Transcriptional profiling of coaggregation interactions between Streptococcus gordonii and Veillonella parvula by Dual RNA-Seq

Abstract

Many oral bacteria form macroscopic clumps known as coaggregates when mixed with a different species. It is thought that these cell-cell interactions are critical for the formation of mixed-species biofilms such as dental plaque. Here, we assessed the impact of coaggregation between two key initial colonizers of dental plaque, Streptococcus gordonii and Veillonella parvula, on gene expression in each partner. These species were shown to coaggregate in buffer or human saliva. To monitor gene regulation, coaggregates were formed in human saliva and, after 30 minutes, whole-transcriptomes were extracted for sequencing and Dual RNA-Seq analysis. In total, 272 genes were regulated in V. parvula, including 39 genes in oxidoreductase processes. In S. gordonii, there was a high degree of inter-sample variation. Nevertheless, 69 genes were identified as potentially regulated by coaggregation, including two phosphotransferase system transporters and several other genes involved in carbohydrate metabolism. Overall, these data indicate that responses of V. parvula to coaggregation with S. gordonii are dominated by oxidative stress-related processes, whereas S. gordonii responses are more focussed on carbohydrate metabolism. We hypothesize that these responses may reflect changes in the local microenvironment in biofilms when S. gordonii or V. parvula immigrate into the system.

Figure 1

Coaggregation between S. gordonii DL1 and V. parvula PK1910. Pico-green-stained S. gordonii cells (green) and propidium iodide-stained V. parvula PK1910 (red) were suspended in human saliva and mixed vigorously to induce coaggregation. Samples were visualized by confocal laser scanning microscopy. A large coaggregated mass is clearly visible, containing S. gordonii and V. parvula cells interspersed throughout the structure. Bar = 30 µm.

Figure 2

Analysis of coaggregates by TEM. Cells of S. gordonii and V. parvula were washed, suspended in human saliva and concentrated cultures were mixed vigorously to induce coaggregation. Samples were embedded in resin and visualized by TEM. At relatively low power (a), large areas containing densely packed S. gordonii and V. parvula cells were visible. S. gordonii (Sg) and V. parvula (Vp) could be more easily distinguished at higher magnification (b). Nearby cells appeared to be connected by extracellular material or fibrils (arrows).

Figure 3

Volcano plots of DESeq results showing gene expression changes. Differentially expressed genes with P_adj value and a log fold changes are shown as green and red dots. The logarithms of the foldchanges of individual genes (x-axis) are plotted against the negative logarithm of their P-value to base 10 (y-axis). Positive log₂ (fold change) values represent up-regulation and negative values represent down-regulation.

Figure 4

Validation of RNA-Seq data by RT-qPCR. The expression of selected genes in coaggregates versus monocultures was assessed by RT-qPCR, normalized to 16S rRNA expression. Graphs show mean and SD from three independent experiments. In general, there were strong correlations between expression levels measured by RNA-Seq (dark bars) versus RT-qPCR (grey bars).

Figure 5

Analysis of interacting gene networks in the ‘oxidoreductase’ category of the V. parvula differentially expressed gene dataset. Where possible, genes were assigned names based on the V. parvula DSM2008 genome nomenclature by homology searching. The gene list was uploaded to the STRING database, which predicts potential interactions based on factors such as gene environment, protein function or text-mining. The strength of evidence for interactions is indicated by the thickness of edges, with thicker lines indicating stronger interactions. Color coding is based on k-means clustering. The largest network is centered upon Vpar_1637, a predicted glutamate synthase.

Figure 6

Interacting gene networks in the S. gordonii differentially expressed gene dataset, as predicted by STRING DB. Interactions are based on a number of different lines of evidence, with stronger evidence indicated by thicker edges. The coloring of nodes is based on k-means clustering. The network of nodes shown in green is primarily related to carbohydrate metabolism.