Description of study areas
Collection of water samples was carried out from three rivers namely; Great Fish, Keiskamma and Tyhume, in the Eastern Cape Province, South Africa between April 2017 and March 2018. The Great Fish River is located in Chris Hani District Municipality in the Eastern Cape Province and it is one of the major rivers used for irrigation and livestock farming in the area. This river is prone to agricultural and municipal runoffs and also serves as the receiving stream of effluents from many wastewater treatment plants (WWTPs), especially those situated in urban communities such as Craddock. Keiskamma and Tyhume rivers are located in the Amathole District Municipality in the Eastern Cape Province and are exposed to different anthropogenic activities from the rural and urban communities along the river courses, such as livestock drinking and irrigation farming. In addition, these rivers receive effluents from wastewater treatment plants (WWTPs) situated close to their banks. Different sampling points on these rivers were selected based on where humans and animals come into direct contact with them, for example, points where they are used for fishing, drinking and swimming purposes, downstream of the WWTPs, points where irrigation water is released to the water bodies and proximity to hospital facilities.
Water samples were collected at five different sites (S1, S2, S3, S4, and S5) from Great Fish, Keiskamma and Tyhume rivers respectively, for a period of one year, which covers the four seasonal patterns in South Africa (autumn, winter, spring, and summer). Water samples were collected aseptically in sterile 1L glass bottles from different sampling points by midstream-dipping of sample bottles at 25–30 cm down the water column, with the mouth tilting against the flow of the river. All water samples were labelled properly and safely taken to the laboratory (in an ice chest) where they were processed within 6 h of collection . Aliquots of water samples were used for isolation of Acinetobacter species based on standard microbiological procedures 
Isolation and purification of presumptive Acinetobacter species
The isolation of the presumptive density of Acinetobacter species in the water samples was determined by membrane filtration technique . Cellulose membrane of pore size 0.45 𝜇m was used to filter three volumes of 100 mL of the water samples under vacuum . These membranes were aseptically placed on plates with Acinetobacter species selective medium-CHROMagar Acinetobacter base plus selective supplement (CHROMagar, Paris, France) which was prepared according to the manufacturer’ instruction. Each sample plate was subjected to incubation at 37oC for 24 h after inoculation. Each sample was analysed in triplicate. All bacterial colonies with red colouration on the CHROMagar plates were counted as presumptive Acinetobacter species and were expressed as CFU/100ml. All isolates were sub-cultured on nutrient agar using a streak plate method (Oxoid, UK) and purified for further species identification. Fifty percent (50%) glycerol stocks of the pure culture was prepared and stored at -80 ∘C.
Molecular identification of Acinetobacter species by PCR assays
Extraction of genomic DNA: Presumptive Acinetobacter spp. in glycerol stocks was first being resuscitated on tryptic soy broth and incubated for 18 to 24h at 37 ∘C. DNA extraction from the bacterial isolates was carried out using the direct boiling method according to . The broth culture was centrifuged at 15000 rpm for 5 min using a Mini Spin Microcentrifuge (Lasec, RSA), then the supernatant was dispensed out and the pellet rinsed with sterile normal saline. The pellet was re-suspended in sterile distilled water and boiled in a heating block for 10 minutes using an AccuBlock (Digital dry bath, Labnet).
Amplification of unique Acinetobacter species DNA
Polymerase chain reaction (PCR) assay was used for the amplification of the Acinetobacter species recA gene as previously described . The forward and reverse primers used were P-rA1 (5′-CCTGAATCTTCTGGTAAAAC-3′) and P-rA2 (5′-GTTTCTGGGCTGCCAAACATTAC-3′) respectively. Briefly, an aliquot of 25 µl containing Taq PCR (12.5 µl) Master Mix (Qiagen, Hilden, Germany), each of the primers (1 µl) (Inqaba, SA), nuclease-free water (6.5 µl) and DNA template (5 µl) was used for the PCR amplification assay. The condition for the amplification included initial denaturation step (94 oC, 5 min), followed by 35 cycles (92 oC, 40 s), annealing (58 oC, 40 s), and the final extension step (72oC, 10 min) was performed using a thermocycler (Bio-Rad Thermal cycler, USA). Five microlitres (5 µl) of the amplicon was subjected to gel (1.5% agarose) electrophoresis at 100 Volts for 45 min in Tris Boric EDTA buffer (pH 8.0) (0.089 M Tris, 0.089 M boric acid, and 0.002 M EDTA). Ethidium bromide (5 µl of 0.5 mg/ml) (Sigma-Aldrich, USA) was used for gel staining and DNA ladder (100 bp) (Thermo Scientific, (EU) Lithuania) was added into the gels as a standard. Finally, DNA bands were visualized under an ultraviolet transilluminator (Alliance 4.7, France).
Delineation of Genus Acinetobacter into species
The confirmed Acinetobacter isolates were delineated into species accordingly [30, 52, 53] using a PCR assay. Firstly, optimization of conditions for A. baumannii, A. nosocomialis and A. pittii was carried out using species-specific primers (Table S1 under supplementary Section), while reference strains DSM-102929, DSM-102856, and DSM-9341 (DSMZ, Germany) were used as positive controls for A. baumannii, A. nosocomialis and A. pittii respectively. The PCR amplification was performed as stated in the previous section.
Detection of virulence genes
Polymerase chain reactions were also carried out for the identification of some Acinetobacter virulence genes including afa/draBC, epsA, fimH, OmpA, PAI, sfa/focDE, and traT genes, which have been previously found in clinical samples [54, 55, 56]. The standard strain of A. baumannii DSM-30007 (DSMZ, Germany) was used as positive control. There was no positive control available for A. nosocomialis. The PCR assay and electrophoresis were conducted as earlier described.
All statistical analyses were performed using the Statistica software v184.108.40.206 (64-bits). A simple factorial ANOVA was performed for the comparison of normally distributed data. The p-values of less than 0.05 were considered statistically significant for all the statistical tests performed.