Post-Transcriptional Noise Control

Abstract

Recent evidence indicates that transcriptional bursts are intrinsically amplified by messenger RNA cytoplasmic processing to generate large stochastic fluctuations in protein levels. These fluctuations can be exploited by cells to enable probabilistic bet-hedging decisions. But large fluctuations in gene expression can also destabilize cell-fate commitment. Thus, it is unclear if cells temporally switch from high to low noise, and what mechanisms enable this switch. Here, the discovery of a post-transcriptional mechanism that attenuates noise in HIV is reviewed. Early in its life cycle, HIV amplifies transcriptional fluctuations to probabilistically select alternate fates, whereas at late times, HIV utilizes a post-transcriptional feedback mechanism to commit to a specific fate. Reanalyzing various reported post-transcriptional negative feedback architectures reveals that they attenuate noise more efficiently than classic transcriptional autorepression, leading to the derivation of an assay to detect post-transcriptional motifs. It is hypothesized that coupling transcriptional and post-transcriptional autoregulation enables efficient temporal noise control to benefit developmental bet-hedging decisions.

Keywords: autoregulation; fate selection; negative feedback; noise control; post-transcriptional; splicing; stochastic noise.

Figure 1.. Transcriptional fluctuations are generally amplified by nuclear export and cytoplasmic mRNA processing, requiring feedbacks to modulate noise.

(a) Probability that a gene will show amplification versus attenuation of noise when comparing the nuclear- to-cytoplasmic noise (σ²/μ) ratio in the absence (left) and presence (right) of bursty cytoplasmic mRNA processing. Increasing red represents increasing noise amplification, while increasing blue represents increasing noise attenuation, and white represents no change in noise from nucleus to cytoplasm. Adapted from ^[56]. (b and c) Large fluctuations in protein expression levels in the absence of a feedback (left) can be amplified with a positive feedback (center) or attenuated with a negative feedback (right).

Figure 2.. Two separate feedbacks, one transcriptional, one post-transcriptional, temporarily regulate noise in HIV to control and stabilize viral fate.

(a) Tat positive feedback amplifies transcriptional fluctuations generating high noise (i.e., low stability) to probabilistically select HIV’s alternate fates early in the viral life cycle (left, green). The Rev negative feedback attenuates noise to drive viral commitment (i.e., high stability) to a specific fate, at later stages in HIVs life cycle (right, blue). (b) In the nucleus unspliced (US) transcripts are post-transcriptionally spliced into multiply spliced (MS) transcripts. Tat and Rev are both produced from MS transcripts. Tat is responsible for a positive feedback acting on the HIV long terminal repeat (LTR) promoter. Rev acts by exporting US transcripts to the cytoplasm, thereby depleting nuclear US mRNA and reducing the amount of MS mRNAs that can be produced, generating an auto-regulatory negative feedback loop. (c) Gillespie simulations in the absence (left, green) and presence (right, blue) of the Rev negative feedback show that the negative feedback efficiently attenuates noise and stochastic ON-OFF switching of the LTR is minimized, stabilizing active gene expression. In the absence of the Rev negative feedback, there is higher gene expression noise and increased stochastic ON-OFF switching of the promoter despite an ~3- fold higher mean expression-level of the ON state. Adapted from ^[47]. (d) Histograms of 1000 simulations of the HIV precursor auto-depletion model (blue) and absent this negative feedback (green) at the end of the simulation run (i.e., t=300hr from c). In the absence of the Rev- negative feedback, substantially more trajectories are expected to be in the GFP OFF state compared to the simulations with the Rev-negative feedback. In other words, the absence of the negative feedback substantially destabilizes commitment to the active state. Adapted from ^[47].

Figure 3.. Post-transcriptional architecture of HIV’s negative feedback suppresses noise to stabilize fate.

(a) i. Representative smFISH images of transcriptional pulse-chase in fixed HIV-infected Jurkat cells. The nucleus is DAPI stained (blue) and the mRNA is visualized using smFISH (white). Across: TNF activation of the HIV promoter was chased 14 min later with the transcriptional elongation inhibitor ActD. Down: smFISH probes were designed to visualize unspliced (US), singly spliced (SS), and multiply spliced (MS) mRNA respectively. ii. Quantification of nuclear mRNA molecules during the TNF pulse (blue) and the ActD chase (red) from i. Adapted from ^[47]. (b) Time-lapse microscopy of WT HIV d2GFP containing the Rev negative feedback (right, blue) and a mutant with enhanced splice-acceptor efficiency lacking the Rev negative feedback (left, green). Insets: mean trajectories normalized to max (to examine overshoot). Adapted from ^[47]. (c) Flow cytometry analysis of active-state stability following a pulse of TNF reactivation for WT HIV d2GFP and splicing mutant lacking the Rev negative feedback; cells were removed from TNF induction at time 0. As predicted in Fig. 2d, after 58 hours, the mutant lacking the Rev negative feedback shows substantially more cells in the GFP OFF (i.e., naïve) state compared to WT. Adapted from ^[47].

Figure 4.. Several feedback architectures can lead to post-transcriptional noise control

(a) Schematics of simplified gene-circuit models used for comparing effects of different negative feedback motifs on noise suppression. (b-f) Outputs of Gillespie simulations for each model shown in (a): (b) two-state model; (c) transcriptional repression; (d) mRNA precursor auto-depletion (mPAD); (e) translational repression; and (f) protein precursor auto-depletion (pPAD). (g) Fano factor (σ²/μ) of stochastic simulations comparing models shown in (a) for bursty promoter regime (k_off >> k_on). Left to right: two-state model; transcriptional (TX) repression; mRNA precursor auto-depletion (mPAD); translational (TL) repression; and protein precursor auto-depletion (pPAD). Mean and standard deviation shown for three simulation (200 iterations each). (h) Fano factor (σ²/μ) of stochastic simulations comparing models shown in (a) for constitutive promoter regime (k_on >> k_off). Left to right: two-state model; transcriptional (TX) repression; mRNA precursor auto-depletion (mPAD); translational (TL) repression; and protein precursor auto-depletion (pPAD). Mean and standard deviation shown for three simulation (200 iterations each).