A Stress Response that Monitors and Regulates mRNA Structure Is Central to Cold Shock Adaptation

Abstract

Temperature influences the structural and functional properties of cellular components, necessitating stress responses to restore homeostasis following temperature shift. Whereas the circuitry controlling the heat shock response is well understood, that controlling the E. coli cold shock adaptation program is not. We found that during the growth arrest phase (acclimation) that follows shift to low temperature, protein synthesis increases, and open reading frame (ORF)-wide mRNA secondary structure decreases. To identify the regulatory system controlling this process, we screened for players required for increased translation. We identified a two-member mRNA surveillance system that enables recovery of translation during acclimation: RNase R assures appropriate mRNA degradation and the Csps dynamically adjust mRNA secondary structure to globally modulate protein expression level. An autoregulatory switch in which Csps tune their own expression to cellular demand enables dynamic control of global translation. The universality of Csps in bacteria suggests broad utilization of this control mechanism.

Keywords: DMS-seq; RNA regulation; RNA secondary structure; cold shock; cold shock proteins; ribosome profiling; translation.

Figure 1. Global change in translation and mRNA structure during the acclimation phase following cold shock

(A) Top: Cell growth after shift to 10°C. Cells gro wing exponentially at 37°C were shifted to 10°C (using dilution to achieve instantaneous tempe rature equilibration indicated by the arrow at t = 0; see Methods), and grown until exponential phase was restored (~25 hours after cold shock). Bottom: Protein synthesis after shift to 10°C. Cultures were simultaneously sampled for OD₄₂₀ and protein synthesis (³⁵S-methionine pulse labeling) at different timepoints (see Methods). Protein synthesis rate is calculated as: ³⁵S-methionine incorporated / minute / OD₄₂₀ / 100µL culture. The acclimation phase is shaded gray and marked by a red arrow. (B) Detailed time course of protein synthesis immediately after shift to 10°C, measured as in (A). (C) Meta-gene analysis of ribosome run-off after cold shock. At 5 min and 15 min after cold shock, the median of ribosome density at each position relative to the ORF start codon was calculated across well-expressed genes (N = 492) and normalized to the median ribosome density within the window between 1000 nt and 1180 nt downstream of start codon. Analysis is limited to ORFs ≥ 1200 nt long, and curves are smoothed by calculating the median within rolling windows of 25 nt (see Methods). (D) Histogram of Gini indices of E. coli ORFs calculated from in vivo DMS-seq data at 37°C (red) or 30 min after cold shock (blue). Gini indices were calculated for 391 well-expressed genes with relatively constant mRNA across the ORF and a mean DMS-seq signal of ≥ 5 reads/nt in all samples (see Figure S1C-D for determination of read cut-off). P-value < 2.2^-16, based on Kolmogorov-Smirnov (K-S) test. (E) Scatter plot comparing Gini indices of individual ORFs at 37°C and 30 min after cold shock to 10°C. (F) Histogram of Gini indices of the 391 ORFs described in (D) at 30 min (blue) or 6 hr (yellow) after cold shock in WT cells. P-value < 2.2^-16, based on K-S test. (G) Plot of the change in Gini indices vs translation efficiency (TE) between 30 min and 6 hr after cold shock. The distribution of changes in RNA structure level is shown in a histogram above the figure and that for TE changes is shown to right. Most genes fall in the upper left quadrant, indicating a decrease in RNA structure level concomitant with an increase in TE. The change of TE during acclimation phase is highly correlated with the change of mRNA structure (Spearman's rank correlation coefficient ρ = −0.65). See also Figure S1 and Table S1.

Figure 2. Key factors involved in the translation recovery circuit during acclimation phase

(A) Scatter plot comparing ORF protein abundance at 37°C and at the end of acclimation (6 hr after cold shock). Grey dashed line: Y = X; red dashed line: 5-fold increase in protein abundance after acclimation relative to 37°C. Red p oints above the line are well expressed ORFs whose protein abundance increased ≥ 5-fold (N = 45), with the labeled ORFs increasing ≥ 10-fold (N = 20). See also Table S2. (B) Deleting RNase R and the Csps inhibits translation recovery during acclimation. Bar graph indicates the ratio of cellular translation (measured by ³⁵S-methionine pulse labeling) at 6 hr vs 30 min after cold shock for WT and mutant strains. Error bars: standard deviation across >2 replicates. (C) Fraction of total protein translation (upper panel) and total mRNA amount (lower panel) for the 45 most highly induced genes at different time points after cold shock. See also Table S3. See also Figure S2.

Figure 3. RNase R facilitates mRNA degradation during acclimation phase

(A-C) RNA content of WT and Δrnr cells at 20 min, 4 hr and 8 hr after cold shock for: (A) stable RNA; (B) tmRNA; and (C) mRNA. RNA content was calculated from the fraction of RNA-seq reads mapping to different types of RNA, normalized to total RNA level measured by continuous labeling of ³H-uridine during 37°C growth and after cold shock (see Methods). (D) mRNA content of WT and Δrnr cells before and after rifampicin (rif) treatment at 10°C. Upper: schematic of experiment. Lower: mRNA / total RNA before and after the 2 hr treatment with rifampicin, calculated from the fraction of RNA-seq reads mapping to mRNA. (E-F) mRNA amount of individual genes (N = 937) in WT (E) or Δrnr cells (F) before and after rif treatment diagrammed in (D). mRNA level was quantified by number of RNA-seq Reads Per Kilobase of transcript per Million mapped reads (RPKM). See also Figure S3 and Table S5.

Figure 4. Csps are required for decreased mRNA structure during acclimation

(A) Histogram of Gini indices of ORFs in WT (orange) or ΔcspABEG cells (green) at 8 hr after cold shock (N = 592). Gini indices were calculated from the in vivo DMS-seq data. P-value < 2.2^-16, based on K-S test. (B) Plot of the difference in Gini indices vs the difference in translation efficiency (TE) between WT and ΔcspABEG cells at 8 hr after cold shock. Histograms of the differences in RNA structure level (above) and of TE differences (right) are shown. Most genes fall in the lower right quadrant, indicating that ΔcspABEG cells have higher mRNA structure level and lower TE than WT cells. The difference of TE between WT and ΔcspABEG cells is highly correlated with the difference of their mRNA structure level (Spearman's rank correlation coefficient ρ = −0.58). (C) Growth curves (OD₄₂₀) of WT, ΔcspA, ΔcspABEG and ΔcspABCEG cells at 37°C in MOPS rich defined media minus methionine on a log scale. Doubling times: WT 27.6 ± 0.2 min; ΔcspA 27.6 ± 0.3 min; ΔcspABEG 28.4 ± 0.2 min; ΔcspABCEG 29.8 ± 0.2 min, showing mean ± SD based on 3 replicates. (D) Distribution of the change in TE in ΔcspABCEG cells compared to WT cells at 37°C. Genes were binned into 9 groups based on their TEs in WT cells, with mean TE of each group indicated on the X-axis. The distribution of changes in TE in ΔcspABCEG vs WT cells was then calculated for each group of ORFs, with the box center indicating the median, and the box length indicating the 25^th ~ 75^th percentile of TE change. See also Table S6.

Figure 5. The 5’UTR of cspA changes in mRNA structure during acclimation

(A) The normalized in vivo DMS-seq signal of A/C bases within cspA 5’UTR in WT cells at 37°C (top), 30 min (middle) or 6 hr (bottom) after cold shock. DMS-seq signals were normalized to the maximum signal within cspA message after removing outliers by 98% Winsorisation (see Methods). The red dashed line represents the signal cutoff (0.24), above which the A/C bases are predicted to be unpaired. The region highlighted in red has long-range interactions with the “cold box” element at 10°C. (B-C) The predicted structure of the cspA 5’ UTR at (B) 37°C or (C) 30 min after cold shock. Structure predictions were generated by constraining a minimum free-energy prediction with in vivo DMS-seq data. The start codon of cspA (green), the conserved “cold box” element (blue) and its long-range interaction region at 10°C (red) are highlighted. (D) Scatter plot comparing Gini indices of ORFs (N = 391) and 5’UTR of cspA, cspB, cspG (red dots) at 30 min vs 6 hr after cold shock. Grey dashed line: Y = X. (E) The predicted structure of the cspA 5’ UTR at 6 hr after cold shock, as shown in (B-C). See also Figures S4, S5 and S6.

Figure 6. Csp protein autoregulates its own 5’ UTR in vitro

(A) The in vitro DMS-MaPseq signal of the A/C bases within cspA 5’UTR at 37°C (top) and at 10°C without CspA protein (middle) or with 0.1mM CspA protein (bottom). The region highlighted in red has long range interactions with the “cold box” element at 10°C. The DMS-MaPseq signals from biological replicates are first normalized to the maximum signal within the cspA 5’UTR, and then averaged. Error bars indicate standard deviation of 2–4 replicates. The red dashed line indicates the threshold (0.27) above which the A/C bases are predicted to be unpaired (see Methods). (B-D) The in vitro structures of cspA 5’UTR at (B) 37°C, (C) 10°C and (D) 10°C with 0.1mM CspA protein, generated as in Figure 5. The start codon of cspA (green), the conserved “cold box” element (blue) and its long-range interaction region at 10°C (red) are highlighted. See also Figure S7.

Figure 7. An mRNA structure surveillance system mediated by Csps and RNase R is central to translation recovery upon cold shock

(A) ORF-wide mRNA structure is the strongest predictor of translation efficiency of endogenous genes in vivo during normal growth at 37°C. Orange ovals: transl ating ribosomes; green arrows: elongation rate. The level of CspA (red ovals) is limited by instability of its RNA mediated by its 5’UTR structure (indicated by black rectangle). (B) Cold shock induces the expression of Csps and RNase R. After cold shock, genome-wide mRNA structure level increases and translation initiation and elongation decrease. The 5’UTR of cspA mRNA undergoes a structural change due to the temperature downshift (indicated by red rectangle), which is critical for increased expression. (C) During acclimation, cellular translation is recovered via an mRNA structure surveillance system mediated by Csps and RNase R. Early after cold shock, Csps predominantly interact with cellular mRNA as RNA chaperones to decrease mRNA structure level and increase translation globally (B). As recovery proceeds and the Csp concentration increases, Csps bind their own 5’UTRs, triggering a conformational change (indicated by a black rectangle) that leads to decreased expression level. RNase R is required for facilitating degradation of the structured mRNA induced by temperature downshift, assisting Csps function by restoring the appropriate Csp/mRNA balance.