Improved in silico and in vitro methods for Escherichia coli LPS outer core typing

Abstract

The Gram-negative bacterial envelope comprises the outer membrane, periplasmic space with the peptidoglycan layer and inner (cytoplasmic) membrane. A lipopolysaccharide (LPS) layer linked to the outer membrane is essential for survival in most species of bacteria, primarily by providing structural stability and regulating selective chemical permeability. These functions make the LPS layer a key pathogenicity determinant, protecting bacteria from host defences. At the same time, LPS serves as a common receptor for multiple bacteriophage orders, making it a crucial point of bacterial vulnerability. LPS outer core typing is traditionally performed using immunoblotting and PCR. With the increasing availability of sequenced genomes, PCR has emerged as the primary method for typing. This study presents a set of nine oligonucleotides designed for typing the five LPS outer core structures of Escherichia coli, R1, R2, R3, R4 and K-12, both in vitro and in silico. Our method was able to successfully type 99% of strains within a comprehensive dataset of 4,549 complete genomes. Using these new methods, we identify previously unreported hybrid LPS structures involving R2 and K-12 outer core types and establish associations between LPS outer core types and E. coli phylogenetic groups, pathogenicity, O-antigen polysaccharides, sequence types and capsule types. We also introduce LPSTyper, a Python-based command-line tool that enables rapid and precise LPS outer core typing in E. coli genome sequences. Together, the expanded dataset of LPS outer core types and enhanced laboratory and computational tools for their detection provide new insights into LPS structural evolution and its role in E. coli ecology and pathogenicity. We anticipate that these tools will improve the efficiency of future research and diagnostics of this important bacterial envelope structure.

Impact Statement.

Bacterial lipopolysaccharide (LPS) layer types are important for determining both pathogenicity potential and phage susceptibility. This study presents a robust and scalable framework for precise typing of Escherichia coli LPS outer core types, enabled by the development of nine novel oligos targeting the waa locus. Using the LPSTyper tool for in silico analysis and a simple PCR-based method for in vitro LPS outer core type confirmation, this approach allows rapid and accurate typing across thousands of genomes, supporting large-scale monitoring of LPS diversity in both commensal and pathogenic E. coli populations. Our findings reveal strong associations between LPS outer core types, O-antigens, capsules, phylogenetic groups, sequence types and pathotypes, highlighting the critical role of LPS structure in virulence, immune evasion and environmental adaptation. The discovery of hybrid and novel waa locus arrangements further underscores the evolutionary plasticity of LPS biosynthesis. Overall, this work provides essential tools and insights for genomic surveillance and deepens our understanding of E. coli surface structure variability central to host–pathogen interactions and antimicrobial resistance.

Data Summary

The LPSTyper source code is freely available on GitHub under the Apache 2.0 licence: https://github.com/ellinium/LPSTyper. Sequencing data for all genomes introduced in this study are accessible via the NCBI database under BioProject accession number PRJNA1243169. Individual genome accession numbers are listed in Tables 1 and S1 (available in the online Supplementary Material). All supporting data are provided in Supplementary Material Files 1, 2, and 3.

Table 1. Reference E. coli strains used in in vitro PCR.

E. coli strain	GenBank accession no.	In vitro LPS outer core type	Reference
NCTC122	NZ_LT906474.1	R1	–
DSMZ13127	JBMNYY000000000.1	R1	This work
ROAR340	JBMNZB000000000.1	R1	This work
C4741	JBMNZA000000000.1	R1	This work
ECOR11	LYAZ00000000.1	R2	[60]
ROAR375	JBMNYX000000000.1	R2	This work
ECOR09	LYBB00000000.1	R3	[60]
M1402	JBMNYZ000000000.1	R3	This work
ECOR04	LYBG00000000.1	R4	[60]
ECOR58	LYCY00000000.1	R4	[60]
MG1655	NC_000913.3	K-12	[18]
ECOR03	LYBH00000000.1	K-12	[60]

Introduction

The Gram-negative bacterial envelope comprises the outer membrane, periplasmic space with a peptidoglycan layer, and the inner membrane. The envelope plays a crucial role in maintaining structural integrity, acting as a selective barrier and contributing to antibiotic resistance and host interactions. Lipopolysaccharides (LPSs) are key structural components of the outer membrane in Gram-negative bacteria. They play critical roles in structural integrity and protection from external stresses, making them vital for bacterial survival [1]. LPS contributes to antibiotic and detergent resistance by forming a permeability barrier that reduces the effectiveness of these compounds [2]. LPS is also an important target for immune recognition and response mediated by toll-like receptor 4, myeloid differentiation factor (MD-2), CD14, LPS-binding proteins, antimicrobial peptides and host defence lectins [1]. In contrast to the protective effects of LPS on physical, chemical and immune system threats, components of the LPS layer also serve as receptors for the receptor-binding proteins of diverse bacteriophage orders, representing a critical point of vulnerability [3,4].

LPS molecules are comprised of three main parts: the lipid A region, a core polysaccharide (subdivided into inner and outer core) and the O-antigen [3]. Lipid A is a glycolipid consisting of acylated and phosphorylated β-1,6-linked glucosamine disaccharides that anchor LPS into the membrane and are recognized by the human immune system as a potent endotoxin [5]. Glucosamines of lipid A are linked to the LPS inner core that is conserved across many bacterial species and consists of a non-repeating oligosaccharide that includes 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo) and l-glycero-d-manno-heptose (Hep) residues [6]. The highly variable O-antigen, composed of one or more repeating polysaccharide units, contributes to antigenic diversity among bacterial strains, enabling Escherichia coli to evade the host immune system [7]. In smooth strains, the O-antigen is attached to the outer core, creating an extended glycan barrier around the bacterium. In rough strains that lack the O-antigen, the LPS outer core represents the terminal structure exposed to the external environment.

The outer core LPS is highly variable across different Gram-negative bacterial strains [8]. This structure is primarily composed of variable hexoses and N-acetylhexosamine. Variations in the structure of the outer core can shield or expose the associated O-antigen or lipid A molecules, influencing immune evasion or activation [7]. Additionally, LPS outer core structural variability can influence bacterial susceptibility to bacteriophages. For example, certain Siphoviridae, Myoviridae, Podoviridae and Microviridae phages use components of the outer core as a receptor for attachment and infection [4]. As these structures are essential for phage infection, the LPS outer core faces conflicting selective pressures, balancing host integrity with avoiding bacteriophage recognition [4,9].

In E. coli, five distinct types of outer core LPS are recognized: R1, R2, R3, R4 and K-12 [10]. The genes encoding these types are located in the waaA genetic locus and are organized in three operons: (1) the waaD operon, which is essential for the biosynthesis and transfer of l,d-heptose (synonyms: rfaD and gmhD); (2) the central waaQ operon containing genes necessary for the biosynthesis and modifications of the outer core; and (3) the waaA operon encoding Kdo transferase and additional non-LPS genes [11].

The landscape of E. coli outer core type distribution was previously identified in several hundred E. coli strains using LPS extraction followed by SDS-PAGE and immunoblotting with mouse monoclonal antibodies [12,13] or PCR [14]. However, no in silico tools are currently available for E. coli LPS outer core typing. Accurate identification of E. coli LPS outer core types can provide valuable insights into immune interactions and bacterial pathogenicity, facilitate the development of targeted vaccines [15] and inform strategies to combat phage and antibiotic resistance.

This work describes new in silico and PCR-based in vitro methods for LPS outer core typing of any known E. coli strain. Using the in silico method, we determined the distribution of LPS outer core types across 4,549 commensal and pathogenic E. coli strains. We validated the PCR-based method on a subset of analysed strains. Along with the identification of R1, R2, R3, R4 and K-12 types, we also identified previously uncharacterized hybrid R2 and K-12 LPS outer core gene sets. With this new dataset, we explored the associations between LPS outer core types and E. coli phylogenetic groups, serotypes and capsule types, uncovering distinct correlations with specific pathogenic groups, ecological niches and capsular and O-antigen polysaccharides. Collectively, this study significantly enhances our understanding of the association of E. coli LPS outer core types, virulence and adaptability, while also providing user-friendly tools for further investigation of these associations.

Methods

Bacterial strains

E. coli reference (ECOR) collection from the Yehl Lab (Miami University) was verified with multiple-locus variable-number tandem-repeat analysis using multiplex PCR [16]. E. coli ROAR340, DSM13127, ROAR375, C4741 and M1402 were kindly provided by Olivier Tenaillon (Institut Cochin, Université Paris Cité). E. coli NCTC122 was from the National Collection of Type Cultures-UK Health Security Agency.

Bacterial genomic data

A total of 4,549 complete E. coli genomes were retrieved from the NCBI Reference Sequence database in March 2024. The dataset included commensal and pathogenic strains. The complete list of genomes is provided in Table S1. Pathogenic strains were subtyped based on the keywords related to pathogenicity in the metadata provided with the genomes from the columns ‘Assembly BioSample Description Title’, ‘Assembly Accession’, ‘Assembly BioSample Description Comment’, ‘Assembly BioProject Lineage Title’, ‘Assembly BioSample Attribute Name’ and ‘Assembly BioSample Attribute Value’. All genomic data were downloaded in FASTA nt format and re-annotated with Prokka v1.13 [17] using the following parameter values: ‘genus’ - Escherichia, ‘species’ - coli, ‘kingdom’ - Bacteria, ‘gcode’ - 11, ‘usegenus’ and ‘proteins’ referring to an MG1655 FASTA aa file (GenBank accession ID U00096.3 [18]) used for the EcoCyc database [19].

LPS outer core identification

Genomic data were parsed using custom Python scripts to obtain waa operon gene sequences from waaD (rfaD) to coaD, including nt and aa sequences. The sequences and their order were compared with previously typed strains and LPS operon gene order described in the literature [11,14, 20] and aligned using the online Clustal Omega service v2.1 [21]. Primers for the LPS outer core were identified semi-manually using Geneious Prime® 2024.0.7. For the genes in question, annotations were compared with the original GenBank data, searched using the NCBI Nucleotide and Protein Blast online services, UniProt blast [22], and aligned with the known genes.

In vitro LPS outer core typing

PCR was performed on a BIO-RAD T100™ Thermal Cycler using OneTaq® Hot Start DNA Polymerase (NEB) according to the manufacturer’s protocol. Primers were used at a final concentration of 0.2 µM (Table 2). The template DNA was prepared as described previously [16], with one μl used per a 10 µl reaction. The PCR programme included an initial denaturation at 94 °C for 30 s, followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 49 °C for 30 s, extension at 68 °C for 90 s and a final extension at 68 °C for 5 min. The PCR products were separated on a 1% agarose gel containing thiazole orange (0.1 µl ml⁻¹) for 40 min at 110 V. The same PCR programme was used for PCR with all LPS primers combined in the same mixture (multiplex PCR).

Table 2. Oligonucleotide primers for LPS outer core typing.

LPS outer core type	Primer name	In vitro oligonucleotide primer (5′−3′)	Product size (bp)	Target LPS gene
R1	R1-F	GAACCGCATACATATCGCA	842	waaV
	R1R4-R	TTCTTGCTGGAATGATGTGG		waaW
R2	R2-F	TTGCGACCTGATAAATCCATG	610	waaK
	R2-R	TACGTTCATTTTACCGCAGAG		waaZ
R3	R3-F	TACATCCACGTTTCAACAGC	1,021	waaD
	R3-R	AATGTCTCATGGGGTATTGATGA		waaJ
R4	R4-F	TCCATCAGTCATTTCTGCG	380	waaX
	R1R4-R	TTCTTGCTGGAATGATGTGG		waaW
K-12	K12-F	AACTGTAAATCCCGCAGATGC	1,442	waaU
	K12-R	CGTAAAGCGAAAGATCGTATTGA		waaY

Determination of E. coli phylogenetic groups, serotypes, sequence types and capsules

E. coli phylogenetic groups were determined using a standalone version of the Clermont phylotyping tool ClermonTyping v23.06 [23]. Serotypes were identified with a command-line version of SerotypeFinder 2.0 [24], and capsules were typed with VirulenceFinder 2.0 [25]. The capsule was identified if both kpsE and kpsM genes were present. E. coli sequence type (ST) was determined using mlst [26,27].

Sequencing

Genomic DNA from E. coli ROAR340, DSM13127, ROAR375, C4741 and M1402 was extracted using Qiagen DNeasy Blood and Tissue Kit (Gram-negative bacterial protocol), sequenced at iThree Institute (University of Technology, Sydney) on Illumina MiSeq in 2×150 bp mode and further assembled and annotated as described previously [28]. The bacterial genomes were deposited in the NCBI BioProject PRJNA1243169. PCR products for R1, R2, R3, R4 and K-12 LPS core types were purified using Monarch® DNA clean-up kit (NEB) and sequenced using Plasmidsaurus service ‘Standard Purified Linear PCR’.

Statistical analysis

χ² and bias-corrected Cramér’s V [29] were used to measure the association between E. coli LPS outer core type and E. coli pathotypes, phylogroups, serotypes and STs.

Results

Surveying current methods for identifying LPS outer core types in E. coli, we found that existing approaches relied solely on wet lab methods, with a noticeable gap in computational tools. We then attempted to adapt previously published PCR oligonucleotides [14] for in silico identification of the LPS outer core genes. Using an E. coli dataset comprising 4,549 genomes, including commensal and pathogenic strains, we found that only 46.81% of the genomes could be LPS outer core typed using these oligonucleotides for in silico detection. Although PCR can tolerate single-point mismatches in primer binding sites in vitro, in silico typing requires identifying all mutations in the primer-binding region beforehand. To address this issue, we aimed to create primers for LPS outer core typing that could be used both in silico and in vitro.

Using aligned nt sequences of the genes within the waa locus of E. coli strains with known LPS outer core types, including DSMZ13127 (R1), ROAR340 (R1), ROAR375 (R2), M1402 (R3) [20], NCTC122 (R1), and the reference strain MG1655 (K-12), we designed nine PCR oligonucleotides (Table 2) to uniquely identify genes diagnostic for R1, R2, R3, R4 and K-12 LPS outer core types. We iteratively refined the oligos in silico on the genomes of 4,549 E. coli strains until 99% of the LPS outer core types were identified.

The final oligo sequences (Table 2) bind to a 842 bp region in R1 types between UDP-galactose-(Galactosyl) LPS α-1,2-galactosyltransferase (waaW) and β-1,3-glucosyltransferase (waaV), a 610 bp region between the gene involved in KdoIII attachment during LPS core biosynthesis (waaZ) and the LPS 1,2-N-acetylglucosamine transferase (waaK) for R2 and a 1,021 bp segment between the UDP-glucose: (galactosyl) LPS α-1,2-glucosyltransferase (waaJ) and ADP-l-glycero-d-mannoheptose 6-epimerase (waaD) genes for R3 (Fig. 1). For R4 outer core typing, the same oligo used for identifying the waaW gene in R1 types was paired with a unique sequence in β-1,4-galactosyltransferase (waaX), yielding a 380 bp product. For the K-12 type, a 1,442 bp region was identified between LPS core heptose (II) kinase waaY and the ADP-l-glycero-β-d-manno-heptose:LPS heptosyltransferase IV (waaU).

Fig. 1. Genomic organization of E. coli LPS operons and corresponding glycan structures across eight outer core types. The outer core types R1-R4 and K-12 represent distinct E. coli variants. K-12/R2 and R2/K-12 illustrate hybrids of E. coli K-12 and R2 types. Genes are aligned to facilitate the comparison of outer core architectures. Gene waaG overlaps with waaQ and waaP in all E. coli core types [61]. Salmonella enterica Typhimurium LPS genes are included for comparison with R2 and K-12 genes. A complete gene annotation is provided in Table S13. Monosaccharides are depicted using the Symbol Nomenclature for Glycans (SNFG) [62]: yellow circles represent galactose (Gal), blue circles represent glucose (Glc), blue squares represent N-acetylglucosamine (GlcNAc), green flat hexagons represent l-glycero-d-manno-heptose (LDmanHep). Glycosidic linkages are indicated by black lines, with numbers specifying the bond positions and anomeric configurations – α1→6(16A), α1→2(12A), α1→3(13A) and β1→4(14B).

In vitro determination of E. coli LPS outer core types using PCR

We evaluated the primers in vitro using E. coli strains NCTC122 (R1 type), ECOR11 (R2), ECOR09 (R3), ECOR04 (R4) and MG1655 (K-12) (Table 1). The PCR products were of the expected sizes (Fig. 2), and sequencing confirmed that the amplified products matched the correct target sequences. We repeated the test with a multiplex primer set combining all nine primers and had the same results. We next performed PCR on additional strains, including those previously typed in Michel et al. [20] (DSMZ13127, ROAR340, C4741 and ROAR375) and two ECOR strains (ECOR58 and ECOR03). In all cases, the results were consistent with those reported in the original publications (Table 1), showcasing the ease and utility of our new PCR primers for LPS outer core typing.

Fig. 2. PCR products amplified from five different E. coli LPS outer core types separated on a 1% agarose gel. The following strains were used for the typing: E. coli NCTC122 (R1), ECOR11 (R2), ECOR09 (R3), ECOR04 (R4) and MG1655 (K-12).

In silico determination of E. coli LPS outer core types

We next developed LPSTyper, a command-line Python tool designed to automatically identify LPS outer core types from nt sequences (https://github.com/ellinium/LPSTyper). The tool accepts a single input parameter: the full path to a directory containing one or more files in FASTA nt sequence format (e.g., FASTA and FNA). It scans these files using our in silico LPS outer core oligos to identify sequences matching with forward and reverse primers and then uses presence/absence patterns to determine the corresponding LPS core types – R1, R2, R3, R4 or K-12. LPSTyper supports recursive processing of subdirectories within the input directory. For files containing contigs, the LPS waa operon genes must either reside on the same contig, or the contig order must follow the sequential order of the genes in the waa operon in either direction. The tool does not work with FNA or FASTA files containing coding sequences and does not correct genome assembly errors.

ECOR collection verification

We used LPSTyper to identify the LPS outer core types in the ECOR collection, which returned results for all strains except ECOR46 (Table S2). We could not initially identify the ECOR46 LPS outer core type from its nt sequence using in silico primers due to the genome sequence not being closed as a single contig and the LPS genes being on different contigs and out of order. Additionally, we discovered two copies of O-antigen ligase waaL: one matching R1 waaL and another matching R4 waaL. Moreover, both R1 waaV and R4 waaX genes were upstream of corresponding waaL genes. We also found a transposase gene insO upstream of waaY, suggesting that LPS gene order was changed and both R1 and R4 genes appeared after a transposition event. This unusual genome configuration could be a reason for the different LPS typing results in Amor et al. [14] where they found R1 versus Gaborieau et al. [30] that found R4 (Table S2). We PCR-typed ECOR46 in vitro using the primers developed in this work (Table 2) and obtained an amplification product matching only the R4 core type, not R1.

Our in silico results for the remaining ECOR strains revealed discrepancies with Amor et al. [14] for strains ECOR07, ECOR43, ECOR44, ECOR46, ECOR47-ECOR50, ECOR58 and ECOR72 and Gaborieau et al. [30] for strains ECOR43 and ECOR46 (Table S2). For ECOR07, both our tool and Gaborieau et al. results showed K-12, while Amor et al. typed this strain as R1. ECOR43 had R4 type in Amor et al. and was non-typed in Gaborieau et al. With LPSTyper and additional manual verification of the LPS gene sequences, we identified it as R2. ECOR44, ECOR47 and ECOR72 all had R3 type, while ECOR48, ECOR49, ECOR50 and ECOR58 had R4 type in our data and Gaborieau et al. By contrast, Amor et al. identified all those strains as R1 type except ECOR58, which had R3 type (Table S2). Overall, these contrasting results may be attributed to different sources of the ECOR collection used in each study and the level of cross-contamination, as previously documented [16].

Hybrid LPS and non-typed structures

Next, we looked closer at 40 E. coli strains (0.9% of the genomic dataset) that did not return an LPS outer core type identification from LPSTyper. Manual genome analysis revealed 12 genomes that should have been identified as R1, R3 or R4 types but had single-point mutations in the gene locations where the in silico primers bound, thus preventing their identification. To address this, we included additional in silico primers (Table S14) in our command-line typing tool LPSTyper. We took an approach of including primers that identified >99% of samples we analysed while avoiding adding additional primers for every possible variation in the LPS genes targeted by PCR as they are highly conserved within each LPS outer core type. While PCR primers can tolerate some mismatches in vitro, in silico tools require stricter matching to ensure specificity. When required, new primers can be incorporated into the LPSTyper tool in the future to account for any newly discovered mutations in these target genes.

R2/K-12 hybrids

While manually identifying the LPS outer core type of E. coli strains that could not be typed using LPSTyper, we discovered several strains that possessed gene sets from both R2 and K-12 types (Table S15). Eight strains had genes identical to K-12 type but with waaU gene (ADP-l-glycero-β-d-manno-heptose:LPS heptosyltransferase IV) replaced by waaK (LPS 1,2-N-acetylglucosamine transferase) from R2 type with 90% nt identity. In these strains, the waaL gene was also 100% identical to waaL of K-12, and waaS was 93.25% identical to K-12 waaS (Fig. 1 and S1).

Two additional genomes, GCA_036870985.1 and GCA_012711215.2, were also identified with gene sequence arrangements identical to the K-12 core genes but with a putative transposase, yhhI, inserted between waaZ and waaY. Furthermore, nt identity analysis revealed that waaL shared 92.86% identity and waaK 89.94% identity with the corresponding waaL and waaK genes of R2 type, while waaS exhibited 93.16% identity to the waaS gene of K-12. Moreover, we identified one strain (GCA_013122565.1) with an R2 core gene arrangement but with waaK replaced by K-12 waaU and waaL and wabA genes matching R2 gene sequences (Fig. 1). These findings suggest that transposition or horizontal gene transfer events have occurred in E. coli strains containing R2 and K-12 outer core structures.

We also identified two strains with duplicated LPS genes. E. coli MSB1_3C (GCA_904863165.1) contained a gene sequence resembling the R2 structure but included two truncated copies of each of the waaK and waaQ genes. Similarly, E. coli O158:H23 (GCA_018986795.1) exhibited a K-12-like structure but with two waaS genes and an additional short peptide sequence between waaA and waaQ. These findings may represent either genome assembly artefacts or genuine gene rearrangement events. There was also one strain (GCA_013892175.1) following K-12 structure with an additional short 96-nt sequence between waaA and waaQ genes matching GCA_018986795.1 strain and encoding an uncharacterized protein (UniProt ID A0A485J7Q0).

LPS typing in other Escherichia species

Interestingly, we found two genomes (GCA_013898955.1 and GCA_013898975.1) that were initially classified as E. coli and possessed K-12 LPS gene sequence order. However, nt alignment showed lower identity to K-12: waaL (48.93%), waaY (90.27%), waaU (82.21%) and waaS (91.25%). We later found that those strains were re-classified as Escherichia marmotae [31], suggesting that LPS biosynthetic pathways share a close genetic relationship between E. coli K-12 and E. marmotae. Additionally, the LPSTyper tool identified the LPS outer core type of Escherichia fergusonii RHB19-C05 (NZ_CP057657.1) as R2. However, the LPS type could not be determined for Escherichia albertii, Escherichia hermannii or Escherichia ruysiae.

Distribution of LPS outer core types within commensal and pathogenic E. coli

To investigate the association between E. coli LPS outer core types and pathogenicity, we categorized the dataset into several pathogenic and non-pathogenic groups, each containing more than 20 strains. Pathogenic groups included enterohaemorrhagic E. coli (EHEC; n=187), enterotoxigenic E. coli (ETEC; n=44), avian pathogenic E. coli (APEC; n=23), neonatal meningitis E. coli (NMEC; n=27), uropathogenic E. coli (UPEC; n=369) and other non-typed extraintestinal pathogenic E. coli (ExPEC; n=21). Additionally, 1,394 strains were classified as pathogenic in the NCBI database without subtype designation. The remaining strains consisted of just under half the dataset (n=2,452) and were identified as commensal strains. Strains with unidentified LPS core types (n=44) were excluded from the analysis.

Our findings indicated that the R1 LPS outer core type was the most prevalent among all E. coli strains analysed, accounting for 35.38% (Fig. 3a). It was followed by R3 (25.79%), K-12 (18.38%), R4 (12.3%) and R2 type (8.15%) as the least common. Commensal E. coli strains showed a similar distribution as the whole dataset, with the R1 LPS type being most prevalent (36.9%), followed by R3 (23.4%), and R2 being the least common (8.7%) (Fig. 3b).

Fig. 3. Distribution of E. coli LPS outer core types within pathogenic and commensal groups. (a) Distributions of LPS outer core types within 4,505 commensal and pathogenic E. coli strains with identified LPS outer core types. (b) Distribution of LPS outer core types within different pathogenic subgroups of E. coli and commensal strains. Asterisks (**) indicate groups with a moderate effect size based on Cramér’s V values between 0.2 and 0.6. The numerical data for the figure are available in Tables S3 and S4.

By contrast, the distribution of LPS outer core types across different E. coli pathogenic types revealed significant variations, underscoring associations between specific LPS structures and E. coli pathotypes (Fig. 3b). EHEC strains were predominantly associated with the R3 core type (97.9%), with the remaining ~2.1% composed of R1 and R4. This group is primarily represented by E. coli O157:H7 strains in databases, which may constrain the observed LPS outer core genetic variability. ETEC strains were predominantly associated with the R1 LPS outer core type (65.9%), with moderate representation of R4 (18.2%) and R3 (13.6%). No ETEC strains were associated with the R2 type. Within the ExPEC pathotype, NMEC strains showed a near-exclusive association with the R1 LPS type (92.6%), with minor representation in R3 (7.4%) and no presence in other LPS types (Fig. 3b). Likewise, APEC strains were primarily associated with the R1 core (69.57%), with a smaller proportion linked to the R4 type (26.09%). The other ExPEC group, UPEC, was also dominated by R1 LPS type (60.4%) but exhibited the remaining LPS outer core types: K-12 (16.8%), R3 (11.6%), R4 (7.9%) and R2 (3.2%). ExPEC strains that could not be classified into a subtype were dominated by R1 type (52.4 %) but appeared to match the overall LPS type pattern of UPEC strains rather than NMEC strains. However, the K-12 LPS type was conspicuously absent from this group, suggesting the presence of an additional unclassified subtype within ExPEC (Fig. 3b). Pathogenic strains that were not split into subgroups displayed a more even distribution, with moderate prevalence in R1 (27.7%), R3 (25.2%), K-12 (21.6%) and R4 (15.3%) and a lower association with R2 (10.2%), likely a result of the mixed nature of the grouping. Across all E. coli populations, the R1 LPS outer core structure was the most dominant except in EHEC strains, while the R2 type was the least common and entirely absent in EHEC, NMEC and ETEC strains. The K-12 LPS structure was also lacking in EHEC and NMEC strains. A chi-square test indicated a significant association between E. coli pathotype and LPS outer core types [χ²(44)=789.58, P<0.001]. The strongest contributor to the association was the EHEC pathotype with the R3 LPS outer core type, accounting for 47.7% of the total chi-square statistic (standardized residual=19.42). The strength of this association was further supported by a moderate effect size (Cramér’s V=0.34). All other pathotype-phylogroup combinations contributed less than 10%. These findings suggest that, with the exception of EHEC, the distribution of LPS outer core types is not strongly associated with specific pathogroups.

Distribution of LPS outer core types within E. coli phylogenetic groups

The established phylogenetic groups of E. coli (A, B1, B2, C, D, E, F and G) represent bacteria occupying different ecological niches with sets of similar characteristics, such as antibiotic resistance, metabolic capacity and pathogenicity [32,33]. Using the LPSTyper results, we compared the LPS outer core distribution across these eight major E. coli phylogroups.

A chi-square test indicated a highly significant association between phylogenetic groups and LPS core types [χ²(28)=4,338.3, P<0.001). Commensal phylogroup A (n=1352) showed the most balanced distribution among the outer core types, with R1, R2 and K-12 present in similar proportions at 28.8%, 25.5% and 28.6%, respectively, and lower representation of R3 (11.9%) and R4 (5.1%)(Fig. 4). However, across all groups, R2 and R3 showed a moderate association with phylogroup A (Cramér’s V=0.41 and 0.21, respectively) (Table S6). Phylogroup B1 (n=891) was dominated by R1 (42.5%, V=0.07, no significant association) and R3 (49.7%, V=0.28, moderate association), while K-12 was rare (0.65%) but still showed a moderate association (V=0.23). Phylogroup B2 (n=696) was dominated by K-12 (43.6%, V=0.36, moderate) and R1 (53.4%, V=0.09, no association). By contrast, phylogroup C (n=165) was almost exclusively associated with R1 (99.4%), which showed a moderate association (V=0.26). Phylogroup D (n=380) exhibited the highest proportion of R4 (60.9%, V=0.47, moderate) and a substantial proportion of R3 (28.9%) although the latter did not display a statistically significant association (V=0). Phylogroup E (n=436) accounted for more than half of the pathogenic strains and was predominantly associated with the R3 core type (84.3%, V=0.45, moderate), while the infrequently observed R1 (2.3%) also showed a moderate association (V=0.23). Phylogroup F (n=128) showed a high proportion of R4 (51.4%, V=0.2, moderate) and R1 (45.7%, V=0.03, no significant association), whereas phylogroup G (n=49) had the highest proportion of R4 (84.9%, V=0.23, moderate) among all groups. These results indicate that while the majority of phylogroups exhibit moderate associations with particular LPS outer core types, their overall distribution patterns tend to overlap. This study also provides, to our knowledge, the first report on LPS outer core types for phylogroup G.

Fig. 4. Distribution of LPS outer core types within E. coli phylogenetic groups. Asterisks (**) indicate phylogenetic groups with a moderate effect size based on Cramér’s V values between 0.2 and 0.6. The numerical data for the figure are available in Tables S5 and S6.

Distribution of LPS outer core types within E. coli serogroups

During LPS assembly, the O-antigen is covalently attached to the outermost hexose of the LPS outer core by the O-antigen ligase WaaL. Due to the structural and chemical requirements of this reaction, we hypothesized that there might be an association between O-antigens and LPS outer core types. To investigate this, we analysed the distribution of O-antigens among E. coli strains with LPS outer core types determined by LPSTyper and examined the pairwise frequency of each core type with predicted O-antigens across our E. coli genome dataset.

We found that all O-antigens were strongly associated with only one LPS outer core type except for O1, O5, O83, O128 and O141, which were associated with multiple core types (Table S7). The O128 serotype frequently co-occurred with R1 (35.0%) and R3 (30.0%), while O83 was evenly distributed between R1 (46.1%) and R4 (53.8%) types. Similarly, O141 showed 42.9% co-occurrence with both R2 and R3. By contrast, the O5 serotype displayed a broader distribution, associating with K-12 (34.3%), R1 (28.6%) and R3 (34.3%) (Table S7). Our analysis revealed that a total of 136 out of 197 (69.0%) predicted O-antigen types co-occurred at least once with the R1 outer core type, 64.0% with R3, 38.1% with R4, 31.5% with R2 and 17.8% with the K-12 outer core type (Table S7). By contrast, K-12 and R2 outer core types were found at low frequencies or were absent for many O-antigen types.

To measure the strength of association between specific LPS outer core types and O-antigen types, we calculated Cramér’s V statistics with a bias correction to quantify the effect size [29]. Although we found 197 serotypes in our data, including mixed types, only 29 serotypes reached abundances necessary to calculate the association with LPS core types (Fig. 5a, Table S8). We identified a statistically significant and very strong association between O16 O-antigen – one of the most common EHEC-associated O-types [34] – and the K-12 LPS outer core type (Cramér’s V=0.62). Similarly, O25, a serogroup frequently found in ETEC strains [34], was also strongly associated with K-12 (0.54), while O157 showed a very strong association with R3 (0.67). Meanwhile, the R2 type was strongly associated with the O89 serogroup (0.79), which has recently been shown to be a rough strain serotype that lacks any expressed O-antigen [7]. We also observed moderate associations between O1 (0.21) serotype that is primarily linked to bloodstream and urinary tract infections [35], O102 (0.36) and O86 (0.29) with R4. Lastly, we found the O157 (0.26), O6 (0.26), O8 (0.27), O16 (0.25) and O25 (0.21) serotypes associated with R1.

Fig. 5. Distribution of E. coli LPS outer core types within select O-antigens, capsule groups, K-antigens, and sequence types. (a) A heatmap of associations between E. coli LPS outer core types and serotypes with calculated Cramér’s V. Cramér’s V association effect is denoted on the heatmaps as *** very strong association (effect size, >0.6) or ** moderate association (0.2<effect size≤0.6). Weak associations are not shown. (b) E. coli LPS outer core type distribution within capsule groups. (c) E. coli LPS outer core type distribution within specific capsule K-antigens. (d) A heatmap showing associations between LPS outer core types and E. coli STs, with Cramér’s V effect sizes indicated as in panel (a). The numerical data for the figure are available in Tables S7–S12.

These findings suggest a potential co-evolutionary relationship between O-antigens and LPS outer core types likely driven by selective pressures ensuring structural compatibility, stability and functional synergy [36].

Distribution of LPS outer core types within E. coli capsule types

The E. coli capsule, typically represented by K-antigens, is structurally and functionally linked to the LPS core and O-antigen polysaccharides through shared biosynthetic pathways that ensure outer membrane integrity [37,38]. This coordination aids immune evasion, as the capsule masks O-antigens, while the LPS core supports the attachment of both capsule and O-antigens [1,37].

We analysed the distribution of LPS core types within strains with group 2 and group 3 capsules, as well as E. coli lacking capsules (Fig. 5b). For strains with group 2 (n=1358) and 3 (n=74) capsules, the R1 outer core type was the most prevalent, accounting for 45.0% and 51.3%, respectively. The R4 core type showed similar frequency distribution in both groups (19.0% and 21.6%), whereas the K-12 core type was significantly more common in group 2 than in group 3 (27.8% vs. 4.0%). Conversely, the R3 core was more frequent in group 3 (21.6%) compared with group 2 (6.8 %).

In strains lacking capsules (or with unidentified group 1 capsules) (n=3117), the R3 outer core type was the most common (34.1%), followed by R1 (30.8%) and K-12 (14.6%), while R4 and R2 were less frequently observed (9.2% and 11.2 %, respectively). Among specific K-antigens, the distribution patterns are more distinct (Fig. 5c). R1 was highly associated with K1, K15, K23 and K96, with frequencies ranging from 63% to 92%. By contrast, K52 and K98 showed the lowest representation of R1 (6.1% and 7.7%). The R4 outer core type was predominantly linked to K4 and K98 (61.5% each), while K52 was most strongly associated with R3 (67.3%).

These findings, showing LPS outer core enrichment in specific capsule types and across broad capsule groups, suggest that the presence and type of capsule might be influenced by the distribution of LPS outer core types, with R1 being particularly enriched in capsule-containing strains. Conversely, the R3 LPS outer core type was more frequently associated with capsule-deficient strains. However, R3 was highly enriched in K52-type capsules, suggesting structural or functional advantages for the co-occurrence of the R3 LPS outer core with at least some capsule structures (Table S10).

Distribution of LPS outer core types within E. coli STs

Sequence Types (STs) in E. coli are defined by multilocus sequence typing, a method that assigns allelic profiles to isolates based on the DNA sequences of conserved housekeeping genes [39]. Comparing the distribution of STs and their associated LPS core types can reveal potential links between the genetic background and surface structure variation in E. coli.

In our dataset, STs formed distinct clusters within LPS outer core types, with certain STs comprising 100% of isolates within specific LPS groups (Fig. 5d, Table S11). For those STs with sufficient counts for statistical analysis, we identified a significant association between LPS outer core types and STs (χ²=8803.1, df=120, P<0.001). One of the strongest associations was observed between ST131, a globally disseminated multidrug-resistant clone strongly associated with both urinary tract and bloodstream infections [40,41], and the K-12 LPS outer core type (Cramér’s V=0.64). Additionally, ST131 formed moderate associations with R1 and R3 LPS outer core types. R1 was also moderately associated with ST10, ST11, ST58, ST73, ST93, ST95, ST410 and ST1193 (Table S12). For R2, the strongest association was observed with ST167 (V = 0.49), an emerging high-risk, multidrug-resistant lineage often carrying carbapenemase genes such as bla_NDM-5 [42]. Moderate associations were also seen with ST10, a common lineage frequently carrying extended-spectrum β-lactamase (ESBL) genes [43], and ST744, an emerging multidrug-resistant clone [44]. R3 was most strongly linked to ST11 (V=0.67), the lineage that includes the O157:H7 serotype and corresponds with our previous finding of O157 association with R3 (Fig. 5a). R3 also displayed a moderate association with ST69 (V=0.32), a pandemic uropathogenic clone prevalent in community-acquired urinary tract infections and occasionally isolated from bloodstream infections [45]. For R4, ST38 was a strongly associated ST (V=0.54). ST38 is a globally disseminated ExPEC clone known for ESBL production and frequent involvement in urinary tract and bloodstream infections [45]. ST117, associated with APEC and zoonotic spillover [46], ST405, an ESBL-producing uropathogenic lineage [47], and ST648 – a high-risk clone linked to both human and animal infections [48] – showed moderate associations with R4 (Table S12).

Discussion

Our study introduces a rapid and comprehensive framework for in silico and in vitro typing of E. coli R1, R2, R3, R4 and K-12 LPS outer core structures by targeting unique regions within the waa locus. Using nine novel oligos, we achieved robust typing with a 99% success rate in silico across 4,549 complete E. coli genomes from the NCBI database using our LPSTyper tool. We successfully validated these new improved oligos (Table 2) on reference strains using a simple in vitro PCR method (Table 1 and Fig. 2).

LPS outer core type patterns in commensal and pathogenic E. coli

Our analysis of LPS outer core types in 4,549 genomes, including commensal and pathogenic strains, revealed distinct distribution patterns. The R1 outer core type emerged as the most prevalent, as reported in previous works [13,14, 49, 50], followed by R3, K-12 and R4, with R2 being the least common. In earlier studies, R1 type was observed in 68.3% of 180 strains typed using mouse monoclonal antibodies [13]; however, nearly 30 years later, it was detected in only 33.9% of 499 strains using PCR typing [50]. Our current study of 4,549 strains identified the R1 type in 35.5% of E. coli strains (Fig. 3a), indicating that while R1 remains a prominent core type, its relative abundance diminishes in larger, more diverse genomic datasets. Our data also reveal a significant enrichment of the R1 LPS core type in specific pathogenic E. coli strains, including UPEC (59.7%), NMEC (92.6%), ETEC (65.9%) and APEC (68.6%). The prevalence of the R1 core structure may confer a selective advantage by facilitating the acquisition of virulence genes, enhancing colonization in various host tissues or improving survival and replication in the circulatory system [14,51]. This hypothesis warrants further investigation to fully elucidate the role of the R1 core in the biology and pathogenesis of E. coli.

Exploring the connection between O-antigens and LPS outer core types in E. coli

The relationship between serotypes and LPS outer core structures in E. coli remains a subject of ongoing investigation, as the O-antigen is covalently attached to a specific residue of the LPS outer core [52]. Previous studies on smaller E. coli genomic datasets lacked sufficient sample sizes to determine statistical significance or identify associations between these features [14,50]. However, our data reveal significant associations between specific O-antigen types and LPS outer core types (Fig. 5a). Notable examples identified in this work include K-12 with O16 and O25 serotypes, R2 with O89, R3 with O157-antigen and R4 showing moderate associations with O102 and O86 serotypes. We also observed that while LPS outer core types can be linked to multiple O-antigen serotypes, each O-antigen type generally exhibits a preference for a single LPS outer core type. However, some O-antigens such as O1, O5, O83, O128 and O141 were compatible with multiple outer core types (Fig. 5a, Table S7). This ability to link diverse O-antigens to each LPS outer core type underscores the dynamic nature of LPS biosynthesis and likely relies on the precise enzymatic transfer of the O-antigen from a pyrophosphate-undecaprenyl carrier to the LPS core by WaaL, an enzyme with broad donor substrate specificity [53]. Further investigation is needed to understand this variability in O-antigen-core linkages fully.

Associations between particular combinations of O-antigen and outer core types may play a role in enhancing bacterial immune evasion and host adaptation. For instance, EHEC strains are predominantly characterized by the R3 outer core type and the O157 serotype, which account for 78.5% of the phylogenetic group E in our data. The prevalence of the R3 and O157 combination suggests that there is a selective advantage for pathogenicity, host colonization or defence against bacteriophage-mediated attack. Beyond bacteriophage resistance, evidence indicates that predatory protozoa, such as Acanthamoeba castellanii, can influence E. coli survival by selectively modulating bacterial consumption based on E. coli LPS outer core type and O-antigen composition [54,55]. This suggests that variations in LPS structure may indirectly enhance virulence and contribute to environmental persistence by reducing susceptibility to predation.

Hybrid and novel LPS outer core structures

In this work, we identified novel non-canonical waa locus gene structures (Table S15), including genomic rearrangements between R2 and K-12 LPS outer cores, gene duplications and the presence of transposable elements, highlighting the dynamic nature of the waa locus (Fig. 1). These findings align with previous studies suggesting recombination and horizontal gene transfer drive LPS variability, potentially influencing host immune evasion and bacterial environmental adaptability [56,58].

Notably, we show that the R2 outer core type, the rarest E. coli LPS type in our dataset, was associated with several hybrid variations involving the K-12 core type (Figs 1 and S1, Table S15 and S21). This association between R2 and K-12 outer core types adds evidence to the notion that the R2 outer core type itself is a hybrid of E. coli K-12 and Salmonella enterica LPS core types [59]. In the K-12/R2 gene arrangements we identified, eight K-12 E. coli strains harboured waaK genes instead of waaU and a waaL gene highly similar to that of R2 but not K-12, whereas one R2 strain contained waaU in place of waaK, along with a waaL gene resembling that of K-12 rather than R2 (Fig. S1, Tables S15 and S21). The waaK gene encodes UDP-N-acetylglucosamine:(glucosyl) LPS α1,2-N-acetylglucosaminyltransferase, which in the R2 outer core plays a critical role by adding a terminal α1,2-linked GlcNAc side branch [59]. This modification is essential for recognition by the O-polysaccharide ligase WaaL. By contrast, in the K-12 outer core, WaaU, an ADP-l-glycero-β-d-manno-heptose:LPS heptosyltransferase IV, attaches a fourth heptose residue to the third glucose, serving as the site for O-antigen attachment. At first glance, these hybrid structures might be expected to resemble either the K-12 or R2 core. However, additional modifications introduced by WaaS and WabA could result in structures that do not precisely match either parental type. Notably, the hybrid K-12/R2 outer cores can accommodate O-antigens not supported by the original K-12 structure, such as O127, O19 and O58, indicating that the replacement of the waaL gene contributes to the modified outer core composition. Interestingly, the single R2/K-12 hybrid carried the O3 antigen (Fig. 1, Table S15), which is typically associated with R1, R2, R3 and R4 outer core types, and no additional O-antigen types unrelated to R2 were incorporated into this chimaera. Due to the limited number of hybrid strains identified, it remains unclear whether specific O-antigens preferentially associate with these hybrid LPS outer cores or whether additional structural constraints influence their compatibility. Further studies incorporating a broader genomic dataset would be necessary to clarify these relationships.

Additionally, our study identified an R1/R4 hybrid in the ECOR collection strain ECOR46. The formation of such a hybrid core is thought to require the co-expression of waaL genes specific to R1 and R4, along with waaV and waaX [14]. Notably, ECOR46 meets these criteria, possessing R1- and R4-specific waaL, waaV from R1 and waaX from R4. This result suggests that ECOR46 should not be strictly classified as R1 or R4. However, the exact structure of its LPS chimaera remains unclear, requiring further experimental characterization.

In conclusion, this study underscores the evolutionary plasticity of the waa locus and highlights the effectiveness of our primers and in silico typing tool in revealing the diversity and distribution of LPS outer core structures in E. coli. The strong associations between LPS types, pathogenicity, phylogenetic groups, STs and genomic rearrangements emphasize the significance of LPS variability in bacterial fitness, adaptation and virulence. Our findings demonstrate that the in silico typing method is the most reliable approach for genome-based analysis and should be prioritized when genomic sequences are available. Moreover, the differences between this work and prior reports [14,50] highlight the need for regular updates to typing tools to accommodate novel genomic variability uncovered by increasing deep sequencing of the biosphere. As more genomic sequences become available, expanding analyses to encompass incomplete genomes will likely reveal a broader diversity of hybrid LPS outer core types. These efforts will strengthen our understanding of LPS structural evolution and its role in E. coli ecology and pathogenicity.

Abbreviations

APEC

avian pathogenic Escherichia coli

ECOR

Escherichia coli reference

EHEC

enterohaemorrhagic Escherichia coli

ESBL

extended-spectrum β-lactamase

ETEC

enterotoxigenic Escherichia coli

ExPEC

extraintestinal pathogenic Escherichia coli

LPS

lipopolysaccharide

NMEC

neonatal meningitis Escherichia coli

ST

sequence type

UPEC

uropathogenic Escherichia coli