The nucleoid protein Dps binds genomic DNA of Escherichia coli in a non-random manner

Abstract

Dps is a multifunctional homododecameric protein that oxidizes Fe2+ ions accumulating them in the form of Fe2O3 within its protein cavity, interacts with DNA tightly condensing bacterial nucleoid upon starvation and performs some other functions. During the last two decades from discovery of this protein, its ferroxidase activity became rather well studied, but the mechanism of Dps interaction with DNA still remains enigmatic. The crucial role of lysine residues in the unstructured N-terminal tails led to the conventional point of view that Dps binds DNA without sequence or structural specificity. However, deletion of dps changed the profile of proteins in starved cells, SELEX screen revealed genomic regions preferentially bound in vitro and certain affinity of Dps for artificial branched molecules was detected by atomic force microscopy. Here we report a non-random distribution of Dps binding sites across the bacterial chromosome in exponentially growing cells and show their enrichment with inverted repeats prone to form secondary structures. We found that the Dps-bound regions overlap with sites occupied by other nucleoid proteins, and contain overrepresented motifs typical for their consensus sequences. Of the two types of genomic domains with extensive protein occupancy, which can be highly expressed or transcriptionally silent only those that are enriched with RNA polymerase molecules were preferentially occupied by Dps. In the dps-null mutant we, therefore, observed a differentially altered expression of several targeted genes and found suppressed transcription from the dps promoter. In most cases this can be explained by the relieved interference with Dps for nucleoid proteins exploiting sequence-specific modes of DNA binding. Thus, protecting bacterial cells from different stresses during exponential growth, Dps can modulate transcriptional integrity of the bacterial chromosome hampering RNA biosynthesis from some genes via competition with RNA polymerase or, vice versa, competing with inhibitors to activate transcription.

Fig 1. Large-scale profiles of the Dps targets correlate with the landscape of direct and inverted repeats and the pattern of Fis binding sites.

A: Distribution of the Dps contact regions along the E. coli MG1655 genome identified by CLC GW in two experiments (the default settings). The areas covered by Dps are combined in 100,000 bp bins and plotted as percentage to the total length of all sites occupied by Dps. B: The same for the contact sites of Fis [49], IHF [51], H-NS [49] and RNAP [50] from the cells grown in conditions similar to those used in our experiments. The plot for Dps shows the distribution of the sites from the combined set (CS) (Page 3, S3 Table). C: The same for direct (5–24 bp separated by 1–15 bp) and inverted (5–18 bp separated by 3–20 bp) repeats collected from the genome of E. coli MG1655 using Unipro UGENE [53].

Fig 2. Dps binds genomic DNA in a non-random manner.

A: Correlation between the number of reads from immunoprecipitated and control sequence libraries calculated for 100 bp bins (first experiment). All points, corresponding to rRNAs operons are marked in red. All other points with R-values higher than 1.4 are plotted in blue. Dashed line shows the bisectrix of the plot. B: Distribution of the Dps-binding sites in the genome. Two outer circles represent the gene map of the top and the bottom strands of the E. coli MG1655 genome. The red ticks on the third circle mark positions of rRNA operons. The profile of R-values and the distribution of reads registered in the control library calculated for a 35 bp running window are plotted on the fourth and fifth circles, respectively.

Fig 3. Dps-binding sites are enriched with inverted repeats.

The overlap between sequences of CS (colored box-plots) and UR (gray boxes) with direct or inverted repeats was characterized by the parameter K_ij as described in the text. Black dot on the right panel shows one outlier. Box-plots with statistically significant differences are provided with corresponding p-values. Regions of bound and unbound sets overlapping with repeated sequences of both types are indicated in S3 Table.

Fig 4. Dps shares its binding sites with other proteins of bacterial nucleoid and has affinity to REP-elements and promoter islands.

Intersection of Dps targets (A) and sites unbound by Dps (B) with structural and functional elements of bacterial genome was estimated as described above for repeated sequences and plotted as fold ratio to the expected values. Bent black arrows on the bottom schematically show areas occupied by CS or UR. Gray rectangles and numerals inside indicate the expected number of common base pairs if compared modules are independently distributed along the genome. Gray and colored bent arrows show registered overlap calculated in 1 bp resolution. Numerals in parenthesis indicate the size of compared sets. Genomic locations of REP elements were taken from KEGG DataBase (http://www.genome.jp/kegg/, [63]), and fold ratios obtained for 302 REP-sequences containing 1–3 REP-modules (14–100 bp) were plotted. Analyzed ChIP-chip and ChIP-seq data sets were obtained from [–51] for cells grown in LB medium (LB), M9 medium with fructose (M9) or MOPS minimal medium with glucose (GMM), harvested at early (EE), middle (ME) or late (LE) exponential phase or upon transition to the steady growth (TS).

Fig 5. Deletion of the dps gene affects rpoA and rpoB expression.

A: Profiles of the Dps binding sites obtained in two experiments (indicated) for the genomic region with three operons of ribosomal genes (running window of nine 35 bp bins). Genes are represented by blue horizontal arrows; magenta lines show ribosomal operons. Vertical arrows mark locations of inverted repeats (if longer than 7 bp). B: Band shift assays performed for indicated genomic loci. The regulatory region of the dps gene was used as a positive control for all band shift assays in this study. Fragment from the lacZ coding sequence was used as a reference gene for qRT-PCR experiments. Positioning of primers for amplification (F and R) is indicated in panel A of this figure, Fig 6A (for the dps regulatory region) and in S3 Fig (for rpoA and lacZ). C: Changes in the expression efficiency of selected genes in response to dps deletion. Primers used for reverse transcription and consecutive PCR are designated as RT and PCR, respectively, here and all other figures. Expression levels were estimated based on 3 and 5 biological samples (3–18 technical repeats in each) for rpoA and rpoD, respectively. Error bars show an average deviation. Statistical significance was assessed using Student’s t-test.

Fig 6. qRT-PCR and reporter assays revealed apparent positive autoregulation of the dps gene.

A: Profiles of the Dps binding sites obtained in two experiments for the dps genomic locus (running window of nine 35 bp bins). Genes are displayed as blue horizontal arrows. Vertical arrows show locations of inverted repeats, if longer than 7 bp (blue). Primer pair F1-R2 was used for cloning the whole regulatory region in pET28b_eGFP. The dps_short fragment (positive control in all band-shift assays) was amplified with F2-R2. B: Scheme of the pET28b-eGFP reporter vector with the insert of the dps promoter region. C: Images of E. coli MG1655 cell colonies transformed by pET28b-eGFP plasmid with or without the promoter insert. Images were obtained by fluorescent microscope Leica with 2 or 0.5 sec exposition for control and experimental cells, respectively. D: Changes in the expression efficiency of gfp in response to dps deletion measured in qRT-PCR using kan as a reference gene. Expression levels were estimated based on 3 biological samples with 3 technical repeats in each. Error bars show an average deviation. Statistical significance was assessed using Student’s t-test.