Phenotypic and genotypic relationships between Shigella species and Escherichia coli O37:H10 isolates in Addis Ababa, Ethiopia

Abstract

Objective

This study aimed to investigate the genotypic and phenotypic differences between Shigella species and E. coli O37:H10, as well as their antimicrobial resistance (AMR) and virulence factors, in children aged under five with diarrhea in Addis Ababa, Ethiopia.

Results

Using whole genome sequencing (WGS), all 28 S. sonnei identified with antisera were also verified as S. sonnei, while all nine isolates of S. flexneri were found to be E. coli O37:H10. One S. flexneri sample, identified serologically, was excluded from WGS analysis because of low genome coverage. Most resistant genes were associated with beta-lactam and trimethoprim drugs. All S. sonnei isolates had genes that encoded blaEC-8 and blaZEG-1. The E. coli O37:H10 had the Extended-spectrum beta-lactamase (ESBL) blaEC-15 for cephalosporins, blaMIX-2, and blaMIX-6 for penicillin. In contrast to S. sonnei, the most prevalent plasmid type in E. coli O37:H10 isolates was IncFII, while virulence genes like gad were frequently detected. Composite transposons, such as cn_43599_IS621 and cn_3785_IS3, were detected in all S. sonnei and E. coli O37:H10. Generally, the study highlighted a significant discrepancy between phenotypic and genotypic AMR, as well as variations in serotypes and phylogenetic relationships with global isolates.

Background

Shigella species are one of the most common enteric bacteria responsible for bacillary dysentery, particularly in humans [1]. Shigella species and E. coli are closely related to Gram-negative bacteria [2]. Differentiating these two bacteria is quite difficult. Shigella species have been shown conclusively to be E. coli clones by sequencing the housekeeping genes [3]. The O antigen, which contains several repeats of an oligosaccharide unit (O unit), is a component of the lipopolysaccharide (LPS) found in Gram-negative bacteria’s outer membranes [4]. Shigella species have O antigen and lack H and K antigens, which are found in E. coli [5]. The genes involved in O antigen production are often present on the chromosome as an O antigen gene cluster, and genetic variation within the gene cluster is the primary reason for the diversity of O antigen types [6]. Although Shigella and E. coli have long been recognized as closely related, it was not until the 1940s that Shigella strains were differentiated from E. coli and formally classified into their genus and subgroups [2]. Shigella is a highly contagious bacterium in comparison to other causes of gastroenteritis, causing an infection with only 10 bacilli [7]. Additionally, the LPS O antigens of known Shigella serotypes (excluding S. sonnei) share one or more O antigen groups of E. coli [5]. In clinical or public health settings, it is essential to differentiate between E. coli and Shigella species and accurately identify the specific Shigella species detected [8]. Conventional microbiological approaches may result in conflicts when distinguishing Shigella strains from E. coli, possibly due to the close genetic relatedness among all four Shigella species, which exhibit high DNA-DNA similarity [9, 10]. Serological detection of bacterial strains that produce Shigella-like colonies on selective agar plates, and biochemical testing, reveals cross-reactivity with other bacteria, such as E. coli and Shigella-specific antibodies [11]. Commercial tests (such as Phoenix and serotyping) may misidentify inactive E. coli isolates as Shigella, necessitating additional testing to achieve a definitive identification [12]. For revealing evolutionary links between strains, researchers are shifting to WGS due to its superior resolution over earlier techniques (e.g., Pulsed-field gel electrophoresis (PFGE)) [13]. The WGS analysis of k-mers and single-nucleotide polymorphisms (SNPs) exhibits high discriminatory power [10]. In Ethiopia and other developing countries, bacteria such as Shigella and E. coli are typically identified using culture and biochemical approaches, which often struggle to distinguish between species with overlapping characteristics. Therefore, this study aimed to investigate the genetic relationship between Shigella species and E. coli O37:H10 using WGS in Addis Ababa, Ethiopia.

Methods

Study sites, recruitment of study participants, and sample collection

The study utilized stored Shigella isolates collected from diarrheic children at four public health facilities (health centers) in Addis Ababa between June 2021 and April 2022. The isolates were obtained as part of a larger study investigating the genomic diversity and AMR profiles of Shigella species among children under five years of age. Fresh stool specimens were collected and cultured on MacConkey (MAC) and xylose lysine deoxycholate (XLD) agar to isolate bacterial growth. Suspected colonies were identified as Shigella through biochemical testing using standard bacteriological methods as described in the earlier study [14]. Further characterization included sero-grouping by slide agglutination with polyvalent antisera and antimicrobial susceptibility testing (AST) using the disk diffusion method.

DNA extraction, library preparation, and WGS

The stored isolates were sub-cultured, and DNA extraction, library preparation, and WGS were performed under identical conditions, following the manufacturer’s instructions as detailed in the previous study [15].

WGS data analysis

The quality of the paired-end raw reads was assessed using standard FAST-QC methods. Low-quality sequences, adapter contaminants, and poor-quality bases at the ends were removed through trimming and filtering. Subsequent steps, including Quality scoring, species confirmation, alignment, and mapping, as well as filtration for phylogenetic tree construction, were performed as stated in the previous study [15]. Prediction of serotypes in total sequenced isolates of E. coli was identified using the SeroTypeFinder tools at the CGE web-based platform with a default threshold. We used genomic data to find plasmid, virulence, and AMR genes using an integrated strategy, including Mykrobe software as detailed in the previous study [15].

Results

Serotypes

Of the 47 Shigella species initially identified based on cultural and biochemical characteristics, 31 were serologically confirmed as S. sonnei and 16 as S. flexneri. For whole-genome sequencing (WGS), 28 S. sonnei and 9 S. flexneri isolates were selected, with the reference strain NZ_CP055292.1 included for analysis. The WGS results showed that all 28 isolates of S. sonnei were identified as S. sonnei SE6-1 strains, with query coverage ranging from 87.61 to 92.43%. In contrast, all nine isolates of S. flexneri were identified as E. coli O37:H10 serotypes, corresponding to two type strains (E. coli JNQH497 & E. coli JNQH498), with Query-coverage ranging from 86.83 to 87.05.

Comparison between phenotypic and genotypic AMR

Among the 28 S. sonnei isolates, 7 and 17 showed discordant results for trimethoprim and cefoxitin, respectively, when comparing phenotypic and genotypic AMR profiles as described in Table 1. In contrast, 8 E. coli O37:H10 isolates showed concordant results for trimethoprim between the phenotype and genotype (Table 1). In E. coli O37:H10, the phenotypic and genotypic AMR data corresponded well for amoxicillin and ampicillin across susceptibility profiles. The majority was susceptible based on AMR data gathered from the genome. However, since these isolates were phenotypically resistant to the respective antimicrobials, being genotypically susceptible, all were classified as major errors.

Table 1.

Evaluation of genotypic analysis for predicting phenotypically susceptible patterns of S. sonnei (n = 28) and E. coli O37:H10 isolates (n = 9)

Species	Number isolates (%)	Antimicrobial tested	Drug class	Resistance status		Resistant gene	Accession
				Phenotype	Genotype
S. sonnei	7 (25)	Cotrimoxazole (trimethoprim)	Folate synthesis inhibitors	Susceptible	Resistant	dfrA1	KU546748.1
	17 (60.7)	Cephalosporin (cefoxitin)	Beta-lactam	Susceptible	Resistant	blaEC-8	NG_049086
	28(100)	Unknown beta-lactam	Beta-lactam	Unknown	Resistant	blaZEG-1	AY265891
E. coli O37:H10	1 (11.1)	Co-trimoxazole (trimethoprim)	folate synthesis inhibitors	Susceptible	Resistant	dfrA1	KU546748.1
	4 (44.4)	Cephalosporin (cefoxitin)	Beta-lactam	Susceptible	Resistant	blaEC-15	NG_049081
	4 (44.4)	Cephalosporin, (cefoxitin)	Beta-lactam	Susceptible	Resistant	blaMIR-2	AY227752
	4 (44.4)	Cephalosporin (cefoxitin)	Beta-lactam	Susceptible	Resistant	blaMIR-6	JQ664733
	9 (100)	Amoxicillin/ ampicillin	beta-lactam	Resistant	Resistant	blaMIR-2& blaMIR-6	AY227752 & JQ664733

Resistance to beta-lactamase and folate synthesis inhibitors

Among the blaEC-8 genes identified in the S. sonnei genotyping analysis, about 60.7% of the corresponding isolates were phenotypically susceptible to cefoxitin, as presented in Table 1. All isolates of E. coli O37:H10 corresponded well for amoxicillin and ampicillin, as shown in Table 1. Extended-spectrum beta-lactamase (ESBL) blaEC-15 for cephalosporin, blaMIX-2, and blaMIX-6 for penicillins were detected in all E. coli O37:H10 isolates (Fig. 1). All isolates carried genes in S. sonnei and E. coli O37:H10 conferring resistance to trimethoprim. The gene isolates conferring resistance to trimethoprim had the dfrA1 gene. Four isolates (B230, B235, B375 and B405) were phenotypically resistance to cefoxitin had blaEC-15, blaMIR-2 and blaMIR-6 genes (Table 1).

Fig. 1.

Genes associated with S. sonnei and E. coli O37:H10 among the clinical isolate samples from under-five children in Addis Ababa (Bio Plot analysis source; https://www.chiplot.online/)

Mobile genetic elements (MGEs)

Seven types of Insertion Sequence (IS) elements, characterized by varying sequence coverage and length, were identified in all sequenced genomes of the isolates. The IS elements contained open reading frames (ORFs) encoding the transposase enzyme. The identified IS elements included IS609, ISEsa1, IS621, IS3, IS30, ISEc1, and ISEc30. Additionally, Miniature Inverted-repeat Transposable Elements (MITEs), such as MITEEc1, were detected as part of a group of non-autonomous Class II transposable elements. Composite transposons were also identified. Specific composite transposons, such as cn_43599_IS621 and cn_3785_IS3, were detected in all S. sonnei and E. coli O37:H10 isolates. Furthermore, the transposon cn_8959_ISEc1 was found exclusively in S. sonnei, while cn_8950_ISEc1 and cn_44020_IS621 were identified in all E. coli O37:H10 isolates. Additionally, cn_44020_IS621 was detected in one isolate of S. sonnei.

Plasmid profile

Using the PlasmidFinder tool, we identified that the IncB/O/K/Z, IncFII(pHN7A8), IncFII_1_Psfo, Col(MG828), Col(BS512), and Col156 plasmid types differed in position and identity in E. coli O37:H10. The plasmid types IncB/O/K/Z and Col (BS512) in E. coli O37:H10 were present in all isolates, while the IncB/O/K/Z plasmid was absent in certain S. sonnei isolates. The Col (MG828) plasmid was identified in all samples except one (B166) in E. coli O37:H10; however, the Col156 plasmid was detected in all S. sonnei but only in one isolate (B230) in E. coli O37:H10. The most prevalent plasmid type in E. coli O37:H10 isolates was IncFII; however, the IncFII plasmid incompatibility element and Col (MG828) plasmids were only identified in one S. sonnei isolate (Fig. 1).

Virulence gene profile

The finding of virulence genes revealed few differences between isolates and no genes generating Shigatoxin. Except for the quantity, the virulence genes identified in S. sonnei and E. coli O37:H10 were the same. Most isolates had virulence genes, including gad, which is involved in glutamate decarboxylase in epithelial cells. All E. coli O37:H10 isolates contained two gad genes, except for one isolate, which had four; In contrast, among the S. sonnei isolates, one had a single gad gene, 11 isolates had three gad genes, and 16 had two gad genes. Additional virulence genes identified in the isolates include csgA, fdeC, fimH, hlyE, nlpl, lpfA, terC, yehA, yehB, yehC, and yehD, as shown in Fig. 1.

Phylogenetic analysis S. sonnei and E. coli O37:H10

Based on SNP with 100% bootstrap values, the phylogenetic tree demonstrated that S. sonnei were distinctly separate from E. coli O37:H10, as shown in Fig. 2. Most strains were more closely related, except one strain (B394) in S. sonnei that was split and showed polyphyletic groups. Some trees were differentiated within S. sonnei and E. coli O37:H10 with 100% bootstrap values. The E. coli O37:H10 isolates diverged early and separated from S. sonnei. The S. sonnei isolates were distinguishable from E. coli O37:H10, and these groups formed a large branch separate from the other S. sonnei isolate (B394). Thus, its evolutionary history could be different from those of S. sonnei. As a result, the number of nucleotide frequencies and DNA sequence diversity in the study differed between isolates, implying numerous origins.

Fig. 2.

Phylogenetic analysis of S. sonnei and E. coli O37:H10

Phylogenetic relationships of E. coli O37:H10 and S. flexneri

The merged tree showed the presence of two different groups: A (S. flexneri) and B (E. coli O37:H10). However, when considering the 100% bootstrap value at the node of separation, the A group’s sub-clusters were more diverse than the B group’s. A study of Ethiopian samples from Tanzania supported the diversity when only S. flexneri isolates were analyzed (Fig. 3). The Indian strains of S. flexneri were more comparable in identity and shared ancestors with the reference than others. Most sub-clusters in B strains had identical sequences that differed somewhat within E. coli O37:H10 isolates. However, B405 and B235 isolates and B139 and B506 isolates of E. coli O37:H10 in the B group appeared to represent ancestral forms of B166 and B388, respectively.

Fig. 3.

Phylogenetic relationship of E. coli O37:H10 and S. flexneri. Sequencing data for the eight global S flexneri strains (black-group A) was retrieved from the NCBI SRA and provided into SNVPhyl with the E. coli O37:H10 (blue) to carry out SNP calling

Phylogenetic relationships of local and global isolates of E. coli

Single-nucleotide polymorphism (SNP) calling was performed, and a maximum-likelihood tree was constructed to show the relatedness of the isolates. Algerian strains emerged earlier than others as shown in Fig. 4. The reference (R) strains were shown to be more closely linked in studies done in Ethiopia and Kenya than in others. The local E. coli O37:H10 strain and Chinese isolates shared a common ancestor. As a result, the E. coli retrieved from China were more closely related to the current inquiry than the other isolates (Fig. 4). Most global isolates formed distinct clusters that differed significantly from the current findings. Therefore, clustering appears to be geographically driven, with separate clusters created by foreign and local E. coli isolates.

Fig. 4.

Phylogenetic relationship of E. coli O37:H10 with the global isolates. Sequencing data for the eight global E. coli strains (red) was acquired from the NCBI SRA and provided to SNVPhyl with the local isolates to carry out SNP calling

Discussion

Distinguishing between Shigella and E. coli can often be biochemically challenging, as there are aerogenic (gas-producing) Shigella and lactose-negative, non-gas-producing, non-motile E. coli strains. Additionally, new varieties of the Shigella serotype that do not agglutinate with commercially available antisera are becoming more common [16]. Like the present study, discrepancies were reported between phenotypic and genotype identification between Shigella species and E. coli in India [17]. In the current study, all nine isolates were serologically confirmed to be S. flexneri and were found to be E. coli O37:H10 through WGS analysis using KmerFinder and SerotypeFinder. This discrepancy may be due to antisera failing quality control procedures, unresolvable cross-reactions, dysfunctional O antigens encoding genes, or novel serogroups [18]. Most Shigella O antigens cross-react serologically with inactive strains of E. coli O antigen [12, 19]. As a result, laboratory identification of Shigella by agglutination is difficult using commercially available polyvalent antisera and may not cover all potential epitopes of the Shigella O-antigen.

The phylogenetic tree may infer the relationships between isolates by constructing clades and branches [19]. Some S. sonnei strains had a significant level of genetic similarity in this study. Almost all isolates from closely related strains formed separate, smaller clusters. However, general grouping is independent of geographic origin, as one main cluster contains both international and local isolates of S. sonnei [15]. Nine E. coli O37:H10 isolates in this study deviated from S. sonnei, indicating distinct origins. Shigella has been separated as a genus but comprises several clones of E. coli [20]. The strong evolutionary relationships between Shigella and E. coli explained unresolved biochemical and serological test results, which were resolved by WGS data [8]. The phylogenetic relationships between the E. coli O37:H10 and S. flexneri revealed that E. coli O37:H10 isolates formed separate branches from S. flexneri. Like pathogenic forms of E. coli O37:H10, S. flexneri isolates lack a single evolutionary origin, implying that Shigella phenotypic traits evolved convergent. Shigella clusters appear to have evolved over 35,000-270,000 years, suggesting that shigellosis was one of the oldest human infectious diseases [3, 20]. Tanzanian strains (sample source from Ethiopia) emerged early, and the Indian strains of S. flexneri were more comparable in identity and shared ancestors with the reference. When comparing the E. coli O37:H10 and Shigella clusters, it was noticeable that the E. coli O37:H10 clusters less diverged than the S. flexneri clusters. Assuming that the housekeeping genes evolve at roughly identical rates in both forms, we can speculate that the E. coli O37:H10 has had less time to diverge within its clusters and may have emerged earlier than the Shigella strains. Similar to the current study, Shigella phenotypes were also found to have evolved individually multiple times [3, 21]. The S. flexneri strains may have diverged more phenotypically than E. coli O37:H10. This may also apply to S. sonnei clusters. Because the cluster branch lengths are long in both cases and the bootstrap value is large, the clusters must have been isolated for a long time, even if they descended from a common ancestor strain. In comparison to non-O157 serotypes, the highly pathogenic and most investigated type of E. coli was serotype O157:H7 [22]. When E. coli O37:H10 strains were compared to other human isolates from around the world, it was found that the Chinese strains were more closely connected to the current study. Most isolates have a variety of genotypes and are not particularly attached to strains from other particular places. According to the phylogenetic tree, E. coli O37:H10 was more diverse, and the serotypes were separated into several groups, implying numerous origins.

Plasmids containing AMR determinants in clinical strains are a significant global issue [23]. The genetic components, such as plasmid Col156 and Col(BS512) detection, identify plasmids containing the colicin E production, immunity, and lysis system [24]. The presence of incompatible plasmids, particularly the IncF plasmid, has been related to the worldwide rise of clinically relevant ESBLs and numerous AMR determinants [25]. The IncFII plasmid was identified in only one isolate, however, the IncB/O/K/Z plasmid was identified in most of the S. sonnei isolates. The most common plasmid type identified in E. coli O37:H10 isolates was IncFII. The ESBL genes are carried via various plasmids (IncFII, IncB/O/K/Z) or even inserted into the chromosome and encode resistance to third-generation cephalosporin [26].

In contrast to the genetic results, most S. sonnei isolates showed higher phenotypic AMR to the antimicrobials tested, as documented in the earlier work [14]. The detected E. coli O37:H10 isolates were resistant to amoxicillin, cotrimoxazole, and other penicillins, with strong genotype-phenotype correlations. However, phenotypically, most isolates were more resistant to antimicrobials than the genotyping results indicated. Resistance genes are frequently plasmid-mediated, and alterations may occur due to the loss of certain plasmids during storage [27]. Phenotypic AMR without genes reveals that it is mediated by different mechanisms [28]. The most concordant results were observed for beta-lactamase combinations, particularly amoxicillin/ampicillin in E. coli O37:H10, suggesting that high levels of beta-lactamase production contribute to the observed high-level phenotypic AMR [29]. The prevalence of ESBL-producing Enterobacteriaceae in healthy individuals varies between 3% in Western European nations to 60% in India [30]. The blaCTX-M-15 gene is the most extensively distributed gene that encodes ESBLs linked to human infections [31]. In other research, this gene was the dominant genotype among ESBL-positive isolates [32, 33]. However, in the current study, all isolates containing blaEC-15 were identified in E. coli O37:H10. The discrepancy might be the difference in O: H serotype combinations between the studies [34]. The ESBL genes have been linked to multiple plasmid incompatibility groups, including IncF, and are commonly found on conjugative plasmids. Like the current study, IncF is a widely known plasmid type identified in E. coli and is thought to be a significant reservoir for this plasmid [35]. Furthermore, the plasmid types IncB/O/K/Z and Col(BS512) were identified as comparable to the study conducted on domestic animals and children in Ecuador [36].

Plasmids are key carriers for transporting MGEs and AMR genes. The virulence genes distributions in the S. sonnei and E. coli O37:H10 isolates were comparable in this study. Bunduki and colleagues [37] discovered numerous pathogenicity-associated islands containing the fimH gene, demonstrating their importance. Another study revealed the presence of virulence genes such as iha, lpfA, yeh(A-D), and gad for adhesion to host cells [38]. In the current study, we identified IS elements in the sequenced genomes of S. sonnei and E. coli O37:H10 isolates. The diversity of these IS components was shown to be influenced by the sequences’ type and duration. This variety reveals strains’ genetic complexity and adaptability, as IS elements play an important role in genome plasticity and evolution. The presentation of various IS components may contribute to strain-specific features and pathogenicity, emphasizing the necessity of future research into their functional and evolutionary implications. These findings suggest a possible role of IS elements and composite transposons in the mobilization and dissemination of resistance genes.

Strengths and limitations

The study was health facility-based and might not fully represent the broader community. Sero-grouping was only conducted using polyvalent antisera. Due to limited resources, WGS analysis was performed on only 37 isolates, which may not capture the full diversity of all isolates. However, this study provides valuable insights by evaluating both the phenotypic and genotypic data, including serotype and AMR profiles, plasmid content, virulence factors, and phylogenetic relatedness. These findings contribute to a better understanding of Shigella species and E. coli O37:H10, and serve as a useful reference for researchers.

Conclusion

The study found all S. sonnei isolates carried blaEC-8 and blaZEG-1 resistance genes, while E. coli O37:H10 strains showed resistance to common antibiotics and harbored blaEC-15 and blaMIX genes. Plasmid types differed between species (Col156 in all S. sonnei vs. IncFII in most E. coli). Genomic analysis revealed close clustering of S. sonnei strains but earlier divergence in E. coli, suggesting different evolutionary paths. The study highlighted significant discrepancies between phenotypic and genotypic results. Therefore, a WGS-based study is essential to identify strains and associated genes, improve treatment effectiveness, and guide targeted interventions in Ethiopia.

Author contributions

B.A. performed data acquisition, investigation, data analysis, and interpretation of results, and wrote the manuscript. T.S.T. and Z.M. contributed to the methodology, validation, and review of the manuscript. D. H.A., K.M., and Sh. M. participated in the investigation and validation. M.F.N. and A.A. contributed to data curation and formal analysis. A.M. and G.B. contributed equally to methodology, resources, validation, and supervision. All authors revised and approved the manuscript for publication.

Funding

No external funding for this study.

Data availability

All the necessary data are included in the manuscript. The datasets used and analyzed in this study are publicly available in the NCBI SRA database under accession number PRJNA1291950.

Declarations

Ethics approval

The study was conducted following the Declaration of Helsinki, and ethical approval was obtained from the Institutional Review Board (IRB) of the Jimma University Institute of Health (IHRPRD/1092/2021). Consent was also obtained from the study sites.

Consent to participate

The mothers/caregivers of the children provided informed, voluntary, written, and signed consent after the study’s purpose was explained in full to them. Parents/caregivers of the children were also informed about the confidentiality of the information gained and their absolute right to refuse or withdraw from participation in the research.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.