Limits...
Effects of post-transcriptional regulation on phenotypic noise in Escherichia coli.

Arbel-Goren R, Tal A, Friedlander T, Meshner S, Costantino N, Court DL, Stavans J - Nucleic Acids Res. (2013)

Bottom Line: Cell-to-cell variations in protein abundance, called noise, give rise to phenotypic variability between isogenic cells.Studies of noise have focused on stochasticity introduced at transcription, yet the effects of post-transcriptional regulatory processes on noise remain unknown.Extrinsic noise provides the dominant contribution to the total protein noise over the total range of RyhB production rates.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics of Complex Systems, Weizmann Institute of Science, Rehovot 76100, Israel.

ABSTRACT
Cell-to-cell variations in protein abundance, called noise, give rise to phenotypic variability between isogenic cells. Studies of noise have focused on stochasticity introduced at transcription, yet the effects of post-transcriptional regulatory processes on noise remain unknown. We study the effects of RyhB, a small-RNA of Escherichia coli produced on iron stress, on the phenotypic variability of two of its downregulated target proteins, using dual chromosomal fusions to fluorescent reporters and measurements in live individual cells. The total noise of each of the target proteins is remarkably constant over a wide range of RyhB production rates despite cells being in stress. In fact, coordinate downregulation of the two target proteins by RyhB reduces the correlation between their levels. Hence, an increase in phenotypic variability under stress is achieved by decoupling the expression of different target proteins in the same cell, rather than by an increase in the total noise of each. Extrinsic noise provides the dominant contribution to the total protein noise over the total range of RyhB production rates. Stochastic simulations reproduce qualitatively key features of our observations and show that a feed-forward loop formed by transcriptional extrinsic noise, an sRNA and its target genes exhibits strong noise filtration capabilities.

Show MeSH

Related in: MedlinePlus

Stochastic simulations of protein variability of two genes downregulated by the same sRNA. (A) Standard deviation of protein number  as a function of the mean protein number  of one of the two target proteins (the behavior of the second protein is qualitatively similar), in simulations including intrinsic noise alone (cyan); intrinsic plus extrinsic noise in translation (red); intrinsic plus extrinsic noise in transcription (blue). The red line is a linear fit to the four points obtained with intrinsic plus extrinsic noise in translation, with largest value of . (B) Pearson correlation coefficient ρ between the protein concentrations of two genes whose transcripts are targets of the same sRNAs, as a function of the mean protein number . Empty circles represent simulation results including in addition to intrinsic noise both transcriptional and translational extrinsic noise of comparable contribution. (C) Transcriptional extrinsic noise affects directly the transcription of target genes, as well as indirectly, via the sRNA in one of the arms of an incoherent feed-forward loop configuration. (D) Total protein noise  as function of  of one of two target proteins for three different levels of transcriptional extrinsic noise, set by three values of the variance of the Gamma distribution out of which random numbers multiplying transcription rates are drawn: 0.10 (asterisks), 0.25 (empty circles) and 0.50 (full circles). Simulations were carried out when sRNA transcription is subject to transcriptional extrinsic noise in addition to intrinsic noise (blue) and when sRNA production fluctuates due to intrinsic noise alone (red). (E) Pearson correlation coefficient ρ between the concentrations of two target proteins as a function of the geometric average of the respective mean concentrations. Colors and symbols are as in D. The simulations include transcriptional bursting, as well as stochasticity effects due to binomially distributed partition of mRNA and proteins during cell division. All the simulation parameters are listed in Supplementary Table S1.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3643596&req=5

gkt184-F6: Stochastic simulations of protein variability of two genes downregulated by the same sRNA. (A) Standard deviation of protein number as a function of the mean protein number of one of the two target proteins (the behavior of the second protein is qualitatively similar), in simulations including intrinsic noise alone (cyan); intrinsic plus extrinsic noise in translation (red); intrinsic plus extrinsic noise in transcription (blue). The red line is a linear fit to the four points obtained with intrinsic plus extrinsic noise in translation, with largest value of . (B) Pearson correlation coefficient ρ between the protein concentrations of two genes whose transcripts are targets of the same sRNAs, as a function of the mean protein number . Empty circles represent simulation results including in addition to intrinsic noise both transcriptional and translational extrinsic noise of comparable contribution. (C) Transcriptional extrinsic noise affects directly the transcription of target genes, as well as indirectly, via the sRNA in one of the arms of an incoherent feed-forward loop configuration. (D) Total protein noise as function of of one of two target proteins for three different levels of transcriptional extrinsic noise, set by three values of the variance of the Gamma distribution out of which random numbers multiplying transcription rates are drawn: 0.10 (asterisks), 0.25 (empty circles) and 0.50 (full circles). Simulations were carried out when sRNA transcription is subject to transcriptional extrinsic noise in addition to intrinsic noise (blue) and when sRNA production fluctuates due to intrinsic noise alone (red). (E) Pearson correlation coefficient ρ between the concentrations of two target proteins as a function of the geometric average of the respective mean concentrations. Colors and symbols are as in D. The simulations include transcriptional bursting, as well as stochasticity effects due to binomially distributed partition of mRNA and proteins during cell division. All the simulation parameters are listed in Supplementary Table S1.

Mentions: We simulated the stochastic dynamics of a network consisting of two protein-coding genes and a third gene encoding for an sRNA that promotes the mutual degradation of each of the target transcripts with itself, using Gillespie’s method (29) (Supplementary Methods). It is noteworthy that the protein-coding genes are transcriptionally independent, as no upstream common regulatory components, such as transcription factors correlate their transcription. Pathway-specific extrinsic noise is only introduced by RyhB. In addition to the intrinsic noise stemming from stochastic effects in transcription, translation, sRNA–mRNA interaction and degradation, we introduced extrinsic cell-to-cell variations in either transcription or translation by multiplying the respective rates in each cell by a random number drawn from a Gamma distribution with mean equal to one and the variance chosen so that protein noise is comparable with the measured values. Plots of as function of the mean protein number in simulations, including intrinsic noise alone, or intrinsic noise with extrinsic noise added either in transcription or in translation, are shown in Figure 6A. The dependence of on when intrinsic noise alone is included exhibits the expected Poissonian scaling . When extrinsic noise in translation is included in addition to intrinsic noise, the behavior of is linear for large enough as experimentally observed and expected from a theoretical argument (Supplementary Text). At small , intrinsic noise becomes dominant, and the data deviate from linearity, so that approaches zero asymptotically with . A straight line fit to the large portion of the data crosses the y-axis at a non-zero value of (Figure 6A), as observed experimentally (Figure 3). Finally, when extrinsic noise in transcription is included, the behavior of approaches linearity asymptotically for large and deviates from linearity at small and intermediate values of because of the coupled sRNA–mRNA degradation terms (21). The behavior of the standard deviation of the transcript distribution as a function of the mean transcript number under the three sources of noise is illustrated in Supplementary Figure S9.Figure 6.


Effects of post-transcriptional regulation on phenotypic noise in Escherichia coli.

Arbel-Goren R, Tal A, Friedlander T, Meshner S, Costantino N, Court DL, Stavans J - Nucleic Acids Res. (2013)

Stochastic simulations of protein variability of two genes downregulated by the same sRNA. (A) Standard deviation of protein number  as a function of the mean protein number  of one of the two target proteins (the behavior of the second protein is qualitatively similar), in simulations including intrinsic noise alone (cyan); intrinsic plus extrinsic noise in translation (red); intrinsic plus extrinsic noise in transcription (blue). The red line is a linear fit to the four points obtained with intrinsic plus extrinsic noise in translation, with largest value of . (B) Pearson correlation coefficient ρ between the protein concentrations of two genes whose transcripts are targets of the same sRNAs, as a function of the mean protein number . Empty circles represent simulation results including in addition to intrinsic noise both transcriptional and translational extrinsic noise of comparable contribution. (C) Transcriptional extrinsic noise affects directly the transcription of target genes, as well as indirectly, via the sRNA in one of the arms of an incoherent feed-forward loop configuration. (D) Total protein noise  as function of  of one of two target proteins for three different levels of transcriptional extrinsic noise, set by three values of the variance of the Gamma distribution out of which random numbers multiplying transcription rates are drawn: 0.10 (asterisks), 0.25 (empty circles) and 0.50 (full circles). Simulations were carried out when sRNA transcription is subject to transcriptional extrinsic noise in addition to intrinsic noise (blue) and when sRNA production fluctuates due to intrinsic noise alone (red). (E) Pearson correlation coefficient ρ between the concentrations of two target proteins as a function of the geometric average of the respective mean concentrations. Colors and symbols are as in D. The simulations include transcriptional bursting, as well as stochasticity effects due to binomially distributed partition of mRNA and proteins during cell division. All the simulation parameters are listed in Supplementary Table S1.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3643596&req=5

gkt184-F6: Stochastic simulations of protein variability of two genes downregulated by the same sRNA. (A) Standard deviation of protein number as a function of the mean protein number of one of the two target proteins (the behavior of the second protein is qualitatively similar), in simulations including intrinsic noise alone (cyan); intrinsic plus extrinsic noise in translation (red); intrinsic plus extrinsic noise in transcription (blue). The red line is a linear fit to the four points obtained with intrinsic plus extrinsic noise in translation, with largest value of . (B) Pearson correlation coefficient ρ between the protein concentrations of two genes whose transcripts are targets of the same sRNAs, as a function of the mean protein number . Empty circles represent simulation results including in addition to intrinsic noise both transcriptional and translational extrinsic noise of comparable contribution. (C) Transcriptional extrinsic noise affects directly the transcription of target genes, as well as indirectly, via the sRNA in one of the arms of an incoherent feed-forward loop configuration. (D) Total protein noise as function of of one of two target proteins for three different levels of transcriptional extrinsic noise, set by three values of the variance of the Gamma distribution out of which random numbers multiplying transcription rates are drawn: 0.10 (asterisks), 0.25 (empty circles) and 0.50 (full circles). Simulations were carried out when sRNA transcription is subject to transcriptional extrinsic noise in addition to intrinsic noise (blue) and when sRNA production fluctuates due to intrinsic noise alone (red). (E) Pearson correlation coefficient ρ between the concentrations of two target proteins as a function of the geometric average of the respective mean concentrations. Colors and symbols are as in D. The simulations include transcriptional bursting, as well as stochasticity effects due to binomially distributed partition of mRNA and proteins during cell division. All the simulation parameters are listed in Supplementary Table S1.
Mentions: We simulated the stochastic dynamics of a network consisting of two protein-coding genes and a third gene encoding for an sRNA that promotes the mutual degradation of each of the target transcripts with itself, using Gillespie’s method (29) (Supplementary Methods). It is noteworthy that the protein-coding genes are transcriptionally independent, as no upstream common regulatory components, such as transcription factors correlate their transcription. Pathway-specific extrinsic noise is only introduced by RyhB. In addition to the intrinsic noise stemming from stochastic effects in transcription, translation, sRNA–mRNA interaction and degradation, we introduced extrinsic cell-to-cell variations in either transcription or translation by multiplying the respective rates in each cell by a random number drawn from a Gamma distribution with mean equal to one and the variance chosen so that protein noise is comparable with the measured values. Plots of as function of the mean protein number in simulations, including intrinsic noise alone, or intrinsic noise with extrinsic noise added either in transcription or in translation, are shown in Figure 6A. The dependence of on when intrinsic noise alone is included exhibits the expected Poissonian scaling . When extrinsic noise in translation is included in addition to intrinsic noise, the behavior of is linear for large enough as experimentally observed and expected from a theoretical argument (Supplementary Text). At small , intrinsic noise becomes dominant, and the data deviate from linearity, so that approaches zero asymptotically with . A straight line fit to the large portion of the data crosses the y-axis at a non-zero value of (Figure 6A), as observed experimentally (Figure 3). Finally, when extrinsic noise in transcription is included, the behavior of approaches linearity asymptotically for large and deviates from linearity at small and intermediate values of because of the coupled sRNA–mRNA degradation terms (21). The behavior of the standard deviation of the transcript distribution as a function of the mean transcript number under the three sources of noise is illustrated in Supplementary Figure S9.Figure 6.

Bottom Line: Cell-to-cell variations in protein abundance, called noise, give rise to phenotypic variability between isogenic cells.Studies of noise have focused on stochasticity introduced at transcription, yet the effects of post-transcriptional regulatory processes on noise remain unknown.Extrinsic noise provides the dominant contribution to the total protein noise over the total range of RyhB production rates.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics of Complex Systems, Weizmann Institute of Science, Rehovot 76100, Israel.

ABSTRACT
Cell-to-cell variations in protein abundance, called noise, give rise to phenotypic variability between isogenic cells. Studies of noise have focused on stochasticity introduced at transcription, yet the effects of post-transcriptional regulatory processes on noise remain unknown. We study the effects of RyhB, a small-RNA of Escherichia coli produced on iron stress, on the phenotypic variability of two of its downregulated target proteins, using dual chromosomal fusions to fluorescent reporters and measurements in live individual cells. The total noise of each of the target proteins is remarkably constant over a wide range of RyhB production rates despite cells being in stress. In fact, coordinate downregulation of the two target proteins by RyhB reduces the correlation between their levels. Hence, an increase in phenotypic variability under stress is achieved by decoupling the expression of different target proteins in the same cell, rather than by an increase in the total noise of each. Extrinsic noise provides the dominant contribution to the total protein noise over the total range of RyhB production rates. Stochastic simulations reproduce qualitatively key features of our observations and show that a feed-forward loop formed by transcriptional extrinsic noise, an sRNA and its target genes exhibits strong noise filtration capabilities.

Show MeSH
Related in: MedlinePlus