Cell Reports Report

Nutrient Deprivation Elicits a Transcriptional and Translational Inflammatory Response Coupled to Decreased Synthesis

Paulo A. Gameiro1,2 and Kevin Struhl1,3,* 1Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, MA 02115, USA 2Present address: Department of Molecular Neuroscience, UCL Institute of Neurology, 10-12 Russell Square House, London WC1B 5EH, UK 3Lead Contact *Correspondence: [email protected] https://doi.org/10.1016/j.celrep.2018.07.021

SUMMARY tionally repressed in an mTOR-dependent manner (Tang et al., 2001; Thoreen et al., 2012). Nutrient deprivation inhibits mRNA In the ISR pathway, diverse stresses lead to the phosphoryla- through mTOR and eIF2a signaling, but it is unclear tion of the translation initiation factor eIF2a by four kinases, each how the translational program is controlled to reflect of which is activated by specific stresses (Holcik and Sonenberg, the degree of a metabolic stress. In a model of breast 2005). Phosphorylation of eIF2a generally inhibits translation, but cellular transformation, various forms of nutrient it enhances the translation of specific mRNAs, such as that en- deprivation differentially affect the rate of protein coding the transcription factor ATF4 or its yeast counterpart Gcn4 (Harding et al., 2000; Holcik and Sonenberg, 2005). Trans- synthesis and its recovery over time. Genome-wide lational control of ATF4 is mediated by short upstream open translational profiling of -deprived cells re- reading frames (uORFs) that block translation of the downstream veals a rapid upregulation of mRNAs containing canonical ORF during normal growth but permit the scanning uORFs and downregulation of to initiate translation at the canonical ORF under mRNAs, which are followed by selective translation starved conditions (Vattem and Wek, 2004). Selective translation of cytokine and inflammatory mRNAs. Transcription of ATF4 is responsible for the transcriptional activation of adap- and translation of inflammatory and cytokine tation genes under amino acid deprivation (Harding et al., 2003). are stimulated in response to diverse metabolic As dividing cells rely differently on (Bauer et al., 2004; stresses and depend on eIF2a phosphorylation, Warburg et al., 1927), glutamine (DeBerardinis et al., 2007; with the extent of stimulation correlating with the Newsholme et al., 1985), glycine (Jain et al., 2012), serine (Mad- decrease in global protein synthesis. In accord with docks et al., 2013), and leucine (Sheen et al., 2011) to support specific metabolic functions, deprivation of some nutrients the inflammatory stimulus, glutamine deprivation may impose more severe constraints on mRNA translation stimulates the migration of transformed cells. Thus, than others. In addition, oncogenes can drive translation of spe- pro-inflammatory expression is coupled to cific mRNAs involved in cell growth, metabolism, invasion, and metabolic stress, and this can affect cancer cell metastasis (Truitt and Ruggero, 2016). However, there is limited behavior upon nutrient limitation. knowledge about how different metabolic stresses control mRNA translation and how the translational program is linked to the degree of metabolic stress. Here, we show that a variety INTRODUCTION of nutrient stresses lead to an inflammatory response, at both the transcriptional and translational levels, that is linked to inhibi- An adequate nutrient supply is essential for optimal mRNA trans- tion of global protein synthesis. lation, and eukaryotic cells have evolved nutrient-sensing path- ways that coordinate protein synthesis with nutritional status. RESULTS Nutrient deprivation inhibits global protein synthesis through modulation of the mechanistic target of rapamycin (mTOR) Nutrient Limitation Differentially Affects Nascent (Wullschleger et al., 2006) and integrated stress response (ISR) Protein Synthesis pathways (Holcik and Sonenberg, 2005). We analyzed the response to nutrient limitation in a model of The mTOR signaling pathway integrates nutritional and energy cellular transformation that involves an immortalized breast signals to regulate global translational rates via phosphorylation epithelial cell line (MCF10A) containing an endoplasmic reticu- of 4E-BP1 and S6K1, both of which are important for cap-depen- lum (ER)-Src fusion gene consisting of the v-Src oncogene and dent translation (Wullschleger et al., 2006). In addition, mTOR the ligand-binding domain of the estrogen receptor (Hirsch preferentially regulates mRNAs containing TOP motifs in their et al., 2010; Iliopoulos et al., 2009). Treatment of these 50 UTR, which often encode ribosomal that are transla- cells with tamoxifen activates Src, which generates a rapid

Cell Reports 24, 1415–1424, August 7, 2018 ª 2018 The Author(s). 1415 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). inflammatory stimulus causing an epigenetic switch from a non- (2 uORFs) inhibits translation through polyamine depletion. transformed state to a stable transformed state (Iliopoulos et al., ATF4 and SAT1 are the only uORF-containing mRNAs upregu- 2009). Using a fluorescent methionine analog, we quantified de lated in both transformed and non-transformed cells. novo protein synthesis in cells subjected to short-term (30 min) deprivation of glucose or amino acids with different bioenergetic Selective Translation of Cytokine and Inflammatory functions. With the exception of glycine and serine deprivation, mRNAs upon Glutamine Deprivation protein synthesis is inhibited to various extents depending on To investigate the gene expression program during the recovery the limiting nutrient (Figures 1A and 1B), with deprivation of from glutamine deprivation, we performed ribosome and mRNA branched-chain amino acids (BCAAs) having the strongest ef- profiling of cells cultured for 4 hr with and without glutamine fect. Protein synthesis is reduced in transformed cells, though (Table S1). In contrast to the 30-min response (Figure 2A), the the relative effects for the various nutrient stresses are similar fold changes in mRNA and ribosome occupancy are strongly in both cell types. correlated (Figure 2B), indicating a strong transcriptional contri- To analyze how cells recover from nutrient deprivation, we bution to the response following 4 hr of glutamine deprivation. In measured protein synthesis after 4 and 6 hr of nutrient depriva- addition, the cells are still metabolically stressed, as evidenced tion (Figure 1C). Protein synthesis returns to normal or near- by decreased translational efficiency of ribosomal proteins and normal levels upon deprivation of cysteine + cysteine or of increased translational efficiency of ATF4 (Figure 2B). glucose, presumably due to metabolic adaptations. In contrast, Interestingly, 34 and 56 mRNAs are translated more efficiently cells deprived of BCAAs and glutamine recover poorly, presum- in transformed and non-transformed cells (Figure 2B), respec- ably because they cannot synthesize these amino acids to suffi- tively, and these are enriched for genes involved in cytokine ac- cient levels. In all these cases, non-transformed and transformed tivity, inflammation, and angiogenesis (Figures 2B and 2C; Table cells behave similarly, though transformed cells seem to recover S2). Some cytokines and inflammatory mRNAs are induced tran- less well under glutamine deprivation (Figure 1C). The impor- scriptionally but do not show increased TE (Figure 2B; Table S2). tance of glutamine in cancer cell metabolism (Altman et al., As TE values tend to be correlated with cytosolic mRNA levels 2016) prompted us to investigate the translational program of (Larsson et al., 2010), we also performed an analysis of the partial transformed and non-transformed cells subjected to glutamine variance (APV) of ribosome occupancy and mRNA levels to iden- deprivation. tify differential translation with higher sensitivity (Larsson et al., 2010). The APV shows increased translation of cytokine mRNAs The Immediate Translational Response to Glutamine mainly in transformed cells (Figure 2C) and increased ribosome Deprivation: Downregulation of mRNAs Involved in occupancy (combined transcription and translation) of inflam- Translation and Upregulation of mRNAs Containing matory mRNAs in both cell types (Figure S1C; Table S2). As ex- uORFs pected from the induction of the ISR pathway, the mRNAs with We performed ribosome profiling (Ingolia et al., 2012), increased ribosome occupancy are also involved in the ER stress sequencing of ribosome-protected RNA, on transformed and response, which includes the transcription factors DDIT3 and non-transformed cells cultured with either complete or gluta- XBP1, the canonical targets CHAC1 and PPP1R15A, and amino- mine-depleted medium for 30 min. Using RibORF (Ji et al., acyl-tRNA synthetases (Figure 2B and S1C). Non-transformed 2015), we identified 5 million unique exon-mapped, ribosome cells also exhibit increased expression of cytokine and angio- footprints corresponding to 5,700 translated ORFs (Table genic mRNAs, but this response is less pronounced and,

S1). For each mRNA, we log2 divided the reads per kilobase of primarily, transcriptionally mediated (Figures 2B, S1C, and transcript per million mapped reads (RPKM) measured by ribo- S1D; Table S2). Thus, glutamine deprivation stimulates inflam- some profiling over the RPKM measured by RNA sequencing matory mRNAs, and this response is both transcriptionally and (RNA-seq) to obtain translation efficiencies (TEs), and then deter- translationally enhanced in transformed cells. mined differential TE induced by glutamine deprivation. We also Although MCF10A cells express glutamine synthetase (GLUL) analyzed the ribosome occupancy (RO) as determined by the and can proliferate without glutamine supplementation in the RPKM of ribosome-protected mRNA fragments. Approximately medium, GLUL is not regulated by glutamine deprivation (Fig- 140 mRNA transcripts have 1.5-fold or lower TE values, and ure 2D). In contrast, genes involved in glutamine catabolism, these translationally inhibited mRNAs are enriched for ribosomal such as glutaminase (GLS) and glutamate-oxaloacetic transam- and translation proteins (Figures 2A and S1A), and most of them inase 1 (GOT1), are induced (Figure 2D). In general, genes en- contain TOP motifs in their 50 UTRs (Figure S1B). This regulatory coding in the tricarboxylic acid (TCA) cycle are mostly response resembles that observed upon inhibiting TOR activity unaffected by glutamine deprivation. (Thoreen et al., 2012). Conversely, only 13 (non-transformed cells) or 22 (transformed Cytokine and Inflammatory Gene Expression Is Coupled cells) mRNAs are translationally stimulated after 30 min of gluta- to Translational Inhibition upon Nutrient Deprivation mine deprivation (Figure 2A), the most dramatic of which en- We performed similar experiments in cells subjected to 4 hr of codes ATF4. Like ATF4, some upregulated transcripts are linked deprivation of glucose, cysteine/cystine, and BCAA. As to translational control and have translated uORFs. PPP1R15B observed upon glutamine deprivation, cytokine and inflamma- (2 uORFs) is a phosphatase that dephosphorylates eIF2a, tory mRNAs are enhanced at both the transcriptional and TE AZIN1 (6 uORFs) is an antizyme inhibitor that stimulates levels upon deprivation of BCAA (Figures 3A and S2A; Table polyamine levels necessary for mRNA translation, and SAT1 S2). Cells deprived of glucose or cysteine/cystine-deprived cells

1416 Cell Reports 24, 1415–1424, August 7, 2018 A

200 μM 200 μM 200 μM 200 μM 200 μM 200 μM

+ Chx + Chx + CChxhx TAM- EtOH treated control

No Glucose No Glutamine Complete DMEM No Glucose No Glutamine Complete DMEM 30’ 30’ 30’ 30’

B EtOH control (non-Transformed) TAM-treated (Transformed) Complete DMEM ¥ * Chx (100ug/ml) * * No Glc * * ¥ No Q * * No Glc/Q * ¥ No Serine / Glycine No Cysteine ¥ No Cystine No Cysteine/Cystine * * No BCAA * * Torin1 (250nM) * 0% 20%* 40% 60% 80% 100% Protein Synthesis after 30 minutes (Relative pixel intensity from FITC channel)

C EtOH Complete DMEM TAM Complete DMEM EtOH No Glc TAM No Glc

100 μM EtOH No Q TAM No Q EtOH No Cysteine/Cystine TAM No Cysteine/Cystine EtOH No BCAA TAM No BCAA 120% y t ) i

s 100% s i n s e t e n h I

t 80% l n e y P = 0.055 x i S

p 60% n i e e v t i t o r a

l 40% P e from FITC channel R ( 20% 30min 4hr 6hrs 0% Duration of nutrient deprivation

Figure 1. Quantification of Protein Synthesis in Response to Metabolic Stresses (A) Nascent proteins containing the ‘‘clickable’’ methionine analog (AHA) visualized by fluorescence microscopy. (B) Quantification of nascent protein synthesis in tamoxifen (TAM)-treated and EtOH control cells under different metabolic conditions. CHX, cycloheximide. (C) Recovery of nascent protein synthesis over time. Error bars represent SD of three or more biological replicates, except for the 6-hr samples, and no cysteine or cystine at 30 min (two biological replicates). *p < 0.05, comparing nutrient-deprived to complete medium; Up < 0.05, comparing TAM-treated to EtOH control cells, as determined by t test. FITC, fluorescein isothiocyanate.

Cell Reports 24, 1415–1424, August 7, 2018 1417 A TAM NoQ 30min EtOH NoQ 30min Unregulated Unregulated RO UP RO UP TE UP TE UP RO DN RO DN TE DN TE DN Ribosomal TE DN Ribosomal TE DN Ribosome Occupancy) Ribosome Occupancy) Δ Δ Log2 ( Log2 (

Log2 (Δ mRNA levels) Log2 (Δ mRNA levels)

B TAMT NoQ 4hr EtOH NoQ 4hr Unregulated Unregulated RO UP RO UP TE UP TE UP Cytokine/Inflam/Ang RO UP Cytokine/Inflam/Ang RO UP Cytokine/Inflam/Ang TE UP Cytokine/Inflam/Angz TE UP ER-stress RO UP ER-stress RO UP RO DN RO DN TE DN TE DN Ribosomal TE DN Ribosomal TE DN Ribosome Occupancy) Δ Ribosome Occupancy) Δ Log2 ( Log2 (

Log2 ( Δ mRNA levels) Log2 (Δ mRNA levels)

C D TAM-treated – NoQ 4hrs 4.5 Analysis by Log2 TE fold change EtOH Ctrl * term ID description log10 p-value dispensability N genes 4.0 EtOH No Q GO:0001819 positive regulation of cytokine production -2.85 0.00 7 3.5 TAM Ctrl GO:0034613 cellular protein localization -2.30 0.00 11 TAM NoQ GO:0060429 epithelium development -2.37 0.00 10 3.0 GO:0071496 cellular response to external stimulus -3.09 0.00 7 2.5 * GO:0072593 reactive oxygen species metabolic process -2.85 0.00 6 GO:2001057 reactive nitrogen species metabolic process -2.75 0.01 4 2.0 Analysis by Anota2seq – “Translation Mode” 1.5 term ID description log10 p-value dispensability N genes (/ RPKM in Ctrl) (/ RPKM GO:0001819 positive regulation of cytokine production -3.21 0.00 6 1.0 * GO:0006094 gluconeogenesis -2.29 0.06 3 0.5

GO:0006984 ER-nucleus signaling pathway -2.98 0.16 3 Relative RibosomeOccupancy GO:1901214 regulation of neuron death -2.17 0.18 4 0.0 GO:0030155 regulation of cell adhesion -2.68 0.19 6 GLUL GLS GOT1 GO:0051270 regulation of cellular component movement -2.37 0.20 6

Figure 2. Regulation of mRNAs upon Glutamine Deprivation (A) Changes in mRNA and ribosome occupancy levels upon 30-min deprivation of glutamine relative to complete medium in transformed (left) and non-trans- formed (right) cells. (B) Similar analysis in cells grown for 4 hr in glutamine-deprived versus complete medium. (legend continued on next page)

1418 Cell Reports 24, 1415–1424, August 7, 2018 behave similarly, although to a lesser extent (Figures S2B and tion, torin1 markedly inhibits the phosphorylation of mTOR tar- S2C). In contrast to the 30-min dataset, the 22 inflammatory gets at 30 min and 4 hr (Figures 4B and S4A) but does not lead mRNAs with increased TE in all stresses (Figure S2D) do not to increased translation of inflammatory genes (Figure 4A). contain many uORFs (Figure S2E; Table S2). As also determined Indeed, the TE of some cytokine mRNAs actually decreases by the APV model, the overall inflammatory response by nutrient upon torin1 treatment, and the level of transcriptional induction depletion is induced both at the transcriptional level (Figure 3B) is considerably below what occurs upon nutrient stresses. and translational level (Figure 3C), and it is more pronounced in In accord with these observations, eIF2a phosphorylation in- the transformed state. Intriguingly, stresses causing the stron- creases upon all forms of nutrient deprivation tested here, but gest inhibition of protein synthesis after 30 min and weakest re- it is not affected by torin1 treatment (Figure 4B). In addition, covery at 4 and 6 hr (Figure 1, depletion of glutamine or BCAA) non-stressed transformed cells exhibit higher levels of eIF2a also cause the most pronounced increase in cytokine and in- phosphorylation and lower levels of protein synthesis than flammatory expression (Figures 3D and S2F). In addition, the non-transformed cells (Figure 4B). To test whether the inflamma- number of TE-downregulated mRNAs is higher in more ‘‘severe’’ tory response is linked to eIF2a phosphorylation, we measured stresses (Figures 2B, 3A, S2B, and S2C). cytokine levels in nutrient-deprived cells treated with ISRIB, a To examine more rigorously whether increased cytokine potent inhibitor of the integrated stress response pathway (Si- expression is linked to translational stress, we performed hierar- drauski et al., 2013). Cytokine protein levels are induced after chical clustering of all conditions based on ribosome occupancy glutamine depletion and in response to the other amino-acid (Figure 3E). Aggregation of all mRNAs into five k-means clusters stresses (Figure S4B). Treatment with the ISRIB inhibitor de- shows that cytokine and inflammatory mRNAs are highly co- creases the induced expression of IL8, IL6, and CCL20 in trans- regulated among the metabolic conditions with the highest upre- formed cells, and of IL8 in non-transformed cells, after 4 hr of gulation in BCAA and glutamine-deprived transformed cells and glutamine deprivation (Figure 4C). In accord with the literature, the lowest expression in non-transformed cells grown with com- ISRIB does not affect the levels of phosphorylated eIF2a (Si- plete, cysteine/cystine- or glucose-free DMEM (Figure 3F; Table drauski et al., 2013). Thus, translational repression by eIF2a S2, cluster 2). Conversely, the cluster with the strongest downre- phosphorylation is necessary for the enhanced cytokine expres- gulation in BCAA and glutamine-deprived transformed cells is sion under metabolic stress. strongly enriched for ribosomal proteins (Figure 3F; Table S2, cluster 5). Cluster 1 consists of immune-related mRNAs and is, Short-Term Nutrient Deprivation Increases Cell overall, upregulated in transformed cells (Iliopoulos et al., Migration 2009). In addition, recovery of protein synthesis 6 hr after the As secreted chemokines can induce the directed migration of stress is inversely correlated with the geometric mean in ribo- leukocytes and neighbor cells, we examined whether glutamine some occupancy of 12 cytokines (Figure 3G; Table S2), but deprivation affects cell migration. Migration of transformed cells not with aminoacyl-tRNA synthetases, whose transcription is through a basement membrane is enhanced toward conditioned induced by amino-acid deprivation (Figure S3A), or ATF4, the ca- medium (CM) from glutamine-deprived cells relative to fresh nonical example of selective mRNA translation (Figure S3B). glutamine-depleted medium, whereas complete CM inhibits Likewise, the ribosome occupancy of all downregulated mRNAs cell migration when compared to fresh complete medium (Fig- (Figure 3H) and 176 ribosomal factor mRNAs in ure 4D). Migration of non-transformed cells is not affected (GO): 0003735 (Figure S3C) is lower in the more severe stresses, through the basement membrane (Figure 4D), but their motility though the pattern is less evident for the latter, as ribosomal fac- is enhanced in a wound healing assay in CM from glutamine- tor mRNAs are proportionally more represented in the glucose or BCAA-deprived cells (Figure S4C). Transformed cells do not stress (Figure S3D). exhibit increased motility in the wound-healing assay (Fig- ure S4C). Thus, short-term glutamine deprivation can drive can- Enhanced Cytokine Expression Requires Translational cer cell behavior independently of proliferation. Inhibition by eIF2a Phosphorylation To examine whether inhibition of protein synthesis, per se, leads DISCUSSION to an inflammatory response, we performed ribosome profiling of transformed cells treated with torin1, an inhibitor of mTOR. As defined by the resulting gene expression patterns, depletion As expected (Thoreen et al., 2012), and in accord with what oc- of nutrients with different metabolic functions cause similar, curs upon nutrient depletion, torin1-treated cells show but not identical, inflammatory responses in transformed and decreased levels of protein synthesis (Figure 1B), with translation non-transformed cells at the transcriptional and translational of ribosomal proteins being particularly affected (Figure 4A). levels. It is likely that the transcriptional response involves the Nutrient deprivation mitigates the phosphorylation of 4EBP1 transcription factor NF-kB (nuclear factor kB), as glutamine and S6K1 at 30 min and 4 hr to different extents (Figures 4B deprivation leads to increased NF-kB activity (Hou et al., 2012; and S4A). However, in contrast to conditions of nutrient deple- Kim et al. 2014). Conversely, glutamine supplementation

(C) GO analysis of translationally regulated mRNAs as determined by log2 TE (top) and APV by anota2seq (bottom). (D) Relative RO of glutamine-catabolizing enzymes. Error bars represent SD of two biological replicates. *p < 0.05, comparing glutamine-deprived to complete medium, as determined by t test. See also Figure S1.

Cell Reports 24, 1415–1424, August 7, 2018 1419 1420 24 mRNAs (FDR B&H: 3.71e-3) B&H: (FDR 24 mRNAs transcription bypolymerase RNA II of Regulation – GO:0006357 E G B A elReports Cell EtOH

Ribosome occupancy of 90 Anota2seq – RO UP inall stresses UP RO – Anota2seq 12 cytokines (Geometric Mean Log2 (Δ Ribosome Occupancy)

RPKM) 111 100 150 200 250 50 24 0 TAM 112 %2%4%6%8%10 120% 100% 80% 60% 40% 20% 0%

No BCAA DN TE Ribosomal TE DN DN RO ER-stress RO UP TE UP Cytokine/Inflam/Ang UP RO Cytokine/Inflam/Ang TE UP UP RO Unregulated 4hr TAM NoBCAA PPP1R15A, DDIT3, PPP1R15A, DDIT3, NFE2L2, ASNS, ATF3, ATF4, 1.85e-10) B&H: (FDR 8 mRNAs response protein unfolded PERK-mediated GO:0036499 – IL1B AHR,FGF2,CTGF, KLF5, EPHA2,JAG1, HERPUD1, PTGS2, RUNX1, CYP1B1,FN1, HBEGF, WARS, XBP1, PPP1R15A, 1.85e-10) B&H: (FDR 21 mRNAs Angiogenesis – GO:0001525 4512,Ags ,2018 7, August 1415–1424, , Adjusted R²=0.76 Adjusted =0.73 R² P =0.035 P =0.041

TAM No Q CCL2, IL1RL1, TNFAIP3, CCL2, IL1RL1,TNFAIP3, IL6,DCD1LG2 CD274, NFKB2, PVR, PIK3R1, NFKB1, IL20, PTX3, IFNGR1, IFI16, TNFRSF21, IRF7, GCH1,BCL2A1,LCP1,DLL1, DAPK3, GBP1, BIRC2,BIRC3, ETS1 SOCS3, TNIP3, ICAM1, ABL2, PDE4D, IL24, EPRS, 9.71e-9) B&H: (FDR 34 mRNAs response Immune GO:0006955 – IL8 LIF No Glc Log2 ( , SAT1, CYR61, NFE2L2, CYR61, , SAT1, Protein synthesis activity at 6 activity synthesishours Protein Leu/Iso/Val (Pixel intensity - FITC channel) FITC - (Pixel intensity

No Ctrl DMEM Δ , CASP4, SHMT2, SHMT2, CASP4, , Leu/Iso/Val

No Cys/Cys levels) mRNA No IL1A, No Q No Cys/Cys

EtOH Ctrl DMEM No Glc No Q

No Glc RCAN1 LDHAL6A, LMO4, PER1,ERRFI1, ETV5, RSRC2, FTH1, PNRC2, DCTN4, JAG1, KRT34, AZIN1, ZBTB38, IER2, PMAIP1, C19orf48, PCTP, BCL10, SAT1, BTG3, DYNC1H1, ITPRIP, KRT16, STX1A, C

No Glc No BCAA No Cys-Cys Anota2seq – Translation UP inall stressesUP Translation – Anota2seq Tamoxifen Ethanol

NoQ PPP1R15B, CCL2, ATF4, DDIT3 4mRNAs (FDRB&H:3.37e-04) unfolded proteinresponse PERK-mediated GO:0036499 – PDE4D, ABL2, FN1, ERRFI1, PER1, CD274 11 (FDRB&H:3.37e-04) mRNAs Regulationofcytokine production GO:0001817 – No Cys-Cys Z-score Ctrl DMEM Ctrl Ctrl DMEMCtrl -2 -1 0 1 2 EtOH 25 ATF4, FN1, GADD45B, NUPR1 GADD45B, FN1, ATF4, HIST1H2BK, HIST1H2AC, HIST1H1C, DDIT3, 817 F TAM 5 4 3 2 1 , CCL2, BCL10, ATF4, BCL10,IL6, ATF4, , CCL2, DDIT3 Log2 (Δ Ribosome Occupancy) No BCAA 7 mRNAs (FDRB&H:5.67e GO:0071396 GO:0071396 FHL2, CCL2, TNC, IL6, ABL2, R3C1 PDE4D, FHL2, CCL2,TNC, H

TAM No Q RibosomalDN TE TE DN T DN RO UP RO ER-stress TEUP Cytokine/Inflam/Ang UP RO Cytokine/Inflam/Ang T TE UP UP RO Unregulated 4hr EtOH NoBCAA – Log2 FC Ribosome Occupancy No Cys/Cys Cellular response tolipid -2 -1 0 1 All TEDownregulated

Ctrl DMEM - 4) No Glc P Log2 ( = 0.1 No BCAA P D ATF4, ASNS, GARS ASNS, ATF4, CARS, PSAT1, AARS, PHGDH, 2.05e-5) B&H: (FDR 7 mRNAs Cellular aminoacid GO:0006520 – = 0.5 ZNF644, SEL1L ZNF644, ERO1LB, RAPH1, RHOF, ALDH1A3, OGT, TM4SF19, IVNS1ABP, PTHLH, CRELD2, SDF2L1, HSPA5. ALDH1L2, PSPH LC7A5, HMOX1, GPT2,

NoQ Δ P mRNA levels) mRNA EtOH No Cys/Cys = 8e-182

No Cys/Cys metabolic process

mRNAs Ctrl DMEM

P No Glc = 4e-86 5 No Glc (1,124 mRNAs) (1,124 P CTGF, CXCL2 CTGF, DUSP5, GADD45A, GO:0003735 -> Structural constituent -> constituent Structural GO:0003735 0 = 3e-58 1 GO:0006955 -> (54) response Immune (11) activity -> cytokine GO:0005125 13 SERPINB2, TAM 17 0 No Leu/Iso/Val 1 P 0 4 = 3e-25 1 CYR61 28 2 0 28 52 TBX3, PPARD 1.03e-4) B&H: (FDR 4 mRNAs myoblast differentiation of regulation Negative GO:0045662 – TAM NoBCAA TAM NoQ TAM NoGlc TAM NoCys NoBCAA EtOH NoQ EtOH NoGlc EtOH EtOH NoCys lgn nnx page) next on (legend frbsm (20) of ribosome PTGS2, DDIT3, EPHA2, IL6, JAG1,IL8 EPHA2, DDIT3, PTGS2, IL1A, PPP1R15A, TNFAIP3, LIF, XBP1, 1.16 B&H: e-6) (FDR 10 mRNAs Angiogenesis – GO:0001944 No Q BIRC3 NFE2L2, TXNRD1, 1.19e-3) B&H: (FDR 9 mRNAs species metabolicprocess oxygen Reactive – GO:0072593 BIRC3, CCL2, KLF4 ICAM1, ETS1, CD274, GBP1, EPRS, NFE2L2, NFKB1, NFKB2, 3.35e-3) B&H: (FDR 11 mRNAs cytokine to Response – GO:0034097 , DLL1 , MMP3, , DDIT3 KLF4 PTX3 , PMAIP1 , , ICAM1, prevents NF-kB activity and cytokine expression (Chen et al., contrast, ISRIB reverses the effects of eIF2a phosphorylation 2008; Singleton et al., 2005). Glutamine deprivation also triggers (Sidrauski et al., 2013) and diminishes enhanced cytokine co-localization of autophagosomes, lysosomes, and the Golgi expression in glutamine-deprived cells, suggesting that inhibi- apparatus into a subcellular structure whose integrity is essential tion of translation initiation by eIF2a phosphorylation triggers for IL-8 secretion, a process regulated by mTOR and JNK ki- the inflammatory response. eIF2a phosphorylation leads to acti- nases (Shanware et al., 2014). vation of NF-kB due to decreased levels of the IkB inhibitor Translational stimulation of cytokine and inflammatory mRNAs arising from the combination of its short half-life and the reduced upon nutrient or other stresses has not been described previ- translation (Deng et al., 2004; Jiang et al., 2003). This mechanism ously. However, based on published ribosome profiling data, may explain why the transcriptional induction of cytokines does we noticed that cells treated with thapsigargin (Reid et al., not occur rapidly upon nutrient deprivation but, rather, takes 2014), a drug that induces ER stress, show increased translation several hours. of inflammatory mRNAs. Our observation that nutrient (and, The inflammatory response arising upon nutrient depletion more generally, ER) stress affects inflammatory genes at both also occurs at the translation level. Though unknown, the mech- the transcriptional and translational levels underscores the bio- anistic basis of this translational control likely involves eIF2a logical importance of this response. phosphorylation and may act in a uORF-independent manner, The relationship between the inflammatory response and as 15 of the 22 TE upregulated inflammatory mRNAs do not metabolic stress is relevant to transformed cells. First, as contain uORFs. Cytokine mRNAs can be stabilized by the NF-kB and STAT3 are critical for inflammation and tumor devel- mitogen-activated protein (MAP) kinase/p38 pathway via au- opment (Grivennikov et al., 2010), increased inflammation in rich element (ARE)-binding proteins that interact with sequences transformed cells is likely to have a greater impact in nutrient- within the 30 UTR (Hoffmann et al., 2002), so ARE- or other RNA- deprived tumors than in normal tissues. As the concentrations binding proteins might interact with translation factors and/or of glucose and glutamine are lower in tumors than in normal tis- to affect the translation of cytokine mRNAs. Whatever sues (Cairns et al., 2011; Warburg et al., 1927), pro-inflammatory the precise mechanisms involved, the opposing effects of translation may be enhanced in a tumor microenvironment. Sec- eIF2a phosphorylation on general translation and inflammatory ond, increased expression of cytokines in a nutrient-limited envi- response provides a mechanism to coordinate these processes ronment would be an autocrine mechanism for tumor cells to in response to a wide range of nutritional states. invade and metastasize. As glutamine depletion, per se, nega- tively modulates cell migration, cytokine production under this EXPERIMENTAL PROCEDURES condition might permit cancer cells to migrate toward a region Cell Culture with higher glutamine concentration. In addition, breast cancer MCF10A-ER-Src cells were grown at 37C with 5% CO in phenol red-free cells selected to grow in the absence of glutamine become 2 DMEM-F12 medium containing 5% charcoal-stripped horse serum, peni- more metastatic and tumorigenic and overexpress cyclooxyge- cillin-streptomycin (13), epidermal growth factor (EGF) (200 ng/mL), hydrocor- nase-2 (COX-2 or PTGS2) (Singh et al., 2012), one of the more tisone (0.5 mg/mL, final), cholera toxin (100 ng/mL), and insulin (10 mg/mL), as efficiently translated mRNAs in glutamine-deprived cells in our described previously (Iliopoulos et al., 2009). Cells were transformed by treat- experiments. ment with 1 mM 4-hydroxy-tamoxifen for 24 hr, and ethanol was used as The inverse relationship between the inflammatory response vehicle control in non-transformed cells. For metabolic experiments, cells were grown in the same medium but lacking nutrients as indicated. and general protein synthesis suggests a mechanistic connec- tion between these two processes. Inhibition of mTOR activity in- Measurement of Nascent Protein Synthesis a duces eIF2 phosphorylation in a context-dependent manner Protein synthesis was assessed by treating cells with the ‘‘clickable’’ methio- (Cherkasova and Hinnebusch, 2003; Gandin et al., 2016; nine analog L-azidohomoalanine (AHA) for 30 min, 4 hr, or 6 hr and detecting Thoreen et al., 2009; Wengrod et al., 2015). Inhibition of mTOR nascent proteins by fluorescence microscopy using the 488-nm laser channel does not explain our findings, because torin1 decreases protein according to the manufacturer’s instructions (Thermo Fisher Scientific, catalog synthesis to an extent comparable to that of many conditions of no. C10289). Details are given in the Supplemental Experimental Procedures. nutrient depletion but does not lead to a similar inflammatory Ribosome Profiling and RNA-Seq response. In fact, torin1-treated cells exhibit a lower translation Details pertaining high-throughput sequencing data acquisition and analyses of various cytokine and inflammatory mRNAs, so basal mTOR are given in the Supplemental Experimental Procedures. Ribosome profiling activity is probably required for the inflammatory response. In and RNA-seq were performed as described previously (Ingolia et al., 2012).

Figure 3. Regulation of mRNAs upon 4 hr of Metabolic Stress Conditions (A) Changes in mRNA and ribosome occupancy levels upon 4-hr deprivation of BCAAs relative to complete medium in transformed (left) and non-transformed (right) cells. Ang, angiogenesis; DN, down-regulated; RO, ribosome occupancy; UP, up-regulated. (B and C) Venn diagram showing the anota2seq analysis of mRNAs with increased (B) transcriptional and (C) translational activity in all metabolic stresses in transformed versus non-transformed cells. (D) Venn diagram showing the mRNAs with increased ribosome occupancy in each metabolic stress condition in transformed cells; cytokines are highlighted in red in non-inflammatory GO terms. (E and F) Shown here: (E) hierarchical and (F) k-means clustering of ribosome occupancy across metabolic conditions. (G) Correlation between protein synthesis and ribosome occupancy of regulated cytokines among nutrient-deprived conditions. (H) Ribosome occupancy of all TE-downregulated mRNAs in each metabolic condition; p values are derived from the Wilcoxon test. See also Figures S2 and S3.

Cell Reports 24, 1415–1424, August 7, 2018 1421 A TAM Torin1 4hr Unregulated RO UP RO DN ) TE DN y

c Ribosomal TE DN

n Cytokine mRNAs a p u c c O e m o s o b i R Δ Log2 (

Log2 (Δ mRNA levels)

B 30 min 30 min 30 min p-eIF2α p-eIF2α EtOH TAM p-S6K1 p-S6K1 p-eIF2α TAM p-4EBP1 EtOH p-4EBP1 eIF2α β-Actin β-Actin No Q No Q No Q No Q Torin1 Torin1 Ctrl DMEM No Gluc Ctrl DMEM No Gluc No BCAA No BCAA Ctrl DMEM Ctrl DMEM No Cys/Cys No Cys/Cys

Transformed (TAM) C D 4 hours Ctrl CM Ctrl No Q CM No Q

EtOH TAM p-eIF2α

IL6 CM = conditioned medium CCL20 TAM EtOH 2.5 P = 0.012 IL8 n.s. 2.0 α-Actin 1.5 P = 0.0032 Ctrl Ctrl No Q No Q 1.0

Ctrl + ISRIB 0.5 Ctrl + ISRIB Ctrl + Relative Cell Migration No Q + No Q + ISRIB No Q + No Q + ISRIB

(/ respective 'fresh medium') 'fresh (/ respective 0.0 CM Ctrl No FBS CM No Q

Figure 4. Cytokine Expression under Inhibition of mTOR and eIF2a Signaling and Migration of Glutamine-Deprived MCF10A-ER-Src Cells (A) Changes in mRNA and ribosome occupancy levels in transformed cells subjected to 4-hr treatment with torin1 (500 nM) relative to complete medium. (B) Immunoblot analysis of mTOR-phosphorylated proteins and eIF2a phosphorylation after 30 min of nutrient deprivation or Torin1 treatment (500 nM). (legend continued on next page)

1422 Cell Reports 24, 1415–1424, August 7, 2018 For determination of gene expression levels of protein-coding genes, we Chen, G., Shi, J., Qi, M., Yin, H., and Hang, C. (2008). Glutamine decreases in- removed refSeq-defined coding regions overlapping with uORFs defined in testinal nuclear factor kappa B activity and pro-inflammatory cytokine expres- Ji et al. (2015), 15 amino acids downstream of the start codons and 5 amino sion after traumatic brain injury in rats. Inflamm. Res. 57, 57–64. acids upstream of the stop codons. The gene expression levels were calcu- Cherkasova, V.A., and Hinnebusch, A.G. (2003). Translational control by TOR lated as RPKMs in the remaining coding regions, and we considered only and TAP42 through dephosphorylation of eIF2a kinase GCN2. Genes Dev. 17, mRNAs expressed at RPKM > 10 in at least one of the metabolic conditions 859–872. (Table S1). Differential translation was assessed by (1) TE, and (2) partial vari- DeBerardinis, R.J., Mancuso, A., Daikhin, E., Nissim, I., Yudkoff, M., Wehrli, S., ance of RO and RNA levels (Larsson et al., 2010). and Thompson, C.B. (2007). Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and Statistical Analyses nucleotide synthesis. Proc. Natl. Acad. Sci. USA 104, 19345–19350. Statistical analyses and visualizations were performed in R or Excel. The hier- archical and k-means clustering was performed using the pheatmap package Deng, J., Lu, P.D., Zhang, Y., Scheuner, D., Kaufman, R.J., Sonenberg, N., with rlog-transformed ribosome occupancy. The association between protein Harding, H.P., and Ron, D. (2004). Translational repression mediates activation synthesis and ribosome occupancy of cytokine mRNAs was tested using the of nuclear factor kappa B by phosphorylated translation initiation factor 2. Mol. Pearson correlation coefficient. Other statistical significance was assessed Cell. Biol. 24, 10161–10168. using either the Wilcoxon rank-sum test or t test, as indicated. Gandin, V., Masvidal, L., Cargnello, M., Gyenis, L., McLaughlan, S., Cai, Y., Tenkerian, C., Morita, M., Balanathan, P., Jean-Jean, O., et al. (2016). DATA AND SOFTWARE AVAILABILITY mTORC1 and CK2 coordinate ternary and eIF4F complex assembly. Nat. Commun. 7, 11127. The accession number for the sequencing data reported in this paper is GEO: Grivennikov, S.I., Greten, F.R., and Karin, M. (2010). Immunity, inflammation, GSE114794. and cancer. Cell 140, 883–899. Harding, H.P., Novoa, I., Zhang, Y., Zeng, H., Wek, R., Schapira, M., and Ron, SUPPLEMENTAL INFORMATION D. (2000). Regulated translation initiation controls stress-induced gene expression in mammalian cells. Mol. Cell 6, 1099–1108. Supplemental Information includes Supplemental Experimental Procedures, Harding, H.P., Zhang, Y., Zeng, H., Novoa, I., Lu, P.D., Calfon, M., Sadri, N., four figures, and two tables and can be found with this article online at Yun, C., Popko, B., Paules, R., et al. (2003). An integrated stress response reg- https://doi.org/10.1016/j.celrep.2018.07.021. ulates amino acid metabolism and resistance to oxidative stress. Mol. Cell 11, 619–633. ACKNOWLEDGMENTS Hirsch, H.A., Iliopoulos, D., Joshi, A., Zhang, Y., Jaeger, S.A., Bulyk, M., Tsichlis, P.N., Liu, X.S., and Struhl, K. (2010). A transcriptional signature and We would like to thank Ji Zhe for help and expert advice with the RibORF pipe- common gene networks link cancer with lipid metabolism and diverse human line and gene expression profile. This work was supported by grant CA 107486 diseases. Cancer Cell 17, 348–361. to K.S. from the NIH. Hoffmann, E., Dittrich-Breiholz, O., Holtmann, H., and Kracht, M. (2002). Mul- AUTHOR CONTRIBUTIONS tiple control of interleukin-8 gene expression. J. Leukoc. Biol. 72, 847–855. Holcik, M., and Sonenberg, N. (2005). Translational control in stress and Conceptualization and Resources, P.A.G. and K.S.; Methodology, Investiga- apoptosis. Nat. Rev. Mol. Cell Biol. 6, 318–327. tion, Validation, Formal Analysis, Data Curation, and Visualization, P.A.G.; Hou, Y.-C., Chiu, W.-C., Yeh, C.-L., and Yeh, S.-L. (2012). Glutamine modu- Writing – Original Draft, P.A.G.; Writing – Review & Editing, P.A.G. and K.S.; Su- lates lipopolysaccharide-induced activation of NF-kB via the Akt/mTOR pervision, Project Administration, and Funding Acquisition, K.S. pathway in lung epithelial cells. Am. J. Physiol. Lung Cell. Mol. Physiol. 302, L174–L183. DECLARATION OF INTERESTS Iliopoulos, D., Hirsch, H.A., and Struhl, K. (2009). An epigenetic switch involving NF-kappaB, Lin28, Let-7 microRNA, and IL6 links inflammation to The authors declare no competing interests. cell transformation. Cell 139, 693–706. Ingolia, N.T., Brar, G.A., Rouskin, S., McGeachy, A.M., and Weissman, J.S. Received: November 22, 2017 (2012). The ribosome profiling strategy for monitoring translation in vivo by Revised: May 22, 2018 deep sequencing of ribosome-protected mRNA fragments. Nat. Protoc. 7, Accepted: July 4, 2018 1534–1550. Published: August 7, 2018 Jain, M., Nilsson, R., Sharma, S., Madhusudhan, N., Kitami, T., Souza, A.L., REFERENCES Kafri, R., Kirschner, M.W., Clish, C.B., and Mootha, V.K. (2012). Metabolite profiling identifies a key role for glycine in rapid cancer cell proliferation. Sci- Altman, B.J., Stine, Z.E., and Dang, C.V. (2016). From Krebs to clinic: gluta- ence 336, 1040–1044. mine metabolism to cancer therapy. Nat. Rev. Cancer 16, 773. Ji, Z., Song, R., Regev, A., and Struhl, K. (2015). Many lncRNAs, 5’UTRs, and Bauer, D.E., Harris, M.H., Plas, D.R., Lum, J.J., Hammerman, P.S., Rathmell, pseudogenes are translated and some are likely to express functional proteins. J.C., Riley, J.L., and Thompson, C.B. (2004). Cytokine stimulation of aerobic eLife 4, e08890. glycolysis in hematopoietic cells exceeds proliferative demand. FASEB J. Jiang, H.-Y., Wek, S.A., McGrath, B.C., Scheuner, D., Kaufman, R.J., Cavener, 18, 1303–1305. D.R., and Wek, R.C. (2003). Phosphorylation of the alpha subunit of eukaryotic Cairns, R.A., Harris, I.S., and Mak, T.W. (2011). Regulation of cancer cell meta- initiation factor 2 is required for activation of NF-kappaB in response to diverse bolism. Nat. Rev. Cancer 11, 85–95. cellular stresses. Mol. Cell. Biol. 23, 5651–5663.

(C) Immunoblot analysis of cytokine protein levels after 4 hr of glutamine deprivation and/or ISRIB treatment (250 nM). (D) Microscopy images and quantification of cell migration through a basement toward a chamber with different test media. Error bars represent SD of three or more biological replicates. The p values are derived from the t test. n.s., not significant. See also Figure S4.

Cell Reports 24, 1415–1424, August 7, 2018 1423 Kim, M.-H., Kim, A., Yu, J.H., Lim, J.W., and Kim, H. (2014). Glutamine depri- Singleton, K.D., Beckey, V.E., and Wischmeyer, P.E. (2005). Glutamine pre- vation induces interleukin-8 expression in ataxia telangiectasia fibroblasts. vents activation of NF-kB and stress kinase pathways, attenuates inflamma- Inflamm. Res. 63, 347–356. tory cytokine release, and prevents acute respiratory distress syndrome Larsson, O., Sonenberg, N., and Nadon, R. (2010). Identification of (ARDS) following sepsis. Shock 24, 583–589. differential translation in genome wide studies. Proc. Natl. Acad. Sci. USA Tang, H., Hornstein, E., Stolovich, M., Levy, G., Livingstone, M., Templeton, D., 107, 21487–21492. Avruch, J., and Meyuhas, O. (2001). Amino acid-induced translation of TOP mRNAs is fully dependent on phosphatidylinositol 3-kinase-mediated Maddocks, O.D.K., Berkers, C.R., Mason, S.M., Zheng, L., Blyth, K., Gottlieb, signaling, is partially inhibited by rapamycin, and is independent of S6K1 E., and Vousden, K.H. (2013). Serine starvation induces stress and p53-depen- and rpS6 phosphorylation. Mol. Cell. Biol. 21, 8671–8683. dent metabolic remodelling in cancer cells. Nature 493, 542–546. Thoreen, C.C., Kang, S.A., Chang, J.W., Liu, Q., Zhang, J., Gao, Y., Reichling, Newsholme, E.A., Crabtree, B., and Ardawi, M.S. (1985). The role of high rates L.J., Sim, T., Sabatini, D.M., and Gray, N.S. (2009). An ATP-competitive of glycolysis and glutamine utilization in rapidly dividing cells. Biosci. Rep. 5, mammalian target of rapamycin inhibitor reveals rapamycin-resistant func- 393–400. tions of mTORC1. J. Biol. Chem. 284, 8023–8032. Reid, D.W., Chen, Q., Tay, A.S.-L., Shenolikar, S., and Nicchitta, C.V. (2014). Thoreen, C.C., Chantranupong, L., Keys, H.R., Wang, T., Gray, N.S., and Sa- The unfolded protein response triggers selective mRNA release from the endo- batini, D.M. (2012). A unifying model for mTORC1-mediated regulation of plasmic reticulum. Cell 158, 1362–1374. mRNA translation. Nature 485, 109–113. Shanware, N.P., Bray, K., Eng, C.H., Wang, F., Follettie, M., Myers, J., Fantin, Truitt, M.L., and Ruggero, D. (2016). New frontiers in translational control of the V.R., and Abraham, R.T. (2014). Glutamine deprivation stimulates mTOR-JNK- cancer genome. Nat. Rev. Cancer 16, 288–304. dependent chemokine secretion. Nat. Commun. 5, 4900. Vattem, K.M., and Wek, R.C. (2004). Reinitiation involving upstream ORFs reg- Sheen, J.-H., Zoncu, R., Kim, D., and Sabatini, D.M. (2011). Defective regula- ulates ATF4 mRNA translation in mammalian cells. Proc. Natl. Acad. Sci. USA tion of autophagy upon leucine deprivation reveals a targetable liability of hu- 101, 11269–11274. man melanoma cells in vitro and in vivo. Cancer Cell 19, 613–628. Warburg, O., Wind, F., and Negelein, E. (1927). The metabolism of tumors in Sidrauski, C., Acosta-Alvear, D., Khoutorsky, A., Vedantham, P., Hearn, B.R., the body. J. Gen. Physiol. 8, 519–530. Li, H., Gamache, K., Gallagher, C.M., Ang, K.K.H., Wilson, C., et al. (2013). Wengrod, J., Wang, D., Weiss, S., Zhong, H., Osman, I., and Gardner, L.B. Pharmacological brake-release of mRNA translation enhances cognitive (2015). Phosphorylation of eIF2a triggered by mTORC1 inhibition and PP6C memory. eLife 2, e00498. activation is required for autophagy and is aberrant in PP6C-mutated mela- Singh, B., Tai, K., Madan, S., Raythatha, M.R., Cady, A.M., Braunlin, M., Irving, noma. Sci. Signal. 8, ra27. L.R., Bajaj, A., and Lucci, A. (2012). Selection of metastatic breast cancer cells Wullschleger, S., Loewith, R., and Hall, M.N. (2006). TOR signaling in growth based on adaptability of their metabolic state. PLoS ONE 7, e36510. and metabolism. Cell 124, 471–484.

1424 Cell Reports 24, 1415–1424, August 7, 2018 Cell Reports, Volume 24

Supplemental Information

Nutrient Deprivation Elicits a Transcriptional and Translational Inflammatory Response Coupled to Decreased Protein Synthesis

Paulo A. Gameiro and Kevin Struhl Supplemental Experimental Procedures

Measurement of nascent protein synthesis. Protein synthesis was assessed using the ‘clickable’ methionine analog L-azidohomoalanine (AHA) that is incorporated into nascent proteins and detected by fluorescence microscopy according to the manufacturer’s instructions (ThermoFisher scientific cat. No. C10289). Non-transformed and transformed cells were rinsed twice with warm PBS, and incubated with methionine-free medium (and lacking other nutrients as indicated) containing AHA for

o 30 minutes, 4 hours, or 6 hours at 37 C in 5% CO2. Cells were washed with PBS, fixed with 3.7 % formaldehyde, incubated with the Alexa Fluor 488 alkyne, and AHA-containing proteins were visualized using the 488nm laser channel. DAPI staining was performed to determine cell number. Quantification of protein synthesis was performed with MetaXpress Analysis Software (Molecular Devices) using custom cell scoring that consider the nuclei number, cell size and background correction for determination of average intensity of Alexa fluorescence per cell. We used ImageJ to set an equal minimum and maximum pixel intensity value in all RGB images for visualization of normalized fluorescence (Figure 1A). Three or more biological replicates were performed (6-12 imaging sites per replicate at 10X magnification) except for the 6 hours samples, no cysteine and no cystine condition at 30 minutes (two replicates).

Ribosome profiling and RNA sequencing. Non-transformed and transformed MCF10A-ER-Src cells were cultured with either complete or nutrient-deprived DMEM as indicated. For ribosome profiling, cells were washed with ice-cold PBS containing 100µg/ml cycloheximide, and flash-frozen with detergent lysis [20mM Tris-HCl (pH 7.4), 150mM NaCl, 5mM MgCl2, 1mM DTT, 1% (vol/vol) triton X-100] containing 25 U/ml Turbo DNase I and 100µg/ml cycloheximide. Ribosome footprinting was performed by adding 8.75 µl of Rnase I (100 U/ml) to 350 µl of DNase I-treated lysates for 45 minutes at room temperature followed by addition of 12 µl of SUPERase·In inhibitor to stop the reaction. The Rna se I-digested lysates were added to 0.90 ml of 1M sucrose cushion, ribosomes were pelleted by centrifugation at 55,000 rpm at 4C overnight, resuspended in 700 µl of Qiazol reagent, and the RNA isolated using the miRNeasy Kit (Qiagen, cat. no. 217004). Ribosome-protected fragments were PAGE purified on a 15% TBE-Urea gel (ThermoFisher scientific, cat. No. EC68852BOX), 3’end ligated to a pre-adenylated universal miRNA linker (5′-rAppCTGTAGGCACCATCAAT–NH2-3′, NEB, cat. no. S1315S), converted into cDNA using Superscript III reverse transcriptase (ThermoFisher scientific, cat. no. 18080-093), circularized, depleted for ribosomal RNA using biotinylated probes, and PCR amplified using Illumina TruSeq indexed primers as previously described (Ingolia et al., 2009, 2012). For RNA-seq, total RNA was extracted using the miRNeasy kit

(Qiagen cat No. 217004), enriched for poly(A) RNA via three rounds of Oligo d(T)25 magnetic bead purification (NEB, cat. No. S1419S), fragmented with 10mM ZnCl2 in 10mM Tris-HCl (pH 7.0) for 5 minutes at 80ºC, PAGE purified on a 15% TBE-Urea gel and processed equally as the ribosome protected fragments for Illumina TruSeq library preparation. The ribosome profiling and RNA-seq experiments were performed in duplicate with the exception of RNA-seq at 30 minutes (single experiments), ribosome profiling in cysteine/cystine-deprived EtOH cells (triplicate experiments), and the Torin1 dataset (single experiment). To process the data for both ribosomal profiling and RNA-seq, we first trimmed the adapter sequences from the raw sequencing reads, and mapped the reads to rRNA reference sequences. Non-ribosomal RNA reads were then aligned to the GENCODE (Harrow et al., 2012) defined transcripts and reference genome (hg19), and only the uniquely mapped reads were used for the subsequent calculations (Table S1). For determination of gene expression levels of protein coding genes, we removed refSeq defined coding regions overlapping with upstream open reading frames (uORFs) defined in (Ji et al., 2015), 15 amino acids downstream of start codons and 5 amino acids upstream stop codons. The gene expression levels were calculated as reads per kilobase of transcript per million mapped reads (RPKM) in the remaining coding regions.

Analyses of differential transcription and translation. For analysis of regulated mRNAs, we treated the 30 minutes and 4 hours datasets independently and considered only mRNAs expressed at RPKM > 10 in at least one condition in each dataset, which corresponds to 5,696 (30 minutes) and 6,478 (4 hours) translated ORFs (Table S1). We applied the exact Poisson test (poisson.test in R, two-sided) to derive the statistical significance of ribosome occupancy (RO) counts between samples of interest, as follows: we compared the observed number of RO count events x (counts in gene A divided by total uniquely mapped counts in the nutrient-deprived condition) to the expected number of RO count events T (counts in gene A divided by total uniquely mapped counts in the complete DMEM condition) across all mRNAs. We defined transcriptional regulation as those mRNAs with an absolute log2 fold change in ribosome occupancy (RO) > 1.0 between the nutrient-deprived and complete DMEM condition and P value < 1e-03 (exact Poisson test). To define translational regulation, we used two analytical approaches. A) Translation efficiency (TE) was calculated as the log2 ratio of RPKM measured by ribosome profiling over RPKM measured by RNA-seq. We used the intersection of TE values with RO fold change to reject mRNAs with TE differences caused mostly by a decrease in mRNAs levels rather than an increase in RO upon the metabolic stress. An absolute difference in TE > 0.5, log2 RO fold change > 0.5 and P value < 1e-03 (exact Poisson test) between the nutrient- deprived and complete DMEM condition were used to define mRNAs with decreased translation, while a more stringent log2 RO fold change threshold (> 1.0) was applied along with the other criteria to define mRNAs with increased translation at 4 hours. B) We applied analysis of partial variance (APV) of normalized and rlog transformed RO and mRNAs levels and weighed the gene-specific variances using a random variance model to exclude possible correlations between mRNAs levels and TE values, as previously described (Larsson et al., 2010). Using the anota2seq package (Oertlin et al., 2018), mRNAs were defined as differentially translated if meeting the following criteria (-0.5 < minSlopeTranslation < 1.5, deltaP > 0.5, deltaPT > 0.15, FDR < 0.05). We applied a variance cutoff of 1e-06, and did not filter genes with zero counts to keep genes expressed in TAM but not in EtOH cells as input for APV.

Motif search, gene ontology and hierarchical clustering. We used RibORF (Ji et al., 2015) to define uORFs, and the MEME suite (Bailey et al., 2009) to performed an unbiased search for motifs in the 5’UTR of TE-regulated mRNAs. Enrichment of gene ontology (GO) terms in regulated mRNAs with corrected P-values (FDR, Benjamini and Hochberg) was determined using the ToppGene Suite (Chen et al., 2009), and representative clusters of GO terms were found using REVIGO (Supek et al., 2011), with an ‘allowed semantic similarity’ of 0.4. Plotting in R was performed using the ggplot2 package (Wickham, 2009), and clustering analysis with the pheatmap package (Kolde, 2012). The ten metabolic conditions were clustered based on the normalized and rlog transformed RPKM values measured by ribosome occupancy of either all mRNAs (Figure 3E) or after aggregation of the mRNAs into 5 k-means (Figure 3F) using ‘correlation’ as a distance measure and the ‘average linkage’ method.

Immunoblotting. MCF10A-ER-Src cells were grown under different media, washed twice with PBS, lysed in RIPA buffer (150mM NaCl, 1.0% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 50mM Tris-HCl pH 8.0) supplemented with protease inhibitors, sonicated, and centrifuged. Supernatant proteins were separated by SDS-PAGE on a polyacrylamide gel (NuPAGE 4-12% Bis-Tris gel, ThermoFisher scientific), and transferred to a nitrocellulose membrane. The membrane was blocked with 5% milk (or 5% BSA for phosphorylated proteins) in TBS + 0.1% Tween-20 (TBS-T), incubated with primary antibody in blocking buffer overnight at 4C, washed 3X with TBS-T, incubated with a

HRP-conjugated secondary mouse or rabbit antibody (Cell Signaling) in blocking buffer, and detected by enhanced chemiluminescence using either an X-ray film or a CCD camera. Primary antibodies: anti-phospho-eIF2α (Cell Signaling cat. No. 9721), anti-eIF2α (Cell Signaling), anti-phospho-p70 S6 kinase Thr389 (Cell Signaling, cat. No. 9205S), anti-phospho-4E-BP1 (Cell Signaling), anti-IL8 (ThermoFisher scientific cat. No. 14-7189-80), anti-IL6 (Novus Biologicals cat. No. NB600-1131SS), anti-CCL20 (antibodies-online GMBH cat. No. ABIN1043784, or ThermoFisher scientific cat. No. MA5-16253), anti-CXCL1 (Abcam cat. No. ab86436), anti-CXCL2 (Abcam cat. No. ab91511), anti-

CXCL3 (Abcam cat. No. ab10064).

Wound-healing motility assay. A single scratch was created using a p20 micropipette tip in confluent non-transformed and transformed cells. Cells were washed with PBS and cultured with either complete, glutamine-free, or BCAA-free DMEM for 6 hours followed by complete DMEM for 24 hours. As negative control, ER-Src cells were cultured with serum-free DMEM for 30 hours. Triplicate images were acquired by phase-contrast microscopy at 0 and 30 hours after wounding, and the percentage of gap closure was determined using TScratch (Gebäck et al., 2009). Three biological replicates were performed except for BCAA-deprived transformed cells (two experiments).

Transwell migration assay. MCF10A-ER-Src cells were trypsinized, resuspended in PBS, aliquoted (1E+05 cells) into complete or glutamine-free medium (200 μl) and seeded in the upper chamber of a Boyden insert placed inside 24 well plates. Conditioned-medium from cells grown in complete or glutamine-free DMEM was mixed with an equal volume of correspondent fresh medium, 750 µL added to the lower chamber, and cell migration via a membrane pore (8 μM) assessed after 18-20 hours of incubation. As control, fresh complete DMEM, glutamine-free DMEM or PBS were added to the lower chamber of other wells. Cells were fixed with 3.7% formaldehyde for 10 minutes, washed with PBS, permeabilized with 100% methanol for 20 minutes, washed with PBS, and stained with 0.1% crystal violet (CV) for 30 minutes at room temperature. After three PBS washes, non-migrated cells were scraped off the upper chamber with cotton swabs and dried. For quantification, we placed the Boyden chamber (containing the migrated cells) in 500 μl of 10% acetic acid to release the CV from cells, and measured absorbance at 590 nm. Relative cell migration was presented as the absorbance ratio of the conditioned medium relative to the respective fresh medium. Five and three biological replicates were performed in transformed and non-transformed cells, respectively. B FIGURE S1 A Molecular Function TAM (TE-Downregulated mRNAs) Structural Constitutent of ribosome 2 Structural molecule activity E = 1.5E-24 RNA binding

TAM NoQ 30' s rRNA binding 1i t Large ribosomal subunit rRNA… TE-DN B 0 10 20 30 40 50 60 -Log10 (FDR B&H) 0 1 2 3 4 5 6 7 8 9 10 11 12 Molecular Function Structural Constitutent of ribosome 2 EtOH (TE-Downregulated mRNAs) RNA binding E = 5.9E-53 E = 2.8E-48 structural molecule activity s

rRNA binding 1 i t

EtOH NoQ 30' B mRNA binding TE-DN 0 20 40 60 80 100 -Log10 (FDR B&H) 0 1 2 3 4 5 6 7 8 9 10 11 12 C EtOH-treated – NoQ 4hrs Analysis by Anota2seq – Ribosome occupancy UP (‘Translated mRNA’ Mode) term ID description log10 p-value dispensability N genes GO:0036499 PERK-mediated unfolded protein response -9.41 0.00 8 GO:0006915 apoptotic process -7.19 0.05 36 GO:0072359 circulatory system development -7.08 0.07 26 GO:0006357 regulation of transcription from RNA polymerase II -7.67 0.18 37 GO:0006520 cellular amino acid metabolic process -5.19 0.22 14 GO:0006984 ER-nucleus signaling pathway -8.37 0.26 9 GO:0009628 response to abiotic stimulus -5.58 0.26 25 GO:0042127 regulation of cell proliferation -6.28 0.27 31 GO:0009719 response to endogenous stimulus -5.49 0.35 30 GO:0034976 response to endoplasmic reticulum stress -7.01 0.39 14

TAM-treated – NoQ 4hrs Analysis by Anota2seq – Ribosome occupancy UP (‘Translated mRNA’ Mode) term ID description log10 p-value dispensability N genes GO:0036499 PERK-mediated unfolded protein response -11.98 0.00 10 GO:0072593 reactive oxygen species metabolic process -6.06 0.02 15 GO:0006564 L-serine biosynthetic process -6.20 0.05 4 GO:0042981 regulation of apoptotic process -7.43 0.16 40 positive regulation of multicellular organismal process -6.97 0.23 GO:0051240 40 GO:0009991 response to extracellular stimulus -6.92 0.23 24 GO:0042127 regulation of cell proliferation -6.27 0.23 39 GO:0006984 ER-nucleus signaling pathway -9.96 0.26 11 GO:0071216 cellular response to biotic stimulus -5.82 0.28 13 GO:0006955 immune response -5.54 0.33 36 GO:0009607 response to biotic stimulus -5.46 0.33 28 GO:0048585 negative regulation of response to stimulus -6.75 0.33 38 GO:0009719 response to endogenous stimulus -5.85 0.35 39 GO:0034976 response to endoplasmic reticulum stress -7.88 0.39 18 GO:0044283 small molecule biosynthetic process -5.75 0.39 21

D ATF4 PTGS2 IL24 IL6 Control.1 Control.2 OH t No Q.1 E No Q.2 Control.1

M Control.2 A

T No Q.1 No Q.2 Normalized Coverage Normalized Coverage Normalized Coverage Normalized Coverage

Figure S1; Related to figure 2: Regulation of mRNAs and their properties under glutamine-deprived transformed (TAM) and non-transformed (EtOH) MCF10A-ER-Src cells. (A) Gene ontology and (B) motif enrichment at 5’ untranslated regions of mRNAs with decreased translation efficiency after 30 minutes. (C) Gene ontology of mRNAs with increased ribosome occupancy after 4 hours as determined by anota2seq (‘translated mRNA mode’). (D) Ribosome occupancy of inflammatory mRNAs regulated after 4 hours. A CXCL3 IL24 PTGS2 FIGURE S2 Control.1

H Control.2 O t No BCAA.1 E No BCAA.2 Control.1

M Control.2 A

T No BCAA.1 No BCAA.2 Normalized Coverage Normalized Coverage Normalized Coverage

B TAM NoG 4hr EtOH NoG 4hr Unregulated Unregulated RO UP RO UP TE UP TE UP ) ) y y Cytokine/Inflam/Ang RO UP Cytokine/Inflam/Ang RO UP c Cytokine/Inflam/Ang TE UP Cytokine/Inflam/Ang TE UP n RO DN a ER-stress RO UP

p RO DN TE DN u TE DN Ribosomal TE DN c cc upan c c Ribosomal TE DN O O

e e m m o o s s o o b i b i R R Δ Δ ( ( Log2 Log2

Log2 (Δ mRNA levels) Log2 (Δ mRNA levels) C TAM NoCys/Cys 4hr EtOH NoCys/Cys 4hr Unregulated Unregulated RO UP RO UP

) TE UP TE UP y ) Cytokine/Inflam/Ang RO UP c ER-stress RO UP y

n Ribosomal TE DN Cytokine/Inflam/Ang TE UP

a ER-stress RO UP p RO DN u

c TE DN

c Ribosomal TE DN cc upan c O

O

e e m m o s o s o o b i i b R R Δ ( Δ ( Log2 Log2

Log2 (Δ mRNA levels) Log2 (Δ mRNA levels)

D E 16 F Coding GO:0030728– Ovulation 15 NoQ 30min – TE UP EtOH 4 mRNAs (FDR B&H: 1.11e-3) All stresses 4hrs – Cytokine/Inflam TE UP PLAU, IL1B, INHBA, SIRT1 GO:0010629 – Negative regulation of gene GO:0036499 – PERK-mediated unfolded 14 TAGLN, CXCL3, expression protein response 13 ITGA5, OGT, TINF2 AKAP12, IFFO2 GO:0001944 – Vasculature 24 mRNAs (FDR B&H: 1.99 e-5) 3 mRNAs (FDR B&H: 1.84 e-4) P = 0.56 Development CDKN1A, EZR, LPIN1, SOX7, JUNB, TBX3, DDIT3, ATF4, PPP1R15A 12 No Glc No Leu/Iso/Val 18 mRNAs (FDR B&H: 2.70e-8)

Fs KRT16 ZBTB38, TRIB3, STC2, RBMX, ANKRD1, NFX1, LIF, CYP1B1, TBX3, EREG, SAT1, EGR1. NFE2L2, HBEGF, JAG1, BTG1, PER1, XBP1, DLX2, CBX4, RFC1, KLF4, SKIL, FERMT2, KDM4, DDIT3, SAMSN1, 11 6 45 OGT, PPP1R15A, HIST1H1C, LIF, P = 0.013 P = 0.059 ERRFI1, AHR, XBP1, PPP1R15A, IL1A, SIRT1, CHAC1, VEGFA, HR, PRDM1 O R 10 RRBP1, GFPT1, MSN, KLF4, DDIT3, IL8, PMAIP1, HIST1H2AC, ATF4, TXNIP, 1 4

u EZR, NCL, KRT6A 66 CXCL2 & CXCL3, IL24, NCL, AZI2 f 9 6 No Q

o 0 5 90 8 GO:0006418– tRNA aminoacylation GO:0006954 – Inflammatory response 0 for protein translation 10 mRNAs (FDR B&H: 1.21e-4) 7 No Cys/Cys 0 6 mRNAs (FDR B&H: 1.24e-4) be r 1 EtOH TAM TNFAIP6, PPARD, IL1B, NUPR1, PTX3, EPRS, AARS, WARS, MARS, NARS, TARS 6 20 0 PTGS2, IL6, NFKB1, NFKB2, ICAM1 m 2 u 5 81 13 36 GO:0019752– Carboxylic acid metabolic DNAJB9, TNFAIP3, PTGS2, CTGF, PRDM1 N process 4 9 mRNAs (FDR B&H: 4.06 e-4) TRIB3, CARS, PSAT1, PCK2, SESN2, DDIT3, HERPUD1 3 PHGDH, ATF4, ASNS, GARS 2 1 0 N = 6,478 N =33 N =22

Figure S2; Related to Figure 3: Regulation of mRNAs in transformed and non-transformed MCF10A-ER-Src cells subjected to 4 hours of metabolic stress conditions. (A) Ribosome occupancy (RO) of inflammatory mRNAs after 4 hours of BCAA deprivation. (B-C) Changes in mRNA and RO levels upon 4 hours deprivation of (B) glucose and (C) cysteine/cystine relative to complete medium in transformed (left) and non-transformed (right) cells. (D) Venn diagram showing the mRNAs with increased translation efficiency (TE) in all metabolic stresses in transformed vs. non-transformed cells; cytokines highlighted in red. (E) Number of upstream open reading frames (uORFs) in all expressed coding mRNAs (grey), mRNAs with increased TE at 30 minutes of glutamine deprivation (red) and all inflammatory mRNAs with increased TE at 4 hours of nutrient deprivation (green). (F) Venn diagram showing the mRNAs with increased RO in each metabolic stress condition in non-transformed cells. FIGURE S3 A B - l 60 No Cys- 300 y No No Q Cys Ethanol Leu/Iso/Val No Cys- ) ) 50 Cys 250 No Tamoxifen

M Leu/Iso/Val No Q PK M s K mi noa c R

P No Cys-Cys a No (

No Q f

R 40 200 No Cys-Cys as e Leu/Iso/Val o e No

n F4 y m a No Glc T Leu/Iso/Val he t c

o No Q e A

n Ctrl DMEM s

n t 30 150 f M a

y o p c Adjusted R² = -0.06 No Glc s ib o i Adjusted R² = 0.31 u y r t R c P = 0.44 e c 20 100 P = 0.19 m O Ctrl DMEM

o No Glc tRN A

e Adjusted R² = -0.14 - upan c Ctrl DMEM e Adjusted R² = 0.05 c m G 10 P = 0.53 Ethanol 50 ( o P = 0.35 Ctrl DMEM s O c No …

o Tamoxifen b i 0 0 R 0% 20% 40% 60% 80% 100% 120% 0% 20% 40% 60% 80% 100% 120% Protein synthesis activity at 6 hours Protein synthesis activity at 6 hours (Relative pixel intensity - FITC channel) (Pixel intensity - FITC channel)

C Ribosomal subunit mRNAs (176 mRNAs) P = 3.4e-04 P = 0.6 P = 1.1e-14 P = 5.6e-05

y 1 P = 8.0e-04 c P = 1.3e-30 n a p u c c

O EtOH NoCys

0

e EtOH NoGlc EtOH NoQ m

o EtOH NoBCAA

s TAM NoCys o TAM NoGlc b i TAM NoQ R

-1 TAM NoBCAA C F

2 g o L -2

D TAM No Glc TE DN TAM No Q TE DN TAM No BCAA TE DN TAM No Cys TE DN

Ribosomal 55 Non-ribosomal 46 51 61 1 1 138 354

EtOH No Glc TE DN EtOH No Q TE DN EtOH No BCAA TE DN EtOH No Cys TE DN

Ribosomal 22 66 3 Non-ribosomal 44 52 84 668 13

Figure S3; Related to Figure 3: Correlation between protein synthesis and the ribosome occupancy of (A) aminoacyl- tRNA-synthetases and (B) ATF4 in MCF10A-ER-Src cells. (C) Ribosome occupancy of all ribosomal subunit mRNAs in each metabolic condition; P values are derived from the Wilcoxon test. (D) Representation of ribosomal subunit mRNAs in the pool of mRNAs with decreased TE in each metabolic stress condition in MCF10A-ER-Src cells. FIGURE S4 A B Ctrl NoG NoQ 4 hours CXCL3 p-S6K1 M A T M β-Actin

A p-4EBP1 T β-Actin Ctrl NoG NoQ NoBCAA p-S6K1

H IL8 M A O t

p-4EBP1 T E β-Actin β-Actin

1 s A M n y uc i Ctrl NoG NoQ NoBCAA l r / C o C A No Q s T CXCL1 DM E

l No G No B

Ct r CXCL2 No C y M A

T CCL20

β-Actin

C EtOH TAM Time0 ~30hrs Time0 ~30hrs

Ctrl DMEM (6hrs) + 100% * Ctrl DMEM (~24hrs) * EtOH 80% TAM ure o s

NoQ (6hrs) + l 60% c Ctrl DMEM (~24hrs) *

a p 40% 20% % G * No BCAA (6hrs) + N o 0% Ct r N o B Ctrl DMEM (~24hrs) No C A F l Q B A S

No FBS (~30hours)

Figure S4; Related to Figure 4: Immunoblot analysis and cell motility assay of nutrient-deprived transformed (TAM) and non-transformed (EtOH) cells. (A) Immunoblot analysis of mTOR-phosphorylated proteins after 4 hours of nutrient deprivation or Torin1 treatment (500nM). (B) Immunoblot analysis of cytokine levels after 24 hours of nutrient deprivation (C) Phase-contrast images (left) and TScratch quantification (right) of nutrient-deprived cells in the wound healing assay. Error bars represent s.d. of three biological replicates (Control, NoQ, NoFBS) or two biological replicates (NoBCAA). * P < 0.05 comparing nutrient-deprived to complete medium as determined by t-test. Supplemental References

Bailey, T.L., Boden, M., Buske, F.A., Frith, M., Grant, C.E., Clementi, L., Ren, J., Li, W.W., and Noble, W.S. (2009). MEME SUITE: tools for motif discovery and searching. Nucleic Acids Res. 37, W202-8. Chen, J., Bardes, E.E., Aronow, B.J., and Jegga, A.G. (2009). ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res. 37, W305-11. Gebäck, T., Schulz, M.M.P., Koumoutsakos, P., and Detmar, M. (2009). TScratch: a novel and simple software tool for automated analysis of monolayer wound healing assays. Biotechniques 46, 265–274. Harrow, J., Frankish, A., Gonzalez, J.M., Tapanari, E., Diekhans, M., Kokocinski, F., Aken, B.L., Barrell, D., Zadissa, A., Searle, S., et al. (2012). GENCODE: the reference annotation for The ENCODE Project. Genome Res. 22, 1760–1774. Ingolia, N.T., Ghaemmaghami, S., Newman, J.R.S., and Weissman, J.S. (2009). Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling. Science 324, 218–223. Ingolia, N.T., Brar, G. a, Rouskin, S., McGeachy, A.M., and Weissman, J.S. (2012). The ribosome profiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments. Nat. Protoc. 7, 1534–1550. Ji, Z., Song, R., Regev, A., and Struhl, K. (2015). Many lncRNAs, 5’UTRs, and pseudogenes are translated and some are likely to express functional proteins. Elife 4, e08890. Kolde, R. (2012). Pheatmap: pretty heatmaps. R Packag. Version 61. Larsson, O., Sonenberg, N., and Nadon, R. (2010). Identification of differential translation in genome wide studies. Proc. Natl. Acad. Sci. 107, 21487–21492. Oertlin, C., Lorent, J., Gandin, V., Murie, C., Masvidal, L., Cargnello, M., Furic, L., Topisirovic, I., and Larsson, O. (2018). Generally applicable transcriptome-wide analysis of translational efficiency using anota2seq. bioRxiv. Supek, F., Bošnjak, M., Škunca, N., and Šmuc, T. (2011). REVIGO Summarizes and Visualizes Long Lists of Gene Ontology Terms. PLoS One 6, e21800. Wickham, H. (2009). ggplot2: Elegant Graphics for Data Analysis (Use R!) (Springer, New York).