The bacterial nucleoid-associated proteins, HU and Dps, condense DNA into context-dependent biphasic or multiphasic complex coacervates

Abstract

The bacterial chromosome, known as its nucleoid, is an amorphous assemblage of globular nucleoprotein domains. It exists in a state of phase separation from the cell’s cytoplasm, as an irregularly-shaped, membrane-less, intracellular compartment. This state (the nature of which remains largely unknown) is maintained through bacterial generations ad infinitum. Here, we show that HU and Dps, two of the most abundant nucleoid-associated proteins (NAPs) of Escherichia coli, undergo spontaneous complex coacervation with different forms of DNA/RNA, both individually and in each other’s presence, to cause accretion and compaction of DNA/RNA into liquid-liquid phase separated condensates in vitro. Upon mixing with nucleic acids, HU-A and HU-B form (a) biphasic heterotypic mixed condensates in which HU-B helps to lower the C_sat of HU-A and also (b) multiphasic heterotypic condensates, with Dps, in which demixed domains display different contents of HU and Dps. We believe that these modes of complex coacervation that are seen in vitro can serve as models for the in vivo relationships among NAPs in nucleoids, involving local and global variations in the relative abundances of the different NAPs, especially in demixed subdomains that are characterized by differing grades of phase separation. Our results clearly demonstrate some quantitative, and some qualitative, differences in the coacervating abilities of different NAPs with DNA, potentially explaining (i) why E. coli has two isoforms of HU, and (ii) why changes in the abundances of HU and Dps facilitate the lag, logarithmic, and stationary phases of E. coli growth.

Biomolecular condensation of proteins and nucleic acids through liquid-liquid phase separation (LLPS) is thought to exert spatiotemporal control over complex intracellular biochemical processes and biological phenomena (1, 2, 3, 4, 5, 6, 7). LLPS is proposed to drive the formation of nonstoichiometric, highly dynamic, liquid-like, mesoscopic, subcellular, membrane-less compartments, both within the cytoplasm and in the nucleus of eukaryotic cells (8, 9), for example, in respect to stress granule formation (10), nucleolar organization (11), replication and transcription (12), and translational regulation (12). In prokaryotic cells too, recent exciting developments have established that critical cellular processes are governed by LLPS, for example, subcellular organization, chromosomal segregation, transcriptional and translation regulation (13, 14), cell division (15), and DNA protection (16). The self-assembly of DNA within mitochondria (eukaryotic cellular organelles with a presumed prokaryotic evolutionary origin) has also very recently been proposed to be LLPS-dependent (17).

Bacterial chromosomes contain DNA (∼80%), proteins (∼10%), and RNA (∼10%) (18). Their leitmotif is their ability to persist through successive rounds of replication and transcription spanning an infinite series of cell generations, while remaining in an extreme state of compaction and phase separation from the cell’s cytoplasm. This compaction facilitates the packing of double-stranded genomic DNA, which is a few millimeters in length, into a dynamic, functionally active entity known as the nucleoid, inside cells that are only about 1 μm in length (19). Such extreme compaction of DNA into the nucleoid (amounting to a ∼2000-fold decrease in persistence length) is believed to occur not merely due to DNA supercoiling, bending, or looping alone but also due to macromolecular crowding (20), potentially associated with LLPS (21). Measurements of the optical refractivity of nucleoids (22), the recovery of fluorescence shown by nucleoids after photobleaching (fluorescence recovery after photobleaching, FRAP) (23), and the physical compressibility of nucleoids (24) suggest that bacterial nucleoids exist in a constant state of LLPS. At the same time, the observed dynamics of bacterial chromosomal segregation (25) and the mechanical properties of bacterial chromosomes (24) suggest that nucleoids exist as chains of globular domains (25) that assemble into irregular shapes that change with time (26).

Clearly, there exists a need to understand both (a) what causes nucleoids to remain in a perpetual state of LLPS and also (b) why nucleoids are seen to exist as irregularly shaped objects, if indeed nucleoids do exist in a state of LLPS, as is currently considered possible. Therefore, our aims in this work were two-fold: (i) to examine whether nucleoid-associated proteins (NAPs) are capable of accreting and compacting DNA/RNA into micron-sized (or submicron-sized) globular entities in vitro, for example, through a complex coacervation-dependent phase-separation of NAPs with nucleic acids and (ii) to examine whether super-resolution microscopic visualization of a fluorescently labeled NAP within a nucleoid suggest the existence of submicron-sized domains assembled into different irregular shapes in different cells.

We began by proposing, based on first principles, that an NAP participating in the creation and/or maintenance of a state of phase separation of the entire length of DNA constituting a bacterial chromosome would need to possess the following properties: (a) be highly conserved across all bacteria, (b) be highly abundant within all bacteria, (c) be non-sequence-specific in the binding of nucleic acids, and (d) be characterized by the presence of some structural disorder, since disorder is understood to play a role in the tendencies of proteins to undergo phase separation (5, 9, 27). Further, we envisaged that (e) two types of NAPs could be required to package phase-separated DNA into a nucleoid: one with a basic pI (isoelectric point), capable of mostly locating within the nucleoid’s interior by being amenable to burial by DNA; and one with an acidic pI, mostly located on the nucleoid–cytoplasm interface, due to its being less amenable to burial by DNA (owing to its overall negative charge in the neutral pH of the bacterial cytoplasm). Consideration of all NAPs present in Escherichia coli suggested to us that two of these individually fulfill four out of the five criteria listed above and together fulfill all five criteria. The two NAPs are: (I) HU [a histone-like protein with a basic pI, first isolated from E. coli strain U93, (28, 29, 30) and constituted of two highly homologous isoforms, HU-A and HU-B, forming homodimers or heterodimers] and (II) Dps [a DNA-binding protein from starved cells, with an acidic pI (31), already known to be a dodecameric homolog of ferritin].

The structures of both HU and Dps contain intrinsically disordered regions (28, 31). The sequences and structures of HU and Dps are highly conserved across all bacteria (29, 32). Both HU and Dps are highly abundant, being among the most abundant of all proteins in bacteria (33). Both HU and Dps bind to multiple physical and chemical forms of nucleic acids (34, 35), with no established sequence-specificity (36, 37). Thus, based on entirely a priori considerations, HU and Dps would appear to be the most ideal, of all bacterial NAP candidates, to drive the formation and maintenance of a phase-separated state involving bacterial chromosomes. Thus, we resolved to examine the abilities of HU (HU-A and HU-B) and Dps to cause phase separation of DNA, keeping in mind that HU and Dps are also known to be interconnected in multiple ways, as mentioned below.

Dps and HU are regulated by an NAP called Fis (38). Dps is a component of HU’s genetic stress response regulon (35). Dps and HU-B display strong connections with stress in terms of cellular physiology. Dps is produced in copious amounts in cells subjected to stress. With HU, although the HU-A homodimer dominates the lag phase of E. coli growth under conditions of little or no stress, the HU-A/B heterodimer dominates the log phase of E. coli growth, while HU-B and Dps are mainly present in the stationary phase associated with starvation-related stress (30, 33, 39). Further, the gene encoding HU-B is regulated by three promoters, two of which are cold-responsive (40), with HU-B implicated in different stress response pathways (35). HU-A and HU-B also differ in their DNA-binding potency. The binding constants of HU-A homodimers, HU-A/HU-B heterodimers, and HU-B homodimers for cruciform DNA have been measured to be 25, 20, and 100 nM⁻¹, respectively, in the presence of 200 mM NaCl (41).

Below, we show that HU and Dps engage in LLPS with DNA to form biomolecular condensates in vitro and that HU can be used to visualize the organization of bacterial nucleoids in vivo into small globular domains. We present the results of investigations demonstrating that HU-A, HU-B, and Dps, individually and collectively, cause condensation of nucleic acids into globular phase-separated condensates, under conditions mimicking the temperature, pH, ionic strength, NAP concentrations, and nucleic acid concentrations of E. coli cells. We show that the three NAPs form complex coacervates that could potentially serve as model ex vivo micron-sized simulacrums of submicron-sized bacterial nucleoids.

Results

Submicron-sized globular entities in nucleoids can be visualized using RFP-HU-A

To directly observe the morphology of nucleoids inside live E. coli cells, we used structured illumination microscopy to image nucleoids in cells overexpressing a fluorescent DNA-binding HU protein construct, RFP-HU-A (previously described) (42, 43, 44). The upper panel of Figure 1A shows a differential interference contrast (DIC) image of a field of cells, with some cells present as filaments owing to their incomplete separation after division (a known tendency of E. coli that is amplified by doubling the levels of HU through expression from a plasmid) (43). The lower panel in Figure 1A shows a magnified DIC image emphasizing the linear nature of filaments. Figure 1B shows a super-resolution image of a single such filament made up of five cells, each containing a fluorescent RFP-HU-A–labeled nucleoid. Each nucleoid appears to consist of multiple (submicron-sized) bead-like entities that, furthermore, appear to be unique to each nucleoid/cell. A time-resolved series of images derived from the super-resolution videographic imaging of the RFP-HU-A–labeled nucleoid of a single (representative) cell present in a large field of such cells is shown in Figure 1C, with the entire video shown in Video S1.

Figure 1.

Images showing Escherichia coli cells and the submicron-sized domains in their nucleoids and the liquid-liquid phase separation behavior of the NAP, HU-B.A, DIC image of unseparated E. coli cells. B, super-resolution (SIM-derived) image of unseparated cells containing nucleoids characterized by submicron-sized domains of different shapes. C, time-series of images of a single nucleoid (extracted from the super-resolution SIM-based videography shown in Video S1), demonstrating that the submicron-sized domains are dynamic. D, image showing turbidity obtained through addition of 4WJ DNA to HU-B. E, confocal fluorescence image demonstrating that the turbidity shown in panel D owes to the formation of droplet-shaped condensates of Alexa 594–labeled HU-B with 4WJ DNA. F, single-frame from a fluorescence video (Video S2) demonstrating the fusion of droplet-shaped condensates of Alexa 594–labeled HU-B. G, evidence of dripping from the condensates of Alexa 594–labeled HU-B. H, evidence of surface wetting by the condensates of Alexa 594–labeled HU-B. I, averaged kinetics of fluorescence recovery after photo-bleaching (FRAP) of Alexa 488–labeled HU-B in droplets of HU-B and 4WJ DNA, with representative images shown for a single experiment. For panels E–I, the concentration of HU-B was 50 μM, and the concentration of 4WJ DNA was 3 μM. The main and outset (magnified) images in panel A have scale bar lengths of 5 μm. Panels C and I have scale bar lengths of 1 μm. 4WJ, 4-way junction; NAP, nucleoid-associated protein; SIM, structured illumination microscopy.

The nucleoid shown in the time series of images in Figure 1C appears to be made up of submicron-sized, globular fluorescent domains that rearrange over a time scale of a few tens of seconds. Such dynamic rearrangements of nucleoid domains have been predicted, or proposed, by others (25). In the present instance, our purpose was to validate their existence through visualization using fluorescent HU-A. Figure 1C and Video S1, which help to visualize the dynamicity of the domains, appear to indicate that the globular domains rearrange through separation and fusion, to the extent that this can be discerned, given the current limits of technology.

The sizes of the domains shown in Figure 1, B and C are more than an order of magnitude smaller than the sizes of (liquid condensate) droplets typically formed by proteins in vitro; however, submicron-sized, phase-separated condensates have already been observed in vivo (45), in particular, in respect of carboxysomes within bacteria (46). Therefore, the above in vivo data encouraged us to examine the possible role of HU and other NAPs in organizing nucleoids into globular domains, through inferences drawn from an in vitro examination of the abilities of such NAPs to undergo complex coacervation with DNA. Below, we describe our initial experiments with recombinant forms of two isoforms of the abundant NAP, HU, known as HU-A, and HU-B, to show that both isoforms cause both the compaction, and the phase separation, of DNA.

HU-A and HU-B trigger instantaneous accretion of DNA in vitro

The likelihood of HU-A and HU-B undergoing phase separation was initially assessed both through analyses of the distribution of net (linear) charge per residue (47) in the amino acid sequences of HU-A and HU-B and through computational prediction of sequence stretches that are likely to exist in disordered state, using the software, PONDR (48) (Fig. S1), since both are indicators of the likelihood of phase separation. Together with the above analyses, the structural disorder already known to exist within the DNA-binding loops in the three-dimensional structure of HU suggest that HU’s content of disorder could encourage phase separation. With a view to examining whether phase separation can be triggered by DNA, we added (unlabeled) HU-A or HU-B to salmon testis DNA (prelabeled with a fluorescent DNA-intercalating dye, called SYBR Green). The labeled DNA, which was initially homogenously dispersed in the entire solution, was observed to instantaneously accrete into compact, spheroidal entities, upon addition of either HU-A or HU-B. Fig. S2 shows a representative confocal image of entities formed through the addition of HU-B, along with a control image of the labeled DNA in the absence of HU-B. Video S2 shows representative fluorescence imaging of the initial stages of this accretion, which occurs over a time scale of seconds.

HU-B and DNA undergo complex coacervation into globular liquid condensates in vitro

We next examined the formation of phase-separated entities through complex coacervation of HU with a covalently-defined form of DNA: a synthetic 4-way junction (4WJ) that is already known to bind to HU (43), constituted of four defined oligonucleotides that associate to form a cruciform structure. We (and numerous others) have used 4WJ DNA previously, to study the binding of DNA by HU, as 4WJ DNA is one of the forms of DNA that displays high affinity to HU, owing to its contents of both curved and bent DNA (29). Our objective in using 4WJ DNA was to use a standard (covalently-defined) form of DNA that would allow us to compare results across different conditions of coacervation. Our objective in using several other additional forms of DNA/RNA (described later), in addition to 4WJ DNA and salmon testis DNA (already described), was to examine HU’s ability to accrete and phase separate DNA/RNA differing in sequence context, strand number(s), backbone chemistry, length, and three-dimensional organization.

As Figure 1D shows, no turbidity resulting from phase separation could be observed in a control HU-B solution in the absence of 4WJ DNA; however, instantaneous development of turbidity was observed, upon addition of 4WJ DNA (final concentration, 3 μM) in a solution of HU-B (final concentration, 50 μM) including the macromolecular crowding agent, polyethylene glycol 6000, or PEG 6000 (final concentration, 2%). Different modes of imaging, including bright-field imaging, differential interference contrast imaging, phase-contrast imaging, and wide-field fluorescence microscopic imaging (Fig. S3), as well as confocal microscopic imaging (Fig. 1E), were then used to establish that the observed turbidity owes to the formation of globular (spherical) condensates of HU-B and DNA. To eliminate the possible causation of any LLPS behavior by the 6xHis affinity tag present on the HU-B polypeptide, we confirmed that tag-lacking HU-B also undergoes phase separation under similar conditions (Fig. S4).

The liquid-like nature of the formed globular condensates was established through observations of condensate-condensate fusion (Fig. 1F and Video S3), condensate dripping (Fig. 1G), surface wetting by condensates (Fig. 1H), and recovery of fluorescence in condensates (FRAP) after photobleaching (Fig. 1I). Using methods described earlier (49), a diffusion coefficient of ∼0.0892 ± 0.0064 m² s⁻¹ was calculated for HU-B molecules inside such condensates, using FRAP data (showing rapid recovery of ∼100% fluorescence in <1 min). We then established that the above condensates also form in the absence of PEG 6000 (Fig. S5), for example, when a higher concentration of 4WJ DNA is used (≥5 μM, instead of only 3 μM), demonstrating that HU and DNA spontaneously undergo phase separation through complex coacervation, as is known to occur with other pairs of polyelectrolytes of opposite charge (50). The important point to be noted here is that phase separation is observed even when no external macromolecular-crowing agent (e.g., PEG) is used to cause an ‘excluded volume’ effect, that is, when the phase separation occurs solely through the mutual effects of the polyelectrolytes upon each other (through their mutual accretion of each other, tantamount to crowding). Having thus confirmed that PEG is not required for the complex coacervation of HU-B and DNA, we continued to use a nominal PEG concentration of 2% in all further experiments (unless otherwise mentioned), to limit the use of synthetic (relatively expensive) 4WJ DNA.

We then proceeded to confirm that just as HU-B and DNA form spherical condensates in the absence of PEG, HU-B also forms spherical condensates by itself, in the complete absence of DNA, when a higher concentration of PEG 6000 (i.e., 8%, instead of 2%) is used (Fig. S6). We would also like to point out here that the concentrations of 4WJ DNA used in the above experiments (3 μM or 5 μM) happen to be substantially lower than the concentration of 4WJ DNA that represents the base pair concentration of chromosomal DNA within an E. coli cell (60–120 μM), owing to the presence of ∼2.25 to ∼4.50 million base pairs of genomic DNA per cell (see calculations in the Experimental procedures section).

Intermolecular interactions driving the complex coacervation of HU-B and DNA

Having established the liquid-like nature of the spherical condensates formed by coacervating HU-B and 4WJ DNA, we next turned our attention to understanding the nature of the intermolecular interactions that are responsible for the observed phase separation, using turbidity as a metric of LLPS condensate formation, per established practice (7, 51). Turbidity data is shown below for all variations of physical and chemical conditions (Fig. 2, A–F). Microscopy data is shown for all variations of chemical conditions (Fig. S7).

Figure 2.

Development of turbidity owing to the formation of liquid-liquid phase-separated droplet-shaped condensates of HU-B, under varying physical or chemical conditions.A, dependence upon concentration of 4WJ DNA. B, dependence upon concentration of HU-B protein (in the presence and absence of DNA). C, dependence upon temperature. D, dependence upon the concentration of 1,6-hexanediol (HEX). E, dependence upon pH. F, dependence upon ionic strength (concentration of the salt, KCl). Unless one of the following parameters was specifically altered, in order to collect the data for any panel, the conditions used were as follows: 3 μM 4WJ DNA; 50 μM HU-B; 37 °C; 0% HEX; pH 7.4; 150 mM KCl; 50 mM Tris; 2% PEG (PEG-6000). 4WJ, 4-way junction.

Effect of varying DNA and protein concentrations

With a 50 μM (fixed) concentration of functional (dimeric) HU-B, turbidity rose progressively with increasing DNA concentration and saturated at ∼5 μM 4WJ (Fig. 2A). With a (fixed) 3 μM concentration of 4WJ DNA, turbidity rose progressively with increasing HU-B concentration, saturating above a concentration of ∼ 80 μM HU-B (Fig. 2B). A magnified section of Figure 2B is shown in Fig. S8, to show that the critical saturation value (C_sat) of HU-B is ∼2.5 μM, in the presence of 3 μM 4WJ DNA.

Effect of temperature

Higher temperatures resulted in higher turbidity (Fig. 2C), presumably owing to either higher molecular kinetic energies (and the consequent increase in protein-protein collisional frequencies) or enhancement of hydrophobic contributions to intramolecular and intermolecular protein interactions (since hydrophobic interactions improve with increasing temperatures; in particular, specific hydrophobic stacking interactions between beta sheets of monomers, in HU-B dimers (52), or between HU-B and nitrogenous bases of DNA exposed through binding of HU-B).

Effect of varying the presence of hydrophobic cosolvents

HEX (1,6-hexanediol) is known to disrupt weak hydrophobic interactions relevant to LLPS condensate formation (53). A very minor effect was observed with HEX, at unusually high concentrations (Fig. 2D), suggesting a minor contribution of weak hydrophobic interactions to the formation of condensates by HU-B and DNA. We have previously shown that hydrophobic interactions are relevant to HU only in respect of interactions between stacked β-sheets at the dimeric interface and interactions that are resistant to disruption by a hydrophobic cosolvent which is even more hydrophobic than HEX, namely dioxane (52).

Effects of electrostatic interactions

There is a rise in turbidity with lowering of pH (Fig. 2E), indicating the importance of the number(s) of positive charges upon HU-B in determining its interactions with negatively charged DNA. As pH is lowered from ∼10.0 (HU-B’s pI), greater numbers of positive charges are anticipated to be present upon HU-B, giving rise to greater scope for interaction with DNA, thus explaining the increase in turbidity. Turbidity also rises with the lowering of salt concentration (Fig. 2F), as would be expected, due to the increase in the interactions between HU-B’s positive charges and DNA’s negative charges, through lower levels of counter-ion–mediated dissociation of DNA from HU-B.

HU-A also undergoes complex coacervation with DNA and forms heterotypic condensates with HU-B

Alignment of the amino acid sequences of E. coli HU-A and HU-B (Fig. 3A) reveals that the two isoforms share a sequence identity of ∼70% (54). Alignment of the structures of the monomeric HU-A and HU-B chains (Fig. 3B) shows that the chains are superimposable with an RMSD of 0.491 Å. These two results establish that HU-A and HU-B are similar. It is reasonable to anticipate, therefore, that HU-A would also display the tendency to form liquid condensates with DNA, like HU-B. We found that HU-A does indeed form liquid condensates with DNA, under conditions (and using DNA concentrations) identical to those used for experiments involving HU-B. Intriguingly, however, the HU-A dimer concentrations required for such condensation to occur exceed ∼20 to 25 μM (Fig. 4G), in contrast to the value of ∼2.5 μM earlier observed with HU-B dimers (Fig. S8). This suggests that HU-A dimers are roughly ten times poorer at forming condensates with 3 μM 4WJ DNA than HU-B dimers. Further evidence for this is presented in the form of a turbidity development-based analysis of HU-A’s ability to undergo phase separation in the presence, and absence, of 3 μM 4WJ DNA (Fig. S9), similar to the analysis shown for HU-B in Figure 3B. A representative image of the condensates formed by the coacervation of HU-A (55 μM) and 4WJ DNA (3 μM) is also presented (Fig. 3C).

Figure 3.

Comparison of the sequences, structures, and phase separation behavior of HU-A and HU-B and demonstration of their ability to coacervate with each other.A, alignment of HU-A and HU-B amino acid sequences. B, alignment of HU-A (green; PDB ID 1MUL) and HU-B (magenta; PDB ID 4P3V) monomeric polypeptide chain structures. C, representative image of small droplet-shaped HU-A (55 μM) condensates with 4WJ DNA (3 μM). D, fluorescence recovery (kinetics/images) after photobleaching of Alexa 488–labeled HU-A in a droplet condensate, with representative images shown for a single experiment. E, phase diagram of HU-A coacervation with 4WJ DNA, as function(s) of varying salt and protein concentration. F, phase diagram of HU-B coacervation with 4WJ DNA, as function(s) of varying salt and protein concentration. G, complex coacervation of HU-A (green) and HU-B (red) with 4WJ DNA (3 μM). H, plots of the Pearson’s and Mander’s colocalization coefficients quantifying the presence of HU-A in the droplets of HU-B (n = 23). Panels C and G have scale bar lengths of 5 μm. Panel D has a scale bar length of 1 μm. 4WJ, 4-way junction.

Figure 4.

Understanding the differential complex coacervation of HU-A and HU-B with 4WJ DNA.A, turbidity arising from the mixing of 4WJ DNA with HU-A, HU-B, and mixtures of HU-A and HU-B, analyzed for triplicates using one-way ANOVA, with ∗∗∗ indicating p < 0.001 and ∗∗ indicating p < 0.01. B, turbidity arising from the mixing of 4WJ DNA with different concentrations of HU-A in the presence (yellow), and absence (green), of 10 μM HU-B. C, control microscopic images of HU-A (5 μM) and HU-B (5 μM). D, experimental microscopic images of the mixtures of HU-A (5 μM) and HU-B (5 μM), imaged individually using fluorescent HU-A or HU-B (with text in black denoting the protein that is not being imaged). E, size-exclusion chromatogram of HU-A (green) and HU-B (red). F, electrophoretic behavior of HU-A and HU-B on native PAGE (lanes as annotated). G, turbidity plots showing differences in the C_sat values of HU-A, HU-A E34K-V42L, and HU-B, with 3 μM 4WJ DNA. H, confocal microscopic images of condensates of different concentrations of HU-A, HU-A E34K-V42L, and HU-B, with 3 μM 4WJ DNA. Panels C, D, and H have scale bar lengths of 5 μm. 4WJ, 4-way junction.

As with HU-B, condensates of HU-A and DNA are spherical, liquid-like, and display rapid recovery of fluorescence in FRAP experiments (Fig. 3D). Phase diagrams presenting the effects of altering salt or protein concentration upon the turbidity of HU-A (Fig. 3E), or HU-B (Fig. 3F), allowed us to further confirm that HU-A is significantly poorer at coacervating with 4WJ DNA than HU-B. Importantly, HU-B and HU-A were also found to be capable of forming mixed heterotypic condensates with each other, and with DNA, when all three species are present together (Fig. 3G). It may be noted that the term ‘heterotypic’ is used here in the same sense in which it is used in the literature, for protein condensates (55). The Mander’s and Pearson’s coefficients of colocalization were calculated for HU-A and HU-B and plotted (Fig. 3H). The images and colocalization plots both show a near-complete colocalization of the two HU isoforms.

The results presented above suggest an interesting possibility, namely that the ratio of HU-A:HU-B in a particular region of genomic DNA could determine the extent of its phase separation, since the abilities of HU-A and HU-B to cause phase separation are clearly very significantly different. In experiments presented in Figure 4, A–D, we explored this possibility further. Figure 4A shows that although concentrations of HU-A, such as 12.5 μM or 20 μM, do not independently show any significant turbidity by themselves, in the presence of 3 μM 4WJ DNA, they add very significantly to the overall turbidity when there is some HU-B also present in the solution with the 4WJ DNA, for example, (1) when an equal concentration of HU-B is present, if the HU-A concentrations is 12.5 μM or (2) when a four times lower concentration of HU-B (5 μM) is present when the concentration of HU-A is higher (20 μM). In both cases, when HU-B is also present, it is very clear from the data presented in Figure 4A that HU-B increases HU-A’s ability to engage in condensate formation and that the obtained turbidity clearly results from the occurrence of some ‘emergent’ phenomenon, since this obtained turbidity is far greater than the sum of the turbidities individually derived from the interaction of 4WJ DNA with HU-A or HU-B.

Figure 4B further explores these mutual effects of HU-B and HU-A upon each other, through an examination of the effects of varying HU-A concentrations over a range of 0 to 40 μM, while keeping the concentration of HU-B fixed at 10 μM. The lower curve in Figure 4B shows a plot of turbidity as a function of the concentration of HU-A alone (in the absence of HU-B, but with 3 μM 4WJ DNA present). The upper curve in Figure 4B plots the turbidity seen in an identical set of solutions, when 10 μM HU-B is present. The upper curve starts with a high value of turbidity corresponding to the 0 μM HU-A concentration data point, due to the formation of condensates driven by the presence of 10 μM HU-B. Interestingly, the turbidity plotted in the upper curve also increases further as a function of increasing HU-A concentration, in a range of concentrations of HU-A (0–40 μM) in which the protein displays no turbidity by itself, in the presence of 4WJ DNA.

Figure 4, C and D capture the above results visually, using concentrations of the two HU-isoforms which are different from those used above but identical to each other and also close to the C_sat value of HU-B (while being far below the C_sat value of HU-A). In Figure 4C, it is seen that there is some condensate formation by 5 μM HU-B in the presence of 3 μM 4WJ DNA, but virtually no condensate formation by 5 μM HU-A. In contrast, in Figure 4D, it is seen that if both HU-A and HU-B are each present at 5 μM concentration, there is (a) a much greater degree of condensate formation and (b) a coincidence of signals from HU-A and HU-B in the very same (heterotypic) condensate entities, in all three figure panels (while imaging signals from HU-A and HU-B either separately or together). Overall, therefore, our data not only establishes that HU-B homodimers are significantly better at undergoing complex coacervation with DNA than HU-A homodimers but also that HU-A and HU-B appear to enhance each other’s abilities to form condensates with DNA (in mole-per-mole comparisons) through an emergent phenomenon, for example, through the well-known heterodimerization of HU-B and HU-A homodimers, to form HU-A/B heterodimers (56). In a separate section, we have shown that a simulacrum of the HU-A/B heterodimer, in the form of a fusion of HU-A and HU-B polypeptides (HU-A/HU-B), also undergoes phase separation.

HU’s multimericity appears to be correlated with the degree of condensate formation

It has been speculated, in respect of LLPS condensates, that a higher degree of multimericity translates into a greater valency of intermolecular interactions and that this further (positively) affects the formation of condensates (57). With the two isoforms of HU, that is, HU-A and HU-B, it has been reported that HU-A mainly exists in the form of dimers, whereas HU-B exists in the form of dimers, tetramers, and octamers (58). We reestablished that recombinant HU-B also displays higher multimericity than recombinant HU-A, in three different ways: (i) in the gel filtration (size exclusion) chromatography data shown in Figure 4E, we use the column’s calibration profile (Fig. S10) to demonstrate that HU-A elutes as a single (dimeric) peak, while HU-B elutes as a mixture of dimeric, tetrameric, and octameric peaks, although elutions at all volumes are observed to contain only HU-A or HU-B, and no other protein, using SDS-PAGE (Fig. S11); (ii) in the acidic native PAGE data shown in Figure 4F, we demonstrate that HU-A migrates as a single electrophoretic band, whereas HU-B migrates as a smeared band (consistent with its interconversion between multimeric states during electrophoresis); (iii) in the EMSA data presented in Fig. S12, A and B, respectively, we demonstrate that binding of HU-A to 4WJ DNA generates discrete mobility-shifted species, whereas binding of HU-B to 4WJ DNA generates both discrete mobility-shifted species, and DNA–protein complexes which are incapable of entering the gel (consistent with HU-B’s formation of larger complexes with DNA, owing to higher multimericity).

Having established that HU-B is characterized by a higher degree of multimericity, we proceeded to examine whether these differences in multimericity correlate with differences in the apparent C_sat values of the two isoforms (i.e., >∼25 μM for HU-A and ∼2.5 μM for HU-B), by generating an HU-A mutant incorporating two point mutations (E34K, V42L). This mutant is reported to form dimers and octamers, but no tetramers (58), unlike HU-A (only dimers) and also unlike HU-B (dimers, tetramers and octamers). Turbidity plots for HU-A, HU-B, and E34K:V42L HU-A are shown in Figure 4G, as functions of varying protein concentrations in the presence of a fixed concentration of 4WJ DNA. The figure shows that there is a dramatic shifting of C_sat value from HU-A (>∼25 μM) to the double mutant, E34K:V42L HU-A (∼10 μM), in the direction of the C_sat value of HU-B (∼2.5 μM). Confocal microscopic images shown in Figure 4H also confirm that the E34K:V42L HU-A's propensity to form spherical condensates is intermediate to those of HU-A and HU-B, suggesting that multimericity correlates with condensate forming ability with DNA.

It is known that a heterodimer of HU-A and HU-B dominates the log phase growth of E. coli cultures (59). We examined the condensate-forming behavior of HU-B/HU-A heterodimers, by examining the behavior of a proxy, or simulacrum, of such a heterodimer, which was created through the genetic fusion of HU-B and HU-A. This fusion construct has been described by us earlier, in a different context (60). The concentration-dependent turbidity plot of this HU-B/HU-A simulacrum (Fig. S13A) shows that it also forms condensates with DNA under conditions and concentrations similar to those used to examine condensate formation by HU-B (Fig. S13B). This establishes that HU in all of its homo-dimeric and hetero-dimeric forms engages in the formation of liquid droplet condensates with DNA.

Dps (the stationary phase NAP) also undergoes complex coacervation, to form condensates with DNA and multiphasic heterotypic condensates with DNA and both HU isoforms

Dps, which is acidic and dodecameric, is a major NAP of the stationary phase of E. coli and other bacteria. It interacts with DNA through a basic N-terminal tail present in each monomer. Dps also contains some IDRs or intrinsically disordered regions (Fig. S14A) and some charge clusters (Fig. S14B) that are present throughout its amino acid sequence (61, 62). Fascinatingly, the enhancement of compaction of the nucleoid during the stationary phase of growth of E. coli is correlated with the increased expression of Dps, which is concomitant with a reduced expression of HU-A and HU-B (24, 63). Stress is known to induce phase separation of Dps into both crystalline and liquid-like crystalline structures in vivo (62, 64, 65, 66). Dps is also known to undergo solid-liquid phase separation into crystals in vitro (62, 64). The concentration of Dps monomers/cell rises to 180,000 (∼15,000 dodecamers) (63), translating into a bacterial cell volume–dependent dodecameric concentration of ∼18.75 to ∼50.0 μM. It is not known whether Dps passes through an LLPS phase, in the earliest parts of the onset of the stationary phase (when it presumably contends with the simultaneous presence of HU, prior to the formation of liquid crystalline and crystalline phases by Dps, in association with DNA, once the presence of the HU isoforms has significantly reduced). Thus, we decided to examine the ability of Dps to form LLPS states and also the compatibility of such states with the LLPS states formed by HU’s isoforms.

We tested the ability of 30 μM Dps to form spherical condensates, in the absence, as well as in the presence, of DNA and in the absence or presence of HU-A and/or HU-B. Fascinatingly, as shown in Figure 5A, Dps displayed phase separation with Salmon testis DNA and formed condensates. These condensates were smaller than those formed by the isoforms of HU, and there was evidence of the formation of both spherical and nonspherical entities. It is worth noting here that nonspherical condensates of DNA have also been observed previously by others (67). The nonspherical entities appear to have formed through the irregular adherence of spherical condensates of different sizes that were unable to fuse in as facile a manner as most HU condensates, presumably owing to a higher interfacial tension on the surfaces of Dps condensates, than on the surfaces of HU condensates, and also due to the consequent, or associated, differences in the shear relaxation characteristics of the two types of condensates. Notably, however, Dps condensates did display rapid FRAP, as shown in Figure 5B, indicating that the bulk of the population was in a liquid state, without having transitioned into a state with more solid-like (liquid crystalline or crystalline) characteristics. Within both the noncanonical (i.e., nonspherical) condensates and the spherical condensates, confocal microscopic imaging confirmed colocalization of Dps protein and Salmon testis DNA (Fig. 5C).

Figure 5.

Phase separation behavior of Dps with salmon testis/4WJ DNA, in the absence/presence of HU-B.A, representative confocal microscopic image of associating spherical condensates of Dps with salmon testis DNA. B, fluorescence recovery (kinetics/images) after photobleaching of Alexa 594–labeled Dps in spherical condensates of Dps with salmon testis DNA. C, assemblages of associating spherical condensates of Alexa 594-labeled Dps (red) and SYBR Green–labeled salmon testis DNA (green). D, Pearson’s correlation coefficient (PCC) and Mander’s (M1/M2) colocalization coefficients of Dps with salmon testis DNA, with M1 representing the fraction of Dps in droplets positive for Salmon testis DNA and M2 representing the fraction of Salmon testis DNA in droplets positive for Dps (n = 13). E, complex coacervation of Alexa 594–labeled Dps (red) and Alexa 488–labeled HU-B (green), with unlabeled 4WJ DNA. F, Pearson’s correlation coefficient (PCC) and Mander’s (M1/M2) colocalization coefficients of Dps with HU-B, with M1 representing the fraction of Dps in droplets positive for HU-B and M2 being the fraction of HU-B in droplets positive for Dps (n = 12). G, enlarged (boxed; rectangular) section of panel E showing Dps-enriched condensates (red/orange) enclosed in demixed form within HU-B–enriched condensates (green). H, a three-dimensional confocal image of condensates of Dps, HU-B, and 4 WJ DNA, similar to panel E. I, complex coacervates of Dps, HU-B, and 4WJ DNA, imaged using Alexa 594–labeled Dps (red), FAM-labeled DNA (green), and Alexa 647–labeled HU-B (blue). Panels A–C have scale bar lengths of 1 μm, while all other panels have scale bar lengths of 5 μm. 4WJ, 4-way junction.

Equally fascinatingly, for a protein like Dps which displays a ready tendency to undergo crystallization, Dps forms LLPS condensates in the presence of 8% PEG 6000 (Fig. S14C), that is, in the absence of any DNA. More fascinatingly, upon addition of 4WJ DNA to solutions containing Dps and an isoform of HU, Dps was mostly found to be colocalized in small condensates that were enclosed within larger condensates that were rich in HU-B (Fig. 5D) or HU-A (Fig. S15), that is, into irregularly-shaped, Dps-rich condensates incorporated within HU-rich condensates, in multiple numbers of such condensates-within-condensates. This result could be attributed to primary differences in the characteristics of HU and Dps condensates, in that both HU and Dps appear to act as scaffolds when they act alone, but Dps appears to be additionally capable of acting as a client in the presence of HU, at least at the relative concentrations used. This possibility may be viewed in light of the speculation that certain proteins can act as client proteins in scaffolds prepared by other proteins, during phase separation (68). What is most remarkable about these complex coacervates of HU-B, 4WJ DNA, and Dps is the clustering of Dps-rich condensates within larger HU-rich condensates, indicative of the formation of phases of different grades of liquidity that do not mix easily (Fig. 5D). A magnified image (see the boxed region in Fig. 5D) of one such group of condensates-within-condensates (Fig. 5E) and a three-dimensional cross section of a field containing many such condensates-within-condensates (Fig. 5F) are also shown.

Interestingly, when HU-B and Dps were subjected to phase separation in the absence of DNA, through the inclusion of 8% PEG, uniform mixing was observed (Fig. S16), giving rise to spherical condensates that contained no distinct regions that could be said to be distinguishably rich in either HU-B or Dps. This suggests that the irregular shapes of both (a) the Dps-DNA condensates and (b) the Dps-dominated condensates enclosed within larger spherical HU-B–dominated condensates probably owe to the manner in which Dps binds to DNA and then either organizes DNA or becomes organized by DNA. As shown in Figure 5G, which shows a confocal image for labeled 4WJ DNA (in addition to images for HU-B and Dps, in complex coacervates of HU-B, Dps, and 4WJ DNA), there is a greater intensity of fluorescence from the DNA present within the Dps-rich condensates-within-condensates than from the DNA present within the HU-rich regions of the larger condensates. This also suggests that Dps either organizes DNA differently (enhancing the quantum yield of fluorescence from the labeled 4WJ DNA) or that Dps causes a greater increase in the physical concentration of the 4WJ DNA (enhancing the observed fluorescence from DNA in these regions). Further, it is clear from Figure 5G that DNA also either organizes Dps differently or causes Dps to be more closely packed, because the brightness of HU-B appears to be more uniform (or similar) in HU-dominated and Dps-dominated regions, in Figure 5G, whereas the brightness of Dps appears to be higher in Dps-dominated regions. Thus, HU-B, DNA, and Dps are all present in all regions of all mixed condensates. The enrichment of Dps in Dps-dominated regions is not reflected in a parallel enrichment of HU-B in HU-B–dominated regions. Rather, the highest brightness for all three species, that is, HU-B, DNA, and Dps, is seen in the Dps-dominated regions.

Our summary view of these varied complex coacervates is that HU and Dps, two major NAPs present in stationary-phase E. coli, are capable of coexisting within the same condensates, just as the two isoforms of HU are capable of such coexistence, albeit in different modes of presentation, respectively, as multiphasic, or biphasic, condensates. The possibility of HU-B and DNA forming a different phase within nucleoid interiors from the phases formed by HU-A and DNA at nucleoid surfaces has also not escaped our notice. The three protein species seem to offer a near-continuum of extents of phase separation and compatibility with each other.

HU-B undergoes LLPS with multiple forms of nucleic acids and DNA polymerase

We explored the ability of HU-B to undergo condensation/coacervation with different forms of DNA/RNA, using experiments simulating the situation inside a nucleoid where different forms of DNA and RNA, and different proteins like HU-B, HU-A, and Dps, exist together with other NAPs and DNA-binding proteins, for example, DNA polymerase. Figure 6 shows independent and merged images of differentially labeled forms of HU-B with different combinations of HU-A, Dps, 4WJ DNA, nicked DNA, dsDNA, and ssDNA, showing the colocalization of multiple and varied species into the same condensates. For the images shown in Figure 6, A–E, Pearson’s and Mander’s coefficients of colocalization are observed to be between 0.6 and 1.0 (Figs. S17–S21). In addition to the above, we also found that HU-B undergoes phase separation with Poly-U RNA (Fig. S22). Notably, knockout of a noncoding HU-binding RNA has been reported to lead to decompaction of the nucleoid (69).

Figure 6.

Complex coacervation of HU-B (labeled by Alexa 594 or Alexa 488) with multiple different forms of DNA (labeled by DAPI, SYBR Green, or 6-fluorescein amidite) and with other NAPs, such as Dps (labeled by Alexa 594) or HU-A (labeled by Alexa 488).A, condensates of HU-B (red) and dsDNA (green). B, condensates of HU-B (red), 4WJ DNA (blue), and nicked DNA (green). C, condensates of HU-B (red), 4WJ DNA (blue), and ds DNA (green). D, condensates of HU-B (red), 4WJ DNA (blue), and ssDNA (green). E, condensates of HU-B (red), 4WJ DNA (blue), and HU-A (green). All panels have scale bar lengths of 10 μm. 4WJ, 4-way junction; NAP, nucleoid-associated protein.

Lastly, we examined whether DNA polymerase can exist within condensates of HU and DNA, through an experiment in which fluorescently labeled Pfu DNA polymerase was mixed with condensates of DNA and Venus-HU-B, a construct that has been described previously (43). DNA polymerase was observed to colocalize with HU-DNA condensates (Fig. S23). This indicates that the HU-mediated phase separation of genomic DNA is not incompatible with the presence of other DNA-binding protein species within LLPS condensates.

Discussion

The idea that the nucleoid is phase-separated from the cytoplasm in bacteria has existed for decades (22). In recent years, evidence has accumulated for the LLPS of various proteins in the company of nucleic acids, for example, HP1 in heterochromatin (70), NPM1 in the nucleolus (71), H1 in nucleosomes (72), and TFAM in mitochondrial nucleoids (17). In bacterial nucleoids, HU and Dps are the most abundant of all NAPs (and also two of the most abundant proteins in E. coli) (33). Recently, there has been some speculation about the possibility that HU and/or Dps could be somehow involved in the formation of a phase-separated state by the nucleoid (73); however, there has neither been any particular rationale presented, in support of such speculation, nor any experimental evidence to indicate such a possibility. With our demonstration of the complex coacervation of HU-A, HU-B, and Dps, with DNA, we now firmly establish both (i) that these NAPs cause the accretion and compaction of DNA into LLPS condensates in ways that are qualitatively, or quantitatively, different, while also being mutually compatible and capable of coexisting within the same condensates and also (ii) that HU-A and HU-B are much more compatible with each other than either one of these HU isoforms is with Dps, since the simultaneous presence of Dps, DNA, and HU-A, or HU-B, leads to the formation of demixed LLPS subdomains that are enriched in Dps.

One important general implication of the demonstration of the complex coacervation of NAPs and DNA is that this highlights how LLPS can effectively function to cause, and maintain, the accretion and compaction of DNA within the cell. The genomic DNA of E. coli (which is 2 mm in length) remains ever-compacted into a tiny volume (∼0.5–3.0 fl), in the form of the nucleoid which has a maximum dimension, in any direction, of a few microns. Within this tiny volume, or upon the surface of the nucleoid which changes shape and constitution constantly due to the occurrence of replication, transcription-coupled translation continues to occur at all times while cells grow in the presence of nutrients. For decades, the view has existed that DNA supercoiling, together with the binding of NAPs, must mediate the bending, looping, bridging, and bunching, of DNA, that is anticipated to be required to facilitate the packaging of so much DNA into such a tiny volume (19). What our work demonstrates is that the mutual induction of the formation of LLPS condensates by NAPs and DNA could also play an equally crucial role (along with the mechanisms mentioned above) to cause, and to help maintain, the accretion and compaction of DNA into tiny volumes. Notably, it has been reported in the past that a double-knockout of the genes, hupA and hupB, encoding HU-A and HU-B, respectively, leads to decompaction of the nucleoid (74, 75, 76). What we propose is that the decompaction of the nucleoid in the absence of HU-A and HU-B occurs not merely because DNA is no longer being bent, looped, bridged, or bunched (since these are activities are also performed by one or more of the other NAPs, including IHF-A, IHF-B, H-NS, Hfq, Fis, and StpA) but also because DNA is no longer being accreted and compacted sufficiently through HU-mediated phase separation.

It is our view that the abundant NAPs mainly help to neutralize DNA’s negative charges, thus annulling the mutually repulsive interactions between the phosphate groups of DNA that otherwise prevent the compaction of DNA into such tiny volumes (with DNA, also reciprocally performing the same service, i.e., charge neutralization, for basic NAPs, like HU-A or HU-B and for acidic NAPs containing basic DNA-binding motifs/domains, like Dps). It is likely that the effects of the neutralization of DNA’s negative charges by abundant, conserved, disorder-containing proteins (that is, the very triggering of mutual self-crowding behavior through charge-neutralization), are not unlike the phase separation behavior that is commonly shown by pairs of oppositely charged polyelectrolytes. DNA is known to undergo LLPS through mutual-crowding in the presence of a positively charged polyelectrolyte such as poly-L lysine (50), which can be considered to be a proxy for NAPs. Similarly, positively charged proteins, which can be considered to be a proxy for NAPs, are known to undergo phase separation through mutual-crowding in the presence of a negatively charge polyelectrolyte such as polyphosphate (77). Thus, HU and DNA would appear to be together driven into LLPS states in vitro through mutual-crowding caused by (i) the accretion of HU molecules by DNA (through binding and charge-neutralization) and (ii) the accretion of DNA by HU molecules (also through binding and charge-neutralization), in the absence of any other crowding agent, in addition to (iii) the accretion of DNA through bending, looping, bridging, and bunching by HU. The supercoiling of DNA could be presumed to further aid the accretion of DNA in vivo.

At the same time, what is probably special about NAPs like HU-A, HU-B, and Dps is their own peculiar ways of binding to DNA to cause mutual- and self-crowding. Our work clearly demonstrates that HU-A and HU-B, the two isoforms of HU in E. coli, both form spherical heterotypic condensates with dsDNA as well as with other forms of nucleic acids, independently and with each other, under physiological conditions of temperature, pH, and ionic strength and at physiological HU concentrations, using subphysiological concentrations of nucleic acids. What is very significant, however, is the fact that there is a >10-fold difference in the abilities of HU-A (C_sat ∼ 25 μM) and HU-B (C_sat ∼ 2.5 μM) to trigger such phase separation in ∼3 μM 4WJ DNA, with higher concentrations of HU-A (in comparison with HU-B) being required to cause phase separation, with all else being equal. The quantitatively different behaviors of HU-A and HU-B could conceivably translate into lower degrees/grades of phase separation by DNA in nucleoids within cells in (i), the early logarithmic phase of growth, that is, when E. coli cells are typically dominated by homodimers of HU-A and when nucleoids are required to remain in a constant state of replication and gene expression (i.e., transcription-coupled translation), during growth under conditions of nutritional excess (whether in laboratory cultures or inside the human intestine).

In contrast, during (ii), the late logarithmic phase of growth, when heterodimers of HU-A/HU-B, together with homodimers of HU-B, dominate E. coli cells and during (iii), the stationary phase of growth, when Dps dominates E. coli cells, and when nutrients are limiting or when there is starvation-induced (or other) stress, the fact (A) that DNA is required to become, and remain, more compact under such conditions appears to be clearly correlated with (B) the demonstrated greater abilities of HU-B and Dps (than HU-A) to cause phase separation at lower concentrations and, therefore, at earlier timepoints during the cell generation and for longer fractions of an ever-lengthening cell generation time.

The “irregular” shapes of nucleoids in vivo that are visible in the super-resolution microscopy data, as well as the irregular (nonspherical) shapes of condensates in vitro that are seen in Dps-DNA condensates and also occasionally in HU-DNA condensates, could result from the fact that multiple NAP-mediated interactions like bridging, and bunching, of DNA could potentially also influence shapes, leading to deviations from sphericality (78). Further, the co-occurrence of LLPS and polymer-polymer phase separation would be expected to give rise a viscoelastic physical nature of condensates (78, 79). This would permit the occurrence of dynamic changes in the nucleoid, such as those observed in the super-resolution microscopic data, without any requirement for the nucleoid in vivo, or NAP-DNA condensates in vitro, to be completely liquid-like and ‘spherical’.

Further, the ability of Dps to form enriched subdomains within condensates of HU and DNA could be conceived to arise from differences in the interaction of DNA with Dps and HU. High local concentrations of Dps are known to cause DNA to form toroidal structures that subsequently arrange into crystals (80). The formation of separate Dps-enriched subdomains could signal the beginnings of the formation of such toroidal structures within condensates. Therefore, the differential behaviors of HU-A, HU-B, and Dps, in terms of the quantitative and qualitative differences in their engagements with DNA to form LLPS condensates, offer interesting insights into why multiple NAPs exist to cause the accretion and compaction of DNA and also into the differential roles played by such NAPs in actively dividing quiescent cells and in cells that are transitioning between these two states.

It is our view that HU and Dps constitute a ‘minimalistic’ set of proteins for phase separation of the bacterial chromosome into a nucleoid, both during the normal growth of cells, when HU dominates the field of NAPs (and Dps is present at only one-twentieth of its concentration in stationary phase cells) (63), and also during the entry of cells into the stationary phase, when Dps dominates the field of NAPs (and the presence of HU is reduced to a minimum). Since HU has a basic pI, it is understandable how HU could potentially also organize DNA into LLPS states within nucleoid interiors, where HU presumably remains buried by bound DNA until such DNA becomes exposed to the cytoplasm during replication. On the other hand, since Dps has an acidic pI (and binds to DNA through a basic N-terminal tail), it could potentially organize DNA into LLPS states, at least initially, at the nucleoid-cytoplasmic interface, since charge-charge repulsions with DNA would prevent Dps from becoming completely buried. However, at later timepoints, and following the conversion of the chromosome into a crystalline state made up of Dps-induced toroids of DNA appearing to fill up the entire bacterial cell, it is quite possible that this crystalline state becomes suffused with the aqueous environment of the cytoplasm, like any crystal that undergoes solid-liquid phase separation. It would be interesting to understand how Dps eventually facilitates the generation of this solid-liquid phase-separated state from the initial LLPS state that it participates in the formation of, along with HU.

Experimental procedures

Nucleic acids

Synthetic oligonucleotides

All oligonucleotides (oligos) were purchased from Sigma-Merck and were resuspended in Tris buffer at pH 7.5.

Fragmented genomic DNA

Salmon testis DNA was purchased from Sigma-Merck (Cat no. D-9156).

Poly(U) RNA

Poly(U) RNA was purchased from Sigma-Merck (Cat no. P9428).

ssDNA

The oligo, 5′-TGATCATGCATCGTTCCACTGTGTCCGCGACATCTACGTC-FAM - 3′ was used as ssDNA.

dsDNA

dsDNA of a length of 20 base pairs was prepared by annealing two oligos, 5′-GTTCAATTGTTGTTAACTTG-3′ and 5′-CAAGTTAACAACAATTGAAC-3′.

Nicked dsDNA

Nicked dsDNA was created by annealing the long oligo 5′- GACGTAGATGTCGCGGACACAGTGGAACGATGCATGATCAGCAAGGACGATCCTGTCTTGGTGGTAAGGGTGCGC-3′ to two short oligos, 5′-TGATCATGCATCGTTCCACTGTGTCCGCGACATCTACGTC-FAM-3′ and 5′-GCGCACCCTTACCACCAAGACAGGATCGTCCTTGC-3′.

4WJ DNA

4WJ DNA was generated through the annealing of four different oligos in equimolar amounts to create a Holliday junction–like (cruciform) structure. The four oligos used were as follows:

5′- CCCTATAACCCCTGCATTGAATTCCAGTCTGATAA 3′,

5′-GTAGTCGTGATAGGTGCAGGGGTTATAGGG-3′,

5′-AACAGTAGCTCTTATTCGAGCTCGCGCCCTATCACGACTA-3′,

5′-TTTATCAGACTGGAATTCAAGCGCGAGCTCGAATAAGAGCTACTGT-3’.

DNA fluorescent labeling

4WJ DNA and dsDNA were fluorescently labeled with DAPI and SYBR green, respectively; free fluorophores were removed through repeated ultrafiltration and washing.

Construction of genes encoding WT or mutant HU/Dps

Strategies for the cloning of fusion constructs involving the fluorescent proteins RFP, or Venus, placed at the N-termini of HU isoforms (with the intent of creating either RFP HU-A, or Venus HU-B) have already been described (43). In the same descriptions of work, strategies for the cloning of an F47W mutant of HU-B and an F79W mutant of HU-A have also been described, together with the creation of a fusion of HU-B and HU-A (HU Simulacrum) incorporating an 11 amino acids-long flexible GS linker placed between the two proteins. We replaced the WT HU-B and HU-A in the HU simulacrum with the said F47W mutant of HU-B and the F79W mutant of HU-A, respectively, to create F47W-HU-B and F79W-HU-A, using standard recombinant DNA techniques. In further developments of mutants, we used the genetic background of the F47W mutation in HU-B, or the F79W mutation in HU-A, to carry out the requisite site-directed mutagenesis, using splicing by overlap extension PCR-based strategies entirely identical to those described earlier (43), as well as mutation-specific primers designed to create HU-B or HU-A genes encoding the following mutations: (i) an N-terminal cysteine in HU-B, (ii) an S35C mutation in HU-B, (iii) an S17C mutation in HU-A, and (iv) an E34K-V42L double mutation in HU-A already containing the F79W mutation. Separately, an S22C mutation was created in Dps which already contains two tryptophan residues and does not require the introduction of a tryptophan (W) residue as a spectroscopic reporter (unlike HU-B or HU-A). The gene encoding Dps was cloned by amplification from the E. coli genome through a PCR reaction in which the forward primer contained an Nde I restriction site, and the reverse primer contained an Xho I restriction site, with the genomic DNA template sourced through lysing of whole E. coli cells by heat (98 °C, 5 min) immediately prior to 35 cycles of PCR (Denaturation: 95 °C, 30 s; Annealing: 55 °C, 45 s; Extension: 72 °C, 60 s) using the Go-Taq Flexi (Promega) thermostable DNA polymerase. The amplicon was extracted from the agarose gel, digested, and ligated between Nde I and Xho I sites of the pET-23a vector, by T4 DNA ligase. DNA sequences of all clones of HU and Dps were confirmed through sequencing by dideoxy chain termination, using a commercial service (Agrigenome).

Protein expression and purification

HU-A and HU-B

Genes encoding HU-A and HU-B, and mutants bearing 6xHis affinity tags at their N-termini, were cloned into the pQE-30 vector (between the Bam HI and Hind III restriction sites) for expression in either the XL-1-Blue strain or the M15 (pRep4-carrying) strain of E. coli. HU-B and its mutants were expressed in the XL1-Blue strain, whereas variants of HU-A were expressed in the M15 strain. Cultures were grown for 8 h, and cells were harvested through pelleting by centrifugation at 9000g. Pelleted cells were resuspended in PBS supplemented with 1 M NaCl, to facilitate separation of HU from DNA (and, consequently, also DNA-bound proteins) during subsequent purification. The resuspended cells were lysed using a sonicator (Qsonica/Misonix). Removal of cell debris was done through centrifugation at 15,000g for 1 h. Supernatants were subjected to Ni-NTA immobilized metal affinity chromatography (IMAC), using standard conditions (Qiagen). Protein purified through Ni-NTA chromatography was subjected to further (polishing-step) purification, using cation exchange chromatography performed on either a HiFliQ S-type (Protein Ark) column or a Bio-Rad (Econo) column packed with Mono-S resin (GE). The purified protein was then dialyzed against 50 mM Tris buffer of pH 7.4, containing 150 mM KCl, concentrated to ∼3 mg/ml, partitioned into aliquots, flash-frozen in liquid nitrogen, and stored at −80 °C.

Dps

Dps and its cysteine-containing variants were cloned into the pET-23a vector between the Nde I and Xho I restriction sites, for expression in the BL21 (DE3) pLysS∗ strain of E. coli, through induction by IPTG (1 mM) at a culture O.D of 1.0, followed by 24 h of growth of culture(s). The method for purification of C-terminally 6xHis-tagged Dps through Ni-NTA (IMAC) chromatography was identical to that used for HU. Since high purity was obtained using IMAC, Dps and its variants were not subjected to ion-exchange chromatography, as was required with HU-B or HU-A.

Estimation of HU/Dps concentrations

Protein concentrations were estimated using the 280 nm absorption of tryptophan and estimated extinction coefficients. Native HU-A and HU-B do not contain the residue tryptophan. Since a concentration-dependent phenomenon like phase separation cannot be studied with precision using tryptophan-lacking proteins (since the estimation of concentration proves to be difficult), we first verified, using roughly estimated concentrations, that WT HU-B and HU-A display LLPS behavior. Thereafter, we performed standard experiments only with the F47W mutant of HU-B, or the F79W mutant of HU-A, with an accurate estimation of protein concentrations based on the tryptophan residue present in these mutants, which have previously been used and shown to be identical to WT HU protein in respect to DNA binding (43).

Electrophoretic mobility shift assay

Protein samples of varying concentrations were mixed with 4WJ DNA to yield a final DNA concentration of 1 μM and loaded onto a 0.5% agarose gel in TAE buffer, using a constant voltage of 60 V, for 1 h. Protein-bound DNA and free DNA on the gel were then stained with ethidium bromide and visualized through UV-induced fluorescence.

Phase separation assays

First standard physical/chemical conditions, and reagent concentrations, were established, as described in the next six subsections. In the seventh subsection, a description is provided of the ranges over which each of these physical/chemical conditions, or reagent concentrations, was varied, while maintaining all the others at a fixed value.

• (1)pH and ionic strength. For examining the ability of coacervates of protein and DNA to form LLPS condensates, a standard pH of 7.4 (50 mM Tris buffer), and a standard salt concentration of 150 mM KCl, were chosen, with the intent of mimicking the pH and ionic strength of the bacterial cytosol. A temperature of 37 °C was chosen, with the intent of mimicking the temperature at which a culture of a bacterium such as E. coli grows optimally.
• (2)HU concentration. A standard HU dimer concentration of 50 μM (i.e., HU monomer concentration of 100 μM) was chosen, with the intent of mimicking an HU concentration of ∼1 mg/ml for LLPS experiments, corresponding to 60,000 HU monomers per femtoliter, since the average physiological concentration of HU varies from 30,000 to 60,000 monomers per cell and since cells display variation of volume from 0.5 to 3.0 fl.
• (3)Dps concentration. A standard Dps dodecamer concentration of 30 μM (i.e., Dps monomer concentration of 360 μM) was chosen to mimic the cellular concentration of Dps (12.5 μM for a cell of 2 fl volume and varying between 9.38 μM and 50 μM for cell volume with volumes varying from 0.5 to 3.0 fl; assuming 180,000 Dps monomers per cell).
• (4)4WJ DNA concentration. A cruciform-shaped DNA Holliday 4WJ (made up of ∼110 base pairs) concentration of 3 μM (∼0.15 mg/ml) was chosen to create a DNA concentration of 112,500 base pairs per femtoliter. This DNA concentration happens to under-represent, by a factor of 20 to 40, the average physiological concentration of ∼2.25 to ∼4.50 million base pairs per femtoliter (corresponding to one half-genome or a full genome per femtoliter).
• (5)ssDNA and plasmid DNA concentrations. Single-stranded salmon testis DNA with an average length of ∼1200 to 1600 bases, and linearized plasmid DNA with a length of 4.2 kilobase pairs, were used at a concentration of 0.15 mg/ml.
• (6)Crowding agent concentration. After confirming that LLPS behavior is also seen in the absence of a crowding agent, a standard (nominal) PEG-6000 concentration of 2% (w/v) was chosen to under-represent concentrations of this crowding agent (5–20%) which are commonly used to simulate cellular conditions and the crowding of macromolecules in the cell cytoplasm.
• (7)Variations of conditions and concentrations. These were employed to map the ranges of conditions under which HU-B forms LLPS condensates, by varying one condition at a time (e.g., temperature, pH, or ionic strength) while maintaining all other standardized conditions (described above). A single readout of turbidity was measured after confirming fully that the turbidity reflects the formation of LLPS condensates. To examine the effect of varying chemical conditions, for example, pH or ionic strength, turbidity measurements were made immediately after mixing of protein with DNA, using a wavelength of 600 nm, on a 96-well plate (Eppendorf cat no. 0030730011) in a plate reader (BMG Labtech POLARstar Omega). For examination of the effects of varying physical conditions, for example, temperature, turbidity plots were based on measurements taken after the completion of two periods: (i) an initial incubation period of protein, and nucleic acid, for 5 min, separately, that is, prior to mixing, to allow the establishment of thermal and other (e.g., conformational) equilibria and (ii) a subsequent incubation of 5 min after mixing of protein and DNA, to allow an LLPS equilibrium to be established, in addition to thermal and chemical equilibria. For turbidity plots of pH-dependence, and for microscopic imaging of the effect of mixing, proteins were subjected to buffer exchange in order to transfer them into various buffer systems using repeated washing and centrifugal-ultrafiltration in Merck centrifugal filter units (cat no. UFC500396). Solutions of different pH were created using citrate buffer (pH 4), Tris buffer (pH 6 and pH 8), and CAPS buffer (pH 10). Buffers of different pH all contained 150 mM KCl, by way of salt.

Acidic native PAGE

All proteins were loaded at a concentration of 1.2 mg/ml on acidic native gels (15% acrylamide) that were cast and run according to the protocol mentioned at http://ubio.bioinfo.cnio.es/data/crystal/local_info/protocols_old/ANATPAGE.html. Gels were run at 60 V on ice in a Bio-Rad Mini-Protean Tetra vertical electrophoresis setup. Gels were run for approximately 3 h and then stained using Coomassie brilliant blue staining solution.

Fluorescence labeling

Cysteine mutants were labeled using Alexa Fluor 594 C₅ Maleimide, excited by 594 nm light, or Alexa Fluor 488 C₅ Maleimide, excited by 488 nm light, or Fluorescein-5-Maleimide, also excited by 488 nm light. Hundred micromolar of protein was incubated with 300 μM of TCEP (Tris(2-carboxyethyl) phosphine hydrochloride, from Merck) and 200 μM of dye. Free dye was removed by ultrafiltration using Merck centrifugal filter units (cat no. UFC500396). Labeling efficiencies were calculated according to the protocols of manufacturers using a scanning Cary 50 Bio UV-Vis spectrophotometer.

Fluorescence microscopy

Samples for microscopy were mixed on the bench and visualized immediately under either on a Leica widefield THUNDER imaging system or on a Zeiss LSM 980 system, for all samples involving DNA or RNA. In general, in all experiments in which protein (HU-B, HU-A, or Dps) was being visualized, the unlabeled form of the protein (99%) was spiked with a fluorescently labeled form of the protein (1%). Makeshift chamber slides were made using double-sided tape, to prevent the crushing of condensates by coverslips. All images were analyzed using Fiji ImageJ. For experiments involving Poly(U) RNA, the Zeiss LSM 980 was used, and 500 nM of Alexa Fluor 594 C₅ Maleimide-labeled HU-B was spiked into the mixture, before the addition of RNA, for imaging of the labeled HU-B protein engaging in condensate formation, with associated turbidity, due to the addition of the Poly(U) RNA, as the RNA itself was not labeled for this experiment. Brightfield, DIC, and phase-contrast images were all collected on an Olympus IX-83 widefield fluorescence microscope.

Super-resolution microscopy

Cells were aliquoted from E. coli cultures, in the log phase of growth, overexpressing RFP-HU-A and then air-dried upon a slide for visualization using the 564 nm laser line of a super-resolution microscope (Zeiss Elyra 7) based on Lattice SIM technology, with use of the Plan Apo 63×/1.40 oil objective, and an sCMOS camera (PCO Edge).

FRAP assays

FRAP assays were carried out on a Zeiss LSM 980 microscope.

HU-B

HU-B (50 μM) was spiked with 1% of Alexa 488–labeled HU-B mutant (containing an N-terminal Cys conjugated with Alexa Fluor 488 C₅ Maleimide from Thermo Fisher Scientific; Cat no. A10254). The 488 nm laser was set at 100% for efficient bleaching and focused on a small circular region with a diameter of 1 μm. Iterations of the incident laser were set at 150. Fluorescence intensities were measured for 100 cycles. Data was collected in triplicates and plotted for the same after normalization.

HU-A

FRAP of HU-A was performed exactly as for HU-B, with the spiking of HU-A with 1% labeled HU-A (i.e., the S17C mutant of HU-A labeled by the Alexa Fluor 488 C₅ Maleimide fluorophore), with a different protein identity (HU-A), different concentration (55 μM), and different laser focus diameter (0.5 μm).

Dps

FRAP of Dps was performed exactly as for HU-B, with the spiking of Dps with 1% labeled Dps (i.e., the S22C mutant of Dps labeled by the Alexa Fluor 594 C₅ Maleimide), with a different protein identity (Dps), different protein concentration (30 μM), different laser line (594 nm, instead of 488 nm), and the same laser focus diameter as used for HU-A (0.5 μm).

Analytical size-exclusion chromatography to determine multimericity

Size-exclusion chromatography was performed on a Superdex-75 Increase column (GE), on an AKTA Purifier 10 workstation (GE), with equilibration of the column with Tris (50 mM) buffer of pH 7.4 containing 150 mM KCl, at room temperature. All protein loading was carried out using 500 μl protein of 110 μM concentration. Monitoring of the elution volume(s) of protein(s) was done through the measurement of the 280 nm UV absorption of the eluent.

HU-induced DNA accretion experiment using salmon testis DNA

Salmon testis DNA (stock concentration of 10 mg/ml; Sigma/Merck Cat no. D-9156) was labeled with SYBR Green dye (Stock concentrations 10000X) by adding dye (final concentration of 1 μl/10 ml in solution with DNA) to DNA (∼100 μl), following which free SYBR Green dye was removed through repeated washing and centrifugal-ultrafiltration in Merck centrifugal filter units (cat no. UFC500396). The resulting labeled DNA (20 μl volume; concentration not determined) was pipetted onto a cover slip and visualized on an inverted Nikon Eclipse Ti-u microscope, as a uniform field of green fluorescence. Following this, with continuing video-based visualization of this drop of 20 μl, using the microscope’s camera, 20 μl of HU-B (100 μM) protein was added to the DNA already present (and being visualized) on the cover slip, and the mixed drop containing HU-B and salmon testis DNA was visualized over time.

Examination of DNA polymerase’s ability to associate with DNA in HU-B condensates

Pfu DNA polymerase (purified from a bacterial clone expressing the polymerase previously produced in our laboratory) was labeled, using Alexa Fluor 594 C₅ Maleimide. The labeled DNA polymerase was then added to Venus-HU-B using standard recommended buffer conditions, and condensates formed after the addition of 4WJ DNA (with associated solution turbidification) were imaged using confocal microscopy, as for other experiments, to examine whether the labeled DNA polymerase colocalized with the HU-B and the DNA, within condensates.

Colocalization image analysis

Images were opened in ImageJ. Max entropy thresholding was applied. The JaCOP plugin was used to calculate the Pearson's correlation coefficient and the Mander’s overlap (MI and M2) coefficient.

Statistical analysis

All experiments were done thrice. The data was then plotted as mean ± SD. Scattered data points from independent experiments were also plotted. The statistical significance analysis was performed by carrying out one-way ANOVAanalysis of variance tests, using the software Origin Pro 2021b. All data analysis, data fitting, and data plotting were performed using Origin Pro 2021b.

Data availability

The Article and Supplementary Information contain all the data for this manuscript.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare no conflict of interests.

Reviewed by members of the JBC Editorial Board. Edited by Patrick Sung