The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), 10.1016/J.Molcel.2013.09.018 Molecular Cell Short Article

Home , CENPC1, CENPF, Nuclear pore, STAG1, WAPAL

Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell Short Article

The Translational Landscape of the Mammalian Cell Cycle

Craig R. Stumpf,1,2 Melissa V. Moreno,1,2 Adam B. Olshen,2,3 Barry S. Taylor,2,3,4,* and Davide Ruggero1,2,* 1Department of Urology, University of California, San Francisco, CA 94158, USA 2Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, CA 94158, USA 3Department of Epidemiology and Biostatistics, University of California, San Francisco, CA 94158, USA 4Department of Medicine, University of California, San Francisco, CA 94158, USA *Correspondence: [email protected] (B.S.T.), [email protected] (D.R.) http://dx.doi.org/10.1016/j.molcel.2013.09.018

SUMMARY through the ubiquitin proteasome pathway, at specific times during the cell cycle is required for progression to subsequent Gene regulation during cell-cycle progression is an phases (Peters, 2006). Systems-level mass spectrometry ap- intricately choreographed process, ensuring accu- proaches are also beginning to elucidate targets of the ubiqui- rate DNA replication and division. However, the tin-proteasome pathway, as well as cell-cycle-specific patterns translational landscape of gene expression underly- of posttranslational modifications on a large number of proteins ing cell-cycle progression remains largely unknown. (Kim et al., 2011; Merbl et al., 2013). While these studies have Employing genome-wide ribosome profiling, we provided great insight into the highly coordinated gene expression program of the cell cycle, a key step in modulating protein uncover widespread translational regulation of levels has remained undefined: the regulation of mRNA hundreds of mRNAs serving as an unexpected translation. mechanism for gene regulation underlying cell-cycle To date, the study of translational control during the mamma- progression. A striking example is the S phase trans- lian cell cycle has generally focused on global reductions in pro- lational regulation of RICTOR, which is associated tein synthesis during mitosis, monitored by a decrease in amino with cell cycle-dependent activation of mammalian acid incorporation into proteins (Fan and Penman, 1970; Konrad, target of rapamycin complex 2 (mTORC2) signaling 1963). Conversely, a relatively modest number of mRNAs have and accurate cell-cycle progression. We further iden- been identified as actively translated during mitosis (Qin and Sar- tified unappreciated coordination in translational now, 2004). Other studies have primarily investigated regulation control of mRNAs within molecular complexes dedi- of single mRNAs, such as the translational regulation of cyclin E, cated to cell-cycle progression, lipid metabolism, which is required for progression into S phase (Lai et al., 2010). While limited in scope, these studies both highlight the impor- and genome integrity. This includes the majority of tance of translational regulation during cell-cycle progression mRNAs comprising the cohesin and condensin com- and underscore the need for an unbiased, genome-wide analysis plexes responsible for maintaining genome organi- of the translational landscape during the mammalian cell cycle. zation, which are coordinately translated during Here, we have employed ribosome profiling (Ingolia et al., 0 specific cell cycle phases via their 5 UTRs. Our find- 2009) to uncover widespread translational regulation during ings illuminate the prevalence and dynamic nature of cell-cycle progression. We defined remarkable levels of transla- translational regulation underlying the mammalian tional control of key cell cycle genes. Importantly, among these cell cycle. mRNAs, we uncovered translational regulation of RICTOR, which correlates with the cell-cycle phase-specific signaling of the mammalian target of rapamycin (mTOR) kinase pathway INTRODUCTION and suggests an important role in cell-cycle progression. More- over, we identify surprising, coordinate regulation in translational During cell division, exquisite temporal control of protein expres- control of functionally related sets of mRNAs, suggesting a reg- sion in distinct phases of the cell cycle underlies fundamental ulatory mechanism that dictates the coordinate expression and checkpoints that ensure accurate completion of chromosome activity of key molecular machinery in the cell. Among these duplication and segregation of a daughter cell. A central para- translationally controlled networks are mRNAs required for digm that has emerged is that rapid, dynamic, and fine-tuned re- metabolism, nuclear transport, and DNA repair, the latter of programming of gene expression occurs during specific phases which ensures genome fidelity. Together, this work highlights of the cell cycle. For example, a large number of mRNAs, both the prevalence and the dynamic nature of translational including those involved in promoting cell-cycle progression, regulation during cell-cycle progression. It further suggests a are transcriptionally activated in a cell-cycle phase-dependent multifaceted mechanism for differential regulation of transcript- manner (Cho et al., 2001; Whitfield et al., 2002). In addition, specific translational control in the execution of distinct steps degradation of many cell-cycle checkpoint proteins, primarily of the cell cycle.

Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. 1 Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

Figure 1. Systematic and Multifaceted Translational Control of Gene Expression during the Mammalian Cell Cycle (A) The cumulative fraction of ribosome-bound mRNA on all expressed transcripts in each phase of the cell cycle is shown as a function of increasing ribosome-bound mRNA. The x axis represents the scaled fraction of total ribosome-bound reads, and the y axis represents the fraction of expressed transcripts. (B) Representative scatter plots illustrate ribosome occupancy as a function of mRNA abundance (measured as sequencing read counts). The dashed line represents the expected level of ribosome occupancy given mRNA abundance (see Supplemental Experimental Procedures). mRNAs with statistically signiﬁcant translational regulation are those with greater or less than the expected levels of ribosome occupancy given their mRNA expression (FDR < 1%; G1 is gray, S phase is blue, and mitosis is green). (C) The total number of genes with signiﬁcantly increased (black) or decreased (gray) ribosome occupancy is shown, including those unique to a given phase or shared between multiple phases of the cell cycle. (legend continued on next page)

2 Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

RESULTS exhibit increased or decreased translation in all phases of the cell cycle, suggesting that they may share a mechanism to main- To understand the extent and impact of translational regulation tain high or low levels of translation throughout the cell cycle (Fig- during cell-cycle progression, we employed ribosome profiling ure 1C). Moreover, the specific patterns obtained by hierarchical to identify individual mRNAs exhibiting unexpected levels of clustering of the 1,255 translationally regulated mRNAs were not ribosome association during each phase of the cell cycle. After observed when comparing corresponding transcript expression synchronizing human HeLa cells in G1, S phase, or mitosis, levels, indicating that translational regulation is indeed a distinct libraries for total mRNA and ribosome-protected RNA fragments regulatory system, uncoupled from transcription, controlling were prepared, sequenced, and analyzed using a sophisticated gene expression during the cell cycle (Figure S1B). statistical framework (see Experimental Procedures)(Figure S1A In addition to defining the relative level of translation for spe- available online) (Olshen et al., 2013). We initially characterized cific genes within each cell cycle phase, we also identified a large global levels of ribosome occupancy in the expressed transcrip- number of mRNAs that undergo significant changes in transla- tome of each phase of the cell cycle. This data revealed a signif- tion between phases of the cell cycle. We identified 353 mRNAs icant decrease in the overall level of ribosome-bound mRNAs with a statistically significant change in translation between any during mitosis compared to either G1 or S phase, which is two phases of the cell cycle (FDR < 5%) (Figure 1E, Table S2). consistent with a decline in cap-dependent translation during Among these, 112 mRNAs were translationally regulated exclu- mitosis (Figure 1A) (Fan and Penman, 1970; Konrad, 1963). sively between specific phases of the cell cycle. These mRNAs Beyond these global changes in translation, our principal aim comprise a number of important cell-cycle regulatory genes, was to understand the extent of gene-specific translational regu- including CLASP2 and KNTC1, which are involved in establish- lation during cell-cycle progression. We therefore developed a ing the mitotic spindle checkpoint. Overall, these data demon- computational framework for determining the statistical signifi- strate that systematic and multifaceted translational control of cance of relative ribosome occupancy of mRNAs within individ- gene expression exists during progression through the mamma- ual phases of the cell cycle. This analysis quantifies levels of lian cell cycle. ribosome occupancy higher or lower than those predicted from We next sought to understand the regulatory logic underlying transcript abundance (Figure 1B, highlighted points; see Exper- general changes in translational regulation observed among imental Procedures). This approach enabled us to quantify the specific phases of the cell cycle. To determine certain parame- landscape of mRNA translation during a given cell cycle phase ters of cell cycle-dependent translational regulation, we exam- or to directly assess differential translational regulation between ined specific structural characteristics of 50 UTRs, including their any two phases of the cell cycle. length, percent of G + C content (%G+C), and minimum free en- Strikingly, we observed extensive and dynamic translational ergy (MFE). In the G1 phase of the cell cycle, there is a significant control of individual mRNAs during different phases of the cell association between ribosome occupancy and the length of 50 cycle (Figure 1B, Table S1). In total, we identified 1,255 mRNAs, UTRs such that mRNAs with shorter UTRs had high levels of representing 12% of the expressed transcripts in these cells, ribosome occupancy. In both G1 and S phase, there is an inverse that exhibit levels of ribosome occupancy in any cell cycle phase relationship between the G+C content of 50 UTRs and ribosome higher or lower than those expected (false discovery rate [FDR] < occupancy levels. Finally, mRNAs with higher ribosome occu- 1%). We next sought to determine how the translation of these pancy in all phases of the cell cycle studied were associated 1,255 mRNAs varies among cell cycle phases. Transcript-spe- with higher predicted MFE, revealing that their 50 UTRs possess cific translational regulation was most prevalent in G1 and S less-complex secondary structures (Figure S1C). Together, phases (Figure 1C). Indeed, genome-wide, there was far greater these findings suggest that several of the defining features of similarity in the level of ribosome occupancy in mRNAs between 50 UTRs may contribute to the translational efficiency of the G1 and S phase than existed when comparing either of these to mammalian genome within specific phases of the cell cycle. mitosis (Figure 1D). Importantly, these differences in transcript- However, these very general characteristics of the 50 UTR do specific translational control during the cell cycle are indepen- not account for all the patterns of translational regulation dent from global changes in the levels of protein synthesis that observed here. Therefore, it is likely that there are additional occur during specific cell cycle phases. Here, we measured important determinants of transcript-specific translational con- the levels of ribosome association for specific mRNAs relative trol (see below). to the background level of ribosome association during each in- The translational dynamics of key cell-cycle regulators are still dividual phase. This is especially relevant during mitosis, when poorly understood. We therefore assessed the pattern of statis- global protein synthesis levels are lower compared to those of tically significant translational regulation among bona fide cell- G1 or S phase (Figure 1A) (Fan and Penman, 1970; Konrad, cycle regulatory genes during each phase of the cell cycle (Fig- 1963). Additionally, we identified core groups of mRNAs that ure 2A) (Subramanian et al., 2005). This analysis uncovered

(D) Density plots of the nominal p value of ribosome occupancy for each gene between any two phases of the cell cycle (colors designate the density of genes in a given region, where red and blue are the most and least dense, respectively). (E) Unsupervised hierarchical clustering of mRNA translation across the phases of the cell cycle using the 353 differentially translationally regulated transcripts between any two phases in direct comparisons. Genes and cell cycle phases are clustered based on the level of normalized ribosome occupancy (mean- centered translational efﬁciency by gene, scales and colors indicate the direction and magnitude of mRNA translation; S is S phase, M is Mitosis). See also Figure S1 and Tables S1 and S2.

Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. 3 Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

Figure 2. Phase-Dependent Translational A B Test Control Control of Key Cell-Cycle Regulators, 5’UTR Firefly Renilla Including RICTOR and mTOR Signaling G1 S phaseMitosis FBXO5 RICTOR (A) A heatmap of the significance of translational SMC4 KIF15 NUAK2 regulation among cell-cycle progression genes POLA1 MRE11A (asterisk: RICTOR). Shading represents the sig- OFD1 nificance level of increased or decreased trans- SASS6 KIF18A lational regulation (red and blue, as indicated). RINT1 0 RAD50 (B) A diagram describing the 5 UTR luciferase AHR Relative translation 0 ROCK1 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 3 5 7 reporter assay (top panel). 5 UTRs cloned into the AHCTF1 firefly reporter are scaled to length. Capped LIN9 NEDD1 mRNAs are transfected into synchronized cells CENPF RICTOR * G1 prior to assaying reporter levels. Levels of trans- CENPC1 RICTOR S phase 0 PLK1 lation directed by RICTOR or NUAK2 5 UTRs are CCNB1 Mitosis shown: y axis is reporter value relative to cells in ASNS PRIM1 G1 (bottom panel). BUB1B TPX2 (C) Western blots showing RICTOR protein levels PRKDC and phosphorylation of mTORC2 targets (AKT- E2F5 NUAK2 BUB1 S473, PKCa-S657). pHistone H3 is a mitosis CUL2 EGF marker. Tubulin is a loading control. PSMA3 NCAPH (D) Mitotic progression of thymidine-synchronized RBL2 DOCK2 MEFs: y axis is the percent of mitotic cells. Bars KIF23 represent the mean ± SD. See also Figure S2. RACGAP1 SYNE1 KIF4A C KIFAP3 G1 S phase Mitosis PSMD14 CCNE2 CASC5 RICTOR KIF11 CENPE mTOR complex 2 (mTORC2) (Laplante MYO6 pAKT S473 ANAPC1 and Sabatini, 2012), displayed a signifi- KIF2A TTK total AKT cant change in the level of ribosome oc- ANLN cupancy, with an almost 3-fold increase LRPPRC pPKCα S657 NUP107 CENPI from G1 to S phase and a greater than SGOL2 total PKCα NUP133 13-fold decrease from S phase to mitosis. PSMD1 SMC3 pHistone H3 S10 Our findings further show an accumula- PCM1 CLASP1 tion of RICTOR protein, but not other SMARCA5 Tubulin CEP135 components of mTORC2, that mirrors HOOK3 CKAP5 this pattern of ribosome occupancy (Fig- RBBP4 STAG1 D 60 ures 2A, 2C, and S2B), which suggests KIF20A wildtype DYNC1I2 that RICTOR protein accumulation is, at CDC26 50 Rictor-/- RHOF least in part, mediated by translational E2F1 CRIPAK 40 control. To further investigate the mecha- ARPC4 RUNX3 nisms responsible for cell cycle-depen- MRAS 30 RAC1 dent translational control of RICTOR,we RPS27 assessed the activity of its 50 UTR. Strik- EVL 20 APITD1 0 MYL6B ingly, the 5 UTR of the RICTOR mRNA APOE 10 is sufficient to direct increased protein

Significance (q value) Percent of cells in Mitosis expression during S phase, which then 0 N.S. decreases upon entry into mitosis, a <1e-10 <1e-10 8 12 24 Decreased Increased Hours post thymidine pattern consistent with the cell cycle- dependent ribosome occupancy we observed (Figures 2A, 2B, and S2A). We extensive translational regulation among these genes, high- compared the translation directed by the RICTOR 50 UTR to lighting the prevalence of translational control during cell-cycle the 50 UTR of an mRNA that exhibits a distinct pattern of cell- progression. For example, we observed high levels of ribosome cycle phase-specific translation, NUAK2 (Figure 2B). In our occupancy in G1 and S phase for CCNE2, a cyclin known to be ribosome profiling experiments, NUAK2 showed a significant translationally regulated. On the contrary, our data revealed low increase in ribosome occupancy during mitosis compared to levels of ribosome occupancy of E2F1, also during G1 and S either G1 or S phase (q values = 9.5 3 104 and 1.3 3 103, phase, which is consistent with cell cycle-dependent miRNA- respectively; Table S2). RICTOR and NUAK2 50 UTRs promote mediated regulation of E2F1 (Pulikkan et al., 2010). Furthermore, unique patterns of translation in the luciferase reporter assay, this analysis identified translational regulation among several suggesting that translational regulation of these mRNAs is cell-cycle mediators, including PLK1 and BUB1, two mitotic specific and the patterns of translation we observe are not due checkpoint kinases. Notably, RICTOR, the defining subunit of to global changes in protein synthesis (Figure 2B).

4 Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

These data suggest that regulation of RICTOR mRNA transla- control may be a key contributor in regulating the response to tion and its accumulation at the protein level during S phase diverse types of DNA damage that arise during cell-cycle could modulate the activity of mTOR during the cell cycle. To progression. test this hypothesis, we assessed phosphorylation levels of Given this striking pattern of translational regulation among mTORC2 targets AKT (S473) and PKCa (S657) in lysates from genes involved in DNA repair, we sought to identify cell-cycle synchronized cells (Figure 2C). We identified a strong correlation phase-specific translational regulation of mRNAs required for between the levels of RICTOR protein and the phosphorylation of genome fidelity. Strikingly, the majority of genes comprising these two mTORC2 targets during the S phase of the cell cycle. the cohesin and condensin complexes are translated in a cell- Moreover, the phosphorylation of AKT during S phase is depen- cycle phase-specific manner (Figure 4A). mRNAs from compo- dent on RICTOR, since it does not occur in Rictor/ mouse em- nents of both the condensin and cohesin complexes exhibit rela- bryonic fibroblasts (MEFs) (Figure S2C). Upstream signaling tively high levels of ribosome occupancy during G1 and S phase pathways, such as phosphatidylinositol 3-kinase (PI3K)/ that decrease in mitosis (Figure 4A). As these complexes are PDPK1, can activate AKT by phosphorylating T308 (Alessi loaded onto DNA during S phase or G2 in order to prepare chro- et al., 1996; Alessi et al., 1997). The fact that phosphorylation mosomes for segregation during mitosis (Hirano, 2012; Wood of AKT at T308 does not change during the cell cycle further sug- et al., 2010), the ribosome occupancy we observe is consistent gests that the RICTOR-dependent phosphorylation of AKT at with the requirement for the condensin and cohesin complexes S473 during S phase is independent from the induction of up- to be produced prior to mitosis. We determined that the molec- stream PI3K signaling (Figure S2B). Importantly, we also observe ular basis for these translational patterns in gene expression is, an increase in the phosphorylation of p27, a downstream target at least in part, through 50 UTR-mediated regulation. For of AKT, specifically during S phase (Figure S2B). Phosphoryla- example, the 50 UTRs of the core components of the condensin tion of p27 by AKT inhibits its activity, thereby promoting cell- complex, SMC2 and SMC4, direct similar patterns of transla- cycle progression (Shin et al., 2002). These findings suggest tional activation during G1 and S phase that decrease in mitosis that the accumulation of RICTOR protein levels in S phase may and mirror the cell cycle ribosome occupancy profile. Likewise, have an important function in cell-cycle progression. Rictor/ the 50 UTRs of SMC3, STAG1, and NIPBL, components of the MEFs have previously been shown to possess a proliferation cohesin complex, direct translation that is high during G1 and defect (Shiota et al., 2006). We therefore determined when the S phase and decreased during mitosis (Figure 4B). On the con- proliferation defect in Rictor/ MEFs manifests during the cell trary, the 50 UTR from RANBP1, a control mRNA whose pattern cycle relative to the increase in translation of RICTOR observed of ribosome occupancy is distinct from components of the con- in S phase. Rictor/ MEFs exhibit a defect in progression densin and cohesin complexes and is functionally unrelated, through mitosis as evidenced by a lag in the accumulation of does not (Figure 4B). diploid cells after synchronization in S phase (Figure 2D), associ- We further identified unique cell cycle-dependent patterns of ated with a marked decrease in S phase cells compared to wild- ribosome occupancy within specific regions of transcripts type (p value = 0.0006, Student’s t test; Figure S2D). Together, belonging to the cohesin complex. For example, the density of our findings suggest that translational regulation, at least in bound ribosomes on the NIPBL mRNA, a member of the cohesin part, leads to an increase in RICTOR protein that underlies cell complex, shifts from the primary initiation codon of the coding cycle-dependent mTORC2 activity and is associated with the region to an upstream open reading frame (uORF) 59 nucleo- progression from S phase to mitosis. tides upstream of the primary initiation codon in a highly evolu- An outstanding question is whether functionally related tionarily conserved region of the 50 UTR as cells progress groups of genes may be translationally coregulated as a means through the cell cycle into mitosis (Figures 4C, left, and S3A). by which to simultaneously control the expression of important uORFs often act to decrease the translation of primary open mRNA networks. Strikingly, we identified numerous clusters of reading frames, which is consistent with the decrease in trans- functionally related genes among translationally coregulated lation observed from the NIPBL 50 UTR in the luciferase reporter mRNAs (Table S3). These include genes central to the control assay (Figure 4B). In the case of WAPAL, a gene that promotes of metabolism, nuclear transport, and DNA repair (Figure 3A, dissociation of cohesin from sister chromatids during mitosis Table S3). Among the metabolism genes identified, there was (Gandhi et al., 2006; Kueng et al., 2006), we identified an unex- a particular enrichment of genes involved in lipid metabolism pected peak of ribosome density downstream of the translation and the tricarboxylic acid cycle (TCA) cycle (Figure 3B). Further- termination codon in the 30 UTR that is only present during more, we identified a significant enrichment of translationally mitosis (Figure 4C, right). This region of the WAPAL 30 UTR is regulated mRNAs, primarily during G1 and S phase, involved highly conserved and contains two additional in-frame stop co- in nuclear-cytoplasmic transport, including a large number of dons, suggesting that WAPAL could produce a C-terminally core scaffolding components of the nuclear pore complex extended protein product during mitosis (Figures 4C, right, (NPC) (Figure 3C). In fact, over 20% of NPC components are and S3B). Together, these data suggest that components of translationally regulated during cell-cycle progression, primarily the condensin and cohesin complexes utilize multiple modes during interphase, when the number of NPCs increases of translational regulation to coordinate their expression during dramatically (Antonin et al., 2008). Furthermore, translationally the cell cycle. Furthermore, translational regulation may help regulated genes are components of multiple DNA repair path- to facilitate the assembly or modulate the activity of large protein ways, including mismatch repair and double-strand break complexes by ensuring that individual components of these repair pathways (Figure 3D). This suggests that translational complexes are coordinately expressed.

Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. 5 Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

G1 G1 S phaseMitosis G1 S phaseMitosis S phase ST6GALNAC6 FH STS IDH1 Mitosis AGPAT1 ACO2 NEU3 PDK2 ECHS1 PDK3 ADM IDH3A Nuclear ETNK1 PDHA2 Transport ALG1 PCK1 Metabolism CPT1A SDHC Cohesin and Lipid TCA Cycle Condensin Metabolism C D

G1 S phaseMitosis G1 S phaseMitosis NUP133 POLA1 NUP205 MRE11A NUP107 MSH2 TPR LIG4 KPNB1 RAD50 NUP93 MSH6 NUP188 MSH3 NUP155 RECQL DNA Nuclear ERCC4 Repair Transport UBE2V1 DNA Repair

Significance (q value) <1e-10 <1e-10

Decreased N.S. Increased

Figure 3. Translational Coregulation of Large Molecular Complexes during Cell-Cycle Progression (A) Among translationally regulated genes during the cell cycle, a network of statistically significant functional enrichments (nodes, nominal p value < 0.05; radius scaled to the number of genes) and their relatedness (edges, spearman correlation r R 0.3) indicate a highly interconnected set of modules of molecular function. Groups of nodes closely related by function are highlighted in yellow and labeled. A Venn diagram overlay represents the relative overlap of enriched molecular function between cell cycle phases (G1 is gray, S phase is blue, and mitosis is green). (B–D) Heatmaps highlight the significance of translational regulation of genes that define representative functional categories, including lipid metabolism and TCA cycle (B), nuclear transport (C), and DNA repair (D). Shading represents the significance level of increased or decreased translational regulation (red and blue, as indicated). See also Table S3.

DISCUSSION and PKCa. Phosphorylation of both AKT and PKCa by mTORC2 during S phase is consistent with their roles in promoting cell Our work delineates the unexpected magnitude and dynamic growth and proliferation and reveals how this process may be nature of translational regulation during the mammalian cell cy- regulated at the level of translational control. Moreover, we did cle. We have presented a comprehensive network of interrelated not observe overt translational regulation of other mTOR com- and coordinately translationally regulated mRNAs underlying plex components, highlighting the specificity in RICTOR 50 this fundamental biological process. These data suggest that UTR translational activation in controlling mTOR signaling during translational control is a particularly well-suited mechanism for cell-cycle progression. Elucidating the precise mechanisms gov- fine-tuning gene expression during dynamic processes such erning translational regulation of RICTOR will be an important as cell-cycle progression. For example, we uncovered unex- area of future research that may play an important role in pected translational regulation of a key component of the mTORC2 signaling during cell-cycle progression. mTOR pathway, a key regulator of cell growth. RICTOR be- One surprising finding from our data is the translational core- comes translationally induced specifically upon transitioning gulation of the molecular machinery responsible for maintaining into S phase of the cell cycle. Although we cannot exclude that genome integrity. A number of translationally regulated genes other mechanisms, such as control of protein stability, may are involved in sensing multiple types of DNA damage, such as also cooperate in modulating RICTOR protein abundance during base pair mismatches and double-strand breaks, that are specif- the cell cycle, our findings show that accumulation of RICTOR ically translationally activated during S phase. Notably, multiple during S phase modulates mTORC2 signaling to promote the orthogonal DNA repair pathways are controlled by translation, phosphorylation of specific downstream targets, including AKT suggesting a critical regulatory mechanism that maintains the

6 Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

A Condensin complex Cohesin complex p = 0.027 p = 0.0015 3 3

2 2

1 1 SMC3

0 0

SMC2 STAG1 Translational efficiency Translational Translational efficiency Translational -1 -1 SMC4 NIPBL -2 -2 WAPAL G1 S phase Mitosis G1 S phase Mitosis

B 1.4

Test Control 1.2 5’UTR Firefly Renilla SMC2 1.0 SMC4 SMC3 0.8 ranslation STAG1 NIPBL 0.6 RANBP1 0.4 Relative t 0.2

0.0 SMC2 SMC4 SMC3 STAG1 NIPBL RANBP1

C 1.0 1.0 0.5 0.5 0.0 0.0 Conservation Conservation NIPBL WAPAL G1 S phase Ribosome-bound Reads Mitosis

36,953,650 uORF 36,954,000 88,194,980 88,197,970

AUG

Figure 4. Condensin and Cohesion Complex Components Are Coordinately and Translationally Regulated during the Cell Cycle at the Level of Their 50 UTR (A) The translational efficiency of components of the condensin (left) and cohesin (right) complexes during each cell cycle phase are indicated (specific genes mentioned elsewhere are outlined for clarity). (B) A diagram of the luciferase reporter assay with 50 UTRs scaled to length (left). Levels of translation directed by specific 50 UTRs are indicated: y axis is the reporter value relative to cells in G1 (right). Bars represent mean ± SD from six replicates. (C) The level of bound ribosome in the 50 UTR of NIPBL (left) and the 30 UTR of WAPAL (right), with peaks of interest denoted by red arrows (evolutionary conservation is indicated). G1 is gray, S phase is blue, and mitosis is green. Representative gene features are indicated: uORF is designated by an orange box; narrow or wide black bars represent UTRs and coding exons, respectively; black lines indicate introns; and arrows indicate the coding strand. Numbers represent absolute genomic positions. See also Figure S3.

Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. 7 Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

fidelity of the genome, adding a robust level of protection in mRNA. Ribosome-protected mRNA fragments were isolated by centrifuga- safeguarding the genome. This may also be true for protein com- tion. Total mRNA was alkaline fragmented and size selected. Both samples plexes that are responsible for organizing the higher-order struc- were processed for small RNA library sequencing. Libraries from two biological replicates per cell cycle phase were sequenced on an Illumina HiSeq ture of chromosomes, which also show coregulated patterns of 2000. translational control. The primary role of the cohesin and condensin complexes is to package the genome to ensure faithful Alignment and Analysis of Ribosome Profiling Data segregation of chromosomes during cell division (Hirano, 2012; Sequencing reads were processed and aligned to the human genome using Wood et al., 2010). The mRNAs comprising the majority of these standard procedures (see Supplemental Experimental Procedures). Transla- two complexes exhibit high levels of translation during both G1 tional regulation was inferred using an errors-in-variables regression model. and S phase that may ensure sufficient, stoichiometric amounts This analytical model estimates p values in each replicate to represent of the proteins required to package chromosomes prior to their ribosome-given-mRNA counts lower or higher than those expected. We simi- larly developed methods for combining p values among replicates within cell segregation. cycle phases and for testing differential translational regulation, all of which Interestingly, cohesins can promote transcription, and disrup- is described in greater detail in the Supplemental Experimental Procedures. tion of the cohesin complex impairs ribosomal RNA transcription, thus leading to defects in protein synthesis (Bose et al., 50 UTR Characterizations 2012). Most notably, mutations in cohesin genes characterize Genes were classified as having a level either higher than, lower than, or ex- an entire class of human disorders termed cohesinopathies. pected of ribosome occupancy given their mRNA levels within each phase Cohesinopathies manifest as developmental disorders with of the cell cycle as described in the Supplemental Experimental Procedures. 0 characteristic limb defects, including oligodactyly, and neurode- The length and %G+C content of the 5 UTR of each expressed gene was computed from Gencode version 11, while the predicted minimum free energy velopmental delay. These features overlap with the phenotypic was computed with the Vienna suite; RNAfold version 1.8.4 (Gruber et al., spectrum of ribosomopathies where ribosome components are 2008). The significance of each 50 UTR feature was determined with Wilcoxon found mutated, suggesting a possible relationship between rank sum test statistics. these two fundamental biological processes. Moreover, it is 0 notable that a mutation in a uORF in the 5 UTR of NIPBL leads Luciferase Reporter Assays to a decrease in NIPBL translation and is associated with a cohe- 50 UTRs from HeLa cDNA were cloned upstream of firefly luciferase in pGL3- sinopathy known as Cornelia de Lange Syndrome, suggesting T7. RNAs were in vitro transcribed using the T7 MEGAscript Kit (Life Technol- that alterations in translational regulation may underlie this hu- ogies). Purified RNA was capped using a vaccinia RNA capping system (New England Biolabs). Luciferase reporter RNA was transfected at a ratio of 20:1 man disease (Borck et al., 2006). This mutation disrupts a 0 with a control Renilla reporter RNA using the TransIT-mRNA Transfection Kit uORF in the 5 UTR of NIPBL, which could be responsible for (Mirus Bio). Luciferase levels were read after a 4 hr incubation using the Dual the observed decrease in translation. This finding is consistent Luciferase Assay (Promega). Firefly luciferase signal was normalized to with our studies showing a critical role for the 50 UTR in regulating Renilla signal for each sample, averaged, and the signal relative to G1 was the translation of mRNAs belonging to the cohesin complex. calculated. These results also suggest a potential feedback mechanism between the cohesin complex and the translation machinery that Western Blotting may be of great importance to the etiology of cohesinopathies. Proteins were analyzed by standard western blotting protocols. Antibodies used were as follows: RICTOR, AKT phospho-S473, total AKT, AKT phos- Together, our studies shed light on the unexpected dynamics pho-T308, total PKCa, and PROTOR-2 (Cell Signaling Technology); PKCa of translational control in the regulation of gene expression dur- phospho-S657 (Santa Cruz); PROTOR-1 (GeneTex); SIN1 (Bethyl Labora- ing fundamental cellular programs, such as the mammalian cell tories); phospho-p27 (R&D Systems); total p27 (BD Biosciences); histone H3 cycle. The magnitude of this translational regulation, involving phospho-S10 (Millipore); and a-tubulin (Sigma-Aldrich). hundreds of mRNAs, suggests that currently unknown regulatory mechanisms and transcript-specific translational regulators may ACCESSION NUMBERS endow remarkable specificity to the posttranscriptional gene expression program that is fundamental for accurate replication Sequencing data were deposited in the NCBI Short Read Archive under acces- and cell division. sion number SRA099816.

EXPERIMENTAL PROCEDURES SUPPLEMENTAL INFORMATION

Tissue Culture and Cell-Cycle Analysis Supplemental Information includes Supplemental Experimental Procedures, HeLa cells and MEFs were cultured under normal growth conditions. Synchro- three figures, and three tables and can be found with this article online at nization in specific cell cycle phases was achieved by release from thymidine http://dx.doi.org/10.1016/j.molcel.2013.09.018. block (G1 and S phase) and nocodazole treatment (mitosis) as previously described (Jackman and O’Connor, 2001). Cell-cycle analysis was performed ACKNOWLEDGMENTS on cells fixed in 80% ethanol prior to ribonuclease (RNase) digestion and stain- ing with 40 mg/ml propidium iodide or with the Click-iT EdU Flow Cytometry We thank Maria Barna and members of the Ruggero lab for critical insights dur- Assay (Life Technologies). ing this work and D. Byrd, K. Tong, and C. Milentis for reading the manuscript. Mark Magnuson provided the Rictor/ MEFs. C.R.S. is supported by NIH/ Library Preparation and Sequencing NRSA (F32CA162634). This work was supported by NIH R01CA140456 Next-generation sequencing libraries were prepared as described previously (D.R.), NIH R01CA154916 (D.R.), and NIH P30CA82103 (A.B.O.). B.S.T. is a (Hsieh et al., 2012). Briefly, synchronized cells were treated with cyclohexi- Prostate Cancer Foundation Young Investigator. D.R. is a Leukemia & Lym- mide, lysed, and split into pools for isolating total mRNA and ribosome-bound phoma Society Scholar.

8 Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. Please cite this article in press as: Stumpf et al., The Translational Landscape of the Mammalian Cell Cycle, Molecular Cell (2013), http://dx.doi.org/ 10.1016/j.molcel.2013.09.018 Molecular Cell The Translational Landscape during the Cell Cycle

Received: May 14, 2013 Kim, W., Bennett, E.J., Huttlin, E.L., Guo, A., Li, J., Possemato, A., Sowa, M.E., Revised: August 5, 2013 Rad, R., Rush, J., Comb, M.J., et al. (2011). Systematic and quantitative Accepted: September 17, 2013 assessment of the ubiquitin-modified proteome. Mol. Cell 44, 325–340. Published: October 10, 2013 Konrad, C.G. (1963). Protein Synthesis and Rna Synthesis during Mitosis in Animal Cells. J. Cell Biol. 19, 267–277. REFERENCES Kueng, S., Hegemann, B., Peters, B.H., Lipp, J.J., Schleiffer, A., Mechtler, K., and Peters, J.M. (2006). Wapl controls the dynamic association of cohesin with Alessi, D.R., Andjelkovic, M., Caudwell, B., Cron, P., Morrice, N., Cohen, P., chromatin. Cell 127, 955–967. and Hemmings, B.A. (1996). Mechanism of activation of protein kinase B by in- sulin and IGF-1. EMBO J. 15, 6541–6551. Lai, M.C., Chang, W.C., Shieh, S.Y., and Tarn, W.Y. (2010). DDX3 regulates cell growth through translational control of cyclin E1. Mol. Cell. Biol. 30, 5444– Alessi, D.R., James, S.R., Downes, C.P., Holmes, A.B., Gaffney, P.R., Reese, 5453. C.B., and Cohen, P. (1997). Characterization of a 3-phosphoinositide-dependent protein kinase which phosphorylates and activates protein kinase Laplante, M., and Sabatini, D.M. (2012). mTOR signaling in growth control and Balpha. Curr. Biol. 7, 261–269. disease. Cell 149, 274–293. Antonin, W., Ellenberg, J., and Dultz, E. (2008). Nuclear pore complex assem- Merbl, Y., Refour, P., Patel, H., Springer, M., and Kirschner, M.W. (2013). bly through the cell cycle: regulation and membrane organization. FEBS Lett. Profiling of ubiquitin-like modifications reveals features of mitotic control. 582, 2004–2016. Cell 152, 1160–1172. Borck, G., Zarhrate, M., Cluzeau, C., Bal, E., Bonnefont, J.P., Munnich, A., Olshen, A.B., Hsieh, A.C., Stumpf, C.R., Olshen, R.A., Ruggero, D., and Taylor, Cormier-Daire, V., and Colleaux, L. (2006). Father-to-daughter transmission B.S. (2013). Assessing gene-level translational control from ribosome profiling. of Cornelia de Lange syndrome caused by a mutation in the 50 untranslated re- Bioinformatics. Published online September 18, 2013. http://dx.doi.org/10. gion of the NIPBL Gene. Hum. Mutat. 27, 731–735. 1093/bioinformatics/btt533. Bose, T., Lee, K.K., Lu, S., Xu, B., Harris, B., Slaughter, B., Unruh, J., Garrett, Peters, J.M. (2006). The anaphase promoting complex/cyclosome: a machine A., McDowell, W., Box, A., et al. (2012). Cohesin proteins promote ribosomal designed to destroy. Nat. Rev. Mol. Cell Biol. 7, 644–656. RNA production and protein translation in yeast and human cells. PLoS Pulikkan, J.A., Dengler, V., Peramangalam, P.S., Peer Zada, A.A., Mu¨ ller- Genet. 8, e1002749. Tidow, C., Bohlander, S.K., Tenen, D.G., and Behre, G. (2010). Cell-cycle regu- Cho, R.J., Huang, M., Campbell, M.J., Dong, H., Steinmetz, L., Sapinoso, L., lator E2F1 and microRNA-223 comprise an autoregulatory negative feedback Hampton, G., Elledge, S.J., Davis, R.W., and Lockhart, D.J. (2001). loop in acute myeloid leukemia. Blood 115, 1768–1778. Transcriptional regulation and function during the human cell cycle. Nat. Qin, X., and Sarnow, P. (2004). Preferential translation of internal ribosome en- Genet. 27, 48–54. try site-containing mRNAs during the mitotic cycle in mammalian cells. J. Biol. Fan, H., and Penman, S. (1970). Regulation of protein synthesis in mammalian Chem. 279, 13721–13728. cells. II. Inhibition of protein synthesis at the level of initiation during mitosis. Shin, I., Yakes, F.M., Rojo, F., Shin, N.Y., Bakin, A.V., Baselga, J., and Arteaga, J. Mol. Biol. 50, 655–670. C.L. (2002). PKB/Akt mediates cell-cycle progression by phosphorylation of Gandhi, R., Gillespie, P.J., and Hirano, T. (2006). Human Wapl is a cohesin- p27(Kip1) at threonine 157 and modulation of its cellular localization. Nat. binding protein that promotes sister-chromatid resolution in mitotic prophase. Med. 8, 1145–1152. Curr. Biol. 16, 2406–2417. Shiota, C., Woo, J.T., Lindner, J., Shelton, K.D., and Magnuson, M.A. (2006). Gruber, A.R., Lorenz, R., Bernhart, S.H., Neubo¨ ck, R., and Hofacker, I.L. Multiallelic disruption of the rictor gene in mice reveals that mTOR complex (2008). The Vienna RNA websuite. Nucleic Acids Res. 36(Web Server issue), 2 is essential for fetal growth and viability. Dev. Cell 11, 583–589. W70–W74. Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Hirano, T. (2012). Condensins: universal organizers of chromosomes with Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and diverse functions. Genes Dev. 26, 1659–1678. Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based Hsieh, A.C., Liu, Y., Edlind, M.P., Ingolia, N.T., Janes, M.R., Sher, A., Shi, E.Y., approach for interpreting genome-wide expression profiles. Proc. Natl. Stumpf, C.R., Christensen, C., Bonham, M.J., et al. (2012). The translational Acad. Sci. USA 102, 15545–15550. landscape of mTOR signalling steers cancer initiation and metastasis. Whitfield, M.L., Sherlock, G., Saldanha, A.J., Murray, J.I., Ball, C.A., Alexander, Nature 485, 55–61. K.E., Matese, J.C., Perou, C.M., Hurt, M.M., Brown, P.O., and Botstein, D. Ingolia, N.T., Ghaemmaghami, S., Newman, J.R., and Weissman, J.S. (2009). (2002). Identification of genes periodically expressed in the human cell cycle Genome-wide analysis in vivo of translation with nucleotide resolution using and their expression in tumors. Mol. Biol. Cell 13, 1977–2000. ribosome profiling. Science 324, 218–223. Wood, A.J., Severson, A.F., and Meyer, B.J. (2010). Condensin and cohesin Jackman, J., and O’Connor, P.M. (2001). Methods for synchronizing cells at complexity: the expanding repertoire of functions. Nat. Rev. Genet. 11, specific stages of the cell cycle. Curr. Protoc. Cell Biol. Chapter 8,3. 391–404.

Molecular Cell 52, 1–9, November 21, 2013 ª2013 Elsevier Inc. 9 Molecular Cell, Volume 52

Supplemental Information

The Translational Landscape of the Mammalian Cell Cycle

Craig R. Stumpf, Melissa V. Moreno, Adam B. Olshen, Barry S. Taylor, and Davide Ruggero Supplemental Materials

Figure S1 A 100 G1 82% S-phase 89% 80 Mitosis 73%

x 60 Ma f o

% 40

0 0 200 400 600 800 1000 PI intensity B Translation G1 S M G1 S M Genes (n=1255)

-1.75 1.75 C 0

** *** *** *** *** *** 10000 1.0 *** 0.0 ***

0.8 p<0.001 1000 ** -0.2 *** p<0.0001

0.6 G1 100 S-phase -0.4 Mitosis 0.4 5' UTR G+C%

5' UTR length (bp) 10 0.2 -0.6 High Low Background High Low Background High Low Background High Low Background High Low Background High Low Background High Low Background High Low Background High Low Background

1 0.0 -0.8

Figure S1. Related to Figure 1. (A) Flow cytometry histogram plotting the distribution of cells as a function of DNA content measured by propidium iodide (PI) fluorescence intensity. (B) Heat maps representing the translational regulation

(left) and mRNA expression levels (right) for the 1255 translationally regulated genes within any phase of the cell cycle. Genes and cell cycle phases are clustered based on the level of normalized ribosome occupancy (mean-centered translational efficiency by gene, left). mRNA expression levels (right) were ordered based on the clustering of translational regulation. Scales and colors indicate the direction and magnitude of mRNA translation or levels, respectively. (C) Box and whisker plots denoting the 5’UTR length (left), %G+C content (middle), and minimum free energy

(MFE) (right) for genes with higher, lower, or expected levels of ribosome occupancy. Y-axes are as labeled. Cell cycle phases are colored: Grey, G1; Blue, S- phase; Green, mitosis. Figure S2

A RICTOR 5’ UTR sequence: GUUUCCGGUGUUGUGACUGAAACCCGUCAAUAUG

G1 S-phaseMitosis

SIN1 PROTOR-1 PROTOR-2 p-AKT T308 total AKT p-p27Kip1 total p27Kip1 Tubulin C

G1 G1 S-phaseS-phaseMitosisMitosis RICTOR: + - + - + - p-AKT S473

total AKT

Tubulin

D Percent of cells 0 20 40 60 80 100

rictor-/- G1 S-phase Mitosis p-value = 0.0006

Figure S2. Related to Figure 2. (A) The sequence of the RICTOR 5’UTR. The start codon for the coding region is highlighted in red. (B) Western blots showing the levels of mTORC2 components and the phosphorylation status of AKT and downstream signaling targets. SIN1, PROTOR-1, and PROTOR-2 are components of mTORC2. Phosphorylation of AKT T308 is independent of mTORC2. p27Kip1 is a phosphorylation target of AKT. Tubulin is a loading control. (C) Western blots illustrating the phosphorylation status of AKT in synchronized wildtype and Rictor-/-

MEFs. (D) Cell cycle analysis of asynchronous MEFs with the percentage of cells in each phase of the cell cycle indicated along the Y-axis. The red box highlights the magnitude of the difference between the number of WT and Rictor-/- MEFs in S- phase. Probability calculated by Student’s t-test. Cell cycle phases are colored as labeled.

Figure S3

A NIPBL 5’UTR sequence: GGGCCCGCGGCGAUGCCCCCCCGGUAGCUCGGGCCCGUGGUCGGGUGUUUGUGAGUGU UUCUAUGUGGGAGAAGGAGGAGGAGGAGGAAGAAGAAGCAACGAUUUGUCUUCUCGGCUG GUCUCCCCCCGGCUCUACAUGUUCCCCGCACUGAGGAGACGGAAGAGGAGCCGUAGCCAC CCCCCCUCCCGGCCCGGAUUAUAGUCUCUCGCCACAGCGGCCUCGGCCUCCCCUUGGAUU CAGACGCCGAUUCGCCCAGUGUUUGGGAAAUGGGAAGUAAUGACAGCUGGCACCUGAACU AAGUACUUUUAUAGGCAACACCAUUCCAGAAAUUCAGGAUG...

B WAPAL 3’UTR sequence: CTGCTTTACCTTTGCTTCAGGTGCTCGGTAATGCTGGAGCTATCCTTAGACAAAGAAAAGTCAAGTC ATGAAAGAAGTCCTTGAAGATATACCAAGAACATTCATCAGTATCATTCGTGTTTGGATTTTTAA...

Putative amino acid sequence of C-terminal extension: LLYLCFRCSVMLELSLDKEKSSHERSP*RYTKNIHQYHSCLDR*

Figure S3. Related to Figure 4. (A) The sequence of the NIPBL 5’UTR. Potential

AUG start codons are bold. The uORF corresponding to the peak of bound ribosome in Figure 4C is orange. The region of the CC > A mutation in Cornelia de Lange syndrome is underlined. (B) The sequence of the WAPAL 3’UTR immediately downstream of the stop codon. The region corresponding to the peak of ribosome binding is underlined. In-frame stop codons are shown in red. The predicted amino acid sequence of the putative C-terminal extension is shown below. The sequence up to the first in-frame stop codon is red.

Supplemental Tables

Table S1. Translationally regulated genes from each phase of the cell cycle, related to Figure 1.

Table S2. Genes with significant changes in translation between G1 and S-phase, G1 and mitosis, and S-phase and mitosis, related to Figure 1.

Table S3. Functional enrichments of genes that are translationally regulated during

G1, S-phase, or mitosis, related to Figure 3.

Supplemental Experimental Procedures

Alignment and analysis of ribosome profiling data

Raw single-end sequencing reads were processed in the following manner. First, reads were clipped of the known adaptor sequence (CTGTAGGCACCATCAAT) and remaining 3’ sequence with the FASTX-Toolkit, retaining only reads of 24bp post- clipping length or longer. Clipped reads were mapped to the human genome assembly (NCBI build 36) with Burrows-Wheeler Alignment (BWA; ver. 0.5.9-r16, default parameters)(Li and Durbin, 2009). Unaligned reads were then mapped to known splice junctions with Tophat (ver. 1.4.1) (Trapnell et al., 2009) using a transcriptome index created from version 11 of the Gencode gene set (Harrow et al.,

2006). Provisional merged BAM files (Li et al., 2009) were created from aligned and unaligned reads from either the genome or spliced alignment and read groups were enforced across lanes and experiments. BAM files were then indexed, coordinate sorted, and had alignment metrics determined, all with the Picard suite

(http://picard.sourceforge.net/). Only those protein-coding transcripts of levels 1 or 2 of Gencode support (verified or manually curated loci respectively) were used for all subsequent analyses. A unified gene model was created for each gene in which the union of coding sequences across all isoforms of a given gene was collapsed into a single model per gene. Gene-level translational regulation was inferred for only those genes estimated to be expressed in the Hela transcriptome, determined by comparing their observed expression level (calculated as a normalized rpkM expression estimate) to a null distribution of similar expression in randomly selected non-coding regions of the genome with a size distribution sampled from the empirical size distribution (Olshen et al. submitted). In total,

10,814 genes were retained for further analysis.

We quantify mRNA translation for genes (ribosome-given-mRNA levels) with a statistical framework called Babel. Briefly, this errors-in-variables regression model, the details of which are presented elsewhere (Olshen et al.), assesses the significance of translational regulation for every gene within and between samples and conditions. For a transcript with a given abundance (mRNA counts), the expected level of ribosome occupancy (RPF counts) is inferred from trimmed least- squares regression (based on the monotonic relationship typified in Figure 1). We model both mRNA levels and the level of bound ribosome given mRNA abundance by the negative binomial distribution. The over-dispersion of the latter is modeled using an iterative algorithm to prevent over-fitting. We then use a parametric bootstrap based on the modeling described above to estimate a p-value for every gene. It tests formally the null hypothesis that the level of bound ribosome is as expected from mRNA abundance. These p-values are estimated for a one-sided test in which both low and high p-values are of interest; low p-values correspond to higher than expected ribosome-given-mRNA counts and high p-values correspond to the opposite.

Since Babel quantifies this mRNA translation (ribosome-given-mRNA levels) as a p- value in each sample of a given condition, we combine these p-values into a single assessment of the significance of translational regulation in each condition (phase of the cell cycle). Here, we developed an alternative methodology to Fisher’s method first proposed by Edgington (Edgington, 1972). This approach, in which the combined p-value is a function of the arithmetic mean of individual p-values, maintains symmetry and convexity between each of two or more tests of hypotheses and their p-values. It produces a small combined p-value from consistently small p-values of ribosome-given-mRNA levels (and the reverse for two or more large p-values). Once p-values are combined, two-sided p-values can be estimated and genes are considered regulated at the translational level if their corresponding p-values are low after adjusting for multiple comparisons [in this study, those corresponding to a false discovery rate (FDR) less than 1% (Storey,

2002)].

Differences in ribosome occupancy given mRNA levels between conditions are assessed utilizing a statistic based upon using the Gaussian quantile function to transform the within condition p-values into z-statistics. Corresponding p-values are estimated by comparison to the Gaussian distribution and again corrected for multiple comparisons as described above. Significant differentially translationally regulated genes are those corresponding to an FDR < 5% in any of the pair-wise comparisons of cell cycle phases. For visualization purposes, the conventional measure of translational efficiency was used and calculated as previously described

(Ingolia et al., 2009).

Supplemental References

Edgington, E.S. (1972). Additive Method for Combining Probability Values from

Independent Experiments. J Psychol 80, 351-&.

Harrow, J., Denoeud, F., Frankish, A., Reymond, A., Chen, C.K., Chrast, J., Lagarde, J.,

Gilbert, J.G., Storey, R., Swarbreck, D., et al. (2006). GENCODE: producing a reference annotation for ENCODE. Genome Biol 7 Suppl 1, S4 1-9.

Ingolia, N.T., Ghaemmaghami, S., Newman, J.R., and Weissman, J.S. (2009). Genome- wide analysis in vivo of translation with nucleotide resolution using ribosome profiling. Science 324, 218-223.

Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with Burrows-

Wheeler transform. Bioinformatics 25, 1754-1760.

Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis,

G., and Durbin, R. (2009). The Sequence Alignment/Map format and SAMtools.

Bioinformatics 25, 2078-2079.

Olshen, A.O., Hsieh, A.C., Stumpf, C.R., Olshen, R.A., Ruggero, D., and Taylor, B.S.

(2013). Assessing gene-level translational control from ribosome profiling.

Bioinformatics (in press).

Storey, J.D. (2002). A direct approach to false discovery rates. J Roy Stat Soc B 64,

479-498. Trapnell, C., Pachter, L., and Salzberg, S.L. (2009). TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 1105-1111.