Potentially pathogenic Escherichia coli from household water in peri-urban Ibadan, Nigeria

Clean water and safe sanitation are prerequisites for good health but many people do not have these basic needs (World Health Organization/United Nation Children Education Fund 2006). In populations with a limited access to municipal water supply, use of alternative water sources such as wells and boreholes for domestic purposes is common (Aboh et al. 2015). Generally, these groundwater sources are among the most vital finite natural sources of fresh water on earth for drinking and presumed to be free from contamination (Avtar et al. 2019). However, due to watershed erosion, drainage from sewage, run-offs and non-enforcement of guidelines for groundwater exploration or well construction, bacterial contamination of groundwater sources does occur (Eduvie et al. 2003). These water sources place people at risk of water-borne diseases like diarrhea and typhoid (Kaplan et al. 2017). Consequently, around 6–8 million people die each year due to water-transmitted diseases (United Nation Water 2013). To estimate the scale of the problem and prevent diarrheal disease morbidity and mortality, it is necessary to investigate and monitor the microbiological quality of drinking water (Momtaz et al. 2013). Fecal contamination of water meant for human use constitutes a threat to public health since such water can be a vehicle to transmit diseases such as infectious diarrhea and typhoid.

Most  Escherichia coli are commensals but some strains are pathogenic. Pathogenic strains are subdivided into intestinal pathogenic E. coli (InPEC), also known as diarrheagenic E. coli (DEC) and extra-intestinal pathogenic E. coli (ExPEC). The six well characterized InPEC pathotypes are Shiga toxin-producing E. coli (STEC), enteropathogenic E. coli (EPEC), enterotoxigenic E. coli (ETEC), enteroaggregative E. coli (EAEC), diffusely adherent E. coli (DAEC) and enteroinvasive E. coli (EIEC). These pathotypes were so categorized based on virulence genes and gastrointestinal pathogenesis mechanisms (Nataro & Kaper 1998; Kaper et al. 2004). ExPEC are classified based on the infection type into four groups: neonatal meningitis E. coli (NMEC), sepsis-associated E. coli (SEPEC), uropathogenic E. coli (UPEC) and avian pathogenic E. coli (APEC) (Riley 2014).

Urban Ibadan households largely rely on wells, springs and boreholes (Adesoji & Ogunjobi 2016) due to inability of the public water supply to meet population needs. Diarrheal disease and enteric fever incidence are not evenly distributed across the city and this has been attributed to differences in water safety and its microbiological quality, in particular (Popoola et al. 2019). Very little information is, however, available from the peri-urban areas. For instance, Akinyele, a peri-urban local government area (LGA) is not covered by the public water distribution network and for the most part, residents of the neighboring Lagelu LGA (LLGA) do not enjoy regular municipal water supply. Previous studies conducted in these areas on flooding, management of wells and drinking water assessment indicated poor water quality. However, these studies were performed through questionnaires or single sample analyses (Ince et al. 2010; Agbola et al. 2012) thus necessitating a longitudinal and laboratory-based investigation of the water sources.

With seasonal exacerbation of risk, we hypothesized that peri-urban household water sources constitute health risk for diarrhea and other enteric infections. Thus, we aimed to longitudinally identify enteric bacteria as a function of bacteriological quality of these sources. In addition, we also aimed to use the whole-genome sequencing (WGS) data to assess the E. coli virulence profiles since fewer studies have predicted pathotypes of E. coli recovered from household water sources.

The total sample size (n) was estimated at 271 determined by the following equation

X = Z (c/100)2r (100–r); n = NX/((N−1) E2 + X); E = √{(N−n) X/n (N−1)} using Raosoft (2004) software, with the following pre-estimated parameters, namely Error margin E, 5%; confidence level 90%; response distribution r, 50%; total estimated population, N, 505, 100 (National Population Commission 2015). Major assumptions are normal distribution and sample size (n) > 30. The study was planned longitudinally to collect household water samples over four seasons – dry and wet – for 2 years. However, three samples were collected each from 96 household water sources – well and borehole from January 2019 to March 2020. The intention to collect a fourth sample was thwarted by movement restrictions due to the COVID-19 pandemic. There is relative abundance of well over borehole water sources, thus, the well to borehole water sampling ratio was 2:1 and all samples were Global Positioning System (GPS)-located. The water samples were collected as previously described (Akinyemi et al. 2006), transported to the laboratory in a cool box and processed for microbiological analysis within 4 h of collection.

Estimates of Ibadan monthly rainfall from January to December (Ibadan Rainfall 2019) were sourced secondarily from the Nigeria Meteorological Agency (NiMet), which is publicly available (https://www.weather-nga.com/en/nigeria/ibadan-weather-january). These estimates were graphed to generate a rainfall time series for the city of Ibadan.

The THC, TCC and TEC were obtained using recommended standard pour plate method (Wohlsen et al. 2006), with modifications. Briefly, THC and TCC were obtained on Nutrient agar (Oxoid) and MacConkey agar (Oxoid), respectively. In order to detect E. coli in the water samples, 1,000 μL of undiluted water was pour-plated with Eosine Methylene Blue (EMB) agar. The plates were observed for growth after incubation at 37 °C for 24 h. The plates were incubated overnight at 37 °C and observed for growth. Characteristic colonies were sub-cultured, after which the isolates were further identified by colonial, morphological and standard biochemical tests. Green metallic sheen on EMB agar was used for E. coli pre-estimation. The E. coli isolates were stored in Luria Bertani-Glycerol (LB Glycerol) broth and frozen at −80 °C until further analysis.

The E. coli isolates were cultured in Luria Bertani broth on a shaker for 24 h at 37 °C. The genomic DNA (gDNA) was extracted using Wizard^® Genomic DNA Purification Kits (Promega) following the manufacturer's procedure.

PCR was performed on extracted gDNA to detect sequences specific for E. coli housekeeping uidA (β-D glucuronidase) gene (Divyashree et al. 2015) with the forward UAL-754 5′AAAACGGCAAGAAAAAGCAG 3′ and reverse primers UAR-900 5′CGCGTGGTTACAGTCTTGCG 3′ at 146 bp (base pairs). Reactions were performed in 20 μL total reaction volumes containing 10 μL One Taq Quick-Load 2 × Master Mix with Standard Buffer (New England, BioLabs). Amplification was conducted in a programmable thermal cycler (Applied Biosystems Veriti 96 Well Thermal Cycler). The reaction was carried out using 1.0 μL of DNA template, 0.1 μL of each forward and reverse primers, and 8.8 μL nuclease-free water. The PCR program was optimized to consist of an initial denaturation at 94 °C for 5 min followed by 30 cycles at 94 °C for 30 s, annealing at 60 °C for 30 s and extension at 72 °C for 30 s with final extension at 72 °C for 10 min. E. coli ATCC 25922 was used as positive control while DNase-free molecular grade water was used as a negative control.

All the PCR products were electrophoresed on 1.5% agarose (w/v) gel at 150 V, 277 mA for 50 min. Amplicon size was estimated with a 100 bp molecular weight marker (Invitrogen) and the products were visualized using an ultraviolet trans-illuminator (UVP GelMax Imager).

To determine the pathotypes of the 25 E. coli, and their inter-relationships, if any, WGS was carried out using Illumina technology. The gDNA standardized to 0.2 ng/μL was submitted to paired-end WGS using MiSeq sequencer (Illumina, San Diego, CA, USA). Genomic libraries were prepared using the Nextera XT kit (Illumina, CA, DNA). The gDNA was fragmented and tagged with sequencing adapters in a single step using Nextera transposome (Nextera XT DNA Library Preparation Kit, Illumina, San Diego, USA). The fragmented and tagged gDNA was amplified (PCR 12-cycle amplification) and cleaned up with AMPure beads. Nextera libraries were measured using Qubit and the size profile was analyzed on 2100 Bio-analyzer using high-sensitivity DNA assay kit (Agilent Technologies, Waldbronn, Germany). Libraries selected for sequencing were normalized to 1 nM and pooled. The 1 nM pooled library was denatured and diluted prior to loading on a MiSeq paired-end 2 × 150 bp (MiSeq reagent kit V2 (300 cycles)) or 2 × 300 bp (MiSeq reagent kit V3 (600 cycles)) sequence run.

Following from WGS, raw sequences were submitted to fastqc and multiqc to assess and aggregate phred scores and cytosine to Guanine (CG) content. The contigs were then annotated using Prokka v1.12 with contigs <200 bp and coverage <10-fold excluded from the analyses. The assemblies were checked for quality strictures while the assembled reads were aligned to E. coli-042 reference genome to create an alignment file using Bcftools, Samtools, Gatk, Bwa, Smalt and Picard. The single nucleotide polymorphism (SNP) was performed on the alignment file using SNP sites to produce the SNP file. The SNP file was used to construct phylogenetic tree using IQ tree tool and the tree was viewed with interactive tree of life (iTOL). The species were determined using Pathogenwatch resource available at https://pathogen.watch/upload/fasta. Genomes downloaded from the European Nucleotide Archive (ENA) database and incorporated into phylogenetic tree are SRR6362403, a soil environment E. coli – Minnesota, USA (North Dakota State University 2015); ERR987377, an enteroaggregative E. coli clinical isolate in Nigeria (Dougan et al. 2018); SRR11654221, a food animal E. coli – USA (United States Department of Agriculture, Food Safety and Inspection Service 2020) and FN554766, an EAEC_042 – Human E. coli isolate (Chaudhuri et al. 2010).

Virulence-associated genes (VAGs) were identified by the center for genomic epidemiology resources tool VirulenceFinder https://cge.cbs.dtu.dk/services/VirulenceFinder/ that includes 38 ExPEC-associated virulence genes (1,890 alleles) recently added to the existing E. coli VirulenceFinder database (Tetzschner et al. 2020). Phylogroups were identified through the ClermonTyper with fasta files uploaded to http://clermontyping.iame-research.center/. Multilocus sequence typing (MLST) was performed by the Achtman MLST scheme that uses seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA and recA) through the website https://cge.cbs.dtu.dk/services/MLST/.

In order to assign ST number to novel STs, the assembled sequences (fasta) of the isolates were first uploaded to the Centre for Genomic Epidemiology platform at MLST – Typing Results (dtu.dk) using MLST-2.0 Server where allele numbers were given. Then, the raw sequences (fastq) were uploaded to Enterobase – Search ecoli (warwick.ac.uk) and allele numbers inputed to yield ST number.

Data were subjected to Pearson correlation and paired sample t-test (IBM SPSS 20 version) to show relationship among the microbiological parameters at levels of 5 and 1%. Biological characteristics of the source water over the seasons were compared using Fischer's exact test.

Of the 96 household water sources sampled each season from Akinyele and Lagelu communities, 66 samples were collected from wells while 30 were collected from boreholes. The number of wells with THC ≥500 CFU/mL and TCC ≥10 CFU/mL in wet (THC = 63, TCC = 60) and dry seasons (THC = 57, TCC = 59) was comparable (Supplementary Material, file S1). Five and 12 boreholes, respectively, had THC ≥500 CFU/mL and TCC ≥10 CFU/mL in wet season. No borehole had THC ≥500 CFU/mL but 10 boreholes had TCC ≥10 CFU/mL in dry seasons (Table 1). Thirteen (19.70%) of the 66 wells were contaminated with E. coli during the wet season. However, borehole samples did not have E. coli contamination in either season (Table 1). The water sources showed season-independent heterotrophic contamination, THC; P = 0.127 (well) and P = 0.052 (borehole), coliform contamination, TCC (P = 0.492 (well) and P = 0.789 (borehole)). The E. coli contamination, TEC, for well water sources was found to be significant (P = 0.0001) across the seasons (Table 1).

The number of samples collected and submitted to total bacteria, coliforms and E. coli testing per months were higher in dry seasons than the wet season (Supplementary Material, files 2S1, 2S2 and 2S3). This is because dry samples were collected in 3 months from January through March while wet samples were collected from April to September (6 months). In order to give a clearer picture of how seasonal variation especially rainfall drives contamination of water sources, the graphed annual rainfall data from January to December were aligned with the total heterotrophic, coliform and E. coli time series and seasonality for wells and boreholes. The time series and seasonality plot for total bacteria and coliforms showed two peaks in April/May and around July to September. The result showed the fluid impact of rainfall on groundwater contamination. The higher total heterotrophic and coliform counts correspond to the wet season as shown in June and September rainfall peaks. However, E. coli time series and seasonality plot specifically showed peaks in April, June and September. The use of the mean data as point data for each location (Supplementary Material, files 3S1 and 3S2) to create maps (total bacteria, coliforms and E. coli) showed fluctuations depicted in a south to north and east to west gradient in total bacteria and coliforms. The area with high E. coli contamination with 700–950 CFU/mL are localized to northern part (Ikereku) and south-eastern locales (Lalupon) of the study area. For these two locations, the total coliform maps where coliforms were greater than or equal to 2,000 CFU/mL is consistent with E. coli contamination (Figure 5).

The 25 E. coli obtained in this study belonged to three phylogroups. Eleven (44%) belonged to phylogroup A, 13 (52%) to group B1, while 1 (4%) belonged to group D. The isolates of this study, namely 5EC, 108EC, 172EC and the E. coli genome with the accession number ERR987377 obtained from European Nucleotide Archive, ENA belonged to ST10. Other ENA strains – FN554766, SRR6362403, SRR11654221 belonged to ST414, ST9005 and ST11, respectively. The 13EC, 14EC, 15EC and 16EC belonged to ST2541, whereas 171EC, 172EC1, 173EC and 175EC belonged to ST2008. ST156 included 69EC, 70EC and 70EC1 strains, ST48 included 166EC and 184EC strains but 167EC and 167EC1 belonged to a novel ST13028. Other strains belonged to the following STs-58, 69, 155, 196, 4981, 11400 and 9987*.

Analysis of the VAGs showed that majority of the strains did not carry virulence genes usually possessed by diarrheagenic pathotypes. The total number of different VAGs harbored by the strains were 33. The E. coli strains possessed an average of six virulence genes (range 3–16) per isolate. ExPEC classification requires possession of the following virulence genes – papA/papC, sfa/foc, Afa/dra, kpsM II, ompT and iutA, whereas EAEC is inclusive of eilA, chuA, fyuA and ompT genes. Among other VAGs, APEC subpathotype possesses iutA, iss, cvaC, IroN and ompT genes. Based on the virulence gene repertoire, three (12%) were potential ExPEC (Johnson et al. 2003), EAEC (Chaudhuri et al. 2010) or APEC (Collingwood et al. 2014), the rest were classified as likely commensals. With the exception of three isolates (191EC, 166EC, 184EC), all the strains possessed an average of 1–3 adhesin genes (afaD, afaA, iha, hra and lpfA). The strain 172EC possessed AfaA/D typical of diffusely adherent E. coli. Among the adhesin genes were iha, EHEC Novel non-hemagglutinin adhesin from O157:H7 (Tarr et al. 2000) and hra2, heat-resistant agglutinin gene that encodes integral outer membrane hemagglutinin (Mancini et al. 2011) possessed by 172EC strain that met the ExPEC classification. The long polar fimbrial structural subunit gene lpfA was contained in 18 (72%) of the E. coli genomes.

Two isolates carried toxin-related genes such as Shigella enterotoxin B gene (senB), astA, hemolysin, colicin and/or microcin (hlyF−, cia, cba, cib, cvaC and cma). The isolate 191EC carried hlyF-, cia, cba, cib, cvaC and cma, whereas 172EC carried senB, astA and hlyF toxin-associated virulence genes, though senB and astA are also seen in DEC. An average of two genes responsible for bacteria iron metabolism such as iron acquisition, yersiniabactin biosynthesis and aerobactins genes; iroN, irp2, fyuA, chuA, iutA, iucC– and traT– were found in nine E. coli strains. The irp2 and fyuA genes that encode yersiniabactin biosynthesis and receptor protein from Yersinia Pathogenic Island (PAI) were possessed by seven isolates as did surface exclusion, traT genes. The capsular genes kpsE and kpsMIII were found in two strains; one (154EC) carried kpsE only while the other (183EC) carried both.

Additional ExPEC-associated virulence factors include glutamate decarboxylase gad, an acid survival factor and tellurite ion resistance terC genes that were possessed by all the strains. Thirteen of these isolates were positive to iron acquisition or transport system sitA, which encodes periplasmic-binding protein, whereas four of the latter possessed plasmid-borne type II secretion system genes, etpD seen in ETEC. Enteroaggregative immunoglobulin repeat protein air and pAA plasmid-encoded eilA, a HilA-like regulator EAEC was carried by 183EC classified as EAEC. Others were outer membrane protein ompT and increased serum survival iss genes respectively found in six and nine strains. Efflux transporter gene etsC and EAEC capU were possessed by one isolate respectively (Figure 6).

VAG profiles were used to classify the E. coli strains into pathotypes. One E. coli strain each was classified as ExPEC (172EC), EAEC (1823EC) and APEC (191EC). E. coli strains not classifiable into any pathotype were regarded as commensals. The Microreact visualization of the virulence genes and metadata are available at https://microreact.org/project/hGKMbAnF2REcU7VZv3SUgc-whole-genome-sequence-analysis-of-e-coli-obtained-from-household-well-water-sources-in-peri-urban-ibadan-nigeria.

SNP was used to investigate the point-source contamination of the well water sources which also infers the diversity and clonality of the E. coli strains in the 13 household water sources where the isolates were recovered. Across the households in the same municipal area (AELCDA); the strains 171EC, 173EC and 175EC were clonal (SNP = 0) even though 171EC and 173EC strains were found in household 8 as against 175EC strain found in household 10. Within the households, four (4) strains (13EC, 14EC, 15EC and 16EC) in the same sampling found in household 2 in the ASLCDA municipal area slightly diverged from one another by two to four nucleotides.

In household 7, the clonality of isolates (69EC and 70EC) recovered (SNP = 0) were similar to what occurred in household 8 where 171EC and 173EC were isolated. The third strain (172EC) also obtained in household 8 had wide genome difference (SNP = 26,816) from 171EC and (SNP = 26,757) from 173EC. In household 9, the three strains (182EC, 183EC and 184EC) obtained from the household well water sources were non-clonal as they had wide nucleotide difference that ranged from 10,487 to 47,637. Strains that were non-clonal (n = 2) found in household 11 also had extensive genome difference (SNP = 13,679). The strains recovered in households 7, 8, 9 and 11 were, however, isolated within the same municipal area (AELCDA) (Figure 6, Supplementary Material files 4S1 and 4S2).

The increasing stress on available water resources worldwide and potential for disease spread require sentinel surveillance of water systems (Invik et al. 2017). It is also important to understand temporal variations in water contamination to effectively predict possible outbreaks, delineate high-risk periods (Hay et al. 2013; Galway et al. 2015) and raise evidence for clean water provision policy. Natural water sources are predisposed to chemical and bacteriological contaminants principally from animal and human waste (Nigerian Standard for Drinking Water Quality 2015) such as pit latrines, domestic waste and leakages from landfills that permeate peri-urban landscapes.

In Nigeria, water intended for domestic use is expected to be free from coliforms above 10 CFU/mL as such water is considered fecally contaminated (NSDWQ 2015). This study found that peri-urban well and borehole from Akinyele and Lagelu households were highly contaminated with bacteria and coliforms. The findings of this study are similar to reports from previous studies in Ibadan (Adesoji & Ogunjobi 2016), as well as other Nigerian cities such as Lagos (Akinyemi et al. 2006) and Zaria (Aboh et al. 2015). Furthermore, water intended for drinking should contain no detectable E. coli per 100 mL sample (WHO 2011; NSDWQ 2015). Standard pour plate and Petri film methods are known to give better and more consistent results than membrane filtration, defined substrate technology, Colilert-18, Quanti-Tray/2000 system and the most probable number (MPN) techniques which produced low recovery or significant variability between replicates in side-by-side evaluations (Wohlsen et al. 2006). In this study, we therefore used standard pour plate method and this yielded higher total E. coli counts of 4,250 CFU/100 mL than 16 CFU/100 mL reported elsewhere in Nigeria (Aboh et al. 2015).

Our observation that 20 (66.7%) of sampled boreholes had total coliform counts below 10 CFU/mL and no detectable E. coli (0 CFU/100 mL) demonstrates that this source was superior to wells, for which only seven (10.6%) were ever found to be potable. This is similar to what was reported in Limpopo, South Africa where Molekoa et al. (2019) found more than 80% of borehole water samples fit for drinking and domestic purposes. Boreholes are expensive to install and therefore beyond the reach of many households and communities in Ibadan. However, the poor quality of well water recorded and the high risk of contracting typhoid in Ibadan (Popoola et al. 2019) suggests that these safer water sources may be more cost-effective overall. As the boreholes did not, however, yield unanimously potable water, even boreholes cannot be presumed to be good sources of water, until assessments have been conducted.

Maps and time series provide more granular information on how temporality and geography affect groundwater contamination. For spatial analysis, we used GPS coordinates which permitted good resolution and visualization of the study area showing high-risk locales (Invik et al. 2017). E. coli contamination of wells in April, June and September corresponds with onset and abundance of rainfall as shown in the June and September rainfall peak (Figures 2–4) as similar observation was not recorded in the preceding and succeeding dry seasons. E. coli-positive wells have been previously reported in Nigeria (Aboh et al. 2015; Adesoji & Ogunjobi 2016), our study reveals that E. coli contamination of wells in Lagelu (Lalupon) and Akinyele (Ikereku) regions of Ibadan are driven by rainfall (Figure 4). Risk could also be worsened by factors that we did not track such as well shallowness, open defecation and agricultural practices that include use of animal manure in farmlands. For future surveillance in the general area, these maps (Figure 5) can be used to anticipate hot spots of microbial contamination in groundwater (Invik et al. 2017).

E. coli presence in water is a marker of fecal contamination, highlighting risk that harder-to-detect pathogens may be present in the water. However, some E. coli are pathogenic and therefore pose a risk in themselves. We examined the lineages and virulence factors of E. coli isolates from this study to determine whether they are potential pathogens. We used WGS because it offers the best resolution for comparison of pathogen genomes and exhaustively identifying virulence factors. Based on phylogrouping, E. coli is classified into eight groups – A, B1, B2, C, D, E, F or cryptic clade I (Clermont et al. 2000). The E. coli recovered in this study belong to phylogroup A, B1 and D classified as commensals (A and B1), pathogenic E. coli (A, B1 or D) and ExPEC (group D) (Clermont et al. 2000). Phylogroup B1 is commonly adapted to water environments and it predominated in the E. coli population in this study, i.e. B1 (52%), A (44%) and D (4%). This is similar to the observation of Nowicki et al. (2021) that used WGS to examine the association of E. coli with recent fecal contamination in Kenyan rural drinking water.

The MLST provides finer subtyping for E. coli than broad phylotyping as it gives clearer understanding of the spread and diversity of strains. Three ST10 isolates were recovered from different zones of the sampling region. The ST10 is a large and diverse clonal complex that includes human isolates and is the most common commensal detected in Africa, Asia, Europe and South-America (Rashid et al. 2015; Matamoros et al. 2017). The current study detected ST10 isolates in Moniya (172EC), Lalupon (108EC) and Arulogun (70EC) communities separated by not less than 20 km apart. This observation is in agreement with the dominance of ST10 lineages within the same setting, though among enteroaggregative E. coli especially from healthy individuals (Okeke et al. 2010). The predominant STs identified in this study were ST2541 and ST2008 as both constituted 32% of the isolates, including two near-identical ST2541 strains (1 SNP distance) that were isolated from the same well water point in the same sampling even though this ST is yet to be reported from water sources. One of the four ST2008 isolates, ID/ST, 172EC1/ST2008 has a wide, 2452, SNP distance from others. However, the other three ST2008 were clonal (SNP distance = 0) even though they were recovered from two different households (above 5 km apart) within the same municipal area (AELCDA) in the same season. It is significant to highlight that the current study reports a novel ST13028 yet to be reported in the literature.

One EAEC isolate was detected in the study and belonged to ST69, which like the E. coli_042 ST414 is in Phylogroup D (Okeke et al. 2010). The EAEC-ST69 isolate carried a similar virulence gene repertoire with Peru EAEC/E. coli_042 strain (Chaudhuri et al. 2010) that causes diarrhea. The ST48, also detected, has been reported from drinking water in France (Madec et al. 2016) as ST69, ST156 and ST10 detected in this study have been found in river water sources (Colomer-Lluch et al. 2013; Rashid et al. 2015). The ST10, ST155 and ST156 E. coli lineages are known to be linked to human E. coli isolate (Matamoros et al. 2017; Zahra et al. 2018) and this indicates cross-niche detection. Similarly, the only ExPEC isolate detected in this study that belonged to phylogroup A are known to cause infections in anatomical sites and sepsis (Nataro & Kaper 1998).

Lugli et al. (2022) recently reported tap water in Italy as a reservoir of bacteria that can colonize human intestine, shape gut microbiota and enable diverse metabolic processes attributed to human health. In the low and middle income countries without municipal water services, the microbiological quality of informal drinking water sources has been less thoroughly investigated (Kaplan et al. 2017; Nowicki et al. 2021). However, the few studies performed have yielded bacteria with pathogenic potential, including Salmonella enterica serovar Typhi, S. Paratyphi A and E. coli, although the virulence of E. coli isolates is rarely investigated. By evidencing high total and coliform counts and recovery of potentially pathogenic E. coli, this study adds to the evidence pointing to unsafe household water in Ibadan, and its potential for exacerbating feco-oral disease transmission.

Peri-urban Ibadan populations that are dependent on wells and boreholes with limited connection to municipal water works in Ibadan, investigated in this study do not have access to safe water as their water sources are largely contaminated with coliforms beyond the acceptable limits. Seasonal variations worsen the quality of the groundwater sources in the wet season. Overall, borehole water sources appeared to be safer, but not uniformly so. There is also a need to further assess these sources for other quality parameters before declaring the few sources without coliforms and E. coli as safe. E. coli recovery from these water sources include potential pathogens such as ExPEC, EAEC and APEC that pose a direct threat of disease spread.

This work is supported by the African Research Leader's Award, awarded to O.I.N. funded by the UK Medical Research Council (MRC) and the UK Department for International Development under the MRC/DFID Concordant agreement and is also part of the EDCTP2 program supported by the European Union. I.N.O. is a Calestous Juma Science Leadership Fellow supported by the Bill and Melinda Gates Foundation. The funders had no role in the content, crafting or submission of this paper.

O.I.F., O.E.F., A.G.R. and I.N.O. conceived and designed the experiments. A.G.R. performed the experiments. O.I.F., I.N.O. and A.G.R. did literature search, and contributed reagents and materials. O.I.F., I.N.O., R.A.D. and A.G.R. analyzed the data and contributed analysis tools. A.G.R. wrote the first draft. All authors approved the final manuscript.

The authors have declared that no competing interests exist.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.