Mechanism of environmentally driven conformational changes that modulate H-NS DNA-bridging activity

Bacteria frequently need to adapt to altered environmental conditions. Adaptation requires changes in gene expression, often mediated by global regulators of transcription. The nucleoid-associated protein H-NS is a key global regulator in Gram-negative bacteria and is believed to be a crucial player in bacterial chromatin organization via its DNA-bridging activity. H-NS activity in vivo is modulated by physico-chemical factors (osmolarity, pH, temperature) and interaction partners. Mechanistically, it is unclear how functional modulation of H-NS by such factors is achieved. Here, we show that a diverse spectrum of H-NS modulators alter the DNA-bridging activity of H-NS. Changes in monovalent and divalent ion concentrations drive an abrupt switch between a bridging and non-bridging DNA-binding mode. Similarly, synergistic and antagonistic co-regulators modulate the DNA-bridging efficiency. Structural studies suggest a conserved mechanism: H-NS switches between a ‘closed’ and an ‘open’, bridging competent, conformation driven by environmental cues and interaction partners.

A key protein in nucleoid organization of Gram-negative bacteria is the Histone-like Nucleoid Structuring protein (H-NS). Genome-wide binding studies have revealed that H-NS binds along the genome in long patches (Grainger et al., 2006;Kahramanoglou et al., 2011;Lucchini et al., 2006;Navarre, 2006;Oshima et al., 2006), which have been proposed to mediate the formation of genomic loops (Noom et al., 2007;van der Valk et al., 2014). H-NS is also an important regulator of global gene expression, implied in mediating global transcriptional responses to environmental stimuli (osmolarity, pH, temperature) (Atlung and Ingmer, 1997), and operating as xenogeneic silencer, silencing horizontally integrated DNA . A large fraction of Escherichia coli and Salmonella genes (5-10%) is thus regulated (usually repressed) by the action of H-NS. H-NS operation is modulated by environmental stimuli and through interplay with other proteins (Atlung and Ingmer, 1997;Stoebel et al., 2008). In solution, the H-NS protein exists as a dimer, which oligomerizes at high concentrations (Ceschini et al., 2000;Spurio et al., 1997). H-NS consists of three structural domains: a C-terminal domain responsible for DNA binding (Shindo et al., 1995), a N-terminal dimerization domain Cerdan et al., 2003;Esposito et al., 2002;Ueguchi et al., 1996) and a central dimer-dimer interaction domain responsible for multimer formation (Arold et al., 2010;Leonard et al., 2009). These two interaction domains are connected by a long a-helix (Arold et al., 2010) (helix a3). H-NS exhibits two seemingly distinct DNA-binding modes: DNA bridging (Dame et al., 2006;Dame et al., 2000;Dame et al., 2001;Dame et al., 2002;Schneider et al., 2001), the condensation of DNA by intra-and inter-molecular DNA binding by H-NS and DNA stiffening, the rigidification of DNA through the formation of a H-NS-DNA filament (Amit et al., 2003;Dame and Wuite, 2003;Liu et al., 2010). These modes have been attributed to the basic functional H-NS unit (a dimer) binding to DNA either in cis or in trans (Wiggins et al., 2009;Joyeux and Vreede, 2013). H-NS paralogues StpA, Sfh, Hfp, and truncated derivatives such as H-NST, have been proposed to modulate H-NS function by forming heteromers with H-NS (Baños et al., 2008;Deighan et al., 2003;Müller et al., 2010;Williams et al., 1996;Williamson and Free, 2005), with DNA-binding properties different from homomeric H-NS. Members of the Hha/YmoA family of proteins are H-NS co-regulators with limited sequence homology to H-NS (Madrid et al., 2007). At many targets along the genome, H-NS and Hha co-localize. Localization of Hha at these sites is strictly H-NS dependent, whereas the genome-wide binding pattern of H-NS is only mildly affected by Hha (Ueda et al., 2013).
Although evidence has been put forward that the concentration of divalent ions determines the binding mode of H-NS (Liu et al., 2010), a mechanistic explanation is lacking. Moreover, the possible effect of co-regulators of H-NS, such as Hha, on these binding modes has remained unexplored. To obtain a better understanding of the molecular basis underlying the H-NS binding modes, it is crucial to determine the effects of ion valence and concentration, as well as the presence of helper proteins on both the stiffening and the bridging mode. Here, we investigate DNA stiffening on short DNA tethers using Tethered Particle Motion (TPM). As intramolecular DNA bridging does not occur on short DNA tethers, DNA stiffening can be uncoupled from DNA bridging. In addition, to accurately determine intermolecular DNA-bridging efficiencies in solution, we developed a sensitive quantitative bulk assay. Using these two assays, we unravel the assembly pathway of bridged DNA-H-NS-DNA complexes and the roles of mono-and divalent ions, helper proteins Hha and YdgT, and truncated H-NS derivatives. Finally, Molecular Dynamics (MD) simulations reveal that ions and interacting proteins directly alter H-NS structure from a 'closed' bridging incapable to an 'open' bridging eLife digest The genetic information every cell needs to work properly is encoded in molecules of DNA that are much longer than the cell itself. A key challenge in biology is to understand how DNA is organized to fit inside each cell, whilst still providing access to the information that it contains. Since the way DNA is organized can influence which genes are active, rearranging DNA plays an important role in controlling how cells behave.
In Escherichia coli and many other bacteria, a protein called H-NS contributes to DNA reorganization by forming or rupturing loops in the DNA in response to changes in temperature, the levels of salt and other aspects of the cell's surroundings. In controlling loop formation, it dictates whether specific genes are switched on or off. However, it remains unclear how H-NS detects the environmental changes.
To address this question, van der Valk et al. used biochemical techniques to study the activity of H-NS from E. coli under different environmental conditions. The experiments show that changes in the environment cause structural changes to H-NS, altering its ability to form DNA loops. A previously unnoticed region of the protein acts as a switch to control these structural changes, and ultimately affects which genes are active in the cell.
These findings shed new light on how bacteria organize their DNA and the strategies they have developed to adapt to different environments. The new protein region identified in H-NS may also be present in similar proteins found in other organisms. In the future, this knowledge may ultimately help to develop new antibiotic drugs that target H-NS proteins in bacteria.

Results
The role of Mg 2+ and H-NS multimerization in DNA bridging and DNA stiffening In order to dissect the role of divalent ions in the formation of bridged and stiffened complexes, we applied a novel, sensitive, and quantitative DNA-bridging assay and carried out TPM experiments (providing a quantitative and selective readout of DNA stiffening). The DNA-bridging assay relies on immobilization of bait DNA on magnetic microparticles and the capture and detection of 32 P labeled prey DNA if DNA-DNA bridge formation occurs (see Figure 1-figure supplement 4b for a schematic depiction of the assay). 80% of initial prey DNA is recovered at high H-NS concentrations (see Figure 1a). In the absence of either the H-NS protein or bait DNA, no prey DNA is recovered under our experimental conditions. Next, we used this assay to quantify the DNA-bridging efficiency of H-NS as a function of the amount of Mg 2+ ions (see Figure 1b), reproducing the qualitative results of Liu et al. (2010) and providing independent confirmation of the previously observed effects. The concentration range from 0 to 10 mM Mg 2+ is considered to be physiologically relevant (Hurwitz and Rosano, 1967). Importantly, the transition from no bridging to complete bridging is abrupt between 4-6 mM Mg 2+ , indicating that changes in Mg 2+ concentration might drive a binary switch.
Earlier studies have shown that H-NS binding along a single DNA molecule results in DNA stiffening (Amit et al., 2003;Liu et al., 2010). However, due to the co-occurrence of DNA bridging, previous studies were incapable of measuring DNA stiffening in the presence of Mg 2+ . Here, we used TPM to investigate DNA stiffening as a function of Mg 2+ concentration. In TPM, the Root Mean Square displacement (RMS) of bead movement is a direct reflection of tether stiffness and length ( Figure 1-figure supplement 4a). DNA-binding proteins can affect both, but previous studies have shown that the DNA contour length is not affected by binding of H-NS (Dame et al., 2006;Dame et al., 2001). Thus, an increase in stiffness due to H-NS binding translates into a higher RMS value of a DNA tether ( Figure 1d). Here, we measured the effects of H-NS on DNA stiffness in the absence and presence of 10 mM Mg 2+ and confirmed that H-NS stiffens DNA (Amit et al., 2003;Liu et al., 2010); importantly, our experiments reveal that Mg 2+ does not affect the stiffness of the fully formed H-NS-DNA complexes at saturation, as in both conditions the RMS is the same (Figure 1d). Analysis of the binding characteristics using the McGhee-von Hippel equation revealed that the association constant of H-NS is somewhat reduced in the presence of Mg 2+ , while cooperativity increases under these conditions (Figure 3-figure supplement 1a,d). The reduction in DNAbinding affinity of H-NS may be attributed to shielding of the negatively charged phosphate backbone by Mg 2+ .
The cooperative binding of H-NS and DNA stiffening observed by TPM suggest that H-NS multimerizes along DNA, likely via the recently defined dimer-dimer interaction domain (Arold et al., 2010). Multimerization along DNA has been previously suggested (Esposito et al., 2002;Williams et al., 1996) but has never been conclusively demonstrated. To test this hypothesis, we generated a mutant, H-NS Y61DM64D , predicted to have disrupted dimer-dimer interaction based on the H-NS 1-83 crystal structure (Arold et al., 2010). Size exclusion chromatography showed that this H-NS mutant indeed exists solely as a dimer in solution independent of protein concentration ( cooperativity due to high local DNA concentration adjacent to existing bridges (Dame et al., 2006;Dame et al., 2000), is insufficient to explain the formation of bridged DNA-H-NS-DNA complexes. The nature of the effect of Mg 2+ on DNA-bridging efficiency is not understood. An increase in affinity would be expected if Mg 2+ would only facilitate interactions between the DNA phosphate backbone and negatively charged residues on H-NS, but this is not observed (

Modulation of DNA bridging by osmotic factors
Although it has long been known that the expression of some H-NS controlled genes (such as the proU operon) is modulated by the osmolarity of the medium (Cairney et al., 1985), the underlying mechanism remains undetermined. Previous studies have revealed that the H-NS DNA stiffening mode is mildly sensitive to the KCl concentration (Amit et al., 2003;Liu et al., 2010). Using TPM, we were able to confirm these observations. The reduction in DNA stiffening is gradual (Dame and Wuite, 2003) and modest ( Figure 1f). It is thus questionable whether the multimerization of H-NS along DNA alone is sufficient to explain its role in repression of transcription (and modulation thereof). Could the modulation of gene repression be due to ionic effects on DNA-bridging efficiency? Using our DNA-bridging assay, we observed complete abolishment of H-NS DNA bridging by KCl at concentrations exceeding 120 mM (Figure 1c), a binary response, similar to what we observed for the Mg 2+ titration. This in vitro observation mirrors the in vivo response of the ProU operon, at which KCl concentrations exceeding 100 mM are required to alleviate H-NS-mediated repression (Cairney et al., 1985). Control experiments using K-glutamate (Figure 1-figure supplement 5) confirm that K + , and not the counter-ion, is responsible for the observed effects, even though the counter ion may affect the DNA-binding affinity of the protein (Figure 1-figure supplements 5c and Leirmo et al., 1987) This strong and abrupt effect on DNA bridging, while leaving DNA stiffening essentially unaffected, might indicate that H-NS reverts to the 'closed' conformation by the addition of K + . To investigate this effect at a structural level we performed MD simulations at In addition to interactions between the DNA-binding domain and the dimerization domain, the two DNA-binding domains in the dimer interact with each other. Moreover, the DNA-binding domains interact with helix a3. In particular regions 98-105 (KYSYVDENGE) and 123-129 (EQGKS) are involved in these interactions. The average distance between the donors and acceptors in the hydrogen bonds within helix a3 (Figure 2b) indicates that that buckles in helix a3 also occur at a high KCl concentration, at the same location as determined by low-salt conditions. These observations indicate that the 'closed' state can have multiple forms (see

Modulation of H-NS by Hha and YdgT
Gene regulation by H-NS often occurs in conjunction with other proteins; these co-regulators are known to interact with H-NS at specific loci along the genome. Two such proteins, Hha, and YdgT, are members of the Hha/YmoA (Madrid et al., 2007) family of proteins. In order to understand the modulation of H-NS function by these proteins we investigated their influence on the H-NS DNA binding modes. We observed that Hha, when added at equimolar concentrations, enhances DNA bridging by H-NS at low Mg 2+ concentrations (Figure 3c).  presence of Mg 2+ and YdgT, H-NS causes 'DNA collapse' in TPM (data not shown), similar to earlier observations for H-NS in the presence of Mg 2+ without (Liu et al., 2010;Wang et al., 2014) or with Hha added (Wang et al., 2014). This 'DNA collapse' is attributed to H-NS-mediated DNA bridge formation. One possible explanation for the effects of Hha and YdgT, is that they effectively increase the DNA-binding affinity of H-NS (Ali et al., 2013). To investigate whether Hha affects H-NS conformation, we performed MD simulations, incorporating structural information from the recently resolved H-NS 1-43 -Hha co-crystal structure (Ali et al., 2013). Our MD simulations reveal that Hha does not prevent buckles in helix a3 (see Figure 2b).

Discussion
It has been known for many years that H-NS binding induces gene silencing. H-NS activity in vivo is modulated by physico-chemical factors (osmolarity, pH, temperature) and interaction partners. These findings support the hypothesis that H-NS plays a role in environmental adaptation. However, mechanistically it is unclear how functional modulation of H-NS by such factors is achieved. Based on our findings, we conclude that H-NS is incapable of bridging or stiffening DNA as dimers. H-NS dimers bind DNA in cis (Dame et al., 2006;Dame et al., 2000) and associate side-by-side along DNA, likely via the recently identified dimer-dimer interaction domain (Arold et al., 2010), resulting in DNA stiffening. This process is cooperative as H-NS dimers interact with neighbors, as well as with DNA. Our studies reveal that H-NS-DNA filaments are structurally very similar, independent of the presence of Mg 2+ (Figure 1d). But functionally, these H-NS-DNA filaments are distinct. H-NS can be 'activated' by Mg 2+ , which promotes a conformational change, rendering both DNA-binding domains of H-NS dimers accessible for DNA bridging. In our model, the assembly of bridged complexes proceeds in distinct steps: (1) nucleation (Lang et al., 2007) (binding of an H-NS dimer at a high affinity site), (2) lateral filament growth by H-NS dimer-dimer interactions (leading to DNA stiffening) and (3) bridging of the assembled filament to bare DNA provided in trans (Figure 4). Each step can potentially be modulated by osmolarity and protein interaction partners. Here, we show that these factors most effectively target DNA bridging.
What are the implications of our observations? Our observations add to the large body of evidence showing that regulation of transcription via H-NS is complex, and that it does not proceed via a single, simple, well-defined mechanism. The most straightforward form of repression by H-NS is via occlusion of RNA polymerase from the promoter (Lim et al., 2012;Prosseda et al., 2004;Gö ransson et al., 1990). Whether this mechanism of repression involves lateral filament formation or bridging is unclear. It is expected that both types of complexes assembled at a promoter site can in principle occlude RNA polymerase. A second mechanism of repression is to trap RNA polymerase at the promoter, preventing promoter escape (Schrö der and Wagner, 2000;Shin et al., 2005). RNA polymerase trapping likely involves DNA bridging of promoter upstream and downstream elements (Dame et al., 2002;Shin et al., 2005). A third mechanism of repression by H-NS is to interfere with RNA polymerase progression during active transcription by intragenic binding (Dole et al., 2004). In this model, both modes could interfere with transcription. However, it was suggested recently that only bridged filaments are capable of interfering with transcription (Kotlajich et al., 2015), with lateral filaments likely being disassembled as RNA polymerase encounters them. Generally, the type of complex formed by H-NS is expected to depend on the type and number of ions, combined with local DNA conformation (dependent on DNA sequence or DNA topology), and the presence of modulating proteins. The interplay between these factors will determine the strength of the complex, and degree of repression. Thus, different H-NS repressed genes are expected to be subject to different types of modulation, providing a key to a coordinated response in gene expression to altered conditions and selectivity for the interplay with specific co-regulators at specific target regions.

Construction of expression vectors H-NS expression vector
The vector pRD18 for expression of H-NS was constructed by inserting the PCR amplified hns gene into the pET3His overexpression vector using NdeI and XhoI restriction sites. Using the XhoI restriction site, the encoded protein does not contain a C-terminal His-tag.

H-NS Y61DM64D expression vector
The vector pRD69 for expression of H-NS Y61DM64D was constructed by inserting a PCR fragment containing the hns gene mutated to encode aspartic acid instead of tyrosine/methionine at position 61 and 64 into pET3His using NdeI and XhoI restriction sites.

Hha expression vector
The vector pRD38 for expression of N-terminally His tagged Hha was constructed by inserting a PCR fragment containing the hha gene into pET3His using XhoI and BamHI restriction sites.

YdgT expression vector
The vector pRD39 for expression of N-terminally His tagged YdgT was constructed by inserting a PCR fragment containing the ydgT gene into pET3His using XhoI and BamHI restriction sites.

Protein overproduction and purification
H-NS/H-NS Y61DM64D /H-NS E43A,E44A,S45A . BL21 (DE3) (RRID: WB-STRAIN: HT115(DE3)) Dhns::kan/frt pLysE (NT201, our lab) cells transformed with plasmids expressing H-NS/H-NS mutants were grown to an OD 600 of 0.4, and induced for 2 hr using IPTG (500 mM). For the H-N E43A,E44A,S45A protein, the cells were co-transformed with pRD252, coding for LacI to help suppress leaky expression of H-N E43A,E44A,S45A . The cells were pelleted and lysed by sonication in 100 mM NH 4 Cl, 20 mM Tris pH 7.2, 10% glycerol, 8 mM b-mercaptoethanol, 3 mM benzamidine). The soluble fraction was loaded onto a P11 column and eluted using a 100 mM-1 M NH 4 Cl gradient, the protein eluted at 280 mM NH 4 Cl. The peak fractions were dialysed to buffer B (identical to buffer A, but containing 130 mM NaCl instead of NH 4 Cl) by overnight dialysis. The dialysate was loaded onto a heparin column (GE Healthcare) and eluted using a 130 mM-1 M NaCl gradient, the protein eluted at 350 mM NaCl. The pooled peak fractions were dialysed to buffer B and concentrated using a 1 ml Resource-Q column (GE Healthcare). The purity of the protein was verified on an SDS-PAGE gel. The protein concentration was determined using a Bicinchoninic Acid assay (Pierce BCA protein assay kit, Thermo Scientific). H-NS 1-58 BL21 (DE3) (RRID: WB-STRAIN: HT115(DE3)) Dhns::kan/frt pLysE (NT210, our lab) cells transformed with plasmids expressing H-NS/H-NS mutants were grown to an OD 600 of 0.4, and induced for 2 hr using IPTG (500 mM). The cells were pelleted and lysed by sonication in 100 mM NaCl, 20 mM Tris pH 7.2, 10% glycerol, 8 mM b-mercaptoethanol, 3 mM benzamidine). The soluble fraction was heated to 65˚C for 10 min and then spun down at 10.000 RPM for 10 min. The supernatant was collected and a 1:1 ratio saturated ammonium sulfate (50 mM Tris pH 7.2, 4M ammonium sulfate) was gradually added to the cooled sample. The sample was spun down at 8.000 RPM for 15 min and a 1:1 ratio of 5 mM Tris pH 7.2, 15% glycerol was added to the supernatant. To remove further impurities, the sample was run through a 1 ml hydrophobic interaction column and 1 ml Blue-agarose column (the protein should not bind to either of these column) before finally binding the protein to 1 ml Resource-Q column (GE Healthcare). The protein was eluted with a 25 mM À1M gradient of NaCl, the protein eluted at roughly 380 mM NaCl. The purity of the protein was verified on an SDS-PAGE gel. The protein concentration was determined using a Bicinchoninic Acid assay (Pierce BCA protein assay kit, Thermo Scientific).

Hha/YdgT
BL21 (DE3) (RRID: WB-STRAIN: HT115(DE3)) Dhns::frt, hha::kan, pLysE (our lab) cells transformed with plasmids pRD38/pRD39 expressing hha/ydgT were grown at 37˚C to an OD 600 of 0.4, and induced for two hours using IPTG (500 mM). The cells were pelleted and lysed in 20 mM HEPES pH 7.9, 1 M KCl, 10% glycerol, 8 mM b-mercaptoethanol. The soluble fraction was loaded onto a Ni-column. The column was first washed with buffer D (20 mM HEPES pH 7.9, 0.5 M KCl, 10% glycerol, 8 mM b-mercaptoethanol). The protein was then eluted using a 0 mM-0.5 M Imidazole gradient, the protein eluted at 300 mM Imidazole. The peak fractions were dialysed to buffer E (identical to buffer D, but containing 100 mM KCl) by overnight dialysis. The sample was then loaded onto pre-equilibrated SP Hi-Trap-column and Ni-column connected in series. After loading the samples on the column, the SP Hi-Trap -column was disconnected and the protein was eluted from the Ni-column using a 0-0.5 M imidazole gradient. The purity of the protein was verified on an SDS-PAGE gel. The protein concentration was determined using a Bicinchoninic Acid assay.

Peptide production and purification
A truncated form of H-NS (H-NS  was synthesized by way of automated solid phase synthesis using standard protocols via Fmoc-strategy. Purification was performed by RP-HPLC with a Gemini 5m C18 reversed phase column. Identity of the peptides was determined via MALDI-MS. The purity was determined by means of analytical RP-HPLC. The peptide was freeze dried and dissolved in 20 mM Tris pH 7.2, 300 mM KCl, 10% glycerol, 8 mM b-mercaptoethanol. The peptide concentration was determined using a Bicinchoninic Acid assay (Pierce BCA protein assay kit, Thermo Scientific).

Size exclusion chromatography
Size exclusion chromatography was done using a Superose-12 column with a flow of 0.3 ml/min, pre-equilibrated with 10 mM Tris-HCl, 50 mM KCl, 5% glycerol containing or lacking 10 mM MgCl 2 . The absorbance of the eluting fractions was measured at 215 nm. These experiments were performed in triplicate.

DNA substrates DNA preparation
All experiments were performed using a random, AT-rich, 685 bp (32% GC) DNA substrate (Laurens et al., 2012). The DNA substrate was generated by PCR, and the PCR products were purified using a GenElute PCR Clean-up kit (Sigma-Aldrich). If required, DNA was 32 P-labeled as described previously (Wagner et al., 2011). resuspended in 200 mL CB containing 100 nM biotinylated DNA. Next, the bead suspensions were incubated for 30 min on a rotary shaker (1000 rpm) at 25˚C. After incubation, the beads were washed twice with Incubation buffer (IB: 10 mM Tris-HCl pH 8.0, 50 mM KCl, 10* mM MgCl 2 , 5% v/v Glycerol, 1 mM DTT and 1 mM Spermidine) before resuspension in IB and addition of ±8000 cpm of radioactively labeled 32 P 685 bp DNA. Radioactive DNA was supplemented with unlabeled 685 bp DNA to maintain a constant (20 nM) DNA concentration. The DNA-bridging protein H-NS (concentrations indicated in the text), and if applicable Hha or YdgT were added and the mixture was incubated for 30 min on a shaker (1000 rpm) at 25˚C. To remove unbridged prey DNA, the beads were washed with IB, before resuspension in 12 mL stop buffer (10 mM Tris pH 8.0, 1 mM EDTA, 200 mM NaCl, 0.2% SDS). All samples were quantified through liquid scintillation counting over 10 min. All values recovered from the DNA-bridging assay were corrected for background signal (using a sample lacking H-NS), and normalized to a reference sample containing the amount of labeled 32 P 685 bp DNA used in the assay. The samples were then run on a 5% 0.5x TBE gel to ensure DNA integrity. DNA bridging was calculated based on a reference sample containing 2 mL of prey DNA. All DNA-bridging experiments were performed in triplicate. Unless indicated otherwise all DNA-bridging experiments were performed in the presence of 10 mM of MgCl 2 and 3,3 mM of H-NS (10% more H-NS than is required for saturation -see Figure 1A). Each experiment contains 50 mM KCl, to which additional KCl or K-glutamate are added depending on the experimental condition tested.

Tethered particle motion experiments
Tethered particle motion experiments were performed as reported previously (Driessen et al., 2014;. Flow cells were prepared as described with minor modifications (Driessen et al., 2014;. Here, before flowing in protein diluted in the experimental buffer (10 mM Tris-HCl pH 8.0, 50 mM KCl, 10 mM MgCl 2 /EDTA, 5% v/v Glycerol, 1 mM DTT) the flow cell was washed using 4 flow cell volumes with the experimental buffer. Next, the flow cell was incubated for 10 min with protein solution before sealing the flow cell. The flow cell was maintained at a constant temperature of 25˚C. Measurements were started 10 min after the introduction of protein solution. TPM experiments were done at least in duplicate. The data were analyzed as previously described (Driessen et al., 2014;. The RMS was first converted to persistence length(L p ) using Equation 1 and d ¼ n# Here, the association constant(K) is described as a function of the protein concentration (c) and a cooperativity parameter (!Þ. van  To this end, weighted orthogonal distance regression (ODR) was performed to estimate the parameters of the nonlinear implicit equation describing the cooperative ligand binding. The binding site size (n) of H-NS was fixed to a value of 30 bp during regression, which corresponds to values determined previously (Dame et al., 2006;Amit et al., 2003). The association constant (K) and the cooperativity parameter (w) were assumed to be positive real numbers. A custom fitting routine was implemented in Fortran and makes use of the ODRPACK library (Boggs et al., 1992).

Molecular dynamics simulations
The starting conformation of the full-length H-NS dimer was constructed as described previously (van der Valk et al., 2014). The system was placed in a periodic dodecahedron box with a distance of at least 0.8 nm between the box edge and the most extended atom of the protein dimer, followed by the addition of water and ions. With this system we performed Molecular Dynamics (MD) simulations of full length H-NS at different concentrations of KCl, and MgCl 2 and with the addition of Hha, summing up to a total of six different systems. Hha was added by aligning the crystal structure containing the H-NS -Hha complex (PDB code 4ICG [Ali et al., 2013]) with the full-length structural model and copying the Hha molecules. System size ranged from 513457 atoms for the Hha systems to around 1,135,000 atoms for the other four systems.
Interactions between atoms were described by the AMBER99-SB-ILDN force field (Lindorff-Larsen et al., 2010), in combination with the TIP3P water model (Jorgensen et al., 1983). Longrange electrostatic interactions were treated via the Particle Mesh Ewald method (Darden et al., 1993;Essmann et al., 1995) with a short-range electrostatic cutoff distance at 1.1 nm. Van der Waals interactions were cut off at 1.1 nm. Preparation of the systems consisted of energy minimization equilibration of the solvent. Energy minimization was performed by the conjugate gradient method. After energy minimization, the positions of water molecules and ions were equilibrated by a 1 ns molecular dynamics run at a temperature of 298 K and a pressure of 1 bar in which the heavy atoms in the protein were position-restrained with a force constant in each direction of 1000 kJ/mol nm. After preparation, we performed 16 50 ns runs for each system, varying initial conditions by assigning new random starting velocities drawn from the Maxwell-Boltzmann distribution at 298 K. All simulations were performed with GROMACS v.4.6.3 (Pronk et al., 2013) at the Dutch National Supercomputer with the leap-frog integration scheme and a time step of 2 fs, using LINCS (Hess et al., 1997) to constrain all bonds. All simulations were performed in the isothermal-isobaric ensemble at a pressure of 1 bar, using the v-rescale thermostat (Bussi et al., 2007) and the Parrinello-Rahman barostat (Parrinello and Rahman, 1981).
Frames were stored every 10 ps. The first 10 ns of each simulation are excluded from analysis, unless stated otherwise. Analysis focused on determining contacts between domains, between H-NS and Hha, between H-NS and ions, and helical hydrogen bonds. Contact maps of interactions between residues in the H-NS dimer system were obtained by first calculating the minimum distance between each residue pair in the system. A residue pair is counted to be in contact if they are at a minimum distance of 0.6 nm or less. The probability of a contact is then calculated as the average over all 16 simulations (excluding the first 10 ns) and displayed as a contact probability matrix. We used a modified version of the g_mdmat tool in GROMACS (Pronk et al., 2013) in combination with Perl scripts to generate contact maps. To determine the location of ions with respect to the H-NS system, we calculated the minimum distance between each residue in the H-NS dimer and the ions and counted a contact if the distance between an H-NS residue and an ion is 0.6 nm or less. These contact probabilities (P Mg2+ , P K+ and P Cl-) are averaged over all 16 simulations and the two monomers. A similar procedure was performed to determine the probability of contacts between H-NS and Hha, P Hha , and between H-NS domains, with P 1-40 indicating the probability of finding the DNAbinding domain (residues 96-137) close to the dimerization domain (residues 1-40). P 96-137 indicates the probability of finding the dimerization domain close to the DNA-binding domain.
To determine the location of the buckle in helix a3, we calculated the helical hydrogen bond distances d O-N for residues 22-67 in each monomer between the backbone carbonyl oxygen O of residue i and the backbone amide nitrogen N of residue i + 4. A hydrogen bond is counted to be in contact if they are at a minimum distance of 0.35 nm or less. These probabilities are averaged over all 16 simulations (excluding the first 10 ns) and the two monomers. Snapshots and movies were generated with PyMOL. van