Mapping the Regulatory Network for Salmonella enterica Serovar Typhimurium Invasion

Abstract

Salmonella enterica pathogenicity island 1 (SPI-1) encodes proteins required for invasion of gut epithelial cells. The timing of invasion is tightly controlled by a complex regulatory network. The transcription factor (TF) HilD is the master regulator of this process and senses environmental signals associated with invasion. HilD activates transcription of genes within and outside SPI-1, including six other TFs. Thus, the transcriptional program associated with host cell invasion is controlled by at least 7 TFs. However, very few of the regulatory targets are known for these TFs, and the extent of the regulatory network is unclear. In this study, we used complementary genomic approaches to map the direct regulatory targets of all 7 TFs. Our data reveal a highly complex and interconnected network that includes many previously undescribed regulatory targets. Moreover, the network extends well beyond the 7 TFs, due to the inclusion of many additional TFs and noncoding RNAs. By comparing gene expression profiles of regulatory targets for the 7 TFs, we identified many uncharacterized genes that are likely to play direct roles in invasion. We also uncovered cross talk between SPI-1 regulation and other regulatory pathways, which, in turn, identified gene clusters that likely share related functions. Our data are freely available through an intuitive online browser and represent a valuable resource for the bacterial research community.

Importance: Invasion of epithelial cells is an early step during infection by Salmonella enterica and requires secretion of specific proteins into host cells via a type III secretion system (T3SS). Most T3SS-associated proteins required for invasion are encoded in a horizontally acquired genomic locus known as Salmonella pathogenicity island 1 (SPI-1). Multiple regulators respond to environmental signals to ensure appropriate timing of SPI-1 gene expression. In particular, there are seven transcription regulators that are known to be involved in coordinating expression of SPI-1 genes. We have used complementary genome-scale approaches to map the gene targets of these seven regulators. Our data reveal a highly complex and interconnected regulatory network that includes many previously undescribed target genes. Moreover, our data functionally implicate many uncharacterized genes in the invasion process and reveal cross talk between SPI-1 regulation and other regulatory pathways. All datasets are freely available through an intuitive online browser.

FIG 1

Direct regulation of the invF-containing transcript by HilD, HilC, and RtsA. Data represent results of RNA-seq and ChIP-seq analysis of HilD, HilC, and RtsA, for the region encompassing invF and invH. Red arrows represent genes. RNA-seq graphs show sequence read density for one replicate experiment for cells lacking the indicated TF or for cells transiently overexpressing the indicated TF. ChIP-seq graphs show sequence read density for one replicate experiment for the indicated TF. The genes are colored red to indicate positive regulation by the TFs. Black arrowheads indicate ChIP-seq peaks. RNA-seq data are normalized according to the genome position with the 90th percentile for sequence read coverage.

FIG 2

Regulatory network associated with SPI-1. The hexagon-shaped nodes represent TF-encoding genes hilD, hilC, rtsA, invF, sprB, and rtsB. The circular nodes represent target genes directly regulated by these TFs. The directed edges indicated with pink lines represent the regulatory relationships among the TFs. The directed edges indicated with black lines represent the regulatory relationships between the TFs and their target genes. The node color indicates the log ratio of the differential levels of expression (red = TF activated; blue = TF repressed), and the color scale is shown. Note that the regulatory targets of HilA are omitted because ChIP-seq for this protein was unsuccessful. Where possible, common gene names are shown. For genes without a common name, the name of the orthologous gene from S. Typhimurium strain LT2 is shown (“STMxxxx” names). For annotated genes without a common name and without an orthologue in S. Typhimurium strain LT2, the S. Typhimurium 14028s gene name is shown (“STM14_xxxx” names). For unannotated genes, we list a brief description of the transcript location (“anti” = antisense to the indicated gene; “sRNA_internal” = initiation within the indicated gene, in the sense orientation relative to that gene; “sRNA_hilD/hilA” = initiation in the intergenic region between hilD and hilA, antisense to the hilA 5′ UTR).

FIG 3

Overlap of the HilD, HilC, and RtsA regulons. (A) Venn diagram showing the degree of overlap of direct regulatory targets between HilD, HilC, and RtsA. (B) Heat maps showing overlap of ChIP-seq peaks for HilD, HilC, and RtsA. Each block shows data for two or three TFs at a single 400-bp genomic region, centered on a ChIP-seq peak for HilC. The combined, normalized sequence read count from the ChIP-seq data for HilC is shown in the first row of each block, with color intensity representing read density. The one or two additional rows in each block show equivalent data for HilD or RtsA or both, as indicated. (C) Sequence motifs for HilC and RtsA. (D) Association of ChIP-seq peaks with regulation for HilD, HilC, and RtsA. Each of the histograms represents all ChIP-seq peaks (i.e., binding sites for the corresponding TF). Each bar represents a single ChIP-seq peak, and the height of the bar indicates the level of binding (FAT score). Binding sites associated with regulation are indicated by red bars. Binding sites not associated with regulation are indicated by black bars.

FIG 4

RtsB binding sites that overlap those of RcsB. Data represent results of comparison of the sequence motif for RtsB to the sequence motif for E. coli RcsB taken from the PRODORIC database (90).

FIG 5

Identification of regulated sRNAs and 5′ UTRs. (A) Data represent results of RNA-seq and ChIP-seq analysis of RtsA, for the region encompassing dacB. The sRNA that initiates within dacB is indicated by a red bracket. (B) RNA-seq and ChIP-seq data for RtsA, for the region encompassing slrP. The sRNA that initiates antisense to slrP is indicated by a red bracket. Note that the RNA-seq data for the minus strand are zoomed 15 times relative to the data for the plus strand. (C) RNA-seq and ChIP-seq data for RtsB, for the region encompassing galF. The 5′ UTR of a galF transcript that is regulated by RtsB is indicated by a blue bracket. In all panels, the genes are colored red or black to indicate positive regulation or no regulation by the TFs. Black arrowheads indicate ChIP-seq peaks.

FIG 6

Extensive cross talk between SPI-1 regulators, and predicted network expansion. (A) The hexagon-shaped nodes represent TF genes hilD, hilC, rtsA, hilA, invF, sprB, and rtsB. The directed edges indicated with pink lines represent the regulatory relationships among these TFs. The node color indicates whether the TF is a positive (red) or negative (blue) regulator. (B) Regulation of known and predicted regulators by SPI-1 TFs. Positive regulation is indicated by lines with arrowheads. Negative regulation is indicated by lines with flat heads.

FIG 7

Hierarchical clustering of gene expression profiles reveals connections between genes regulated by SPI-1-associated TFs. The central heat map shows pairwise correlations between gene expression profiles (data from reference 4) for every direct regulatory target of HilD, HilC, RtsA, InvF, SprB, RtsB, and HilA. Each row/column represents a gene, with genes arrayed in identical orders in rows and columns (note the symmetry of the heat map). Stronger yellow colors indicate higher correlation coefficients, as indicated on the scale (bottom right). Columns indicated by arrows represent genes whose expression profile has an average correlation coefficient of >0.5 for comparisons to all genes located within SPI-1. Blue arrows indicate known invasion genes. Red arrows indicate likely novel invasion genes. The four dashed squares highlight clusters of genes whose expression profiles correlate strongly. These clusters are enlarged and annotated below the heat map.