biopsy samples, will also provide an original visualization of the immune cell metabolism in different pathogenic landscapes, such as tumor beds. Indeed, many cancers have the ability to produce factors (e.g. indoleamine 2,3-dioxygenase, or arginase I and II) able to inhibit protein synthesis in T cells and favor immune tolerance towards the tumors (Munn and Mellor, 2016). A technique like SunRiSE would now allow us to obtain a snapshot of tumor-associated immune cells with regard to their metabolic activity, with a prognostic value on their fitness and the tolerogenic impact of their immediate tumor environment. This technique could be further adapted to different physiological situations by using other inhibitors of the first translation elongation cycle, like lactimidomycin, which could be used as an alternative to harringtonine, to increase translation arrest efficiency in specific cell types or tissues, and thus further improve the accuracy of the method (Garreau de Loubresse et al., 2014). SunRiSE complements existing approaches to comparatively monitor protein synthesis, and its use in combination with previously existing techniques will assist in deciphering the complexity of the post-transcriptional regulation mechanisms that underlie the regulation of protein expression in most biological systems.

WT and eEF2K^−/− MEFs were obtained from Alexey Ryazanov (Rutgers University, Piscataway, NJ). MEFs and mouse splenocytes were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 5% fetal calf serum (FCS) and 50 μM of 2-mercaptoethanol (MCCM; mouse cell culture medium) at 37°C under 5% CO₂. HeLa cells were cultured in DMEM with 10% FCS and 50 μM of 2-mercaptoethanol. To obtain splenocytes, 8-week-old wild type C57BL/6J mice were killed by cervical dislocation and splenectomized. Single-cells suspensions from the spleens were generated and cultured in MCCM. Cells were starved in Hank's balanced salt solution (HBSS) medium (Invitrogen, cat. no. 14185052) for 4 h, as previously described (Ingolia et al., 2012). Splenocytes were cultured in the absence (control) or presence of 0.1 μg/ml of lipopolysaccharide (LPS), or both PMA (5 ng/ml; Sigma, cat. no. P-8139) and ionomycin (500 ng/ml; Sigma, cat. no. I-0634) overnight. For ER stress induction, cells were treated with tunicamycin (100 ng/ml) for 16 h, thapsigargin for 3 h (500 nM), or with or without ISRIB (750 nM).

Cells from human blood, cell lines from in vitro culture or ex vivo mouse cells, were divided into 24-well plates (Fig. 1). Each well sample was treated with 2 μg/ml of harringtonine (Abcam, cat. ab141941), and after different time points, treated with puro (Sigma, cat. P7255) at 10 μg/ml (16.7 μM) (i.e. every 30 s for 7 min). Similar results were obtained by treating cells with harringtonine at different time points and adding puro (final concentration of 10 μg/ml) at the same time in all wells. After puro treatment, cells were washed in cold PBS and either stained (i.e. mouse splenocytes and fetal liver cells) with a combination of fluorophore-conjugated antibodies (see flow cytometry section) or directly fixed and permeabilized with BD Cytofix/Cytoperm™ (cat. no. 554714) followed by intracellular staining of puro using fluorescently labeled anti-puro antibodies for 1 h (1:1000, Clone 12D10, Merck, cat. no. MABE343).

Polysomal mRNA molecules were enriched by sucrose gradient fractionation following the original protocol (del Prete et al., 2007). Briefly, 5×10⁶ MEFs were lysed in 1 ml of polysome buffer [10 mM Tris-HCl pH 8, 140 mM NaCl, 1.5 mM MgCl₂, 0.5% NP40, 0.1 mg/ml cycloheximide, and 500 units/ml RNasin (Promega, Madison, WI)]. After 10 min on ice, lysates were quickly centrifuged (10,000 g for 10 s at 4°C) and the supernatant was resuspended in a stabilizing solution (0.2 mg/ml cycloheximide, 0.7 mg/ml heparin, 1 mM phenylmethanesulfonyl fluoride). After a quick centrifugation (12,000 g for 2 min at 4°C) to remove mitochondria and membrane debris, the resulting supernatant was layered on a 15% to 40% sucrose gradient. Gradients were then ultracentrifugated (35,000 g for 2 h at 4°C, SW41 rotor) and, after centrifugation, 20 fractions of 550 ml were collected, starting from the top of the gradient. All the fractions were then digested with proteinase K (200 mg/ml) in the presence of 1% SDS and 10 mM EDTA. RNA was then extracted with phenol/chloroform/isoamylalcohol (volume ratio 25:24:1) and precipitated with 2.5 volumes of 100% ethanol in the presence of 0.8 M lithium chloride. After precipitation, RNA was resuspended in RNase-free water. The correct fractionation of the polysomes was tested by measuring the ratio of different rRNA types on 1% denaturing agarose gel.

20 µg of NP-40 soluble material was separated by 4–12% gradient SDS-PAGE prior to immunoblotting and chemiluminescence detection (Pierce™). After performing SunRiSE on MEFs, puromycinilated peptide staining on immunoblots was performed using the monoclonal anti-puro antibody (1:5000; clone 12D10). Rabbit polyclonal antibodies anti-P-eEF2 and anti-eEF2 antibodies (Cell Signaling, cat. nos 2331 and 2332, both at 1:1000) were also used.

Cell suspensions were incubated with the antibody cocktail diluted in cold FACS buffer (PBS, 2% FCS, 2 mM EDTA) for 30 min at 4°C. Cells were fixed and permeabilized with cytofix/cytoperm buffer (BD Biosciences™), following the manufacturer's instructions, and stained for 1 h at 4°C with Alexa Fluor 488-conjugated anti-puro (puro-AF488) in PermWash. Flow cytometry was conducted by using BD LSR II and BD LSR Fortessa X-20 machines (BD Biosciences™), and data were analyzed with FlowJo (Tree Star™ or Cytobank). The antibodies used to stain cells were: puro-AF488 (1:100, Merck, cat. no. MABE343), phospho-S6-PE (1:600, Cell Signaling Technology, cat. no. #5316), Ki67 PE-eFluor 610 (1:400, eBioscience™, cat. no. 61-5698-82), anti-CD4-APC-eF780 (1:400, eBioscience™, cat. no. 47-0042-82), CD8-APC (1:600, eBioscience™, cat. no. 17-0081-83), CD80-PercPCy5.5 (1:600, Biolegend™, cat. no. 104722), anti-B220-BV421 (1:400, Biolegend™, cat. no. 103251), MHC-II-AF700 (1:800, eBioscience™, cat. no. 56-5321-82), and LIVE/DEAD™ Fixable Aqua Dead Cell Stain (Invitrogen™, cat. no. L34957). For proliferation assays, freshly isolated splenocyte suspensions were stained with CFSE. Briefly 1×10⁷ cells per ml were incubated for 20 min with CellTrace Violet in 1× PBS following the manufacturer’s instructions (Thermofisher Scientific, cat. no. 34557). Afterwards, the 1×10⁷ splenocytes were treated with LPS or PMA/ionomycin or not (control), as mentioned above. After 48 h, SunRiSE was performed separately, and cells were stained with antibodies against CD4, CD3, CD19, MHC-II, CD11c, puro and Ki67.

Wild-type C57BL/6 mice were purchased from Charles River and maintained in the animal facility of CIML under specific pathogen-free conditions. C57Bl/6 embryos at E12.5 were depleted of all organs and brain, and dissociated in Liberase (0.5 mg/ml, Roche) and DNaseI (0.2 mg/ml, Roche) in PBS for 15 min at 37°C while stirring constantly. Cell suspensions were washed with RPMI (Thermo Scientific), supplemented with 2% heat-inactivated FCS, 100 U/ml penicillin and 100 mg/ml streptomycin prior to immunostaining and flow cytometry analysis. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals, the French Ministry of Agriculture and of the European Union. Animals were housed in the CIML animal facilities accredited by the French Ministry of Agriculture to perform experiments on live mice. All animal experiments were approved by Direction Départementale des Services Vétérinaires des Bouches du Rhône (Approval number A13-543). All efforts were made to minimize animal suffering.

Statistical analysis was performed with GraphPad Prism software. For comparison of several conditions, we performed a one-way ANOVA, followed by a Tukey range test to assess the significance among pairs of conditions. When only two conditions were compared, we performed a Student's t-test or Welch t-test, according the validity of the homoscedasticity hypothesis. *P<0.05, **P<0.01, ***P<0.005.

We thank all CIML cytometry and Imaging core facilities for expert assistance and Dr A. Ryazanov for his kind gift of reagents. We acknowledge financial support from the Agence Nationale de la Recherche ANR (ANR-10-INBS-04-01) to the France Bio Imaging and the ImagImm CIML imaging core facility.