Single-molecule tracking reveals that the nucleoid-associated protein HU plays a dual-role in maintaining proper nucleoid volume through differential interactions with chromosomal DNA

Summary

HU is the most conserved nucleoid-associated protein in eubacteria, but how it impacts global chromosome organization is poorly understood. Using single molecule tracking, we demonstrate that HU exhibits nonspecific, weak, and transitory interactions with the chromosomal DNA. These interactions are largely mediated by three conserved, surface-exposed lysine residues (triK), which were previously shown to be responsible for nonspecific binding to DNA. The loss of these weak, transitory interactions in a HUα(triKA) mutant results in an over-condensed and mis-segregated nucleoid. Mutating a conserved proline residue (P63A) in the HUα subunit, deleting the HUβ subunit, or deleting nucleoid-associated naRNAs, each previously implicated in HU’s high-affinity binding to kinked or cruciform DNA, leads to less dramatically altered interacting dynamics of HU compared to the HUα(triKA) mutant, but highly expanded nucleoids. Our results suggest HU plays a dual role in maintaining proper nucleoid volume through its differential interactions with chromosomal DNA. On one hand, HU compacts the nucleoid through specific DNA structure-binding interactions. On the other hand, it decondenses the nucleoid through many nonspecific, weak, and transitory interactions with the bulk chromosome. Such dynamic interactions may contribute to the viscoelastic properties and fluidity of the bacterial nucleoid to facilitate proper chromosome functions.

Graphical Abstract

The most conserved chromosome-associated protein in bacteria, HU, has two roles in maintaining a proper chromosome volume: (1) condensation, mediated by HU’s high affinity binding to looped or bent DNA, and (2) expansion, mediated by the sum of multiple weak and transitory interactions of HU with linear DNA.

Introduction

The E. coli chromosome of 4.6 million base pairs must condense over a thousand-fold to be packed into a cell of a few micrometers in size. At the same time, the chromosome must also be “loose” enough to allow DNA replication, chromosome segregation, and transcription to occur without hindrance. This properly compacted chromosome, together with its associated proteins and RNAs, is termed the nucleoid. The precise molecular mechanism of how the nucleoid volume is maintained is not well understood, but several key factors have been identified. A group of proteins known as nucleoid-associated proteins (NAPs), such as H-NS, IHF, Fis, and HU have been shown to play important roles (reviewed in (Browning et al., 2010)). Among these NAPs, HU is the most conserved across eubacteria and one of the most abundant in E. coli ( ~30,000 copies per cell during middle exponential phase growth (Claret and Rouviere-Yaniv, 1997; Ali Azam et al., 1999)). E. coli HU has two subunits α and β, ~9 kDa each and mainly exists as dimers, as monomers of HU are not detectable (Claret and Rouviere-Yaniv, 1997; Ali Azam et al., 1999). HUα₂ homodimers are predominant in the early log growth phase, whereas HUαβ heterodimers are mainly in the late log and stationary phase. HUβ₂ homodimers are negligible under both log and stationary phase growth (Claret and Rouviere-Yaniv, 1997). HU impacts numerous cellular processes including the regulation of replication (Dixon and Kornberg, 1984; Bonnefoy and Rouviere-Yaniv, 1992), transcription (Prieto et al., 2012) and translation (Balandina et al., 2001), and has been implicated in pathogenesis (Kar et al., 2005), likely through its role in organizing the chromosome structure.

Despite its importance, how HU mediates nucleoid organization has remained elusive. In vitro data demonstrates HU binds specifically with high affinity to distorted DNAs such as kinked or cruciform DNAs that are recombination intermediates (Kamashev and Rouviere-Yaniv, 2000), but for bulk chromosomal double-stranded DNA (dsDNA) the binding is weak and nonspecific (Pinson et al., 1999). A crystal structure of HU with nicked or artificially bent dsDNA revealed that the high-affinity binding of HU was stabilized through intercalation of a conserved proline residue from the beta branch “arm” of HU. This structure, together with the fact that the gal repression loop requires HU’s binding to a bent DNA segment, led to the theory that HU induces and or stabilizes bends in distorted dsDNA (Aki and Adhya, 1996; Aki and Adhya, 1997). More recently, HU was crystallized with unbent dsDNA, which was bound to a patch of positively charged lysine residues on the side of the HU dimers. These different binding modes suggest that HU may play multiple roles in organizing the nucleoid depending on its interacting surface. Another model, based on the compatible interfaces of HUα₂ and HUαβ dimers found in these crystal structures, suggests that HU compacts the nucleoid through enhanced DNA bundling by oligomerizing on dsDNA (Hammel et al., 2016). However, HU oligomerization has yet to be observed in vivo.

Most recently, a study demonstrated that HU binds to small RNAs (Balandina et al., 2001; Balandina et al., 2002; Macvanin et al., 2012), and that in the presence of HU, post-processed small non-coding RNAs (renamed as nucleoid-associated RNAs, naRNAs) from repetitive extragenic palindromic (REP) elements (Macvanin et al., 2012; Qian et al., 2015; Qian et al., 2017) condensed plasmids in vitro and promoted chromosomal contacts between various REP elements in vivo (Qian et al., 2015). These observations lead to a catalytic chaperone model, in which HU deposits naRNAs onto cruciform DNA structures to bridge DNA-DNA contacts and mediates chromosome compaction (Qian et al., 2017). Because of these seemingly disparate observations of HU behavior, a better understanding of how this abundant NAP mediates chromosome organization in vivo is needed.

In this work, we examined the in vivo interaction dynamics of HU with the chromosome using single-molecule tracking (SMT) under a set of wild-type (WT) and altered conditions. Our results suggest that HU may play a dual role in both decondensing and compacting the nucleoid through its differential interactions with the chromosomal DNA and specific DNA structures.

Results

HU-PAmCherry exhibited two rapidly switching diffusive states

To probe the dynamics of HU in the nucleoid, we used an E. coli strain in which the chromosomal copy of hupA gene encoding for the α subunit of HU was replaced by a 3’ tagged photoactivatable fluorescent protein fusion gene hupA-PAmCherry (Wang et al., 2014). We verified that the fusion protein HUα-PAmCherry was expressed at the expected full length (Figure S1A) and supported WT-like growth as the sole cellular source of HUα (Figures S1B and S1C). Furthermore, HUα-PAmCherry supported HU-dependent mini-P1 plasmid replication and Mu phage growth in the ΔhupB background at levels indistinguishable from those of the WT parental strain MG1655 (Macvanin et al., 2012; Qian et al., 2015) (Figures S1D and S1E). These results suggested HUα-PAmCherry could replace the endogenous HUα for its function.

To perform single-molecule tracking on HUα-PAmCherry, we imaged mid log-phase live E. coli cells grown in EZ rich defined media (EZRDM) at room temperature (24 °C) on agarose gel pads. We activated individual HUα-PAmCherry molecules using a low level of 405 nm UV light and tracked their cellular locations over time with a frame rate of ~150 Hz (Δt = 6.74ms) with laser excitation at 568 nm. This imaging rate allowed us to determine the diffusion coefficients of individual molecules up to 3 μm²/s with high confidence (Supplemental Note 1). In Figure 1A, we showed a few representative trajectories superimposed with the corresponding bright-field image of the cell. Individual trajectories displayed different diffusive properties: some were confined within small regions (blue trajectories), some traversed larger cell areas (brown trajectories), and some switched in between (rainbow trajectories). Plotting the apparent diffusion coefficients calculated from single-step displacements of all molecules suggested that a heterogeneous distribution of HUα-PAmCherry molecules with a wide range of diffusion coefficients indeed existed (Figure S2A).

Figure 1: Single molecule tracking (SMT) of HUα-PAmCherry in live E. coli cells grown in rich media showed two diffusive states, rapid transition kinetics, and nucleoid-like localizations.

(A) Representative SMT trajectories superimposed on top of the corresponding brightfield image of the cell (gray). Blue trajectories represented State I and red State II. Molecules transitioned between the two states were colored in rainbow. Scale bar = 1 μm. (B) Two diffusive states of HUα-PAmCherry molecules with respective population percentages (size of the pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (C) Mean squared displacement (MSD) plots of State I (light blue) and State II (dark blue) HUα-PAmCherry trajectories as a function of time; the plateau signified confined diffusion of molecules in both States. (D) Two-dimensional (2D) histogram of all cellular HUα-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for HUα-PAmCherry (total 59,432 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary unit) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images).

To analyze these different populations quantitatively, we used a Bayesian-based hidden Markov Model (HMM) (Persson et al., 2013), which determines the optimal number of diffusive states represented in the data, the diffusion coefficients of those states, the percentage of molecules in each state, and the apparent transition rates between states. We found that the data fit best to a two-state model (Figure 1B, Table 1), with diffusion coefficients of State I D₁ = 0.14 ± 0.004 μm²/s, and State II D₂ = 0.39 ± 0.006 μm²/s, (μ ± s.e.m., n = 60,432 displacements). The percentages of the HU molecules in State I and II were ~45% and ~55% respectively. Using several other classic methods (Vrljic et al., 2007), we obtained the same two states with essentially identical diffusion coefficients and population percentages (Figures S2A, S3A, and S4A; Table S3; Supplemental Note 1). The two distinct diffusive states likely represent the bound and unbound states of HUα-PAmCherry to the chromosome (Kamashev and Rouviere-Yaniv, 2000; Balandina et al., 2001; Lal et al., 2016), which we discuss in the next section. Furthermore, using HMM, we found that HUα-PAmCherry molecules switched between the slow diffusing State I and the fast diffusing State II with apparent rates of k₁₂ = 10.9 s⁻¹ and k₂₁ = 14.9 s⁻¹ respectively (Figure 1B). The dwell time of States I and II were at 100 ± 15 ms and 74 ± 7 ms, respectively (Table 1). The fast switching kinetics of HU-PAmCherry between the two states indicate that these interactions were rapid and transitory.

Table 1:

Diffusion parameters extracted from single-molecule tracking trajectories of HUα-PAmCherry under different experimental conditions using the Hidden Markov Model (HMM). All values expressed as μ ± s.e.m.

	State I			State II
	D1 (μm2/s)	Occupation (%)	Dwell time (ms)	D2 (μm2/s)	Occupation (%)	Dwell time (ms)	n (displacements)
HUα-PAmCherry	0.14 ± 0.004	45 ± 2	100 ± 15	0.38 ± 0.006	55 ± 2	74 ± 7	60,432
HUα(triKA)-PAmCherry	0.18 ± 0.01	5 ± 0.2	107 ± 23	1.08 ± 0.04	95 ± 0.2	852 ± 142	193,678
HUα(P63A)-PAmCherry	0.25 ± 0.01	32 ± 1	129 ± 15	0.84 ± 0.01	68 ± 1	114 ± 8	90,406
HUα-PAmCherry + chlor	0.13 ± 0.003	40 ± 1	80 ± 7	0.35 ± 0.003	60 ± 1	70 ±4	108,846
HUα-PAmCherry + rif	0.14 ± 0.003	22 ± 2	80 ± 8	0.39 ± 0.001	78 ± 2	100 ± 7	137,019
HUα-PAmCherry + ΔREP325	0.13 ± 0.004	30 ± 1	72 ± 9	0.36 ± 0.003	70 ± 1	96 ± 9	43,413
HUα-PAmCherry + ΔhupB	0.24 ± 0.01	38 ± 2	168 ± 43	0.88 ± 0.02	62 ± 2	155 ± 29	22,449
HUα(triKA)-PAmCherry + ΔhupB	0.41 ± 0.01	42 ± 0.8	196 ± 23	2.42 ± 0.03	58 ± 0.8	84 ± 6	58,867
HUα(P63A)-PAmCherry + ΔhupB	0.29 ± 0.02	30 ± 1	100 ± 8	1.02 ± 0.05	70 ± 1	110 ± 6	132,631

The two diffusive states reflected HU-PAmCherry’s transient and nonspecific binding with chromosomal DNA

Based on the values of the apparent diffusion coefficients, we hypothesized that the slow diffusive State I (D₁ = 0.14 ± 0.004 μm²/s) likely represented HUα-PAmCherry molecules bound to DNA, whereas the fast diffusive State II molecules (D₂ = 0.39 ± 0.006 μm²/s) represented the unbound population diffusing within the nucleoid. Note that the dwell times of both states (~ 100 ms and ~ 75 ms for State I and II respectively) were significantly shorter than what would be expected for the specific binding of most DNA binding proteins (Jen-Jacobson et al., 2000) (Table 1), or the known nanomolar affinity of HU to nicked, cruciform, and kinked DNA structures (Kamashev and Rouviere-Yaniv, 2000; Balandina et al., 2002), but were consistent with the weak binding nature of HU to bulk DNA (Pinson et al., 1999). Supporting the possibility that State I represents HU non-specifically bound to DNA, we found that the mean-squared displacement (MSD) curve of State I molecules plateaued at a level significantly lower than that of State II molecules at relatively long time scales (> 0.1 s, Figure 1C), suggesting that State I molecules experienced more restricted diffusion than State II molecules (Figure 1C). Using a confined diffusion model assuming a finite circular boundary (Supplemental Note 2), we estimated the diameter of the confinement zone of single State I molecules at ~230 nm. The size of the confinement zone suggests that State I HU dimers were unable to move significantly afar, consistent with being bound to a large, relatively immobile molecule such as DNA. However, it is also significantly larger than that determined from stationary HUα-PAmCherry molecules bound to DNA in chemically fixed cells (Figure S5, ~110 nm, Supplemental Note 2). This observation can be explained through two possibilities: (1) that State I HUα-PAmCherry molecules did not remain bound to DNA during the ~100 ms dwell time but “hopped” on adjacent sequences transiently, or (2) that they remained bound, but the bound chromosomal DNA had intrinsic motions on the same time scale in live cells.

To differentiate these two possibilities, we tracked the diffusion of a chromosomal DNA segment labeled with six tetO operator sites (tetO⁶) tightly bound by TetR-mCherry fusion protein molecules under the same cell growth and imaging conditions as HUα-PAmCherry (Figure S6A). Note that compared to other fluorescent repressor-operator systems (FROS) using tandem arrays of hundreds of DNA binding sites (Lau et al., 2003), the small footprint of the tetO⁶ site (~200 bp) enabled us to pinpoint the position of the labeled chromosomal DNA with high accuracy and negligible perturbations (Hensel et al., 2013). We observed that TetR-PAmCherry formed distinct fluorescent spots in cells (Figure S6A), consistent with its specific DNA binding property. In contrast, the cellular distribution of HUα-PAmCherry mimicked that of the nucleoid (Figure 1D), indicating that the binding sites of HU are distributed throughout the chromosome, consistent with the distribution of HU observed in C. crescentus (Wang, et al., 2011). The average apparent diffusion coefficient D_app of TetR-PAmCherry molecules (0.09 ± 0.04 μm²/s, μ ± s.e., n = 1808 displacements, Figure S6B) and the corresponding confinement zone (~200 nm, Figure S6C) were nearly identical to those of State I HUα-PAmCherry molecules. Note that the D_app of ~ 0.09 was also comparable to what was previously reported on labeled E. coli chromosomal foci (~ 0.1 μm²/s) (Javer et al., 2013). Together, these results suggest that State I HUα-PAmCherry molecules remained DNA-bound, even though the binding was short-lived, and that the chromosomal DNA segments were intrinsically dynamic.

For State II HUα-PAmCherry molecules, the D_app at ~0.4 μm²/s was one order of magnitude less than what would be expected from a freely diffusing, non-DNA-interacting molecule of similar size in the E. coli cytoplasm (1 to 10 μm²/s) (Elowitz et al., 1999), but comparable to those of the “unbound” population of many other DNA-binding proteins in E. coli, such as LacI (Elf et al., 2007), RNA polymerase (Stracy et al., 2015), and UvrA (Stracy et al., 2016). Furthermore, the apparent confinement zone of State II HUα-PAmCherry molecules was ~ 300 nm, smaller than the size of the nucleoid, which we estimated at ~550 nm (FWHM) along the short axis of cell by superresolution structured illumination microscopy (SIM) (Gustafsson, 2000) imaging under the same growth condition (Figure 1D and S7). These results suggest that when dissociated from DNA, a HUα-PAmCherry molecule likely did not diffuse far before it encountered another nonspecific binding site, which was effectively any other DNA sequence nearby. These frequent interactions are not resolvable under our imaging time resolution (frame rate of 150 Hz, Δt = 6.74ms) and hence the observed apparent diffusion coefficient reflects the overall, averaged nonspecific interactions of HU with chromosomal DNA. Less frequent interactions would speed up the diffusion and lead to a larger apparent diffusion coefficient and vice versa. The slowed diffusion of HU and other DNA binding proteins in nucleoid due to frequent, transitory DNA interactions is reminiscent of random diffusion of molecules in liquid condensates, suggesting that the nucleoid may “trap” DNA-binding proteins using these types of non-specific interactions.

Using the apparent on-rate (k₂₁ from fast State II to slow State I) and off-rate (k₁₂ from slow State I to fast State II) measured from SMT and an average concentration of nonspecific chromosomal DNA binding sites at ~ 5 mM under our growth condition (i.e. average 2 chromosomes per cell, with each chromosome having ~4 × 10⁶ nucleotides, representing a total number of ~ 8 × 10⁶ nonspecific binding sites in a total volume of ~ 2 fL), we estimated that the in vivo non-specific binding affinity K_d of HUα-PAmCherry to DNA was ~ 4 mM and the association rate constant was ~2 × 10³ M⁻¹s⁻¹ (Supplemental Table S6). Both constants were two to three orders of magnitude lower than other specific DNA or RNA binding proteins (Dayton et al., 1984; Fei et al., 2015). Taken together, these results suggest that HU primarily interacts with chromosomal DNA in a weak and transitory manner (Sarkar et al., 2007; Dame et al., 2013).

Surface lysine and proline residues have distinct roles in HU’s chromosomal DNA interaction dynamics and binding affinities

Previous crystallographic studies identified sets of residues on HU that are important to its DNA interactions (Figure S8). Three positively-charged lysine residues on the side surface of HU were shown to form electrostatic interactions with the negatively-charged backbone of DNA to stabilize its nonspecific binding (Hammel et al., 2016). A nearly universally conserved proline residue (P63) located in the β-branch arm of each HU subunit is proposed to intercalate into the minor groove of DNA to stabilize specific DNA conformations such as loops, bends and supercoils (Aki and Adhya, 1997; Swinger et al., 2003; Becker et al., 2007; Guo and Adhya, 2007; Hammel et al., 2016) (Figure S8). To examine how these residues contributed to the transitory binding dynamics of HUα-PAmCherry, we constructed two mutant fusions, one with the three lysines on the surface of HUα mutated to alanines (K3A K18A K83A, referred to as triKA) and the other with the conserved proline residue mutated to alanine (P63A). We then replaced the chromosomal hupA gene with the hupA(triKA)-PAmCherry or hupA(P63A)- PAmcherry fusion gene and investigated the corresponding binding dynamics of the mutant fusions.

We first performed SMT on HUα(triKA)-PAmCherry using the same growth and imaging conditions as WT HUα-PAmCherry cells. We found that HUα(triKA)-PAmCherry molecules displayed dramatically altered dynamics as compared to WT HU-PAmCherry (Figure 2A and B). Nearly all HUα(triKA)-PAmCherry molecules (~ 95% ) exhibited fast, random Brownian-like diffusion (D_app = 1.08 μm²/s, n = 193,678 displacements from 61 cells, Figure 2A and B, dark gray), with a very minor population (~ 5%) diffusing at an intermediate D_app = 0.24 μm²/s (Figure 2A and B, light gray, Table 1). The mutant State II diffusion coefficient and confinement zone are significantly larger than those of the State II unbound population in the WT condition, suggesting that HUα(triKA)-PAmCherry largely lost its nonspecific interactions with chromosomal DNA. Furthermore, while the apparent off-rate (k₁₂) was negligibly affected, the apparent on-rate (k₂₁) was reduced ~ 10-fold compared to that of WT HUα-PAmCherry, resulting in a ~ 10-fold higher apparent in vivo K_d (Figure 2A, Table 1). These observations demonstrate a near-complete loss in the ability of HUα(triKA) to bind DNA, and thus HUα(triKA) diffuses freely throughout the nucleoid with minimal interactions with the chromosome. Correspondingly, the cellular distribution of the fast diffusing population of HUα(triKA)-PAmCherry expanded significantly toward the cytoplasm compared to that of WT HUα-PAmCherry and no longer colocalized with the nucleoid (Figure 2C). Thus, the electrostatic interactions between the triple lysine residues on the HUα surface and the negatively charged backbone of chromosomal DNA appear to be responsible for HU’s weak interactions with nonspecific DNA.

Figure 2: The triKA and P63A mutations affected HU dynamics differentially.

(A) Two diffusive states of HUα(triKA)-PAmCherry with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (B) Mean squared displacement (MSD) plots of State I (light gray) and State II (dark gray) HUα(triKA)-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (C) Two-dimensional (2D) histogram of all cellular HUα(triKA)-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for HUα(triKA)-PAmCherry (total 193,678 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images). (D) Two diffusive states of HUα(P63A)-PAmCherry with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (E) Mean squared displacement (MSD) plots of State I (light orange) and State II (dark orange) HUα(triKA)-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (F) Two-dimensional (2D) histogram of all cellular HUα(P63A)-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for HUα(triKA)-PAmCherry (total 90,406 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary unit) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images).

Additionally, we observed detrimental defects of HUα(triKA) on cell growth and nucleoid morphology. Cells expressing HUα(triKA)-PAmCherry had significantly wider distributions in cell length and width and were on average longer and skinnier than WT cells (Figure S9A, left). Most interestingly, the nucleoid of HUα(triKA) mutant cells appeared to be significantly more compact compared to that of WT HU cells (Table S7), and the chromosomal DNA distribution in the mutant strain was no longer symmetric, representing a nucleoid segregation defect (Figure S7). These results suggest that the dynamic interactions of HU with DNA mediated by the triple lysine residues on HUα’s surface, although weak and transitory, play a significant role in avoiding an over-condensed nucleoid volume, which is important for proper chromosome segregation.

Next, we performed SMT on HUα(P63A)-PAmCherry under the same growth and imaging conditions as WT. HUα(P63A)-PAmCherry also exhibited two populations with apparent diffusion coefficients similar to HUα(triKA)-PAmCherry (D_app¹ = 0.25 ± 0.007 μm²/s and D_app² = 0.88 ± 0.008 μm²/s respectively, Figure 2D, Table 1, n = 90,406 displacements from 60 cells). However, HUα(P63A)-PAmCherry had a larger slow diffusing population ( ~ 30%), and the apparent on-rate (k₂₁ = 9.3 s⁻¹) was closer to that of WT cells compared to HUα(triKA)-PAmCherry. In contrast to the expanded cellular distribution of HUα(triKA)-PAmCherry, the cellular distribution of HUα(P63A)-PAmCherry remained nucleoid-centered (Fig. 2F) and the apparent K_d of HUα(P63A)-PAmCherry remained similar to that of WT HUα-PAmCherry (Supplemental Table S6), suggesting that HUα(P63A)-PAmCherry retained its nonspecific binding to DNA. Further, while HUα(P63A)-PAmCherry expressing cells exhibited relatively normal growth rates and cell length distributions (Figures S1B and S9), their nucleoids expanded significantly compared to that of WT cells, opposite of what was observed in HUα(triKA)-PAmCherry expressing cells, and both nucleoid morphology and segregation appeared minimally affected (Figure S7, Table S7).

Taken together, the increased diffusion of the fast State II for both triKA and P63A compared to WT indicates that P63-mediated DNA interactions (likely on bent, cruciform, or altered DNA structures such as that produced during replication, transcription, and recombination (Dixon and Kornberg, 1984; Bonnefoy and Rouviere-Yaniv, 1992; Aki et al., 1996; Kamashev and Rouviere-Yaniv, 2000)) and triK-mediated DNA interactions (likely on bulk linear DNA) both contribute to the transitory interactions of HU with the chromosome. Their differential effects on nucleoid compaction and apparent DNA binding affinity, however, suggest that binding of HU to specific DNA structures may be important for nucleoid compaction while the nonspecific binding of HU to linear DNA may be important for avoiding an overly compacted nucleoid, which would impede DNA segregation.

HUαβ heterodimer exhibits different dynamics from HUα₂ homodimer and may play an important role in compacting the nucleoid

As stated before, HU exists as a mixture of both HUα₂ homodimers and HUαβ heterodimers in exponentially growing WT cells, with HUβ₂ dimers nearly undetectable (Claret and Rouviere-Yaniv, 1997). As cells enter the stationary phase, HUα₂ recedes and HUαβ becomes the predominant form. It was suggested that although both HUα₂ and HUαβ may function equally to organize the nucleoid, the heterodimer may play an essential role in protecting DNA from radiation or starvation-mediated DNA damages, possibly through an increased affinity to bind specialized DNA structures (Kamashev and Rouviere-Yaniv, 2000). To better understand the potentially disparate roles of the two HU dimers, we performed SMT on HUα-PAmCherry molecules in a ΔhupB background, in which HUα could only form homodimers of HUα₂. Like WT HUα-PAmCherry, ΔhupB HUα-PAmCherry cells grew similarly compared to a WT parental background strain and supported HU-dependent mini-P1 and Mu phage replication (Figure S1).

However, we found that HUα₂-PAmCherry molecules in the ΔhupB background diffused faster compared to that in the WT condition (Figure 3A, B), while its apparent binding affinity to DNA (determined by the transition kinetics, Table 1) was not significantly reduced (Supplementary Table S6). These observations suggest that under our WT growth conditions (i.e. mid-log phase growth at room temperature, in EZ rich defined media), the HUαβ heterodimer is likely the major species interacting with DNA.

Figure 3: dynamics of HU when HU heterodimerization was perturbed demonstrates functional aspects of HU dimers.

(A) Two diffusive states of HUα-PAmCherry in ΔhupB background with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (B) Mean squared displacement (MSD) plots of State I (mint) and State II (dark olive) ΔhupB HUα-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (C) Two-dimensional (2D) histogram of all cellular ΔhupB HUα-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for ΔhupB HUα-PAmCherry (total 30,739 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images). (D) Two diffusive states of ΔhupB HUα(P63A)-PAmCherry with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (E) Mean squared displacement (MSD) plots of State I (light purple) and State II (dark purple) ΔhupB HUα(P63A)-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (F) Two-dimensional (2D) histogram of all ΔhupB HUα(triKA)-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for ΔhupB HUα(P63A)-PAmCherry (total 173,740 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images). (G) Two diffusive states of ΔhupB HUα(triKA)-PAmCherry with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (H) Mean squared displacement (MSD) plots of State I (light pink) and State II (dark pink) ΔhupB HUα(triKA)-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (I) Two-dimensional (2D) histogram of all ΔhupB HUα(triKA)-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for HUα(triKA)-PAmCherry (total 80,749 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images).

Most interestingly, the apparent diffusion coefficients and transition kinetics of HUα₂ observed in ΔhupB HUα-PAmCherry cells were remarkably similar to those of HUα(P63A)-PAmCherry molecules with (Figure 2D) or without HUβ (Figure 3D, E). This suggests that mutation of P63A in the HUαβ heterodimer altered the transitory interactions of the heterodimer with DNA, or caused HUαβ to dissociate so only HUα₂ homodimer could form. Recalling that HUα(P63A) is implicated in recognizing bent or distorted DNAs in the forms of replication or transcription intermediates, it is possible that HUαβ heterodimer preferentially interacts with these specific DNA structures through P63-mediated dynamics, which may, in turn, play an important role in compacting the nucleoid. Consistent with this expectation, the nucleoid volume of ΔhupB HUα-PAmCherry cells expanded substantially compared to that of WT cells (Figure S7B, Table S7), indicating that the transitory interaction of the HUαβ heterodimer with DNA is important for maintaining a compact nucleoid.

Next, we monitored the dynamics of HUα(triKA)-PAmCherry homodimers in the ΔhupB background. We observed the largest increase in the diffusion coefficients (0.4 ± 0.01 μm²/s and 2.42 ± 0.03 μm²/s, μ ± s.e.m., n = 58,867 displacements from 68 cells) of the two significant populations (42% and 58%, Figure 3G). Consistent with larger diffusion coefficients, HUα(triKA)-PAmCherry homodimers were largely cytoplasmically localized (Figure 3I), mimicking that of non-DNA-interacting protein molecules freely diffusing through the nucleoid. Accompanying these dramatically abolished DNA interactions was the more compacted nucleoid observed in HUα(triKA)₂ cells, further supporting a role of the dynamic interactions of HU (mediated by triK) with DNA in decondensing the nucleoid to maintain its normal size, consistent with what was observed on the HUα(triKA)β heterodimer (Figure 2A, Figure S7B). Note that HUα(P63A)₂ cells showed more expanded nucleoids compared to HUα(triKA)₂ cells, consistent with the difference observed between HUα(P63A)β and HUα(triKA)β heterodimers (Figure S7B, Table S7). Importantly, both HUα(triKA)₂ and HUα(P63A)₂ cells had abnormal, single-lobed nucleoid morphology (Figure 3F, 3I) and severe cell growth defects (Figure S9, S12, Table S4), suggesting that the presence of at least one WT copy of either α or β subunit is important for decondensing the nucleoid, which is required for proper chromosomal DNA segregation.

HU dynamics are not significantly affected by the loss of naRNAs or nucleoid morphology

In addition to binding DNA, HU has also been shown to bind many RNA species (Balandina et al., 2001; Balandina et al., 2002; Macvanin et al., 2012; Qian et al., 2015). In particular, a set of highly homologous small non-coding naRNAs expressed from the REP325 element was shown to bind to HU and subsequently be deposited onto cruciform DNAs to condense relaxed plasmid DNA in vitro (Qian et al., 2015; Qian et al., 2017). Cells deleted of the REP325 element exhibited expanded nucleoids (Qian et al., 2015) (Figure S7), a phenotype that can be reversed by exogenous expression of naRNA4 (Qian et al., 2015). To determine if HU’s binding to naRNAs contribute to its DNA interaction dynamics, we integrated the hupA-PAmCherry fusion gene into a ΔREP325 strain (Qian et al., 2015) replacing the endogenous hupA gene. Single-molecule fluorescence in-situ hybridization (smFISH) confirmed that in the ΔREP325 strain there was a significant reduction of the corresponding naRNAs (Figure S10).

However, single-molecule tracking of HUα-PAmCherry in the ΔREP325 background only showed slight changes in HU dynamics (Figure 4A). The two diffusion populations of HUα-PAmCherry in ΔREP325 background had nearly identical apparent diffusion coefficients compared to WT cells, while the population percentage of State I molecules and transition kinetics showed changes on the order of 15 – 30% (Figure 3A, Table 1). The MSD plots of both State I and State II HUα-PAmCherry molecules in the ΔREP325 strain showed relatively larger sub-diffusion zones compared to those of WT cells (Figure 4B, Table S5), consistent with the slightly more dispersed cellular distribution of HUα-PAmCherry localizations (Figure 4C). Nevertheless, the nucleoid volume of the ΔREP325 was more expanded than WT cells and comparable to hupA(P63A)-PAmCherry cells (Figure S7B). These results suggest that HU’s diffusive dynamics in the nucleoid are not significantly altered by the absence of naRNAs, but naRNAs deposited by HU on the cruciform structures of the chromosome facilitate nucleoid condensation, as previous studies suggested (Qian et al., 2015; Qian et al., 2017).

Figure 4: Depletion of ncRNAs did not affect HU dynamics significantly.

(A) Two diffusive states of HUα-PAmCherry in ΔREP325 background with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (B) Mean squared displacement (MSD) plots of State I (light gold) and State II (dark gold) ΔREP325 HUα-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (C) Two-dimensional (2D) histogram of all cellular ΔREP325 HUα-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for ΔREP325 HUα-PAmCherry (total 108,846 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images). (D) Two diffusive states of HUα-PAmCherry under rifampicin treatment with respective population percentages (size of pie piece), transition rates, and diffusion coefficients (height of pie piece) as identified by the HMM. (E) Mean squared displacement (MSD) plots of State I (light red) and State II (dark red) rifampicin-treated HUα-PAmCherry trajectories as a function of time compared to State I and II molecules in the WT condition (light and dark blue). (F) Two-dimensional (2D) histogram of all cellular rifampicin-treated HUα-PAmCherry localizations from SMT (top) and aggregated nucleoid morphology from SIM (structured illumination microscopy) imaging (bottom). The pixel size of both was 100 × 100 nm. The top color bar indicated the number of localizations used in each bin for HUα-PAmCherry (total 137,019 localizations). The bottom color bar indicated the normalized fluorescence level (in arbitrary units) of the nucleoid-intercalating dye Hoechst 33342 (total 20 fluorescence images).

Because there are hundreds of highly homologous copies of REP elements in the genome (Dimri et al., 1992; Blattner et al., 1997) and HU can bind multiple species of naRNAs (Macvanin et al., 2012), a single ΔREP325 background might not be sufficient to examine the effect of naRNA binding on HU-DNA interaction dynamics. To test this possibility, we treated HUα-PAmCherry cells with rifampicin (200 μg/mL, 15 min), under which we observed the complete elimination of all cellular naRNAs from REP elements using smFISH (Figure S10). Interestingly, we still observed similar HUα-PAmCherry diffusion dynamics compared to the ΔREP325 strain (Figure 4D, E, Table 1). These results further confirmed that RNA-binding of HU did not contribute significantly to HU’s interacting dynamics with the chromosome. However, in rifampicin-treated cells, the nucleoid morphology changed significantly (Figure S7); the 2D histogram of the cellular distribution of HUα-PAmCherry localizations exhibited a three-lobed pattern, significantly different from untreated cells (Figure 4F). These dramatic morphologic changes likely resulted not from loss of naRNAs, but rather the loss of transcription which has been repeatedly shown to lead to global reorganization of the nucleoid (Cabrera et al., 2003; Cabrera et al., 2009; Bakshi et al., 2014).

As an additional control for whether altered nucleoid morphology affects HU-DNA interaction dynamics, we treated cells with the translation inhibitor chloramphenicol, which condenses the nucleoid through the inhibition of transcription-coupled translation (Cabrera et al., 2009). We observed similar diffusion dynamics of HUα-PAmCherry compared to untreated cells, yet highly condensed nucleoid morphology (Figure S11). Correspondingly, the cellular distribution of all HUα-PAmCherry localizations was highly compact as well (Figure S11C). These results demonstrated that the diffusive dynamics of HU and its weak and transitory binding to DNA were not affected by RNA binding or the underlying nucleoid morphology, and that these interactions of HU alone are necessary but not sufficient to maintain a proper nucleoid volume.

Discussion

Using single-molecule tracking of HUα-PAmCherry molecules in live E. coli cells, we demonstrated that HUα-PAmCherry exhibits nonspecific, weak, and transitory interactions with chromosomal DNA. These interactions are largely mediated by conserved lysine residues, as the triple lysine mutant HUα(triKA)-PAmCherry abolished these interaction dynamics and HU became predominantly cytoplasmic. The loss of such interactions coincides with an over-condensed and mis-segregated nucleoid, which was further exacerbated by the loss of the HUβ subunit in the mutant background (ΔhupB HUα(triKA)-PAmCherry). Because the triple lysine residues were shown previously to be responsible for binding to linear DNA nonspecifically and weakly in vitro (Hammel et al., 2016), these results suggest that the weak, transitory interactions of HUαβ heterodimer with the chromosome may play an important role in expanding the nucleoid to avoid an overly condensed nucleoid volume, which would be detrimental for nucleoid segregation (Fig. 5A).

Figure 5: HU plays a dual role in maintaining a proper nucleoid volume through differential interactions with chromosomal DNAs.

Specific DNA structure binding of HU, mostly likely mediated by the conserved P63 residue, the β subunit, and naRNA-binding, leads to the condensation of the nucleoid (top), whereas nonspecific, weak and transitory interactions of HU with chromosomal DNAs, most likely mediated by the surface lysine residues, leads to the expansion of the nucleoid (bottom). The balance between these two types of differential interactions maintain a proper nucleoid volume (middle). The loss of one does not affect the other, but leads to either an overly condensed nucleoid (top) or an expanded nucleoid (bottom).

Interestingly, when we mutated a conserved proline residue (P63) that was previously shown to bind kinked or cruciform dsDNA with high affinity (Swinger et al., 2003) we observed less dramatically altered diffusive behavior of HUα(P63A)-PAmCherry compared to that of HUα(triKA)-PAmCherry, but an expanded nucleoid volume. The further loss of the HUβ subunit in the P63A mutant background (ΔhupB HUα(P63A)-PAmCherry) did not significantly further aggravate the loss of dynamic interactions or nucleoid volume defect, suggesting that the P63 residue and the presence of the β subunit may be implicated in the same function of compacting the nucleoid through specific DNA structure binding.

Consistent with this hypothesis, we observed that in a ΔREP325 background where the expression of naRNAs was eliminated and the nucleoid significantly expanded, HUα-PAmCherry showed essentially unchanged interaction dynamics compared to that of the WT background. These results strongly suggest that the specific DNA structure binding of HU or the DNA-DNA contacts mediated by naRNAs between such DNA sites (Qian et. al., 2017) is important for nucleoid compaction, but has a negligible effect on the nonspecific, weak and transitory binding dynamics of HU to the bulk of chromosomal DNA.

Taken together, our results point to a model where HU plays a dual role in maintaining a proper nucleoid volume through its differential interactions with the chromosomal DNAs (Fig. 5). On one hand, HU compacts the nucleoid through its specific DNA structure binding ability, likely mediated by the conserved P63 residue, the presence of the β subunit, and/or the binding of naRNAs (Figures 2B, 3, 4, S7, S9, and S10). These high affinity binding events do not appear to contribute significantly to the transitory interaction dynamics of WT HU we observed in this work.

On the other hand, HU expands the nucleoid through its nonspecific, weak, and transitory interactions with the bulk chromosomal DNA. This role is most likely mediated by its three surface conserved lysine residues (Figure 2A, Figure 3G). The collective sum of a large number of weak, transitory interactions with chromosomal DNAs may contribute to the viscoelastic properties and fluidity of the bacterial nucleoid to maintain a properly expanded nucleoid that is essential for chromosome segregation (Jaffé et al., 1997) and other DNA transactions (Dixon and Kornberg, 1984; Dri et al., 1992; Li and Waters, 1998; Prieto et al., 2012). Such a fluid chromosome could also facilitate rapid (on the time scale of seconds) global changes in nucleoid morphology (Fisher et al., 2013). Note that other factors must also contribute to a properly expanded nucleoid in addition to HU, as chloramphenicol-treated cells maintained WT HU dynamics but had highly condensed nucleoid (Figures S7, S11), most likely due to the loss of protein transertion (Cabrera et al., 2009).

The viscoelastic maintenance model has been demonstrated for other systems, for example, the non-specific, transitory interactions of the muscle protein α-actinin modulate the mechanical properties of the actin filament network to form a viscoelastic gel (Xu et al., 1998). Interestingly, actin filament networks, due to their dynamic nature, demonstrate properties of liquid phase separation (Weirich et al., 2017). In the case of HU, this could suggest that the dynamic nature of HU binding to DNA could maintain the nucleoid as a separate “macro-compartment” that is physically distinct from the cytosol, and could explain why HU deletion strains have increased nucleoid volumes. Overall, our findings underscore the importance of weak, transitory non-specific binding on the biophysical and physiological properties of the cell and provide strong evidence for the contribution of this weak binding to the maintenance of the mechanical properties of the chromosome.

Experimental Procedures

Cell growth

Strains were inoculated from single colonies from freshly-streaked LB plates into EZ Rich Defined Media (EZRDM, Teknova; comprised of MOPS buffer, dipotassium phosphate, nucleoside solution, and amino acid solution) using 0.4% glucose as the carbon source with chloramphenicol (Sigma-Aldrich C0378) added to 150 μg/mL, and carbenicillen (Sigma-Aldrich C3416) added to 60 μg/mL respectively. Cells were grown overnight at room temperature (RT), shaking at 240 rpm. The following morning, saturated cultures were re-inoculated into fresh EZRDM and grow for several hours at RT until mid-log phase growth (OD600 ~0.4). For rifampicin drug treatment, rifampicin antibiotic (Sigma-Aldrich R3501) was added to a final concentration of 200 μg/mL for 15 minutes before cells were harvested for imaging (see Materials and Methods: SMT data collection below). For chloramphenicol drug treatment, additional chloramphenicol was added to a final concentration of 600 μg/mL for 30 minutes before cells were harvested for imaging. For both drug treatments, the drug was added when cells were at mid-log phase (OD600 ~ 0.4). For imaging of the galP fluorescent reporter operator system, cells were spun down at 4.5 relative centrifugal force (rcf) for 5 minutes and resuspended in EZRDM media with 0.4% arabinose to induce expression of the reporter gene. Induction was done at RT for 2 hours with shaking, then cells were again spun down at 4.5 rcf for 5 minutes, washed twice with fresh EZRDM media, then resuspended in fresh EZRDM media and grown at 30°C for 1 hour with shaking to allow for maturation of the mCherry fluorophore, after which cells were harvested as described in the section below.

Growth curves

For growth curves, distinct colonies on agar plates were inoculated into 2mL of EZRDM media for each strain and grown overnight at 24°C, shaking at 240 rpm. After ~16 hours of growth, saturated cultures were diluted 1 to 200 into a 96 well plate (100uL volume total for each well) using EZRDM media alone as a control. The 96 well plate was placed into a microplate reader (Life Sciences Tecan) and the absorbance at 600nm was measured every 30 minutes for a total of 12 hours. Reads were normalized against the control wells.

Construction of strains

Strains used in this study are mentioned in Supplementary Table 1. The construction of the hupA-PAmCherry strain (KB0026) has been described elsewhere (Wang et al., 2014). To test the functionality of the PAmCherry tagged HUα in the absence of HUβ, the ΔhupB mutation from the donor strain SCV152 was transduced into the HupA-PAmCherry strain using standard P1vir transduction protocol. Before P1vir transduction, the chloramphenicol resistance cassette linked to PAmCherry was removed using the temperature sensitive plasmid pCP20 (Cherepanov and Wackernagel, 1995) that expresses yeast flippase (FLP). For this, the HupA-PAmCherry strain was transformed with pCP20 plasmid and grown overnight at 42^oC to induce removal of the chloramphenicol resistance cassette flanked by short flippase recognition target (FRT) sites by the FLP recombinase.

hupA(triKA)-PAmCherry (SCV149) and hupA(P63A)-PAmCherry (SCV146) were constructed by lambda red recombineering using pSIM6 plasmid (Datta et al., 2006). In the first recombineering step, a kan-pBAD-ccdB cassette was introduced immediately after the stop codon of wild-type hupA gene of MG1655. In the second recombineering step, the wild-type hupA gene was replaced with the hupA gene encoding HUα(P63A) or HUα(triKA) protein. A synthetic double stranded DNA synthesized by Integrated Technologies (IDT) with the mutation was used for the recombination. The presence of the mutation was confirmed by Sanger sequencing. In the third recombineering step, PAmCherry linked with chloramphenicol resistance cassette was amplified from KB0026 and was fused in-frame to the hupAP63A gene after the last codon. The ΔhupB726::Kan^R from the strain JW0430–3 (obtained from Coli Genetic Stock Center (CGSC) at Yale University) was transduced into SCV146 and SCV149 to yield the strains ΔhupB hupA(P63A)-PAmCherry (SCV150) and ΔhupB hupA(triKA)-PAmCherry (SCV153). For the hupA-PAmCherry strain in the MG1655 background (SCV148), PAmCherry linked with chloramphenicol resistance cassette was fused in-frame to the wild-type hupA gene of MG1655 after the last codon. Subsequently, the ΔhupB726::Kan^R from the strain JW0430–3 was transduced into SCV148 to yield the strain ΔhupB hupA-PamCherry (SCV152).

Mu phage assay

Spot dilution plates were performed initially to estimate proper Mu phage lysate dilution to use for phage plaque counting. Cells were grown overnight in Luria Broth (LB) at 37^oC, and the next day reinoculated into LB with 1 mM CaCl₂ and 2.5 mM MgCl₂. Cells were grown until O.D. ~ 0.4, then spun down and resuspended so the final O.D. for later use in the protocol is ~ 1. Next, 100 μl of cells were mixed with 10 μl of the appropriate dilution of Mu phage lysate. This was mixed quickly into 3 ml of 0.7% top agar (warmed to 55 ^oC) and immediately poured onto TB plates (warmed to 37^oC) and redistributed evenly prior to hardening. Plates were let dry at RT and incubated overnight at 37^oC. The next day, plates were examined and the plaques were counted by eye. The PFU/ml (plaque forming units, Mu/mL) was calculated using the following equation:

$$\frac{PFU}{ml}=\frac{Averagenumberofplaques}{DxV}$$

D = dilution

V = volume of diluted virus added to plate

Mini-P1 plasmid transformation assay

Electrocompetent cells were made using standard protocol. In brief, 6 ml of E. coli cells were grown until O.D. ~0.4 in LB, the cultures were put on ice for 30 min. Cells were spun down at 4100 rpm at 4^oC for 10 min, cell pellet was resuspended in 6 ml ice cold H₂O. Cells were spun down again and resuspended in 3 ml of ice cold 10% glycerol. This was repeated once more, then cells were resuspended in 1.5 ml ice cold 10% glycerol and spun down again. The final cell pellet was resuspended in ~300 μl of GYT media, and 6 individual aliquots of 50 μl of competent cells were made. Cells were quickly frozen in dry ice, and stored at −80^oC until later use.

The same batch of competent cells were used for all the transformation experiments for consistency. The pUC19 transformation control was performed in order to normalize the transformation efficiency of competent cells between the different E. coli strains. An appropriate volume of transformation outgrowth was plated on appropriate antibiotic selection LB agar plates and incubated O/N at 37^oC. The next day, the numbers of colonies on agar plates were counted and recorded. The mini-P1 CFU/μg DNA was normalized to the same strain’s pUC19 CFU/μg DNA to account for competent cell transformation efficiency differences. The WT MG1655 strain was set as the reference standard for comparison with other E. coli strains.

SMT data collection

3% w/v agarose gel pads were made using a low-melting-temperature agarose (SeaPlaque, Lonza) and EZRDM media. 1.5mL of cells were harvested by spinning down in a bench-top microcentrifuge at 4.5 rcf for 5 minutes and resuspended in ~100μL of fresh EZRDM. Cells were pipetted onto agarose pads and sandwiched between the agarose pad and a #1 coverslip. Immobilized cells were imaged on an Olympus IX71 inverted microscope with a 100x oil objective (PN, NA =1.45) with 1.6x additional magnification. Photons from cells were collected with an Andor EMCCD camera using MetaMorph imaging software (Molecular Devices). Fluorescence from cells was obtained using solid-state lasers at 405nm and 568nm wavelengths (Coherent). All SMT images were collected using 5ms exposure with 1.74ms of cycle time for a total frame length of 6.74ms. Three movies consisting of 2500 frames each were taken for each cell. No cell was imaged longer than five minutes to avoid phototoxicity effects. Activation of the fluorescent proteins was continuous throughout imaging, and no changes were made to either the activation or excitation power throughout the imaging session.

SMT analysis

For single-molecule tracking analysis, tiff stacks of cell images were imported into the single-molecule tracking software UTrack version 3.1 (Jaqaman et al., 2008) within the Matlab 2017a software. We performed the detection and tracking of molecules on individual movies using the Gaussian Mixture Model setting within the UTrack software. For detection, an α value of 0.01 was used; for frame-to-frame linking, 15 frames was the maximum time and 0 to 2 pixels (or up to 6 pixels with increased linking cost) was the maximum distance to link trajectories together. Linked trajectories were then filtered by their intensity (between 0.5–1.5x the average single-molecule intensity to exclude molecules that had temporarily colocalized) and their localization within cells (i.e. any trajectory outside of cells was excluded) and exported as MATLAB mat files. We then only used consecutive frame trajectories for subsequent analysis; when gaps were encountered, they were made into separate trajectories. These trajectories were analyzed using custom in-house code (available upon request) to assess their diffusive properties. Exact equations used to analyze the trajectories are explained in more detail in the supplemental information. To build the HMM, we installed and used the software vbSPT, version 1.1 on MATLAB version 2014a (Persson et al., 2013). Cells were rotated such that the long cell-axis corresponded to the x-axis. We used displacements in the x-direction for analysis to limit the effects of confinement due to the small size of the bacterial cell.

Localizations of HU-PAmCherry

Detections of HU-PAmCherry were determined as above using the UTrack software. Cells were rotated such that the long cell axis corresponded to the x-axis, and the middle of the cell corresponded to the origin. We determined the spatial coordinates based on this axis and added spatial coordinates of all cells together. For single-molecule tracking experiments, cells were chosen to fit within the imaging view (40 × 40 pixels) and thus have a narrow distribution of cell lengths and widths. We binned these displacements into 100 × 100 nanometer bins and applied the red-hot colormap for viewing. Images were cropped to 4 microns long and 1.5 microns wide to enable direct comparison across conditions.

Imaging of cell nucleoids

To image nucleoids, strains were grown as above, then stained with 5 μM of Hoechst 33342 dye for 10 minutes. Cells were washed twice with fresh media to remove unincorporated dye, then fixed in 3.7% paraformaldehyde (15 min). Cells were washed twice with PBS and resuspended in a final volume of 200uL and stored in a light-proof box at 4°C until ready to image. Cells were kept for no longer than one week. Fixed and stained cells were mixed 1:1 with anti-fading solution (20% NPG (n-propyl gallate), 60% glycerol, 1x PBS) and adhered to coverslips treated with 0.01% poly-L-lysine solution. Coverslips were sealed onto a microscope slide using clear nail polish and imaged on the GE OMX SR structured illumination microscope (excitation: 405, camera channel: 488, exposure 50ms, 5% of highest laser intensity). Images were reconstructed using the standard parameters for the GE OMX SR microscope within the GE SRx software.

Analysis of nucleoid images

For each strain, individual cell images were cropped and rotated such that the long axis of the cell corresponded to the x-axis. Black space was manually added to the edges of images to center cells, such that the middle of the cell corresponded to the axis origin. Each set of images for each strain was then added together and re-binned such that each bin corresponded to 100 × 100 nm to allow for direct comparison to HUα-PAmCherry localizations. All images were cropped to 4 microns long and 1.5 microns wide. For better visual comparison, matrix values were normalized to 1 and then multiplied by the total localizations of HUα-PAmCherry.

To determine the area of the nucleoid, we first normalized images using WT HUα-PAmCherry nucleoids as a template using histogram matching combined with gradient information (Sintorn et al., 2010) to account for intensity and noise differences inherent from day-to-day imaging. Next, we determined a threshold based on Otsu’s method (graythresh.m in MATLAB), ensuring the effectiveness was greater than 0.85, to generate a binary image. The nucleoid area was then the number of pixels above the calculated threshold times the pixel size (40 × 40 nm). To estimate the error, we used bootstrapping (boostrap.m in MATLAB) to re-sample the average nucleoid intensities and repeated the area measurement as before. Bootstrapping was completed 100 times per strain. These bootstrapped means were used to perform a one-way ANOVA test to determine if these nucleoid areas were statistically significant from each other.

Total cellular RNA extraction

Cell growth conditions are same as described above. Starting with an identical number of cells for WT condition (no drug treatment), rifampicin treatment (200 μg/mL, 60 min) and chloramphenicol treatment (600 μg/mL, 60 min). Take volume of cells to an O.D. equivalent of 8.5, pellet cells (10 min, 4100 rpm, 4^oC), resuspend the pellet in 1 mL 1xPBS, and transfer to an Eppendorf tube. The cells were pelleted again (2 min, 14000 rpm, 4^oC), the pellet was quickly frozen on dry ice and kept at −80^oC until the later steps. For RNA extraction, 600 μL of Trizol was added to the cell pellet after thawing, then the protocol from the Direct-zol RNA Miniprep Kit (Zymo Research) was followed, an in-column DNase I digestion was performed as suggested for 15 min. Finally, RNA was eluted with 50 μl of DNase/RNase free H₂O and quantified by measuring on Nanodrop.

RNA FISH

Cells were grown the same as for imaging (described above; grown in rich defined media at 24°C). At mid-log phase growth corresponding to OD₆₀₀ ~ 0.4, cells were fixed in 2.6% paraformaldehyde, 0.8% glutaraldehyde in 1 x PBS for 15 minutes at room temperature. Cells were then washed twice in 1 x PBS and resuspended into a small volume of GTE buffer (2mL of cells were spun into 50uL final volume). Coverslips were treated with 0.1% poly-L-lysine for 5 to 10 minutes, then rinsed and dried. Cells were spotted onto treated coverslips and left to dry for 20 minutes. Non-adhered cells were washed away with PBS. Cells were then treated with ice-cold 70% ethanol and placed at −20°C for 10 minutes to permeabilize the cells. After permeabilization, cells were washed twice in 2 x SSCT (2 x SSC + 0.1% Tween-20) for five minutes each. Cells were then placed in pre-hybridization solution (2 x SSC, 40% formamide, 0.1% Tween-20) for 30 to 45 minutes at 37°C. Pre-hybridization solution was removed, then a small volume of hybridization solution (2 x SSC, 40% formamide, dextran sulfate, yeast tRNAs, 1μM of Cy3-labelled probe) was spotted onto the coverslip, and covered with a clean piece of saran wrap to set and prevent evaporation. Coverslips were placed in a sterile petri dish and submerged in a water bath set to 94°C for two minutes, then immediately placed at 42°C for 15 to 18 hours to hybridize. The probe sequence used was: Cy3-GTTGCCGGATGCGGCGTAAACGCCTTATCCGGCC. Cells were washed in 40% wash solution (2 x SSC, 40% formamide, 0.1% Tween-20) for 30 minutes at 37°C, washed in 20% wash solution (2 x SSC, 20% formamide, 0.1% Tween-20) for 10 minutes at room temperature, washed in 2 x SSCT solution for 10 minutes at room temperature, and finally transferred to 2 x SSC solution until imaging. Coverslips were mounted onto imaging chambers and anti-fading media was perfused into the system. Cells were imaged using the same set-up as for SMT tracking. Z-stacks of cells were collected using 50ms exposure per frame with an EM gain of 300. Z-stacks were imported to ImageJ, where cells were manually identified. Fluorescence in each cell was corrected by subtraction of the background (a random area devoid of cells for each image). Although the sequence was against REP325 naRNA4, signal remained above background in the ΔREP325 background strain due to high homology of the REP325 RNAs with hundreds of other REP elements in the cell. For each of the two biological replicates, over 100 cells were analyzed.