SUMMARY
Selective recruitment and concentration of signaling proteins within membrane-less compartments is a ubiquitous mechanism for subcellular organization. However, little is known about the effects of such a dynamic recruitment mechanism on intracellular signaling. Here, we combined transcriptional profiling, reaction-diffusion modeling, and single-molecule tracking to study signal exchange in and out of a microdomain at the cell pole of the asymmetrically dividing bacterium Caulobacter crescentus. Our study revealed that the microdomain is selectively permeable, and that each protein in the signaling pathway that activates the cell fate transcription factor CtrA is sequestered and uniformly concentrated within the microdomain or its proximal membrane. Restricted rates of entry into and escape from the microdomain enhance phospho-signaling, leading to a sublinear gradient of CtrA~P along the long axis of the cell. The spatial patterning of CtrA~P creates a gradient of transcriptional activation that serves to prime asymmetric development of the two daughter cells.
INTRODUCTION
The ability of a cell to recruit and coalesce biochemically-related components at specific subcellular sites is crucial for achieving functional complexity and different cell fates across all kingdoms of life (Misteli, 2007; Rudner and Losick, 2010; Shapiro et al., 2009; Slaughter et al., 2009; Wodarz, 2002). While the mechanisms controlling the formation and composition of subcellular domains are increasingly understood, less is known about the spatial and temporal implications of reactions taking place within them. In Caulobacter crescentus, the disordered polar organizing protein PopZ establishes a space-filling ~100-200 nm microdomain adjacent to the cell pole that is not encapsulated by a protein shell or a membrane (Bowman et al., 2010; Bowman et al., 2008; Ebersbach et al., 2008; Gahlmann et al., 2013). Here we show that this polar microdomain selectively recruits members of the phospho-signaling pathway that culminates in the activation of the master transcription factor CtrA. We further provide a mechanism for how the concentration and motion of these proteins within the microdomain regulates spatially constrained gene expression and the establishment of asymmetry in the predivisional cell.
Caulobacter divides to produce two morphologically distinct daughter cells: a sessile stalked cell that replicates its chromosome following division, and a motile swarmer cell in which chromosome replication is delayed until it differentiates into a stalked cell (Figure 1A, right). This developmental asymmetry is governed by differentially–localized groups of signaling proteins at the two cell poles (Lasker et al., 2016). The localization of these signaling proteins is dependent on the polar organizing protein PopZ that self-assembles into a branched network of filaments in vitro and creates a space-filling microdomain at the cell poles in vivo (Bowman et al., 2010; Bowman et al., 2008; Gahlmann et al., 2013). PopZ directly binds at least nine proteins that temporally and spatially regulate cell cycle progression, and deletion of the popZ gene leads to delocalization of these proteins and to severe cell cycle defects (Bowman et al., 2010; Bowman et al., 2008; Ebersbach et al., 2008; Holmes et al., 2016; Mignolet et al., 2016; Ptacin et al., 2014). Among these binding partners are two signaling proteins that together play a key role in controlling the levels and activity of the master transcription factor CtrA: the membrane-bound hybrid histidine kinase CckA, which acts as the phosphate source for CtrA, and the small cytoplasmic protein ChpT, which shuttles phosphate from CckA to the master transcription factor CtrA (Figure 1A). In its active phosphorylated form (CtrA~P), CtrA controls the transcription of over 100 genes, including those that are necessary for the formation of the nascent swarmer cell, including the flagellar and chemotaxis transcriptional hierarchy (Figure 1A) (Ardissone and Viollier, 2015; Laub et al., 2002; Laub et al., 2000). CtrA~P also serves to inhibit the initiation of DNA replication in the swarmer cell by binding to the chromosome origin (Quon et al., 1998). CtrA levels and activity vary as a function of the cell cycle. In swarmer cells, high CtrA~P levels promote the swarmer fate. During the swarmer to stalk cell transition CtrA~P is cleared from the cell to allow initiation of DNA replication. In predivisional cells, CtrA proteolysis ceases and its synthesis and activation resumes. Upon cell compartmentalization but prior to division, CtrA~P is proteolyzed in the stalked compartment, while active CtrA~P remains in the swarmer compartment.
CckA’s auto-kinase activity is density-dependent and occurs only when CckA is concentrated at the new pole microdomain of the predivisional cell, where it drives phosphorylation of CtrA (Iniesta et al., 2010a; Jacobs et al., 2003; Jonas et al., 2011; Mann et al., 2016; Tsokos et al., 2011). In contrast, CckA acts as a phosphatase everywhere else in the cell with density-independent activity (Mann et al., 2016). The antagonism between CckA acting as a kinase and as a phosphatase resembles the process by which concentration gradients are formed in much larger eukaryotic cells and tissues (Wartlick et al., 2009).
For the cytosolic ChpT phospho-transfer protein to interact with CckA at the membrane of the new pole, it must pass through the space-filling PopZ microdomain. ChpT~P must then phosphorylate CtrA to complete the signaling pathway. Thus, critical questions are if (and how) the PopZ microdomain influences access of ChpT to the CckA kinase, where ChpT~P encounters CtrA, and ultimately, what spatial distribution of CtrA~P is established as it emanates from the new pole microdomain. With spatially resolved transcriptional measurements as a proxy for CtrA~P activity, we demonstrate that predivisional cells exhibit a sublinear (approximately exponential) gradient of activated CtrA~P emanating from the new cell pole. By imaging the single-molecule trajectories of CckA, ChpT and CtrA relative to the super-resolved PopZ microdomain, we show that in addition to acting as a localization factor, the microdomain sequesters each member of the CtrA phospho-transfer pathway and restricts the rate at which molecules within the microdomain exchange with the rest of the cell, while restricting polar access for other molecules. Integrating this knowledge with other biochemical data in a comprehensive reaction-diffusion model, we show that both high concentration and slow turnover at the poles are necessary to generate a gradient of CtrA~P activity, in quantitative agreement with our transcriptional measurements. These results demonstrate that a bacterial microdomain adjacent to the cell pole selectively sequesters a phospho-signaling pathway to establish a spatial trajectory of information transfer that controls chromosome readout.
RESULTS
The CtrA activation pathway is sequestered within the PopZ microdomain
To understand the influence of the PopZ microdomain on the CtrA activation pathway, we began by determining to what extent the pathway’s proteins were colocalized with PopZ at the old and new poles of predivisional cells. We replaced native popZ with an allele for the (photo-activatable) fluorescent protein fusion PAmCherry-popZ (Figure S1A, Tables S1-S3). We observed that PopZ forms foci at both poles in predivisional cells as previously shown (Bowman et al., 2008; Ebersbach et al., 2008), with 55% of its signal localized to the old pole and 31% to the new pole on average (Figures 1B-D, S1B). Using strains expressing enhanced yellow fluorescent protein (eYFP) fused to either CckA, ChpT, or CtrA, we measured the co-localization of each pathway component with respect to PAmCherry-PopZ (Figure S1) (at diffraction-limited 200 nm resolution, also see Figures 4 and 5). Consistent with previous studies (Ebersbach et al., 2008), we found that CckA-eYFP co-localized with PopZ at both the new pole and the old pole, with 60% of total CckA fluorescence intensity localized to the new pole and 27% localized to the old pole (Figure 1B). While CckA clearly associates with PopZ (Holmes et al., 2016), CckA exhibits a greater concentration at the new pole due to additional pole-specific localization factors (Iniesta et al., 2010a). The cytosolic phosphotransferase ChpT, like its interacting membrane histidine kinase CckA, was shown to bind PopZ in vivo using the heterologous Escherichia coli system (Holmes et al., 2016). We thus speculated that PopZ-binding interactions would also cause ChpT to be enriched within the PopZ microdomain in Caulobacter. Indeed, ChpT formed foci at both poles of the cells, with 22% of ChpT signal localized to the new pole and 34% of the signal localized to the old pole (Figures 1C, S1B). While ChpT weakly binds CckA in vitro (Blair et al., 2013), ChpT did not show greater localization at the new pole, where most of CckA molecules reside. These results suggest that PopZ is a critical polar localization factor for ChpT in Caulobacter.
Unlike ChpT and CckA, CtrA does not colocalize with PopZ when heterologously expressed in E. coli, suggesting that binding interactions between CtrA and PopZ are weak or nonexistent (Holmes et al., 2016). We asked whether CtrA forms a focus within the PopZ microdomain in Caulobacter without direct binding to PopZ, as does the cytosolic protein DivK (Holmes et al., 2016; Lam et al., 2003). Previous imaging of an N-terminal YFP fusion to CtrA in predivisional cells showed a large diffuse population of CtrA with transient accumulation at the old cell pole prior to proteolysis, independent of CtrA’s phosphorylation state or its degradation motif (Ryan et al., 2004), while no CtrA accumulation was detected at the new pole. Because the N-terminal domain of CtrA is important for function, we repeated this experiment using a sandwich fusion that integrated eYFP between CtrA’s DNA binding domain and its 14-residue C-terminal degradation tag (Figure S1A) (CtrA-eYFP-14). We observed a localization profile of CtrA-eYFP-14 in synchronized predivisional cells (Figure S1C) in which 17% of the fluorescent signal was at the new pole and 28% of the signal was located at the old pole. Altogether, our imaging results show that all three members of the phospho-signaling pathway are greatly enriched within the PopZ microdomains of the Caulobacter cell.
CtrA exhibits a gradient of activity in predivisional cells
We designed a transcription assay as a proxy for measuring the CtrA~P spatial profile and used it to determine whether polar accumulation of the proteins in the CtrA activation pathway regulates the distribution of CtrA~P across the cell. We introduced a CtrA~P activated promoter (P350, from the CCNA_00350 operon) driving an eyfp gene at four different sites along the right arm of Caulobacter’s single circular chromosome (loci L1-L4) (Figure 2A, Table S4). Locus L1 is adjacent to the replication origin, which is positioned near the cell pole, while locus L4 is positioned near the chromosome terminus (Figure 2A). Because the position of any given gene on the Caulobacter chromosome reflects its position within the cell (Viollier et al., 2004), we can approximate the spatial position of loci L1-L4 within the cell and their time of replication (SI, Table S4).
We used reverse transcription quantitative PCR (RT-qPCR) (Heid et al., 1996) to measure the reporter eyfp mRNA transcribed from each locus L1-L4 in synchronized populations beginning from swarmer cells (Figure 2A). This allowed us to quantitatively and directly compare the activity of CtrA~P regulated promoters from each spatial position over the course of the cell cycle. For clarity, we report the transcriptional output relative to the 5-minutes time point of L1 (1 AU). Over the course of the experiment, the single chromosome replicates once. Thus, the replication of the single starting chromosome completes once, thus the output of each locus should double upon replication of that locus on the chromosome. While transcription from all four integration loci remained low during the first 60 minutes of the cell cycle, the loci began to diverge at the onset of CtrA activation in pre-divisional cells. Critically, at the 90-minutes time point, transcriptional activity at site L1 peaked to 5.5±0.11 AU, while transcription from site 2 peaked to 2.5±0.25 AU, even though both sites had been duplicated far earlier (Figure 2A, Table S4). Transcription from sites L2-L4 peaked at the 120-minute time point, upon completion of DNA replication, with 1.6±0.67 AU and 1.8±0.42 AU for sites L3 and L4 (Figure 2A). As a control for CtrA-independent regulation, the transcription of eyfp under the xylose-inducible PxylX promoter showed a ≤ 2-fold increase resulting from chromosome duplication (Figure S2A). In addition, the transcription of eyfp under another CtrA regulated promoter PpilA showed chromosome position-specific effects (Figure S2B). Finally, comparing the cumulative occurrences of CtrA ChIP-seq peaks and CtrA binding motifs we observed that despite uniform distribution of CtrA binding motifs across the chromosome, CtrA distribution on the chromosome is not uniform and exhibits a larger concentration of CtrA molecules in the origin-proximal region (Figure 2SC). These results indicate that the activated CtrA~P transcription factor exhibits spatial control of promoter activity, with highest transcription from loci positioned near the newly replicated origin of replication at the new cell pole.
To test our hypothesis that the gradient of CtrA activity results from the spatial regulation of CckA and ChpT activity, we deleted the native ctrA gene in the presence of a plasmid-borne, xylose-inducible phosphomimetic mutant, ctrA(D51E) (Figure 2B). This mutant cannot accept a phosphate, but the glutamate replacing the aspartate results in constitutive CtrA activity, (Domian et al., 1997), decoupling it from the CckA-ChpT phospho-transfer pathway. When ctrA(D51E) was expressed as the sole copy of CtrA, position-specific effects on transcription from the CtrA-activated P350 promoter were lost, demonstrating that the gradient in CtrA activity depended on the signaling cascade emanating from the cell pole. This effect was not due to changing protein levels, as xylose-induced expression of a plasmid-borne wildtype ctrA still resulted in a gradient of P350 activity (Figure 2B). Collectively, these experiments argue that localized activity of the CckA phospho-signaling pathway leads to a gradient in CtrA~P and spatial control of CtrA-controlled promoters.
Accumulation of phospho-relay components in the PopZ microdomain governs CtrA~P levels and spatial distribution
Using previously published biochemical data, we quantitatively interpreted the results of our RT-qPCR experiments to define potential CtrA~P distributions within the cell. Our observation that CtrA~P-regulated promoter activity depended on the promoter’s position within the cell indicates a gradient of CtrA~P emanating from the new cell pole. However, the distribution of CtrA~P may not be identical to the gradient in transcriptional activity: CtrA~P binds cooperatively to the PccrM and PfliQ promoters (Reisenauer et al., 1999; Siam and Marczynski, 2000), and because CtrA~P binds P350 as a dimer (Zhou et al., 2015), it is likely to exhibit cooperative activity when promoting the transcription of 350 mRNA as well. Accordingly, we modeled the transcriptional response from P350 as a Hill function, with fitting parameters based on previously measured CtrA copy number and cell cycle dependent transcription of 350 (Schrader et al., 2016) (Equation S7, Figure S3A). We examined the effect of several different shapes of CtrA~P gradients on the transcriptional output from the four (L1-L4) integration sites (Figures 3A and 3B). An exponential decrease in CtrA~P levels best reflected the 2-fold decrease in mRNA levels from L2 and a 5 fold decrease in mRNA levels from L3 and L4, with respect to mRNA levels from locus 1 (Figures 3A and 3B III). Thus, we concluded that the observed drop in transcriptional activity as a function of distance of the integration site from the origin is likely the result of an exponential or other sublinear decay in CtrA~P levels from the new pole to the old pole.
The canonical model for forming a gradient of the phosphorylated state of a signal requires that phosphorylation and dephosphorylation events occur at separate cellular locations and that the kinetics of phosphatase activity is fast relative to the diffusion of the signal (Barkai and Shilo, 2009; Brown and Kholodenko, 1999; Kiekebusch and Thanbichler, 2014; Wartlick et al., 2009). As all members of the CtrA activation pathway accumulate inside the PopZ microdomain (Figure 1B-D) we turned to mathematical modeling to explore how this accumulation affects gradient formation properties. We defined a reaction-diffusion model in which a set of partial differential equations track the concentrations of CckA, ChpT, CtrA as a function of diffusion and biochemical interactions inside and outside of the PopZ microdomains (Figures 3C and S3, Table S6). To identify the biochemical parameters that are most important for setting the gradient of CtrA activity, we performed a global sensitivity analysis of our model (SI). We explored the sensitivity of the amplitude and shape of the CtrA~P gradient to changes in the binding affinity to the microdomain, rates of phospho-transfer, and diffusion coefficients inside and outside the microdomain. To reflect the possibility that the PopZ microdomain environment is different at the two cell poles, we also allowed for differential diffusivity and binding rates inside the two microdomains. By comparing the steady-state CtrA~P profiles across these different parameter values, we calculated ‘amplitude sensitivity’ (total CtrA~P copies in the cell), and ‘distribution sensitivity’ (the ratio between sum of CtrA~P molecules from the two halves of the cell) for each parameter we had varied.
The amplitude of CtrA~P was highly sensitive to alterations in the KD of binding between CckA and the new pole microdomain as well as the phospho-transfer rate between CckA to ChpT (Figure 3C, upper heatmap, Figure 3D). Changing the KD between CckA and the new pole microdomain alters the concentration of CckA, and because CckA’s kinase activity increases nonlinearly with increasing concentration (Mann et al., 2016), this greatly increases the rate at which phosphate is introduced into the pathway (Figure 3D). Meanwhile, faster phospho-transfer rates between CckA and ChpT and between ChpT and CtrA increase the fraction of CtrA~P molecules that is generated at the new pole microdomain, near the CckA source of phosphate.
The distribution of CtrA~P was highly sensitive to changes in ChpT and CtrA binding to the microdomain at the new pole, as well as to CtrA diffusion in the body of the cell (Figure 3C lower heatmap, Figure 3E). Weakening the KD of ChpT or CtrA for the new pole microdomain led to a shallower gradient across the cell, with a larger contribution from ChpT. Strong interactions between both proteins and the new pole were critical to achieve a CtrA~P distribution that, insilico, recapitulates the transcriptional measurements (Figure 3D). These results suggest that the spatial position of the phospho-transfer between ChpT and CtrA can modulate the shape of the gradient. Additionally, it was critical that CtrA diffusivity in the body not be substantially higher than the effective kinetics of phospho-transfer, as this would disrupt any gradient. Collectively, our model predicts a critical role of the PopZ microdomain in modulating the behavior of the components of phospho-transfer cascade.
CckA molecules are concentrated and their diffusion is slowed within the polar PopZ microdomain
The results of our sensitivity analysis (Figure 3) strongly suggested that the formation of an exponential gradient of CtrA~P depends on the dense concentration and slow diffusion of phospho-signal pathway proteins within the PopZ microdomain. To directly visualize these properties, we turned to single-molecule tracking of the CtrA pathway proteins combined with super-resolution microscopy of the static PopZ microdomain, beginning with CckA. Because CckA is associated with the curved inner membrane, single-molecule tracks generated from two-dimensional (2D) measurements would appear distorted at the “sides” of the cell, leading to erroneously low estimates of diffusivity in these regions (Figure S4A). Therefore, we used an engineered point spread function (the DH-PSF) (Gahlmann et al., 2013) to localize and follow the motion of CckA-eYFP molecules in three dimensions (3D), avoiding the 30% errors that would be present for 2D trajectories of membrane proteins (Figures S4B and S4C).
By correlating our tracking measurements with 3D super-resolution imaging of a xylose-induced copy of PAmCherry-PopZ in a merodiploid background, we precisely defined what parts of the CckA-eYFP trajectories explored the membrane area of the PopZ microdomains at the new and old poles (Figure 4A). Using mean-squared-displacement (MSD) analysis, we found that CckA diffused most rapidly in the cell body, 2-fold slower in the old pole, and 4-fold slower in the new pole (D = 0.0082 ±0.0020, 0.0040 ± 0.0014, and 0.0022 ± 0.0013 μm2/s, respectively) (Figure 4B). One certain cause for the reduction in diffusivity at the poles is polar binding interactions, both to PopZ and to CckA-specific localization factors at the new pole (Holmes et al., 2016; Iniesta et al., 2010a). Another likely cause is the formation of higher-order CckA assemblies (Mann et al., 2016), which would presumably be enhanced at the high CckA concentrations within the new pole (and to a lesser extent, within the old pole).
To define the average nanoscale distribution of CckA at the new and old poles, we combined CckA and PopZ high-resolution localization data from many individual microdomains (SI). We found that CckA molecules were distributed roughly evenly throughout the 3D membrane region adjacent to PopZ, and that the CckA concentration dropped off sharply away from the microdomain (Figure 4C, Figure S4D). While the uniform density in the averaged CckA distribution does not preclude the possibility of individual CckA clusters within microdomains (as has been suggested in PopZ overexpression experiments (Ebersbach et al., 2008)) it clearly indicates that CckA does not have a preference for particular positions within the PopZ microdomain. CckA formed a hemispherical cup surrounding PopZ in old poles (Figure 4C), while in new poles, the PopZ microdomain took up a smaller volume, and CckA occupied a proportionally smaller area on the new pole membrane (Figure S4D). Notably, the area taken up by CckA at the new pole was lower despite having a higher total number of CckA molecules (Figure 1C). Combining our measurements of polar radii (r = 108 nm for old poles, and r = 82 nm for new poles) with estimates of the total number of CckA molecules at both poles, we estimated a CckA concentration of ~5,000-10,000 and 1,250 molecules/μm2 for the new and old poles, respectively, more than 100x greater than the concentration in the cell body (9 molecules/μm2) (SI). The estimated concentration of CckA at the new pole agrees with the previously determined CckA concentration on liposomes in vitro that leads to maximum autokinase activity, while the estimated CckA concentration at the old pole was shown in vitro to lead to little or no autokinase activity (Mann et al., 2016).
To determine whether CckA was free to enter and exit the pole over a longer timescale than seconds, we used a diffraction-limited confocal microscope to photobleach CckA at the old pole. We then measured the timescale of fluorescence recovery and loss at the old and new poles (Figure 4D and Figure S5). While targeted photobleaching depleted ~30% of the total cell fluorescence, we found that the ratio of CckA within the new and old poles was restored close to its pre-bleach value after 10 minutes. This was consistent with reaction-diffusion simulations using the experimentally measured diffusivities and assuming transient binding of CckA to species within the microdomain (Figure S5). Thus, while CckA recovery was slower than would be expected for free diffusion, CckA was still mobile and not irreversibly bound within the polar microdomain. Collectively, the CckA tracking and photobleaching experiments indicate that CckA is concentrated throughout the polar membrane cap of the PopZ microdomain by crowding and reversible binding to other proteins within the microdomain.
ChpT and CtrA molecules are transiently sequestered within the PopZ microdomain volume
We used our correlative single-molecule tracking and super-resolution imaging approach to study the cytoplasmic members of the signaling pathway, ChpT and CtrA, to ascertain their diffusion coefficients, distributions, and residence times within the polar PopZ microdomains. ChpT-eYFP molecules outside the poles diffused rapidly throughout the cell, but were captured within the PopZ volume upon reaching the poles (Figure 5A). While most captured ChpT molecules appeared to diffuse throughout the microdomain volume, a fraction of ChpT-eYFP trajectories exhibited motion only in the plane of the polar membrane (Figure 5B), suggestive of binding to CckA. Similar to ChpT, we observed that CtrA-eYFP-14 entered and was slowed within the PopZ microdomain (Figure 5C). CtrA and ChpT molecules exited the microdomain after a short time period (Figure 5D; ChpT not shown) and immediately regained their previous diffusivity within the cell body.
We combined high-resolution single-molecule data from many predivisional cells to generate average localization profiles of CtrA, ChpT, and PopZ (Figure 5E and Figure S6). As in our diffraction-limited images (Figure 1D-E), while ChpT and CtrA were present throughout the cell, they were much more concentrated at the cell poles, with ChpT being somewhat more concentrated than CtrA. In addition, the higher-resolution single-molecule distribution clearly indicated that ChpT and CtrA entered the microdomain and were distributed throughout the PopZ volume. As CtrA does not have affinity for PopZ but does bind ChpT (Blair et al., 2013; Holmes et al., 2016), this suggests that ChpT molecules embedded in the microdomain act as binding partners to indirectly recruit CtrA. We also observed a slight enrichment of CtrA at the cytoplasmic face of the microdomain at both the new and old poles, ~50-100 nm from the PopZ centroid (shoulders on green curves in Figure 5E). This may result from CtrA binding to the chromosome origin sequence ori, which CtrA binds with nM affinity (Siam and Marczynski, 2000), and is likely within 150 nm of PopZ based on previous super-resolution and chromosome imaging experiments. (Hong et al., 2013; Ptacin et al., 2014).
We used MSD analysis to quantify the diffusivity of ChpT and CtrA, projecting motion onto the 1D cell axis to avoid confinement effects arising from the narrow cell width (Figure 5F). Both ChpT and CtrA exhibited anomalous subdiffusive motion in the cell body, apparent as a slope less than 1 in the log-log MSD plot. Given the short length of our eYFP trajectories, fits to these curves only allowed us to bound the subdiffusive expontent α, indicating α < 0.7 and a short-timescale apparent diffusivity Dapp = 1.8 μm2/s (SI, Figure S6). We also observed anomalous subdiffusion in trajectories of free eYFP, which we do not expect to directly bind any targets in Caulobacter (Figure S6). Thus, subdiffusion in the cell body likely results from general properties such as crowding in the cytoplasm and obstruction by the nucleoid.
Within the poles, both ChpT and CtrA exhibited MSD values an order of magnitude lower than in the cell body (Figure 5F). ChpT molecules explored the pole on a timescale comparable to our 20 ms integration time (Figure 5A), causing the MSD to immediately reach an asymptotic limit of ~10−2.5 μm2 (3x higher than our 24 nm localization precision) independent of lag, implying D ≥ 0.1 μm2/s. We analytically confirmed that this limit is to be expected given the geometry of the PopZ microdomain, and calculated that free motion within spheroids of 150-200 nm diameter would give rise to asymptotic MSDs of 10−2.6 to 10−2.4 (SI). In contrast, polar CtrA exhibited a steady increase of its MSD between the limits of error and free diffusion, as expected from our observation of tracks that slowly explored the PopZ microdomain (Figure 5D), and fits to these data gave D ≈ 0.01 μm2/s (Figure S6). Overall, these results show that both ChpT and CtrA were relatively free to explore the polar space in which they were concentrated.
We determined the rate of ChpT and CtrA turnover at the poles by calculating the dwell times of eYFP-labeled molecules within the microdomain before returning to the cell body (Figure 5G). The apparent exit rates of both ChpT and CtrA were increased due to competition between true exit events and the rapid photobleaching of eYFP labels. We compensated for this effect by approximating photobleaching and exit as competing exponential processes, scaling the rates by the proportion of tracks that exited before photobleaching (SI). This approximation gave similar dwell times of ChpT and CtrA within the poles: 132±39 ms and 132±28 ms (95% CI), respectively. This was three-fold slower than the rate of exit in simulations of free diffusion, reflecting binding events and slowed diffusion within the crowded PopZ matrix.
The PopZ microdomain creates a barrier to entry for non-client proteins
Ribosomes and chromosomal DNA do not enter the PopZ microdomain, an effect ascribed to such large molecules being unable to pass through the fine pores of a filamentous PopZ matrix (Bowman et al., 2010; Ebersbach et al., 2008). We wondered whether the slowed diffusion and escape rates of ChpT and CtrA within the PopZ microdomain was due in part to percolation through such a matrix. To isolate this effect, we performed single-molecule tracking of proteins heterologous to Caulobacter, for which binding interactions to polar localization factors should be minimal. We selected two proteins: eYFP by itself, and eYFP fused to a 100 amino acid fragment of A. thaliana PIF6 which we term fPIF (Levskaya et al., 2009). As eYFP was 0.5 times and fPIF-eYFP was 0.75 times the mass of ChpT-eYFP and CtrA-eYFP-14 (Table S1), as reflected by their faster diffusion in the cell body (Figure S6), we predicted that both proteins would penetrate the PopZ matrix. Yet surprisingly, both eYFP and fPIF-eYFP only explored the 3D volume of the cell up to the edge of PopZ, and did not enter the microdomain (Figure 5H, Figure S6).
We quantified the polar recruitment and exclusion of ChpT, CtrA, fPIF, and eYFP by counting single-molecule localizations that appeared within the super-resolved reconstruction of PopZ, and comparing these patterns to the expected distribution in the absence of microdomain interactions from simulations of free diffusion throughout the cell volume (Figure 5I, Figure S6). Relative to the “free diffusion” case, ChpT and CtrA were concentrated ~3x at the old pole and ~2x at the new pole, reflecting their direct (PopZ-binding) and indirect (ChpT-binding) recruitment. By contrast, fPIF and eYFP were not only not enriched, but were actively excluded from the poles; while a small fraction (4.0 +/- 0.6%) of fPIF molecules were scored as polar, these molecules were generally located at the cytoplasmic interface of the microdomain, not the interior (Figure 5H). As fPIF and eYFP were excluded, despite being smaller than ChpT and CtrA, pore size cannot explain this effect. In contrast to our experiments with wildtype cells, fluorescent proteins are able to enter the PopZ microdomain when PopZ is heterologously expressed in E. coli (i.e., in the absence of known clients), or when PopZ is enlarged 10-20 fold to 2-4 μm in Caulobacter, greatly increasing the available volume (Ebersbach et al., 2008; Laloux and Jacobs-Wagner, 2013). This suggests that in addition to the presence or absence of specific binding interactions to PopZ or other polar proteins, general properties such as the volume and client occupancy of the PopZ microdomain may also control the microdomain permeability in a nonspecific way.
CtrA activation occurs within and modulated by the new pole microdomain
Using our quantitative microscopy measurements of the distributions and dynamics of CckA, ChpT, and CtrA, we comprehensively simulated signaling in the predivisional cell within our reaction-diffusion framework. The simulation included: (i) a calculated effective KD between the microdomain and CckA, ChpT, and CtrA to match our localization profiles, (ii) the measured diffusion coefficients, and (iii) an adjusted unbinding rate between CckA and the microdomain to match the measured exit rates (Figure S4). In our model, steady state distributions of CckA, ChpT, and CtrA recapitulated localization profiles observed both by diffraction-limited and singlemolecule imaging (Figures 1B-D, 5E, S7A). Similarly, steady state profiles of all three proteins revealed the concentration of the phosphorylated state at the new cell pole microdomain (93% of CckA~P, 68% of ChpT~P, and 10% of CtrA~P). ChpT~P concentration sharply dropped from the new pole microdomain and continued to decline with a shallower sublinear gradient (Figure 6A). In contrast, the CtrA~P concentration declined smoothly away from the new pole microdomain with a sublinear distribution (Figure 6A). Following our calculation of transcriptional response from the P350 promoter integrated at the four chromosomal loci (Figure 3A,B), we calculated transcriptional output in response to the modeled CtrA~P distribution. The calculated eyfp mRNA values recapitulated the measured mRNA levels in our model (Figures 3A, S7A). Collectively, based on the consistency between multiple sets of independent measurements (biochemical rates from the literature, diffusion coefficients from single-molecule tracking, and transcription activity based on qRT-PCR), we concluded that CtrA~P exhibits a roughly exponential gradient in predivisional cells.
Sequestration of the phosphorylated form of all three proteins to the new pole microdomain suggests that forward phospho-transfer events happen within the microdomain. We used our reaction-diffusion framework to count transfer events as a function of time and cellular position. We observed that more than 95% of forward transfer events (from CckA~P to ChpT and from ChpT~P to CtrA) occur within the new pole microdomain, while back transfer events (from CtrA~P to ChpT and from ChpT~P to CckA) occur everywhere away from the new pole (Figure 6B, Figure S7). Our sensitivity analysis predicted that the binding affinity between ChpT and the new pole microdomain strongly affects the CtrA~P gradient (Figure 3C). Counting phospho-transfer events in an in-silico mutant with a weakened binding affinity between ChpT and the microdomain (KD between ChpT and the microdomain reduced from 10 nM to 10 μM), we observed a drastic change in the spatial distribution of forward phospho-transfer events (Figure 6B). 46% of ChpT~P molecules left the pole before passing the phosphate to CtrA, which led to 60% of ChpT~P to CtrA transfer events occurring outside of the new pole hub. Collectively, phospho-transfer count analysis provides evidence that CtrA is phosphorylated within the new pole microdomain and that binding to the microdomain facilitates this process.
Our single-molecule measurements showing protein exclusion from the pole demonstrate that selective entry into the limited cytosolic polar volume may be controlled not only by the strength of specific binding interactions, but by mechanisms that control global permeability for all potential clients. One potential mechanism is client density: if many clients bind within the microdomain, they may occupy free volume or binding sites and prevent competing proteins from entering. To model such (occupancy-based) universal regulation of microdomain permeability, we altered the density of cytosolic client proteins inside the PopZ microdomain by introducing a separate cytosolic protein with high binding affinity for the microdomain while keeping the binding affinities of CckA, ChpT and CtrA to the microdomain constant (SI). This perturbation led to competition for limited binding sites within the PopZ microdomain, mimicking the effect of changing the global permeability of the microdomain. The simulated reduction in the microdomain permeability led to a 50% decrease in the relative proportion of ChpT and CtrA inside the poles, driving a dramatic reduction in the effectiveness of CtrA~P signaling (quantified by CtrA~P amplitude and CtrA~P distribution, Figure 6C). These simulations illustrate the critical role microdomain permeability plays in modulating CtrA~P amplitude and shape by regulation of the concentration of members of the activation pathway.
CtrA molecules are cleared from the cell during differentiation and resynthesized in predivisional cells over the course of 60 minutes (Domian et al., 1997; Quon et al., 1996). We used our model to determine how the CtrA~P gradient evolves from newly synthesized CtrA molecules in predivisional cells. We found that for a given number of total CtrA molecules, CtrA is rapidly phosphorylated and that the CtrA~P amplitude reaches steady state within five minutes of simulation time (Figure 6D). The nonlinear gradient in CtrA~P distribution is established immediately at the beginning of the simulation and is maintained until reaching its steady state. Thus, the quasi-steady-state CtrA~P gradient is established early in predivisional cells as CtrA molecules are synthesized, robustly priming the nascent daughter cells for asymmetry.
DISCUSSION
Membrane-enclosed compartments, such as the nucleus (Meldi and Brickner, 2011), mitochondria (Nunnari and Suomalainen, 2012), and chloroplasts (Eberhard et al., 2008) in eukaryotes, and thylakoids (Nevo et al., 2007), magnetosomes (Komeili et al., 2006), and the periplasmic space (Bos et al., 2007) in bacteria, are an effective means for concentrating functional complexes to isolated environments within the cell. Alternatively, membrane-less compartments have emerged as a widespread organizational unit. These include membrane-less organelles in eukaryotes (Alberti, 2017; Banani et al., 2017; Shin and Brangwynne, 2017) such as Cajal bodies (Mao et al., 2011) and germ granules (Voronina et al., 2011), and protein-encapsulated microcompartments in prokaryotes and archaea (Kerfeld and Erbilgin, 2015; Kerfeld et al., 2010) such as carboxysomes (Kerfeld and Melnicki, 2016) and encapsulins (Giessen, 2016). Bacterial cells also employ low complexity multivalent proteins as membrane-less localization platforms to direct cellular development (Ben-Yehuda et al., 2003; Bowman et al., 2008; Ebersbach et al., 2008; Lin et al., 2017; Yoshiharu et al., 2012).
Eukaryotic membraneless compartments are held together by weak interactions between their components that form agglomerated condensates (Molliex et al., 2015; Nott et al., 2015). Unlike structurally defined macromolecular complexes, such as the ribosome or the 26S proteasome, the size and internal organization of such condensates are both heterogeneous and dynamic (Shin and Brangwynne, 2017). Here, we showed that the Caulobacter PopZ microdomain, which lacks both a membrane enclosure and a protein shell, differs from bacterial microcompartments while sharing key qualities with eukaryotic condensed compartments. Most significantly for function, the PopZ microdomain size is dynamic, client entry into the microdomain is selective, and clients dynamically exchange with the cytosol. We further showed that these properties facilitate accelerated phospho-transfer reactions and modulation of the amplitude and distribution of CtrA~P in the predivisional cell.
Organization and selective permeability of the space-filling PopZ microdomain
PopZ self-assembles to organize a space-filling microdomain (Bowman et al., 2013; Gahlmann et al., 2013; Ptacin et al., 2014) that coordinates a suite of cell polarity functions including the phospho-transfer cascade controlling CtrA activity (Berge and Viollier, 2017; Ptacin and Shapiro, 2013). Whereas PopZ itself is cytoplasmic, both transmembrane and cytosolic client proteins are (to our imaging resolution) uniformly distributed adjacent to and throughout the space-filling PopZ microdomain for transmembrane and cytosolic proteins, respectively (Figure 4, Figure 5) (Ptacin et al., 2014). PopZ is recruited to the membrane by direct binding to several membrane-bound factors, and mislocalization of these factors can seed ectopic PopZ microdomains (Berge et al., 2016; Holmes et al., 2016; Perez et al., 2017). Thus, the shape of the PopZ microdomain appears to be defined by a condensation and wetting process that balances cohesive forces (self-interaction) and adhesive forces (membrane-protein interactions), similar to that observed for tau protein droplets on microtubules (Hernandez-Vega et al., 2017). This balance allows PopZ to lie across a substantial membrane surface area while still creating a large but porous space-filling volume, bridging the membrane and the cytoplasm and jointly concentrating both membrane and cytoplasmic proteins.
The PopZ microdomain selectively captures and concentrates target proteins at the cell pole, including the three CtrA pathway proteins. Proteins are recruited to the PopZ microdomain either by direct binding to PopZ, as shown for at least nine client proteins (Berge et al., 2016; Holmes et al., 2016), or indirectly by binding to one or more proteins which bind PopZ, as shown for the membrane protein DivJ (Perez et al., 2017) and the cytosolic protein CtrA (Figures 1D, 5, and 7). Whereas large molecules and complexes such as DNA and ribosomes (Bowman et al., 2008; Ebersbach et al., 2008) are excluded from the microdomain, we have shown that small proteins lacking a microdomain binding partner, such as free eYFP and fPIF-eYFP, cannot penetrate the PopZ microdomain despite being similar in size and charge to proteins that do enter (Figure 5H, Figure 7, Table S1). Notably, when the PopZ microdomain is drastically expanded by exclusively overexpressing popZ, fluorescent proteins, including eYFP, can enter the microdomain (Ebersbach et al., 2008; Laloux and Jacobs-Wagner, 2013), suggesting that changes in the microdomain such as occupancy can regulate its permeability prosperities. Indeed, studies of eukaryotic condensates have identified multiple factors that affect permeability, including occupancy by additional binding partners (Banani et al., 2017), charged proteins (Nott et al., 2016), or RNAs (Wei et al., 2017), as well as environmental properties including viscosity, dielectric constant, and scaffold-scaffold interaction strength within the condensate. Our observation that eYFP and fPIF-eYFP cannot enter the wildtype microdomain interior even transiently (Figure 5H) demonstrate that even without membrane barriers or a protein shell, the PopZ microdomain forms a compartmentalized volume of strictly defined composition.
Efficient phospho-transfer within the PopZ microdomain
Joint sequestration of CckA, ChpT, and CtrA acts to locally enhance pathway activity within the PopZ microdomain (Figures 3, 5, 6C, and 7). We previously reported in vitro reconstitution experiments showing that high densities of CckA on liposomes enhance its autokinase activity (Mann et al., 2016). In keeping with these results, CckA functions as a kinase in vivo solely when sequestered at the new pole of predivisional cells (Iniesta et al., 2010a; Jacobs et al., 2003; Jacobs et al., 1999), where the new pole-specific factor DivL is recruited to allosterically modulate CckA kinase activity (Childers et al., 2014; Iniesta et al., 2010a; Tsokos et al., 2011). CckA positioned elsewhere in the cell primarily acts as a phosphatase (Iniesta et al., 2010a; Mann et al., 2016). Indeed, the dependence of CckA autokinase activity on concentration implies that the ~500-1,000x higher CckA density achieved within the new pole is necessary for CckA to act as a phosphate source (SI) (Mann et al., 2016). Further, the elevated concentrations of all three proteins increase the probability of intermolecular binding and phosphate transfer events between CckA and ChpT, and ChpT and CtrA via mass action (Figure 3D), and the vast majority of phospho-transfer events take place within the microdomain (Figure 6B). Similarly, a model for the function of the chromosome partitioning protein ParA (Ptacin et al., 2014) suggests that high ParA concentration and thus an increased rate of homodimerization within the PopZ microdomain is critical for chromosome segregation. More broadly, membrane-less compartments are emerging as an important mechanism for creating specialized condensed signaling environments (Chong and Forman-Kay, 2016; Li et al., 2012; Su et al., 2016). This function is reminiscent of scaffolding molecules that bind multiple signaling proteins simultaneously to regulate selectivity and adaptation to stress (Atay and Skotheim, 2017; Good et al., 2011). Quantitative characterization of the unique properties of each architecture would provide insight into the advantages of using either subcellular organization strategy.
Our results show that in addition to concentrating proteins, the intrinsic character of the PopZ microdomain slows turnover between the pole and the cell body. The observed rates of ChpT and CtrA exit from the pole, on the order of 100 ms (Figure 5), are several times slower than would be expected without sequestration within the microdomain. This slow turnover results from a combination of binding to PopZ and other polar proteins (as implied by localization to the pole), and to percolation through the microdomain volume (reflected by the slow, anomalous polar diffusion of CtrA, Figures 5D, 5F, and 7). Similarly, while CckA still diffuses freely (though more slowly) at the polar membrane, binding interactions slow the exchange of CckA between the polar microdomain and the rest of the cell (Figure 4D). Tracking experiments with gold nanobead labels in mammalian cells have shown that physical obstruction by cortical actin networks causes transmembrane proteins to exhibit “hop diffusion” between nanoscale domains on the sub-ms timescale (Fujiwara et al., 2016). While such phenomena are below the resolution of our experiments, obstruction by PopZ polymers close to the membrane may slow CckA motion at the poles.
A sequestered phospho-signaling cascade drives a gradient of active CtrA
Sensitivity analysis in our reaction-diffusion model showed that sequestration of all three proteins within the new pole microdomain is critical for establishing a CtrA~P gradient (Figure 3C). Both ChpT and CtrA must be localized at high density for rapid phospho-transfer within the pole, while allowing ChpT to transfer its phosphate to CtrA outside of the microdomain would reduce or eliminate the gradient (Figure 3E and 6). Critically, we found replacing ctrA with its phosphomimetic mutant ctrA(D51E), which is active independent of the polar signaling cascade, abolished the gradient of transcriptional activity and disrupts normal cell cycle progression (Figure 2B).
Spatial separation of opposing enzymes in activation–deactivation cycles of protein modification is necessary for maintaining a protein activity gradient (Brown and Kholodenko, 1999). Because the phospho-transfer cascade that activates CtrA is conducted via the intermediate cytosolic protein ChpT, the requirements for maintaining a steady state gradient of CtrA~P are more complex than a canonical gradient established by two factors. In systems with one phospho-transfer step, in which the phosphatase is far from saturating, the gradient decays almost exponentially as a function of the ratio between diffusion of signal and phosphatase rate (Barkai and Shilo, 2009; Tropini et al., 2012; Wartlick et al., 2009). Examples include GTPase RAN (Kalab et al., 2002), the yeast mAPK Fus3 (Maeder et al., 2007), and the yeast protein kinase Pom1 (Moseley et al., 2009). In systems with a cascade of phospho-transfer steps, co-localization of the pathway components plays a key role in gradient shape (Kholodenko et al., 2010). Scaffolding proteins have been shown to modulate signal processing and propagation by recruitment of multiple components to a restricted space and the induction of allosteric effects. Examples include β-arrestin that coordinates multiple pathways downstream of G-proteins coupled receptors (DeWire et al., 2007) and Ste5 that functions in the yeast mating MAPK pathway (Atay and Skotheim, 2017; Bhattacharyya et al., 2006; Good et al., 2009). Here we demonstrated that a bacterial microdomain exhibits a similar function to eukaryotic scaffolds by sequestering all members of the CtrA activation pathway and effectively combining the forward phospho-transfer activities of CckA and ChpT into one localized enhanced source, generating a gradient of CtrA~P across the entire cell (Figure 6 and Figure 7).
A stable CtrA~P gradient facilitates asymmetric division
Caulobacter asymmetric cell division yields daughter cells that exhibits different genetic readouts despite having identical genomes (McAdams and Shapiro, 2009). Differences in gene expression profiles stem from the asymmetric inheritance of the CtrA~P transcription factor controlling the expression of over 100 cycle-regulated genes, many of which encode swarmer cell-specific functions (Laub et al., 2002; Laub et al., 2000). We showed that the CtrA~P distribution is not only skewed toward the new pole but decreases rapidly away from the new pole forming a stable sublinear CtrA~P gradient (Figure 2, Figure 6A). In our simulations, the CtrA~P gradient is formed rapidly and maintains its shape, showing that this output is reached quickly and reliably (Figure 6D). We demonstrated using RT-qPCR that timing of the CtrA-regulated gene expression can be modulated by altering the position of the gene on the chromosome (Figure 2A). These results suggest that the induced asymmetry in CtrA~P concentration before division may be used by the cell as regulator of gene expression timing. Indeed, the relative positions of CtrA regulated genes is conserved across alpha-proteobacteria (Ash et al., 2014) and movement of the ctrA transcriptional unit from the vicinity of the origin to the vicinity of the termini disrupts normal cell cycle progression (Reisenauer and Shapiro, 2002).
The skewed inheritance of CtrA~P, primed by the gradient, is critical for the differential ability of the two daughter cells to initiate chromosome replication. The daughter stalked cell, in which inherited CtrA~P levels are low, initiates a new round of DNA replication immediately after cytokinesis, whereas the daughter swarmer cell, with high levels of CtrA~P, is arrested in G1 phase until it differentiates into a stalked cell (Quon et al., 1996; Quon et al., 1998). In its phosphorylated form, CtrA~P inhibits initiation of DNA replication by binding to specific sites in the origin region (Quon et al., 1998). Indeed, it was previously suggested that CtrA~P is skewed towards the new cell pole based on the observation that additional replication events happened more often at the stalked cell pole when normal cytokinesis and cell cycle progression was disrupted (Chen et al., 2011). Maintaining a sharp gradient of CtrA~P concentration in the predivisional cell therefore facilitates the robustness of asymmetric division by controlling both differential gene expression and replication initiation.
Outlook
It was recently shown that signaling proteins in eukaryotic cells can not only spontaneously separate into liquid-like agglomerated condensates with distinct physical and biochemical properties, but that these condensates can selectively exclude proteins with antagonist function (Su et al., 2016). Our work highlights how selective sequestration of a signaling pathway within a bacterial membrane-less microdomain provides an environment for accelerated biochemistry that directs signaling across the entire cell with exquisite spatial control culminating in differential activity of a key transcription factor that controls an asymmetric division. Although we do not know the physical nature of the microdomain, nor whether the microdomain separates into a liquid- or gel-like phase, its major constituent, PopZ, is a low complexity protein that forms a mesh-like structure in vitro, and self assembles into a distinct cytoplasmic microdomain in vivo. The selective permeability of the PopZ microdomain is reminiscent of the eukaryotic nuclear pore complex (NPC), which mediates active transport between nucleus and cytoplasm (Knockenhauer and Schwartz, 2016). The NPC gains a remarkable sorting selectivity from disordered domains that come together in a meshwork and selectively bind to pockets in the proteins that are marked for transport, suggesting a possible functional parallel to the organization and function of the PopZ microdomain (Fu et al., 2017). Notably, similarities have been drawn between the NPC low complexity domains and P granules, a well-studied membrane-less organelle (Schmidt and Gorlich, 2016). Thus, further structural and functional characterizations of the bacterial PopZ microdomain will be pivotal for revealing fundamental principles for subcellular organization on the mesoscale.
ACKNOWLEDGMENTS
We thank Justin W. Kern for help with plasmid design and construction, Michael D. Melfi for providing a RT-qPCR protocol, Alberto Lovell and Michael R. Eckart at Stanford PAN Facility for their support in conducting RT-qPCR, Jared M. Schrader for sharing unpublished data on mRNA half-life in Caulobacter, Andrew Olson at the Stanford Neuroscience Microscopy Service for assistance with photobleaching experiments, as well as Allison Squires for critical feedback on the manuscript. We also thank Harley H. McAdams for helpful discussions on modeling of signal transduction, Darshankumar Pathak and all members of the Shapiro and Moerner labs for helpful discussions throughout the project. We acknowledge support from the Gordon and Betty Moore Function GBMF 2550.03 to Life Sciences Research Foundation [to K.L.], the Weizmann Institute of Science National Postdoctoral Award Program for Advancing Women in Science [to K.L.], and from NIGMS of the National Institutes of Health under award numbers T32GM007276 [to T.H.M], R01-GM086196 [to W.E.M. and L.S.], R35-GM118067 [to W.E.M.], and R35-GM118071 [to L.S.]. L.S. is a Chan Zuckerberg Biohub Investigator.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
↵4 Co-first author