Direct comparison of spatial transcriptional heterogeneity across diverse Bacillus subtilis biofilm communities

Abstract

Bacillus subtilis can form various types of spatially organised communities on surfaces, such as colonies, pellicles and submerged biofilms. These communities share similarities and differences, and phenotypic heterogeneity has been reported for each type of community. Here, we studied spatial transcriptional heterogeneity across the three types of surface-associated communities. Using RNA-seq analysis of different regions or populations for each community type, we identified genes that are specifically expressed within each selected population. We constructed fluorescent transcriptional fusions for 17 of these genes, and observed their expression in submerged biofilms using time-lapse confocal laser scanning microscopy (CLSM). We found mosaic expression patterns for some genes; in particular, we observed spatially segregated cells displaying opposite regulation of carbon metabolism genes (gapA and gapB), indicative of distinct glycolytic or gluconeogenic regimes coexisting in the same biofilm region. Overall, our study provides a direct comparison of spatial transcriptional heterogeneity, at different scales, for the three main models of B. subtilis surface-associated communities.

Fig. 1. An overview of spatial transcriptome remodelling between the different selected populations of B. subtilis.

a Schematic illustration of the differently localised spatial populations selected. From the planktonic culture, the exponential (EX) and stationary (ST) phases were selected. From the static liquid model, the pellicle (PL) formed at the liquid-air interface, the submerged biofilm (SB) formed at the solid‒liquid interface, and the free detached cells (DC) between these two populations were collected separately. From the swarming model, four localised populations were collected separately: the mother colony (MC), the inoculation site from which the swarm developed as a mature macrocolony; the base (BS) of the dendrites as an earlier biofilm form; the dendrites (DT), a monolayer of cells ready to later form the biofilm; and the tips (TP) formed by motile and rapidly dividing cells. For each population, distinguished by a specific distinct colour, three independent samples were taken as biological replicates. (b) The pairwise distance (Spearman) between RNA-seq profiles is summarised by a hierarchical clustering tree highlighting the divergence of the mother colony (MC) and the stationary phase (ST) between/along other selected populations and the proximity of adjacent spatial populations of either the static liquid model (SB, DC and PL) or the swarmer populations (BS, DT and TP), which share a closer genetic expression profile with the exponential phase (EX). (c) Projection of the 27 transcriptomes on the 2 main axes of the principal component analysis (PCA) plot. These axes account for up to 65% of the total variability of the data. (d) Global heatmap representation of the 4028 genes present in NDmed across the spatially selected surface-associated populations. The colour code reflects the comparison to the mean calculated for each gene (log2 ratio), taking as reference the mean of all conditions, except the planktonic ones (EX and ST). The hierarchical clustering tree shown on the left side of the heatmap (average link) was cut at an average Pearson correlation of 0.7 (dashed red line) to define the expression clusters shown as rectangles on the right side of the heatmap. Clusters were named by decreasing size (from G1 to G321), and only those containing more than 100 genes are highlighted (number of genes in black, cluster name in red).

Fig. 2. Spatial transcriptomic remodelling with in situ imaging reveals heterogeneous differential expression of central carbon metabolism.

a Heatmap representation of the relative variation in expression levels between populations. The colour code reflects the comparison to the mean calculated for each gene across all populations, except the planktonic populations (EX and ST) (log2 ratio). Genes were selected from SubtiWiki categories specific to glycolysis or gluconeogenesis. The purple box highlights central genes specific for glycolysis (gapA, cggR) and gluconeogenesis (gapB, pckA). (b) Spatial confocal imaging for the different selected populations from the swarming (MC, BS, DT, TP) and static liquid (SB, DC, PL) models after 24 h at 30 °C, using strain GM3900 reporting the transcription of cggR-gapA by mCherry (in red) and of gapB by Gfp (in green), with the same protocol as for the transcriptome analysis, except for the use of a 96-well microplate instead of a 12-well microplate for the static liquid model. Three replicate observations were performed independently for each model. (c) Sections from real-time confocal imaging (x 50 μm, y 50 μm, z 80 μm) for 48 h, imaged every one and a half h, using strain GM3900 reporting the transcription of cggR-gapA) by mCherry (in red) and of gapB by Gfp (in green) (Supplementary Movie 4). Kymographs representing the intensity of the expression of the transcriptional reporter fusions to the cggR and gapB promoters along a spatiotemporal scale by colour coding. Three biological replicates were performed (Supplementary Fig. 13). (d) A multidimensional kymograph representing the intensity of dead cells, obtained from the ratio of dead/living cells, as a function of their spatial localisation and time (Supplementary Movie 5). The kymograph is representative of at least three replicates (Supplementary Fig. 12).

Fig. 3. Space-time kymographs of fluorescent reporters for the transcription of 15 genes in the submerged biofilm (SB).

a Kymograph of the fluorescent signal in arbitrary units (a.u.) of the constitutively expressed Gfpin the NDmed-GFP strain. b Kymographs for the 15 transcriptionally reported genes/operons, representing the intensity of their expression. The white dotted line in each kymograph represents the time (24 h) corresponding to the RNA-seq analysis. The white spots in the graphs indicate intensity values below a threshold limit, representing no significant signal for that specific tile. Each kymograph presented is one of three independent biological replicates shown in Supplementary Fig. 13.