Ribosome demand links transcriptional bursts to protein expression noise
- Department of Bioscience and Biotechnology, IIT Kharagpur, India
Figures
Figure 1 with 3 supplements
High mRNA noise combined with high translational efficiency leads to high protein noise.
(A) Existing model describing the impact of translational efficiency on protein expression noise. (B) Genes were classified into 16 classes according to the quartiles of mean mRNA expression, calculated from Nadal-Ribelles et al., 2019, and then by the quartiles of protein synthesis rates per mRNA from Riba et al., 2019. The protein noise values for genes in each of the classes were obtained from Newman et al., 2006, and the measure distance-to-median (DM) value, as derived in their work, was considered as the measure of noise. (C) Two-state model of gene expression with the transition rates Kon and Koff between transcriptional ON and OFF states was used for stochastic simulations. (D, E) Relationship between mean protein expression and protein noise (coefficient of variation, CV) obtained from stochastic modeling using the two-state model. The panel (D) describes the results of stochastic simulations obtained at different starting mRNA numbers of a gene to test whether the mRNA expression level of a gene can explain the positive relationship between mean protein expression and protein noise. The panel (E) shows the results obtained from stochastic simulations at different transcriptional burst frequencies, but keeping the starting mRNA number of the gene constant. (F) Mean-adjusted mRNA expression noise calculated from the single-cell RNA-seq data Nadal-Ribelles et al., 2019 in 16 classes of genes classified according to the quartiles of mean mRNA expression and the quartiles of translational efficiency based on the data on protein synthesis rate per mRNA from Riba et al., 2019. Q1, Q2, and Q3 represent first, second, and third quartiles.
Figure 1—figure supplement 1
Density distributions of experimental parameters utilized in the present study.
Distribution of (A) mean-mRNA expression from Nadal-Ribelles et al., 2019, (B) protein synthesis rate per mRNA from Riba et al., 2019, and (C) protein expression noise (distance-to-median, DM) from Newman et al., 2006.
Figure 1—figure supplement 2
Correlation of protein expression noise, distance-to-median (DM) from Newman et al., 2006, (A) with mean mRNA expression, calculated from Nadal-Ribelles et al., 2019, and (B) with protein synthesis rate per mRNA from Riba et al., 2019.
Figure 1—figure supplement 3
Calculation of mRNA expression noise in yeast and its correlation with mean mRNA expression and protein expression noise.
(A) Calculation of mean-adjusted mRNA expression noise from the plot of coefficient of variation vs ln(mean mRNA expression), calculated from Nadal-Ribelles et al., 2019. (B) Distribution of mean-adjusted mRNA expression noise for 5500 genes in yeast. (C) Correlation between mean RNA expression and mRNA expression noise, calculated from Nadal-Ribelles et al., 2019. (D) Correlation between protein expression noise, distance-to-median (DM) (Newman et al., 2006), and mean-adjusted mRNA expression noise. The solid black line represents the linear regression line between mean-adjusted mRNA expression noise and protein expression noise (DM). The gray dotted lines represent ±1 s.d. lines and the brown dotted lines represent ±2 s.d. lines. The orange points show genes which fall outside the ±2 s.d. lines.
Figure 2 with 4 supplements
Stochastic fluctuation in mRNA expression, originating from transcriptional bursts, combined with high translational efficiency generates high protein noise.
(A) The new working model postulated that the genes with bursty transcription (low transcriptional burst frequency) and high translational efficiency were likely to exhibit higher protein expression noise compared to the genes with bursty transcription but low translational efficiency. (B) Estimation of parameters of two-state model of gene expression from single-cell RNA-seq data as described by Kim and Marioni, 2013. (C) Protein noise of genes with different levels of transcriptional burst frequencies and translational efficiency, estimated by protein synthesis rates per mRNA (Nadal-Ribelles et al., 2019). (D) Protein expression noise (distance-to-median [DM] values from Newman et al., 2006) of genes classified into 16 classes based on burst frequency and translational efficiency (protein synthesis rate per mRNA; Riba et al., 2019). Q1, Q2, and Q3 represent first, second, and third quartiles, respectively.
Figure 2—figure supplement 1
Distributions of burst parameters for yeast genes, estimated using the two-state model of transcription, and correlations between them.
(A) Distribution of values of parameters of transcriptional bursts (Kon, Koff, and βm) for yeast genes estimated from Nadal-Ribelles et al., 2019. (B) Relationship between Kon, Koff, and mean-adjusted mRNA expression noise for yeast genes. (C) Relationship between Kon, βm, and mean-adjusted mRNA expression noise for yeast genes.
Figure 2—figure supplement 2
Relationship between transcriptional burst frequency (Kon), protein synthesis rate per mRNA (Riba et al., 2019), and mean-adjusted mRNA expression noise.
Figure 2—figure supplement 3
Distribution of tRNA adaptation index (tAI) across yeast genes and its correlation with experimentally measured protein synthesis rate per mRNA molecule.
(A) Distribution of tRNA adaptation index (tAI) values of yeast genes calculated from Sabi and Tuller, 2014 according to the method described in Tuller et al., 2010. (B) Correlation between tAI and protein synthesis rate per mRNA (Riba et al., 2019) for yeast genes.
Figure 2—figure supplement 4
Correlation of tAI with burst parameters and its association with protein expression noise.
(A) Relationship between transcriptional burst frequency (Kon), tRNA adaptation index (tAI), and mean-adjusted mRNA expression noise. (B) Relationship between transcriptional burst frequency (Kon), tAI, and protein expression noise (distance-to-median, DM) (Newman et al., 2006). (C) Protein expression noise (DM) in 16 classes of yeast genes classified based on the quartiles of transcriptional burst frequency (Kon) and then by the quartiles of tAI.
Figure 3
The model combining transcriptional and translational bursting does not explain the positive correlation between translational efficiency and protein noise.
(A) Schematic diagram depicting the integrated model of transcriptional and translational bursting. (B) Ribosomal traversal speed along an mRNA molecule is given by V at a position L in the mRNA molecule. As multiple ribosomes could translate a single mRNA molecule at the same time, a second translation initiation happened only when the preceding ribosome traversed at least 10 codons, to account for steric interaction between ribosomes (Steitz, 1969; Ingolia et al., 2009). The ribosome traversal speed was modeled using a Hill function as several studies have shown the presence of a gradually increasing profile of translational efficiency or ramp in the 5′ end of coding regions of genes (Tuller et al., 2010; Weinberg et al., 2016). (C) Traversal time calculated as a function of KHill from the Hill function for a gene with 300 codons, and the maximum traversal speed of 100 codons per minute. (D) Relationship between KHill, translation initiation rate, and ribosome traversal time. Faster ribosome traversal enabled higher translation initiation rate (TLinit) (Barrington et al., 2023). A and B are parameters of the model. (E) The results obtained from stochastic simulations using the combined model of transcriptional and translational bursting. Protein noise changes with changes in translational efficiency and transcriptional burst frequency but does not reveal a positive correlation between mean protein expression and protein noise. (F) Protein noise obtained from the combined model changes with changes in translation initiation rate and translational burst frequency but does not explain the positive correlation between mean protein expression and protein noise.
Figure 4 with 1 supplement
Inclusion of ribosome demand associated with translation of mRNA molecules of a gene can reveal positive correlation between translational efficiency and protein noise.
(A) Schematic diagram depicting how ribosome demand for translation of mRNA molecules can vary with bursty transcription and bursty translation. As mRNA numbers of the gene fluctuate due to bursty transcription, high translational efficiency can lead to intermittent elevated ribosome demand for translation of the mRNA molecules of that gene. This can lead to increased inter-individual variation in protein numbers. (B) Uniform transcription of a gene does not lead to a sudden elevated ribosome demand for translation, thereby reducing inter-individual variation in protein numbers. (C) Results from simulations with three different functions (function 11, 12, and 16 from Supplementary file 1) that model the impact of ribosome demand on translational efficiency. Results from simulations with other functions to model ribosome demand are shown in Figure 4—figure supplement 1. (D) The relationship between mean protein expression and protein noise at different transcriptional burst frequencies obtained from stochastic simulations with the model incorporating ribosome demand along with transcriptional and translational bursting. The ribosome demand was modeled using function 16 (Supplementary file 1). For each transcriptional burst frequency, the translational efficiency was altered by changing the translation initiation rate (TLinit) while keeping the rest of the parameters constant. The figures show mean ± 1 s.d. values obtained from simulations.
Figure 4—figure supplement 1
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations based on different mathematical functions to model ribosome demand (Supplementary file 1).
Functions 11, 12, and 16 are repeated from Figure 4 for sake of completeness. The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5 with 10 supplements
Ribosome demand is necessary for the positive correlation between mean protein expression and protein noise.
(A) The relationship between mean protein expression and protein noise at different translational burst frequencies obtained from stochastic simulations with the model incorporating ribosome demand along with transcriptional and translational bursting. The ribosome demand was modeled using function 16 (Supplementary file 1). For each transcriptional burst frequency, the translational efficiency was altered by changing the translation initiation rate (TLinit) while the rest of the parameters were kept constant. (B) Stochastic simulations where mean protein expression was altered by changing base translation initiation rate (TLinit), thus altering ribosome demand, but keeping ribosome traversal speed constant, maintained positive correlation between mean protein expression and noise. This was done by keeping KHill constant at specific values during simulations that constrained the traversal time (Equation 10). (C) Stochastic simulations where mean protein expression was altered by changing the ribosome traversal speed (Equation 10) but keeping the base translation initiation rate (TLinit) constant, and thus, not allowing variation in ribosome demand with changes in ribosome traversal rate, abolished the positive correlation between mean protein expression and protein noise. The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 1
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter A.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 2
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter B.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 3
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter Vmax.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 4
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter Koff.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 5
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter βm.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 6
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for different values of the parameter TLoff.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 7
Relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for sets of parameter values obtained by random sampling of parameter space.
The figures show mean± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 8
The relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for sets of parameter values obtained by random sampling of parameter space.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 9
The relationship between mean protein expression and protein noise (coefficient of variation, CV) derived from stochastic simulations for sets of parameter values obtained by random sampling of parameter space.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 5—figure supplement 10
Stochastic simulations where mean protein expression was altered by changing the ribosome traversal speed (Equation 10) but keeping the base translation initiation rate (TLinit) constant, and thus, not allowing variation in ribosome demand with changes in ribosome traversal speed abolished the positive correlation between mean protein expression and protein noise.
The figures show mean ± 1 s.d. values obtained from simulations.
Figure 6 with 6 supplements
Impact of changes in translational efficiency on protein noise is dependent on the transcriptional burst characteristics of promoters.
(A) Gene–promoter construct for genomic integration and noise measurement. (B) Noise estimation from a homogeneous group of cells with similar cell size and complexity. (C) Protein noise (distance-to-median, DM) (Newman et al., 2006) vs burst frequency values for yeast genes. Burst frequency values were estimated from single-cell RNA-seq data (Nadal-Ribelles et al., 2019) using the method described by Kim and Marioni, 2013. (D) Measured mean protein expression and protein noise of the promoters of RPL35A, RPG1, CPA2, and QCR2 with the wild-type GFP gene. (E) Relationship between normalized protein noise and normalized mean expression of the GFP mutants in the bursty promoters CPA2 and QCR2. (F) Relationship between normalized protein noise and normalized mean expression of the GFP mutants in the non-bursty promoters RPG1 and RPL35A. The figures show mean ± 1 s.d. values quantified from flow cytometry experiments. Three clones for each mutant were measured.
Figure 6—figure supplement 1
Structure of the promoter–GFP constructs along with auxotrophic marker (HIS3MX6) and homologous recombination sites (His3L and His3R) for genomic integration into yeast.
The restriction sites for building the construct are shown by arrows.
Figure 6—figure supplement 2
Snapshot of a flow cytometry experiment for measurement of mean protein expression and protein noise.
Figure 6—figure supplement 3
Fitting of ellipses (shown by red color) with different values of major and minor axes (‘a’ and ‘b’) and identification of the best fit that chooses a homogenous set of cells and contains at least ~20,000 cells for quantification of noise.
Figure 6—figure supplement 4
Experimental measurement of mean protein expression and protein noise for GFP mutants under the regulatory control of bursty and non-bursty promoters.
Normalized mean protein expression (A) and normalized protein noise (B) of GFP variants under the regulation of two bursty promoters CPA2 and QCR2, and two non-bursty promoters RPG1 and RPL35A. The p-values show the results of the Mann–Whitney U-test to test for differences in normalized mean expression of normalized protein noise between a GFP variant and the wild-type. The figures show mean ± 1 s.d. values obtained from flow cytometry measurements. Three clones for each mutant were measured.
Figure 6—figure supplement 5
Correlation between protein expression noise and protein synthesis rate and tAI for classes of genes with different busrt frequency parameter value.
(A) Correlation between protein synthesis rate per mRNA (Riba et al., 2019) and protein noise, distance-to-median (DM) (Newman et al., 2006), for groups of genes with different values of transcriptional burst frequency (Kon). (B) Correlation between tRNA adaptation index (tAI) and protein noise, DM (Newman et al., 2006), for groups of genes with different values of transcriptional burst frequency (Kon). Q1, Q2, and Q3 represent quartiles.
Figure 6—figure supplement 6
Construction of GFP mutant variants along with different promoters for genomic integration into yeast by overlap-extension PCR.
Additional files
-
Supplementary file 1
List of mathematical functions explored to model ribosome demand.
- https://cdn.elifesciences.org/articles/99322/elife-99322-supp1-v1.docx
-
Supplementary file 2
List of primers used.
- https://cdn.elifesciences.org/articles/99322/elife-99322-supp2-v1.docx
-
Supplementary file 3
Fragment size and nucleotide sequence of the parts of the promoter–GFP constructs.
- https://cdn.elifesciences.org/articles/99322/elife-99322-supp3-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/99322/elife-99322-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Ribosome demand links transcriptional bursts to protein expression noise
eLife 13:RP99322.
https://doi.org/10.7554/eLife.99322.3
