DNA sequence-directed cooperation between nucleoid-associated proteins.

Nucleoid-associated proteins (NAPs) are a class of highly abundant DNA-binding proteins in bacteria and archaea. While both the composition and relative abundance of the NAPs change during the bacterial growth cycle, surprisingly little is known about their crosstalk in mutually binding and stabilizing higher-order nucleoprotein complexes in the bacterial chromosome. Here, we use atomic force microscopy and solid-state nanopores to investigate long-range nucleoprotein structures formed by the binding of two major NAPs, FIS and H-NS, to DNA molecules with distinct binding site arrangements. We find that spatial organization of the protein binding sites can govern the higher-order architecture of the nucleoprotein complexes. Based on sequence arrangement the complexes differed in their global shape and compaction as well as the extent of FIS and H-NS binding. Our observations highlight the important role the DNA sequence plays in driving structural differentiation within the bacterial chromosome.

Introduction

Nucleoid-associated proteins (NAPs) represent a small class of highly abundant DNA architectural proteins involved in shaping the bacterial chromatin and in regulating the gene expression in prokaryotes and archaea (Dame et al., 2020; Dillon and Dorman, 2010; Driessen and Dame, 2011; Luijsterburg et al., 2006; Travers and Muskhelishvili, 2005). During the bacterial growth cycle, these proteins are expressed in a growth-phase-dependent manner to coordinate the chromosome structure with the metabolic state (Azam et al., 1999; Sobetzko et al., 2012; Sonnenschein et al., 2011; Travers and Muskhelishvili, 2020). NAPs bind DNA with varying affinities from nanomolar to micromolar concentrations and affect the gene expression by acting as bona fide transcription factors as well as so-called “topological homeostats” (Muskhelishvili and Travers, 2003; Travers et al., 2001). The regulation of genomic transcription by NAPs is closely coupled with their propensity to modulate the availability of free or “unconstrained” DNA superhelicity, as NAPs constrain the DNA acting both as supercoil repositories and topological barriers to supercoil diffusion (Berger et al., 2016; Dages et al., 2020; Hardy and Cozzarelli, 2005; Hatfield and Benham, 2002; Muskhelishvili and Travers, 2003).

Factor for Inversion Stimulation (FIS) protein is the most abundant NAP during the exponential growth phase in Escherichia coli, while its concentration quickly drops to undetectable levels upon the transition of cells to stationary phase (Azam et al., 1999; Ball et al., 1992; Ninnemann et al., 1992). FIS has a global nucleoid-structuring function (Fisher et al., 2013; HadizadehYazdi et al., 2012; Wu et al., 2019a, 2019b) as well as local accessory roles in the assembly of synaptic complexes by site-specific recombinases (Ball and Johnson, 1991; Mertens et al., 1988) and transcription-initiation complexes at various promoters, including the exceptionally strong RNA (rRNA and tRNA) promoters (Bokal et al., 1995; Lazarus and Travers, 1993). The latter are characterized by upstream activating sequences (UAS) containing multiple FIS-binding sites that are arranged in a helical register (Hirvonen et al., 2001). FIS is a helix-turn-helix (HTH) DNA-bending protein (Stella et al., 2010) which upon binding at the phased sites in UAS forms a coherently bent DNA loop that associates with RNA polymerase (Maurer et al., 2006). In general, FIS nucleoprotein complexes form loops and stabilize branches in supercoiled DNA (Japaridze et al., 2017a; Schneider et al., 2001; Skoko et al., 2006; Travers et al., 2001).

In contrast to FIS, the histone-like nucleoid-structuring (H-NS) protein is an NAP expressed throughout the entire bacterial growth cycle (Azam et al., 1999), slightly increasing in concentration toward the later growth stages. While overproduction of H-NS in vivo is lethal for the cell (Spurio et al., 1992), the deletion of the H-NS gene does not result in large-scale restructuring of the nucleoid (Brunetti et al., 2001; Lioy et al., 2018; Wu et al., 2019a). H-NS is a DNA-bridging protein that binds preferentially to AT-rich DNA sequences (Dorman, 2007; Grainger et al., 2006; Lang et al., 2007; Lucchini et al., 2006; Navarre et al., 2006), in part mediated by an A-T hook motif that interacts with the DNA minor groove (Badaut et al., 2002; Gordon et al., 2011). It was shown that H-NS can bridge two DNA helices within a rigid filament (Dame et al., 2000; Schneider et al., 2001), trapping RNA polymerase (Dame et al., 2001; Schröder and Wagner, 2000). It is assumed that binding of H-NS nucleates at high-affinity sites and subsequently spreads along the DNA strands, leading to gene silencing (Bouffartigues et al., 2007; Kahramanoglou et al., 2011; Shin et al., 2012). While H-NS can polymerize on a single DNA duplex, resulting in its stiffening (Liu et al., 2010), it is the DNA-bridging mode facilitated by Mg2+ ions that has been primarily implicated in transcriptional repression (Kotlajich et al., 2015; Lim et al., 2012; van der Valk et al., 2017).

While the structural role of FIS, H-NS, and other DNA-binding NAPs has been intensively studied (Dame et al., 2000; Fisher et al., 2013; Hadizadeh Yazdi et al., 2012; Maurer et al., 2006; Schneider et al., 2001; Wu et al., 2019a, 2019b), surprisingly little is known about putative cooperative binding effects and the architecture of the ensuing long-range DNA structures. From NAP expression patterns (Azam et al., 1999) it is clear that distinct combinations of these proteins interact with the genomic DNA during the different stages of the cell cycle (Travers and Muskhelishvili, 2020). Indeed, exploration of their cooperative binding effects appears indispensable for understanding the growth phase-dependent gene regulation. Interestingly, previous studies using atomic force microscopy (AFM) showed that cooperative binding of various combinations of NAPs to the linear phage λ-DNA led to regular structures that were quite distinct from those observed with individual proteins (Maurer et al., 2009), while on binding to large supercoiled molecules, NAPs did phase separate, forming domain-like regions (Japaridze et al., 2017a). Resolving specific higher-order structures formed by cooperative binding of NAPs is challenging, but it can be achieved by using model DNA sequences that contain a few high-affinity binding sites that facilitate the nucleation of long-range nucleoprotein complexes.

Here, we explore whether the binding of a combination of two major bacterial NAPs, FIS and H-NS, leads to the emergence of distinct nucleoprotein structures that are more than the mere sum of those formed by the individual NAPs. To address this question, we employed DNA sequences with various arrangements of FIS and H-NS binding sites and studied the resulting higher-order nucleoprotein complexes using AFM and nanopore experiments. We find that sequential organization of the binding sites directs the assembly of peculiar nanometer-sized hairpin-like DNA architectures by FIS and H-NS, which are not observed with either FIS or H-NS alone. Furthermore, we find that FIS and H-NS self-organize to separate domains along the DNA molecules when both proteins are present. Our results exemplify the type of structural rearrangements that local DNA regions can undergo during bacterial growth cycle, highlighting the crucial role of the DNA sequence in this process.

Results

DNA sequence directs cooperative binding of FIS and H-NS

To study the combined effects of native FIS and H-NS proteins (Lautier and Nasser, 2007, Nasser and Reverchon, 2002) binding to DNA, we used two ∼4kb plasmids of ∼1300nm length that differed in the sequence arrangement of the NAP-binding sites. In one construct, we arranged an FIS-binding sequence UAS and an H-NS-binding sequence NRE in a head-to-tail (HT) fashion (HT: UAS-NRE-UAS-NRE). Here, UAS is the upstream activating sequence of tyrosyl-tRNA gene promoter (tyrT UAS) (Lamond and Travers, 1983) and NRE is the negative regulatory element of proV gene (proV NRE) from an osmoregulatory operon (Gowrishankar, 1985). In the second construct, the same NAP-binding sequences were arranged in a head-to-head (HH) fashion (HH: UAS-NRE-NRE-UAS). These sequences were inserted into a 2.9kb plasmid backbone devoid of any strong FIS or H-NS binding sites (Japaridze et al., 2017b) (Figures 1A and 1E, see transparent methods). The 397bp-long tyrT UAS region contained three specific FIS binding sites (with K values ranging between 7.5 nM and 60 nM) arranged in a helical register (Lazarus and Travers, 1993), while the 264 bp long proV NRE contained two high-affinity H-NS sites (K values between 15 nM and 25 nM) separated by about 10 helical turns and many low affinity sites (Bouffartigues et al., 2007). The constructs with HT and HH arrangements have been described in detail in a previous study (Japaridze et al., 2017b), which showed that the binding of H-NS to various arrangements of high-affinity binding sites led to the formation of distinct long-range plectonemically coiled structures. Increasing concentrations of bound H-NS resulted in gradual shortening of the DNA contour length, indicating that H-NS protein upon binding was braiding the DNA into a twisted structure rather than just bridging the duplexes (which would not result in a reduction of the contour length). These nucleoprotein structures differed in their shape, compaction, and capacity to constrain DNA supercoils.

Figure 1

AFM images of FIS nucleoprotein complexes formed on circular nicked DNA constructs

Schematic depiction of (A) the head-to-tail (HT) and (E) head-to-head (HH) circular constructs.

AFM images of (B) control HT and (F) control HH plasmids deposited on mica without proteins with the respective number of DNA loops (right panels in B and F).

Upon the addition of FIS protein (FIS:bp ratio = 0.0026) to (C) the HT plasmid as well as to the (G) HH plasmid, the number of DNA loops increased about threefold. Solid lines denote Gaussian fits. Scale bars 200nm. Zoomed in AFM images of FIS looped structures for (D) HT and (H) HH plasmids . Scale bars 100nm.

First, we investigated nucleoprotein structures formed by the binding of the native FIS protein to these constructs. We incubated FIS with either the nicked circular HT or HH plasmid and subsequently imaged by AFM in air (see transparent methods). We found that binding of the protein induced global changes in the shapes of both DNA constructs. FIS stabilized DNA crossovers and loops (Figures 1B–1D and 1F–1H). This is in sharp contrast to the plectonemic DNA structures stabilized by H-NS (Japaridze et al., 2017b) (see also Figure S1). Upon deposition on the surface, naked plasmids displayed, on average, a single DNA crossing (Figures 1B and 1F, average number of crossings is 1.0 for HT (N = 108) and 0.9 for HH (N = 129)). Upon co-incubation with FIS, however, the number of loops increased almost three times to 2.8 for HT (N = 60) and 2.5 for HH (N = 63). Furthermore, the typical height of the DNA crossings stabilized by FIS in both constructs was consistently larger (hFIS = 1.35 ± 0.15 nm, N = 60; mean ± SD) than that of DNA crossings in the control samples without protein (hDNA = 1.05 ± 0.15 nm, N = 60; mean ± SD) (Figure S2). This enabled us to distinguish the FIS binding position along the DNA molecules. Next, we looked at the typical loop sizes (defined as the loop contour length) formed by the binding of FIS protein. The average length of the DNA loops in FIS bound samples was 135 ± 45nm (N = 50; mean ± SD) for the HH construct and 165 ± 45nm for HT (N = 50; mean ± SD), values that in both cases were smaller than the loop size in the control sample (loop length HH control = 230 ± 70nm; N = 40; loop length HT control = 240 ± 80nm, N = 40; mean ± SD).

The nanometer resolution of the AFM enabled us to also very precisely measure the various DNA shape parameters, such as the total measured contour length, the radius of gyration, and the effective persistence length of molecules – providing an insight into the structural changes induced by the FIS binding (summarized in Table 1). The radius of gyration (Rg) describes the average globular size of the molecules (Grosberg and Khokhlov, 1994) while the persistence length (lp) (Rubinstein and Colby, 2003), describes the stiffness of the DNA molecules. For both constructs, binding of FIS decreased the radius of gyration significantly, by roughly 30%, from the initial size of ∼150 nm down to ∼105 nm (Table 1). Interestingly, when we incubated FIS with a control pBR322 plasmid of a similar size (4.4kb) that was devoid of any strong FIS binding sites (Figure S3), a smaller compaction rate was observed. At similar FIS concentrations, we observed only a 17% compaction. The average number of loops, however, increased by ∼3 times (i.e., similar as in the HT and HH constructs) from 0.6 (N = 118) to 1.9 (N = 85), indicating FIS was still binding the DNA molecule independent of high-affinity binding sites. The effective persistence length of the DNA molecules was also affected by the binding of FIS, decreasing from 52 ± 3 nm (N = 35, error is SD) to 40 ± 5 nm (N = 30, error is SD) as expected by the observed looping of DNA. Notably, the contour length of the DNA was not much changed by FIS binding (Table 1), in contrast for H-NS binding where we saw a clear compaction.

Table 1

Statistical parameters of circular DNA molecules bound by FIS and H-NS (mean ± SD)

Circular DNA shape parameters
Sample	Number	L(contour) [nm]	ΔL (%)	lp [nm]	Rg [nm]	ΔRg (%)
Head-to-head
Control	60	1280 ± 45	–	53 ± 3	145 ± 20	/
+ FIS: kb = 2.6	45	1240 ± 65	−3	35 ± 5	107 ± 15	−26
+ FIS: H-NS: kb= 2.6 : 6.8 : 1	35	1265 ± 65	−1.0	30 ± 5	105 ± 15	−27
+ H-NS: kb = 0.9	35	1110 ± 105	−13	41 ± 5	121 ± 15	−16
Head-to-tail
Control	65	1295 ± 65	–	53 ± 3	152 ± 15	/
+ FIS: kb = 2.6	45	1260 ± 45	−2.5	27 ± 5	102 ± 10	−33
+ FIS: H-NS: kb= 2.6 : 6.8 : 1	43	1290 ± 65	−0.5	34 ± 5	105 ± 10	−31
+ H-NS: kb = 0.8	30	1180 ± 75	−9	42 ± 5	125 ± 15	−18
pBR322 control (4.4kb)
Control	35	1370 ± 30	–	52 ± 3	162 ± 20	/
+ FIS: kb = 2.4	30	1420 ± 70	+3.5	40 ± 5	135 ± 20	−17
+ H-NS: kb = 3.7	55	1405 ± 35	+2.5	53 ± 5	155 ± 15	−4

Next, we imaged the nucleoprotein complexes formed in the presence of both FIS and H-NS with the HT and HH constructs. In our experiments, we used a FIS monomer to DNA bp ratio of 0.0026, and H-NS monomer to DNA bp ratio of 0.0066, protein concentrations that were lower by about an order of magnitude than the maximum physiological levels in the cell (Azam et al., 1999). On both constructs, we observed regions where the DNA duplexes were either bridged into plectonemic braids (as based on the DNA contour length shortening) or were crossing each other to form loops (Figure 2), that is, the characteristic shapes induced by H-NS and FIS nucleoprotein complexes, respectively. Thus, at the used near-physiological protein concentrations, both FIS and H-NS could bind naked DNA independently, forming their characteristic nucleoprotein structures. Measurements indicated that the number of DNA crossovers on both constructs was comparable to each other (Figure 2C) as well as to the number of crossovers obtained with FIS alone (Figures 1C and 1D). By contrast, contour length measurements showed that, on average, the bridged duplex regions formed on HH constructs were significantly more extended compared to HT (Figure 2D). On average, in HH DNA the region with bridged duplexes was 47 ± 12nm in length, (N = 43, mean ± SD), while the bridged regions observed in HT constructs were much shorter in length, 33 ± 12 nm (N = 56, mean ± SD). These latter bridged regions are shorter in contour length compared to the bridged regions formed by H-NS only, as reported earlier (Japaridze et al., 2017b).

Figure 2

AFM images of FIS and H-NS nucleoprotein complexes formed on HT and HH constructs

(A) The HT construct forms looped structures indicative of FIS binding, adjoined by DNA bridges indicative of H-NS binding.

(B) The HH construct forms hairpin-like structures where the two proteins appear organized in a specific binding arrangement. Scale bars: 200nm for large scale images and 100nm for magnified images.

(C) The number of DNA crossings is similar for both HT and HH constructs. d. The length of the bridged DNA segments is larger for the HH (N = 43) than for the HT (N = 56) constructs. Solid lines denote Gaussian fits. e. Zoomed-in AFM image of the hairpin structure formed on the HH construct. f. Corresponding height cross section (indicated by dotted line on panel e).

Furthermore, on the HH construct, peculiar hairpin-like structures consisting of a “stem” of bridged DNA duplexes with a protruding loop were observed (Figures 2B and 2E), while no such structures were observed with the HT construct, suggesting that they were produced by a cooperative binding effect of FIS and H-NS to DNA with a specific (HH) arrangement of cognate binding sites.

Notably, earlier AFM studies measuring the DNA plectonemes stabilized by H-NS on both HH and HT constructs as well as on linear phage lambda DNA, reported uniform height values of ∼1 nm along the H-NS bridged regions (Japaridze et al., 2017b; Maurer et al., 2009). By contrast, height measurements of H-NS bridged duplexes in the hairpin structures formed on the HH constructs with the mixture of FIS and H-NS demonstrated non-uniform heights that varied between the stem of the bridged DNA and the base of the loop, i.e., the point where the DNA duplexes became disjointed (Figures 2E and 2F). The stem was typically ∼ 1nm in height, while the height of the crossings abutting the diverging duplexes was 1.45 ± 0.2 nm (N = 60, error is SD) (Figure 2F), remarkably close to the values obtained for FIS binding (Figure S2). Given the propensity of FIS to stabilize DNA loops and high crossovers (Japaridze et al., 2017a; Maurer et al., 2006; Schneider et al., 2001; Skoko et al., 2006) and that of H-NS to form bridged DNA duplexes of uniform height, it is reasonable to assume that the peculiar DNA hairpins observed in HH constructs represent nucleoprotein complexes formed by cooperative binding of H-NS and FIS at this DNA sequence. We estimate that FIS stabilizes the base of the loop whereas H-NS bridges the DNA duplexes and forms the stem of the hairpin. We emphasize that the HT and HH constructs differ only in the spatial arrangement of the sequences with the H-NS and FIS binding sites. Therefore, it is remarkable that binding of H-NS and FIS to these two constructs can lead to the formation of such distinct structures only on one (HH) construct but not on the other (HT).

Nanopores show that nucleoprotein complexes are different in bulk

To rule out that the differences observed between the HT and HH nucleoprotein complexes represent an artifact of AFM imaging on the mica surface, we investigated the formation of these nucleoprotein complexes in solution with the use of solid-state nanopores. Apart from serving as a complementary technique to AFM in assessing the formation of nucleoprotein complexes in solution, nanopores also enabled us to gather good statistics (hundreds of DNA translocation events for each experiment). Essentially, the method involves measuring the electrophoretically driven translocation of DNA molecules across a nanometer-sized pore (15 nm diameter in our case) with an applied voltage (see transparent methods), as shown in Figure 3A. During the translocation of the molecules across the pore, the DNA temporarily disrupts the flow of ions (LiCl solution in this case) which leads to a current blockade. As schematically illustrated in Figure 3B, one can distinguish the level of folding and compaction of DNA molecules based on their current blockade levels (Kumar Sharma et al., 2019; Plesa and Dekker, 2015; Plesa et al., 2016) and see if the molecules are bound by proteins (Yang et al., 2018).

Figure 3

Nanopore experiments

(A) Schematics depicting the translocation of a DNA through a nanopore.

(B) Typical blockade events for HT DNA construct passing through a nanopore.

(C) Typical nanopore traces for the HT + FIS & H-NS samples (1ng/μl final DNA concentration).

(D) Typical nanopore traces for HH + FIS & H-NS samples (1ng/μl final DNA concentration). Upon the protein addition, the blockades become deeper, indicative of protein binding. The blockade events for the HH + FIS & H-NS samples are uniform and homogeneous, indicating that the DNA-protein complexes are similarly organized.

Normalized blockade events for the (E) HT and (F) HH constructs, showing that the deeper events occur much more frequently for the HT construct (in 38% of total events, NHT+ Protein = 277 out of 737 total events), but not for the HH construct (in 18% of total events, NHH+Protein = 329 out of 1812 total events). Errors bars on panels (E) and (F) indicate five times the rms of the signal noise.

If the HH and HT nucleoprotein complexes were different in their levels of compaction, one would expect to see different blockade levels between the two constructs. As expected, both the HT and HH-DNA plasmid-only controls yielded consistent DNA events with a regular depth in the current blockades (Figures S4 and S5). After preincubation of the HH and HT DNA molecules with FIS and H-NS (see transparent methods), we observed that the HT-DNA showed deeper blockade events (Figure 3C) compared to the HH-DNA (Figure 3D) indicative of more extensive protein binding (Figure S4). Furthermore, for HH-DNA (as well as for HT), the blockade events in the presence of both FIS and H-NS were uniform and homogeneous, whereas separate addition of individual NAPs to the HH construct resulted in heterogeneous blockade levels (Figure S5). We set a threshold to select for protein bound molecules and then we quantified the percentage of events as compared to the control. Figures 3E and 3F show a relative increase in the number of deep events for the HT and HH cases in the presence of both proteins. We observed a 1.65 ± 0.06-fold increase in the number of deep events for the HT construct, but only a 1.19 ± 0.07-fold increase for the HH construct. These data are consistent with the AFM data showing that the NAPs binding of the HT construct results in an increased compaction compared to the HH construct, consistent with longer hairpins, irrespective of the used threshold (Figure S5K).

Taken together, our results indicate that spatial organization of the FIS and H-NS binding sites in the DNA is determining the 3D architecture of the nucleoprotein complexes.

DNA sequence arrangement governs the 3D nucleoprotein structure

The only difference between the nicked circular HH and HT constructs resides in the 1.4kb DNA region comprising different arrangements of the UAS and NRE sequences. We studied these 1.4kb regions on linear DNA fragments to test whether they could drive organization of different nucleoprotein assemblies, and whether the observed distinct hairpin structures could still form with the HH (UAS-NRE-NRE-UAS) as opposed to the HT (UAS-NRE-UAS-NRE) arrangement of the DNA sequence.

Hence, we incubated the 1.4 kb linear HH and HT fragments with either FIS or H-NS, or a mixture of both under conditions similar to those used for the circular substrates. We found that binding of FIS alone stabilized the looped structures (Figures 4 and S6C), but we observed clear differences between the two constructs. The HT fragment typically formed single ‘lasso’ type structures, while the HH fragment formed ‘butterfly’ structures with multiple loops (Figures 4C and S6C). The ‘lasso’ structure is consistent with interaction between the FIS-UAS nucleoprotein complexes formed by binding of FIS at the UAS regions located one at the middle and one at the end of the HT fragment, whereas the ‘butterfly’ structure is consistent with interactions between the FIS molecules binding at the UAS regions located at the extremities of the HH fragment. The average number of DNA crossings formed upon FIS binding also increased differently for the two constructs. For the HH fragment, the looping number increased by almost three times (from 0.67 (N = 104) to 1.95 crossings (N = 80)), while for the HT it increased by more than two times (from 0.55 (N = 111) to 1.25 crossings (N = 95)) compared to samples without the added protein. Interestingly, for both constructs, FIS formed DNA loops of a similar size ∼100nm in contour length (HH Lloop_HH = 110 ± 35 nm (N = 43, error is SD) and HT Lloop_HT = 100 ± 25 nm (N = 40, error is SD), i.e., smaller in size compared to the loops formed by simple deposition of the naked DNA on surface Lloop_control = 150 ± 60 nm (N = 85, error is SD) (Figure S7). Similarly, as in the case of nicked circular DNA molecules, the binding of FIS compacted the structures, thereby reducing the radius of gyration and the effective persistence length (Table S1).

Figure 4

AFM images of nucleoprotein complexes formed on linear 1.4kb fragments

(A–H) (A) Schematic depiction of the head-to-tail (HT) and head-to-head (HH) constructs. Blue indicates the NRE and red UAS regions. b-h. AFM images with corresponding schematic depictions of the nucleoprotein organization for (B) control, (C) FIS-bound and (D) H-NS-bound linear fragments, (E) nucleoprotein complex formed by HT fragment bound by FIS & H-NS, and (F–H) nucleoprotein complex formed by HH fragment bound by FIS & H-NS.

(G) Height measurement of the nucleoprotein complex formed upon binding of FIS & H-NS to the HH fragment shown on (F). Scale bars 100nm.

For the H-NS only condition, we also observed different bridged DNA duplex structures on the HT and HH linear fragments (Figure S6D). For the HT fragment, the bridged duplexes were associated with protruding ends of unequal lengths, consistent with H-NS bridging of the NRE sequences located once at the end and once in the middle of the fragment. For the HH fragment, the bridged duplexes did not demonstrate any protruding ends, consistent with a previously observed configuration (Japaridze et al., 2017b), in which the two adjacent NRE regions engaging all the four consecutive high-affinity nucleation sites are collinearly intertwined (Figure 4D). Again, the bridged regions were of different lengths, being shorter in HT arrangement, Lbridge_HT = 45 ± 25 nm (N = 43, error is SD) compared to HH Lbridge_HH = 65 ± 40 nm (N = 36, error is SD). What was surprising to see was that on both the HT and HH fragments, the looped DNA regions connecting the H-NS-bridged duplexes appeared to be also covered by H-NS (Figure S8). This demonstrated that under the same environmental conditions H-NS could both polymerize on individual DNA duplexes (Lim et al., 2012) as well as form bridges between two DNA double strands (Dame et al., 2000; Japaridze et al., 2017a, 2017b; Maurer et al., 2009; van der Valk et al., 2017).

When a mixture of FIS and H-NS was incubated with linear HT fragments, no regular structures were observed, as was the case with circular HT DNA (Figure 4E). Strikingly, with HH fragments however, the binding of FIS and H-NS again formed hairpin-like structures consisting of a stem of bridged DNA duplexes associated with a loop, resembling those formed with circular HH DNA (Figures 4F and 5). Measurements revealed uniform heights across the H-NS bridged regions, whereas blobs of greater heights were associated with the loops, indicative of FIS binding. Interestingly, in contrast to circular DNA, FIS was binding not at the base but rather at the apex of the loop. We infer that the differences observed between the structural organization of nucleoprotein complexes formed with HT and HH constructs are driven by distinct spatial arrangements of the UAS and NRE elements containing the FIS and H-NS binding sites (Figures 4H and 5).

Figure 5

FIS and H-NS proteins phase separate when binding to the DNA

(A) Schematic depiction of the Head-to-Head (HH) construct. Blue indicates the NRE and red UAS regions.

(B–E) Typical magnified AFM images of HH fragments with simultaneously bound H-NS and FIS proteins. White arrows indicate the position of FIS proteins (indicated in blue). Scale bars 100nm.

Discussion

While major factors involved in regulating the global architecture of the bacterial genome become progressively resolved (Japaridze et al., 2020; Lioy et al., 2018; Mäkelä and Sherratt, 2020; Wu et al., 2019a, 2019b), the smaller-scale local architectures remains largely obscure. Given the large variations in the abundance and composition of NAPs during the bacterial growth cycle (Azam et al., 1999), it is conceivable that the nucleoid undergoes reshaping (Meyer and Grainger, 2013) with corresponding changes in the local architecture depending on the metabolic state of the cell.

In this study, we set out to investigate the role of local DNA sequence organization in the assembly of higher-order nucleoprotein structures formed by simultaneous binding of two abundant bacterial NAPs, FIS and H-NS, to different spatial arrangements of their cognate DNA binding sites. We chose FIS and H-NS because of their competing functions in vivo. The global “genomic silencer” H-NS forms tightly interwound plectonemic structures which impede transcription (Chen and Wu, 2005; Dame et al., 2001; Dillon and Dorman, 2010; Kotlajich et al., 2015), whereas more open DNA structures, such as toroids and DNA loops stabilized by the global transcriptional activator FIS are conducive to transcription initiation (Maurer et al., 2006; Muskhelishvili et al., 1995; Skoko et al., 2006). Furthermore, both FIS and H-NS were found to be simultaneously present in transcriptionally active regions in vivo (Grainger et al., 2006) where they apparently compete with each other for binding by stabilizing different DNA topologies (Afflerbach et al., 1999).

To study the nucleoprotein complexes formed by the combination of FIS and H-NS, we used two nicked DNA plasmids with a spatially different HH or HT arrangement of the NAP binding sites. Binding of individual NAPs to the circular plasmids demonstrated a conspicuous difference between the architectures of the FIS and H-NS nucleoprotein complexes, fully consistent with previous microscopic observations using supercoiled DNA (Japaridze et al., 2017a; Schneider et al., 2001). Interestingly, in contrast to H-NS, we did not find any clear differences between nucleoprotein complexes formed on the two substrates with FIS alone. However, when a combination of FIS and H-NS was used, we found that various arrangements led to structurally distinct nucleoprotein complexes. In particular, binding of FIS and H-NS to the HH arrangement of sites produced specific hairpin-like structures in which the bridged duplexes stabilized by H-NS were associated with a bent DNA loop apparently stabilized by the binding of FIS, as indicated by the AFM height measurements (Figures 2 and S2). This observation is consistent with earlier studies showing that FIS can stabilize both DNA loops and crossovers (Schneider et al., 2001; Skoko et al., 2006) and with alternative DNA curvatures implicated in the competition between FIS and H-NS binding (Afflerbach et al., 1999). When we repeated the experiments with linear DNA fragments containing just the sequences with high-affinity protein binding sites, distinct hairpin structures were observed with the HH but not the HT substrate, although in this case FIS was apparently binding at the apex rather than at the base of the loop (Figures 4 and 5). We surmise that this difference might be explained by the so-called “antenna effect,” where increase in the DNA substrate length can lead to an increased probability of protein binding to specific sequences (Shimamoto et al., 2020). However, since we could not observe such regular structures with the HT arrangement of binding sites (nor with a control PBR DNA without any strong binding sites) with either the circular or linear DNA substrates, we infer that their peculiar architecture is a result of crosstalk between FIS and H-NS driven by the specific spatial organization of the binding sequences.

In nanopore experiments, we found that the HH nucleoprotein complexes formed by FIS and H-NS demonstrated regular current blockade levels in contrast to HT (Figures 3 and S5) consistent with the AFM data suggesting higher structural regularity of the HH compared to that of the HT nucleoprotein complexes.

The organization of HH nucleoprotein complexes was shown to be due to H-NS bridging of the two adjacent NRE elements that contain high-affinity H-NS binding sites and plectonemic coiling of the DNA (Japaridze et al., 2017b). This configuration leads to tightly bent small DNA loops that are bound by FIS, as observed in our experiments (Figures 2 and 4). While at the present stage we cannot distinguish the temporal pattern of H-NS and FIS binding, such tight loops might be especially attractive for further FIS binding. First, a tightly bent DNA loop would facilitate FIS binding due to the unusually short distance between the helix-turn-helix motifs of FIS and, accordingly, a strong preference for tightly bent DNA substrates with compressed minor groove (Stella et al., 2010). Second, while the NREs demonstrate a sequence periodicity of ∼11 bp, H-NS-bridging of the adjacent NREs in HH construct constrains negative superhelicity, most likely due to the right-handed plectonemic coiling and stabilization of high negative twist (Japaridze et al., 2017b). This would lead to compensatory over-twisting of the protruding loop reducing the helical repeat to values < 10.5 bp, consistent with the assumed preferential value for FIS binding of ∼10.2 bp (Travers and Muskhelishvili, 2013). This over-twisting would also counteract the H-NS binding within the tight loop stabilized by FIS and lead to the organization of structures observed in Figure 5, in which FIS is positioned in the looped DNA regions surrounded by H-NS bridged DNA filaments. The loop formed between the adjacent NRE elements in HH constructs is devoid of strong FIS binding sites and we observed that FIS binds at different positions in the loop, supporting the view that it is the DNA shape rather than DNA sequence that dictates FIS binding. Interestingly, in the absence of FIS, H-NS is capable of polymerizing on the tightly bent loop connecting the bridged helices (Figures 4D and S8), meaning that FIS hinders the spread of H-NS. While this process may depend on the size and topology of the DNA substrate, further studies of nucleoprotein complexes formed with successive addition of FIS and H-NS on both circular and linear substrates will help to clarify this question. The exact mechanism of this cooperation between FIS and H-NS is not yet resolved, but it is likely mediated by the DNA structure and topology. The local arrangement of FIS and H-NS proteins in the HH hairpins is non-random. This suggests that some form of DNA-driven phase separation could drive a differentiation of the bacterial chromatin architecture (Barbieri et al., 2012; Brackley and Marenduzzo, 2020). The heterogeneity in binding site arrangement could be utilized by abundant DNA architectural proteins to drive the differentiation of chromatin architecture and gene transcription as observed for H-NS filaments both in vivo and in vitro by the Landick group (Shen et al., 2020). Notably, recent studies demonstrated that the physiological B-form DNA can undergo structural changes resulting in non-canonical DNA forms under the influence of DNA supercoiling (Pyne et al., 2021), spatial confinement (Japaridze et al., 2017c), and DNA sequence (Du and Zhou, 2013; Dumat et al., 2016), indicating that the DNA structure itself is highly heterogeneous and can be favorable for protein binding based on structure rather than sequence. Our observations are in keeping with the reported capacity of FIS and H-NS to form topological barriers in the chromosome (Hardy and Cozzarelli, 2005). Looped regions stabilized by FIS that are not bound by H-NS, could potentially create patches of “open” and “closed” DNA regions differentially accessible to transcription machinery (Kane and Dorman, 2011). These coexisting structures are in line with the predominantly repressing and activating roles of H-NS and FIS on transcription, respectively. Since the distribution of the NAP binding sites in genomes of bacteria is non-random (Dillon et al., 2012; Lang et al., 2007; Ussery et al., 2001), we hypothesize that the growth-phase-dependent regulation of transcription (Blot et al., 2006) involves global alterations of chromatin accessibility. This accessibility is modulated by a crosstalk between temporally changing composition of NAPs and a non-random distribution of their cognate genomic binding sites, thus leading to spatiotemporal organization of the “open” and “closed” DNA configurations pertinent to genomic expression.

Limitations of the study

AFM imaging does not permit base-pair DNA resolution and gives only the overall morphology without knowing the exact bp positions where the proteins are bound. Furthermore, DNA in bacteria is typically supercoiled, indicating that conclusions made for linear DNA should be considered with caution when extrapolating to supercoiled DNA. The protein-DNA binding modes for H-NS and FIS as well as DNA binding strength are strongly salt dependent. High-salt experiments are required for the signal in nanopore sensing. For these nanopore experiments, the proteins were pre-bound and then transferred to the high salt condition, which in turn can lead to the dissociation of proteins from the DNA over time which places a limitation of measurement time.

To assess the relevancy of the in vivo observations, it would be useful to have an estimate of the natural occurrence of such HH and HT configurations in bacterial genomes. These proposals merit further independent investigations.

Resource availability

Lead contact

Information and requests for resources should be directed to and will be fulfilled by the lead contact, Aleksandre Japaridze (a.japaridze@tudelft.nl).

Materials availability

This study generated no unique materials.

Data and code availability

All relevant data are available from the lead contact upon request. Original AFM data underlying the data reported in Figures 1,2,4 and 5 in the main text are available online via Mendeley Data repository with DOI link at https://doi.org/10.17632/4xvnyx47b5.1.

Methods

All methods can be found in the accompanying Transparentmethods supplemental file.