Characterization of Escherichia coli from Water and Food Sold on the Streets of Maputo: Molecular Typing, Virulence Genes, and Antibiotic Resistance

: The aim of this study was to investigate the pathogenic potential and antibiotic resistance of 59 Escherichia coli isolates from ready-to-eat (RTE) street food ( n = 31) and drinking water ( n = 28) sold in the city of Maputo, Mozambique. The isolates were characterized by XbaI subtyping analysis via pulsed ﬁeld gel electrophoresis. Multiplex PCRs were performed targeting ﬁve virulence genes (stx, lt, st, astA, and eae) and three groups of antibiotic-resistant genes, namely ß-lactamases ( extended-spectrum ß-lactamase and AmpC ), tetracycline ( tetA , tetB , and tetM ) and sul-famethoxazole/trimethoprim ( sul1 , sul2 , and sul3 ). The stx virulence gene, encoding the Shiga / Vero (VT) toxin produced by the verotoxin-producing E. coli (VTEC), was identiﬁed with similar frequency in isolates from food (5/31) and water (6/28). The highest percentages of resistant isolates from food and water were found for ß-lactams imipenem (35.5 and 39.3%, respectively) and ampicillin (39.3 and 46.4%, respectively). Multidrug resistance was observed in 31.3% of the isolates, being higher in E. coli isolates from water (45.5%) compared to RTE street food isolates (19.2%). Virulence genes were detected in 73% of the multidrug-resistant isolates. Concerning antibiotic-resistant genes, ESBL was the most frequent (57.7%) among β -lactamases while tetA was the most frequent (50%) among non- β -lactamases.


Introduction
In developing countries, enteric infections and diarrheal diseases represent some of the main causes of hospitalization and a serious health problem [1,2]. Their occurrence is closely associated with deficiencies in the water supply and sanitation systems and poor personal hygiene conditions, which are common in developing countries. These diarrheal diseases can be caused by bacteria, parasites, and viruses [3,4]. The exposure to these pathogens in sub-Saharan Africa is favored by the climate and is mainly related to the lack of adequate hygiene and sanitation conditions and the consumption of contaminated ready-to-eat (RTE) foods and vegetables [1,2,5].
The bacterial species E. coli represents one of the main hazards in food and water sources [3,6,7]. However, not all E. coli cause diarrheal diseases [8]. Its association with the occurrence of diarrhea is based on the presence of genes that encode different virulence factors such as stx, eae, and ehxA in the Shiga toxin-producing E. coli or enterohemorrhagic E. coli (STEC/EHEC), bfp (bundle-forming pilus), and eae (intimin) [8,9]. The presence and/or association of these virulence factors determine the E. coli pathotypes that cause

Material and Methods
Overall, 59 E. coli isolates from ready-to-eat street food and water were investigated. Of these, 31 food isolates were recovered from 83 street food samples and isolated according to the ISO 16649-2 [20], as described in Salamandane et al. [1]. The other 28 E. coli isolates were recovered according to ISO 9308-1 [21], from a total of 118 water samples consumed in the city of Maputo, including home-bottled street water, tap water, and water from supply wells, as described previously [22]. The evaluation of the microbiological quality of the samples indicated that in general, the samples were unsatisfactory for consumption. From each plate that presented well isolated colonies and characteristics of E. coli, a CFU was selected.

E. coli Isolates
Biochemical tests were carried out to characterize these presumptive E. coli isolates from chromogenic medium, namely Gram staining, catalase, and oxidase tests [1,22]. Subsequently, molecular identification was performed, targeting the 16S rRNA gene (401 bp) with the AB035924 forward primer (5 CCC CCT GGA CGA AGA CTG A-3 ) and the AB035924 reverse primer (5 -ACC GCT GGC AAC AAA GGA T-3 ), as previously reported [1,22]. Subsequently, the 59 isolates were characterized by antibiotic resistance profiling, PFGE subtyping, and the detection of virulence genes and antibiotic resistance genes as described below.

Pulsed Field Gel Electrophoresis (PFGE)
The 59 E. coli isolates were characterized by DNA macrorestriction analysis using PFGE in accordance with PulseNet International [23]. Briefly, bacterial isolates were inoculated onto Tryptone Soy Agar plates (Biokar Diagnostics, Beauvais, France) and incubated at 37 • C for 14 ± 2 h. From these plates, bacterial cultures were embedded in SeaKem Gold agarose (Bio-Rad Laboratories, Milan, Italy), lysed, washed, and the DNA was in situ digested with 50 U of XbaI (Thermo Fisher Scientific, Fermentas; Waltham, MA, USA) at 37 • C for 3 h. Resolution of the generated DNA fragments was obtained with 1% (m/v) SeaKem Gold Agarose gels in 0.5× Tris-Borate EDTA Buffer (Bio-Rad Laboratories, Milan, Italy) at 14 • C and 6 V/cm with time ramped for 6.7/35.3 s over 18 h, using a CHEF DRII System (Bio-Rad Laboratories, Milan, Italy); pattern images were acquired in the Gel Doc™ EZ System (Bio-Rad Laboratories, Milan, Italy). The TIF images were normalized by aligning the peaks of the XbaI DNA fragments of the size standard strain Salmonella enterica serovar Braenderup H9812 DNA, which was loaded onto two lanes in each gel.

Multiplex-PCR (MPCR) for Identification of Virulence and Antibiotic Resistance Genes
As PCR templates, six bacterial colonies grown on TSA plates for 18 ± 2 h at 37 • C were suspended in 300 µL of sterile ultrapure water and incubated in a boiling bath for 15 min [24]. Subsequently, the tubes were centrifuged at 16,000× g for 12 min, and the lysates supernatants (lysates) were stored at −20 • C. Amplification reactions were performed in a final volume of 25 µL, containing 12.5 µL of Taq DNA Polymerase NZYTaq II2× Colorless Master Mix (MZTech, Lisbon, Portugal) and 1 µL of 50-mM MgCl 2 .
The five targeted virulence genes coding for the respective virulence factors were stx (shiga/vero toxin, VT), lt (heat-labile enterotoxin, LT), st (heat-stable enterotoxin, ST), astA (heat-stable enterotoxin 1, EAST1), and eae (intimin). The respective five primer sets (Table 1) were added at a final concentration of 0.4 µM each. The PCR was performed with an initial denaturation step at 95 • C for 5 min, 35 cycles of 30 s at 95 • C, 40 s at 50 • C, and 72 • C for 1 min, with a final cycle at 72 • C for 10 min. For the screening of antibiotic resistance genes, 16 primer sets were used in four separate MPCR reactions. For the detection of extended-spectrum ß-lactamases (ESBL) genes, five pairs of primers (TEM variants, SHV variants, bla OXA variants, CTX-M Group 1 variants, and CTX-M Group 9) were selected. Five primer pairs (ACC variants, FOX variants, MOX variants, CIT variants, and DHA variants) were selected for the detection of AmpC β-lactamase genes. For the detection of tetracycline resistance genes, three primer pairs, tetA, tetB, and tetM, were selected. For the detection of sulfonamide resistance genes, the primers sul1, sul2, and sul3 were selected. The used primers and the respective sizes of the PCR products, as well as the final concentrations, are listed in Table 2.   For the detection of ESBL and AmpC β-lactamase genes, the reaction mixtures were subjected to the following amplification program: initial denaturation at 94 • C for 10 min, followed by 30 cycles at 94 • C for 40 s, 60 • C for 40 s, 72 • C for 1 min, and a final elongation step at 72 • C for 7 min. For tetracycline resistance genes, the reaction mixtures were subjected to the following amplification program: initial denaturation at 94 • C for 5 min, followed by 35 cycles at 94 • C for 60 s, 55 • C for 60 s, 72 • C for 1.5 min, and a final elongation step at 72 • C for 7 min. For sulfonamide resistance genes, the reaction mixtures were subjected to the following amplification program: initial denaturation at 94 • C for 5 min, followed by 30 cycles at 94 • C for 15 s, 69 • C for 30 s, 72 • C for 1 min, and a final elongation step at 72 • C for 7 min.
All PCR reactions were run in a thermocycler GeneAmp ® PCR System 9700; Applied Biosystems (Bio-Rad Laboratories, Segrate, Milan, Italy). The resulting PCR products were resolved on 2% (m/v) agarose gels in 1xTAE buffer in an EC330 Thermo Fisher Scientific tank (Georgia, USA) at 6 V/cm for 90 min. The gels were then stained with GelRed (Frilabo, Maia, Portugal) and analyzed using the Gel Doc™ EZ System (Bio-Rad Laboratories, Segrate, Milan, Italy). For calculating the size of the PCR products, the molecular marker 100-bp DNA Ladder (InVitrogen, Carlsbad, CA, USA) was used.

Data Analysis and Interpretation
The PFGE patterns were analyzed with BioNumerics ® version 6.6 (Applied Maths, Belgium). The levels of similarity were based on the Dice correlation coefficient, with tolerance and optimization values of 1 and 0.5%, respectively. For cluster analysis of the patterns, the unweighted-pair group matching algorithm (UPGMA) was used. Ten isolates of E. coli were used to determine the experimental variation between duplicate experiments. The minimum level of repeatability of the macrorestriction conditions was calculated by running DNA samples from duplicated restrictions of the DNA of each strain. Based on the obtained results, a 95% cut-off value of similarity was established for typing identical isolates with identical outputs.
For the evaluation of antibiotic resistance, the diameters of the inhibition halos (mm) were measured and compared to those described by the CLSI, 2021 [31]. The isolates were considered non-susceptible to a certain antibiotic when they showed intermediate or full resistance to that antibiotic. Multidrug resistance was considered to be resistant to more than two unrelated antimicrobial agents.

Pulse Typing and the Presence of Virulence Genes
The PFGE of the DNAs of the 59 isolates digested with XbaI showed 12-21 fragments, ranging from approximately 20.5 to 1135 kb in size. The relationships among E. coli isolates based on their XbaI PFGE profiles are shown in the dendrogram in Figure 1. Based on data obtained from the duplicate experiments, a cut-off level of 95% similarity was established for the definition of a PFGE pattern. The PFGE of the XbaI macrorestriction fragments differentiated the 59 isolates into 51 different patterns, showing a great diversity among the isolates regardless of their food, water, or municipal district origin. These results suggest a great diversity of sources of contamination.

Pulse Typing and the Presence of Virulence Genes
The PFGE of the DNAs of the 59 isolates digested with XbaI showed 12-21 fragments, ranging from approximately 20.5 to 1135 kb in size. The relationships among E. coli isolates based on their XbaI PFGE profiles are shown in the dendrogram in Figure 1. Based on data obtained from the duplicate experiments, a cut-off level of 95% similarity was established for the definition of a PFGE pattern. The PFGE of the XbaI macrorestriction fragments differentiated the 59 isolates into 51 different patterns, showing a great diversity among the isolates regardless of their food, water, or municipal district origin. These results suggest a great diversity of sources of contamination. coli XbaI PFGE patterns for 59 isolates. BioNumerics ® version 6.6 was used with a band position tolerance of 1% and an optimization of 0.5%. The reference of the isolates as well as their municipal district, origin, antibiotic resistance, and presence of virulence genes are indicated. Red are isolates from drinking water; green are isolates from street food; Non = the virulence genes searched were not detected; non = susceptible for all antibiotic detected; Mult = multi drug resistance. coli XbaI PFGE patterns for 59 isolates. BioNumerics ® version 6.6 was used with a band position tolerance of 1% and an optimization of 0.5%. The reference of the isolates as well as their municipal district, origin, antibiotic resistance, and presence of virulence genes are indicated. Red are isolates from drinking water; green are isolates from street food; Non = the virulence genes searched were not detected; non = susceptible for all antibiotic detected; Mult = multi drug resistance.
However, at 85% similarity, there were some clusters with a few isolates from street food from adjoining municipal districts, although the isolates in these clusters may differ regarding the presence of virulence genes and antibiotic resistance profiles.
In water, the most predominant genes were stx (35.3%), st (35.3%), lt (11.8%), astA (11.8%), and eae (5.9%). In one water isolate, the combination of stx (VT), lt (LT), and eae (intimin) was observed. The percentage of isolates that did not present any of the searched virulence genes was 16.1% in those recovered from RTE street food and 39.3% in those recovered from water ( Figure 1).

Antibiotic Resistance Genes
A total of 116 fragments genes were amplified from 16 variants of antibiotic resistance genes screened in the 48 non-susceptible E. coli isolates. In food isolates, 59 fragments of antibiotic resistance genes were detected, whereas in water isolates, this number was 57 (Table 5). Among the 59 genes detected in isolates from food, 38 (64.4%) were ß-lactam resistance genes and 21 (35.6%) were genes encoding non-ß-lactam resistance. From the 38 β-lactamase genes, 22 (57.9%) were ESBL genes, and 16 (42.1%) were potential AmpC producers. From the 21 non-β-lactamase genes, 16 (76.2%) were tet genes and five (23.8%) were sul genes. Each of the bla SHV variants and the ACC variants accounted for 23.7% of the total ß-lactam resistance genes, making them the most frequent genes, followed by the FOX variants (18.4%), MCTX-M Group 9 variants, and bla TEM variants (15.8%, respectively). The tetA gene was the most frequent non-ß-lactamase resistance gene (71.4%), followed by the sul2 gene (14.3%) and sul3 (9.5%) ( Table 5). In six (19.4%) isolates from RTE food and eight (28.6%) from water, there was the simultaneous occurrence of ESBL and AmpC genes. In two (6.5%) isolates from RTE and in six (21.4%) from water, we found the simultaneous occurrence of ESBL, AmpC, and sul coding genes. The simultaneous presence of ESBL and sul coding genes was observed in one isolate from RTE street food and water, respectively. The simultaneous occurrence of AmpC and tet coding genes was found in two isolates from RTE street food.
Simultaneous occurrences of more than one variant of ESBL gene were found in eight isolates from RTE street food and in four isolates from water. In RTE street food, bla TEM variants and bla MG9 variants were simultaneous founded in three isolates, variants bla TEM and bla SHV in three isolates, bla MG 9 variants and bla-SHV variants in four isolates, and bla Mg1 variants and bla SHV variants in one isolate. In water, bla TEM variants and bla MG9 variants were simultaneous founded in two isolates from water, The variants bla TEM and bla SHV one from water bla MG 9 variants and bla SHV variants one from water. bla TEM variants and bla MG1 variants, bla OXA variants and bla MG1 variants in one isolate from water. Regard AmpC variants, FOX variants and ACC variants were found in four isolates from RTE street food and in two isolates from water. The variants ACC and CIT were found in simultaneous in three isolates from water. In nom-ß-lactamases gens, sul2 and sul3 were simultaneous founded in two isolates from RTE and four from water.

Discussion
The PFGE subtyping results show a high diversity of isolates, either among those collected from RTE street food or from street water for human consumption. It was not possible to establish a relationship between the origin of the isolates (food or water) or the municipal district where they were sold, confirming the high diversity of fecal contamination sources of food and water sold on the streets, as previously reported [1,20]. In fact, Maputo municipal districts are non-urbanized areas, with numerous sanitation problems and water supply difficulties, contributing to the spread of pathogenic microorganisms [5,15,32]. In some cases, isolates gathered in clusters with a high level of similarity (>85%), presenting different virulence genes or different antibiotic resistance profiles, which suggests that the isolates may have had the same clonal origin but evolved in different environments where different genes were acquired, depending on adaptive mutations to cope with lethal stresses [33]. In general, isolates from food had a higher prevalence of virulence genes when compared with those from water. Nevertheless, stx coding for the shiga/vero toxin (VT) produced by the verotoxin-producing E. coli (VTEC) was identified with similar frequencies in isolates from food (5/31) and from water (6/28). The VTECs are considered a great threat in foodborne diseases. For example, E. coli O157:H7 became the first of several VTECs referred to as enterohaemorrhagic E. coli (EHEC), which can produce one or more shiga toxins (VTs). The main reservoirs for VTECs are ruminants, which continually shed bacteria into the environment contaminating food and water [34,35]. These strains can survive in fresh ground beef and on fresh leafy green vegetables [34]. Currently, E. coli is the most widespread indicator of fecal contamination in food handling processes, and its presence usually correlates with inadequate hygiene or cross contamination [36,37]. The isolates analyzed here were from street food and water with high fecal contamination (food E. coli counts above 3.5 log CFU/g [1], and between 1.6-2.8 log CFU/100 mL in water samples [22]). The occurrence of potential DECs among the isolates strongly supports the hypothesis of fecal contamination of food and water [38]. In fact, reports of lack of toilets and poor water supply at points of sale have already been published [1,2,5].
Among the street food isolates, 17 of the 26 antibiotic-resistant isolates carried one or more resistance gene. Of these, three had the gene encoding VT and were, simultaneously, multidrug-resistant. Nine had genes encoding EAST and showed high resistance to betalactams. All isolates with the eae gene were resistant to at least one antibiotic tested.
Among the water isolates, 14 out of 22 antibiotic-resistant isolates had virulence genes. Four of these isolates were multidrug-resistant with genes encoding VT, Lt and ST. All water isolates carrying the gene encoding EAST were resistant to antibiotics. The increased use of antimicrobials for prophylaxis and therapy against pathogens and for growth promotion in livestock, aquaculture and agricultural production may contribute to the increase of antibiotic resistance strains in the environment [39,40]. Multidrug-resistant Gram-negative bacteria represent significant global public health threats [41].
Information on antibiotic resistance in southern Africa is scarce, and the little existing data were obtained from clinical isolates, with a global prevalence between 25 and 50% [42]. In the present study, the highest percentages of resistant isolates from food and water were found for the ß-lactams IPM (35.5 and 39.3%, respectively) and AMP (39.3 and 46.4%, respectively). Regarding the non ß-lactams, the highest percentages of resistant isolates were found for TET (58.1% from food and 50% from water). Multidrug resistance was higher in E. coli isolates from water (45.5%) compared to RTE street food isolates (19.2%). Overall, three out of five and eight out of 10 multidrug-resistant isolates, from food and water, respectively, had virulence genes. The prevalence of the multidrug resistance reported here is lower than those observed for hospital isolates [15,18,43,44] but similar to those reported in South Africa in isolates from food and water [45,46]. In isolates from beef and water, Jaja et al. [45] reported a prevalence of resistance to β-lactams between 20 and 45% and a prevalence of 24% to the non-β-lactam TET. Malema et al. [46] found 50 to 75% of resistance to AMP, AUG, and TET in E. coli recovered in water, and multidrug resistance was 52% of the isolates [46]. In Tanzania, studies revealed an increasing trend toward the emergence of multidrug-resistant and pathogenic E. coli found in food animals [39].
The search for antibiotic resistance genes showed that ESBLs were the most frequent β-lactam genes, with 57.9% in isolates from RTE street food and 57.5% in water sources. The bla SHV variants (23.7%), CTX-M Group 9 variants (15.8%), and ba TEM -variants (15.8%) were the most predominant ESBL genes. Similar results were obtained by Faife et al. [47], who reported bla SHV (85.0%), followed by the Bla CTX-M group (62.9%) and bla TEM (44.4%), among the E. coli isolated from imported and locally produced chicken in Mozambique. However, according to Faife et al. [47], it is difficult to understand the contribution of local isolates to resistance genes, since some of the chickens were imported. Among E. coli isolates from children with enteric diarrhea, at the Maputo Central Hospital, Mozambique, Berendes et al. [15] found 68% of the isolates with the bla SHV variants and 20% with the prevalence of the CTX-M Group 9. Estaleva et al. [43] also performed studies with isolates from clinical cases and found 77% with the CTX-M Group and 52% with the bla SHV variants. The coexistence of several ESBL and AmpC genes were found in this study. A similar study reported a large number of clinical isolates that co-expressed bla CTM-1 , bla CTM-15 , bla CMY-2 , bla NDM-1 , bla NDM-5 and some other ß-lactamases [41]. ESBL-producing strains represent a major health problem worldwide [40,41]. The most common ESBL variants include blaTEM, blaSHV, and bla CTX-M , which are considered the main causes of hospital and community infections [41].
In this study, AmpC ACC variants and FOX variants were found in RTE street food (23.7 and 18.4%, respectively) and in water isolates (12.5 and 10%, respectively), whereas CIT variants were found only in water isolates (20%). In some isolates from the Enterobacteriaceae family, such as E. coli and Shigella spp., ampC is expressed at clinically irrelevant levels [48]. These isolates do not have an ampR gene [48,49], and AmpC regulation is accomplished by a weak promoter and a strong attenuator [48]. Thus, these isolates are normally sensitive to cefoxitin and are non-inducible, not altering the inhibitory effect of β-lactams unless by mutations in the promoter region [48,49]. However, these mutations must be hyper-producing of ampR, making the isolates resistant to cefoxitin and thirdgeneration cephalosporins [50]. These mutations are sporadic and much less frequent than the resistance mechanism in E. coli through the acquisition of ESBL gene [50].
In this study, DHA variant and MOX variant genes were not detected in food and water isolates. In isolates from clinical cases in Maputo, Chirindze et al. [51] observed a high prevalence of DHA variants (14.4%) and MOX variants (9.6%). Regarding tetracycline resistance genes, tetA, with 71.4% in food and 23.5% in water isolates, was the most frequent non-ß-lactam gene. Similar results were obtained by Berendes et al. [15], who reported a prevalence of tetracycline resistance genes (tetA and tetB) above 85% in isolates from children with enteric diarrhea in Maputo. Mandomando et al. [52] also found a high prevalence (71%) of tetA in Shigella and Salmonella isolates from hospitalized children. Although there are no Mozambican studies investigating the occurrence of tetracycline resistance genes in isolates from water or food, several studies in other African countries report a high prevalence of these genes in food-borne isolates [45,[53][54][55]. Furthermore, 13 out of 32 tetracycline resistance isolates were negative for the three tet genes searched in the present study, which leads us to infer that the enzymes involved in the phenotypic resistance to the antibiotic tetracycline are coded by genes different from those investigated here, making it necessary to search for other, less common genes.
Microbiological criteria define the acceptability of foods based on the number of microorganisms and their pathogenic potential [56]. The occurrence of potentially pathogenic E. coli isolates reported in the aforementioned studies [1,22] represents an increased risk to public health. The occurrence, for example, of potential VTEC in food (5/31) and water (6/28) may represent an increase in clinical cases in local hospitals. These cases are of particular concern as some of these potentially pathogenic isolates (3/5 and 5/6, from food and water, respectively) were multi-resistant isolates. The consumption of untreated water represents the main cause of occurrence of diarrheal diseases in developing countries, especially in children under five years old [57]. In low-income countries, these diarrheal diseases transmitted by water and food account for more than 35% of hospital admissions of these children [58].

Conclusions
A previous assessment of the microbiological quality of water and food sold on the streets of Maputo indicated that they were generally unsatisfactory for consumption. In the present work, although it was not possible to determine a relationship between food contamination and water contamination, a higher percentage of antibiotic resistance was found among the water isolates. This raises concerns about the possibility of contamination of water sources by antibiotics used in medical and veterinary treatments. Multidrug resistance was observed, some of these isolates also had DEC virulence genes. Other genes associated with other pathotypes, such as urinary tract infection or meningitis, have not been evaluated here. Furthermore, the antibiotic resistance profile was designed based on the most locally used antibiotics. The results obtained in this work suggest the evaluation of new generation antibiotics for diseases associated with DECs in Mozambique.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable. Data Availability Statement: All data supporting the findings of this study are available within the article and its supplementary materials.