Manipulation of topoisomerase expression inhibits cell division but not growth and reveals a distinctive promoter structure in Synechocystis

Abstract

In cyanobacteria DNA supercoiling varies over the diurnal cycle and is integrated with temporal programs of transcription and replication. We manipulated DNA supercoiling in Synechocystis sp. PCC 6803 by CRISPRi-based knockdown of gyrase subunits and overexpression of topoisomerase I (TopoI). Cell division was blocked but cell growth continued in all strains. The small endogenous plasmids were only transiently relaxed, then became strongly supercoiled in the TopoI overexpression strain. Transcript abundances showed a pronounced 5'/3' gradient along transcription units, incl. the rRNA genes, in the gyrase knockdown strains. These observations are consistent with the basic tenets of the homeostasis and twin-domain models of supercoiling in bacteria. TopoI induction initially led to downregulation of G+C-rich and upregulation of A+T-rich genes. The transcriptional response quickly bifurcated into six groups which overlap with diurnally co-expressed gene groups. Each group shows distinct deviations from a common core promoter structure, where helically phased A-tracts are in phase with the transcription start site. Together, our data show that major co-expression groups (regulons) in Synechocystis all respond differentially to DNA supercoiling, and suggest to re-evaluate the long-standing question of the role of A-tracts in bacterial promoters.

Figure 1.

Homeostasis and Twin-Domain Models of DNA Supercoiling. (A) Global homeostasis of supercoiling by direct feedback on the expression of topoisomerases (GYR: Gyrase holoenzyme; TopoI: topoisomerase I) and G+C-rich anabolic/growth genes and A+T-rich catabolic and stress-response genes. The gray coils reflect relaxed (left) or supercoiled DNA (right). Dashed arrows indicate transcription and solid arrows catalytic conversions. Green arrows indicate the manipulations of this core regulatory hub studied in this work and the underlying hypothesis that these could be used to redirect metabolic energy towards desired products. (B) Transcription-dependent DNA supercoiling accumulates downstream (positive) and upstream (negative) of the RNA polymerase, widely known as the twin-domain model. If unresolved by TopoI and gyrase, this would lead to RNA polymerase stalling (blue arrow) and R-loop formation. (C) The torsional stress exerted by transcription can lead to long-distance cooperative and antagonistic effects, where negative supercoiling upstream facilitates and positive supercoiling downstream blocks transcription from adjacent loci.

Figure 2.

Batch culture endpoint measurements. Overexpression and knockdown strains of this study where grown for 5 days in BG11 medium supplemented with all required antibiotics, and all inducers for the plasmid constructs in each experiment (100 ng/ml aTc, 1 mM l-rhamnose). (A) The optical density at 750 nm (OD₇₅₀) was measured daily and cell dry weight (CDW) determined directly after the last measurement on day 5. (B, C) Electropherograms of chloroquine-supplemented agarose gels (1.2% agarose, 20 μg ml⁻¹ chloroquine) of plasmids extracted at harvest time (B) of the cultures in (A), or as a time series (growth curve, Supplementary Figure S2D) of the topA^OX strain (C). The migration direction of more supercoiled and more relaxed topoisomers is indicated. See Supplementary Figure S2 for the original gel images. (D) Cell counts and size distributions were measured daily in the CASY cell counter and plotted as a gray-scale gradient (black: more cells at this volume). (E) Absorption spectra after the harvest on day 5. See Supplementary Figure S1B for spectra at inoculation time. All spectra were divided by the absorption at 750 nm. (F) Glycogen content at harvest time was determined by a colorimetric assay after harvest, and boxplots of 18 technical replicates (three samples, each measured 3× in two assays) are shown. (G) ATP and ATP+ADP contents at harvest time were determined by a luciferase-based assay, and boxplots of six technical replicates (three samples and two measurements) are shown.

Figure 3.

Microscopy and Flow Cytometry Confirm the Volume Growth Phenotype. (A) Fluorescence microscopy images of typical round and dividing cells, after 4 days of growth with or without the inducers (Supplementary Figure S3). The bar indicates 2 μm. Chlorophyll-specific fluorescence is shown in blue and DNA-specific (HOECHST 33342) fluorescence in red. Bright-field and single channel images are provided in Supplementary Figure S4. (B, C) Flow cytometry after 6 days of growth in the presence of the inducers (Supplementary Figure S6). The natural logarithms of forward scatter, side scatter (B) and nucleic acid stain Syto9 (C) were calculated and 2D distributions plotted as contour plots (flow cytometry raw data: Supplementary Figure S7).

Figure 4.

Global transcriptome changes and homeostatic regulation of topoisomerase genes. (A) Electropherograms of the capillary gel electrophoresis analysis of extracted RNA used for RNAseq. The fluorescence signal of each lane was normalized by the total RNA content as reported by the Bioanalyzer 2100 software (Supplementary Figure S6C). Lines are the means of three replicates (Supplementary Figures S8 and S9). Locations of the 16S, the 23S and the large fragment of the 23S rRNA (23S.L) are indicated on the x-axis. Arrows indicate the maxima of the 16S rRNA peaks. (B, C) Expression changes of coding genes in induced strains relative to the control strain (EVC) were derived as the log₂ ratio of RPKM normalized read counts and then compared between the three different strains by 2D histograms (yellow: highest and purple: lowest local density of genes). The Pearson correlations (r) are indicated in the bottom right corner. (B) gyrA^kd (y-axis) versus gyrB^kd (x-axis) strains. (C) gyrA^kd (y-axis) versus topA^OX (x-axis). The induction/repression and the homeostatic responses of gyrA, gyrB and topA are highlighted by arrows from the origin to indicate the direction of change. (D) Expression changes of the targeted topoisomerase genes, the gyrA/parC homolog sll1941, the HU protein (sll1712), the qPCR reference rpoA, and the predicted CRISPRi off-targets (indicated by colored stripes). Error bars are standard errors reported by DESeq2.

Figure 5.

Pulsed induction in continuous culture. (A) Photobioreactor growth of the topA^OX strain (1 l BG11 medium, 0.5% CO₂, illumination ≈90 μmol m⁻² m⁻¹ per OD₇₅₀). Optical density was recorded online (OD_λ) and post-calibrated to offline OD₇₅₀. The arrows indicate inoc.: inoculation; cnt.: onset of continuous culture (rate ϕ = 0.01 h⁻¹); IND.: induction of topA by pulse-addition of rhamnose to 2 mM (0.33 g l⁻¹) at time 0 day; and batch: switch-off of dilution. The dashed black line shows the theoretical wash-out curve of rhamnose ( g l⁻¹). Cell dry weight (CDW, g l⁻¹, red) and glycogen content ( g l⁻¹, blue) were measured at the indicated times (points), and LOESS regressions are shown (solid lines) with 95% confidence intervals (dashed lines). The CASY-based cell volume distributions (Supplementary Figure S11A) are shown as a background in gray-scale for reference. (B) The detrended OD_λ signal (red line, Supplementary Figure S11D) shows a ≈24 h trend throughout batch phase and continuous culture before induction (IND.) A wavelet analysis of the dominant periods in the signal is shown as gray-scale background (right axis).

Figure 6.

Cluster analysis of the transcriptome time series data. (A) Cluster medians of transcript abundances (solid lines), relative to the mean of two pre-induction samples. The transparent ranges indicate the 25%/75% quantiles; points and ticks on upper axis indicate the RNAseq sampling times. Cluster labels (1–6) and sizes (number of genes) are indicated in the legend. (B) Sorted enrichment profile of the six clusters with the CyanoBase ‘categories’ gene annotation. The numbers are the gene counts in each overlap, and gray scale indicates the statistical significance (enrichment) of these counts (black field: p_min ≤ 10⁻¹⁰; white text: p_txt ≤ 10⁻⁵). Only overlaps with p_sort ≤ 0.01 are shown (full contingency table in Supplementary Figure S14A). (C) Enrichment profiles (gray scale as in (B)) with other published gene classifications (see text) and t-value profiles (red-blue scale, Supplementary Figure S13) of clusters in the end-point transcriptome experiments. Blue indicates upregulation (t > 0) and red downregulation (t < 0). (D) Cluster medians as in (A) but zoomed in on the first 5 h after induction. (E) Cluster enrichment profile (gray scale as in (B)) with genes upregulated (up), downregulated (down) or without change (nc), 5–20 min (left) or 2.5–3 days (right) after induction.

Figure 7.

Promoter and Transcription Unit (TU) Structure. (A–D) Cluster nucleotide frequencies around transcription start sites (TSS) (Supplementary Figure S18); only TU on the main genome were considered and the legend in (A) provides the number of TU in each cluster. The G+C content in (A) was calculated in 66 bp windows at each position, all others in 5 bp windows. Point sizes (B–D) scale with −log₂(p) from local motif enrichment (filled points) and deprivation (open circles) tests, and the minimal p-values in each plot are indicated in the legends. The sigma factor binding region (σ, −35 to −10), the location of the open bubble (, −10 to 0) and the transcript (RNA, from 0) are indicated. See Supplementary Figures S19–S23 for the full analysis. (E, F) The Jensen–Shannon (JS) divergence (Supplementary Figures S24–S25) between the position weight matrices of time series clusters 1 and 3 (E) and of immediate response clusters ‘up’ and ‘down’ (F); * indicates p < 0.05 (Supplementary Figure S24) (95). The short horizontal bar in (F) indicates the GC discriminator region −6 to −3. (G) Graded response along TU with ≥4 genes in the batch culture experiments in Figure 4. The y-axis shows the difference of the log₂ fold changes between the first and last transcribed gene of each TU. The left panel shows all strains and the right panel the gyrB^kd strain and TUs by their cluster association. See Supplementary Figure S27 for all strains and all TU with ≥2 genes. (H) An example TU from cluster 1 (red) with a transcript abundance gradient in the gyr^kd strains but not in topA^OX strain. The genes on TU865 are, from 5’ to 3’, rps20, tatD, rpoB and rpoC2. The color scheme (viridis) in the strain tracks reflects the log₂ fold-changes (Figure 4), where yellow indicates higher and blue lower expression than the control strain (EVC). The colors of genes and TU reflect their time series cluster association.