An experimental framework to assess biomolecular condensates in bacteria

High-resolution imaging of biomolecular condensates in living cells is essential for correlating their properties to those observed through in vitro assays. However, such experiments are limited in bacteria due to resolution limitations. Here we present an experimental framework that probes the formation, reversibility, and dynamics of condensate-forming proteins in Escherichia coli as a means to determine the nature of biomolecular condensates in bacteria. We demonstrate that condensates form after passing a threshold concentration, maintain a soluble fraction, dissolve upon shifts in temperature and concentration, and exhibit dynamics consistent with internal rearrangement and exchange between condensed and soluble fractions. We also discovered that an established marker for insoluble protein aggregates, IbpA, has different colocalization patterns with bacterial condensates and aggregates, demonstrating its applicability as a reporter to differentiate the two in vivo. Overall, this framework provides a generalizable, accessible, and rigorous set of experiments to probe the nature of biomolecular condensates on the sub-micron scale in bacterial cells.

Here, we present an experimental framework ( Fig. 1) to determine whether a bacterial biomolecular condensate forms through phase separation in vivo. We developed and adapted a suite of molecular and cell biology methodologies along with super-resolution imaging techniques in Escherichia coli to characterize the formation, solubility, and dynamic exchange of condensates. We examined the protein McdB, a component of the carboxysome positioning system that robustly forms phase-separated droplets in vitro [23][24][25] , alongside well-established control proteins that form condensates 26 and insoluble aggregates 27,28 . First, we qualitatively probed the liquid-like properties of McdB condensates using overexpression assays. Next, we used tunable expression promoters to quantify the conditions for condensate formation and probed condensate reversibility. Moreover, we measured the dynamic exchange of condensate constituents and inferred the material properties of condensates with singlemolecule tracking. Finally, we implemented the heat-shock chaperone IbpA as a molecular sensor that differentiates between condensates and aggregates. Our framework ensures a low barrier to general applicability by users, consolidates an array of adapted and new assays, and complements widely accessible methods with more advanced techniques such as super-resolution imaging and singlemolecule tracking to validate method performance.

Results
Probing the liquid-like properties of foci formed by protein overexpression Bacterial condensates have been mainly studied in vitro, and it is largely unexplored if these proteins form foci with liquid-like properties inside the cell. Overexpression and imaging of uorescent fusions provide a quick and useful method for gaining insight into the material state of a bacterial focus. The Maintenance of carboxysome distribution protein B (McdB) undergoes robust phase separation in vitro, but it is unknown whether this activity occurs in the cell 24 . Therefore, we rst overexpressed a fully functional mNeonGreen-McdB (mNG-McdB) fusion in E. coli BL21 cells and performed time-lapse uorescence microscopy to qualitatively invesitgate the behaviors of the McdB protein in vivo. Focus formation was observed in less than one hour post induction with IPTG at 16°C (Supplementary Fig. 1a and Supplementary Video 1). Some proximal mNG-McdB foci fused into larger structures on the seconds timescale ( Supplementary Fig. 1b). DAPI staining of the nucleoid showed that these larger mNG-McdB foci were nucleoid-excluded ( Supplementary Fig. 1c). Strikingly, after three hours, larger mNG-McdB foci localized to the inner membrane near sites of local cell curvature (white arrows in Supplementary Fig. 1a and Supplementary Video 1). We speculate mNG-McdB foci wet to the membrane via nonspeci c electrostatic associations and locally occlude cell wall synthesis. The resulting asymmetry in cell wall growth thereby induces cell curvature. Together, we nd that mNG-McdB forms small foci upon reaching a threshold cellular concentration, and that these small foci fuse to form large, nucleoid-excluded structures that appear to wet to the inner membrane and cause local changes in cell morphology.
A hallmark of phase-separated condensates is their immediate responsiveness to changes in the cellular environment, such as variations in temperature or cell volume. We rst examined the reversibility of mNG-McdB foci via time-lapse microscopy as the stage-top temperature increased from 25 to 37°C. Despite continued overexpression of mNG-McdB, the foci disappeared within eight minutes, indicative of reversal into a single homogenous phase in the cytoplasm (Supplementary Fig. 1d and Supplementary Video 2). To probe the concentration dependency of mNG-McdB foci formation in vivo, we treated the cells with A22 which caused a rod-to-sphere transition and corresponding increase in cell volume. During this transition, mNG-McdB foci dissolved into a homogenous phase in the cytoplasm ( Supplementary Fig. 1e, Video 3), suggesting that the cellular levels of mNG-McdB dropped below c sat .
Because temperature shifts and drug-induced changes to cell volume are gradual and may involve pleiotropic effects, we developed an approach to probe the dynamic properties of mNG-McdB foci upon an instantaneous change to cell volume. In this localized-lysis method, a high-intensity laser was focused on one cell pole to lyse the cell. Localized lysis caused cell contents to unidirectionally rush out of the cell. Strikingly, upon cell lysis, the mNG-McdB focus signal dispersed towards the opposing open cell pole (Supplementary Fig. 1f and Supplementary Video 4). The observed dynamics resemble the observed jetting of P granules under sheer stress 29 . Altogether, these results demonstrate that mNG-McdB foci in E. coli BL21 exhibit properties consistent with those of phase-separated condensates.
Using tunable promoters to probe the formation of biomolecular condensates In eukaryotic cell biology, the majority of condensate studies perform in vivo measurements using ectopic overexpression 30 . However, phase-separating systems are exceedingly sensitive to changes in concentration, and overexpression may introduce signi cant caveats in the extrapolation that a protein forms condensates when expressed at lower endogenous levels. We set out to nd additional metrics other than overexpression to support claims that a bacterial focus is indeed phase-separated.
We turned to a heterologous expression system to observe condensate formation during controlled protein expression. We fused the uorescent protein mCherry to the N-terminus of McdB(mCherry-mcdB) and placed it under the control of an IPTG-inducible promoter on a pTrc99A expression vector (pTrc99A-mCherry-mcdB). Along with our protein of interest, we probed a series of well-established control proteins, all fused to mCherry. The protein cI agg is a truncated and mutated version of the Lambda cI repressor that is well-known to form insoluble aggregates in E. coli 28 . On the other hand, uorescent protein fusions to PopTag proteins form condensates via phase separation with tunable material properties that depend on the length of the linker between PopTag and the fusion protein 26 . We engineered two versions of the PopTag fusion: PopTag SL with a short (six-amino acid) GS repeat linker (pTrc99A-mCherry-L6-PopTag) and PopTag LL with the native linker of the PopZ protein (78-amino acid) (pTrc99A-mCherry-L78-PopTag).
Based on the linker lengths, PopTag LL -mCherry condensates should be more uid than PopTag SL -mCherry condensates, which we expect to be more viscous or in a gel-like state 26 . Additionally, we probed a solubilized McdB mutant (McdB sol ), previously shown to be abrogated in its phase separation activity both in vitro and in vivo 25 . Finally, mCherry alone was used as a control for a completely soluble protein.
All mCherry fusion proteins showed a lower degradation level (less than 20%) compared to mCherry alone when expressed with the pTrc99A promoter ( Supplementary Fig. 2).
Induced expression of these proteins in E. coli cells showed focus formation as a function of increasing protein concentration over time (Fig. 2a-b and Supplemental Fig. 3). After one hour of expression, approximately 60% of mCherry-cI agg and mCherry-PopTag SL cells had a focus, while no foci were observed in mCherry-PopTag LL , -McdB, and -McdB sol cells (Fig. 2b). Between one and ve hours of expression, the percentage of mCherry-cI agg and mCherry-PopTag SL cells with a focus slightly increased from 60 to 80%, while that of the mCherry-PopTag LL and mCherry-McdB cells signi cantly increased to 70% and 60%, respectively. The proportion of the mCherry-McdB sol cells with a focus was two-fold lower than that of mCherry-McdB at 5 hours. Strikingly, a notable fraction of the uorescence signal was localized to the cytoplasm of cells with a mCherry-PopTag SL , -PopTag LL , -McdB, or -McdB sol focus, but not in the mCherry-cI agg cells (Fig. 2a). The focus and the cytoplasmic fraction resemble the dense phase and dilute phase in a two-phase system.
Next, we used quantitative uorescence microscopy to determine the apparent in vivo saturation concentration (c sat_app ) of our protein of interest, McdB. We rst determined the intensity of single mCherry molecules that were spatially isolated prior to photobleaching of uorescence in the cell. These single-molecule localization images were quanti ed to measure the number of photons detected per molecule per imaging frame ( Supplementary Fig. 4). Next, we calculated the cellular concentration of McdB by integrating the total cellular uorescence intensity per imaging frame and dividing this value by the mCherry single-molecule intensity and the cellular volume. Cells were then classi ed by the presence or absence of a focus. After a 4-hour induction, cells expressing mCherry-McdB without a detected focus had an average cellular concentration of 92 ± 29 µM, while the concentration in cells with a focus was 113 ± 37 µM (Fig. 2c). This observation of condensates in cells with a slightly higher total protein concentration is consistent with McdB undergoing a nearly immediate and concentration-dependent transition to form a focus. We performed the same analysis for cells expressing mCherry-PopTag LL . The average cellular concentration of this uid-condensate control was 32 ± 20 µM in cells without a focus while cells with a detected focus had an average concentration of 40 ± 20 µM (Fig. 2c). PopTag SL formed condensates prior to IPTG induction, which prevented similar analyses. Together, the results provide a c sat_app at which condensates form; for mCherry-PopTag LL and mCherry-McdB, we estimate the c sat_app to be between 32-40 µM and 92-113 µM, respectively.
Condensates coexist with a soluble phase A hallmark of phase separation is the existence of a soluble fraction in the cytoplasm. To quantify the ratio of mCherry protein fusion concentration in the focus (dense phase) to the concentration in the cytoplasmic fraction (dilute phase), the partitioning of each fusion was measured (Fig. 2d). The largest partitioning was measured for the aggregator control, mCherry-cI agg , in which there was no detectable uorescent signal in the cytoplasm. The mCherry-PopTag SL condensate control partitioned to a greater extent than the more uid condensate control mCherry-PopTag LL . The partitioning of our protein of interest, mCherry-McdB, was intermediate relative to the PopTag controls. We also analyzed condensation in cells by normalizing all pixel intensities in hundreds of cells per protein and calculating the fraction of pixels below a normalized intensity threshold of 30% ( Supplementary Fig. 5) 16,31 . The relative condensation coe cients acquired from this analysis (Fig. 2e) were consistent with the partition ratios (see Fig. 2d).
To further inspect the partitioning of proteins to polar foci, we titrated photoactivatable (PA) mCherry fusions of each protein and performed single-molecule localization super-resolution microscopy to generate normalized localization density heat maps (Fig. 2f). Consistent with our previous results, PAmCherry-cI agg had minimal localization density in the cytoplasm under all conditions, which suggests nearly all protein is recruited to the polar aggregates. On the other hand, the polar density of PAmCherry-PopTag SL increased with increasing protein concentration, but these cells also maintained a protein fraction in the cytoplasm. Strikingly, PAmCherry-PopTag LL and -McdB displayed a transition between no induction and two hours post induction, indicative of focus formation. Intriguingly, PAmCherry-McdB sol also formed high-density regions at the poles, consistent with bulk uorescence measurements (see  25 ; therefore, we speculate that the localization pattern of the uidized PAmCherry-McdB condensate is due to nucleoid exclusion by repulsive electrostatic interactions. Consistently, we found that mCherry-McdB sol foci expanded and encroached throughout the cytoplasm as the nucleoid was compacted via drug treatment (Supplementary Fig. 6 and Supplementary Video 5). Collectively, our results demonstrate the use of a tunable expression system that shows the concentration-dependent formation of condensates that exhibit a two-phase behavior, indicative of phase separation. Phase separation theory predicts that while many small condensates form at the initiation of phase separation, this number decreases, and the sizes of those that persist increase via coalescence 32,33 .
Our ndings here are consistent with the expected nal ground state being a single large condensate that coexists with a dilute phase. Some cells had foci at opposing poles. However, this can be attributed to the steric effects of nucleoid exclusion that prevent condensates from physically interacting and coalescing.
Bacterial cell growth and division probe the reversibility of condensates Condensates should dissolve if cellular levels of the protein drop below c sat 7 . To probe condensate reversibility in bacteria, we implemented two approaches that lower the protein concentration in the system following focus formation. We expected that driving the protein concentration below the protein c sat would dissolve the condensates and thus demonstrate solubility as a driving force for condensate formation. First, each protein was expressed via IPTG induction, and once foci formed, expression was stopped to maintain a constant cellular protein level. Cells were then allowed to grow and divide, and in doing so, dilute the concentration of the mCherry fusion proteins. Noticeably, mCherry-cI agg foci remained at the pole from which they formed even as cells grew and divided over 15 hours (∼9 generations), resulting in an average focus lifespan of 14.3 ± 2.2 hours (Fig. 3a-c and Supplementary Video 6). This result indicates that there is no concentration dependence in the formation and maintenance of mCherry-cI agg foci, and further supports the previously reported insolubility of cI agg foci 28 .
In contrast, mCherry-PopTag SL , mCherry-PopTag LL , and mCherry-McdB foci all exhibited concentrationdependent dissolution. mCherry-PopTag SL foci had an average lifespan of 11.0 ± 4.4 hours (Fig. 3b-c, Supplementary Fig. 7a), and the total focus intensity dropped approximately two-fold over this time period, suggesting a decrease in focus size. The generational dilution effect was more immediate with the mCherry-PopTag LL ( Supplementary Fig. 7a), which exhibited a rapid decrease in total focus intensity and full dissolution within one to three cell divisions and an average focus lifespan of 2.3 ± 0.8 hours.
When compared to these controls, mCherry-McdB focus dissolution mirrored that of the uid mCherry-PopTag LL condensate, with an average focus lifespan of 2.6 ± 1.3 hours or within one to three cell divisions. Indeed, many mCherry-McdB foci dissolved following a single cell division event (Supplementary Video 6). Interestingly, we observed some mCherry-McdB and mCherry-PopTag SL foci reforming immediately after cell division ( Supplementary Fig. 8). This observation is consistent with the low c sat of mCherry-McdB and mCherry-PopTag SL . As the cell divides, the volume decreases, leading to an increase in the concentration back above c sat in the daughter cell, which ultimately results in the reformation of a focus.
We also probed condensate solubility by diluting protein concentration via an increase in cell length without further protein expression. E. coli cells were prepared as described above, but with the inclusion of the cell division inhibitor cephalexin (10 µg/ml). We expected similar trends for dilution via cell elongation as in cell division, and indeed mCherry-cI agg foci were observed to persist throughout the duration of the experiment (10 hours Strikingly, the average lifespan of mCherry-McdB foci was 5.6 ± 3.1 hours, which was roughly twice as long as that found in the cell division experiments above. This increase in average lifetime was in uenced by the persistence of some foci over the duration of imaging, which resulted in a bimodal distribution of foci lifespan for mCherry-McdB (Fig. 3f). The reason for this persistence of some foci is unknown. But we speculate that the altered cytoplasmic conditions (e.g. effective volume and crowding) of multi-nucleate elongated cells, compared to cells undergoing vegetative growth, play a role in the observed differences in focus dissolution.
Together, using our assays of protein dilution via cell division or cell elongation, we observed decreases in focus size toward its eventual dissolution, and a corresponding turnover of protein into the cytoplasmic phase. The ndings are consistent with the reversal to a one-phase system once the concentration has decreased below c sat . These characteristics further support condensate formation by the PopTag fusions via phase separation and suggest a similar formation process for McdB condensates.
Probing the dynamic rearrangement, con nement, and exchange of biomolecular condensates To probe the dynamic rearrangement of molecules within a focus and their exchange with the surrounding cytoplasm, we rst implemented uorescence recovery after photobleaching (FRAP) on the mCherry fusion foci. The aggregator control, mCherry-cI agg , exhibited no uorescence recovery ( Fig. 4a-b), supporting the static nature of the proteins within these aggregates and the absence of protein exchange with the cytoplasm. The PopTag fusions partially recovered to an extent consistent with their respective uidity levels: mCherry-PopTag SL and -PopTag LL recovered to approximately 10% and 40%, respectively, of the initial uorescence intensity (Fig. 4b). Recovery of mCherry-McdB foci was once again similar to that of the uid mCherry-PopTag LL condensate. Taken together, the data suggest that McdB exhibits dynamics similar to that of the uid PopTag LL , while the minimal recovery of PopTag SL is consistent with its predicted gel-like state. Given the limited number of pixels that make up the photobleached foci, we note that the uorescence recovery contributions from internal rearrangements of molecules within a focus cannot be distinguished from exchange of molecules with the surrounding milieu 30 . While FRAP is commonly used to determine if a compartment is liquid-like, this method is solely a measure of exchange dynamics, so FRAP alone is not a reliable "gold standard" measure of the material state of a focus.  Fig. S5). PAmCherry-PopTag LL also exhibited nucleoid exclusion (Fig. 4c), despite remaining highly mobile. As stated earlier, we speculate that this behavior is caused by charge repulsion with the nucleoid.
By comparing the average mobility of D app, slow populations across all fusions, we nd that PAmCherry-cI agg molecules within a focus were essentially static (Fig. 4d), consistent with being an aggregate. was a minor fraction of the total population ( Fig. 4e and Supplementary Figure S7). Collectively, the data parse the dynamic exchange of molecules in foci and the surrounding cytoplasm, and delineate the spectrum of possible mobilities within a focus, thus allowing for inference of its material state.
IbpA as a reporter to differentiate between condensates and aggregates The heat shock chaperone, IbpA, has been shown to colocalize with insoluble aggregates in E. coli 28 . We therefore initially thought that IbpA may work as an in vivo reporter that discriminates between aggregates and condensates, colocalizing with the aggregator control and not recognizing PopTag or McdB condensates. We expressed the mCherry fusion proteins under focus-forming conditions in an E. coli strain that expressed a chromosomal uorescent reporter of IbpA (IbpA-msfGFP) and then observed colocalization patterns of the mCherry fusion proteins with IbpA. Unexpectedly, the IbpA foci strongly colocalized with all proteins surveyed ( Supplementary Fig. 10a), showing that IbpA does not speci cally localize to misfolded protein aggregates in the cell as ascribed by current models 28 . Upon further investigation, we observed that the localization pattern of IbpA association with the aggregate control versus the condensates was quanti ably different. We quanti ed and compared the diameters of the IbpA signal projections relative to each mCherry fusion ( Supplementary Fig. 10b) and found that the IbpA signal was spread over a larger area than the mCherry-cI agg foci. We also found that in many instances IbpA appeared to coat, as opposed to penetrate, the mCherry-cI agg foci ( Supplementary Fig. 10a).
We extracted from the wide-eld uorescence data an intensity pro le line plot of foci in both color channels. When normalized to the maximum intensity of the mCherry signal, we observed signi cant differences in the relative amounts of IbpA penetrating each of the mCherry fusion foci (Fig. 5a). On average, the max intensity of IbpA was 20 ± 9% of that of mCherry-cI agg foci, while it was 58 ± 33, 188 ± 131, and 33 ± 35% for mCherry-PopTag SL , -PopTag LL , and -McdB foci, respectively. The limited amount of IbpA relative to cI agg suggested to us that while IbpA can sense the aggregate, it cannot penetrate it well. In contrast, the higher amounts of IbpA in all other foci pointed to its ability to penetrate more uid assemblies. This result demonstrates varying degrees of IbpA penetration within the corresponding mCherry fusion foci which correlate well with their hypothesized material state and suggest a different mode of colocalization of IbpA with biomolecular condensates versus aggregates.
To better resolve the colocalization patterns between IbpA and the mCherry fusion foci, we imaged cells by 2D structured illumination microscopy (SIM) under the same conditions as described earlier. With the increased resolution in both channels, the localization patterns of IbpA relative to the mCherry fusion foci were more easily observed (Fig. 5b). Using a projection of all detected foci for each mCherry fusion superimposed onto one another, we observed that IbpA forms a rosette pattern around the mCherry-cI agg foci, a punctate focus within mCherry-PopTag SL foci, and an amorphous focus pattern with mCherry-PopTag LL and mCherry-McdB (Fig. 5c). This result demonstrates that IbpA exhibits a different colocalization pattern with biomolecular condensates compared to aggregates and better penetrates more uid condensates. These patterns serve as a proof of concept that IbpA is a reporter capable of differentiating between these macromolecular assemblies in vivo.

Discussion
In this study, we developed an experimental framework to assess the material state of uorescent foci in bacteria; speci cally, whether a focus can be described as a phase-separated condensate. Inducercontrolled protein expression combined with quantitative uorescence microscopy was shown to have general applicability in identifying the in vivo c sat for condensate formation. Decreasing cellular protein levels back below c sat , via changes in cell shape or by generational dilution, can then probe the reversibility of condensate formation. We further show that protein dynamics within both the cytoplasm and the focus can infer its material state. Finally, we identi ed the heat-shock chaperone, IbpA as a molecular sensor that surrounds solid aggregates but penetrates condensates. We demonstrated the versatility of these approaches by using foci-forming control proteins that span the spectrum of material states from a solid aggregate to a highly uid condensate. When compared to these control proteins, we nd that our protein of interest, McdB, robustly phase-separates into liquid-like condensates in vivo. As shown with the uidized mutant of McdB, this framework can also be combined with mutagenesis studies to determine the regions and residues of a protein governing its material state and phase separation behavior in a bacterial cell.
Our framework overcomes current limitations in identifying the material state of a uorescent focus and can be used to assess the phase separation activity of expressed recombinant proteins and bacterial inclusion bodies (IBs). IBs are mesoscale protein aggregates, once strictly proposed as being composed of nonfunctional and misfolded proteins 34 . Phase separation has only recently been considered in the assembly and organization of IBs 35 . For example, several condensates have been shown to mature into gels and solid amyloids 1 . However, direct determination of whether a bacterial IB is a liquid, gel, solid, or a mixture of these states remains to be demonstrated.
Inclusion body binding protein A (IbpA) of E.coli belongs to the conserved family of ATP-independent small heat shock proteins, well-established in binding protein aggregates and driving them towards reactivation-prone assemblies 36-38 . As such, we presumed IbpA would serve as a molecular sensor that would selectively associate with protein aggregates, but not condensates. Instead, we found that IbpA surrounded protein aggregates and penetrated condensates. Moreover, the degree to which IbpA colocalized with condensates strongly correlated with increasing uidity. Consistent with our ndings, the Drummond group has recently shown that condensates are dispersed by chaperones far more rapidly than misfolded aggregates 39 . These ndings warrant a reevaluation of the function of chaperone systems governing protein homeostasis and demonstrate the utility of IbpA, and potentially other chaperones, as molecular sensors for the material state of uorescent foci in bacteria.
The framework was built by contrasting the principles of aggregation versus phase separation behaviors of proteins. While not an exhaustive list, the approaches used here are highly accessible and probe several aspects of condensate assembly, maintenance, and dissolution. The framework is not without limitations. First, the approaches used here do not provide mechanistic insights into the process of phase separation, that is, whether condensate formation is driven purely by liquid-liquid phase separation (LLPS), phase separation coupled to percolation (PSCP), or phase separation coupled to other phase transitions (PS++) 9 . The current framework also does not include an assessment of the boundary between the dense and dilute phases, which is important in understanding the nite interfacial tension that hinders macromolecular transport across the boundary. Also, for broad use and accessibility, we developed the experimental framework using heterologous protein expression in E. coli, which may not accurately re ect the in vivo conditions of the native host. Finally, low-level leaky expression of inducible promoters may preclude the examination of proteins that phase separate at very low in vivo c sat . Despite these limitations, this framework provides broad-use and systematic approaches that address ongoing debates over the rigor and standardization of phase separation assessments in bacterial cells.

Methods
Bacterial strains, plasmids, and primers Strains, plasmids, and primers used in this study are listed in Tables S1 and S2. All constructs were made using Gibson assembly 40 from PCR fragments or synthesized dsDNA (Integrated DNA Technologies) and veri ed by Sanger sequencing. For example, plasmid pCA3 was generated from plasmid pTrc99A-mCherry-cI78 EP by replacing mCherry with PAmCherry. The PAmCherry fragment was generated using primers YH1 F and YH1 R. The plasmid pTrc99A-mCherry-cI78 EP8 was ampli ed using YH2 F and YH2 R to generate the second fragment. The two fragments were then added to a Gibson assembly reaction to enzymatically join the overlapping DNA sequences. Other plasmids were generated using similar methods with primers indicated in Table S2. When relevant, homology regions for Gibson assembly are indicated in blue. Plasmids were introduced into their respective host strains by chemical transformation and selection for antibiotic resistance encoded by the plasmid. All plasmids are available on AddGene.

Growth conditions
Lysogeny broth (LB) and AB media were used as either a broth or solid for culturing bacteria. LB medium was used to grow E. coli BL21 Arctic Express (AE) and overnight cultures of E. coli MG1665. The minimal AB medium was used when inducing protein expression in E. coli MG1665 to ensure protein expression reproducibility, as all the components are de ned; as opposed to a complex medium such as LB 28,41 . AB medium was supplemented with 0.2% of a carbon source (glycerol for growth or glucose to also inhibit basal protein expression from the P trc promoter), 0.2% casamino acids, 10 µg/ml thiamine, and 25 µg/ml uracil 28 . E. coli was grown in a 15-ml tube overnight for approximately 15 h in 5 ml of LB broth at 37°C on an orbital shaker at 200-225 rpm. Exponential phase cultures were prepared by diluting overnight cultures 1:100 and further incubated until an OD 600 of 0.2-0.6 was reached. When appropriate, the following chemicals were added to the medium at the indicated nal concentrations: carbenicillin (100 µg/ml) for selecting the plasmids in culture, IPTG (0.1-1 mM) or L-arabinose (0.2%) for protein induction.

Total protein and immunoblot analyses
A 0.4-ml aliquot of E. coli cells (OD 600 : 0.2-0.4) was lysed using a Qsonica cupped-horn sonication system (20 cycles, 30 s on, 10 s off at 30% power) and centrifuged at 10,000 x g for 1 min at 4°C. The protein content in the supernatant was measured using a Bradford assay kit according to the manufacturer's instructions. To prepare samples for immunoblot analysis, an equal volume of 4x Laemmli sample buffer was added to E. coli culture prior to boiling 20 min. One microgram of total protein from each sample was loaded in each lane of a 4-12% Bis-Tris NuPAGE gel. Gels were transferred onto a mini-size polyvinylidene di uoride membrane using a Trans-Blot Turbo system (Bio-Rad). The membrane was immunoprobed using a rabbit polyclonal antiserum against mCherry (1:2000). The membrane was then incubated with the goat anti-rabbit IgG Secondary Antibody IRDye 800CW.
Membrane signals were visualized and quanti ed using LI-COR Image Studio. The mCherry band of each lane was normalized to the total intensity of the lane to calculate the degradation level of mCherry fusion proteins.

Wide-eld uorescence and phase-contrast imaging
Wide-eld uorescence and phase-contrast imaging were performed using a Nikon

Time-lapse videos of protein induction in E. coli BL21
Gene mNG-McdB and mNG-cI agg were cloned into the multiple cloning site of the vector pET11a to create the pJB37 and pYH73 plasmids respectively, used for inducible expression under the control of a bacteriophage T7 promoter. pJB37and pYH73 were transformed into BL21 (AE) cells and a 5 ml overnight culture containing 100 µg/ml of carbenicillin in LB medium was grown at 37°C with shaking at 225 rpm. The overnight culture (50 µl) was used to inoculate 5 ml of LB supplemented with 100 µg/ml of carbenicillin in a 15-ml tube. The cells were grown at 37°C with shaking at 225 rpm to an OD 600 of 0.  Temperature shift E. coli BL21 (AE) cells with plasmid pJB37 were grown to an OD of 0.5 and then asks were plunged in an ice bath as described above for 2 min. mNG-McdB expression was then induced by the addition of 1 mM IPTG/0.2% Arabinose solution to the ask. Cells (2 µl) were then immediately spotted on a 1 cm diameter agarose pad. After two minutes the cell-containing side of the pad was ipped onto a 35-mm glass-bottom dish and mounted onto a stage-top incubator with temperature control. mNG-McdB expression and focus formation was imaged using the "YFP" lter set (see above). Once mNG-McdB initiated focus formation, the stage-top temperature was ramped up from 25 to 37°C. An image series was captured every 5 s for 30 min to observe focus dissolution. Cellular shape change in E. coli BL21 E. coli BL21 (AE) cells with plasmid pJB37 and pYH73 were induced with 1mM IPTG/0.2% Arabinose as described in the "Time-lapse videos of protein induction in E. coli BL21" section. The MreB inhibitor A22 (10 µg/ml) was added to the exponential phase cultures at time t = 0. The cultures were then incubated at 37°C with shaking at 225 rpm for 6 hours. Images were taken at 6 h post-treatment with the "YFP" lter set (see the "Wide-eld uorescence and phase-contrast imaging" section).

Tunable induction of mCherry fusion focus
Gene mCherry-cI agg , mCherry-PopTag SL , mCherry-PopTag LL , mCherry-mcdB, mCherry-mcdB sol , and mCherry were cloned into the multiple cloning site of the vector pTrc99A to create the pTrc99A-mCherry-cI78 EP8 , pYH75, pYH80, pYH71, pYH86, and pYH77 plasmids respectively. The above plasmids were transformed into MG1665 cells and a 5 ml overnight culture containing 100 µg/ml of carbenicillin in AB medium (with 0.2% glycerol as the carbon source) was grown at 37°C with shaking at 200 rpm. The overnight culture (50 µl) was used to inoculate 5 ml of AB (with 0.2% glycerol as the carbon source) supplemented with 100 µg/ml of carbenicillin in a 15-ml tube. The cells were grown at 37°C with shaking at 200 rpm to an OD 600 of 0.2-0.6. 500 µM of IPTG was added to the exponential phase culture to induce uorescent fusion protein expression. After each hour, 2 µL of the MG1665 culture were spotted on an agarose round pad. The pads were prepared by dissolving Ultrapure agarose in AB to a nal concentration of 1.5%. A series of images were taken with the "Texas Red" lter set (see the "Wide-eld uorescence and phase-contrast imaging" section) every 1 hour for 5 hours.
Nucleoid compaction E. coli MG1665 with pYH86 plasmid was induced with 1 mM IPTG for 2 hours. Cells were then stained with 2 µM DAPI and spotted on an agarose pad that has 50 µM of cipro oxacin. A series of images were taken with the "Texas Red" and "DAPI" lter sets (see the "Wide-eld uorescence and phase-contrast imaging" section) every 15 min for 7 hours.
Time-lapse videos that examined focus reversibility in E. coli MG1665 E. coli MG1665 with pTrc99A-mCherry-cI78 EP8 , pYH75, pYH80, pYH71, pYH86, and pYH77 plasmids were induced with 1mM IPTG for 2 h. The cells were then washed three times with 10X volume of fresh AB media supplemented with 0.2% glucose and incubated for 30 min at room temperature before imaging. After 30 min, 2 µL of the MG1665 culture were spotted on an agarose round pad. The pads were prepared by dissolving Ultrapure agarose in AB to a nal concentration of 1.5% (the AB medium contains 0.2% glucose and 10 µg/ml cephalexin when speci ed). A series of images were taken with the "Texas Red" lter set (see the "Wide-eld uorescence and phase-contrast imaging" section) every 15  normalized such that the pre-bleach signal is one and the rst frame post-bleaching is zero.
Imaging the IbpA protein with wide-eld uorescence microscopy The plasmids pTrc99A-mCherry-cI78 EP8 , pYH75, pYH80, pYH71, and pYH77 plasmids were transformed into E. coli MG1665 that expresses IbpA-msfGFP by its native promoter. A 5 ml overnight culture containing 100 µg/ml of carbenicillin in AB medium (with 0.2% glycerol as the carbon source) of each strain was grown at 37°C with shaking at 200 rpm. The overnight culture (50 µl) was used to inoculate 5 ml of AB (with 0.2% glycerol as the carbon source) supplemented with 100 µg/ml of carbenicillin in a 15ml tube. The cells were grown at 37°C with shaking at 200 rpm to an OD 600 of 0.2-0.6. 1mM of IPTG was added to the exponential phase culture to induce uorescent fusion protein expression for 2 h. 2 µL of the MG1665 cultures were spotted on an agarose round pad. The pads were prepared by dissolving Ultrapure agarose in AB to a nal concentration of 1.5%. Images were taken with the "GFP" and "Texas Red" lter set (see the "Wide-eld uorescence and phase-contrast imaging" section) at 2 h post-induction.
Imaging the IbpA protein with 2D Structured illumination microscopy (SIM) The agarose pads with induced MG1665 cells were prepared similar to those in the "Imaging the IbpA protein with wide-eld uorescence microscopy" section. SIM images were acquired on a Nikon N-SIM system equipped with a Nikon SR HP Apo TIRF 100X 1.49NA objective, a Hamamatsu ORCA-Flash4.0 camera (65 nm per pixel), and 488 nm and 561 nm lasers from a Nikon LU-NV laser launch. Cells were identi ed using DIC to avoid photobleaching. For each 2D-SIM image, nine images were acquired in different phases via the built-in 2D SIM modes. Super-resolution image reconstruction was performed using the Nikon Elements SIM module.
Single-molecule uorescence microscopy Cells expressing protein fusions to PAmCherry of a plasmid with the pTrc inducible promoter were grown in 3 ml of lysogeny broth (LB) medium in a culture tube for ~ 14 h, with shaking at 225 rpm at 37°C. The following day, cells were diluted 1:100 into 3 ml of fresh AB medium supplemented with 0.2% glucose or glycerol and grown to OD 600 ~ 0.3 before imaging either as is ("no induction" condition) or inducing with 100 µM IPTG ("inducing" condition). After a two-or four-hour induction, cells were washed three times with 1 ml of AB medium supplemented with 0.2% glucose. All "no induction" and "inducing" cells were resuspended in M9 medium supplemented with 0.2% glucose for imaging. Agarose pads were made at 2% (w/v) with M9 minimal medium supplemented with 0.2% glucose. An aliquot of 2.5 µL of cells was loaded onto an agarose pad and sandwiched between two coverslips. Cells were imaged at room temperature with a 100× 1.40 numerical aperture oil-immersion objective. A 406-nm laser (Coherent Cube 405 − 100; 0.2 W/cm 2 ) was used for PAmCherry photoactivation and a 561-nm laser (Coherent-Sapphire 561 − 50: 88.4 W/cm 2 ) was used for imaging. Given the difference in protein expression level, 200-400 ms activation doses were used for cells not induced and 50 ms activation doses were used for the two hour induction sample. The four hour sample had many preactivated molecules and molecules activated spontaneously without activation via 405-nm laser, which limited our ability to con dently localize single molecules. To reduce the number of molecules per imaging frame, an initial 405-nm laser activation (10-15 s) was followed by photobleaching through illumination with a 561-nm laser with the same power density as above for 15-20 minutes or until spatially resolved single molecules were observed. The uorescence emission was ltered to eliminate the 561-nm excitation source and imaged at a rate of 40 ms/frame using a 512 x 512-pixel Photometric Evolve EMCCD camera.

Single-molecule data analysis
Phase-contrast microscopy images were used to segment bacterial cells (see Cell segmentation for details) prior to localization and tracking. Single molecules were detected and localized with a 2D Gaussian tting by the SMALL-LABS algorithm 42 and connected into trajectories using the Hungarian algorithm 43 . Localization heat maps (Fig. 2f) were made by normalizing the segmented cells and rotating them onto their long axes, followed by projection, binning, and symmetrization of the single-molecule localizations onto the normalized cell 44,45 . In this case, D app is the apparent diffusion coe cient, τ is the time lag, and describes the localization precision. Only ts with R 2 ≥ 0.7 were kept. The log distribution histograms of the D app were each t to a two-state Gaussian mixture to determine the D app and the associated weight fraction for the slow and fast diffusion modes ( Supplementary Fig. 7).

Measuring single-cell protein concentrations
To determine the average number of photons detected from a single mCherry molecule per imaging frame, cells expressing the indicated mCherry fusion proteins were grown in 3 ml of AB medium supplemented with 0.2% glucose to OD ~ 0.3 and washed once in 1 ml of M9 minimal medium prior to imaging. Cells were pre-bleached with a 561-nm laser until only a few isolated molecules were observed.
Images were then recorded at 40-ms exposure with a 561-nm laser (110.5 W/cm 2 ) and an input EM gain of 600 (NIS-Elements software setting). Single molecules were detected as described earlier. The number of photons detected per single molecule per imaging frame was obtained by calculating the integrated intensity counts of the 2D Gaussian t from the localization step 47 , which was then converted into the number of photons using the following camera calibrations.
The conversion gain (number of photoelectrons per uorescence intensity count) calibration was performed as described in a Teledyne Photometrics Technical Note 48 . Brie y, multiple images of a white business card were acquired for 10-, 20-, 40-, 80-, 160-, and 320-ms exposure times with no electron multiplication (EM) gain. To account for the camera bias, 100 frames were acquired with nominally 0-ms exposure and a shuttered camera path. The average of these 100 frames was subtracted from every subsequently analyzed image. A plot of the mean signal of the image for each exposure time versus the variance of the same image gives a straight line with a slope that equals the conversion gain ( Supplementary Fig. 3a). Pixel-to-pixel nonuniformity effects were removed by recording two images at each exposure time 49 . The conversion gain of our camera was found to be 1.40 electrons per intensity count.
To determine the EM gain (the number of electrons per photoelectron), two images were recorded: one long-exposure image (1 s) with no EM gain and one short-exposure image (10 ms) with an arbitrary EM gain. The bias was subtracted from both images. The EM gain multiplication factor is the factor difference in signal per time unit between the corrected images. We repeated this procedure for various software setting EM gains and found a linear t in the range of 5-600 input EM gain with a conversion factor of 0.15 between nominal EM gain and output EM gain (Supplementary Fig. 3b).
To calculate the number of photons from the uorescence image, we therefore multiplied the intensity counts recorded by the conversion gain of 1.40 and divided this quantity by the input EM gain multiplied by the EM gain conversion factor of 0.15. The distribution of photons per molecule per imaging frame was t to a Gamma distribution resulting in a peak of ~ 90 photons per molecule per imaging frame ( Supplementary Fig. 3c). σ To determine the apparent cellular concentration of mCherry-McdB and mCherry-PopTag LL when foci are present, cells expressing these proteins were grown and prepared as described above. The mCherry-McdB strain was induced with 1 mM IPTG for 2 h prior to imaging and the mCherry-PopTag LL was induced with 100 µM IPTG for 2 h prior to imaging. Cells were imaged with a 561-nm laser (110.5 W/cm 2 ) at 20 ms exposure and 10x (software setting) EM gain. To minimize the effect of photobleaching on the photon counting measurement, image acquisition was started prior to laser illumination.
The brightest ve frames in the movie were averaged to determine the cell brightness before photobleaching. The integrated total uorescence emission within a cell is linearly related to the total quantity of McdB or PopTag LL molecules per cell: this intensity value was divided by the number of photons per mCherry molecule per imaging frame to obtain the McdB or PopTag LL copy number per cell.
To determine the cellular concentration of the proteins, the volume of the cell was estimated by modeling it as a cylinder with spherical caps 50 .

Image analysis Cell segmentation
Cell segmentation was performed with the Cellpose 51 package in Python. To train the model optimally for bacterial cell morphology, 26 raw phase contrast images of cells were manually annotated. To segment the cells using the trained model, a Gaussian blur (standard deviation of Gaussian = 0.066 µm) was applied to the bacterial cell phase-contrast images and the blurred cells were segmented. Cells touching the borders of the image were ignored. Erroneous segmentations were manually corrected using the Cellpose GUI or excluded from further analysis.

Condensation analysis
Condensation coe cients were calculated as described previously 16,31,52 . Brie y, the uorescence intensity for each pixel in a cell was corrected for background by subtracting the median value of all pixels in an image outside of the cell regions and normalized by the minimum and maximum pixel intensity values in the corresponding cell I n = (I -I min ) / (I max -I min ) (2) These normalized pixel intensities were then binned to generate histograms that represent the localization pattern for each protein (Supplementary Fig. 4a-f).
Next, cells were classi ed by the presence or lack of a focus. An ROI was de ned for each cell using the segmentation phase mask. Each ROI was applied a Gaussian blur (SD = 0.066 µm) and normalized using Eq. (2). Next, foci were detected by generating a binary image where only pixels above a speci ed intensity threshold were assigned a non-zero value. The putative foci were then ltered by area and eccentricity (see Supplementary Table 5 for parameters used). Homogeneously distributed protein mCherry or cells without a detected focus display a at distribution whereas the strongly clustered protein cI agg displays a strongly left-skewed distribution ( Supplementary Fig. 4a-f). To quantify the differences between the proteins measured, we calculated the fraction of pixels with a normalized intensity below a threshold value, I < 0.3, 0.5, and 0.7, for each cell (Supplementary Fig. 4g). These values were then normalized to the condensation coe cient distribution of mCherry as our experimental representative of a homogenous distribution.

Partition ratio analysis
Foci partition ratios were calculated from cells with a detected focus (see Condensation analysis) as the ratio of the average background-corrected pixel intensity within a focus to the average backgroundcorrected pixel intensity of the cells excluding the focus region. For cells with two or more detected foci, the average intensity of all foci was used for the ratio.

Focus dissolution analysis
Fluorescent foci were tracked during the reversibility experiment for images acquired over the course of bacterial cell division events or during cell elongation with cephalexin treatment as described above. We detected uorescent foci by the LoG method in TrackMate 53 using an estimated blob diameter of 600 nm and a quality threshold of 80 for cI agg , PopTag SL , PopTag LL , and 225 for McdB. Spot detections were subsequently linked to generate trajectories using the Simple Lap Track option with a maximum linking distance of three pixels (198 nm), a maximum gap-closing distance of 3 pixels (198 nm), and a maximum gap of 2 frames (30 min). The following trajectory lters were used: (i) only trajectories present in the rst frame were tracked in subsequent frames, (ii) trajectories of foci in cells that move out of the eld of view during the movie were excluded, (iii) false positive detections in the rst frame were removed, and (iv) trajectories of foci in cells that did not grow (cephalexin treatment) or divide (generational dilution) were excluded.

IbpA association analysis
Images of dual-color labeled cells expressing msGFP-IbpA and mCherry fusion proteins were analyzed by detecting uorescent foci in the mCherry channel using an intensity-based threshold of 30% of the maximum intensity in the image. mCherry fusion protein spot detections were subsequently ltered based on a minimum area threshold of 4 pixels and an eccentricity threshold of 0.75. The centroid of these spots was used to de ne the center of a 23 x 23 pixel ROI. This ROI was then used to analyze the IbpA channel; rst an intensity pro le was collected using the centroid of the ROI as the center point for both channels and subsequently normalized to the max intensity values in the mCherry channel. Next, the normalized ROIs for each channel were averaged to generate the projection images (See Fig. 5c). The resulting intensity patterns for both channels were then t with a 2D Gaussian to measure the full width at half max (FWHM) of the intensity. Finally, we calculated the ratio of FWHM IbpA /FWHM mCherry .    Bacterial cell growth and division dissolve condensates by dropping the cellular concentration below the apparent c sat . a. Generational-dilution dissolves condensates. The phase contrast channel (blue) and the mCherry channel (green) were merged and shown for indicated time points. White arrows demarcate the cellular location of the same focus over time. Blank arrows demarcate the same cellular position now absent of a focus. Images are representative of four biological replicates. Scale bar: 2 µm. b.
Quanti cation of focus intensity through cell generations. Colored solid lines and shading are the average and standard deviation, respectively, of the normalized total focus intensity for each strain indicated. The solid line color gradient over time indicates a decrease in the total number of foci detected at that indicated time point due to dissolution. cI agg : n = 427 foci; PopTag SL : n = 98 foci; PopTag LL : n = 81 foci; McdB: n = 48 foci. Dashed vertical lines indicate cell division events. c. Lifespan of protein foci during cell division. Lifespan of individual foci was determined by the particle trajectory length (see Methods). Data points correspond to individual foci. The curves next to the scatter plots are obtained via kernel density estimation. ***p < 0.001 by Welch's t test; n.s. indicates no statistically signi cant difference between the samples. d. Cell elongation dissolves phase-separated condensates. As in (a), except cells were treated with 10 µg/ml cephalexin to block cell division.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.