The synthesis of cellular proteins by mRNA translation is an essential process regulated by multiple interactions between cis-acting sequences and trans-acting factors. Translation initiation is highly regulated to control the rate of protein synthesis. During canonical 5’ cap-dependent initiation, pre-initiation complexes (PICs) assemble at mRNA 5’ ends and scan directionally in search of start codons (Hinnebusch et al., 2016). Due to this directional scanning, mRNA sequences and structures in 5’ transcript leaders have a direct impact on initiation frequency. In particular, upstream Open Reading Frames (uORFs), short coding sequences between the 5’ cap and primary protein coding sequence (CDS), generally decrease the frequency of initiation at downstream CDSs (Wethmar, 2014). Termination at uORF stop codons can also induce nonsense mediated decay (NMD), a major mRNA turnover pathway. As a result, most uORFs are expected to reduce gene expression. Despite this general view, the extent of uORF repression can vary greatly and the underlying causes for this remain unclear.

Despite their general repressive nature, some uORFs enhance expression of their downstream CDS, especially in response to stress (Young and Wek, 2016). A classic example of a gene harboring enhancer uORFs is yeast GCN4, whose transcript leader harbors four uORFs (Mueller and Hinnebusch, 1986). After translation of uORF1, the small subunit of the ribosome remains attached to GCN4 mRNA, reforming a PIC that subsequently resumes scanning. In the absence of stress, the resulting PIC is rapidly recharged with tRNA-Met in a ternary complex with eIF2 and GTP, which leads to translation of uORFs 2–4 and a corresponding inhibition of GCN4 CDS translation. However, most stress conditions cause phosphorylation of the eIF2α subunit, which allows rescanning PICs to scan past uORFs 2–4 to translate the GCN4 CDS (Hinnebusch, 2005). A similar multi-uORF reinitiation system allows stress-dependent translation of mammalian ATF4, in which resumption of scanning after translating a short uORF allows subsequent leaky scanning past a more repressive uORF (Vattem and Wek, 2004). In theory, the same mechanism could allow short uORFs to insulate against longer, more repressive uORFs even in the absence of stress (Lin et al., 2019). However, the frequency of such stress-independent uORF enhancers remains unknown.

Although once thought to be relatively rare, genomic studies have revealed that AUG-initiated uORFs are common, affecting ~15% of yeast (Ingolia et al., 2009; McManus et al., 2014; Zhang and Dietrich, 2005a) and ~50% of human genes (McGillivray et al., 2018; Wethmar et al., 2014). In addition to these canonical elements, ribosome profiling studies using drugs that stall initiating ribosomes, have identified even more uORFs that initiate at non-AUG codons (Brar et al., 2011; Ingolia et al., 2011; Spealman et al., 2018). However, other work suggests the drugs used in those studies may exaggerate the frequency of uORF usage (Gerashchenko and Gladyshev, 2014; Kearse et al., 2019; Lareau et al., 2014). Despite the large number of predicted AUG and non-AUG uORFs, most studies focus solely on identifying these elements and few have been experimentally tested (Wethmar et al., 2014). The few non-AUG uORFs that have been tested had relatively modest influences on expression (Spealman et al., 2018; Zhang and Hinnebusch, 2011). Thus, the relative influence of AUG- and non-AUG uORFs on mRNA translation has not been systematically evaluated.

Due to their frequency and functional importance, uORF evolution has been the subject of multiple studies (Zhang et al., 2019). Early work identified 38 genes with AUG-initiated uORFs that were deeply conserved among yeast species by stringent comparisons of draft genome sequences (Zhang and Dietrich, 2005a). Functional evaluations of nine S. cerevisiae uORFs showed six altered the expression in a luciferase reporter assay. More recently, genome-wide studies using ribosome profiling identified thousands of AUG and non-AUG uORFs in multiple Saccharomyces species, some of which were translated in multiple species (Spealman et al., 2018). Ribosome profiling was also used to identify uORFs used throughout Drosophila development, many of which appear to be adaptive due to signs of positive selection detected by comparative genomics (Zhang et al., 2018). Additional analyses of human, mouse, and zebrafish ribosome profiling data estimated uORF regulatory activity using the ratio (U/C) of ribosome footprints in 5’ transcript leaders (UTRs) to footprints in coding sequences (CDS) (Chew et al., 2016; Johnstone et al., 2016). These studies found modest, though statistically significant correlations of U/C ratios in data from mouse and human tissue culture cells and bulk brain tissue, suggesting that uORF activities are somewhat conserved in vertebrates. However, to our knowledge, the regulatory functions of homologous uORFs from different species have not been experimentally compared. Thus, the extent to which uORF functions are quantitatively conserved across species remains unclear.

Currently, the magnitude of uORF activity is thought to depend primarily on the extent to which uORF start codons match the Kozak consensus sequence (Cuperus et al., 2017; Dvir et al., 2013; Noderer et al., 2014; Sample et al., 2019). Analyses of ribosome profiling data from vertebrates found that the U/C ratio was higher for uORFs with start codons in strong Kozak contexts (Chew et al., 2016; Johnstone et al., 2016). Massively parallel reporter assays (MPRAs) of random transcript leaders confirmed that AUG-uORFs in strong Kozak contexts are repressive (Cuperus et al., 2017; Dvir et al., 2013; Ferreira et al., 2014; Noderer et al., 2014; Sample et al., 2019). However, other features have also been shown to affect uORF activity. For example, a uORF-specific MPRA (FACS-uORF) study found rare codons and dicodons within a uORF can determine whether the uORF enhances or represses expression of the CDS (Lin et al., 2019). The structural accessibility of start codons was also found to play a key role in determining the impact of uORFs from the human α–1-antitrypsin mRNA leader (Corley et al., 2017). Other work has shown that uORF activity can depend on the location of uORFs relative to main protein-coding ORF start codons (Beznosková et al., 2013; Grant et al., 2012). However, the relative influences of these features on natural uORF activities have not been determined, primarily because few such uORFs have been functionally assayed.

Similarly, many questions remain regarding uORFs and NMD. The extent to which NMD contributes to typical uORF-mediated gene repression remain unclear, as do the roles of uORF features in determining NMD susceptibility. While uORFs are clearly correlated with NMD (Celik et al., 2017; Smith et al., 2014), the yeast GCN4 and YAP1 uORFs appear to resist NMD (Ruiz-Echevarría and Peltz, 2000). Early studies suggested yeast NMD requires AU-rich cis-acting downstream sequence elements bound by Hrp1p (Culbertson and Leeds, 2003; Czaplinski et al., 1999; González et al., 2000; Peltz et al., 1993; Ruiz-Echevarría et al., 1998); however, such elements are not well defined and appear to be missing from many NMD targets (Meaux et al., 2008). Other work found premature termination codons (PTCs) were less likely to induce NMD when inserted closer to the 5’ end of a yeast PGK1 mRNA (Muhlrad and Parker, 1999). NMD induction was also shown to be greatly reduced by low-frequency stop codon readthrough at PTCs (Keeling et al., 2004); however, high-frequency readthrough did not prevent NMD in another context (Gorgoni et al., 2019). Other studies showed mRNA can be protected from NMD by positioning poly(A)-binding protein close to the termination codon (Amrani et al., 2004; Eberle et al., 2008). Additional work in human cells found translation initiation factors inhibit NMD, as does reinitiation after termination at PTCs (Lindeboom et al., 2016; Raimondeau et al., 2018; Zhang and Maquat, 1997). While these studies helped explain NMD induction by PTCs in coding genes, most did not consider uORFs, which may behave differently due to their extreme 5’ locations in transcript leaders and generally short lengths. Thus, the features that influence uORF induction of NMD remain unclear. Furthermore, the relative importance of NMD and translation inhibition in natural uORF activity has not been systematically evaluated.

Here, we used two MPRA systems, FACS-uORF and PoLib-seq, to quantify the impact of thousands of natural yeast AUG- and non-AUG uORFs on protein expression and ribosome loading. Our results show that most non-AUG uORFs have small impacts on expression compared to AUG-uORFs. Leveraging the massive scale of our results, we evaluated the influence of sequence and positional features of natural AUG-uORFs on their gene regulatory effects. Using a strain deleted of the NMD factor UPF1 (∆upf1), we showed NMD accounts for roughly a third of uORF repression in yeast, and further investigated how uORF features impact the propensity for NMD. Finally, we used elastic-net regression modeling to select and weight features that correlate with uORF activity, revealing that uORF locations are as predictive for uORF function as Kozak contexts. Surprisingly, we found uORF activity often depends on the site of transcription initiation, as alternative transcription start sites can dramatically change the magnitude of uORF repression.

Thousands of AUG and non-AUG uORFs have been identified from ribosome profiling studies in all major model organisms (Kearse and Wilusz, 2017). Despite concerns that cycloheximide and other drugs used in ribosome profiling may exaggerate occupancy on transcript leaders (Gerashchenko and Gladyshev, 2014; Kearse et al., 2019; Lareau et al., 2014), very few natural uORFs have been experimentally tested because traditional uORF assays are performed on a gene-by-gene basis (Wethmar et al., 2014). Here, we used two Massively Parallel Reporter Assay systems, FACS-uORF and PoLib-seq, to quantify the impact of thousands of yeast AUG- and non-AUG uORFs on reporter expression and translation efficiency in wildtype and NMD-deficient yeast strains. We leveraged the resulting data to investigate the importance of multiple uORF features on their regulatory functions. Our analyses shed new light on uORF functions and have multiple implications.

It has long been recognized that eukaryotic translation can initiate at non-AUG (so-called) near-cognate codons (Kearse and Wilusz, 2017; Kozak, 1991a; Tang et al., 2004). Ribosome profiling studies also support the translation of many non-AUG uORFs in yeast (Brar et al., 2011; Eisenberg et al., 2020; Ingolia et al., 2009; Spealman et al., 2018). However, the few non-AUG uORFs that have been experimentally tested had relatively modest regulatory effects in reporter systems (Spealman et al., 2018; Zhang and Hinnebusch, 2011). Our results show non-AUG uORFs have relatively mild impacts on gene expression. Furthermore, a considerable number of non-AUG uORFs whose mutation significantly changed reporter expression may not be translated, as insertion of the stalling CGA dicodon sequence did not reduce expression. Thus, at least during log-phase growth, many non-AUG uORFs appear to have little influence on gene expression, consistent with the relatively low rate of translation initiation at near-cognate start codons (Kolitz et al., 2009; Takacs et al., 2011). As non-AUG uORFs have been predicted in many other species, our results suggest that most of them may also play limited roles in regulating expression. However, non-AUG uORFs with start codons in strong Kozak contexts (Diaz de Arce et al., 2017) with accompanying downstream mRNA structure (Kozak, 1990) may more have more substantial effects.

Although most uORFs are expected to repress expression, examples of enhancer uORFs have been identified in multiple species. Most notably, the short upstream uORFs in GCN4 (yeast) and ATF4 (metazoans) have been shown to cause stress-dependent upregulation of main ORF translation (Hinnebusch, 2005; Vattem and Wek, 2004). This upregulation results from delayed reinitiation under stress, such that PICs that resume scanning bypass more repressive uORFs. In our system, only two percent of uORFs increased expression. Notably, the oligonucleotide synthesis technology used to generate our reporter library is limited in length, such that the longest leaders that we tested were 180 nucleotides. While this included most yeast transcript leaders, longer leaders may have more enhancers. As such, our numbers may underestimate the frequency of enhancers. Because most of the enhancers we found were in multi-uORF transcript leaders, we propose that, much like GCN4 uORF1, they function by reducing usage of other, more repressive uORFs. Although we performed our experiments under unstressed conditions, reinitiation might still allow leaky scanning past more repressive uORFs near the enhancer stop codon. In other cases, translation of enhancer uORFs might alter transcript leader structure or modulate scanning by upstream PICs. Regardless of the mechanisms used, our data indicate that uORFs rarely act as enhancers, at least under the conditions we assayed.

It is well known that uORF activity depends on the strength of start codon recognition. However, the influence of other factors has been less clear. Our computational modeling of a massive set of yeast uORFs compared the predictive power of these and other features on uORF activity. While the correlation of Kozak context and uORF activity was strong, surprisingly, uORF location had similar predictive power. Because PIC assembly involves unwinding the mRNA around the 5’cap, the increased repression we observed for uORFs that start near the cap might result from increased start codon accessibility. Cap-proximal uORFs might also sterically block the loading of additional PICs. Previous work in yeast found that cap-proximal start codons were often skipped (Arribere and Gilbert, 2013); however, kozak context and RNA structure were not considered. Indeed, a similar tendency to skip cap-proximal start codons in wheat germ extracts could be overcome by the presence of an RNA stemloop downstream (Kozak, 1991b). We also observed stronger repression for longer uORFs and uORFs that terminate closer to the main ORF start codon. Such uORFs may permit less reinitiation at the main ORF, as ribosomes lose initiation factors after translating PICs that resume scanning would have less time to reacquire ternary complex before encountering the main ORF start codon. If so, our results suggest translation reinitiation is more common in yeast than currently appreciated (Gunišová et al., 2017). Future work is needed to evaluate these and other potential mechanisms underlying the importance of uORF location.

Despite the insights gained by our ENR modeling, a large amount of the variance among uORFs cannot currently be explained. This suggests that other features may have important influences on uORF activity. For example, previous work found that structural accessibility of uORF start codons, as measured by SHAPE probing, led to accurate estimates of uORF activity for the human antitrypsin-alpha gene (Corley et al., 2017). Based on this work, it was proposed that ribosomes often shunt past uORFs whose start codons are occluded in stable RNA structures (Mustoe et al., 2018). Although intriguing, this model seems to negate the role of helicases (e.g. DED1 and eIF4A) which unwind such structures during PIC scanning (Guenther et al., 2018; Sen et al., 2019; Sen et al., 2015; Sharma and Jankowsky, 2014). While we included predicted energies of unwinding around the start codon, our purely computational structural predictions contributed only minimally to our models. Given the large amount of uncertainty surrounding computational estimates of RNA structure, we expect that detailed measurements of structural accessibility would be beneficial to future modeling.

By comparing the magnitude of uORF repression in wildtype and upf1∆ yeast, we found the contribution of NMD to uORF repression (%NMD) ranges from 0 to 100%, with a median of 35%. Our work also provides insight into how uORF features impact their ability to induce NMD. Reinitiation after uORF translation can protect human mRNA from NMD (Lindeboom et al., 2016; Zhang and Maquat, 1997), and short uORFs are known to reinitiate more efficiently than long ones (Gunišová et al., 2017, Gunišová et al., 2016) likely due to increased retention of initiation factors (Bohlen et al., 2020; Wagner et al., 2020). Consistent with this, we found short uORFs induce less NMD than long uORFs. We also found NMD induction was lowest for uORFs terminating in UGAC. Since this stop codon context allows high rates of readthrough, it is possible such readthrough protects mRNA from uORF-induced NMD consistent with previous work (Keeling et al., 2004). Notably, AU-rich elements downstream of stop codons were associated with stronger uORF repression in wildtype, but not upf1∆ yeast. Such sequences may function as NMD-inducing cis-elements formerly thought to promote NMD by binding Hrp1p (Culbertson and Leeds, 2003; Czaplinski et al., 1999; González et al., 2000; Peltz et al., 1993; Ruiz-Echevarría et al., 1998). Thus, our work supports significant roles for many sequence features in determining uORF NMD induction.

We also investigated the relationship between uORF location in transcript leaders and NMD induction. Previous work using synthetic NMD targets found that NMD was more efficient when stop codons were located closer to the 5’ end of the PGK1 coding sequence (Cao and Parker, 2003). We found the opposite relationship among uORF stop codons, such that stop codons were less likely to induce decay the closer they were located to the 5’ cap. Although these results at first appear discordant, this may reflect much larger distances between the cap and stop codon in this study (>66, 426, 675, and 957 nucleotides) than found in our library (0–150 nt). It is also possible that the effect of stop location we observed may instead reflect uORF length, as longer uORFs terminate, on average, farther from the transcription start site than shorter ones. Thus, while our work suggests uORF position may also be an important feature impacting the efficiency of NMD, other features may underly this correlation.

Eukaryotic transcription often initiates at alternative sites, even in the relatively simple yeast (Lu and Lin, 2019; Pelechano et al., 2013; Zhang and Dietrich, 2005b). Such alternative transcription start sites alter the translation efficiency of their downstream genes (Arribere and Gilbert, 2013; Rojas-Duran and Gilbert, 2012; Zydowicz-Machtel et al., 2018). However, the extent to which uORF function depends on transcription start site usage has not been previously evaluated to our knowledge. Our results indicate that uORF activity often varies with alternative transcription start site usage. The dependence of uORF activity on transcript start sites may impart different translation efficiency and turnover rates for alternative transcript isoforms. This was broadly consistent with our modeling data, in that uORFs were more repressive when present in transcript leaders that initiated closer to the uORF start codon. If uORF start codons near the 5’ cap are more structurally accessible, uORFs that initiate further downstream may be bypassed by ribosomal shunting, as occurs in the Cauliflower Mosaic Virus uORF and has been proposed recently to be more common (Corley et al., 2017; Mustoe et al., 2018; Ryabova and Hohn, 2000). Alternatively, PIC scanning may become more processive over time, such that more distal start codons are more easily skipped. Regardless of the underlying mechanisms, the dependence of uORF activity on location implies that caution must be taken when studying uORFs using ribosome profiling data because current methods do not connect uORF occupancy with specific transcript isoforms.

Finally, our study included the first direct comparisons of gene regulation from homologous uORFs from two species in their corresponding, native transcript leaders. We found the magnitude of uORF regulation varied substantially for many uORFs, with some examples consistent with changes in uORF features. Our results also show a conserved uORF in yeast ECM7 confers strong repression in both S. cerevisiae and S. paradoxus. Although previous work reported the uORF had no impact on regulation (Zhang and Dietrich, 2005a), this may have been due to non-physiological transcription initiation site usage. The luciferase reporter used in the previous study was driven by a different promoter which may have initiated transcription at alternative site(s). Earlier studies comparing ribosome profiling data from zebrafish, mouse, and human cell lines indicated that conserved uORFs tended to have weaker Kozak sequences than non-conserved uORFs (Chew et al., 2016; Johnstone et al., 2016). They proposed that this allowed conserved uORFs to be activated by trans-acting factors. In addition, conserved uORFs were associated with smaller decreases in ribosome occupancy on downstream main ORFs. By directly assaying uORF effects, we found that conserved uORFs indeed have more modest effects on gene expression than non-conserved uORFs. We also found uORFs with conserved start codons had weaker Kozak sequences and were located further from the transcription start site. These results suggest that both uORF Kozak context and location are subject to purifying selection, underscoring the importance of uORF position we identified by computational modeling. Future work could more fully investigate the evolution of yeast uORFs by evaluating the frequency of uORF-related mutations in the 1000 S. cerevisiae genome project data (Peter et al., 2018). However, such analysis would carry the caveat that the locations of transcript leaders in these strains have not been experimentally determined.

In providing the first, to our knowledge, high-throughput functional analysis of natural uORF regulatory activities, our work defines the range of uORF activity and reveals location is as important as Kozak strength in determining yeast uORF functions. Recent work has also highlighted the large number of human uORFs (Barbosa et al., 2013; Calvo et al., 2009; Lee et al., 2020; McGillivray et al., 2018; Wethmar et al., 2010). While many aspects of translation initiation are deeply conserved, the mechanisms involved in transcript leader scanning may differ substantially in human cells and tissues. Thus, studies using similar methods in human cells are needed to characterize human uORF regulatory functions, and further understand their potential involvement in cellular stress responses.

Sequencing data have been deposited in NCBI SRA under accession PRJNA721222.

1. 1. May GE
   2. Akirtava C
   3. Matthew Agar-Johnson M
   4. Micic J
   5. Woolford J
   6. McManus J

   (2021) NCBI BioProject

   ID PRJNA721222. Saccharomyces cerevisiae translation massively parallel reporter assay raw reads.

   https://www.ncbi.nlm.nih.gov/bioproject/PRJNA721222

1. 1. Amrani N
   2. Ganesan R
   3. Kervestin S
   4. Mangus DA
   5. Ghosh S
   6. Jacobson A

   (2004) A Faux 3′-UTR promotes aberrant termination and triggers nonsense-mediated mRNA decay

   Nature 432:112–118.

   https://doi.org/10.1038/nature03060

   • PubMed
   • Google Scholar
2. 1. Arribere JA
   2. Gilbert WV

   (2013) Roles for transcript leaders in translation and mRNA decay revealed by transcript leader sequencing

   Genome Research 23:977–987.

   https://doi.org/10.1101/gr.150342.112

   • Google Scholar
3. 1. Artieri CG
   2. Fraser HB

   (2014) Accounting for biases in Riboprofiling data indicates a major role for Proline in stalling translation

   Genome Research 24:2011–2021.

   https://doi.org/10.1101/gr.175893.114

   • Google Scholar
4. 1. Barbosa C
   2. Peixeiro I
   3. Romão L
   4. Fisher EMC

   (2013) Gene expression regulation by upstream open reading frames and human disease

   PLOS Genetics 9:e1003529.

   https://doi.org/10.1371/journal.pgen.1003529

   • PubMed
   • Google Scholar
5. 1. Beznosková P
   2. Cuchalová L
   3. Wagner S
   4. Shoemaker CJ
   5. Gunišová S
   6. von der Haar T
   7. Valášek LS

   (2013) Translation initiation factors Eif3 and Hcr1 control translation termination and stop Codon read-through in yeast cells

   PLOS Genetics 9:e1003962.

   https://doi.org/10.1371/journal.pgen.1003962

   • PubMed
   • Google Scholar
6. 1. Bohlen J
   2. Fenzl K
   3. Kramer G
   4. Bukau B
   5. Teleman AA

   (2020) Selective 40s Footprinting reveals cap-tethered Ribosome scanning in human cells

   Molecular Cell 79:561–574.

   https://doi.org/10.1016/j.molcel.2020.06.005

   • PubMed
   • Google Scholar
7. 1. Bonetti B
   2. Fu L
   3. Moon J
   4. Bedwell DM

   (1995) The efficiency of translation termination is determined by a synergistic interplay between upstream and downstream sequences in Saccharomyces cerevisiae

   Journal of Molecular Biology 251:334–345.

   https://doi.org/10.1006/jmbi.1995.0438

   • PubMed
   • Google Scholar
8. 1. Brar GA
   2. Yassour M
   3. Friedman N
   4. Regev A
   5. Ingolia NT
   6. Weissman JS

   (2011) High-resolution view of the yeast Meiotic program revealed by Ribosome profiling

   Science 335:552–557.

   https://doi.org/10.1126/science.1215110

   • PubMed
   • Google Scholar
9. 1. Brown A
   2. Shao S
   3. Murray J
   4. Hegde RS
   5. Ramakrishnan V

   (2015) Structural basis for stop Codon recognition in Eukaryotes

   Nature 524:493–496.

   https://doi.org/10.1038/nature14896

   • Google Scholar
10. 1. Calvo SE
    2. Pagliarini DJ
    3. Mootha VK

    (2009) Upstream open reading frames cause widespread reduction of protein expression and are polymorphic among humans

    PNAS 106:7507–7512.

    https://doi.org/10.1073/pnas.0810916106

    • PubMed
    • Google Scholar
11. 1. Cao D
    2. Parker R

    (2003) Computational modeling and experimental analysis of nonsense-mediated decay in yeast

    Cell 113:533–545.

    https://doi.org/10.1016/s0092-8674(03)00353-2

    • PubMed
    • Google Scholar
12. 1. Celik A
    2. Baker R
    3. He F
    4. Jacobson A

    (2017) High-resolution profiling of NMD targets in yeast reveals Translational Fidelity as a basis for substrate selection

    RNA 23:735–748.

    https://doi.org/10.1261/rna.060541.116

    • PubMed
    • Google Scholar
13. 1. Chew GL
    2. Pauli A
    3. Schier AF

    (2016) Conservation of uORF Repressiveness and sequence features in mouse, human and Zebrafish

    Nature Communications 7:11663.

    https://doi.org/10.1038/ncomms11663

    • PubMed
    • Google Scholar
14. 1. Corley M
    2. Solem A
    3. Phillips G
    4. Lackey L
    5. Ziehr B
    6. Vincent HA
    7. Mustoe AM
    8. Ramos SBV
    9. Weeks KM
    10. Moorman NJ
    11. Laederach A

    (2017) An RNA structure-mediated, Posttranscriptional model of human Α-1-Antitrypsin expression

    PNAS 114:E10244–E10253.

    https://doi.org/10.1073/pnas.1706539114

    • PubMed
    • Google Scholar
15. 1. Cridge AG
    2. Crowe-McAuliffe C
    3. Mathew SF
    4. Tate WP

    (2018) Eukaryotic Translational termination efficiency is influenced by the 3′ Nucleotides within the Ribosomal mRNA channel

    Nucleic Acids Research 46:1927–1944.

    https://doi.org/10.1093/nar/gkx1315

    • PubMed
    • Google Scholar
16. 1. Culbertson MR
    2. Leeds PF

    (2003) Looking at mRNA decay pathways through the window of molecular evolution

    Current Opinion in Genetics & Development 13:207–214.

    https://doi.org/10.1016/S0959-437X(03)00014-5

    • Google Scholar
17. 1. Cuperus JT
    2. Groves B
    3. Kuchina A
    4. Rosenberg AB
    5. Jojic N
    6. Fields S
    7. Seelig G

    (2017) Deep learning of the regulatory grammar of yeast 5’ untranslated regions from 500,000 random sequences

    Genome Research 27:2015–2024.

    https://doi.org/10.1101/gr.224964.117

    • Google Scholar
18. 1. Czaplinski K
    2. Ruiz-Echevarria MJ
    3. González CI
    4. Peltz SW

    (1999) Should we kill the messenger? the role of the surveillance complex in translation termination and mRNA turnover

    BioEssays 21:685–696.

    https://doi.org/10.1002/(SICI)1521-1878(199908)21:8<685::AID-BIES8>3.0.CO;2-4

    • PubMed
    • Google Scholar
19. 1. Diaz de Arce AJ
    2. Noderer WL
    3. Wang CL

    (2017) Complete motif analysis of sequence requirements for translation initiation at non-AUG start Codons

    Nucleic Acids Research 46:985–994.

    https://doi.org/10.1093/nar/gkx1114

    • PubMed
    • Google Scholar
20. 1. Dvir S
    2. Velten L
    3. Sharon E
    4. Zeevi D
    5. Carey LB
    6. Weinberger A
    7. Segal E

    (2013) Deciphering the rules by which 5'-UTR sequences affect protein expression in yeast

    PNAS 110:E2792–E2801.

    https://doi.org/10.1073/pnas.1222534110

    • PubMed
    • Google Scholar
21. 1. Eberle AB
    2. Stalder L
    3. Mathys H
    4. Orozco RZ
    5. Mühlemann O

    (2008) Posttranscriptional Gene regulation by spatial rearrangement of the 3' untranslated region

    PLOS Biology 6:e92.

    https://doi.org/10.1371/journal.pbio.0060092

    • PubMed
    • Google Scholar
22. 1. Eisenberg AR
    2. Higdon AL
    3. Hollerer I
    4. Fields AP
    5. Jungreis I
    6. Diamond PD
    7. Kellis M
    8. Jovanovic M
    9. Brar GA

    (2020) Translation initiation site profiling reveals widespread synthesis of non-AUG-initiated protein Isoforms in yeast

    Cell Systems 11:145–160.

    https://doi.org/10.1016/j.cels.2020.06.011

    • PubMed
    • Google Scholar
23. 1. Ferreira JP
    2. Noderer WL
    3. Diaz de Arce AJ
    4. Wang CL

    (2014) Engineering Ribosomal leaky scanning and upstream open reading frames for precise control of protein translation

    Bioengineered 5:186–192.

    https://doi.org/10.4161/bioe.27607

    • PubMed
    • Google Scholar
24. 1. Gerashchenko MV
    2. Gladyshev VN

    (2014) Translation inhibitors cause abnormalities in Ribosome profiling experiments

    Nucleic Acids Research 42:e134.

    https://doi.org/10.1093/nar/gku671

    • PubMed
    • Google Scholar
25. 1. González CI
    2. Ruiz-Echevarría MJ
    3. Vasudevan S
    4. Henry MF
    5. Peltz SW

    (2000) The yeast hnRNP-like protein Hrp1/Nab4 marks a transcript for nonsense-mediated mRNA decay

    Molecular Cell 5:489–499.

    https://doi.org/10.1016/s1097-2765(00)80443-8

    • PubMed
    • Google Scholar
26. 1. Gorgoni B
    2. Zhao YB
    3. Krishnan J
    4. Stansfield I

    (2019) Destabilization of Eukaryote mRNAs by 5’ proximal stop Codons can occur independently of the nonsense-mediated mRNA decay pathway

    Cells 8:800.

    https://doi.org/10.3390/cells8080800

    • PubMed
    • Google Scholar
27. 1. Grant CM
    2. Miller PF
    3. Hinnebusch AG

    (2012) Requirements for Intercistronic distance and level of Eukaryotic initiation factor 2 activity in Reinitiation on Gcn4 mRNA vary with the downstream Cistron

    Molecular and Cellular Biology 14:2616–2628.

    https://doi.org/10.1128/mcb.14.4.2616-2628.1994

    • PubMed
    • Google Scholar
28. 1. Guenther UP
    2. Weinberg DE
    3. Zubradt MM
    4. Tedeschi FA
    5. Stawicki BN
    6. Zagore LL
    7. Brar GA
    8. Licatalosi DD
    9. Bartel DP
    10. Weissman JS
    11. Jankowsky E

    (2018) The Helicase Ded1P controls use of near-cognate translation initiation Codons in 5′ Utrs

    Nature 559:130–134.

    https://doi.org/10.1038/s41586-018-0258-0

    • PubMed
    • Google Scholar
29. 1. Gunišová S
    2. Beznosková P
    3. Mohammad MP
    4. Vlčková V
    5. Valášek LS

    (2016) In-depth analysis of cis-determinants that either promote or inhibit Reinitiation on Gcn4Mrna after translation of its four short uORFs

    RNA 22:542–558.

    https://doi.org/10.1261/rna.055046.115

    • PubMed
    • Google Scholar
30. 1. Gunišová S
    2. Hronová V
    3. Mohammad MP
    4. Hinnebusch AG
    5. Valášek LS

    (2017) Translation reinitiation in microbes and higher eukaryotes

    FEMS Microbiology Reviews 42:165–192.

    https://doi.org/10.1093/femsre/fux059

    • PubMed
    • Google Scholar
31. Book

    1. Hastie T
    2. Tibshirani R
    3. Friedman J

    (2009) The Elements of Statistical Learning: Data Mining, Inference, and Prediction

    New York, NY: Springer.

    https://doi.org/10.1007/978-0-387-84858-7

    • Google Scholar
32. 1. Hinnebusch AG

    (2005) Translational regulation of Gcn4 and the general amino acid control of yeast

    Annual Review of Microbiology 59:407–450.

    https://doi.org/10.1146/annurev.micro.59.031805.133833

    • PubMed
    • Google Scholar
33. 1. Hinnebusch AG
    2. Ivanov IP
    3. Sonenberg N

    (2016) Translational control by 5’-Untranslated regions of Eukaryotic mRNAs

    Science 352:1413–1416.

    https://doi.org/10.1126/science.aad9868

    • Google Scholar
34. 1. Ingolia NT
    2. Ghaemmaghami S
    3. Newman JRS
    4. Weissman JS

    (2009) Genome-wide analysis in vivo of translation with nucleotide resolution using Ribosome profiling

    Science 324:218–223.

    https://doi.org/10.1126/science.1168978

    • PubMed
    • Google Scholar
35. 1. Ingolia NT
    2. Lareau LF
    3. Weissman JS

    (2011) Ribosome profiling of mouse embryonic stem cells reveals the complexity and Dynamics of mammalian Proteomes

    Cell 147:789–802.

    https://doi.org/10.1016/j.cell.2011.10.002

    • PubMed
    • Google Scholar
36. 1. Johnstone TG
    2. Bazzini AA
    3. Giraldez AJ

    (2016) Upstream Orfs are prevalent Translational Repressors in vertebrates

    The EMBO Journal 35:706–723.

    https://doi.org/10.15252/embj.201592759

    • PubMed
    • Google Scholar
37. 1. Kearse MG
    2. Wilusz JE

    (2017) Non-AUG translation: a new start for protein synthesis in Eukaryotes

    Genes & Development 31:1717–1731.

    https://doi.org/10.1101/gad.305250.117

    • PubMed
    • Google Scholar
38. 1. Kearse MG
    2. Goldman DH
    3. Choi J
    4. Nwaezeapu C
    5. Liang D
    6. Green KM
    7. Goldstrohm AC
    8. Todd PK
    9. Green R
    10. Wilusz JE

    (2019) Ribosome Queuing enables non-AUG translation to be resistant to multiple protein synthesis inhibitors

    Genes & Development 33:871–885.

    https://doi.org/10.1101/gad.324715.119

    • PubMed
    • Google Scholar
39. 1. Keeling KM
    2. Lanier J
    3. Du M
    4. Salas-Marco J
    5. Gao L
    6. Kaenjak-Angeletti A
    7. Bedwell DM

    (2004) Leaky termination at premature stop Codons Antagonizes nonsense-mediated mRNA decay in S. cerevisiae

    RNA 10:691–703.

    https://doi.org/10.1261/rna.5147804

    • PubMed
    • Google Scholar
40. 1. Kolitz SE
    2. Takacs JE
    3. Lorsch JR

    (2009) Kinetic and thermodynamic analysis of the role of start Codon/Anticodon base Pairing during Eukaryotic translation initiation

    RNA 15:138–152.

    https://doi.org/10.1261/rna.1318509

    • PubMed
    • Google Scholar
41. 1. Kozak M

    (1990) Downstream secondary structure facilitates recognition of initiator Codons by Eukaryotic Ribosomes

    PNAS 87:8301–8305.

    https://doi.org/10.1073/pnas.87.21.8301

    • PubMed
    • Google Scholar
42. 1. Kozak M

    (1991a)

    Structural features in Eukaryotic mRNAs that modulate the initiation of translation

    The Journal of Biological Chemistry 266:19867–19870.

    • PubMed
    • Google Scholar
43. 1. Kozak M

    (1991b)

    A short leader sequence impairs the Fidelity of initiation by Eukaryotic Ribosomes

    Gene Expression 1:111–115.

    • PubMed
    • Google Scholar
44. 1. Lai W-JC
    2. Kayedkhordeh M
    3. Cornell EV
    4. Farah E
    5. Bellaousov S
    6. Rietmeijer R
    7. Salsi E
    8. Mathews DH
    9. Ermolenko DN

    (2018) mRNAs and lncRNAs intrinsically form secondary structures with short end-to-end distances

    Nature Communications 9:4328.

    https://doi.org/10.1038/s41467-018-06792-z

    • PubMed
    • Google Scholar
45. 1. Lareau LF
    2. Hite DH
    3. Hogan GJ
    4. Brown PO

    (2014) Distinct stages of the translation elongation cycle revealed by sequencing Ribosome-protected mRNA fragments

    eLife 3:e01257.

    https://doi.org/10.7554/eLife.01257

    • PubMed
    • Google Scholar
46. Preprint

    1. Lee DSM
    2. Park J
    3. Kromer A
    4. Rader DJ
    5. Ritchie MD
    6. Ghanem LR
    7. Barash Y
    8. Regeneron Genetics Center

    (2020) Disrupting Upstream Translation in MRNAs Leads to Loss-of-Function Associated with Human Disease

    bioRxiv.

    https://doi.org/10.1101/2020.09.09.287912

    • Google Scholar
47. 1. Letzring DP
    2. Wolf AS
    3. Brule CE
    4. Grayhack EJ

    (2013) Translation of CGA Codon repeats in yeast involves quality control components and Ribosomal protein L1

    RNA 19:1208–1217.

    https://doi.org/10.1261/rna.039446.113

    • PubMed
    • Google Scholar
48. 1. Li JJ
    2. Chew GL
    3. Biggin MD

    (2019) Quantitative principles of cis-Translational control by General mRNA sequence features in Eukaryotes

    Genome Biology 20:162.

    https://doi.org/10.1186/s13059-019-1761-9

    • PubMed
    • Google Scholar
49. 1. Lin Y
    2. May GE
    3. Kready H
    4. Nazzaro L
    5. Mao M
    6. Spealman P
    7. Creeger Y
    8. McManus CJ
    9. Joel McManus C

    (2019) Impacts of uORF Codon identity and position on translation regulation

    Nucleic Acids Research 47:9358–9367.

    https://doi.org/10.1093/nar/gkz681

    • PubMed
    • Google Scholar
50. 1. Lindeboom RGH
    2. Supek F
    3. Lehner B

    (2016) The rules and impact of nonsense-mediated mRNA decay in human cancers

    Nature Genetics 48:1112–1118.

    https://doi.org/10.1038/ng.3664

    • PubMed
    • Google Scholar
51. 1. Lorenz R
    2. Bernhart SH
    3. Höner Zu Siederdissen C
    4. Tafer H
    5. Flamm C
    6. Stadler PF
    7. Hofacker IL

    (2011) Viennarna package 2.0

    Algorithms for Molecular Biology 6:26.

    https://doi.org/10.1186/1748-7188-6-26

    • PubMed
    • Google Scholar
52. 1. Lu Z
    2. Lin Z

    (2019) Pervasive and dynamic transcription initiation in Saccharomyces cerevisiae

    Genome Research 29:1198–1210.

    https://doi.org/10.1101/gr.245456.118

    • PubMed
    • Google Scholar
53. 1. Madeira F
    2. Pearce M
    3. Tivey ARN
    4. Basutkar P
    5. Lee J
    6. Edbali O
    7. Madhusoodanan N
    8. Kolesnikov A
    9. Lopez R

    (2022) Search and sequence analysis tools services from EMBL-EBI in 2022

    Nucleic Acids Research 50:W276–W279.

    https://doi.org/10.1093/nar/gkac240

    • Google Scholar
54. 1. Magoč T
    2. Salzberg SL

    (2011) FLASH: fast length adjustment of short reads to improve genome assemblies

    Bioinformatics 27:2957–2963.

    https://doi.org/10.1093/bioinformatics/btr507

    • PubMed
    • Google Scholar
55. 1. Martin M

    (2011) Cutadapt removes adapter sequences from high-throughput sequencing reads

    EMBnet.Journal 17:10.

    https://doi.org/10.14806/ej.17.1.200

    • Google Scholar
56. 1. McGillivray P
    2. Ault R
    3. Pawashe M
    4. Kitchen R
    5. Balasubramanian S
    6. Gerstein M

    (2018) A comprehensive catalog of predicted functional upstream open reading frames in humans

    Nucleic Acids Research 46:3326–3338.

    https://doi.org/10.1093/nar/gky188

    • PubMed
    • Google Scholar
57. 1. McManus CJ
    2. May GE
    3. Spealman P
    4. Shteyman A

    (2014) Ribosome profiling reveals post-transcriptional Buffering of divergent gene expression in yeast

    Genome Research 24:422–430.

    https://doi.org/10.1101/gr.164996.113

    • Google Scholar
58. 1. Meaux S
    2. van Hoof A
    3. Baker KE

    (2008) Nonsense-mediated mRNA decay in yeast does not require Pab1 or a Poly(A) tail

    Molecular Cell 29:134–140.

    https://doi.org/10.1016/j.molcel.2007.10.031

    • PubMed
    • Google Scholar
59. 1. Mueller PP
    2. Hinnebusch AG

    (1986) Multiple upstream AUG Codons mediate translational control of Gcn4

    Cell 45:201–207.

    https://doi.org/10.1016/0092-8674(86)90384-3

    • PubMed
    • Google Scholar
60. 1. Muhlrad D
    2. Parker R

    (1999) Aberrant mRNAs with extended 3’ Utrs are substrates for rapid degradation by mRNA surveillance

    RNA 5:1299–1307.

    https://doi.org/10.1017/s1355838299990829

    • PubMed
    • Google Scholar
61. 1. Mustoe AM
    2. Corley M
    3. Laederach A
    4. Weeks KM

    (2018) Messenger RNA structure regulates translation initiation: A mechanism exploited from bacteria to humans

    Biochemistry 57:3537–3539.

    https://doi.org/10.1021/acs.biochem.8b00395

    • PubMed
    • Google Scholar
62. 1. Namy O
    2. Hatin I
    3. Rousset JP

    (2001) Impact of the six Nucleotides downstream of the stop Codon on translation termination

    EMBO Reports 2:787–793.

    https://doi.org/10.1093/embo-reports/kve176

    • PubMed
    • Google Scholar
63. 1. Noderer WL
    2. Flockhart RJ
    3. Bhaduri A
    4. Diaz de Arce AJ
    5. Zhang J
    6. Khavari PA
    7. Wang CL

    (2014) Quantitative analysis of mammalian translation initiation sites by FACS-Seq

    Molecular Systems Biology 10:748.

    https://doi.org/10.15252/msb.20145136

    • PubMed
    • Google Scholar
64. Preprint

    1. Oertlin C
    2. Lorent J
    3. Gandin V
    4. Murie C
    5. Masvidal L
    6. Cargnello M
    7. Furic L
    8. Topisirovic I
    9. Larsson O

    (2017) Genome-Wide Analysis of Differential Translation and Differential Translational Buffering Using Anota2seq

    bioRxiv.

    https://doi.org/10.1101/106922

    • Google Scholar
65. 1. Pelechano V
    2. Wei W
    3. Steinmetz LM

    (2013) Extensive transcriptional heterogeneity revealed by Isoform profiling

    Nature 497:127–131.

    https://doi.org/10.1038/nature12121

    • PubMed
    • Google Scholar
66. 1. Peltz SW
    2. Brown AH
    3. Jacobson A

    (1993) mRNA Destabilization triggered by premature Translational termination depends on at least three cis-acting sequence elements and one Trans-acting factor

    Genes & Development 7:1737–1754.

    https://doi.org/10.1101/gad.7.9.1737

    • PubMed
    • Google Scholar
67. 1. Peter J
    2. De Chiara M
    3. Friedrich A
    4. Yue J-X
    5. Pflieger D
    6. Bergström A
    7. Sigwalt A
    8. Barre B
    9. Freel K
    10. Llored A
    11. Cruaud C
    12. Labadie K
    13. Aury J-M
    14. Istace B
    15. Lebrigand K
    16. Barbry P
    17. Engelen S
    18. Lemainque A
    19. Wincker P
    20. Liti G
    21. Schacherer J

    (2018) Genome evolution across 1,011 Saccharomyces cerevisiae isolates

    Nature 556:339–344.

    https://doi.org/10.1038/s41586-018-0030-5

    • PubMed
    • Google Scholar
68. 1. Raimondeau E
    2. Bufton JC
    3. Schaffitzel C

    (2018) New insights into the interplay between the translation machinery and nonsense-mediated mRNA decay factors

    Biochemical Society Transactions 46:503–512.

    https://doi.org/10.1042/BST20170427

    • PubMed
    • Google Scholar
69. 1. Requião RD
    2. de Souza HJA
    3. Rossetto S
    4. Domitrovic T
    5. Palhano FL

    (2016) Increased Ribosome density associated to positively charged residues is evident in Ribosome profiling experiments performed in the absence of translation inhibitors

    RNA Biology 13:561–568.

    https://doi.org/10.1080/15476286.2016.1172755

    • PubMed
    • Google Scholar
70. 1. Rojas-Duran MF
    2. Gilbert WV

    (2012) Alternative transcription start site selection leads to large differences in translation activity in yeast

    RNA 18:2299–2305.

    https://doi.org/10.1261/rna.035865.112

    • PubMed
    • Google Scholar
71. 1. Ruiz-Echevarría MJ
    2. González CI
    3. Peltz SW

    (1998) Identifying the right stop: determining how the surveillance complex recognizes and Degrades an aberrant mRNA

    The EMBO Journal 17:575–589.

    https://doi.org/10.1093/emboj/17.2.575

    • PubMed
    • Google Scholar
72. 1. Ruiz-Echevarría MJ
    2. Peltz SW

    (2000) The RNA binding protein Pub1 modulates the stability of transcripts containing upstream open reading frames

    Cell 101:741–751.

    https://doi.org/10.1016/s0092-8674(00)80886-7

    • PubMed
    • Google Scholar
73. 1. Ryabova LA
    2. Hohn T

    (2000)

    Ribosome Shunting in the cauliflower mosaic virus 35S RNA leader is a special case of Reinitiation of translation functioning in plant and animal systems

    Genes & Development 14:817–829.

    • PubMed
    • Google Scholar
74. 1. Sample PJ
    2. Wang B
    3. Reid DW
    4. Presnyak V
    5. McFadyen IJ
    6. Morris DR
    7. Seelig G

    (2019) Human 5′ UTR design and variant effect prediction from a Massively parallel translation assay

    Nature Biotechnology 37:803–809.

    https://doi.org/10.1038/s41587-019-0164-5

    • PubMed
    • Google Scholar
75. 1. Scannell DR
    2. Zill OA
    3. Rokas A
    4. Payen C
    5. Dunham MJ
    6. Eisen MB
    7. Rine J
    8. Johnston M
    9. Hittinger CT

    (2011) The awesome power of yeast evolutionary Genetics: new genome sequences and strain resources for the Saccharomyces Sensu Stricto genus

    G3: Genes, Genomes, Genetics 1:11–25.

    https://doi.org/10.1534/g3.111.000273

    • PubMed
    • Google Scholar
76. 1. Sen ND
    2. Zhou F
    3. Ingolia NT
    4. Hinnebusch AG

    (2015) Genome-wide analysis of Translational efficiency reveals distinct but overlapping functions of yeast dead-box RNA Helicases Ded1 and Eif4A

    Genome Research 25:1196–1205.

    https://doi.org/10.1101/gr.191601.115

    • PubMed
    • Google Scholar
77. 1. Sen ND
    2. Gupta N
    3. K Archer S
    4. Preiss T
    5. Lorsch JR
    6. Hinnebusch AG

    (2019) Functional interplay between DEAD-box RNA Helicases Ded1 and Dbp1 in Preinitiation complex attachment and scanning on structured mRNAs in vivo

    Nucleic Acids Research 47:8785–8806.

    https://doi.org/10.1093/nar/gkz595

    • PubMed
    • Google Scholar
78. 1. Sharma D
    2. Jankowsky E

    (2014) The Ded1/Ddx3 subfamily of DEAD-box RNA helicases

    Critical Reviews in Biochemistry and Molecular Biology 49:343–360.

    https://doi.org/10.3109/10409238.2014.931339

    • PubMed
    • Google Scholar
79. 1. Siepel A
    2. Bejerano G
    3. Pedersen JS
    4. Hinrichs AS
    5. Hou M
    6. Rosenbloom K
    7. Clawson H
    8. Spieth J
    9. Hillier LW
    10. Richards S
    11. Weinstock GM
    12. Wilson RK
    13. Gibbs RA
    14. Kent WJ
    15. Miller W
    16. Haussler D

    (2005) Evolutionarily conserved elements in vertebrate, insect, worm, and yeast Genomes

    Genome Research 15:1034–1050.

    https://doi.org/10.1101/gr.3715005

    • PubMed
    • Google Scholar
80. 1. Smith JE
    2. Alvarez-Dominguez JR
    3. Kline N
    4. Huynh NJ
    5. Geisler S
    6. Hu W
    7. Coller J
    8. Baker KE

    (2014) Translation of small open reading frames within Unannotated RNA transcripts in Saccharomyces cerevisiae

    Cell Reports 7:1858–1866.

    https://doi.org/10.1016/j.celrep.2014.05.023

    • Google Scholar
81. 1. Spealman P
    2. Naik AW
    3. May GE
    4. Kuersten S
    5. Freeberg L
    6. Murphy RF
    7. McManus J

    (2018) Conserved non-AUG uORFs revealed by a novel regression analysis of Ribosome profiling data

    Genome Research 28:214–222.

    https://doi.org/10.1101/gr.221507.117

    • PubMed
    • Google Scholar
82. 1. Takacs JE
    2. Neary TB
    3. Ingolia NT
    4. Saini AK
    5. Martin-Marcos P
    6. Pelletier J
    7. Hinnebusch AG
    8. Lorsch JR

    (2011) Identification of compounds that decrease the Fidelity of start Codon recognition by the Eukaryotic Translational machinery

    RNA 17:439–452.

    https://doi.org/10.1261/rna.2475211

    • PubMed
    • Google Scholar
83. 1. Tang HL
    2. Yeh LS
    3. Chen NK
    4. Ripmaster T
    5. Schimmel P
    6. Wang CC

    (2004) Translation of a yeast mitochondrial tRNA synthetase initiated at redundant non-AUG Codons

    The Journal of Biological Chemistry 279:49656–49663.

    https://doi.org/10.1074/jbc.M408081200

    • PubMed
    • Google Scholar
84. 1. Vattem KM
    2. Wek RC

    (2004) Reinitiation involving upstream Orfs regulates Atf4 mRNA translation in mammalian cells

    PNAS 101:11269–11274.

    https://doi.org/10.1073/pnas.0400541101

    • PubMed
    • Google Scholar
85. 1. Wagner S
    2. Herrmannová A
    3. Hronová V
    4. Gunišová S
    5. Sen ND
    6. Hannan RD
    7. Hinnebusch AG
    8. Shirokikh NE
    9. Preiss T
    10. Valášek LS

    (2020) Selective translation complex profiling reveals staged initiation and Co-Translational assembly of initiation factor complexes

    Molecular Cell 79:546–560.

    https://doi.org/10.1016/j.molcel.2020.06.004

    • PubMed
    • Google Scholar
86. 1. Wethmar K
    2. Smink JJ
    3. Leutz A

    (2010) Upstream open reading frames: molecular switches in (Patho)Physiology

    BioEssays 32:885–893.

    https://doi.org/10.1002/bies.201000037

    • PubMed
    • Google Scholar
87. 1. Wethmar K

    (2014) The regulatory potential of upstream open reading frames in Eukaryotic gene expression

    Wiley Interdisciplinary Reviews. RNA 5:765–778.

    https://doi.org/10.1002/wrna.1245

    • PubMed
    • Google Scholar
88. 1. Wethmar K
    2. Barbosa-Silva A
    3. Andrade-Navarro MA
    4. Leutz A

    (2014) uORFdb--a comprehensive literature database on Eukaryotic uORF biology

    Nucleic Acids Research 42:D60–D67.

    https://doi.org/10.1093/nar/gkt952

    • PubMed
    • Google Scholar
89. 1. Young SK
    2. Wek RC

    (2016) Upstream open reading frames Differentially regulate Gene-specific translation in the integrated stress response

    Journal of Biological Chemistry 291:16927–16935.

    https://doi.org/10.1074/jbc.R116.733899

    • Google Scholar
90. 1. Zhang J
    2. Maquat LE

    (1997) Evidence that translation Reinitiation Abrogates nonsense-mediated mRNA decay in mammalian cells

    The EMBO Journal 16:826–833.

    https://doi.org/10.1093/emboj/16.4.826

    • Google Scholar
91. 1. Zhang Z
    2. Dietrich FS

    (2005a) Identification and characterization of upstream open reading frames (uORF) in the 5' untranslated regions (UTR) of genes in Saccharomyces cerevisiae

    Current Genetics 48:77–87.

    https://doi.org/10.1007/s00294-005-0001-x

    • PubMed
    • Google Scholar
92. 1. Zhang Z
    2. Dietrich FS

    (2005b) Mapping of transcription start sites in Saccharomyces cerevisiae using 5’ SAGE

    Nucleic Acids Research 33:2838–2851.

    https://doi.org/10.1093/nar/gki583

    • PubMed
    • Google Scholar
93. 1. Zhang F
    2. Hinnebusch AG

    (2011) An upstream ORF with non-AUG start Codon is translated in vivo but Dispensable for Translational control of Gcn4 mRNA

    Nucleic Acids Research 39:3128–3140.

    https://doi.org/10.1093/nar/gkq1251

    • PubMed
    • Google Scholar
94. 1. Zhang H
    2. Dou S
    3. He F
    4. Luo J
    5. Wei L
    6. Lu J
    7. Chen X

    (2018) Genome-wide maps of Ribosomal occupancy provide insights into adaptive evolution and regulatory roles of uORFs during Drosophila development

    PLOS Biology 16:e2003903.

    https://doi.org/10.1371/journal.pbio.2003903

    • Google Scholar
95. 1. Zhang H
    2. Wang Y
    3. Lu J

    (2019) Function and evolution of upstream Orfs in Eukaryotes

    Trends in Biochemical Sciences 44:782–794.

    https://doi.org/10.1016/j.tibs.2019.03.002

    • Google Scholar
96. 1. Zydowicz-Machtel P
    2. Swiatkowska A
    3. Popenda Ł
    4. Gorska A
    5. Ciesiołka J

    (2018) Variants of the 5'-Terminal region of P53 mRNA influence the Ribosomal scanning and translation efficiency

    Scientific Reports 8:1533.

    https://doi.org/10.1038/s41598-018-20010-2

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Unraveling the influences of sequence and position on yeast uORF activity using massively parallel reporter systems and machine learning" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Naama Barkai as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

The essential revisions are outlined in the referees' reports and summarized here:

(1) Estimate the extent to which the effects measured are independent of the reporter used.

(2) Strengthen the analysis of evolution of uORFs between S. cerevisiae and S. paradoxus, for instance by testing their interpretations.

(3) Show that the reporter mRNA initiates at the predicted TSS.

(4) Demonstrate experimentally that the distance from the cap causes the effects revealed from the statistical analyses.

Reviewer #1 (Recommendations for the authors):

This is a very impressive set of experiments and a timely paper.

I have two main comments. One is already explained in the public review and regards the potential dependency of the effects measured on the main ORF. I think this could be discussed but it could be also rather simple to test a subset of uORFs across the range of effects, from strongly repressive to strongly enhancing, and use a different reporter sequence to show that at least the ranking or sign of effects is preserved. Could be done by swapping mCherry and YFP for instance in the plasmid.

This is probably because of my background but I found the section on evolution really interesting and broadening the scope of the paper quite a lot. I wonder if the authors could not push further their analysis for instance by analyzing the polymorphism data available for 100s of strains in S. cerevisiae, for instance by looking at the frequency spectrum of mutations that would enhance the repressive effects of uORFs. Mutational bias could also be considered. With this part, it would also be interesting to know whether there are interactions between uORFs and the main ORFs because changes in magnitude or even signs of uORFs effects could be compensated by other regulatory changes, in a coevolution akin to what the same authors have reported on before regarding transcription and translation.

Reviewer #2 (Recommendations for the authors):

– They have not verified that a random sampling of reporter mRNAs generally have 5' ends corresponding to those predicted by using the ENO2 promoter in their library design. This is important because the position of uORFs from the 5' end is claimed to be an important determinant of uORF function, and represents one of the most novel findings of the study.

– They have not tested their conclusions about the importance of distance of the uORF from the mRNA cap on uORF function by follow-up experiments on a few individual repressive uORFs to examine whether increasing the distance of the uORF ATG from the cap by insertion of unstructured sequences (eg. CAA repeats) reduces inhibition by the uORF; or whether deleting sequences to move a uORF closer to the 5' end increases its inhibitory effect. This is important because the conclusion is based only on correlation analysis and hence could be influenced by other features of the mRNA that vary in parallel with distance of the uORF from the cap.

– Regarding the distance from the cap effect, it seems possible that repression by an uORF would be diminished by a location further downstream from the cap owing to inclusion of other AUG or near-cognate start sites located upstream of the uORF, which would reduce translation initiation at the uORF of interest. This possibility should be addressed bioinformatically, and also by examining individual reporters in which all AUG or near-cognate triplets are removed upstream of the uORF in question to determine if this increases repression by the uORF without changing its position relative to the cap.

– A point of major confusion: Most of the plots and supplementary tables list the reporter expression ratios as log2(AAG/WT); however, I believe that this incorrect and the log2(WT/AAG) ratios have been plotted instead, as the values for the majority of uORFs, which are repressive, are <0.

­– p. 13: The sentence: "In general, repressor uORFs were less repressive in the upf1∆ strain, including 92% of AUG-uORFs and 59% of non-AUG uORFs." requires statistical analysis to bolster the claims, particularly for the small subset of repressive non-AUG uORFs

– p.14: The sentence "Consistent with this, we found median %NMD was higher for uORFs terminating with UGA (37.5%) than those terminating with UAA (35.3%) or UAG (33.2%) (Figure 3D)." is not justified, as the median values for the three stop codons in Figure 3D do not appear to differ significantly. Moreover, assuming it was valid, is there any way to rationalize it based on known termination efficiencies at the three stop codons?

– Regarding the claim on p. 14 that "Unexpectedly, the location of the stop codon relative to the transcript leader cap was positively correlated with the %NMD, such that uORFs that terminate further from the cap were more likely to induce NMD than those that terminate adjacent to the cap (Figure 3F; R^2 = 0.065; P = 7.44x10^-8)", it should be noted that this is a very weak trend that explains only ~7% of the variance in %NMD values. As such, the sentence in the Discussion "We found the opposite relationship among uORF stop codons, such that stop codons were less likely to induce decay the closer they were located to the 5' cap" appears to be an overstatement. Also, isn't it possible that the trend can be explained as originating from uORF length effects shown in panel F, as longer uORFs will have greater cap to stop codon separations? (Note also that panels F and G were incorrectly cited in text.)

– pp. 16-17 and Figure 4D: The description of the parameters included in the ENR model is wholly inadequate. How was Kozak context quantified and for what sequence interval around the AUG? How was uORF Start and Stop and CDS sequence conservation quantified (and for what species)? What sequence interval around the uORF Start was employed to calculate the deltaG folding energies, and how were the calculations made? How far downstream was %AU calculated. In general sufficient information has to provided in the legends or Methods to allow the analysis to be repeated in full by other workers. A related comment is that the authors cite no literature to justify their analyses of these different parameters which they seem to pull out of the hat, eg. why Pro and Gly codons?

– p.19 and Figure 5A: the claim that "In general, uORFs were more repressive when they were closer to the TSS" is not convincing, as many uORFs shown in blue in Figure 5A show the opposite trend; and there are only a few outliers that conform to the stated trend. No statistical analysis of the trend was provided. As such, it is an overinterpretation of the data to claim that these results independently support the importance of uORF distance from the TSS, indicated by the ENR analysis. It also seems important to provide the YFP expression for the different leaders with WT and mutant uORFs, rather than just the WT/AGG ratios, in order to evaluate whether there are other sequences besides uORFs in the longer 5'UTRs that are affecting expression. This information should be added to Figure 5B in particular.

– p.20-21" Regarding the text: "For example, an S. cerevisiae oORF in SEC1 was 1.5 to 2-times more repressive than its S. paradoxus homolog, potentially owing to a deletion in S. paradoxus that results in an earlier stop codon that shortens the oORF. In another case, an S. paradoxus oORF in the AIM22 leader was approximately four-fold more repressive than its S. cerevisiae homolog, possibly due to the presence of more adenosines in its Kozak sequence." The words "potentially" and "possibly" in these sentences presumably reflect the fact that there are other sequence differences in the 5'UTRs between the two species that contribute to the differences in uORF function. These interpretations should be tested by mutational analysis of the reporters for these genes to determine whether shortening the oORF in S.c. SEC1 and improving the Kozak context of the S.c AIM22 oORF are sufficient to confer the altered repressive functions of the orthologous oORFs in S.p.

–

Figure 6D and related text: It is not convincing that conserved uORFs have a statistically significant poorer AUG context compared to nonconserved uORFs, even if one focuses (as they do) on only the -3 position rather than calculating the context score for the entire sequence interval surrounding the AUG. This is being overinterpreted.

Essential revisions:

The essential revisions are outlined in the referees' reports and summarized here:

(1) Estimate the extent to which the effects measured are independent of the reporter used.

To estimate the influence of the reporter protein on uORF activity measurements, we made reporters to test 6 uORFs with various effect sizes in which we swapped the fluorescent reporters, placing the uORF upstream of mCherry, while YFP became the internal control. We also recloned the same uORFs upstream of YFP, with mCherry as an internal control. We then tested all of the reporters using a TECAN fluorometer. This was done to ensure the experimental measurements were equivalent. Considering all 6 uORFs, the Pearson’s correlation constant (R) was 0.814 (R-squared = 0.662). However, one of the reporters had very low expression and a high variance by TECAN measurement. Excluding that reporter uORF, the R-square was 0.869 and R was 0.932. We conclude our results show that the reporter protein has minimal impact on the uORF effect size. These results are now included in the new Figure 1—figure supplement 3 and noted on page 8 of the revised manuscript.

(2) Strengthen the analysis of evolution of uORFs between S. cerevisiae and S. paradoxus, for instance by testing their interpretations.

To further evaluate the evolutionary differences between S. cerevisiae and S. paradoxus uORFs, we swapped the Kozak regions of the homologous uORF in gene YJL046W (AIM22) and compared their activity by TECAN fluorometer measurement. Consistent with our expectation, the S. cerevisiae Kozak region decreased repression in the S. paradoxus uORF. Similarly, the S. paradoxus Kozak sequence increased repression in the S. cerevisiae uORF. This result has been added to the manuscript in the new Figure 5—figure supplement 1. We also revised the analysis of evolutionary changes in uORF activity to compare numerical Kozak context scores for conserved, non-conserved, and de novo uORFs. These analyses supported our initial interpretations from the first submission.

(3) Show that the reporter mRNA initiates at the predicted TSS.

Transcription start sites can be identified by the presence of a 5’ 7-methyl-guanosine (m7G) cap immediately before the site of initiation. Notably, the reverse transcriptase we used in our assay is known to add a cytosine at the site of the m7G cap during cDNA synthesis, followed by two additional nucleotides, primarily CNN (Wulf et al., JBC 2019; Figure 3; PMID 31640989), where the underlined C is complementary to the m7G cap. Because we generated our RNA libraries by ligating an oligo to the 3’ end of the resulting cDNA, reads resulting from capped mRNA contain the complementary sequence “NNG” before the transcription start site, where the “G” reflects the m7G cap nucleotide. In essence, our RNA-seq approach is also a form of CAGE-seq. The transcript leaders in our library are designed to initiate with “AAGC” (from ENO2) immediately before each transcript leader sequence. Thus, “capped” RNA-seq reads will begin with “NNGAAGC” on their 5’ ends.

We reasoned we could use this property of our RNA-seq data to evaluate the frequency at which the capped mRNA initiated at their designed transcription start sites To identify capped RNA-seq reads, we first filtered out reads that aligned perfectly end-to-end to the library design. Then we selected all of the remaining reads that had an “NNG” at their 5’ ends, and trimmed off the NNG. These “cap possible” reads were then aligned to the designed library sequences. Reads that only matched the library design after trimming a 5’ NNG sequence were considered “capped” reads, such that their first base is the transcription start site. In each replicate, ~ 96.6% of the capped reads mapped precisely to the designed AAGC transcription start site. Thus, we find ~ 97% of transcripts from the reporter library initiate transcription at the designed location. We added a new supplemental figure (Figure 1—figure supplement 2) showing this analysis and updated the text accordingly (page 8 of the revision).

(4) Demonstrate experimentally that the distance from the cap causes the effects revealed from the statistical analyses.

Our model suggested uORFs that are far from the TSS will become more repressive when closer to the TSS, and vice versa. To test this, we extended the UTRs of S. cerevisiae YDR396W and S. paradoxus YHR039C, which have strong and medium repressor uORFs, by adding 72-nt of sequence to the UTR 5’ end. The added sequence was designed to prevent the formation of RNA structure using computational predictions (Reuter and Mathews, 2010). The YDR396W uORF became less repressive, while the repressiveness of the YHR039C uORF did not change. Reciprocally, we deleted the 5’ sequence of the YML007W gene’s 5’ UTR to place the uORF 10 nt from the 5’ cap. This did not change the repressiveness of the uORF. We also reduced the UTR length of YPR059C, a weak repressor, to place its uORF 10 nt from the 5’ cap. Contrary to our expectations, the uORF became less repressive. These results have been added to the new supplemental Figure 4—figure supplement 2 and discussed on page 17 of the revised manuscript.

While these results show that uORF repression varies with location in the 5’ UTR, they are not completely consistent with our regression model. One uORF changed as expected, two did not change and one changed in the opposite direction. This may reflect additional effects of changing the 5’ UTR that are not entirely related to uORF position. For example, though we attempted to design the UTR extensions to be unstructured, they may still alter the structural context of uORF or YFP start codons. Some of these sequences could also recruit RNA binding proteins. Similarly, our truncation mutations may have altered the structural accessibility of start codons or removed sites of trans-acting factors. We believe our regression modeling accounts for such additional factors, by leveraging results from thousands of uORFs in diverse sequence, structural, and location contexts. We also performed further computational analysis in response to a comment from reviewer #2, and found that the location effect was seen for single uORF transcript leaders, indicating it is not due to interactions among uORFs.

This highlights an important aspect of computational modeling in general – that correlations are not necessarily causative. Models are useful to the extent that they allow predictions of outcomes. We have revised our manuscript to make this more apparent, now explicitly noting that uORF locations are predictive of their function (revisions on pgs 2, 7, 17, and 25).

Reviewer #1 (Recommendations for the authors):

This is a very impressive set of experiments and a timely paper.

We thank the reviewer for these encouraging comments

I have two main comments. One is already explained in the public review and regards the potential dependency of the effects measured on the main ORF. I think this could be discussed but it could be also rather simple to test a subset of uORFs across the range of effects, from strongly repressive to strongly enhancing, and use a different reporter sequence to show that at least the ranking or sign of effects is preserved. Could be done by swapping mCherry and YFP for instance in the plasmid.

We carried out the suggested swap for six uORF reporters (wildtype and mutant) and assayed the effects of the uORF on reporter protein expression using a fluorometer. The measurements were largely consistent, as shown in the new Figure 1—figure supplement 3 of the revised manuscript.

This is probably because of my background but I found the section on evolution really interesting and broadening the scope of the paper quite a lot. I wonder if the authors could not push further their analysis for instance by analyzing the polymorphism data available for 100s of strains in S. cerevisiae, for instance by looking at the frequency spectrum of mutations that would enhance the repressive effects of uORFs. Mutational bias could also be considered. With this part, it would also be interesting to know whether there are interactions between uORFs and the main ORFs because changes in magnitude or even signs of uORFs effects could be compensated by other regulatory changes, in a coevolution akin to what the same authors have reported on before regarding transcription and translation.

We thank the reviewer for these suggestions. While we agree this is an interesting avenue to explore, this is complicated by uncertainties in locating transcription start site positions in the other genome sequenced strains of S. cerevisiae. Thus, we feel it would be beyond the scope of this initial study and is better suited for future work.

Reviewer #2 (Recommendations for the authors):

– They have not verified that a random sampling of reporter mRNAs generally have 5' ends corresponding to those predicted by using the ENO2 promoter in their library design. This is important because the position of uORFs from the 5' end is claimed to be an important determinant of uORF function, and represents one of the most novel findings of the study.

We agree with the reviewer. Our RNA-seq data were generated by 5’ end ligation. Because reverse transcription adds an untemplated “CNN” at m7G modified 5’ ends (Wulf et al., JBC 2019; Figure 3; PMID 31640989), our RNA-seq data can also be analyzed essentially as 5’ CAGE-seq data. We used this to identify the transcription start sites of our YFP reporter mRNAs. Reassuringly, we found that ~97% of our reporter mRNAs initiated at the designed transcription start site. We added the new Figure 1—figure supplement 2 describing these results and updated the main text (page 8), which are also described in more detail in the response to the editor’s essential revisions list above.

– They have not tested their conclusions about the importance of distance of the uORF from the mRNA cap on uORF function by follow-up experiments on a few individual repressive uORFs to examine whether increasing the distance of the uORF ATG from the cap by insertion of unstructured sequences (eg. CAA repeats) reduces inhibition by the uORF; or whether deleting sequences to move a uORF closer to the 5' end increases its inhibitory effect. This is important because the conclusion is based only on correlation analysis and hence could be influenced by other features of the mRNA that vary in parallel with distance of the uORF from the cap.

– Regarding the distance from the cap effect, it seems possible that repression by an uORF would be diminished by a location further downstream from the cap owing to inclusion of other AUG or near-cognate start sites located upstream of the uORF, which would reduce translation initiation at the uORF of interest. This possibility should be addressed bioinformatically, and also by examining individual reporters in which all AUG or near-cognate triplets are removed upstream of the uORF in question to determine if this increases repression by the uORF without changing its position relative to the cap.

We thank the reviewer for these suggestions. We found that uORFs closer to the cap are more repressive. The reviewer rightly notes that this is based on a regression analysis of thousands of uORFs and had not been experimentally tested. To address this, we extended the 5’ UTRs upstream of two cap-proximal strong uORFs and deleted the 5’ UTRs upstream of two cap-distal mild uORFs. These mutations altered the effects of two of the uORFs. While one extension mutation did reduce uORF repression (as predicted by our ENR modeling), the deletion mutation resulted in a slightly less repressive uORF rather than a more repressive uORF (see new Figure 4—figure supplement 2). As detailed in our response to the reviewing editorial comments, we believe this may reflect additional effects of the mutations outside of changing uORF location (e.g. structural or transacting effects).

We further computationally investigated the possibility that the location effect results from having other upstream uORFs by restricting our correlation analysis to single AUG uORF transcript leaders (501 total; Figure 4—figure supplement 2). As observed with all uORFs, single AUG-uORFs are less repressive with increasing distance from the 5’ cap. The difference is significant with a one-tailed test (P = 0.04, P = 0.03), which is appropriate as our hypothesis is that the cap proximal group is more repressive (as opposed to simply “different”) than the cap distal groups. Stronger repression was also seen for uORFs that end closer to the CDS start codon in single AUG uORF transcript leaders, however this did not reach statistical significance (P = 0.6327 and P = 0.1884). Together, these analyses of single uORF transcript leaders are generally consistent with our interpretations from multiple uORF transcript leaders.

The reviewer also raised an important point regarding the activities of uORFs that are downstream of other uORFs. Our assay isolates repression of individual uORFs by comparing individual uORF AAG mutations to the wildtype UTR. Thus, we expect our measurements will be largely independent of upstream or downstream uORFs. To investigate this, we compared the strength of repression for first, second, third and fourth uORFs in multiple uORF transcript leaders. Strikingly, each had similar median levels of repression (Figure S4—figure supplement 2).

More broadly, the reviewer’s comments raise the important point that correlations from models are not necessarily causative. We agree and have revised the manuscript to better reflect the fact that the model identifies features that are predictive of uORF function (see pages 2, 7, 18, and 26). We believe this more accurately represents the results of the model.

– A point of major confusion: Most of the plots and supplementary tables list the reporter expression ratios as log2(AAG/WT); however, I believe that this incorrect and the log2(WT/AAG) ratios have been plotted instead, as the values for the majority of uORFs, which are repressive, are <0.

Thank you for pointing out this issue. It has been corrected in the revised figures.

­– p. 13: The sentence: "In general, repressor uORFs were less repressive in the upf1∆ strain, including 92% of AUG-uORFs and 59% of non-AUG uORFs." requires statistical analysis to bolster the claims, particularly for the small subset of repressive non-AUG uORFs

We apologize for omitting statistical analysis for this. We used a binomial exact test, with a null hypothesis that equal numbers of more repressive and less repressive uORFs would be seen in the upf1∆ strain. This results in p < 2.2 x 10^-16 for AUG uORFs (N=835) and p = 0.07135 (N = 136) for NCC uORFs. However, this analysis includes a large number of extremely weak NCC uORFs. After filtering out weak uORFs that reduced expression less than 15% in wildtype yeast, we found 93% of AUG (611 / 654, p < 2.2 x 10^-16) and 85% of NCC (22 / 26, p = 0.0005335) repressors were less repressive in the upf1∆ strain. We have added this to the revised manuscript (pg 12).

– p.14: The sentence "Consistent with this, we found median %NMD was higher for uORFs terminating with UGA (37.5%) than those terminating with UAA (35.3%) or UAG (33.2%) (Figure 3D)." is not justified, as the median values for the three stop codons in Figure 3D do not appear to differ significantly. Moreover, assuming it was valid, is there any way to rationalize it based on known termination efficiencies at the three stop codons?

The text notes that it has been reported that UGA is the least efficient stop codon (Bonetti et al., 1995). We also noted that previous studies showed UGAC is the least efficient context for termination. Thus, the variation we observe in NMD induction is consistent with the current model that NMD is induced by inefficient termination. However, as the reviewer correctly states, the difference in %NMD among the three stop codons was not statistically significant. We have updated the text to clarify that point (pg 14).

– Regarding the claim on p. 14 that "Unexpectedly, the location of the stop codon relative to the transcript leader cap was positively correlated with the %NMD, such that uORFs that terminate further from the cap were more likely to induce NMD than those that terminate adjacent to the cap (Figure 3F; R^2 = 0.065; P = 7.44x10^-8)", it should be noted that this is a very weak trend that explains only ~7% of the variance in %NMD values. As such, the sentence in the Discussion "We found the opposite relationship among uORF stop codons, such that stop codons were less likely to induce decay the closer they were located to the 5' cap" appears to be an overstatement. Also, isn't it possible that the trend can be explained as originating from uORF length effects shown in panel F, as longer uORFs will have greater cap to stop codon separations? (Note also that panels F and G were incorrectly cited in text.)

We agree, and revised the results and discussion to include the reviewer’s more nuanced interpretation and avoid overstating the correlation between uORF stop codon location and %NMD (pg 15 and 16). We also corrected citations to panels F and G.

– pp. 16-17 and Figure 4D: The description of the parameters included in the ENR model is wholly inadequate. How was Kozak context quantified and for what sequence interval around the AUG? How was uORF Start and Stop and CDS sequence conservation quantified (and for what species)? What sequence interval around the uORF Start was employed to calculate the deltaG folding energies, and how were the calculations made? How far downstream was %AU calculated. In general sufficient information has to provided in the legends or Methods to allow the analysis to be repeated in full by other workers. A related comment is that the authors cite no literature to justify their analyses of these different parameters which they seem to pull out of the hat, eg. why Pro and Gly codons?

We apologize for these omissions. We have included two new supplemental tables (Supplementary files 1j and 1k) with the features used for modeling and have revised the methods section extensively to describe feature sources, including citations (page 38). We measured relative Kozak scores using our YFP / mCherry reporter plasmid (new Figure 4—figure supplement 1). During revision, we found the Kozak scores used in the initial submission were mistakenly taken from a yeast strain with a different genetic background. We updated the Kozak scores using the wildtype (BY4741) strain and reran the ENR modeling for the resubmission. We also updated the %AU after uORF stop codon feature, as it was missing from uORFs that terminated close to the YFP start codon in the original submission. While the general results were the same (with location being as predictive as Kozak context), the ENR regression weights changed somewhat after modeling with these updated features, such that the wildtype and upf1∆ models were more similar than in the initial submission. We updated the Results section to reflect this.

– p.19 and Figure 5A: the claim that "In general, uORFs were more repressive when they were closer to the TSS" is not convincing, as many uORFs shown in blue in Figure 5A show the opposite trend; and there are only a few outliers that conform to the stated trend. No statistical analysis of the trend was provided. As such, it is an overinterpretation of the data to claim that these results independently support the importance of uORF distance from the TSS, indicated by the ENR analysis.

We agree that this required statistical evaluation. Of uORFS whose activity differed significantly by at least two-fold in the two transcript leaders, thirty-four were more repressive in the shorter transcript leader while twelve were more repressive in the longer transcript leader. We used the binomial exact test to evaluate this result under the null hypothesis that equal numbers of uORFs would be more repressive and less repressive in the context of the shorter transcript leader (P = 0.001641) and added this to the revised manuscript.

It also seems important to provide the YFP expression for the different leaders with WT and mutant uORFs, rather than just the WT/AGG ratios, in order to evaluate whether there are other sequences besides uORFs in the longer 5'UTRs that are affecting expression. This information should be added to Figure 5B in particular.

We agree that raw YFP values should be included with the study and added these to the supplemental table (Supplementary file 1g in the revised submission) and as bar graphs to Figure 5B as requested.

– p.20-21" Regarding the text: "For example, an S. cerevisiae oORF in SEC1 was 1.5 to 2-times more repressive than its S. paradoxus homolog, potentially owing to a deletion in S. paradoxus that results in an earlier stop codon that shortens the oORF. In another case, an S. paradoxus oORF in the AIM22 leader was approximately four-fold more repressive than its S. cerevisiae homolog, possibly due to the presence of more adenosines in its Kozak sequence." The words "potentially" and "possibly" in these sentences presumably reflect the fact that there are other sequence differences in the 5'UTRs between the two species that contribute to the differences in uORF function. These interpretations should be tested by mutational analysis of the reporters for these genes to determine whether shortening the oORF in S.c. SEC1 and improving the Kozak context of the S.c AIM22 oORF are sufficient to confer the altered repressive functions of the orthologous oORFs in S.p.

Although we suggested potential mechanisms for two species differences in uORF activity, other sequence or RNA structural features could certainly contribute. We cloned and tested additional reporters of the AIM22 uORF, swapping the Kozak sequences from the two species. The results were consistent with our suggestion that Kozak sequence differences contribute to these differences in activity (new Figure 5—figure supplement 1), however the Kozak sequence swap did not fully account for the difference. We revised the text to clarify that these are suggested potential contributors to uORF differences, and not necessarily the sole reasons underlying their differences. We believe dissecting the exact mechanisms underlying additional species differences in uORF magnitudes would be an interesting area for future work.

– Figure 6D and related text: It is not convincing that conserved uORFs have a statistically significant poorer AUG context compared to nonconserved uORFs, even if one focuses (as they do) on only the -3 position rather than calculating the context score for the entire sequence interval surrounding the AUG. This is being overinterpreted.

We agree with the reviewer, so we replaced the motif logo comparison with comparisons of Kozak context scores. This includes Wilcoxon rank-sum tests, which show that the lower Kozak scores in conserved uORFs are statistically significant.

Author details

1. Gemma E May

   Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Formal analysis, Investigation, Visualization, Methodology, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

2. Christina Akirtava

   Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Data curation, Software, Formal analysis

   Competing interests

   No competing interests declared

3. Matthew Agar-Johnson

   Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Data curation, Software, Formal analysis, Visualization, Methodology

   Competing interests

   No competing interests declared

4. Jelena Micic

   Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

5. John Woolford

   Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Supervision, Funding acquisition, Writing – review and editing

   Competing interests

   No competing interests declared

6. Joel McManus

   1. Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, United States
   2. Computational Biology Department, Carnegie Mellon University, Pittsburgh, United States

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Visualization, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   mcmanus@andrew.cmu.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6605-2642

• 650

  Page views

• 134

  Downloads

• 1

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.