Fig 1.
High initiation rates decrease protein output of stall-containing S. cerevisiae mRNAs.
(A) Schematic of 3×FLAG-PGK1*-YFP reporters used in (C) and (D). The −4 to −1 nt region preceding the ATG start codon had one of eight different Kozak sequences as indicated. One of three different stall sequences or their respective controls were inserted at the end of PGK1*. PGK1* was modified from wild-type PGK1 sequence based on a previous study [11] (see Strain and plasmid construction). (B) Protein levels of 3×FLAG-PGK1-YFP control reporters with no insert and wild-type PGK1 sequence. (C) Protein levels of the constructs shown in (A). Protein levels are quantified as the mean YFP fluorescence of 10,000 cells for each strain as measured using flow cytometry. Protein levels are expressed as a.u. relative to the mean RFP levels from a constitutively expressed mKate2 control. Error bars in (B) and (C) show standard error of the mean over three or four independent transformants. Many error bars are smaller than data markers. Two of the total 192 strains were clear outliers and were removed after manual inspection. (D) Western blots of reporters with low (CTGC), medium (CCAC), or high (CAAA) initiation rates and with indicated stall sequences or controls using antibody against the FLAG epitope at the N terminus. Histone H3 level is shown as loading control. Numbers for each lane indicate the ratio of the FLAG signal to the H3 signal and are normalized to a maximum of 10 within each blot. The underlying data for panels B and C can be found at https://github.com/rasilab/ribosome_collisions_yeast. a.u., arbitrary unit; chrI, chromosome I; nt, nucleotide; YFP, yellow fluorescent protein.
Fig 2.
High initiation rates decrease stability of stall-containing S. cerevisiae mRNAs.
(A) Schematic of deep-sequencing assay used for measuring mRNA levels. Reporters with different 5′ UTR mutations and stall or control inserts were tagged with four unique 8-nt barcodes in their 3′ UTR. The reporters were pooled and transformed into S. cerevisiae. cDNA and gDNA from the pooled reporters were amplified, and the barcodes were counted by high-throughput sequencing. Barcodes that had fewer than 100 counts for either cDNA or gDNA were discarded during analysis. mRNA level was calculated as cDNA counts normalized by gDNA counts for each barcode and was median normalized across all reporters in each pool. (B) mRNA levels of 3×FLAG-PGK1-YFP control reporters with no insert and wild-type PGK1 sequence. The leftmost point is for a reporter with the ATG start codon mutated to CTG. (C) mRNA levels of 3×FLAG-PGK1*-YFP constructs with stall or control inserts. Error bars in (B) and (C) show standard error over 3 or 4 distinct barcodes for each reporter variant. One reporter variant (10×AGA, ACGC) had only two barcodes with read counts higher than the threshold for discarding. Most error bars are smaller than data markers. The underlying data for panels B and C can be found at https://github.com/rasilab/ribosome_collisions_yeast. a.u., arbitrary unit; gDNA, genomic DNA; nt, nucleotide; YFP, yellow fluorescent protein.
Fig 3.
Collision-stimulated abortive termination model predicts reduced protein output at high initiation rates.
(A) Schematic of different abortive termination models simulated in (B). ka denotes the nonzero rate of abortive termination of ribosomes from indicated configurations. In the TJ model, ribosomes do not abort translation. In the SAT model, only ribosomes that have not undergone collision from the mRNA entry side abort. In the CAT model, only ribosomes that have undergone collision from the mRNA entry side (“trailing” ribosomes in a collision) abort. In the CSAT model, only ribosomes that have undergone collision from the mRNA exit side (“leading” ribosomes in a collision) abort. (B) Average protein synthesis rate as a function of initiation rate predicted using stochastic simulations of the four models in (A). The simulations were of a 650-codon mRNA corresponding to the 3×FLAG-PGK1*-YFP reporters in our experiments. Ribosome stalls were simulated as a consecutive stretch of six slowly translated codons with a net elongation rate of 0.1 s−1. All other codons had an elongation rate of 10 s−1. The stalls were located after 400 codons, corresponding to their approximate location in our experiments. (C) Effect of varying the number of tandem (consecutive) slowly translated codons encoding the stall in the CSAT model. The elongation rate of each slowly translated codon was scaled to maintain a net elongation rate of 0.1 s−1 across the stall. mRNA decay rate was set to zero in these simulations to isolate the effect of abortive termination. All other model parameters are listed in S3 Table. Standard errors from repeating stochastic simulations with different initial random seeds are smaller than data markers in (B) and (C). The underlying data for panels B and C can be found at https://github.com/rasilab/ribosome_collisions_yeast. CAT, collide and abortive termination; CSAT, collision-stimulated abortive termination; SAT, simple abortive termination; TJ, traffic jam; YFP, yellow fluorescent protein.
Fig 4.
CSEC model predicts decrease in mRNA lifetime at high initiation rates.
(A) Schematic of SEC and CSEC mRNA cleavage models simulated in (B) and (C). (B, C) Average mRNA lifetime and protein synthesis rate over 106 s as a function of initiation rate predicted using stochastic simulations of the models in (A). Reporters were simulated as in Fig 3, but with a nonzero endonucleolytic cleavage rate of 0.001 s−1 and a net elongation rate of 0.1 s−1 across the stall. Canonical mRNA decay was allowed, whereas abortive termination rate was set to zero. All other model parameters are listed in S3 Table. Standard errors from repeating stochastic simulations with different initial random seeds are smaller than data markers in (B) and (C). The underlying data for panels B and C can be found at https://github.com/rasilab/ribosome_collisions_yeast. CSEC, collision-stimulated endonucleolytic cleavage; SEC, simple endonucleolytic cleavage.
Fig 5.
Hel2 and Asc1 attenuate gene expression from stall-containing mRNAs only at high initiation rates.
(A) Protein levels of 3×FLAG-PGK1*-YFP reporters (see Fig 1A) with low (CTGC), medium (CCAC), or high (CAAA) initiation rates and with stall (8×CCG) or control (8×CCA) repeats in different deletion backgrounds. (B) Western blots of reporters from (A) performed and quantified as in Fig 1. (C) mRNA levels of 3×FLAG-PGK1*-YFP reporters. Mean mRNA levels and their standard errors quantified as in Fig 2. The ΔLTN1-8×CCA-AAAA variant has only two barcodes. (D) Protein levels of the 8×CCG stall reporter expressed in ΔHEL2 (top) or ΔASC1 (bottom) strain and complemented with the indicated Hel2 or Asc1 variant. Error bars in (A) and (D) calculated as in Fig 1. ** in (D) denotes P < 0.01 (two-tailed t test) for comparison between indicated Hel2/Asc1 variant and the parent strain (“None”) for the same 5′ UTR sequence. The underlying data for panels A, C, and D can be found at https://github.com/rasilab/ribosome_collisions_yeast. a.u., arbitrary unit; nt, nucleotide; WT, wild-type; YFP, yellow fluorescent protein.
Fig 6.
Endogenous mRNAs with stall sequences show signatures of inefficient translation initiation.
(A) Mean ribosome density around stall-encoding sequences in S. cerevisiae mRNAs using data from Weinberg and colleagues [62]. Stall-encoding sequences are defined as 10-codon windows that have either a minimum of 6 lysine and arginine codons or a minimum of 6 proline codons. The first nucleotide of the 10-codon window is at distance 1 nt. In total, 1,251 S. cerevisiae mRNAs have at least one stall-encoding sequence. Arrow indicates peak in ribosome density at +24 nt that is consistent with Brandman and colleagues [16]. (B) Effect of endogenous stalls on protein expression from reporter mRNAs. Protein levels measured by flow cytometry and their mean and standard error quantified as in Fig 1. ***, *, N.S. indicate P < 0.001, P < 0.05, P > 0.05 (one-tailed t test) for higher ΔASC1/WT protein-level ratio of the stall-encoding reporters in comparison with the no-stall control. (C) TE of S. cerevisiae mRNAs that either contain or do not contain stall-encoding sequences. TE is defined as the normalized ratio of ribosome footprint counts to total mRNA counts as measured by Weinberg and colleagues [62]. (D) Fold change in mRNA levels [57] between both ΔHEL2 and ΔASC1 strains and WT strain. (C, D) Box plots shows mean and standard deviation within each gene group; violin plot shows the gene density at each y-axis value; P values calculated using the two-sided Wilcoxon rank-sum test. The underlying data for panels A, B, C, and D can be found at https://github.com/rasilab/ribosome_collisions_yeast. a.u., arbitrary unit; nt, nucleotide; N.S., not significant; TE, translation efficiency; WT, wild-type.