Translation inhibitors cause abnormalities in ribosome profiling experiments

Abstract

Ribosome profiling and high-throughput sequencing provide unprecedented opportunities for the analysis of mRNA translation. Using this novel method, several studies have demonstrated the widespread role of short upstream reading frames in translational control as well as slower elongation at the beginning of open reading frames in response to stress. Based on the initial studies, the importance of adding or omitting translation inhibitors, such as cycloheximide, was noted as it markedly affected ribosome coverage profiles. For that reason, many recent studies omitted translation inhibitors in the culture medium. Here, we investigate the influence of ranging cycloheximide concentrations on ribosome profiles in Saccharomyces cerevisiae and demonstrate that increasing the drug concentration can overcome some of the artifacts. We subjected cells to various manipulations and show that neither oxidative stress nor heat shock nor amino acid starvation affect translation elongation. Instead, the observations in the initial studies are the result of cycloheximide-inflicted artifacts. Likewise, we find little support for short upstream reading frames to be involved in widespread protein synthesis regulation under stress conditions. Our study highlights the need for better standardization of ribosome profiling methods.

Figure 1.

Ribosome profiling. Cell lysis releases a mixture of individual ribosomal subunits, assembled ribosomes in complex with mRNA and blank ribosomes with no RNA attached. Sucrose gradient fractionation allows separation and isolation of these components. Captured mRNA fragments are then sequenced on an Illumina platform.

Figure 2.

Ribosomal occupancy profiles and the effect of stress and drug treatment. (A) Control yeast cells and cells treated with hydrogen peroxide (0.2 mM) in the presence or absence of 100 μg/ml cycloheximide in the media. Nucleotide position count is relative to start codon. (B) Ribosome occupancy profiles of yeast cells undergoing heat shock (42°C, 20 min). The peak appears only when cycloheximide is added to the medium. (C) None of the three tested stress types lead to a significant increase of ribosomes at the 5′ proximities of reading frames in the absence of cycloheximide. Refer to Figure 3A and B for additional details on amino acid starvation. (D and E) Concentration of cycloheximide in the medium affects the shape of the profile, pointing to a passive diffusion model of cycloheximide entering live cells. Cells were grown in YPD medium in the absence of stress (D) or subjected to oxidative stress (0.2 mM hydrogen peroxide, 30 min) (E). Cycloheximide concentration does not immediately reach the threshold, under which all ribosomes are inhibited with 100% efficiency, instead increasing gradually. Therefore, following the treatment some ribosomes initiating translation continue protein synthesis until they encounter the drug, leading to a broad cumulative peak in the ribosomal occupancy profile.

Figure 3.

In-depth investigation of amino acid starvation and changes in the ribosomal profile. (A) Repeating the experiment as was done in (1) but without cycloheximide pre-treatment still leads to a slightly different ribosomal occupancy profiles (gray and dark brown lines on the graph). However, the yeast strain BY4741, used in that study is auxotroph in histidine, leucine and methionine, which are used as selective markers. Depletion of culture medium of all amino acids cannot be considered as starvation, because the lack of three essential amino acids will lead to cell death rather than to metabolism switching toward synthesis of its own amino acids. Therefore, we supplemented the medium without amino acids with normal levels of His, Met and Leu. As a result, the difference in ribosomal profiles between starved and non-starved conditions disappeared. Thus, amino acid starvation does not cause the accumulation of ribosomes at the beginning of ORFs or uORFs. The only scenario when this accumulation was observed is the absolute lack of the essential amino acids, leading to ribosome stalling at corresponding codons. This is a very extreme case, which has little in common with regulation per se. (B) Ribo-seq data published in (1) were processed with our analytical approaches (cycloheximide was present in the culture medium). Brown line is based on our experiment (same as in Figure 3A). (C) Footprint coverage of GCN4 in response to amino acid stress derived from our experiments.

Figure 4.

uORF and 5′ UTR coverage in ribosome profiling experiments. (A) There is a dramatic difference in cumulative 5′ UTR occupancy if cycloheximide treatment is omitted. (B) None of the three examined stresses significantly increase 5′ UTR ribosomal occupancy. Although oxidative stress and amino acid starvation do slightly increase 5′ UTR occupancy, the effect is minimal compared to what was previously found (1,2). (C) Addition of translation initiation inhibitors does not affect cumulative ribosomal occupancy at the 5′ UTR.

Figure 5.

Ribo-seq of the small ribosomal subunit. (A) Dissociation of monosomes and polysomes into subunits. Fractions corresponding to the 40S small ribosomal subunit were collected for sequencing. (B) Shares of reads aligned to 18S and 25S rRNAs in Ribo-seq samples. The upper chart shows typical shares of reads in 80S fraction in control cells (YPD media, log growth phase, no stress, no cycloheximide pretreatment), and the lower shares of reads in the 40S fraction. Although not precisely quantitative, it gives an estimation of 40S fraction impurity. The ‘other’ category combines footprints and unaligned reads. (C) Representative footprint coverage profiles from normal Ribo-seq (gray) and 40S fraction (black). The dashed line marks the location of the GCN4 uORFs. (D) Footprint length distribution in 80S and 40S fractions. Reads aligned to 18S rRNA are given as a size selection control for 40S fraction. Note: each distribution class has its own scale. (E) Cumulative coverage of footprints derived from 40S fraction. No normalization was applied prior to graph plotting. All genes regardless of their length were aligned by their start codon.