Selective Silencing of Foreign DNA with Low GC Content by the H-NS Protein in Salmonella

See allHide authors and affiliations

Science  14 Jul 2006:
Vol. 313, Issue 5784, pp. 236-238
DOI: 10.1126/science.1128794


Horizontal gene transfer plays a major role in microbial evolution. However, newly acquired sequences can decrease fitness unless integrated into preexisting regulatory networks. We found that the histone-like nucleoid structuring protein (H-NS) selectively silences horizontally acquired genes by targeting sequences with GC content lower than the resident genome. Mutations in hns are lethal in Salmonella unless accompanied by compensatory mutations in other regulatory loci. Thus, H-NS provides a previously unrecognized mechanism of bacterial defense against foreign DNA, enabling the acquisition of DNA from exogenous sources while avoiding detrimental consequences from unregulated expression of newly acquired genes. Characteristic GC/AT ratios of bacterial genomes may facilitate discrimination between a cell's own DNA and foreign DNA.

H-NS (encoded by hns) belongs to a family of small abundant nucleoid-associated proteins of Gram-negative bacteria that have the ability to bind DNA with relatively low sequence specificity (1). H-NS has been shown to act as a transcriptional repressor and can multimerize into higher order bridging complexes after DNA binding (1, 2). H-NS also affects local supercoiling, leading to the suggestion that H-NS and other nucleoid binding proteins represent the bacterial functional equivalent of histones or mediate the global modulation of gene expression in response to changes in temperature or osmolarity (1, 3, 4). H-NS has higher affinity for curved DNA, and no consensus sequence has been identified, although the few H-NS binding sites mapped to date are rich in AT.

To determine the Salmonella genes controlled by H-NS, construction of an hns null mutation was attempted in Salmonella enterica sv. Typhimurium (S. Typhimurium). However, hns mutant strains were found to be nonviable unless additional mutations were present in either rpoS encoding the alternative sigma factor σS32) or phoP encoding the virulence gene regulator PhoP. These mutants could tolerate an hns mutation but exhibited a reduced growth rate, whereas an hns mutation in an rpoS phoP double-mutant background displayed growth similar to that of the wild type (fig. S1). This suggests that the detrimental effect of an hns mutation is due to derepression of one or more σS- and PhoP-activated loci and might explain why hns mutations are not lethal in some laboratory strains, given that rpoS mutant alleles are commonly acquired after laboratory passage (5). To enable studies of H-NS function in Salmonella, we constructed an hns mutation in S. Typhimurium carrying a spontaneous rpoS mutation that confers diminished σS activity (6).

Salmonella genes regulated by H-NS were identified by comparing transcript levels in hns+ and hns strains by cDNA microarray analysis (Fig. 1 and table S1). Of 4529 open reading frames (ORFs) represented on the array (4422 from the chromosome and 107 from the virulence plasmid), transcript levels of 178 ORFs exhibited a reduction in abundance in the hns mutant to a level less than one-third that of the wild type, whereas 409 transcripts were more abundant in the hns mutant. As previously reported for Escherichia coli, many H-NS–activated genes are involved in chemotaxis and motility (7). Among genes repressed by H-NS are many known virulence loci of Salmonella, including the Salmonella pathogenicity island (SPI) 2, SPI-3, and SPI-5, most characterized virulence islets, and the plasmid spv genes.

Fig. 1.

H-NS–mediated repression correlates with regions acquired by means of horizontal transfer and low GC content. (A) A cDNA microarray analysis comparing the Salmonella wild type and hns mutant strains for the first third of the S. Typhimurium chromosome (8) (see table S1 and fig. S4 for data and plot of the entire Salmonella genome). The x axis corresponds to the nucleotide numbers of the published S. Typhimurium LT2 genome. The y axis represents fold hns-dependent induction (up) or repression (down) of various ORFs. Gray boxes indicate regions of the S. Typhimurium chromosome that appear to have been acquired by means of horizontal transfer, as determined by comparison with related enteric genomes (8). S. Typhimurium pathogenicity islands SPI-2 and SPI-5 are indicated by blue boxes; Gifsy-2 prophage region is indicated by an orange box; a region of the LT2 genome absent from the strain used in these studies is marked by a dark vertical stripe. (B) Scatter plot of H-NS–dependent expression (x axis) of 4696 S. Typhimurium ORFs plotted against their corresponding GC content (y axis). Genes whose expression varied by less than threefold in the absence of H-NS are shaded in gray. The average GC content of S. Typhimurium is indicated (dashed line), and the region ± 2.5% from the average is shaded in blue.

A large number of H-NS–repressed genes bear the hallmarks of acquisition from a foreign source—i.e., they are not universally present in the genomes of closely related enteric bacteria and possess substantially reduced GC content compared with that of the resident genome (8). Of 409 ORFs exhibiting repression by H-NS, only 40 (9.8%) are common to all reference genomes, whereas 265 (64.7%) are found exclusively in Salmonella (table S1). Most H-NS–repressed genes have GC content that is lower than the overall genome: For ORFs in which H-NS was found to repress expression to one-third or less of the original level, the average GC content is 46.8%, whereas the average GC content of the entire Salmonella LT2 genome is 52.2% (Fig. 2A).

Fig. 2.

H-NS binding is directed to low-GC regions of the chromosome. (A) H-NS binding measured by ChIP-on-chip (pink) plotted against the corresponding local average GC content (2000-nucleotide window, blue; for clarity only regions where GC content is <50% are included) for one-third of the Salmonella chromosome. Table S2 plots of the entire Salmonella chromosome including all GC content percentage values. H-NS binding (y axis, right) is the fold difference in signal intensity over the negative control (6). The LT2 genome region absent from the strain used in these studies is indicated (dark stripe). (B) Results of ChIP-on-chip analysis plotted against the GC content of respective Salmonella ORFs. ORFs are arranged in rank order according to H-NS binding (6) (table S2). The dark blue curve shows the moving average of ORF GC content (window = 20 ORFs). ORFs displaying greater than threefold binding to H-NS are shaded in light blue. The average GC content of S. Typhimurium is indicated (dashed line).

Microarray analysis of cDNA provides only indirect evidence that a transcription factor interacts directly with a given sequence, because many regulatory interactions are dependent on a cascade of transcription events. For example, the apparent activation of flagellar genes by H-NS most likely occurs by means of H-NS–mediated repression of hdfR, a repressor of the flagellar regulators flhDC (1). Salmonella genes that interact directly with H-NS were therefore determined by chromatin immunoprecipitation (ChIP) of in vivo cross-linked H-NS–DNA complexes followed by microarray (ChIP-on-chip) analysis on either a custom ORF array or a tiled oligonucleotide array with 385,000 features [NimbleGen, Madison, WI (6)].

Of the 4438 chromosomal genes covered in the ORF array, 745 (16%) coimmunoprecipitated with H-NS. A notable correlation was observed between H-NS binding predicted in silico by low GC content and binding measured experimentally by ChIP (Fig. 2). Only 5 of 745 precipitated sequences (0.7%) were not situated within 1000 nucleotides of a chromosomal region displaying an average GC content of ≤49% (averaged over a 1000-nucleotide span) (Fig. 2 and table S2). Of 615 ORFs in the annotated Salmonella genome with a GC content of <47%, 433 (70.4%) coprecipitated with H-NS (table S2). The oligonucleotide array provided detailed resolution of H-NS binding sites and revealed a strong correlation between H-NS binding and regional AT content, whether or not the site was in a promoter (fig. S2). Some horizontally transferred sequences (notably SPI-1 and SPI-4) in which cDNA microarray analysis demonstrated only slight transcriptional repression were found to interact directly with H-NS, suggesting either that H-NS binding is not an effective silencer at all binding sites or that conditions used during the cDNA analysis did not favor expression or silencing of these genes.

To prospectively test whether H-NS is capable of targeting AT-rich DNA from a foreign source, a gene from Helicobacter pylori [hp0226, GC content = 39.7%] was recombined along with its promoter into a nonessential region of the Salmonella chromosome with a uniform average GC content of >50% and no demonstrable interaction with H-NS (Fig. 3A and fig. S3). ChIP with quantitative polymerase chain reaction (ChIP/Q-PCR) revealed significant association of H-NS with hp0226 but not with the adjacent gene stm1033 (GC content = 52.5%) (Fig. 3B). Reverse transcriptase/Q-PCR measurement of transcript levels revealed significantly higher (>15-fold) hp0226 expression in the hns mutant compared with that of the wild type (Fig. 3C). We concluded that AT-rich content per se is sufficient for H-NS–mediated silencing and that H-NS can target AT-rich sequences irrespective of chromosomal location, a finding that could be exploited to optimize the expression of foreign genes in enteric bacteria for applications in biotechnology.

Fig. 3.

H-NS binds and silences an experimentally introduced AT-rich foreign gene. (A) Region of the Salmonella chromosome into which hp0226, from Helicobacter pylori, was recombined (6). Q-PCR primers for hp0226 and stm1033 are designated with orange triangles. The 3′ fragment of hp0227 (hatched arrow) and Flp recombinase target site (FRT) are indicated. Putative hp0226 and stm1032 promoters are indicated (thin dashed arrows). Lower panel is the moving average (300-nucleotide window) of the regional GC content. Area corresponding to the Helicobacter sequence is shaded in gray. A transcriptional terminator (ter) was included to prevent read-through from the promoter upstream of the stm1032 gene into the Helicobacter sequence. The average GC content of S. Typhimurium is indicated (dashed line), and the region ± 2.5% from the average is shaded in blue. (B) ChIP analysis of H-NS–DNA complexes. Data are shown as fold enrichment normalized to the marA gene, which does not interact with H-NS (6). H-NS coprecipitates with proV (positive control) and hp0226 but not stm1033. Error bars encompass a 95% CI. (C) H-NS–mediated silencing of hp0226 and proV but not marA (negative control). Reverse transcriptase/Q-PCR analysis of transcript levels in hns and isogenic wild-type Salmonella normalized to mRNA of the gyrB housekeeping gene (6). Fold derepression is the amount of transcript in the hns mutant divided by wild-type transcript levels. Error bars encompass a 95% CI.

Our observations suggest a previously unrecognized role for H-NS in recognizing AT-rich sequences as foreign and preventing their expression. Such “xenogeneic silencing” would protect the cell from detrimental consequences of invading DNA. An association has been noted between H-NS and some horizontally transferred genes in pathogenic E. coli (912), Shigella spp. (13, 14), Vibrio cholerae (15), Proteus mirabilis (16), Yersinia spp. (17), and Erwinia chrysanthemi (18), but it has not been satisfactorily explained (1). A number of virulence regulators [e.g., SlyA (19), RovA (17), Ler (20), CRP-PapB (21), and ToxT (22)] act as antisilencers by displacing H-NS at specific promoters. A model in which H-NS exploits low GC content to silence horizontally acquired sequences provides a unifying explanation for these disparate observations across several bacterial species. The evolutionary development of selective countersilencing mechanisms provides a means by which an organism is protected from adverse consequences of foreign DNA but can nevertheless selectively activate individual loci that confer a fitness advantage.

Our model provides a bacterial analog to the silencing of transposons and mobile genetic elements by heterochromatin/RNA interference in eukaryotes (23). Similar to heterochromatin, H-NS appears to function as a silencer of potentially harmful sequences and to exert control over local nucleoid structure. However, whereas histones are highly conserved among eukaryotes, the primary sequence of H-NS is poorly conserved among bacteria outside of the Enterobacteriaceae (24). Furthermore, the almost exclusive restriction of H-NS binding to AT-rich regions and the maintenance of nucleoid domain structure despite the absence of H-NS (25, 26) suggest that the primary role of H-NS is to silence foreign DNA. A degenerate recognition sequence and the ability to polymerize along and bridge adjacent stretches of DNA ideally suit H-NS for this role and likely account for many of its reported functions, including alteration of recombination events and local supercoiling (1, 27, 28).

The genome-wide average GC content of various bacterial genera can vary from 25 to 75%, and attempts have been made to explain why bacterial genomes maintain their distinctive GC bias (29). The reason for the relative AT richness of horizontally transferred DNA in enteric bacteria has also been enigmatic. We posit that the cell's ability to discriminate between its own DNA and foreign DNA on the basis of differences in GC content can provide a fitness advantage and that xenogeneic silencing has likely shaped bacterial genomes by facilitating the acquisition and preservation of AT-rich DNA. Interestingly, some AT-rich bacteriophages, pathogenicity islands, and mobile genetic elements encode H-NS antagonists (1, 30), indicating that “selfish” genetic elements have evolved countermechanisms to escape H-NS–mediated silencing. Further work will be required to determine whether additional mechanisms of xenogeneic silencing exist in other bacterial species.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1 and S2

References and Notes

View Abstract

Navigate This Article