Chromosomal origin of replication coordinates logically distinct types of bacterial genetic regulation

Abstract

For a long time it has been hypothesized that bacterial gene regulation involves an intricate interplay of the transcriptional regulatory network (TRN) and the spatial organization of genes in the chromosome. Here we explore this hypothesis both on a structural and on a functional level. On the structural level, we study the TRN as a spatially embedded network. On the functional level, we analyze gene expression patterns from a network perspective ("digital control"), as well as from the perspective of the spatial organization of the chromosome ("analog control"). Our structural analysis reveals the outstanding relevance of the symmetry axis defined by the origin (Ori) and terminus (Ter) of replication for the network embedding and, thus, suggests the co-evolution of two regulatory infrastructures, namely the transcriptional regulatory network and the spatial arrangement of genes on the chromosome, to optimize the cross-talk between two fundamental biological processes: genomic expression and replication. This observation is confirmed by the functional analysis based on the differential gene expression patterns of more than 4000 pairs of microarray and RNA-Seq datasets for E. coli from the Colombos Database using complex network and machine learning methods. This large-scale analysis supports the notion that two logically distinct types of genetic control are cooperating to regulate gene expression in a complementary manner. Moreover, we find that the position of the gene relative to the Ori is a feature of very high predictive value for gene expression, indicating that the Ori-Ter symmetry axis coordinates the action of distinct genetic control mechanisms.

Fig. 1. Summary of our investigation and overview of the workflow.

On the structural level (obtained from RegulonDB,), the spatial embedding of the TRN within the circular chromosome is evaluated via the EDURA (Edge Distribution Under Rotation of an Axis) method. The functional level is contributed by the COLOMBOS database and analyzed jointly with the structural information using the concepts of digital and analog control strengths, as well as decision trees.

Fig. 2. Illustration of the edge categories used in the subsequent analysis.

The light blue circle represents the circular chromosome, while the dots represent genes (red: right chromosomal arm; blue: left chromosomal arm). Directed edges indicate interactions between genes. The dashed blue line denotes the axes used for the assignment of edge categories (with the longer end representing Ori).

Fig. 3. Analysis of edge categories in the E. coli TRN.

a Representation of the chromosomally embedded E. coli TRN. As in Fig. 1 the large light blue circle represents the circular chromosome and the blue line indicates the Ori–Ter axis. Black dashes on the chromosome indicate genes. b Edge categories for the chromosomally embedded E. coli TRN from a as a function of the axis position. The true Ori and Ter positions are indicated as a reference.

Fig. 4. Edge category asymmetry analysis for the real E. coli TRN.

Asymmetries as a function of the assumed axis position (upper panel). Correlation coefficient of $A_{\pm }^r$ and $A_{\pm }^l$ as a function of the assumed axis position (lower panel).

Fig. 5. Digital vs. analog control for gene-level RNA-Seq data.

Data for 104 effective networks resulting from contrasts of RNA-Seq experiments have been analyzed. Central panel: Scatter plot of the digital vs. analog control strengths. Top panel: Histogram of the distribution of analog control strength. Right panel: Histogram of the distribution of digital control strength.

Fig. 6. Digital vs. analog control for gene-level microarray data.

Data for 3969 effective networks resulting from contrasts of microarray experiments have been analyzed. Central panel: Scatter plot of the digital vs. analog control strengths. Top panel: Histogram of the distribution of analog control strength. Right panel: Histogram of the distribution of digital control strength.

Fig. 7. Feature importance correlations for RNA-Seq data.

(Left) Spearman correlation coefficient between analogCTC and each of the features for networks derived from the rnaseq experiments. (Right) The same for the digitalCTC. The color code above each bar is: blue—analog control feature, red—digital control feature, green—dual feature (related to the Ori–Ter axis).

Fig. 8. Feature importance correlations microarray data.

(Left) Spearman correlation coefficient between analogCTC and each of the features for networks derived from the microarray experiments. (Right) The same for the digitalCTC. The color code above each bar is: blue—analog control feature, red—digital control feature, green—dual feature (related to the Ori–Ter axis).