Humans and other commonly used model organisms are resistant to cycloheximide-mediated biases in ribosome profiling experiments

Abstract

Ribosome profiling measures genome-wide translation dynamics at sub-codon resolution. Cycloheximide (CHX), a widely used translation inhibitor to arrest ribosomes in these experiments, has been shown to induce biases in yeast, questioning its use. However, whether such biases are present in datasets of other organisms including humans is unknown. Here we compare different CHX-treatment conditions in human cells and yeast in parallel experiments using an optimized protocol. We find that human ribosomes are not susceptible to conformational restrictions by CHX, nor does it distort gene-level measurements of ribosome occupancy, measured decoding speed or the translational ramp. Furthermore, CHX-induced codon-specific biases on ribosome occupancy are not detectable in human cells or other model organisms. This shows that reported biases of CHX are species-specific and that CHX does not affect the outcome of ribosome profiling experiments in most settings. Our findings provide a solid framework to conduct and analyze ribosome profiling experiments.

Fig. 1. Cycloheximide (CHX) does not affect ribosome footprint length distribution in HEK 293T cells.

a Schematic overview of the harvesting and CHX-treatment conditions used for HEK 293T cells. CHX treatment conditions in this and all figures are indicated by color: +/+, green; −/+, yellow; −/−, pink. b Representative histograms showing the influence of CHX on footprint length and reading-frame distribution in HEK 293T libraries. The reading frame in all figures is indicated by color: 0, purple; 1, green; 2, yellow. c Same as b for HEK 293T cells treated with different CHX concentrations and incubation period. Footprints were excised between 18 and 32 nt.

Fig. 2. Cycloheximide (CHX) does not affect the translation ramp or global translation levels.

a Normalized ribosomal A-site coverage observed in long footprints (29–31 nt) for the first 200 (left) and last 200 (right) codons in HEK 293T cells in highly expressed genes (>64 reads). The solid line depicts the mean and shaded areas represent 95% confidence intervals for three biological replicates (n = 3). b Differential ribosome occupancy of open reading frames (ORFs) in HEK 293T cells across inhibitor treatments were identified using DESeq2, which uses a negative binomial distribution model with fitted mean. Uniquely mapped reads were split into short footprints (bottom left; 21–22 nt) and long footprints (top right; 28–32 nt). ORFs, after excluding the first 15 codons, were tested for differential translation (adjusted p-value ≤ 0.05) and for unaltered translation (adjusted p-value ≤ 0.05). Significantly altered ORFs are indicated in blue, unaltered transcripts in red.

Fig. 3. Cycloheximide (CHX) does not alter mammalian ribosome occupancy at the A-site and the P-site.

a Correlation analysis of transcriptome-wide A-site codon occupancy in HEK 293T cells across different CHX treatments for short and long footprints. The solid line depicts the fitted line and shaded areas represent 95% confidence intervals for three biological replicates (n = 3). Each black dot represents a codon. The size of the box indicates p-values. Correlations with a p-value > 0.05 are crossed out. b Like a for P-site codon occupancy. c Like a for E-site codon occupancy. Footprints were excised between 18 and 32 nt but are represented according to size: short footprints (21–22 nt) and long footprints (29–31 nt).

Fig. 4. Cycloheximide (CHX) pre-treatment does not alter ribosome occupancy downstream of rare codons in most species.

a Transcriptome-wide ribosome enrichment profiles according to Hussmann et al. using libraries generated from E14 mouse embryonic stem cells, zebrafish embryos, and wild-type yeast (this study) surrounding CGA and UUA codons and using different CHX treatment regimens. b Same as a for long footprints of HEK 293T cells. c Same as a for published human ribosome profiling datasets^{, , –}. d Same as a for the CGA codon in C. albicans and S. pombe (this study). This plot only uses long footprints (28–32 nt).

Fig. 5. Cycloheximide (CHX) does not explain the poor correlation between various datasets in humans.

Correlation analysis of transcriptome-wide A-site codon occupancy across data from this study and published datasets for human cells (HEK 293 and HEK 293T cells) using different CHX treatments^{, , –}. The solid line depicts the fitted line and shaded areas represent 95% confidence intervals for three biological replicates (n = 3). Each black dot represents a codon. The size of the box indicates p-value. Correlations with a p-value > 0.05 are crossed out.