Transcriptomic buffering of cryptic genetic variation contributes to meningococcal virulence

Abstract

Background: Commensal bacteria like Neisseria meningitidis sometimes cause serious disease. However, genomic comparison of hyperinvasive and apathogenic lineages did not reveal unambiguous hints towards indispensable virulence factors. Here, in a systems biological approach we compared gene expression of the invasive strain MC58 and the carriage strain α522 under different ex vivo conditions mimicking commensal and virulence compartments to assess the strain-specific impact of gene regulation on meningococcal virulence.

Results: Despite indistinguishable ex vivo phenotypes, both strains differed in the expression of over 500 genes under infection mimicking conditions. These differences comprised in particular metabolic and information processing genes as well as genes known to be involved in host-damage such as the nitrite reductase and numerous LOS biosynthesis genes. A model based analysis of the transcriptomic differences in human blood suggested ensuing metabolic flux differences in energy, glutamine and cysteine metabolic pathways along with differences in the activation of the stringent response in both strains. In support of the computational findings, experimental analyses revealed differences in cysteine and glutamine auxotrophy in both strains as well as a strain and condition dependent essentiality of the (p)ppGpp synthetase gene relA and of a short non-coding AT-rich repeat element in its promoter region.

Conclusions: Our data suggest that meningococcal virulence is linked to transcriptional buffering of cryptic genetic variation in metabolic genes including global stress responses. They further highlight the role of regulatory elements for bacterial virulence and the limitations of model strain approaches when studying such genetically diverse species as N. meningitidis.

Keywords: Cryptic genetic variation; MITE; Metabolism; Neisseria meningitidis; Regulatory evolution; RelA; Stringent response; Systems biology; Virulence.

Fig. 1

Genes significantly differently expressed and/or regulated in cross-condition and/or cross-strain comparisons. a Venn diagram comparing sets of genes in strain MC58 differently expressed between conditions as indicated. The total number of genes compared was 1987. b Venn diagrams comparing sets of genes differently expressed between strains as indicated with each diagram. The total number of genes compared in each panel was 1450. c Heatmap depicting cross-condition and cross-strain gene expression differences and hierarchical clustering of significantly differently expressed genes. Average linkage clustering based on the Spearman rank correlation of all 828 genes significantly differently expressed and/or regulated in at least one cross-condition and/or cross-strain comparison (FDR < 0.05). Grey lines correspond to genes that were absent in the α522 genome sequence and therefore excluded from the cross-strain comparisons

Fig. 2

Gene set enrichment analysis of differently expressed genes. a Non-directional comparison of significantly differently expressed gene sets according to the COG functional classification scheme for cross-condition and cross-strain comparisons, respectively, indicating significantly overrepresented functional categories among the significantly differently expressed genes. b Directional comparison of significantly differently expressed gene sets according to the COG functional classification scheme for cross-condition and cross-strain comparisons, indicating significant gene expression asymmetries. In both panels, the heat map depicts significantly enriched COG functional categories in red coloring. The kind of comparison (cross-condition for strain MC58 and cross-strain for each condition) is indicated for each column of the heat maps, and the corresponding FDRs are color coded and given in the respective inserts. The associated tree is based on average linkage clustering of the functional categories using the Spearman rank correlation coefficient

Fig. 3

Analysis of gene expression data based on protein-protein interaction networks. a Integrative network analysis of differently expressed genes between strain MC58 and α522 in human whole blood based on the STRING protein-protein interaction network for strain MC58 (FDR < 10⁻⁹). A subnetwork that consists predominantly (30/35) of genes that were expressed at higher levels in MC58 than in α522 and that code for metabolic genes and in particular for genes involved in energy production and conversion (85%) is shaded in orange. The remaining part of the network comprising 128 protein-coding genes is shaded in light blue. b Integrative network analysis of gene regulation differences between both strains upon transition from saliva to blood (FDR < 10⁻⁷). The two modules consisting predominantly of genes either upregulated in α522 (15/25) or MC58 (47/59) upon transition from saliva to blood are shaded in green and orange, respectively. Only 21% of the genes in the left subnetwork consisting mainly of genes that were upregulated in α522 code for proteins involved in (energy) metabolism compared to over 66% of the genes that were upregulated in MC58 upon transition from saliva to blood. For each gene, the respective expression differences between conditions and strains, respectively, are color coded and indicated in each panel. White boxes indicate genes that were not differently expressed but are part of a subnetwork as identified by the integrative network analysis. Pie charts next to the sub-networks in each panel show the distribution of proteins in the respective subnetwork over the different COG functional classes

Fig. 4

Analysis of gene expression data based on a metabolic model for strain MC58. a Comparison of elementary mode activities in MC58 and α522. The histogram depicts differences in the elementary mode activities (ordinate) for each of the elementary metabolic modes (abscissa) as defined in the Additional file 3: S2 for strain MC58 (red) and α522 (blue) based on gene expression data in human blood. b Inferred differences in metabolic fluxes between strain MC58 and strain α522 in blood based on a metabolic model for strain MC58. Internal metabolites which are considered to have balanced concentrations are given by dark green spheres, external metabolites which are allowed to accumulate or to be consumed by green cones, and reactions together with their corresponding numbers as light green boxes. The reactions for all reaction numbers are given in the Additional file 3: S2. Arrows connect reaction with metabolites. Red coloring indicates higher fluxes in MC58 compared to strain α522, whereas blue colouring indicates that the flux is slightly enhanced in α522. Asterisks along with pink colouring indicate that the reaction has an opposite direction in both strains

Fig. 5

Growth phenotypes of N. meningitidis MC58 and α522 wild-type and mutant strains. a In vitro growth phenotypes. Growth as quantified by the optical density (OD_600nm) is given on the ordinate and the time in hours on the abscissa. b Ex vivo growth phenotypes. The number of colony forming units for each time point (N(t)) relative to the initial number (N(0)) is given on the ordinate and the time in minutes on the abscissa. For each strain and condition the respective growth curves are coded as indicated in the insert in each panel, and the genotypes of the respective strains compared are shown along with the corresponding growth curves. In each experiment rich medium (PPM+) was used as growth control. The arrow at the top of panel b indicates the time when total RNA was extracted for microarray analysis

Fig. 6

GC content variation in potential promoter regions based on the MC58 genome sequence. a Scatter plot of the GC content variation averaged over a 5-bp sliding window within 100 bp upstream regions for genes highly expressed in MC58 (red and yellow lines) or α522 (light and dark blue lines) in human blood. The black line gives the GC content of the respective upstream regions for genes not differently expressed. Regulatory regions are indicated at the top of the panel based on the average length of 5’-untranslated regions in N. gonorrhoeae [75]. The insert gives the number of genes in each gene set. b Box-and-whiskers plot depicting differences in the mean GC content of the putative discriminator (left) and Hfq-binding regions (right) between genes highly expressed in MC58 (red) or α522 (blue) in human blood as depicted in panel (a). The line within each box gives the median and the upper and lower margins the upper and the lower quartile, respectively. The whiskers denote the highest and the lowest values, respectively, and the open circles outliers. *: p < 0.05, **: p < 0.01 (Wilcoxon test)

Fig. 7

Genomic distribution of ATRs and the relA locus in N. meningitidis. a The intergenic region between grxB and relA. The integration site of a copy of an ATR repeat element upstream of relA (ATR_relA) in strain α522 is indicated with respect to the MC58 locus. The transcriptional start sites as determined by 5’-RACE in both strains are indicated along with the deduced −35 and −10 boxes and the computationally predicted promoter regions using PPP [117]. DR: direct repeat. b Alignment of both the MC58 (upper lane) and α522 (lower lane) genomes as visualized with the Artemis comparison tool based on a BLASTN comparison. The linearized MC58 and α522 genomes are shown in the upper and lower panel as gray bars, and regions syntenic in both genomes are connected via red and inverted regions via blue lines, respectively. The location of ATRs is indicated by small arrows in each genome, and the relA region is highlighted in yellow

Fig. 8

Graphical summary and hypothesis relating major findings of this work and published data. The figure is not intended to give a comprehensive overview of the entire metabolism and stress responses in N. meningitidis but to illustrate pathways that link metabolism, protein sequence and gene expression differences of selected (virulence-associated) genes and the pathogenesis of IMD as described in the main text. Accordingly, genes and pathways that were highly expressed in MC58 in blood and/or that are strongly upregulated between saliva and blood in MC58 are depicted in red, and genes and pathways that are highly expressed in α522 or that are strongly upregulated between saliva and blood in α522 are depicted in green. Asterisks next to enzyme or protein names indicate that the corresponding proteins have a less than average sequence similarity (BSRP < 0.958) or are entirely missing in strain α522. Arrows with plus signs indicate (predominantly) activating regulatory interactions, and arrows with minus signs (predominantly) inhibitory regulatory interactions. For further details and abbreviations see main text. The literature cited in the figure is indicated by bracketed numerals next to the respective arrows: (1) Newcombe et al. (2005) [33], (2) Delany et al. (2006) [70], (3) Fantappie et al. (2009) [74], (4) Monaco et al. (2006) [87], (5) Huis in’t Veld et al. (2011) [72], (6) Tala et al. (2011) [86], (7) Takahashi et al. (2015) [124], (8) Gunesekere et al. [71], (9) Criss and Seifert (2012) [38], (10) Seib et al. (2006) [84], (11) Schmitt et al. (2009) [83], (12) Stevanin et al. (2007) [52], (13) Kobsar et al. (2011) [51], (14) Coureuil et al. (2014) [125], (15) Virji (2009) [49], (16) Hellerud et al. (2015) [48]