Pangenome guided pharmacophore modelling of enterohemorrhagic Escherichia coli sdiA

Introduction

One of the more prominent strain of Escherichia coli is the enterohemorrhagic (EHEC) pathotype associated with global outbreaks of bloody diarrhea and hemolytic uremic syndrome¹. EHEC is one of the classical of examples of one health disease as the interface of animal health spills into human health. Within the cattle reservoir, sdiA gene is required by E. coli to survive the acidic rumen environment. SdiA is used by E. coli to sense acyl homoserine in a quorum sensing system². However, it is considered as an orphan as the cognate acyl homoserine synthase is absent, and hence sdiA is considered an environmental sensor to sense the nearby microbial community. SdiA is stabilized by acyl homoserine lactone and acts as transcription factor glutamate decarboxylase needed for survival in the acidic environment. Hence blocking the ability of EHEC to survive the acidic ruminal environment is a proposed mechanism to control shedding in the cattle reservoir.

Whole genome sequencing of bacterial pathogens, particularly EHEC, is quickly transforming the workflows of epidemiological investigations. However, most bioinformatic pipelines used in clinical investigation use data reduction of genomes and artificially reducing diversity due to comparison of a limited number of housekeeping genes³. While wgMLST attempts to increase the number of genes, the assignment of a single reference genome appears to be inadequate in light of the pangenome. Various studies have shown that a significant number of genes that are present to the entire universe of genes within a species is missed for variant calling if only a single reference gene is used⁴. In this study we applied multi-scale approach to generate genome wide clustering using pangenome phylogeny using genome wide gene presence absence variation (PAV). We used this genome wide clusters as guide in generating the pharmacophore of sdiA as potential design strategy to control shedding by reducing the acid survival in the rumen.

We applied the concept of the pangenome, which represents the entirety of the genes that are present within a species, to a pathotype level. The EHEC pangenome represents the combination of genes seen in the EHEC pathotype. While the pangenome of E. coli was published in 2008 contained 8 genomes, we generated an updated EHEC pangenome with 153 genomes. The pangenome enables clustering of isolates using gene presence and absence. We used the genome wide comparison to generate clusters and genomic clade specific pharmacophore model of sdiA. This strategy enables to capture the pangenome wide variation of sdiA and ensures all conserved variants are targeted by the drug discovery pipeline enabling a pangenome to pharmacophore approach.

Methods

Results and discussion

Three main genomic clades were generated using the pangenome PAV of genes (Figure 1A). Clade I includes the genomes isolated from the earliest reported outbreak in 1982 as well as the more recent cases. Clade II includes genomes from European (predominantly Czech and German) enterohemorrhagic E. coli isolates (EHEC) belonging to the O26 serotype, while Clade III includes cluster of cases associated with acute renal failure in children. Clade III is associated with hybrid pathotype of serogroup O80 has aside from Shiga toxin, an extra-intestinal virulence plasmid (pS88), is currently emerging in France. Specific gene clusters are found in cluster 1 and noticeably absent in cluster III, particularly prophages and associated genes (Underlying data: Table 2, Figure 1B). This indicates that specific genome wide gene presence and absence can be used for functional cluster instructive of the underlying epidemiological dynamics. The acquisition of genomic islands unique to individual isolates are well defined in the pangenome gene presence absence matrix (Figure 1B). The core genome is 2145 (Table 1) and total gene count within the EHEC pangenome is 17152, which is not that far off the total E. coli pangenome 22,000¹⁹. This enormous difference between the core gene and total gene highlights the variation between the different isolates, which can be strain specific and individual isolate specific as indicated by the pangenome data.

Figure 1A. EHEC E. coli pangenome (152 genomes) accessory genome phylogeny showing three major clades.

Figure 1B. EHEC E coli pangenome (152 genomes) showing genomic diversity with the presence absence variation (PAV) matrix.

All of the examined genomes contain sdiA gene, which is contained within the 2145 core genome. Using sdiA of Escherichia coli O157:H7 str. Sakai for variant calling, we identified the nonsynonymous mutations that are widely distributed across the EHEC isolates (Figure 3). 45.0% (49/109) of the nonsynonymous mutation is due to conversion of arginine to lysine at position 189 of sdiA (Figure 2A). This amino acid is located with the α-6 domain, adjacent to the amino acid clusters associated with sdiA dimerization. Previous protein modelling determined role of the guanidinium group of arginine which enables interactions in three different directions enabling a more complex electrostatic interaction versus lysine as well as the higher pKa value in arginine that can yield a more stable ionic interaction compared to lysine²⁰. The arginine to lysine mutations are found mostly in clade III isolates. The second ranked nonsynonymous mutation asparagine to serine at amino acid position 101 with 17.4% (19/109) which is located adjacent to η-4 phenylalanine which is associated with the ligand and distributed among the different clades (Figure 2B). β-5 domain alanine to threonine change at amino acid position 140 is the third ranked nonsynonymous mutation with 9.2 % (10/109) (Figure 2C). The fourth ranked nonsynonymous mutation is at amino acid position 138 with the conversion of cysteine to serine (Figure 2D) also within the β-5 domain. Both nonsynonymous variants of amino acid position 140 and 138 are found within clade II and III. None of the highly ranked nonsynonymous mutations impact the ligand interaction, indicating the conservation of the sdiA motif across the population in geographic and temporal distribution, which suggests the possibility of targeting sdiA for quorum sensing inhibition.

Figure 2. sdiA nonsynonymous variants protein modelling.

Figure 3. EHEC E coli sdiA allelic variants showing synonymous and nonsynonymous mutations.

Conclusion

While EHEC pangenome is remarkably diverse, the allelic variants of sdiA, particularly nonsynonymous mutants, indicate the conservation of quorum sensing domain, indicating that targeting this structure can be effective across the different lineages of EHEC pathotype.

Data availability

All underlying and extended data available from Open Science Framework: Supplemental Data for Pangenome guided pharmacophore modelling of enterohemorrhagic Escherichia coli sdiA, https://doi.org/10.17605/OSF.IO/BNZ85⁷

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).