H-NS forms a superhelical protein scaffold for DNA condensation

Abstract

The histone-like nucleoid structuring (H-NS) protein plays a fundamental role in DNA condensation and is a key regulator of enterobacterial gene expression in response to changes in osmolarity, pH, and temperature. The protein is capable of high-order self-association via interactions of its oligomerization domain. Using crystallography, we have solved the structure of this complete domain in an oligomerized state. The observed superhelical structure establishes a mechanism for the self-association of H-NS via both an N-terminal antiparallel coiled-coil and a second, hitherto unidentified, helix-turn-helix dimerization interface at the C-terminal end of the oligomerization domain. The helical scaffold suggests the formation of a H-NS:plectonemic DNA nucleoprotein complex that is capable of explaining published biophysical and functional data, and establishes a unifying structural basis for coordinating the DNA packaging and transcription repression functions of H-NS.

Enterobacteria condense DNA through binding to a protein scaffold. As with eukaryotic DNA, gene compaction via protein condensation and the resulting changes in accessibility to promoter regions provide general mechanisms to control gene expression within the bacterial nucleoid. H-NS is a highly abundant, ubiquitous DNA-binding protein that controls expression of over 200 genes in response to environmental factors by acting on DNA topology (1, 2). For example, constraining of the supercoiling of DNA by H-NS is linked to bacterial response to changes in osmolarity and temperature (3–6). By binding cooperatively to adjacent promoter regions, H-NS is able to repress gene expression (6–8). H-NS is subdivided into two discrete and functionally independent domains joined through a flexible linker; the N-terminal oligomerization domain (residues 1–83) and the C-terminal DNA-binding domain (residues 91–137). Structural models for dimerization of truncated forms of the oligomerization domain have been determined from different organisms (Salmonella typhimurium, residues 2–58 (9) and 2–65 (10); Escherichia coli, residues 2–47 (11); and Vibrio cholerae (H-NS homologue VicH), residues 2–51 (12)). In common in all of these structures the dimers are supported by interactions between two short N-terminal helices (H1 and H2 (Site 1)). In addition, the structure of the C-terminal winged helix-turn-helix (wHTH) DNA-binding domain of E. coli H-NS (residues 91–137) has been determined (13). To date the existing structural detail has failed to produce a satisfactory mechanism for high-order oligomerization or the concomitant DNA condensation.

Results & Discussion

Following an extensive biophysical and biochemical characterization of various H-NS fragments we were able to design a tailored crystallization screen that allowed us to obtain crystals of residues 1–83 of the S. typhimurium H-NS C21S mutant, H-NS_1–83 (14, 15). The structurally conservative C21S mutation removes the possibility of self-association by disulphide bond formation in oxidizing conditions, but has no discernable effect on the biophysical properties of the polypeptide (16). This construct lacks the C-terminal DNA-binding domain, but contains the secondary dimerization site (Site 2—between residues 67 and 83) responsible for oligomerization of the previously elucidated N-terminal dimers (14). A 3.7 Å X-ray structure was determined from these crystals (see Table 1, Methods, and SI Appendix). The structure reveals the presence of an elongated third α-helix, H3 (residues 23–67), followed by a kink and a fourth α-helix (Fig. 1 A, C). These features coincide perfectly with previously reported secondary structure predictions (14).

Table 1.

Crystallographic data collection and structure refinement statistics

A) Data collection *
Space group	P65
Cell dimensions a = b, c (Å)	130.4, 55.06
Resolution range (Å)	32–3.7 (3.83–3.7)
Completeness (%)	96.9 (96.4)
Total number of reflections	15,448
Number of unique reflections	5,676
Multiplicity	2.72 (2.49)
Rmerge †	0.10 (0.44)
I/σI	6.7 (1.8)
Number of molecules in ASU	2
Solvent Content (%)	83.6
Matthews Coefficient	7.5
B) Refinement
Resolution range (Å)	32–3.7
Number of reflections used	5,676
Rcryst (%) ‡	27.1
Rfree (%) §	30.4
Number of nonH atoms	1,292
rmsd from ideal, bond lengths (Å)	0.014
rmsd from ideal, angles (°)	1.55

^*Values in parentheses are for the highest resolution shell

^†R_merge = Σ_hklΣ_i|I_i(hkl) - 〈I(hkl)〉|/Σ_hklΣ_iI_i(hkl), where I_i(hkl) and 〈I(hkl)〉 are the observed individual and mean intensities of a reflection with the indices hkl, respectively, Σ_i is the sum over the individual measurements of a reflection with indices hkl and Σ_hkl is the sum over all reflections.

^‡All measured reflections were included in refinement except for the 8% that were used for calculation of the R_free

^§

Fig. 1.

Structural basis for H-NS oligomerization. A, An oligomer of three symmetry-related (N-terminal oligomerization domain) dimers (Site 1), connected through their secondary dimer interface (Site 2). One dimer has been highlighted in color. B, Detailed view of dimerization Site 2. One protomer is displayed as secondary structure and stick model, the other as molecular surface. The surface is colored to highlight different properties of residues: blue, positively charged atoms; red, negatively charged atoms; green, hydrophobic atoms; salmon, polar oxygens; marine, polar nitrogens; yellow, sulfur. C, Amino acid sequences of H-NS from E. coli and S. typhymurium. The sequences are also shown for the H-NS paralogue StpA from S. typhymurium and the H-NS-like VicH from V. cholerae. Secondary structural features and residue numbers are included for reference.

Structural Basis for H-NS Oligomerization.

The second site of dimerization (Site 2) is formed by H3 and H4 forming a helix-turn-helix motif (between residues 57–83) which interlocks the C termini of the two protomers in an antiparallel fashion (Fig. 1A). This secondary tail-to-tail dimer buries a total of 1,800 Å² of protein surface, and is stabilized by a hydrophobic core (formed by residues Leu58, Tyr61, Met64, Leu65, Leu75, Leu76, and Met79), flanked by salt bridges (Lys57-Asp68*, Arg54-Glu74*, where *refers to the symmetry-related protomer) (Fig. 1B). The Site 2 interface is recognized as a stable “bona fide” dimer by the protein interfaces, surface, and assemblies service PISA at the European Bioinformatics Institute (17) (ΔG^int - 14 kcal/mol, ΔG^diss 4.0 kcal/mol) in agreement with biophysical data (14, 18). The destabilizing effect of mutations at Arg54 (19, 20) and Lys57 (21) and the resulting hns null mutant phenotype can be explained by the presence of these amino acids in this site of dimer formation. When projecting the H-NS_1–83 structure onto our previously published sequence alignment (12), it is apparent that key residues for the dimer interface are conserved in close H-NS homologues such as StpA, Sfh, or VicH, suggesting that these proteins are able to form oligomers in a similar manner to H-NS. Conversely in more remote homologues such as Ler, Bph3, Hvra, or Spb, the region Gln59—Asp68 appears to be missing. This region contains important hydrophobic residues that form the core of the second dimerization site and Asp68 which stabilizes the structure by an ionic bond with Lys75 of the other protomer. Additionally, in these remote homologues either Gly69 or Pro72 is substituted. These residues are important for forming the loop between helices 3 and 4. Moreover the E74LLXXL/M sequence necessary for stabilizing Site 2 is not conserved. It is therefore unlikely that these remote homologues can propagate to the high-order structures seen with H-NS.

The “head-to-head” (Site 1) and “tail-to-tail” (Site 2) dimer interfaces together create a chain of linked H-NS molecules that forms a superhelix (six molecules per turn, about 190 Å diameter, and 280 Å rise) which runs through the crystal, following the P6₅ screw axis (Fig. 2 A, B). This macromolecular structure supports reported data on deletions or mutations at Sites 1 or 2 which fail to oligomerize giving rise to stable dimers (8, 9, 14, 22). Superimposition of the two chains of an isolated H-NS dimer reveals a ∼5° variation in the orientation of the axis of H3 after residue 41, suggesting that the crystal packing imposes a limited deformation on the H-NS superhelix, and illustrating a certain malleability of the structure between residues 40 and 55.

Fig. 2.

Superhelical structure of H-NS in crystal lattice. A and B, 90° views of the superhelix formed by a chain of head-to-head and tail-to-tail linked H-NS molecules. Orientation of the molecules is taken from the crystal lattice. R15, R19, and K32 forming the positively charged, DNA-repelling surface are shown in blue.

At Site 2 the H-NS oligomer orientates H4 (and hence the linker to the DNA-binding domain) towards the outside of the superhelix. Thus, if the DNA-binding domain is directed towards the outside of this structure it suggests that DNA binds on this exterior surface (Fig. 3 A, B). This model for DNA binding is based on the previously reported binding of two strands of DNA to the protein filament (23, 24). Furthermore, the model incorporates the maximum number of protein-DNA contacts possible, making it thermodynamically stable and consistent with potential H-NS nucleation on the oligonucleotide (6). The model of DNA binding to the outside of the superhelix also provides some structural reconciliation of the observed influence of Arg12 on gene expression (22). Although this residue resides within Site 1, mutation of this residue does not affect H-NS structure or oligomerization (14, 20) so the dominant negative H-NS phenotype is likely caused by disrupting a direct interaction with the DNA. This residue forms a positively charged surface with three other highly conserved residues (Arg15, Arg19, and Lys32) (Fig. 1C) (25) capable of compensating for the repulsive effect of bringing two DNA duplex strands into close proximity. Introduction of a negative charge to this region (H-NS A18E mutant) gives rise to a dominant negative H-NS phenotype (26).

Fig. 3.

Proposed plectonemic DNA compaction through H-NS. A, Model of H-NS superhelix (yellow: putative position of the wHTH domains; orange: H-NS_1–83 oligomer; blue: R15, R19, and K32) accommodating two DNA double helices (gray). For comparison the toroid compaction in the nuclear core particle is shown (green: histone core; red: DNA). B. 90° rotation about horizontal axis.

Temperature Dependence of Oligomerization and DNA Binding.

H-NS plays a defined role in DNA supercoiling. For example, differences in DNA supercoiling observed between wild-type E. coli and a hns null mutant are more marked at the site of H-NS binding than for the bacterial chromosome as a whole (27). The superhelix observed in the crystal lattice provides an ideal scaffold for both the compaction of the DNA as a plectonemic supercoil (23) and the mediation of gene repression via binding of double-stranded DNA to each outside edge of the protein core (Fig. 3 A, B). The structure and proposed mode of interaction with DNA provides a rationale for data on the nucleation (28, 29) and bridging (24, 30) models of gene repression by H-NS and is consistent with biophysical studies (31) as well as imaging of H-NS:DNA nucleoprotein complexes by atomic force microscopy and electron microscopy (23, 32, 33). Small angle X-ray scattering (SAXS) analysis showed that at 10 °C and 0.3–0.4 mM H-NS_1–83 adopts a conformation in solution that is comparable to the oligomer present in the crystal; under these conditions extending to an average of eight monomers or four dimeric units (Fig. 4 A, B). Based on the concentration dependence of the size of these oligomers (SI Appendix: Fig. S2), at physiological concentrations of H-NS (∼0.1 mM) preformed protein structures capable of driving DNA supercoiling are unlikely. Consistent with previous observations, increasing the temperature to 40 °C results in disassembly of the oligomer to a species that structurally resembles that expected for an H-NS_1–83 dimer (Fig. 4 B, C), thus enhancing the view that gene repression is released in response to elevated temperature by dissociation of the protein oligomer (14, 34).

Fig. 4.

SAXS analysis of self-association at two temperatures. A, SAXS pattern calculated from an oligomer of four linked H-NS_1–83 dimers (green) and a single Site 1 H-NS_1–82 dimer (red), fitted to experimental data of 0.4 mM H-NS_1–83 recorded at 10 °C (χ = 5.8), and 40 °C (χ = 2.4). B, ∼90° views of 10 °C ab initio sphere models (gray) superimposed onto four linked H-NS_1–82 dimers (orange). C, ∼90° views of 40 °C ab initio sphere models (gray) superimposed onto one H-NS_1–82 site 1-linked dimer (cyan and blue). Ab initio models were produced based on corresponding SAXS data shown in A.

A Model for Plectonemic DNA Condensation.

The model presented here (Fig. 2 A, B) provides a structural basis for plectonemic condensation of DNA, (which is the dominant form of DNA in bacteria during exponential growth), as opposed to the toroidal form observed in eukaryotic nucleosomes. The H-NS protein scaffold supports a more extended DNA conformation compared to the nucleosome core (Fig. 3), in agreement with observations that H-NS does not have a strong capacity to actively deform DNA and binds preferentially to adenosine and thymidine (AT)-rich sequences (29, 35). Furthermore the constraining of DNA by H-NS appears to be less structurally robust than found in binding to the eukaryotic nucleosome. Indeed it would appear that unless stabilized by other factors, the strength with which H-NS can bend DNA cannot exceed the stability of the helix between residues 42 and 54. This limited stability is consistent with the finding that only a moderate increase in temperature is sufficient to disrupt the H-NS oligomer interaction and alleviate transcription repression (8, 34, 36, 37).

In summary, the self-association of H-NS to a concentration-dependent, poly disperse, high-order macrostructure encompasses a fine-tuned, environmentally susceptible mechanism for, on the one hand, condensing DNA into the nucleoid and on the other controlling access of the protein machinery required for gene transcription.

Methods

Crystal Structure Determination.

Protein production and crystallization are described elsewhere (15). Crystallographic data was collected on a Rigaku RAXIS-IV image plate system with rotating-anode source using osmic focusing mirrors to produce monochromatic Cu Kα radiation at a wavelength of 1.5418 Å. The 3.7 Å resolution structure of H-NS 1–83 C21S (H-NS_1–83) was determined by molecular replacement using the program BALBES (38), which chose the N-terminal dimer of VicH (PDB entry 1OV9) as the best search model. BALBES produced a 99% certainty-solution in space group P6₅, requiring only one dimer in the asymmetric unit (R and R_free were 0.36 and 0.37 after initial rigid body and positional refinement of this solution). Unambiguous electron density was observed accounting for the residues not included in the search model (SI Appendix: Fig. S1). H-NS 1–83 was reconstructed using manual (re)building (COOT (39)) and maximum-likelihood refinement (REFMAC (40)). Residues 44–67 could be modeled without ambiguity in the residual 2Fo-Fc electron density (SI Appendix: Fig. S1), because these residues were simply a continuation of the third α-helix of H-NS (H3) of which residues 23–44 were present in the search model. After residue 67 the electron density map showed a kink, followed by a region with further α-helical nature (H4). These electron density features coincided perfectly with the predicted loop-helix structure for residues 68–83. The high solvent content (over 83%), the simplicity of the fold, and the P2 symmetry allowed chain tracing without ambiguity. Model building was assisted by the high solvent content of over 83% allowing for solvent flattening, in combination with twofold electron density averaging. A map sharpening factor of 100 helped to obtain electron densities for about 85% of the side chains. In initial rounds of refinement, the B-factor was initially fixed to 50 Angstrom (2), and five translation-liberation-screw (TLS) domains were chosen for each protomer, based on the structure. Moreover shift dampening (D = 0.3) was used in REFMAC and strong noncrystallographic-symmetry (ncs) constraints were applied to the backbone, and medium ncs constraints were applied to the side chains. For the final cycle of TLS refinement, all these constraints were released. The final model contains ten residues that appear as outliers of the Ramachandran plot (MolProbity (41)). Structure and diffraction data were deposited at the PDB, with accession code 3NR7.

SAXS Data Analysis.

Proteins for SAXS were prepared as described previously (14), except that 370 mM NaCl, 20 mM Tris pH 7.4, 1 mM EDTA was used for H-NS_1–83. SAXS data were collected at the SIBYLS beamline at the Advanced Light Source, Berkeley, USA at a wavelength of 1 Å. Every sample was exposed successively for 0.5, 5.0, and 0.5 s. No radiation damage was observed, as assessed comparing data from first and last 0.5 s exposure. H-NS_1–83 sample concentrations ranged from 1.0–4.4 mg/mL (110–460 μM). Data were recorded at 10, 25, and 40 °C. SAXS patterns at 10 °C and 25 °C were the same (SI Appendix: Fig. S3). Data were analyzed using PRIMUS, GASBOR, DAMMIF, and DAMAVER from the ATSAS v2.3 software package (42). Model SAXS patterns were calculated and fitted to data using FoXS (43). Simple normal mode refinement (carried out using models generated by ElNemo (44) allowed to decrease the χ value for the fit of the H-NS_1–82 model to SAXS data by over 10% (from χ = 5.8 to 5.2 for H-NS_1–83 at 10 °C), suggested that the crystal has packing slightly distorted the oligomer.