Characterization and Phylogenetic Analysis of Campylobacter Species Isolated from Paediatric Stool and Water Samples in the Northwest Province, South Africa

Antibiotic-resistant Campylobacter could adversely affect treatment outcomes, especially in children. We investigated the antibiotic susceptibility profiles, virulence potentials and genetic relatedness of Campylobacter spp. from paediatric and water samples in the North West Province, South Africa. Overall, 237 human and 20 water isolates were identified using culture and real-time polymerase chain reaction (PCR). The antibiotic susceptibility profiles were determined using the disk diffusion method. Gradient strips were used to determine the minimum inhibitory concentration of each antibiotic. Antibiotic resistance (gryA, tetO and 23S rRNA 2075G and 2074C) and virulence (cadF and ciaB) genes were also investigated using PCR. A phylogenetic tree to ascertain the clonality between water and clinical isolates was constructed using MEGA 7. Overall, 95% (water) and 64.7% (human) of the isolates were resistant to at least one antibiotic tested. The highest resistance was against clarithromycin (95%) for water and ampicillin (60.7%) for human isolates. The 23S rRNA 2075G/2074C mutation was the most expressed resistance gene. Phylogenetic reconstruction revealed eight intermixed clades within water and human Campylobacter isolates. This study suggests the possible circulation of potentially pathogenic antibiotic-resistant Campylobacter in the Northwest Province, South Africa with drinking water being a possible vector for disease transmission in this area.


Introduction
Campylobacter are small, spirally curved, Gram-negative, non-spore forming, bacteria with a single polar flagellum [1]. There are currently 32 species and 13 subspecies of the genus [2]. They are the most prevalent and frequent causes of food-related infections worldwide [1]. Their ability to Campylobacter from paediatric samples and water samples collected from the North West Province, South Africa.

Ethical Statement
Approval to carry out the research was obtained from the College of Agriculture and Environmental Sciences (CAES) UNISA (2016/CAES/033), North-West Department of Health and Brits District Hospital. Written informed consent was obtained from parents/guardians of the participants, after which the participant's age, gender, and clinical signs were recorded using a questionnaire before collecting the samples.

Stool and Water Sample Collection
Stool specimens were collected from September 2016 to December 2017. During this period, 505 fecal specimens were collected from diarrhoeal and non-diarrhoeal babies and toddlers under the age of 5 years that were attending the Brits District Hospital, Oukasie Clinic, Lethabeleng Clinic and Bopang Clinic in the Madibeng District of the North-West Province, South Africa.
A total of 92 water samples were aseptically collected from September 2016 to December 2017. These water samples were collected from different households that allowed the researcher access to the premises. The only inclusion criterion was that the household had a child that was less than five years old. A sample of stored water that was intended for drinking or cooking was collected from each household in a 5-L bottle by the researcher. A total of 88 water samples were collected from different homes of which eight were directly from a municipal tap and 80 from water storage containers. Of the stored water samples, 38 were fetched from the municipal tap while 42 were from a well/underground water or rainwater harvested from rooftops. Also, four samples were collected from the Crocodile River.
All collected water and stool samples were transported on ice to the Microbiology Laboratory of the Council for Scientific and Industrial Research (CSIR), Pretoria, for analysis.

Isolation of Campylobacter from Drinking Water and Fecal Materials
Campylobacter was isolated from stool samples using conventional methods that included the morphological appearance and catalase tests as previously published by Bessède et al. [27], with a little modification as discussed in Section 3.5. For the isolation of Campylobacter from water samples, methods described by Jokinen et al. [28] and Talay et al. [29] were used. Briefly, water samples were filtered through a 0.45µm sterile membrane in a vacuum filter. Membrane filters were placed in Bolton broth (BB) and incubated at 42 • C in a microaerophilic environment (MAE) for 24 h. After that, 100 µL of the enriched broth was placed onto a 0.6 µm membrane filter placed on tryptose blood agar (TBA) and incubated at 25 • C for 20 min. The membrane filters were then rolled out, and TBA plates were incubated at 42 • C in an MAE for 48 h.

Campylobacter Species Identification
Presumptive colonies from the culture plates were confirmed as being Campylobacter by targeting the genus-specific 16sRNA gene using real-time PCR [30]. Real-time PCR was also used to confirm species using the primer glyA for C. coli and C. upsaliensis [31,32] and hipO for C. jejuni [32]. DNA was extracted using the heat lysis method [33]. The purity and concentration of the extracted DNA were determined spectrophotometrically using the Nanodrop Lite Spectrophotometer (Thermos Scientific, Waltham, MA, USA), and all samples had an A 260/280 ratio ranging from 1.7 to 2.1.
Real-time PCR was performed using a Corbett Life Science Rotor-Gene TM 6000 Cycler (Qiagen, Hilden, Germany). The primers, 0.5 µL, (Forward and Reverse; final concentration 0.5 µM each), nuclease-free water (1 µL) and sample DNA (3 µL) were added to 10 µL of 2x SensiFAST TM High-Resolution Melt (HRM) mix (SF) (Bioline GmbH, Luckenwalde, Germany), to obtain a final reaction volume of 15 µL. The cycling conditions included an initial activation at 95 • C for 10 min, followed by 40 cycles of denaturation at 95 • C for 10 s, annealing at 60 • C for 15 s and an extension at 72 • C for 20 s. The final step was an extension at 72 • C for 5 min. A melt curve was prepared by ramping up the melting temperature from 72 • C to 95 • C. Melt curve analysis was performed using the Rotor-Gene™ real-time analysis software, version 6.1 (build 93) (Corbett Life Science (Pty) Ltd., Sydney, Australia).

Antibiotic Susceptibility Testing
The E-test strip method (Oxoid Ltd., Basingstoke, UK) was used to detect the MIC of the antibiotics [38]. The strip contained the required antibiotics at appropriate concentrations. The dilution range of the antimicrobial tested were: ciprofloxacin (32-0.015 µg/mL), erythromycin (256-0.015 µg/mL), tetracycline (256-0.015 µg/mL), ampicillin (256-0.015 µg/mL) and gentamycin (32-0.002 µg/mL). In brief, confirmed isolates from water and stool samples were grown on BA at 42 • C for 48 h. After incubation, a suspension was prepared in normal saline and adjusted to a 0.5 McFarland standard. The suspensions were spread onto a 5% sheep blood Muller Hinton agar using a sterile cotton swab, and then the antibiotic strips with different antibiotic concentration gradients were placed onto the agar plates and incubated for 24 h at 42 • C in a MAE. After the incubation, the MICs were measured, and the results were interpreted according to the National Committee for Clinical Laboratory Standards to Enterobacteriaceae [35,36].
The MIC was defined as the lowest concentration of an antibiotic that completely inhibited visible growth and was read at the point where the elliptical zone of inhibition intersected against the MIC scale on the strip. MIC 50 and MIC 90 were described in this study as the MICs that completely inhibited visible growth of 50% and 90% of the strains, respectively. The break-point criteria used for erythromycin and tetracycline were those of the CLSI for Staphylococcus species while for other drugs, criteria recommended for Enterobacteriaceae were used [39].

Detection of Antibiotic Resistance Genes
Genes conferring resistance to macrolides, fluoroquinolones, and tetracycline were screened on all water isolates and 206 clinical isolates, using the extracted DNA template. For the presence of quinolone resistance, the Thr-86-lle mutations that are found in the quinolone resistance-determining region (QRDR) of the gyrA gene in Campylobacter spp. was amplified [40]. The tetO gene responsible for tetracycline resistance was amplified, and for the presence of macrolides (erythromycin) resistance, point mutations were detected at position 2075 and 2074 in the 23S rRNA gene [12,41] using specific primers that targeted the desired fragments (Table 1).

Sequence Assembly and Alignment
To determine the relatedness between the isolates, all isolates from water and 40 isolates from human samples were amplified using primers targeting the conserved region of the 16S rRNA gene for identification of the Campylobacter genus [30]. The PCR products were purified and sequenced using Sanger sequencing at Inqaba Biotech TM (Pretoria, South Africa). Sequence fragments were generated using Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). A BLAST search was performed, and the sequences were compared to known Campylobacter sequences in the GenBank. After that the sequences were analysed, and a phylogenetic tree based on 16S rRNA was created using MEGA7 [44,45].

Results
The participants in the current study included 257 males and 248 females. Of the 505 participants, 184 were on exclusive breastfeeding while 321 were on mixed feeding. Also, children with diarrhoea (321, with 82 having bloody diarrhoea) were more than those without (155). One hundred and fifty-nine participants reported vomiting while 209 had fever.

Detection of Campylobacter spp.
Overall, 108 C. jejuni, 89 C. coli and 40 C. upsaliensis were isolated from 505 paediatric diarrhoea and non-diarrhoea stool specimens. Of these, 81 C. jejuni, 78 C. coli and 29 C. upsaliensis were from diarrhoea and 27 C. jejuni, 11 C. coli and 11 C. upsaliensis were from non-diarrhoea stool samples. The detection of Campylobacter from individual water sources are shown in Table 2. The overall Campylobacter recovery from water samples was 21.7% (20/92). The highest Campylobacter recovery (35.7%) was detected in well/harvested rainwater that was stored in a container.
Campylobacter was not detected from river water and the water samples that were collected directly from the municipal tap.

Antimicrobial Susceptibility of Campylobacter Isolates
Of the 14 antibiotics tested, the highest phenotypic resistance displayed by Campylobacter isolates from water samples was against clarithromycin (95%), while complete susceptibility (100%) was observed against imipenem. From the human samples, the highest resistance was observed against amoxicillin/clavulanic acid (64.7%), while resistance to imipenem was the least observed (15.3%) Table 3. Note: EUCAST interpretation criteria for erythromycin on C. jejuni (<20) and C. coli (<24) was used in interpreting results for the macrolides. CLSI breakpoint for Enterobacteriaceae was used for aminoglycosides, carbapenems and fluoroquinolones.
The distribution of antibiotic resistance according to species showed different patterns (Table 4) with resistance ranging from 13.3% to 95%.

Determination of the Minimum Inhibitory Concentration (MIC)
The MICs of the five antibiotics tested are shown in Table 5. All isolates were highly resistant to ampicillin. C. upsaliensis isolates from human samples were susceptible to ciprofloxacin and gentamicin while the isolate from water was only susceptible to erythromycin.

Prevalence of Multiple-Antibiotics Resistance (MAR)
The overall prevalence of MAR revealed that 76% stool and 90% water Campylobacter isolates were simultaneously resistant to more than three of the antibiotics tested ( Figure 1). MAR in this study was defined as the resistance of Campylobacter to two or more antibiotics [13]. Two isolates each from water and human stools were concurrently resistant to up to 11 antibiotics. The presence of MAR was observed more in C. jejuni isolates.

Determination of the Minimum Inhibitory Concentration (MIC)
The MICs of the five antibiotics tested are shown in Table 5. All isolates were highly resistant to ampicillin. C. upsaliensis isolates from human samples were susceptible to ciprofloxacin and gentamicin while the isolate from water was only susceptible to erythromycin.

Prevalence of Multiple-Antibiotics Resistance (MAR)
The overall prevalence of MAR revealed that 76% stool and 90% water Campylobacter isolates were simultaneously resistant to more than three of the antibiotics tested ( Figure 1). MAR in this study was defined as the resistance of Campylobacter to two or more antibiotics [13]. Two isolates each from water and human stools were concurrently resistant to up to 11 antibiotics. The presence of MAR was observed more in C. jejuni isolates.  Table 6 shows the distributions of the antibiotic resistance genes among the species; 28.6% of the human isolates harboured one or more genes, while more than 90% was found from the water samples.    Table 6 shows the distributions of the antibiotic resistance genes among the species; 28.6% of the human isolates harboured one or more genes, while more than 90% was found from the water samples.

Expression of antibiotic resistance genes by Campylobacter isolates
The tetO gene was the most amplified gene. It was found in 59/206 (28.6%) of the human samples tested (45 isolates that were simultaneously resistant to tetracycline and tigecycline). Also, 14 of those that did not undergo a susceptibility test expressed the tetO gene.
All the isolates phenotypically resistant to tigecycline harboured the tetO gene. Also, 45/48 (94%) of the human isolates that were resistant to tetracycline expressed the tetO gene. Among the species, the tetO gene was found more in C. jejuni (31.8%; 29/91) compared to C. coli 30.8% (25/81) and C. upsaliensis 14.7% (5/34). From water samples, the tetO gene was found in 40% (8/20) of the isolates, and all the isolates that expressed this gene were phenotypically resistant to tetracycline. At the species level, the tetO gene was found more in C. jejuni isolated in human samples, while in water samples it was amplified more in C. coli isolates.  (75) Transitional mutations at position A2075 and A2074 in the V region of the 23S rRNA gene was the most amplified gene in the water isolates, while from the human samples it was the second most amplified gene. From the water samples, 85% of the isolates (17/20) that were phenotypically resistant to erythromycin had an A2074 point mutation, while 16/20 (80%) isolates that were resistant to both erythromycin and clarithromycin showed mutations at position A2074 and A2075 in the V region of the 23S rRNA. From clinical strains, mutations at A2075 occurred in 19.4% (40/206) of the isolates, while a mutation at position A2074 was found in 18.4% (38/206) of the isolates. Also, 18.4% (38/206) of the isolates that were concurrently resistant to the macrolides tested expressed mutations at both positions (A2074/2075) of the V region. Furthermore, two human isolates that were susceptible to erythromycin but resistant to clarithromycin also expressed a mutation at position A2075. C. coli isolates 24.6% (20/81) expressed the highest mutation rate compared to C. jejuni 18.6% (17/91).
The quinolone resistance-determining region (QRDR) of the gryA gene was amplified in 18.4% (38/206) of the clinical isolates and 25% (5/20) of the water isolates. All strains were concurrently resistant to ciprofloxacin and norfloxacin, that is, all fluoroquinolone-resistant isolates expressed the gryA gene. Also, 11 isolates that were excluded during the susceptibility test also expressed the gryA gene. Distribution of the gryA gene according to species is shown in Table 6. Isolates haboring multiple resistance genes were also observed in the clinical as well as the water samples. Thus, from the clinical setting, tetO and gryA genes were found in 6.8% (14/206) of the isolates. The tetO gene and mutation at position A2075G/ A2074C were found in 9.7% (20/206) of the samples. Of these, 12% (11/91) were C. jejuni, 8.6% (7/81) C. coli and 5% C. upsaliensis. C. upsaliensis identified in the water samples harboured all the resistance genes tested. Sequenced samples all showed similarities with known gryA and tetO genes of Campylobacter jejuni and Campylobacter coli in the GenBank.

Expression of Virulence Genes among Campylobacter Species
The result of the virulence genes showed that 85% of the water harboured the ciaB and cadF gene, while from human samples, ciaB was expressed in 38% and cadF in 51% of the isolates (Table 7). From the human samples, the ciaB gene was expressed more in C. coli isolates (39.3%), while the cadF gene was found more in C. jejuni (54.6%). A combination of the ciaB and cadF genes was found in 24% of the isolates, of which 22.2% (24/108) of the combinations were found in C. jejuni, 23.5% (21/89) in C. coli strains and 30% (12/40) in C. upsaliensis. From the water samples, 90.9% of C. jejuni expressed the cadF gene.

Phylogenetic Relationship of Campylobacter Strains by Partial Genome Sequencing
Phylogenetic reconstruction revealed five different clades. These clades were placed into eight groups (Groups I-VIII) according to how closely related the strains were. Groups I, III, VI and VII contained the sequences that were intermixed with Campylobacter spp. isolated from both human and water samples (Figure 2). Groups II, V and VIII, consisted of Campylobacter strains that were circulating within the studied human population. Group IV were strains that only existed in water.

Discussion
Campylobacter spp. are identified as etiologic agents in outbreaks and sporadic cases of diarrhoea and gastrointestinal infections worldwide. A recent report by the Global Enteric Multicentre Study

Discussion
Campylobacter spp. are identified as etiologic agents in outbreaks and sporadic cases of diarrhoea and gastrointestinal infections worldwide. A recent report by the Global Enteric Multicentre Study group (GEMS) indicated Campylobacter as one of the primary agents that causes diarrhoea in developing countries [46], and infection is usually limited to children [47,48]. Untreated drinking water has been noted as a significant source of Campylobacter infections and outbreak [22]. This study, therefore, investigated the genetic similarity of 257 Campylobacter strains isolated from paediatric stool and household drinking water samples in the Northwest Province of South Africa. Overall, 119 C. jejuni (108 from paediatric stools and 11 from water), 97 C. coli (89 strains from stools and 8 from water) and 41 isolates of C. upsaliensis (40 from stools and 1 strain from water) were isolated and screened against 14 different antibiotics. Human isolates exhibited different levels of resistance against all the antibiotics tested, while some water isolates were simultaneously resistant to up to 13 different antibiotics. The 23S rRNA 2075G/2074C mutation and tetO gene were the most expressed of all the resistance genes analysed. Phylogenetic reconstruction revealed eight clades that were intermixed within Campylobacter spp. isolated from both water and human samples.

Detection of Campylobacter spp.
Clinical Campylobacter species were isolated from paediatric patients with diarrhoea and those without diarrhoea, while water isolates were from stored household water and municipal tap water sources. Twenty-one percent of the water samples were contaminated with diverse Campylobacter species. Several studies have assessed the prevalence of Campylobacter spp in different water sources [1,[49][50][51]. In South Africa, 13% was reported in surface and groundwater [52], and in New Zealand, 75% and 29.2% were found in groundwater and drinking water, respectively [53]. The contributions of water to the burden of sporadic cases of Campylobacter infections might be unknown because not all cases lead to severe illness and most often a milder degree of illness might not require medical attention [22]. As such, infected people might not report to a hospital, thus affecting the overall prevalence within a given community. However, most outbreaks have been mainly attributed to the drinking of contaminated water [22], indicating the role water plays in the transmission of Campylobacter infections.
No Campylobacter contamination was recorded in the water samples collected from the Crocodile River. These results were surprising considering that river water has been reported as a reservoir of Campylobacter spp. and prevalence ranging from 60% to 79% have been found [53,54]. However, it should be noted that Campylobacters have been reported as non-indigenous to aquatic environments, mainly because of their growth requirements, and their presence is indicative of recent faecal contamination [55]. Hellein et al. [56] reported that aquatic Campylobacter contamination reflected sewage effluent contamination and agricultural runoff. Thus, it could be assumed that the inability to isolate Campylobacter from the river water sample might be associated with the choice of site because the river water samples analysed in this study were collected from areas that were less impacted by human activities. Campylobacters, like many other enteric pathogens, do not occur in high concentrations in aquatic environments, so isolation usually requires the concentration of larger volumes of water, particular growth requirements and more extended incubation periods; in the absence of the above-mentioned conditions, results in most cases are usually false negatives [57]. Also, it has been reported that Campylobacters can enter the viable but non-culturable (VBNC) state when exposed to prolonged poor nutrient and unfavourable temperature conditions in aquatic environments [52]. These factors, coupled with the few numbers of samples included in the current study, could have affected the results obtained.
Most Campylobacter outbreaks usually highlight infections emanating from inadequate disinfection and filtration or sewage contamination and drinking contaminated water is the most accepted cause of Campylobacter enteritis outbreaks [22,52]. In the present study, water samples collected directly from the municipal taps were all negative for Campylobacter, indicating that municipal taps might not be the source of water contamination. It could also mean that treatment at the waterworks could effectively remove Campylobacter from raw water, as studies have reported on the sensitivity of Campylobacter strains to disinfection [58,59]. The highest Campylobacter contamination in the present study was found in water samples collected from stored containers (rainwater, well and municipal taps). Several factors have been linked to the poor microbial quality of stored household water [50]. Although treated municipal water may contain residual chlorine to ensure the safety of the water during storage, prolonged storage within the house could lead to recontamination of the water. The type of container may also allow for recontamination of previously treated water [50,60]. For example, wide open-neck containers would allow for recontamination during extraction of the water using dirty containers. Another critical factor is the source of water. Most of the samples analysed in the current study were collected from wells or rain. These water sources have been reported to contain substantial numbers of microorganisms, including pathogenic ones [60,61]. Thus, storing such water in the household without pre-treatment could favour the growth of the already present microorganisms.

Antibiotic Resistance Profiles of Campylobacter Species
Antibiotic resistance has been documented as a global pressing public health concern. In developing countries, the situation is deteriorating more rapidly because of the widespread and uncontrolled use of antimicrobial agents [13,62]. Although Campylobacter infections are self-limiting, antibiotics may be prescribed to patients with unusually severe and prolonged symptoms or in immunocompromised patients [14]. In recent years, antibiotic resistance (ABR) has been reported in some Campylobacter spp. in most countries [47]. In the present study, 150 isolates from paediatric stool samples and 20 isolates from domestic water samples were tested against 14 antibiotics. About 65% of clinical isolates and 95% of isolates from the water samples were resistant to at least one antibiotic tested. In previous reports, antibiotics in the carbapenem group have shown an excellent in vitro activity against Campylobacter spp. [63,64]. Correspondingly, in this study, a low antibiotic resistance was observed to imipenem and meropenem. A relatively low resistance was also observed against gentamicin among the clinical isolates although gentamicin has been reported by the WHO as an alternative in cases of sepsis and some neonatal bacteraemia [65]. However, the prolonged use of gentamicin can lead to renal tubular dysfunction in children [66]. Compared to the water isolates, all clinical isolates showed the highest resistance to macrolides and penicillin with resistance rates of 80% and 85% respectively. Minimal resistance was experienced against other antibiotics except imipenem with 100% susceptibility.
Based on the results of this study, no currently tested antibiotic reliably covered all the clinical Campylobacter strains identified in this study. Among the empirical drugs, antibiotic resistance was slightly lower against the fluoroquinolones. Norfloxacin exhibited a lower resistance (13.3%) and offered a better alternative than ciprofloxacin (18%). Fluoroquinolones are among the recommended drugs for the treatment of campylobacteriosis [16]. However, treatment with fluoroquinolones has become quite a challenge as some Campylobacter spp. have developed resistance to this class of antibiotics [67]. Infections due to fluoroquinolone-resistant Campylobacter strains are usually severe and last longer. Fluoroquinolone-resistant Campylobacter have been reported to be responsible for 23% of all campylobacteriosis in the United States and cause an estimated 310,000 illnesses per year [68]. In Campylobacter and other Gram-negative bacteria, the fluoroquinolones act by inhibiting the function of topoisomerase enzymes (topoisomerase II & IV) and DNA gyrase. Studies conducted in many countries have shown that alteration or mutation in the gyrase A (gyrA gene) of Gram-negative bacteria might result in an automatic resistance to fluoroquinolones [13,14]. The percentage resistance recorded for the clinical and water isolates against ciprofloxacin and norfloxacin agrees with a previous prevalence report in South Africa, where Campylobacter resistance rates to fluoroquinolones between 14.8% and 51.3% were recorded [69]. These results illustrate that Campylobacter resistance to fluoroquinolones might not have increased over the years. However, constant monitoring is required as Campylobacter spp. can mutate [70]. Antibiotic resistance prevalence among the species showed that clinical C. jejuni isolates were highly resistant to ciprofloxacin (24.2%) and norfloxacin (16.6%) compared to C. coli (18.6%) and (8.4%), respectively, contrary to the pattern observed with the water isolates. Only one isolate of C. upsaliensis was identified from the water samples, and the strain was resistant to fluoroquinolones. Also, from the clinical specimens, 4% of the C. upsaliensis was resistant to norfloxacin. A previous study in Denmark recorded resistance to fluoroquinolone ranging from 48.2% (C. jejuni) to 66.7% (C. coli) [71]. Similarly, 42% of C. jejuni and 83% of C. coli isolates, recovered from patients with travelers were resistant to fluoroquinolones in a study conducted in Finland [72], while 63.2% of resistance to fluoroquinolones has been reported in patients with severe diarrhoea in the United Arab Emirates [11]. Previous reports have speculated that the spread of fluoroquinolone resistance in human isolates might have originated from the excessive use of veterinary fluoroquinolones (enrofloxacin and danofloxacin) in food-producing animals [73]. For example, in Australia, where the use of fluoroquinolone is banned in food-producing animals, a rate of 0-2% resistance was reported [74]. However, antibiotic resistance of Campylobacter isolates in food-producing animals has not been studied in many developing countries, including South Africa. Thus, conducting similar studies on poultry, for example, would enhance the understanding of possible sources of antibiotic resistance in the study locality.
Given the increasingly high incidence of fluoroquinolone resistance in Campylobacter spp., macrolides were considered the alternative drugs of choice for the treatment of human campylobacteriosis [67]. The incidence of Campylobacter resistance to macrolides in clinical isolates was previously rare, especially in developed countries [41]. Recent reports from different parts of the world have, however, shown that Campylobacter spp. have acquired resistance to this class of antibiotics. The isolates in the current study showed varying resistance rates against the macrolides tested (erythromycin; 26.7% and clarithromycin; 29.3%). Previous studies in South Africa have reported up to 53% resistance to macrolides [12,69], while a resistance of 86.1% has been reported in India [13]. Studies from other parts of the world have reported that C. coli strains often show increased resistance to macrolides compared to C. jejuni [72]. However, C. jejuni was the most resistant to macrolides compared to other species in the current study. A previous study in South Africa recorded resistance to ciprofloxacin and erythromycin in 33.3% and 38.9% of C. coli and 20% and 31.5% of C. jejuni, respectively [12]. An increase in resistance to both erythromycin and ciprofloxacin has also been documented in other developing countries [47], and it is speculated to be primarily influenced by the use of macrolides for infections other than gastrointestinal diseases and the pressing issue of self-medication [75,76]. It might also be due to horizontal transfer of resistance genes from animals to humans as it has been documented that macrolides like spiramycin, erythromycin and tylosin prevent infection in animals or act as growth promoters [67,75]. Reports from the Centers for Disease Control and Prevention (CDC) showed that macrolide-resistant Campylobacter strains are responsible for 2% of campylobacteriosis in the United States and cause an estimated 22,000 illnesses and up to 5 deaths annually [68]. Thus, the high resistance to macrolides recorded in the current study calls for more stringent measures to prevent the spread of these bacteria, especially within settings with limited resources such as the North West Province. Also, the higher prevalence of macrolide resistance recorded in water compared to the clinical isolates is a call for concern, given that the consumption of untreated contaminated water has been linked with numerous waterborne Campylobacter disease outbreaks around the world.
Increase in tetracycline resistance has been reported to emerge from the extensive use of these antibiotics as prophylaxis and therapy of human and animal infections and in promoting animal growth [77]. Thus, Campylobacter resistance to tetracycline has been frequently reported in humans, animals and aquatic environments [78][79][80]. The current study observed a resistance rate of 32% of clinical Campylobacter isolates to tetracycline and 55% from water isolates. Among the species, C. jejuni strains from human samples showed a higher resistance to tetracycline compared to other strains, while from water samples, resistance to tetracycline was observed more among C. coli isolates. A study in Quebec, Canada, reported a 50% tetracycline resistance among C. coli isolates and 39% among C. jejuni isolates [78]. In Spain, C. coli (94%) and C. jejuni (36%) isolated from water samples were reported to be resistant to tetracycline [80]. Moreover, in South Africa, 55% resistance among C. coli isolates and 25.9% among C. jejuni isolates have been reported [12]. Contrarily, C. jejuni isolates showed the highest resistance in the current study compared to the other species. The resistance of Campylobacter spp. exhibited against tetracyclines in the current study suggests a potentially high risk of treatment failure in Campylobacter infections and highlights the importance of monitoring antibiotics and the quest for alternative strategies to treat bacterial infections.

Multi-Antibiotic Resistance (MAR)
There is an increased trend in the occurrence of MAR pathogenic organisms worldwide [81]. MAR, especially to macrolides, fluoroquinolones and tetracyclines, is a considerable concern, and it is considered highly undesirable in Campylobacter as these three classes are generally advocated for as the first-line drugs for the treatment of Campylobacter enteritis [35]. The first ever reported case of MAR Campylobacter in South Africa was observed in children less than five years old at the Red Cross Children's Hospital in Cape Town [82]. Since then, MAR Campylobacter strains have been frequently encountered in human, animals and water samples [79,83]. MAR Campylobacter strains were believed to emerge from the consumption of poultry meat because antibiotics are used in poultry production as growth promoters [72,84]. However, studies have shown that the excessive use of antibiotics by humans can also lead to the development of MAR in human isolates [62]. The present study recorded MAR in 76% and 90% of clinical and water isolates, respectively. These isolates were resistant to more than three antibiotics agents. All three Campylobacter species analysed in the current study exhibited MAR. However, the highest resistance observed from the clinical samples was found in C. upsaliensis (94.4%). In Qatar, United Arab Emirates, 6.3% of MAR Campylobacter strains were isolated from patients with bloody diarrhoea [11]. MAR strains of C. jejuni (52.6%) and C. coli (47.4%) recovered from Finish patients were found to be co-resistant to tetracycline and amoxicillin/clavulanic acid [72]. In Ghana, 100% MAR was recorded in water samples [81]. Resistance to relevant therapeutic agents poses a risk when there is no effective antimicrobial regimen for Campylobacter infection making treatment unattainable. Studies have shown that patients infected with MAR Campylobacter strains have an increased risk of an adverse event compared to patients infected with a susceptible Campylobacter strain [71].

Distribution of Antibiotic Resistance Genes
Antibiotics in the tetracycline family act by inhibiting bacterial protein synthesis. They achieve this by preventing the attachment of aminoacyl-tRNA to the ribosomal acceptor (A) site [60,62,85]. Resistance to tetracyclines in most bacteria is often due to the acquisition of new conjugate genes that are associated with plasmids or transposons [77]. In Campylobacter, the principal determinant of tetracycline resistance is a plasmid-borne gene, belonging to the tet family of proteins (tetO), which confers resistance by displacing tetracycline from its primary binding site on the ribosome [62]. Studies have shown that tetO protein reduces the susceptibility of ribosomes to the action of tetracyclines when guanosine triphosphate (GTP) is present [62]. The tetO gene was the most amplified resistance gene in the human samples and the second most amplified gene in water isolates in the current study. All the human isolates that were phenotypically resistant to tetracycline harboured the tetO gene, while only 40% out of the 50% of the water isolates that were phenotypically resistant to tetracycline expressed the tetO gene. The prevalence of the tetO gene in C. jejuni strains was higher compared to other species. It should be noted that C. jejuni isolates in this study were mostly from paediatric patients with severe clinical symptoms, suggesting that the resistant strains could have contributed to the severity of the infections caused. C. coli and C. upsaliensis also expressed the tetO gene at 20% (25/89) and 12% (5/40), respectively. In Canada, a higher prevalence of the tetO gene has also been reported among human isolates [78]. High prevalence of the tetO gene poses a threat in the treatment of campylobacteriosis given that the tetO gene can be transferred rapidly from a resistant isolate to a susceptible isolate [85].
Just like tetracyclines, macrolides also inhibit bacterial growth by binding to the bacterial ribosomes and interfering with protein synthesis. In Campylobacter, macrolide resistance is chromosomally mediated and is associated with a reduction in macrolide binding affinity to the ribosomal 23S subunit [86]. Thus, Campylobacter spp. may evade macrolides by altering the antibiotic's target site at the V region of the 23S rRNA. Most often, alterations at position 2074C or 2075G proffer high-level resistance. It has been reported that Campylobacter strains carrying these mutations are stable in culture, and also maintain their ability to colonize their host [63]. In the current study, 55 out of the 62 human isolates (88.7%) analysed and 15 out of the 17 water isolates (88.2%) had this mutation. These isolates had very high MICs and expressed both mutations at A2074C and A2075G. Isolates that were either resistant to clarithromycin or erythromycin had a mutation at position A2075G only. Campylobacter isolates that express mutations at position A2075G and A2074C extend high-level resistance [87]. The findings of the current study corroborate the report of Vacher et al. [88] who observed point mutations at position A2074C and A2075G in the 23S rRNA gene in 99.3% of C. coli and C. jejuni isolates in their study [88]. Similar to the present study, single mutations at positions A2075G or A2074C have also been reported previously [14]. Also, combined mutations at A2075G and A2074C have been reported to confer a higher level of erythromycin resistance among Campylobacter isolates [67,87]. Previous studies reported that resistance in C. coli strains was associated with a mutation at position A2075G and A2074C [87]. This report agrees with the present study as the rate of expression of mutations at position A2075G/ A2074C was higher in C. coli isolates compared to C. jejuni and C. upsaliensis.
Members of quinolone antibiotics target two large bacterial enzymes, the DNA gyrase and the topoisomerase IV. Studies have shown that binding of quinolones to these enzymes inhibits the synthesis of bacterial DNA, which ultimately causes cell death [39,89]. However, some bacteria have developed resistance to this set of antibiotics by substituting amino acids at the quinolone resistance-determining region (QRDR) of the topoisomerase [3]. In Campylobacter, resistance to quinolones is primarily mediated by a single point mutation in the QRDR of the gyrA gene at codon 86 (that is, an alteration of the nucleotide from ACA to ATA), leading to isoleucine substitution for threonine [90,91]. Although there are different types of amino acid substitutions, the most frequently observed is the C 257 T mutation in the gyrA gene which leads to Thr86Ile substitution in the gyrase and confers a high level of resistance to this class of antibiotics [90]. In the current study, 18.4% of the human and 25% of the water Campylobacter isolates harboured the gryA gene. The prevalence of the gryA gene was higher in C. upsaliensis, and C. jejuni compared to C. coli. Previous studies have reported a similarly high occurrence of the gryA gene in C. jejuni strains. For example, a study conducted in Europe reported that over 41% of C. jejuni isolated from humans, poultry products, water and wild bird carcasses expressed the gyrA gene [90]. Pere-boto et al. also reported that the gyrA gene is the most prevalent resistance gene in clinical isolates collected from 10 different provinces in Spain, and was expressed mostly in C. jejuni isolates that exhibited high ciprofloxacin MIC [14]. Previous studies have reported that most ciprofloxacin-resistant Campylobacter spp., especially C. jejuni strains, express the Thr86Ile amino acid substitution in the QRDR of gyrA [40,90,92]. The CDC, in 2013, reported that fluoroquinolone resistance among Campylobacter strains comes with a heavy economic burden because infections caused by fluoroquinolone-resistant strains stay longer and can in most cases lead to death [68].

Detection of Campylobacter Virulence-Associated Genes
The mechanism by which Campylobacter causes human diseases is believed to be multifactorial [10]. Specific genes involved in adhesion, colonisation, invasion and toxin production are necessary for the process of infection [93]. To determine the pathogenic potentials of the Campylobacter isolates in the current study, the presence of two essential genes coding for virulence determinants such as the adhesive (cadF) and invasive (ciaB) genes in the isolates was investigated. These virulence factors were expressed more in water isolates compared to clinical isolates. The higher prevalence of virulence genes noticed among the water isolates in this study contradicts other published studies in which relatively higher number of virulence genes were identified in human clinical samples [94,95]. These results, therefore, indicate that the Campylobacter spp. isolated from water samples in this study might be highly virulent, and could attach and invade the host epithelial cells [43].
Interestingly, not all diarrhoeal isolates in the present study expressed the virulence genes. Given that the cadF gene aids Campylobacter to adhere to the host gastrointestinal epithelium and for internal colonization [95], while the ciaB gene is required for maximal invasion of intestinal epithelial cells [96], it would be expected that all isolates identified from diarrhoea cases would harbour the virulence genes. It has been previously reported that Campylobacter strains that lack the cadF gene were unable to colonise in chicken models and their internalisation ability was compromised [97]. The results of the current study are in agreement with those of Koolman et al. who tested for the presence of a series of virulence genes in Campylobacter isolates and observed that not all strains possessed adhesin proteins and that some strains that possessed the genes could invade Caco-2 cells [93]. Similar reports had earlier been published by Ziprin et al. [43], where the C. jejuni cadF mutants were unable to colonise chickens. It has been suggested that the lack of the cadF gene and inability to colonise and bind intimately to the host cells as exhibited by some Campylobacter strains is due to their inability to overcome different biological barriers and stressors encountered in the host cell, including increased temperature of the host and the acidity level of the stomach [10,43].
The low prevalence of the cadF and ciaB genes in the clinical isolates, however, contradicts other previously published data reporting 100% detection of these virulence determinants in their isolates [42,95,98,99]. The results of the present study, therefore, confirm the argument that not all Campylobacter strains harbouring the ciaB or cadF gene can adhere or invade intestinal cells. Some invasive and adherence factors other than those coded by the ciaB and cadF genes have been reported on Campylobacter surfaces [10], and this could have been the case in the current study population. Similar to our study, a lower prevalence of 76.4% in ciaB and 63.9% in cadF has been reported in Qatar, 66.7% and 51.5% in the Arabian Peninsula and 71.4% and 52.4% in Asia [11]. Most of the cadF genes in our study were expressed from the samples of children with diarrhoea (80.9%) and fever (69%).
Virulence genes were also expressed among the antibiotic susceptible isolates more than in the resistant isolates. These results corroborate the report of Al-Mahmeed et al. [100], and Rozynek et al. [101], where the adhesin genes tested were significantly associated with the antibiotic-susceptible strains. It, therefore, means that although a particular Campylobacter strain may be susceptible to a range of antibiotics, it may harbour virulence genes allowing it to elicit an infection of the same magnitude as a strain bearing a resistance gene. Studies have shown that there is a positive relationship between multi-virulence genes and the severity and duration of clinical symptoms [15].

Genetic Relatedness of Campylobacter Isolates from Human and Water Samples
A Newick tree was constructed to determine the genetic relatedness between Campylobacter isolates from water and stool. The tree revealed eight different groups in which, Groups II, V and VIII were Campylobacter strains that were exclusively circulating in paediatric populations, while groups IV were only found in water samples. These groups did not suggest any Campylobacter movement from human to water or vice versa. Four groups (I, II, VI and VII) were intermixed with Campylobacter strains isolated from both water and paediatric samples. These results show that these strains were closely related and may belong to the same lineage, suggesting that there was a possible transmission of Campylobacter infection from water to humans within the study population. These results support the notion that water is a significant source of human Campylobacter infections [53,102]. Seeing that strains from water and strains from humans shared the same group, it could be assumed that human isolates most likely originated from the water. Although Campylobacter species are known to be transmitted through the consumption of contaminated water, further studies involving a larger number of water samples would be needed to establish the association between stored household drinking water and the transmission of Campylobacter infections in the study area.

Conclusions
Control and prevention of campylobacteriosis in humans requires knowledge of transmission routes, antibiotic resistance profiles and virulence capacities of isolates. The results obtained in the present study showed the presence of three Campylobacter species in the studied communities displaying varying degrees of resistance, especially to the empirical drugs used for the treatment of Campylobacter infections. Also, while strains isolated in the study carried virulence genes, the detection rate of these genes was higher in the water samples than in the clinical isolates. Phylogenetic analysis revealed that Campylobacter infection from the studied communities might have been acquired through the consumption of contaminated water. It is, therefore, necessary to undertake continuous monitoring of the prevalence of Campylobacter and its associated virulence genes and antibiotic resistance profile to inform effective treatment regimens for Campylobacter infections. Finally, it is very important to emphasise that the presence of virulence genes is indicative and may not predict precisely how virulent a Campylobacter strain might be. Also, a negative result by real-time-PCR might not necessarily mean the absence of a gene but could be attributed to the sequence variation at the primer binding site or existence of another gene with a similar role. Funding: This research received no external funding; however, the student was supported by NRF and UNISA Research Directorate.