Molecular basis for a protein-mediated DNA bridging mechanism that functions in condensation of the E. coli chromosome

Summary

The E. coli chromosome is condensed into insulated regions termed macrodomains (MDs), which are essential for genomic packaging. How chromosomal MDs are specifically organized and compacted is unknown. Here we report studies revealing the molecular basis for Terminus containing (Ter) chromosome condensation by the Ter-specific factor MatP. MatP contains a tripartite fold with a four-helix bundle DNA-binding motif, ribbon-helix-helix and C-terminal coiled-coil. Strikingly, MatP-matS structures show that the MatP coiled-coils form bridged tetramers that flexibly link distant matS sites. Atomic force microscopy and electron microscopy studies demonstrate that MatP alone loops DNA. Mutation of key coiled-coil residues destroys looping and causes a loss of Ter condensation in vivo. Thus, these data reveal the molecular basis for a protein-mediated DNA-bridging mechanism that mediates condensation of a large chromosomal domain in enterobacteria.

Introduction

The genomic DNA of all organisms must be condensed to fit within the confines of cells that are typically several thousand-fold smaller than the DNA itself. Compaction of bacterial chromosomes results in the formation of a structure called the nucleoid (Thanbichler and Shapiro, 2006). Early studies on isolated nucleoids from Escherichia coli showed that they consist of closed duplex structures with a series of loops (Kavenoff and Bowen, 1976). More recent studies have revealed a multilayered organization that includes long-range chromosome condensation based on large defined regions termed macrodomains (MDs) (Niki et al., 2000; Valens et al., 2004; Espéli et al., 2008). These studies indicated that there are four MDs, Ori, Right, Left, and Ter, which can be classified as separately organized portions of the nucleoid, enclosed within cages that are 0.2 μm in diameter (Espéli et al., 2008). The Ori domain contains the replication origin, oriC, and the Ter domain contains the replication termination site. The Left and Right MDs are adjacent to Ter and are separated from the Ori MD by two so-called non-structured (NS) regions. Studies showed that chromosomal loci have defined cellular positions throughout the cell cycle and that MD formation is critical for chromosome segregation (Li et al., 2003; Wang et al., 2005; Nielsen et al., 2006; Adachi et al., 2008; Espéli et al., 2008; Mercier et al., 2008).

Since the discovery of chromosomal domains, an intriguing yet unknown question has been how are each of the MDs organized into specific regions? While the highly abundant nucleoid associated proteins appear to provide some level of DNA condensation they have been shown to bind all MDs equally in chromatin immunoprecipitation (ChIP) experiments and therefore are not candidates for MD specific organization (Grainger et al., 2006; Dame et al., 2011). Recently, the first MD specific DNA binding protein was identified. This protein, MatP (Macrodomain Ter protein), binds to DNA sequences found only within the 800 kb Ter region and studies indicate that it is responsible for Ter MD organization (Mercier et al., 2008). MatP binds specifically to a Ter signature DNA sequence, GTGACRNYGTCAC called matS for Macrodomain Ter sequence (Mercier et al., 2008). This sequence is found 23 times in the Ter MD between E. coli chromosome coordinates 1135 kb to 1900 kb. Notably, the MatP/matS system is conserved in Enterobacteriaceae and Vibrionaceae species, underscoring its key biological role in bacterial chromosome organization (Mercier et al., 2008). Recently, the interaction of MatP with the division apparatus associated protein ZapB has been shown to promote the anchoring of the Ter MD at midcell (Espeli et al., 2012). While deletion of zapB alters the segregation of the Ter MD it does not influence the condensation of the Ter MD.

MatP is a 17 kD protein that shows no sequence homology to any previously characterized protein, hence how it functions at the molecular level is unclear. Previous studies showed that MatP is highly specific (Mercier et al., 2008) and does not seem to spread between matS sites on long distances as observed for ParB or Spo0J proteins (Rodionov and Yarmolinksy, 2004; Murray et al., 2006). Because matS sites are found every 35 kb on average, MatP must somehow exert its effects on Ter organization over large distances. To gain insight into the molecular mechanism of MatP-mediated Ter condensation we carried out genetic, biochemical and structural studies. These studies unveiled the detailed molecular mechanism by which MatP flexibly links distant matS DNA sites to compact the large Ter MD within enterobacteria chromosomes.

Results

Structure determination of MatP-matS complexes

A combination of structural, biochemical, genetic and in vivo approaches was used to dissect the function of MatP. MatP proteins and matS Ter sites are highly conserved in enterobacteria. Hence, MatP proteins from several Gram-negative bacteria were obtained and purified for crystallographic studies. Structures were obtained of the Yersinia pestis MatP and E. coli MatP proteins in complex with matS DNA; Fluorescence polarization (FP) experiments showed that Y. pestis MatP binds the matS site with affinity similar to that of E. coli MatP (K_d=2–5 nM) (Experimental Procedures; Figure S1). Three different crystal structures of Yersinia pestis MatP-matS were obtained and one E. coli MatP-matS structure was solved, providing multiple independent views of the MatP-matS interaction. Two Y. pestis MatP-matS structures were obtained with the 16mer matS site, TCGTGACATTGTCACG (where the matS consensus is underlined) to resolutions of 2.80 Å and 3.50 Å resolution and a structure of Y. pestis MatP-matS bound to the 23mer, AGTTCGTGACATTGTCACGAACT was solved to 2.55 Å resolution (Figure 1A; Table 1). A structure of the E. coli MatP-matS complex was obtained with the 19mer, TTCGTGACATTGTCACGAA to 3.55 Å (Jones et al., 1991; Brünger et al., 1998; Terwilliger and Berendzen, 1999) (Experimental Procedures; Table 1).

Figure 1. Structure, sequence conservation and functional analysis of MatP.

(A) Ribbon diagram of the Y. pestis MatP-23mer complex. The three MatP regions, the four-helix bundle, the RHH and the coiled-coil are colored red, cyan and yellow, respectively. The secondary structural elements are labeled for one subunit. The DNA is shown as sticks. To the right of the ribbon diagram, shown in the same orientation is an electrostatic representation of MatP, with blue and red indicating electropositive and electronegative regions, respectively. This Figure and Figures 2C–D, 3A–B, 5A and 6A were made using PyMOL (DeLano, 2002). (B) Sequence alignment of MatP homologues from ten enterobacteria and Vibrio cholera highlighting conserved regions. The name of the species is indicated on the left. Secondary structural elements are shown above the sequence. Residues important for different activities are labeled as indicated below the alignment. (C) Representative bacterial two hybrid analysis used to identify residues involved in MatP dimerization. Blue colonies reveal the ability of deleted/altered MatP to interact with wt MatP. Eight representative streaks are shown: pUT18C empty, pUT18C wt matP, matPL104G, matPQ109G, matPE127G, matPΔC31, matPΔC20 and matPα5–6 (truncation from 97 to 150). Plates were places on a pink background to highlight the blue/white phenotypes.

Table 1.

Selected Crystallographic Statistics format-DNA Complexes

Complex	Y. pestis MatP-16-mer	Y. pestis MatP-16-mer	Y. pestis MatP-23-mer	E. coli MatP-19-mer
Data Collection
Space group	P6522	I222	C2221	P212121
a, b, c (Å)	a = b = 102.3, c = 193.4	a = 103.6, b = 129.0, c = 180.1	a = 73.9, b = 111.3, c = 169.1	a = 114.9, b= 180.1 c = 185.0
α, β, γ (°)	90, 90, 120	90, 90, 90	90, 90, 90	90, 90, 90
Resolution (Å)	96.8–2.80	103.4–3.50	61.6–2.55	180.1–3.55
Rmerge (%)a	3.4 (46.1)b	12.0 (44.9)	5.0 (30.9)	12.3 (64.8)
|/δ|	12.2 (2.1)	5.2 (1.5)	13.8 (2.5)	7.9 (1.8)
Total reflections (#)	49449	41824	72566	172722
Unique reflections (#)	14766	13290	20349	43265
Refinement
Rwork/Rfree (%)c	26.0/28.3	28.9/33.5	21.9/24.9	29.0/31.2
Ramachandran Analysis
Most favored (%/#)	89.5/255	76.0/428	93.8/271	82.1/912
Add. favored (%/#)	10.5/30	23.3/130	6.2/18	17.9/199
Allowed (%/#)	0.0/0	0.7/4	0.0/0	0.0/0
Outlier (%/#)	0.0/0	0.0/0	0.0/0	0.0/0
Rmsd
Bond lengths (Å)	0.009	0.014	0.016	0.009
Bond angles (Å)	1.15	1.45	1.73	1.02

^aR_sym = ƩƩ|I_hkl-I_hkl(j)|/ƩI_hkl. where is I_hkl observed intensity and I_hkl(j)| is the final average value of intensity.

^bValues in parentheses are for the highest resolution shell.

^cR_work = Ʃ||F_obs| - |F_calc||/Ʃ|F_obs| and R_free = Ʃ||F_obs| - |F_calc||/Ʃ|F_obs|; where all reflections belong to a test set of 5% randomly selected data.

Overall structure of the MatP-matS complex: a three domain organization

The E. coli and Y. pestis MatP-matS structures revealed the same overall modular fold comprised of an N-terminal four-helix bundle (Y. pestis residues 14–98, E. coli residues 1–85), connected to a central β-strand-helix-helix (Y. pestis residues 104–145, E. coli residues 91–132) and a C-terminal coiled-coil (Y. pestis residues 146–160, E. coli residues 133–148) (Figure 1A–B). The N-terminal four-helix bundle and strand-helix-helix regions are essentially identical and can be superimposed with root mean squared deviations (rmsd) for corresponding Cα atoms of < 0.8 Å for all Y. pestis and E. coli structures. By contrast, the coiled-coil region displays flexibility and its orientation varies between structures (see below). Data-base searches reveal that MatP contains an overall fold that is distinct from any previously observed structure; only individual domains of MatP show similarity to previously solved structures. In particular, the centrally embedded β-strand-helix-helix region of MatP shows homology to ribbon-helix-helix (RHH) structures found in many prokaryotic DNA-binding proteins (Schreiter and Drennan, 2007) and the four-helix bundle of MatP displays weak structural similarity to the iceberg protein, which is a eukaryotic protein that inhibits interleukin-1β (Figure S2A and S2B).

Molecular basis for dimer formation

The structures show that MatP binds the matS site as a dimer. While contacts between the two-fold related helices of the four-helix bundle provide some dimer interactions, the RHH domain, which alone buries an extensive 4016 Ų of total protein surface from solvent, functions as the main dimerization motif. Bacterial two-hybrid analyses using various segments of MatP confirmed that the RHH region functions as the principal MatP dimerization region. In these assays blue colonies indicate the ability of deleted/altered MatP to interact with wt MatP (Experimental Procedures; Figure 1C). These studies revealed that the region encompassing E. coli residues 97–129 (residues 120–142 in Y. pestis MatP) is essential for MatP dimerization. Single point mutations that abrogated MatP interaction, L104G, Q109G and E127G (Figure 1B), map to α5 and the N-terminal and central regions of α6.

MatP binding to matS

The MatP-DNA structures show that the MatP-bound matS sites are not significantly deformed; an overall bend angle of 8° is observed. The only notable DNA distortion is a general widening of the major groove, which is required to accommodate the two distinct but continuous DNA binding modules of MatP, the four-helix bundle and the RHH. The two α4 helices of the dimer grip the major grooves on each end of the matS site while the RHH sits between the two four-helix bundles of the dimer and clasps the central region of the matS major groove. Residues from the N-terminal arm, helix α1 and α4 of the four-helix bundle make extensive phosphate interactions with the DNA.

The function of MatP as a specific chromosome MD organizer necessitates that it binds exclusively to matS sites, which are located only in the Ter MD. Consistent with this stipulation, FP analyses showed that oligonucleotides lacking the matS sequence show no binding to MatP and mutation of any of the conserved nucleotides in the matS consensus lead to a 10-fold or greater loss in DNA binding affinity (Figure 2A). The crystal structures show that the nucleotides in the 13 base pair (bp) matS consensus are specified by direct contacts from residues in MatP, thus revealing the basis for the exquisite selectivity of MatP for matS (Figure 2B). ChIP-seq analyses demonstrated that matS are the main targets of MatP and that in vivo MatP protects a 15–17 bp region centered on the 13 bp matS consensus sequence within each site (Figure S3A–B and Experimental Procedures). Notably, unlike typical RHH DNA binding proteins, which utilize residues from their ribbons for making selective interactions with nucleobases, MatP uses residues from α4 of its four-helix bundle to mediate base specificity (Figure 2B and 2C). To anchor α4 onto successive major grooves MatP employs a plethora of DNA backbone contacts from its N-terminal arm and four-helix bundle (Figure 2B and 2D). Base contacts from α4 are provided by Gln85, Arg88, Ala89 and Arg93, which specify positions 7/17, 6/18, 9/15 and 10/14 of each half site of the matS signature (Figure 1B and 2B). The close approach of the Gln85 side chain to bp 8/16 requires that a cytosine be located at this position, as this contact would discriminate against a thymine or purine. Thus, interactions from α4 of the four-helix bundle completely specify the matS signature. The RHH of MatP is C-terminal to and contiguous with the four-helix bundle. RHH residues that participate in phosphate interactions are Lys104, Lys105, Trp114, Arg122 and Thr127. The RHH provides two non-specific base contacts. Notably, the same base contacts are observed in the E. coli and Y. pestis structures consistent with the fact that residues that contact the DNA, in particular those found on α1, α4 and β1, are the most conserved in these proteins (Figure 1B).

Figure 2. MatP-matS interactions.

(A) Fluorescence polarization (FP) binding isotherms of Y. pestis MatP binding to wt and mutant matS fluoresceinated 20-mer oligonucleotides (red circles=wt DNA, K_d=2.2 +/ 0.6 nM; blue boxes=A/T and T/A substitutions for G/C and C/G (bps 6/18 and bp 18/6), K_d=58 +/ 12 nM; green diamonds=T/A and A/T substitutions for C/G and G/C (bps 10/14 and 14/10), K_d=23 +/ 7 nM; black crosses=A/T; T/A substations for G/C and C/G (bps 8/16 and 16/8), K_d=20 +/ 1 nM and pink + symbols= non-matS site, no binding). Y. pestis MatP concentration and millipolarization units (mP) are plotted along the X-axis and Y-axis, respectively. Note, DNA binding was not affected by the presence of a his-tag on MatP proteins. (B) Schematic representation of the Y. pestis MatP-matS interactions (numbering according to the Y. pestis protein) showing the contacts made by one MatP dimer to a matS duplex. The matS signature is boxed. (C) Ribbon diagram showing key MatP-DNA phosphate interactions. (D) Ribbon diagram showing base specific contacts made by residues in α4. Interactions to only one half site are shown. (E) Functional analyses revealing MatP regions important for DNA binding in vivo. Shown are plates (in the presence and in the absence of streptomycin) with six representative colonies (wt matP, matP(W15G), matP(T73A), matPΔC31, matPΔC20, and matP(E39V)). Of the mutants shown, MatP(W15G), MatP(T73A) and MatPΔ31 were defective in matS binding (as revealed by lack of streptomycin resistance).

Genetic screens were performed to delineate regions of MatP important in DNA binding in vivo. In these assays, E. coli matP mutants were constructed, both randomly and site specifically, and the ability of the resultant proteins to bind matS was tested by using a streptomycin selection screen (Experimental Procedures). In total, 81 mutants were sequenced and 24 streptomycin sensitive mutants were identified that abrogate matS binding. Figure 2E shows a representative experiment testing the ability of wt MatP, MatP(W15G; Y. pestis residue W28), MatP(T73A; Y. pestis residue T86), C-terminal MatP truncations missing the last 31 and 20 residues (MatPΔC31 and MatPΔC20), and MatP(E39V) (Y. pestis residue D52) to bind matS. MatP(W15G), MatP(T73A) and MatP(ΔC31) were clearly deficient in matS binding. These data are consistent with the structure, which shows that Thr73 makes important phosphate contacts and Trp15 is a key residue is the formation of the hydrophobic core of the four-helix bundle. By contrast, Glu39 points out into the solvent and is not involved in folding or DNA binding. Aside from a few mutants that map to α2 and likely also affect MatP folding, the majority of matS binding defective mutants are located on the N-terminal arm, α1, α4 and β1, which the structures shows form the DNA binding surface of MatP (Figure 1A–B).

MatP tetramerization via coiled-coils interactions

Data show that MatP inactivation leads to decondensation of the Ter region (Mercier et al., 2008). However, the mechanism by which MatP may mediate organization remains unclear. Strikingly, the MatP-matS structures all reveal the presence of a long, flexible tetramer linkage between two coiled-coils that allows MatP dimers to bridge between distant DNA sites (Figure 3A). The longest distance measured between DNA sites in the crystal structures is 135 Å. The tetramer interaction is observed in all the Y. pestis and E. coli MatP-matS structures and the large surface area, 1460 Ų, buried upon its creation, further suggests its physiological relevance (Figure 3A). The formation of the MatP coiled-coil tetramer is mediated by four hydrophobic residues, Leu152, Leu156 and Leu160 and Ile159 (Leu139, Leu143, Leu146 and Leu147 in E. coli MatP), which are highly conserved in MatP proteins (Figure 1B, Figure S4A). In the MatP structure, α6 is a long helix formed by the continuation of the last helix of the RHH with the C-terminal coiled-coil (Figure 1A). Interestingly, the residues that form the RHH/coiled-coil linkage display a marked propensity to kink. This is revealed by a comparison of the multiple crystal structures, which show different degrees of bending at this pivot point (Figure 3A–B; Figure S4B). The large size of the juxtaposed side chains of Lys142 and Tyr145 with their dimer mates in the center of the interacting helices appears to prevent tight packing of the continuous helices, leading to this disruption. Cross contact salt bridges between Lys142 with Asp138, however, stabilize the helical interactions (Figure S4B). Thus, the MatP-matS structures suggest an attractive mechanism that would explain the chromosome organization function of MatP, which is long range DNA bridging via a MatP tetramer. Importantly, the tetramer flexibility observed in multiple MatP-matS structures would also explain how MatP can link distant matS sites within the Ter MD.

Figure 3. MatP forms a flexible tetramer that links distant matS sites.

(A) The structure shows the organization of the MatP-matS tetramer, in which two dimeric complexes are linked via their coiled-coil regions. The crystallographically distinct complexes from three different crystal structures (which contained a total of six crystallographically independent complexes) of Y. pestis MatP-matS and one E. coli MatP-matS structure are shown and labeled. The cyan and green subunits of each MatP tetramer are shown in the same orientation to underscore the flexibility afforded by the coiled-coil interaction. (B) Superimposition of the four-helix bundle-RHH regions of one dimer of each of the independent MatP-matS structures highlighting the malleable coiled-coil interactions between tetramers. (C) Size exclusion chromatography studies on MatP and MatP-matS. In the presence of matS DNA, the MatP-matS complex elutes as a tetramer. The MatP PHI mutant elutes exclusively as a dimer, in the absence and in the presence of matS DNA. The control mutant MatP(V43Q/L53K) in its apo and DNA bound forms eluted the same as wt MatP.

A MatP tetramer model: DNA bridging

To test the proposed bridging model, biochemical, structural and genetic experiments were conducted. In these studies the wt protein was compared to mutant proteins in which the coiled-coil region was either deleted (mutant ΔC20) or in which conserved hydrophobic residues that mediate coiled-coil formation were substituted for glycines (L139G/L143G/L146G/L147G; herein called mutant PHI). Size exclusion chromatography studies supported the tetramer model and clearly demonstrated that the coiled-coil region is required for tetramer formation (Figure 3C). Specifically, in these studies MatP-matS were present exclusively as tetramers, while in the absence of matS, MatP elutes as a mixture of dimer and tetramer forms. A control MatP surface mutant, MatP(V43Q, L43K) eluted the same as wt. By contrast, the MatP PHI mutant eluted entirely as a dimer even in the presence of matS DNA. This result was not due to a loss in DNA binding as this mutant bound the matS site with the same affinity as wt (Figure S1). Hence, these studies show that MatP-matS complexes are tetrameric structures in solution and that the conserved hydrophobic residues in the coiled-coil are essential for tetramerization.

An induced fit mechanism for MatP-matS binding

In order for the MatP dimer to grasp and encase the DNA in the manner observed in the structures necessitates that conformational change(s) in the protein or DNA must occur. Because the MatP-bound matS site shows little deformation, the changes are likely in MatP. To examine the MatP-matS interaction in solution, small-angle x-ray scattering (SAXS) experiments were performed. SAXS envelopes obtained from MatP-matS complex data could only be fit with a MatP-matS tetramer structure, supporting the oligomer organization observed in the crystal structures (Figure 4A). Interestingly, the R_G of ~33 Å determined for apo MatP was much larger than that calculated based on the dimeric MatP structure (22 Å), indicating that apo MatP is less compact than the dimer bound to DNA. Modeling shows that the apo MatP envelope accommodates two MatP molecules, but to fit the data, two MatP subunits had to be individually docked into the density and the subunits had to be divided into sub-regions (Figure 4B). Specifically, residues 99–104 and 125–129 (E. coli numbering) were modeled as flexible regions and the N-terminal four-helix bundle and C-terminal coiled-coil regions had to be treated as independent domains (Figure 4B). These data indicate that DNA-binding by MatP involves a large rotation of the four-helix bundles, which permits the α4 helices to insert into successive DNA major grooves (Movie S1). Such an induced fit DNA-binding mechanism would explain how MatP encases the DNA as observed in the structures (Figure 1A and 3A). This model is supported by circular dichroism (CD), which revealed that while the tertiary structure of MatP is affected by DNA-binding, the secondary structure is unaltered (Figure S5).

Figure 4. SAXS analysis of E. coli apo MatP and E. coli MatP-matS complex.

(A) Model and SAXS analysis of the E. coli MatP-matS complex. (B) Model and SAXS analyses of apo E. coli MatP. In each case, the SAXS envelopes were calculated with DAMMIN and GASBOR and the average envelopes (determined by DAMAVER) are shown as gray spheres. The models were calculated using BUNCH and the Y. pestis MatP structure was used for docking into the SAXS envelopes.

In vivo test of the matS bridging model for Ter MD condensation

Chromosome DNA condensation can be visualized in vivo by a decrease in the interfocal distance (ID) between fluorescent markers in the chromosome region of interest, such as the Ter MD (Mercier et al., 2008). To test the model that the C-terminal coiled-coil is directly involved in Ter condensation, the MatP truncation mutant, MatPΔC20, and mutant PHI were analyzed for their affects on the ID between 103 kb distant markers Ter2 and Ter4 (Figure 5A–B). In a wt MatP strain, the mean ID was close to 0.13 μm as observed before in condensed DNA (Mercier et al., 2008). Strikingly, however, in strains expressing matPΔC20 and the matP PHI mutants (Figure 5B), the mean IDs were close to 0.29 μm and 0.25 μm, values similar to that obtained in the absence of MatP (0.31 μm). The MatP PHI and MatPΔC20 proteins fused with mCherry fluorophore still formed foci (Figure 5C), indicating that these coiled-coil mutants are still localized to the Ter MD. The two mutants were still able to specify a Ter MD as the interactions between markers in the Right and Ter MDs, although reduced in the tetramer mutants, were not completely ablated as in ΔmatP cells (Table S1; Figure 5D). Inactivation of matP leads to a number of phenotypes (Mercier et al., 2008; Espeli et al., 2012). Indeed, while the tetramerization mutants resulted in a marked increase mobility for markers in the Ter MD (Table S1), they behave as wt MatP in other readouts (absence of anucleated or filamentous cells in rich medium, level of colocalization of Ter sister markers). These combined data indicate that MatP tetramerization mutants are still able to form a MD, but that the C-terminal coiled-coil is required for MatP’s ability to condense Ter chromosomal DNA. Specifically, colocalization of markers 120 kb apart is reduced in a MatP mutant that cannot loop.

Figure 5. In vivo Ter condensation by MatP and MatP coiled-coil mutants.

(A) Ribbon diagram of the MatP-matS complex showing the location of the conserved hydrophobic residues in the coiled-coil mediating tetramerization. (B) Genetic map of the Ter region. The positions of the matS sites, marker Ter2 (position 1341 kb) and Ter4 (1444 kb) are indicated. Representative views are shown (scale bars represent 2 μm). The interfocal distance (ID) is estimated in wt matP, matPΔC20, matP-PHI, or ΔmatP context. Histogram shows the number of cells in each category of ID (180 cells were counted). The mean distance is indicated. (C) Long distance DNA interactions revealed by λ Int recombination assay. The recombination frequency between attR located at 17’ in the Right MD and four attL sites in the Right and Ter MDs was measured upon induction of the Xis and Int recombinases in wt and mutant matP cells (matP-PHI, matPΔC20, matP^-). The error bars represent standard deviations of three independent experiments. (D) Fluorescence localization of MatP-mCherry fusion protein (scale bars represent 2 μm). The foci number and distribution are indicated for mCherry labelled wt MatP, MatPΔC20 and MatP-PHI.

Visualization of matS bridging via atomic force microscopy and electron microscopy

Our studies demonstrate that the MatP coiled-coil region is essential for its ability to condense the Ter MD. Further, the combined data suggest that MatP might organize the Ter MD by forming dimer bridges between distant matS sites (Figure 6A). To provide a visual assessment of the MatP bridging model, atomic force microscopy (AFM) and electron microscopy (EM) were carried out. To simulate a section of the Ter MD, a 900 bp DNA strand (matS900) was generated with two matS sites on opposite ends of the DNA separated by 500 bp (Figure 6B). In these experiments, both wt E. coli and Y. pestis MatP proteins (visualized as dots of density accrued on the DNA) mediated looping of the matS900 site as did a control E. coli MatP surface mutant, MatP(V43Q/L53K). The size of the loops and distances between the matS sites are consistent with the locations of the dots of density (MatP) located at the loop juncture. Hence, these images indicate that tetramerized MatP dimers bind to matS sites near the ends of the DNA leading to DNA looping. Strikingly, the MatP PHI protein was unable to mediate loops in the DNA. This was not due to an inability to bind DNA in this assay as multiple images revealed clear density for MatP at each end of the DNA sites (Figure 6B).

Figure 6. DNA loop formation by MatP as ascertained by AFM and EM.

(A) Molecular model for MatP-mediated DNA looping. (B) AFM images of E. coli MatP bound to a 900 bp DNA site containing two matS sites (matS900). The DNA construct used in the AFM analysis is shown (top). AFM images of wt E. coli and Y. pestis MatP bound to matS900 DNA showed looped DNA (~500 bp) between a MatP bound juncture. MatP molecules are evident as regions of high density at the juncture. The MatP V43Q/L53Q (control mutant) and MatP PHI mutants were also visualized by AFM in the presence of matS900. The mutant control still formed loops while the PHI mutant was defective for DNA looping. (C) Images of the 3821 bp fragment carrying 2 matS sites separated by 1060 bp. The scheme of the fragment is shown. (a) Image obtained in the absence of MatP. (b) Image obtained in the presence of wt E. coli MatP bound to the substrate. Different examples are shown (b1-b4). (c) Images obtained in the presence of MatPΔC20. (d) Quantitative analysis of loop formation in the absence (grey bar) or in the presence of 250 nM and 500 nM of wt (green) and ΔC20 (red) MatP forms. (e) Loop size distribution obtained in the absence or in the presence of 500 nM wt (green) or ΔC20 (red) MatP forms. (f-h) Images of the same DNA fragment present as a supercoiled plasmid in the absence or in the presence of wt and ΔC20 MatP forms. The incubation of wt MatP leads to the formation of MatP aggregates coincident with multiple loops. Such loops were barely detected with MatPΔC20. (i) Quantification of the number of supercoiled DNA loops. Loop formation is the same in the absence of MatP and MatPΔC20, whereas in presence of wt MatP, the distribution is quite different with a dramatic increase in the number of average loops to 6. The scale bars in images represents 200 nm.

To further visualize DNA looping by MatP, EM was employed (Figure 6C). MatP was able to induce loop formation on a linear DNA substrate, the proportion of which was dependent on MatP concentration (Figure 6C panels a-d). MatPΔC20 lost this looping property as observed by a significant decrease in loops percentage. As in AFM experiments, MatP bound to DNA were observed as dots of density. Substrates used were long, allowing spontaneous intramolecular crossing even if MatP was absent (16.5% of crossing). To ensure that the wt MatP induced DNA loops were formed by MatP dependant connection of the two mats sites, loops sizes were measured (Figure 6C, panels b and e). The most important class exhibited loops with sizes between 300 and 400 nm, the expected value being 360 nm for a 1.06 kb distance between the two matS sites. On the other hand, in the experiment without MatP or in presence of the truncated MatPΔC20 protein, the loop size was randomly distributed between 100 and 600 nm. The presence of MatPΔC20 bound to DNA was observed as the appearance of bright dots (Figure 6C panel c). The behavior of MatP and MatPΔC20 were also tested on supercoiled plasmids containing two 1.06 kb distant matS sites (Figure 6C panel f-h). While wt MatP promoted DNA compaction by forming loops organized around one nucleation point, MatPΔC20 bound onto DNA without inducing DNA looping. Indeed, loop formation was the same in the absence of MatP and MatPΔC20, whereas in presence of wt MatP, there was a dramatic increase (to 6) in the average number of loops (Figure 6C, panel i). Thus, both AFM and EM experiments revealed that MatP tetramer formation is required for its DNA looping function, supporting the model suggested from our structural studies.

Discussion

MatP is comprised of three contiguous domains

Recent data have indicated that the bacterial chromosome is condensed into four large MDs. Thus far, MatP is the only bacterial MD specific binding protein that has been identified (Mercier et al., 2008). MatP shows no significant sequence homology to any structurally characterized protein and hence how it, or any protein, may specifically bind and compact a chromosomal domain has been unknown. The detailed structural and functional analyses reported here reveals significant insight into that question. Specifically, these studies show that MatP harbors a three domain organization: a large N-terminal four-helix bundle domain required for specific matS binding, a central RHH domain required for MatP dimerization and a C-terminal coiled-coil domain involved in Ter MD condensation. RHH proteins that have been structurally characterized to date use residues from their RHH motif to mediate DNA binding specificity (Schreiter and Drennan, 2007). By sharp contrast, MatP binds the 13 bp matS consensus with high selectivity as a dimer by employing its four-helix bundle domain, which is a heretofore unseen DNA binding motif. In fact, the MatP four-helix bundle provides all base specific contacts, while the RHH makes phosphate contacts and mediates dimerization.

The four-helix bundles completely encase the DNA in the MatP-matS structures indicating that structural alterations must occur in the protein to allow matS binding, as the MatP bound DNA shows little deformation. Comparisons of SAXS data from apo MatP and matS bound MatP indicate that dramatic conformational changes occur in the MatP dimer upon DNA binding, most notably, a large outward rotation of each of the four-helix bundle regions (Movie S1). Interestingly, the DNA induced conformational changes do not appear to involve the RHH. Because the RHH is not required for base specificity it likely functions as an initial docking platform. Under this scenario, RHH binding would anchor the four-helix bundles onto a DNA site. Once docked, the four-helix bundles could then rapidly “test” the DNA for authentication of matS sequences, which once encountered would favor the closed state. Unlike many DNA binding proteins that bind DNA with high specificity, this rotation-insertion-readout mechanism would not require a large scale folding event. Consistent with this, CD studies show that the apo and DNA bound forms of MatP contain the same secondary structures. This induced fit mechanism, which also does not require DNA deformation, may have been selected to avoid nucleotide bending that could impede DNA flexibility around matS sites. Indeed, MatP does not condense DNA by applying a bending mechanism, instead it uses a tetramerization interaction to link distant matS sites.

A MatP-promoted DNA bridging mechanism for Ter MD condensation

Our combined biochemical and SAXS analyses suggest that DNA binding favors the tetramer form of MatP by altering the coiled-coil arrangement (Figure 3C: Movie S1). Importantly, the structures also showed that the coiled-coil is flexible and hence can adopt multiple conformations while still permitting tetramer formation. This may allow MatP to mediate DNA looping at various distances and orientations between matS sites. DNA looping between closely located sites is a process that has received considerable attention since its discovery in the control of transcription in different prokaryotic systems and has since been implicated in a number of additional DNA metabolic processes including replication and recombination (Dunn et al., 1984; Hochschild, 2001). Long-range interactions attributable to proteins that bind to distant enhancer and promoter sites or to insulator sites have also been shown in eukaryotic cells (Sexton et al., 2009). This study provides key molecular insights into how a specific chromosome domain can be compacted by a sequence-specific binding protein via a DNA looping mechanism.

There are likely several levels of organization required to fully condense the bacterial chromosome (for review, Thanbichler and Shapiro, 2006). Recent studies have implicated the H-NS protein in the formation of two clusters per chromosome in which H-NS regulated genes are sequestered (Wang et al., 2011). Because around 5% of genes in E. coli can be regulated by H-NS, the formation of such clusters may impact the general folding of the E. coli nucleoid. Another hierarchy of DNA organization involves the formation of MDs. Our data show that chromosomes harboring MatP tetramer mutations, although having less defined Ter MD foci, still appear to form Ter MD (as revealed by genetic localization studies). However, ablation of MatP tetramerization prevents condensation of the Ter MD as evidenced by colocalization of markers. Thus, MatP appears to mediate two levels of MD organization. How MatP-mediated Ter MD structuring is integrated in the control of DNA metabolism is currently unknown. In the 800 kb long Ter MD, matS sites are located ∼35 kp from each other. MatP bound to matS sites has been reported to inhibit supercoiling slithering in the Salmonella chromosome (Booker et al., 2010). However, as suggested by the unbiased distribution of matS sites in open reading frames or in intergenic regions, MatP does not perturb gene expression. Interestingly, matS sites in enterobacteria, although all located within the Ter MD, do not appear to be specifically arranged relative to one another, suggesting that a fixed “pattern” of organization is not likely. Thus, it is interesting to speculate that the tetramer bridging contacts between MatP dimers may be somewhat dynamic, or subject to slow exchange as this would permit DNA compaction while still allowing biological processes to occur without impediment. In addition, chromosomal DNA in prokaryotes exists as supercoiled plectonemes that have been estimated to encompass regions ranging from 10 to 100 kb (Higgins et al., 1996; Stein et al., 2005; Postow et al., 2004). Hence, MatP bound to matS sites may alter the dynamics of supercoiled DNA over the entire Ter MD and impose a more or less rigid plectonemic bouquet. A future challenge is to achieve a full understanding of how MatP modulates the spatial organization of the DNA molecule inside the Ter MD at a molecular level.

Experimental Procedures

in vivo binding of MatP to matS.

All strains and plasmids used in this study are listed in Tables S2 and S3. E. coli K12 strains are all derivatives of MG1655. Binding of mutant MatP proteins to the matS site was assayed as previously described (Mercier et al., 2008). pTSA plasmids carrying p(matP)-matP-lacZ were co-transformed with the pRM1 plasmid (where matS is inserted in the rpsL promoter) into a MG1656 strain carrying the rpsL20 allele.

Bacterial two-hybrid assay.

Two-hybrid studies were performed using the bacterial adenylate cyclase assay (Karimova et al., 1998). For MatP dimerization assays, full-length wt matP was cloned into the pKT25 plasmid whereas mutant or deleted matPs were cloned into the pUT18C plasmid.

Light and Fluorescence Microscopy.

For fluorescence microscopy, samples were imaged with a Leica DM6000 microscope equipped with a heated stage at 30 °C and a coolsnap HQ CCD camera and Metamorph software. Analyses were performed using ImageJ software. The Stackreg and the MtrackJ plugins were used for the analysis of time-lapse images and the Pointpicker plugin was used for snapshot analyses. Custom-made Excel sheets were utilized to monitor the cell size, focus positions, ID, MSD and signal intensity. For each snapshot experiment, at least 200 cells were counted.

Purification of MatP proteins, crystallization of MatP-matS complexes and structure determination.

Artificial genes encoding the E. coli and Y. pestis MatP proteins were codon optimized for E. coli expression and purchased from Genscript Corporation, Piscataway, NJ, USA; Web:www.genscript.com. Both genes were subcloned into the pET15b vector such that the N-terminal hexa-his tag was expressed for purification. The vectors were transformed into E. coli BL21(DE3) cells for expression. Proteins were purified in a single step using Ni-NTA chromatography. Crystals of Y. pestis MatP-matS 16mer complexes were obtained with PEG 3000, 0.1 Tris pH 7.0 and 0.2 mM CaCl₂ (P6₅222 form) and PEG 4000, 0.1 M Mes pH 6.5, 0.1 M calcium acetate (I222 form). Crystals of Y. pestis MatP-matS 23mer in which the hexa-histidine tag had been cleaved from MatP were obtained using PEG 400, 20 mM MgCl₂ and 50 mM Mes 6.5. Crystals of E. coli MatP in which the hexa-histidine tag had been cleaved were obtained with a 19mer matS site using 5% PEG 1000, 20 100 mM acetate pH 5.0, 40% ethylene glycol. All data were collected at ALS beamline 8.3.1 and processed with MOSFLM (Table 1). The P6₅22 Y. pestis MatP-matS 16mer structure was solved by MIR and the other structures by molecular replacement.

Small angle X-ray scattering analyses on MatP and MatP-matS complexes.

SAXS data on the E. coli MatP and E. coli MatP-matS (ratio of 1 MatP dimer:1 DNA duplex) complex were collected at ALS beamline 12.3.1 for each sample over a concentration range of 1 to 6 mg/ml. The radius of gyration (Rg) was determined by the Guinier approximation. The program GNOM was used to compute the pair distance distribution function, P(r), of each and the overall shapes were restored from the experimental data using the program DAMMIN or GASBOR (Svergun et al., 2001). BUNCH was used to produce and evaluate 3-D models in which the crystal structure was docked into density envelopes (Petoukhov and Svergun, 2005).

Fluorescence polarization (FP).

Fluorescence polarization assays were performed with a PanVera Beacon 2000 fluorescence polarization system. Samples were excited at 490 nm, and fluorescence emission was measured at 520 nm. All oligonucleotides used in these assays contain a 5´ fluorescein label. Each assay was carried out with 1 nM oligo in the binding buffer (150 mM NaCl and 20 mM Tris-HCl pH 7.5). The polarization data was analyzed with KaleidaGraph and fitted to a simple bimolecular binding model by nonlinear regression.

Atomic force microscopy (AFM) and electron microscopy (EM) analyses on MatP-DNA complexes.

The 900 bp DNA construct (matS900) used in AFM studies contains two MatP binding sites with the sequence 5′-GTGACAATGTCAC-3′. These binding sites were introduced in to the pET15b plasmid at positions 1501 and 2001 via QuikChange to produce the plasmid matS2S5. Subsequently the matS900 construct was generated through PCR amplifying of position 1300–2200 with the following forward and reserve primers: 5′-CGGCGTTATTTCTTGATGT-3′ and 5′-ACCAGTGACGAAGGCTTGAG-3′. Wt MatP and mutant proteins were mixed with DNA substrate (matS900) at the protein/DNA ratio 80:1 in the presence of binding buffer (150 mM NaCl, 20 mM Tris pH 7.5) at room temperature for 5 min. For AFM, wt and mutant MatP proteins were deposited on specially modified mica surfaces (APS mica) obtained by incubation of freshly cleaved mica as described (Shlyakhtenko et al., 2003; Lyubchenko et al., 1997). AFM images in air were acquired using MultiMode AFM NanoScopIV system operating in tapping mode. For EM studies, DNA PCR products containing one or two matS sites were cloned into pUC18. Negative supercoiled pUC18-matS plasmids were purified using cesium chloride. pUC18 with two matS sites was digested by SspI in order to obtain the linear pUC18-2matS substrate. wt MatP or MatPΔC20 proteins were incubated with DNA substrates with or without matS sites in a buffer consisting of 20 mM Tris pH 7.5 and 150 mM NaCl at a protein to DNA ratio of 1:50 in a total volume of 20 μl for 10 min at 37° C. For EM studies wt MatP or MatPΔC20 proteins were incubated with DNA substrates with or without matS sites. TEM sample preparation was performed as previously described (Dupaigne et al., 2008).

ChIP-seq analysis

ChIP assays were performed as described before (Mercier et al., 2008). DNA segments bound by MatP were identified by sequencing on an Illumina GAIIX genome analyzer (Imagif Platform, CNRS, Gif-sur-Yvette).

Highlights.

• ►MatP is a specific DNA-binding protein that compacts enterobacteria chromosomal DNA.

• ►MatP contains a modular fold with a four-helix bundle DNA binding motif.

• ►MatP coiled-coils form tetramers that link distant DNA sites.

• ►Tetramerization is required for MatP-mediated in vivo chromosomal condensation.