Clusters of bacterial RNA polymerase are biomolecular condensates that assemble through liquid-liquid phase separation

Abstract

Once described as mere "bags of enzymes," bacterial cells are in fact highly organized, with many macromolecules exhibiting nonuniform localization patterns. Yet the physical and biochemical mechanisms that govern this spatial heterogeneity remain largely unknown. Here, we identify liquid-liquid phase separation (LLPS) as a mechanism for organizing clusters of RNA polymerase (RNAP) in Escherichia coli Using fluorescence imaging, we show that RNAP quickly transitions from a dispersed to clustered localization pattern as cells enter log phase in nutrient-rich media. RNAP clusters are sensitive to hexanediol, a chemical that dissolves liquid-like compartments in eukaryotic cells. In addition, we find that the transcription antitermination factor NusA forms droplets in vitro and in vivo, suggesting that it may nucleate RNAP clusters. Finally, we use single-molecule tracking to characterize the dynamics of cluster components. Our results indicate that RNAP and NusA molecules move inside clusters, with mobilities faster than a DNA locus but slower than bulk diffusion through the nucleoid. We conclude that RNAP clusters are biomolecular condensates that assemble through LLPS. This work provides direct evidence for LLPS in bacteria and demonstrates that this process can serve as a mechanism for intracellular organization in prokaryotes and eukaryotes alike.

Keywords: RNA polymerase; biomolecular condensate; liquid–liquid phase separation; single-molecule tracking; spatial organization.

Fig. 1.

(A) Fluorescence images of fixed cells expressing RpoC-mCherry collected during outgrowth at 37 °C in rich (LB), minimal (M9), or defined (EZ) media. (B) Fluorescence images of fixed cells expressing HupA-mCherry or LacI-mCherry collected during outgrowth at 37 °C in LB. (C) Normalized pixel intensity histograms for conditions highlighted in A and B, i–vi. (D) Clustering of RpoC during outgrowth in LB (n = 6), M9 (n = 3), and EZ (n = 3). Data points correspond to individual cells; black bars represent means, error bars are SEM across n biological replicates. Triangles represent mean ± SEM for HupA and LacI controls (n = 3 each). Dashed lines are the means for HupA and LacI across all time points.

Fig. 2.

(A) Possible mechanisms for RNAP clustering. The DNA-binding hypothesis proposes that RNAP molecules cluster through direct binding to DNA. Alternatively, RNAP molecules may instead cluster through LLPS. (B) Fluorescence images of fixed cells expressing RpoC-mCherry from its native promoter (endogenous), LacI-mCherry, or RpoC-GFP from an IPTG-inducible promoter on a high copy number plasmid. Cells were treated with media (Mock) or 5% Hex for 5 min, and subsequently washed with fresh media to remove Hex (Wash). (C) Quantification of clustering of fluorescent proteins for each treatment condition. Data points correspond to individual cells; black bars represent means, error bars are SEM across n = 3 biological replicates. P values calculated by ANOVA and Tukey–Kramer post hoc test.

Fig. 3.

(A) Charge vs. hydrophobicity plot for the E. coli proteome, with proteins from Table 1 highlighted. The dashed line represents an empirical border between proteins that are natively unfolded (left) and folded (right). (B) Clustering and fluorescence images of WT and deletion mutants expressing RpoC-mCherry. P values were calculated by ANOVA and Tukey–Kramer post hoc test. (C) Domain structure and predicted disorder (by IUPred) of NusA. (D) Phase diagram for purified NusA in the presence of 10% dextran. Open circles indicate conditions in which the protein is dissolved; closed circles indicate conditions in which the protein is condensed. (E) Montage of NusA droplet fusion. A third droplet falls from solution at t = 3 min (marked by *). (F) Fluorescence images of live cells expressing LacI-mNeonGreen or LacI-NusA-mNeonGreen. (G) Histogram of number of foci per cell for each protein construct (n = 3). P value was calculated by unpaired, two-tailed Student’s t test.

Fig. 4.

(A) Schematic of a single-molecule tracking experiment. Cells expressing mMaple3 fusion proteins are continuously activated with 405-nm light and converted molecules are tracked over time. (B) Maximum-intensity projections of cells expressing RpoC-mMaple3, NusA-mMaple3 or LacI-mMaple3. The density and mobility, respectively, of tracks are overlaid. (C) Distribution of D_app for all tracks. (D) Classification of tracks for each protein under three growth conditions. The In tracks are localized to clusters; In/Out tracks exchange between clusters and the bulk nucleoid; Out tracks do not interact with clusters. (E) Distribution of D_app for each class of RpoC tracks. (F) Distribution of D_app for each class of NusA tracks.