ATP-responsive biomolecular condensates tune bacterial kinase signaling

Abstract

Biomolecular condensates formed via liquid-liquid phase separation enable spatial and temporal organization of enzyme activity. Phase separation in many eukaryotic condensates has been shown to be responsive to intracellular adenosine triphosphate (ATP) levels, although the consequences of these mechanisms for enzymes sequestered within the condensates are unknown. Here, we show that ATP depletion promotes phase separation in bacterial condensates composed of intrinsically disordered proteins. Enhanced phase separation promotes the sequestration and activity of a client kinase enabling robust signaling and maintenance of viability under the stress posed by nutrient scarcity. We propose that a diverse repertoire of condensates can serve as control knobs to tune enzyme sequestration and reactivity in response to the metabolic state of bacterial cells.

Fig. 1.. SpmX-IDR sequesters the DivJ kinase at the stalked pole microdomain.

(A) A schematic of the Caulobacter cell cycle showing the key steps of cell division. Zoomed-in schematic above shows the pole during differentiation from swarmer to stalked cell. PopZ (green), SpmX (blue), and the DivJ kinase (red). Shown below is the domain organization and disorder prediction for SpmX calculated using IUPred [H1 and H2 are transmembrane (TM) helices]. Asymmetric localization of the disordered protein SpmX and the DivJ kinase at the stalk-bearing pole is critical for cell cycle progression. (B) Top: Representative white light images of live Caulobacter cells expressing an endogenous DivJ-HaloTag fusion in strains bearing different SpmX perturbations. Representative single DivJ trajectories are overlaid on the white light images. Approximate cell boundary is indicated using black dashed line. Bottom: A zoomed in view of four polar trajectories (for WT-SpmX and SpmXΔIDR) and nonpolar trajectories (for ΔSpmX). Each trajectory is color-coded and shown as a 2D projection. Polar DivJ exhibits tightly localized trajectories in WT cells while more diffuse trajectories in SpmXΔIDR cells. (C) CDF of displacements in 20-ms intervals was obtained from DivJ trajectories in WT, SpmXΔIDR, and ΔSpmX cells. Inset shows a zoomed-in view of the graph highlighting the overlapping CDFs for SpmXΔIDR and ΔSpmX (N = 974, 443, and 291, polar trajectories for WT, SpmXΔIDR, and ΔSpmX cells, respectively). In the absence of SpmX-IDR, DivJ exhibited faster diffusivity in the polar microdomain.

Fig. 2.. SpmX-IDR stimulates the DivJ kinase.

(A) In vivo phosphorylation measurement of DivJ in SpmX perturbations. DivJ phosphorylation levels in the bar plot are expressed as a ratio of DivJ~P {intensity in the top gel [ATP(γ − ³²P)]} to DivJ concentration (immunoprecipitation using an antibody against DivJ-HaloTag) and are normalized to the ratio observed in WT cells. ΔIDR denotes SpmXΔIDR. Error bars denote SEM from three biological replicates. SpmX-IDR stimulates DivJ activity in vivo. (B) Distribution of lengths for Caulobacter cells grown in M2 minimal media supplemented with 0.2 mM glucose for 4 hours, followed by phase contrast microscopy. The white dot on dark gray bars within violin plots denotes the median value of the distribution. ΔIDR denotes SpmXΔIDR; coil denotes a strain endogenously expressing the CoilZ fusion of SpmXΔIDR and a CoilY fusion of DivJ; ΔIDR + SpmX denotes the SpmXΔIDR strain with a mild expression of WT SpmX; DivJ* denotes DivJ(H338A) catalytic mutant. A total of 2000 to 7000 cells were analyzed in each case from three biological replicates. SpmXΔIDR cells mimic a DivJ mutant phenotype under glucose depletion.

Fig. 3.. SpmX and PopZ form condensates via LLPS.

(A) Pie charts depicting the percentage of hydrophobic (gray), prolines (yellow), acidic (red), and basic (blue) residues in SpmX and PopZ IDRs. (B) Representative fluorescence images of purified DivJ (ΔTM, 1% labeled with Atto488), SpmX (ΔTM, 5% labeled with Cy3), and PopZ (1% labeled with Atto488, heated at 95°C for 8 min) reconstituted in a physiological buffer [50 mM Hepes-KOH (pH 7.4) and 0.1 M KCl] at 5 μM concentration. DivJ exhibited a diffuse signal, while SpmX and PopZ formed round condensates. (C) Fluorescence micrographs showing 5 μM DivJ (ΔTM, 1% labeled with Atto488) incubated with 5 μM SpmX (ΔTM, 5% labeled with Cy3) (top) or 5 μM DivJ (ΔTM, 1% labeled with Cy3) incubated with 5 μM eYFP-tagged SpmX-IDR (bottom) in a physiological buffer for 30 min. DivJ is enriched within SpmX and SpmX-IDR condensates. (D) Internal rearrangement of SpmX-eYFP clusters in live Caulobacter cells (ΔpopZ) assayed by fluorescence recovery after photobleaching (FRAP). Yellow arrow in adjacent image shows the bleached spot in the first frame after photobleaching. Recovery plots are shown for low (0.03 % xylose, dashed line, N = 11 cells) or high induction (0.3% xylose, solid line, N = 12 cells) of SpmX. Gray color represents the SEM. SpmX clusters exhibited concentration-dependent internal dynamics in vivo. (E) Internal rearrangement of eYFP-PopZ condensates in live Caulobacter cells grown in rich or minimal media assayed by FRAP. Yellow arrow in adjacent image shows the bleached spot. White dashed line denotes an approximate cell boundary. Recovery plots are shown for rich media (blue, N = 11 cells) and minimal media (black, N = 11 cells). Gray color represents the SEM. PopZ exhibits nutrient-dependent internal rearrangements. Scale bars [(B) through (E)], 5 μm.

Fig. 4.. SpmX and PopZ form coexisting yet demixed condensates.

(A) Representative superresolution images of 5 μM PopZ (1% Atto488 labeled, heated at 95°C for 8 min) incubated with 500 nM (top) SpmX (ΔTM, 5% labeled with Cy3) or (bottom) SpmXΔIDR (ΔTM, 5% labeled with Cy3) proteins in a physiological buffer for 30 min. Adjacent bar plot indicates the percentage of PopZ condensates that exhibited more than one SpmX cluster, serving as a proxy for demixing (N ~ 400 condensates per sample). Error bars denote the SD from three biological replicates. SpmX condensate demixing within PopZ condensates is driven by the SpmX-IDR in vitro. Scale bars, 5 μm. (B) Superresolution images of cells expressing mCherry-PopZ on a high-copy plasmid and endogenously tagged SpmX-dL5 (top) or SpmXΔIDR-dL5 (bottom). WT SpmX formed multiple clusters associated with the PopZ microdomain, while SpmXΔIDR remained uniformly distributed within PopZ microdomain. Bar plot on the right indicates the percentage of cells that exhibited more than one SpmX cluster within the PopZ microdomain (N ~ 230 cells per sample). Error bars denote the SD calculated from three biological replicates. SpmX clusters are demixed within the PopZ microdomain in vivo. Scale bar, 5 μm. (C) Left: Manually annotated ribosome excluded region representing the PopZ microdomain from cryogenic electron tomography of four cells. White dashed line denotes an approximate cell boundary. Middle: Localizations of SpmX-PAmKate fusion from correlative cryogenic single-molecule imaging pooled from the same four aligned stalked cells (pixel size, ~14 nm). Color bar to the left of SpmX image denotes the number of localizations binned into each pixel [range 1 (dark) to 3 (light) localizations]. Right: In the overlaid image, SpmX localizations exhibit a nonuniform distribution on one side of the PopZ microdomain, suggestive of demixing. Scale bar, 100 nm.

Fig. 5.. Solutes tune LLPS in a protein-specific manner in vitro.

(A) Fluorescence micrographs showing 5 μM PopZ (top, 1% Atto488 labeled) or 5 μM SpmX (bottom, ΔTM, 5% Cy3 labeled) incubated with indicated ATP concentrations for 1 hour in a physiological buffer supplemented with 5 mM MgCl₂. Violin plots (right) show the distribution of the difference in fluorescence signal within the condensate and that outside the condensate, divided by the average fluorescence outside the condensate from a field of view (N ~ 600 to 1100 condensates for each nondissolving condition across three biological replicates). ATP can dissolve PopZ and SpmX condensates in vitro. (B) Fluorescence micrographs showing 5 μM PopZ (top, 1% Atto488 labeled) or 5 μM SpmX (bottom, ΔTM, 5% Cy3 labeled) incubated with indicated LA concentrations for 30 min in a physiological buffer. Violin plots on the right show the distribution of the fluorescence signal measured as in (A) (N ~ 700 to 1100 condensates across three biological replicates). LA can dissolve PopZ and SpmX condensates in vitro. (C) Fluorescence micrographs showing 5 μM PopZ (top, 1% Atto488 labeled) or 5 μM SpmX (bottom, 5% Cy3 labeled) incubated with indicated 1,6-HD concentrations for 30 min in a physiological buffer. Violin plots on the right show the distribution of the fluorescence signal measured as in (A) (N ~ 600 to 1500 condensates across three biological replicates). Differences between distributions in (A) to (C) are statistically significant (P < 0.0005) based on a two-sample t test unless denoted by n.s. (not significant). 1,6-HD promotes PopZ condensation while dissolving SpmX condensates in vitro. a.u., arbitrary units. Scale bars, 5 μm.

Fig. 6.. Solutes tune LLPS in a protein-specific manner in vivo.

(A) Fluorescence and phase contrast micrographs of Caulobacter cells overexpressing eYFP-labeled PopZ, untreated (M2G) or treated with 100 μM CCCP (10 min), 5 μM LA, or 5% (v/v) 1,6-HD (30 min, each). Middle panel shows violin plots with distribution of the ratio of localized to diffuse fluorescence within the cell (N ~ 900 cells per condition). Right panel shows pairwise statistical significance (P values) calculated for each distribution (normal for all cases, lognormal for LA) using a two sample t test. In vivo, PopZ multivalency was promoted under ATP depletion and inhibited by LA treatment. 1,6-HD–treated cells showed minimal difference from untreated cells. (B) Fluorescence and phase contrast micrographs of Caulobacter cells harboring a popZ deletion and overexpressing eYFP-labeled SpmX, untreated (M2G) or treated with 100 μM CCCP (10 min), 5 μM LA, or 5% (v/v) 1,6-HD (30 min, each). Middle panel shows the distribution of the ratio of localized to diffuse fluorescence within the cell (N ~ 700 cells per condition). Right panel shows pairwise statistical significance (P values) calculated for each distribution (normal for all cases, lognormal for LA) using a two-sample t test. SpmX multivalency was promoted under ATP depletion and inhibited by LA and 1,6-HD treatment in vivo. Scale bars, 5 μm.

Fig. 7.. ATP-dependent modulation of DivJ activity via SpmX LLPS.

(A) Representative phase contrast images of WT cells (top), SpmXΔIDR cells, DivJ(H338A) cells, and SpmXΔIDR cells with mild expression of WT SpmX (bottom), grown in minimal medium with varying glucose concentrations [0.2 mM (left) to 1 mM (right)]. Red arrow in the ΔIDR strain imaged under 0.2 mM glucose denotes an aberrantly dividing cell. Violin plots on the right show the cell length distribution as a function of glucose and intracellular ATP concentrations. Note the differences in the ordinate for WT (1 to 5 μm), SpmXΔIDR (1 to 15 μm), and the DivJ mutant (1 to 25 μm). SpmXΔIDR exhibits an ATP-dependent cell length phenotype, phenocopying WT cells under high ATP, and the DivJ kinase mutant under low ATP concentrations. Mild expression of WT SpmX in cells harboring SpmXΔIDR rescues the division phenotype. Scale bar, 5 μm. (B) Top: Schematic showing the DivJ kinase reaction. SpmX-IDR and ATP promote DivJ phosphorylation while DivK dephosphorylates DivJ. Bottom and left: PhosTag SDS–polyacrylamide gel electrophoresis (SDS-PAGE) immunoblots showing DivK phosphorylation in extracts from stalked and predivisional WT (top) or SpmXΔIDR cells (bottom) as a function of ATP levels. DivK was fused to a Flag peptide and detected via an anti-Flag antibody. Bottom and right: Bar plots show the ratio of phosphorylated and unphosphorylated DivK measured from three replicates of the PhosTag assay. Inset shows pairwise statistical significance (P values) calculated using a two-sample t test. DivK (and, by extension, DivJ) phosphorylation is robust to ATP fluctuations in WT cells but sensitive to ATP levels in SpmXΔIDR cells. (C) Model showing the topological relationship between SpmX and PopZ condensates. PopZ and SpmX condensates exhibit nutrient-dependent dynamics. The SpmX condensate renders ATP-dependent feedback on DivJ, resulting in robust modulation of DivJ under glucose depletion.