Molecular Basis for the Adaptive Evolution of Environment Sensing by H-NS Proteins

The DNA-binding protein H-NS is a pleiotropic gene regulator in gram-negative bacteria. Through its capacity to sense temperature and other environmental factors, H-NS allows pathogens like Salmonella to adapt their gene expression, and hence toxicity and biological responses, to their presence inside or outside warm-blooded hosts. To investigate how this sensing mechanism may have evolved to fit different bacterial lifestyles, we compared H-NS orthologs from bacteria that infect humans, plants, and insects, and from bacteria that live on a deep-sea hypothermal vent. The combination of biophysical characterization, high-resolution proton-less NMR spectroscopy and molecular simulations revealed, at an atomistic level, how the same general mechanism was adapted to specific habitats and lifestyles. In particular, we demonstrate how environment-sensing characteristics arise from specifically positioned intra- or intermolecular electrostatic interactions. Our integrative approach clarified the mechanism for H-NS–mediated environmental sensing and suggests that it resulted from the exaptation of an ancestral protein feature.


INTRODUCTION
The histone-like nucleoid-structuring (H-NS) protein is a central controller of the gene regulatory networks in enterobacteria (1). H-NS inhibits gene transcription by coating and/or condensing DNA; an environment-sensing mechanism allows H-NS to liberate these DNA regions for gene expression in response to physicochemical changes (2)(3)(4). H-NS preferentially binds to AT-rich sequences, which enables its dual role in (i) the organization of the bacterial chromosome and (ii) the silencing of horizontally acquired foreign DNAs (5)(6)(7)(8). The latter mechanism allows bacteria to assimilate foreign DNAs, which, however, are only expressed as a last resort in case of acute threats or stresses (8). Thus, H-NS plays a crucial role in the adaptation, survivability, and antibiotic resistance of bacteria. Given the growing threat of multidrug resistance, H-NS has attracted increasing research interest, with a particular focus on elucidating the molecular mechanisms of adaptive evolution (9)(10)(11). H NS possesses two dimerization domains (site1, residues 1-44; site2, resides 52-82; the numbering of Salmonella typhimurium is adopted throughout the text), and a C-terminal DNAbinding domain (DNAbd, residues 93-137) that is connected through a flexible region (linker, residues 83-92) to site2 (Fig. 1) (5,(13)(14)(15)(16). The combination of site1 and site2 dimers allows H-NS to form multimers for a stable concerted DNA association that results in gene silencing (15).
In a previous study, we showed that site2 of S. typhimurium H-NS is the primary response element to temperature changes (17). Site2 unfolds at human body temperature, allowing the linker-DNAbd region to associate with site1 to adapt an autoinhibited conformation incapable of binding to DNA. Salinity and pH can also influence the stability of site2 dimers, and hence may also affect gene repression by H-NS (17,18). Thus, the sensitivity of H-NS to temperature and other physiochemical changes allows human pathogens such as S. typhimurium, Vibrio cholerae, and enterohaemorrhagic Escherichia coli to sense when they enter a homothermic host and adapt their gene expression profiles accordingly.
To date, studies to elucidate environment sensing of H-NS were almost exclusively conducted with proteins from two model systems, S. typhimurium (e.g. (9,(19)(20)(21)) and E. coli (e.g. (18,(22)(23)(24)(25)(26)), both of which infect humans. Yet, H-NS orthologs are also present in enterobacteria that do not have warm-blooded hosts, raising the question of what biological role H-NS plays in these species. Answering this question requires to determine the structural basis for environmentsensing in H-NS orthologs with drastically different lifestyles. However, the molecular dynamics and multidomain composition of H-NS hamper conventional structural analysis. Therefore, we combined large-scale molecular simulations and spectroscopic approaches to elucidate how environment-sensing by H-NS may have adapted in different species. This pluridisciplinary approach yielded an atomic-level understanding of how H-NS orthologs evolved specific residue substitutions to adapt environment-sensing to their bacterial habitats, and may open new avenues for strategies to combat antibiotic resistance.

RESULTS
To investigate the adaptation of environment-sensing by H-NS, we selected four H-NS orthologs from ~3000 H-NS-like sequences available in the Uniprot database: 1) H-NSST from S. typhimurium (a pathogen of mammals), 2) H-NSEA from Erwinia amylovora (a plant pathogen that infects apples and pears), 3) H-NSBA from Buchnera aphidicola (an endosymbiont of aphids), and 4) H-NSIL from Idiomarina loiheinsis (a free-living bacterium from a deep-sea hypothermal vent).
H-NSEA and H-NSBA share more than 60% sequence identity with H-NSST, whereas H-NSIL is only 40% identical to H-NSST (Fig. 1A). Across the orthologs, the least conserved regions are residues 45-56 in 3, and residues 77-86 (end of 4 and beginning of the linker). The variable residues on 3 and linker may act as simple spacers (Fig. 1), suggesting 4 as a prime candidate for mediating adaptations to the environment.
The site1 dimer is markedly more stable than the site2 dimer in the H-NS orthologs. H-NSST site1 and site2 form homodimers to enable H-NS multimerization in a head-to-head/tail-to-tail fashion ( Fig. 1C) (15). In concert with the DNA interaction of the individual domains, this homooligomerization is required for tight DNA binding and hence gene repression. In our previous study, we showed that only H-NSST site2 dimers unfold and dissociated within a biologically relevant temperature range, whereas site1 dimers remain unaffected (17). The higher stability of the site1 dimer of H-NSST is explained by a substantially larger contact surface between the two monomers (ca. 3,300 Å2 compared to ca. 850 Å2 for site1 and site2, respectively, according to PDBePISA (27)).
To investigate whether this mechanism is conserved in other H-NS orthologs, we built homology models for H-NSEA, H-NSBA, and H-NSIL using the crystal structure of the H-NSST site1-site2 fragment as a template (PDB ID: 3NR7) (15). Next, we constructed a tetrameric model as a minimal representation that conserves all features of the H-NS multimer. This tetramer contained two full-length H-NS monomers (residues 1-137, with templates PDB IDs: 3NR7 and 2L93) and two partial monomers, truncated before site2 (residues 1-52) (Fig. 1C). To probe differences in environmental responses of the orthologs, we first used conventional full-atom molecular dynamics (MD). We simulated all four tetramers (a ~100,000 atom system; see The tetramer simulations at 0.15 M NaCl and 293 K produced a lower residue fluctuation level in site1 (local root-mean-squared fluctuation, RMSF 0.4 to 1.9 Å) than in site2 (local RMSF 0.5 to 4.4 Å) for all four orthologs (Supplementary File 1B). The higher stability of the site1 dimer is explained by the generally higher number of nonpolar contacts than in the site2 dimer (Fig. 2,   Figure Supplement S1A). These contacts involved conserved hydrophobic amino acid residues, notably L5 (or I5) and L8 of 1, L14 of 2, and L23, L26, V36, and V37 (or I37) of 3 (Fig. 2).
These interactions remained formed in all site1 dimers in our tetramer simulations (at 0.15 M NaCl at 293 K) and tetramer simulations at higher salinity (at 0.50 M NaCl at 293 K) or higher temperature (at 0.15 M NaCl at 313 K). Hence, we found that the stability of the site1 dimers resulted mainly from strong and conserved nonpolar packing. Collectively, these data conclude that the mechanism observed for H-NSSTwhere site1 remains stable, and the site2 stability is affected by the environmentis conserved in H-NSEA, H-NSBA, and H-NSIL.  Our simulations show how specific protein dynamics might modulate the ortholog's response to salinity or temperature. For example, we observed increased bending of the 3 backbone (annotated by the black arrow in Fig. 3C) at high temperature (313 K) or high salinity (0.50 M NaCl) (Figure Supplement S2). Although 3 bending occurred in all orthologs, it only significantly affected the site2 dimer of H-NSST by separating R54 from E74' or D71', suggesting that this mechanism contributed to the salt and temperature sensitivity of H-NSST site2, whereas it was not strong enough to significantly affect site2 stability in other orthologs.
Another example was given by H-NSEA, where an alternative R54-D71' salt bridge formed whenever the R54-E74' contact was broken at 313 K. This alternative R54-D71' salt bridge stabilized the H-NSEA site2 dimer at the higher temperature, suggesting that this compensatory mechanism resulted in a decreased sensitivity to temperature (Fig. 3C). H-NSIL provided a final example for a specific response. Compared with the R54-E74' salt bridge (Fig. 3A), the K57-D68' salt bridge only varied slightly in all our simulations (Fig. 3B). However, the substitution D68A in H-NSIL supplanted the electrostatic interaction with a nonpolar interaction, which was broken at 313 K in our simulations (Fig. 3D). This effect suggested that H-NSIL had a reduced sensitivity to salinity, while remaining sensitive to temperature. To complement the dynamics of H-NS orthologs from our conventional MD simulations, we used extensive simulations with umbrella sampling to quantitate the overall site2 stability. We calculated the potential of mean force (PMF) for site2 dimer dissociation (residues 50-82, ca. 46,000 atoms) of the four H-NS orthologs for three different conditions (low salinity/low temperature, high salinity, or high temperature). The site2 monomers were not constrained and remained structurally flexible during the dissociation process. To ensure convergence in the PMFs, we employed long windows (54 ns) in simulations totalling 52 s (details provided in the SI; see Figure Supplement S3 for resulting histograms and PMFs along the dissociation coordinate). According to the free energy difference between the dimerization and dissociation states (G=Gdimer-Gdissociation), we estimated the energetic impact from increased salinity and temperature as G=Ghigh salinity or temperature -G293K, 0.15M NaCl (Fig. 3E). Notably, high salinity (0.50 M NaCl) or temperature (313 K) decreased the stability of the H-NSST site2 dimer by 2.2 kcal/mol. H-NSBA displayed a similar sensitivity to temperature but a lower sensitivity to salinity, which destabilized the dimer by 1.5 kcal/mol. Interestingly, our data indicated that H-NSEA was only sensitive to salinity, whereas raising the temperature had little impact on the stability of the H-NSEA site2 dimer. Conversely, H-NSIL only responded to temperature, whereas the increased salinity did not affect the stability of its site2 dimer (G ~ 0 kcal/mol). Collectively, our conventional MD simulations and PMF calculations suggested how, on the atomic level, changes in the site2 sequence may alter the sensitivity of the H-NS orthologs to different environmental changes.

The autoinhibited H-NS conformation is maintained through dynamic electrostatic
interactions. In a previous study (17), we had shown that melting and dissociation of site2 dimers allow H-NSST to adapt a closed conformation in which the linker-DNAbd fragment interacts with a negatively charged region on site1 3 (Figure Supplement S4A), and that this auto-inhibitory interaction is incompatible with DNA interactions. However, due to extensive signal broadening of mainly linker amides exchanging with water, our conventional proton-detected NMR analysis based on exchangeable amide H/N-observed correlations did not allow confident mapping of the binding site on the C-terminal region (17) ( Figure Supplement S4B). Herein, we overcame this limitation by using proton-less 13C-detected NMR analysis to complete the resonance assignment of the linker-DNAbd fragment ( Fig. 4 and Figure Supplement S4C). These complete carbon chemical shifts allowed us to elucidate the structural mechanism of H-NSST autoinhibition fully, and, in a second step, to use this understanding to investigate the existence of this closed conformation in the orthologs.
To further probe the local dynamics of the polypeptide chain, we determined the random coil index order parameter RCI-S2 based on the fully assigned 13C-resonances for each residue for the ligand-free and site1-saturated CtST. The dynamics of the well-ordered DNAbd domain remained unchanged with or without site1 present, in agreement with its only minor involvement in the auto-association (Fig. 4D). Conversely, the linker residues 84-95 were disordered without regular secondary motifs in the absence of site1 (RCI-S2 < 0.35). Upon addition of site1, the local dynamics decreased, particularly within the stretch of four amino acids K98-R90-A91-A92 (RCI-S2 > 0.6) that form the type VIII β-turn according to MICS. Nonetheless, the overall RCI-S2 of the linker remained low, demonstrating that the association with site1 did not substantially restrict the linker's movements.
Collectively, our analysis established that the autoinhibitory site1:CtST association was driven by oppositely charged residues located on site1 and the linker, and involved only a small region of the DNAbd. The resulting intramolecular interaction was maintained through 'fuzzy' charge-pairing that did not fix the partners into a structurally stable complex.  (Fig. 5A). This level of conservation was expected, given that this region is also required for DNA association (16)  To test this prediction, we carried out in vitro binding experiments using microscale thermophoresis (MST).
In vitro binding experiments between site1 and Ct corroborated that the strength of the auto-association was similar for H-NSST and H-NSEA (Fig. 5B). H-NSBA has mostly lost its capacity for auto-association, in agreement with its altered site1 surface characteristic. Moreover, H-NSBACt has a proline residue (P91) in position 3 of the -turn region, which is highly unfavorable for this secondary structure (29). Conversely, the auto-association was tenfold stronger in H-NSIL than in H-NSST, despite a less acidic site1 and despite the presence of a mildly unfavourable proline in -turn position 4 (P92). Hence, autoinhibition in H-NSIL might include additional and/or different interactions. Increasing the temperature decreased the self-association strength 2-3 fold in H-NSST and H-NSIL, and more than tenfold in H-NSEA (Fig. 5B). It also decreased the Kd for H-NSBA to values beyond the measurement range. We concluded that the strength of the autoinhibitory conformation is mostly modulated by the electrostatic surface characteristics of site1, with additional influence from -turn-breaking proline residues in the Ct, which otherwise preserves its basic characteristic needed for DNA interactions.  (Fig. 5C) (17). The marked decrease of RH for curves at 0.15, 0.25, and 0.50 M NaCl indicated a strong inverse correlation between salinity and site2 stability.
All three H-NS orthologs displayed a similar behaviour overall, further supporting that the general mechanism of site2-mediated multimerization and environment sensing was preserved.
However, we noted important differences in the orthologs' response characteristics (Fig. 5C)

: (i)
Of the four orthologs, H-NSST responded most strongly to salinity and temperature, consistent with the broken R54-E74' salt bridge and large site2 RMSF in our high-salinity or high-temperature simulations. (ii) H-NSEA was less temperature-sensitive and showed weaker multimerization than the other orthologs. Indeed, our simulations suggested that H-NSEA site2 can rearrange the interdimer salt bridge and form either R54-E74' or R54-D71' to maintain site2 stability at higher temperatures. (iii) H-NSBA had the highest tendency to multimerize among all the orthologs tested, which might partly be explained by an absence of the autoinhibitory conformation. Compared to H-NSST, our PMF calculations showed a slightly higher sensitivity to temperature and a slightly reduced sensitivity to salinity. Although these tendencies were apparent in our DLS data, these data were also affected by the fact that H-NSBA required more than 150 mM NaCl to stay in solution, but H-NSBA nonetheless aggregated at 30 ˚C. (iv) H-NSIL showed a decreased sensitivity to salinity compared to H-NSST, as suggested by our computational analysis (i.e., the lack of the site2 K57-R68' salt bridge, the lack of salt-promoted free energy changes, and the attenuated electrostatic site1 surface). Collectively, our experimental observations revealed significant differences in response to physicochemical parameters, which were in agreement with our predictions based on the molecular features of the H-NS orthologs.
Conclusions. Environment-sensing through the pleiotropic gene regulator H-NS helps S. typhimurium to adapt when it is present inside its host mammal. In a previous study, we had shown that an increase in temperature, and to some extent salinity, dissociates the second dimerization element (site2), which produces two effects: firstly, it impedes synergistic DNA binding of H-NS multimers, and secondly, it allows H-NS to adopt an autoinhibitory conformation where DNA binding residues on the C-terminal linker-DNAbd fragment associate with the Nterminal site1 dimerization domain (17). In this study, we confirmed key aspects of this model, namely that site2 is the element that senses changes in physicochemical parameters. We also uncovered additional aspects of this process. In particular, proton-less NMR fully revealed the position and dynamics of the linker-DNAbd residues involved in the autoinhibitory association with site1. The -turn linker residues 89-91 critical for autoinhibition cannot reach site1 without site2 dissociation (Figure Supplement S1B), confirming that the closed autoinhibited conformation is mutually exclusive with H-NS multimerization along DNA. Our NMR analysis also demonstrated that this autoinhibition is achieved at a low entropic cost, maintaining a high flexibility with respect to the exact distribution of the interacting charges on both site1 and the linker-DNAbd fragment.
On the one hand, avoiding the entropic penalty helps the autoinhibitory interaction to prevail against the competing DNA association. (Of note, the covalent link between site1 and the Ct will enhance their local concentration and hence their apparent affinity compared to our measurement based on separate domains in Fig. 5B). On the other hand, the fuzziness of the charge-charge interactions facilitates preserving the autoinhibition during bacterial evolution and adaptation. Although other factors inside bacteria can modify the in vitro behaviour of the isolated protein, it is interesting to consider these idiosyncrasies with respect to the bacteria's habitats ( Fig. 6): (i) H-NSST had the highest sensitivity to temperature and salt, in agreement with the critical role of H-NSST in helping Salmonella adapt its gene expression profile depending on if it is inside or outside a warm-blooded mammal.
(ii) In comparison, we found that the response to temperature was markedly attenuated in H-NSEA. E. amylovora is the causing agent of fire blight, a contagious disease that mostly affects apples and pears (30). The reduced sensitivity of H-NS to temperature may reflect the minor importance of this factor in an environment of ambient temperature in temperate climate zones.
(iii) B. aphidicola is an intracellular symbiont of aphids that is maternally transmitted to the next generation via the ovaries (31). B. aphidicola co-evolved with aphids for more than 150 million years, and despite having the highest sequence identity (61%) to H-NSST of all orthologs, H-NSBA showed the least conserved features among the orthologs tested, indicating that adaptive evolution was achieved by only minor changes. H-NSBA site2 interactions were stronger than those of other orthologs, and the features promoting the autoinhibitory form were compromised.
Hence, H-NSBA may provide a stronger and more robust repression of the genes that it controls.
In vitro, H-NSBA was the least stable ortholog tested, and had already started to aggregate above

Simulation Details.
Using the CHARMM36 all-atom force field, we performed conventional MD simulations of H-NS tetramers and umbrella sampling of site2 dimers in GROMACS. For DLS and MST, recombinant protein production and measurements were adapted from (17). However, we fluorescently labelled site1 for DLS, instead of the Ct. For NMR, 13C,15N-labelled S. typhimurium H-NS84-137 was expressed in minimal M9 media with 5 g/L of U-13C glucose and 1 g/L of 15NH4Cl salt. Proton and low-γ detected high-resolution NMR spectroscopy was carried out on a 700-MH Bruker NEO spectrometer equipped with a 5-mm cryogenic TXO direct observe 15N,13C-optimized probe at 25 oC. Further details are shown as below.

Simulation setup:
Our simulations were carried out by GROMACS (34) (MD simulations of tetramers and Potential of Mean Force (PMF) simulations of site2 dimers). All the models were solvated in a TIP3P water box, with counterions to neutralize the charges and additional NaCl for the desired salinity. Each tetramer system contains ca. 33,000 TIP3P water molecules, counter ions, and 150 or 500 mM NaCl, totaling ca. 100,000 atoms in a periodic box 1399 nm3. All simulations were performed following a minimization, 250 ps equilibration in the NVT and NPT ensemble with Berendsen temperature and pressure coupling, and a production stage NPT (293 or 313 K, 1 bar). The CHARMM36 force field (35) was used with the cmap correction. The particle mesh Ewald (PME) technique (36) was used for the electrostatic calculations. The van der Waals and short-range electrostatics were cutoff at 12.0 Å with switch at 10 Å.
The PMF simulations were carried out with the MD program GROMACS (34) using the umbrellaing sampling (US) technique. The CHARMM36 force field was also used. Each site2 monomer of the center of mass (COM) distance was chosen as the disassociation pathway and used for enhanced sampling. After 500 ps equilibrium with the NPT ensemble, initial structures for windows along the reaction coordinates were generated with steered MD. In the steered MD simulation, one chain was pulled away along in the direction of increasing the COM distance with force constant of 12 kcal mol-1 Å-2, until the COM distance reached 25 Å. The windows were taken within a range of 0-25 Å. The umbrella windows were optimized at the 0.3 Å interval to ensure sufficient overlap. There are about 80 windows per simulation, and each window was simulated with force constant of 1.2 kcal mol-1 Å-2. All PMF simulations converge in 54 ns per window (Supplementary Fig. S1). mM Bis-TRIS at pH 6.5 and 0.002% NaN3.  (41). The low-γ, so 13C-detected experiments mentioned above were started with 1H-excitation in order to enhance the sensitivity and recorded in in-phase and anti-phase (IPAP) mode for the virtual decoupling.

Dynamic Light
All spectra were processed in NMRpipe and analyzed in CARA and Sparky software. The random-coil-index order parameters RCI-S2 and secondary motifs, like β-turn, for Apo and H-NS1-57 saturated (1.5 mM) forms were determined from complete lists of Cα, Cβ (except glycines), N, C' chemical shifts with the TalosN and MICS programs, respectively. NMR chemical shift assignments for the H-NSSTCt in its apo and H-NSST site1-bound states are deposited at the BMRB with the IDs 50239 and 50240, respectively.

Micro Scale Thermophoresis (MST) for protein-protein interactions:
H-NS (residue 1-57) from S. typhimurium, E. amylovora, B. aphidicola and I. loihiensis were individually labeled Nterminally with fluorescent Alexa488-TFP (Thermo Scientific) and then unlabeled C-term of those proteins were titrated against Alexa488 labeled N-term correspondingly and the final results were plotted as described previously (17).     . The separation distance was measured as the C  -C  distance in the tetramer model (within chain B/C, which were modelled as full length). The color scheme annotates different simulation conditions: 293 K, 0.15 M NaCl (Green), 293 K, 0.50 M NaCl (Blue), and 313 K, 0.15 M NaCl (Red). For clarity, we show the smoothed data of two replicas for each system (solid and dash lines respectively). Since complete unfolding of site2 was not observed in these simulations (presumably due to the short timescale and the difficulty of sampling), these plots indicate a minimum separation of 15 to 20 Å between the N and C termini, demonstrating that site2 has to unfold to allow closer contacts between the termini.