High performance liquid chromatographic-grade methanol was purchased from Sigma Aldrich (Johannesburg, South Africa) while ultra-high purity water was prepared in the laboratory using Milli-QRO4 system 117 (Millipore, Bedford, MA, United States). The quantified antibiotics (sulfamethizole, sulfamethazine, sulfamethoxazole, trimethoprim and flumequine) were all analytical grade and supplied in powder form by Merck Chemicals (Pty) Ltd. (Johannesburg, South Africa).

In this study, a stream with effluents from a WWTP and leachates from a municipal dumpsite was investigated. The WWTP and the municipal dumpsite are within 1 km of each other (Figure 1). The stream pours into the Klip River, whose catchment is in western side of Johannesburg city and passes through a cultural township, Soweto. The stream however enters the river after the river has passed Soweto. There is a massive wetland at the point where the stream meets with the river (Figure 1). The Klip River has various wetlands along its length artificially designed for the removal of its pollutants.

The WWTP, the municipal dumpsite and the holding dam are all in a protected area and access by external persons is not permitted. Samples were, therefore, collected as the stream exits the holding dam (Site 1), within 5 km of the holding dam (Site 2), within 10 km of the holding dam (Site 3) as well as where the stream meets another from a separate WWTP just before entering the Klip River (Site 4). Site 4 is within 100 m prior to the stream-river confluence. Another sample was collected along the Klip River within a kilometer before meeting the stream from the WWTPs and the dumpsite (Site 5). Site 6 is within 5 km after the Klip River has combined with the effluent stream. Further down the river is Site 7 which is within 1 km of Site 6 but occurs after another large artificial wetland. The choice of sampling sites was influenced by the distance between the WWTP/dumpsite and sampling site of interest as well as accessibility into the stream.

Grab samples (100 ml) were collected at 30 min intervals over a 4 h period during the day from morning to afternoon, filtered through 0.45-micron hydrophilic polypropylene membrane filters obtained from Pall Corporation (Michigan, United States) and poured into the same 2 L brown bottle. The samples were then transported to the laboratory in cooler boxes where they were homogenized by shaking and immediately subjected to filtration for removal of suspended particles prior to solid phase extraction (SPE). The extracts of SPE were injected into the UHPLC-QTOF-MS system for analysis.

The automated SPE system was a Dionex™ Auto Trace™ 280 SPE from Thermo Fisher Scientific (Waltham, MA, United States). Extraction was performed using 200 mg Oasis HLB cartridges (6 cc) from Waters Corporation (Milford, MA, United States) as the sorbents. The cartridges were initially pre-conditioned with 5 ml of methanol followed by equilibration with 5 ml ultra-high purity water. Both conditioning and equilibration solvents were percolated through the cartridge at 1 ml min⁻¹. The water samples (500 ml) were then loaded at 15 ml min⁻¹. The cartridges were then rinsed with 2 ml of ultra-high purity water at 5 ml min⁻¹ followed by drying with a stream of nitrogen gas at 1 ml min⁻¹ for 5 min. Finally, the analytes were eluted with 8 ml methanol at 1 ml min⁻¹. The eluent was evaporated to near dryness with a slow stream of nitrogen gas and then re-constituted with 1 ml of 0.1% formic acid in methanol and taken to the UHPLC-QTOF-MS for analysis.

Health risk assessment was done based on the ratio of the maximum environmental concentration (MEC) to the predicted non-effect concentration (PNEC). The PNEC-resistance and PNEC-ecotoxicity that cater for antibiotic resistance and ecotoxicity, respectively were considered in predicting potential impact of the quantified antibiotics (Tell et al., 2019). The PNEC-resistance data represents values that would result in antibiotic resistance based on the lowest minimal inhibitory concentrations (MICs) (Bengtsson-Palme and Larsson, 2016). The lowest MIC values presented in the EUCAST database are for up to 25 families of clinical and environmental bacteria (https://mic.eucast.org/search/). The PNEC-ecotoxicity values were adapted from Bengtsson-Palme and Larsson 2016 and confirmed from the original source, the Farmaceutiska Specialiteter i Sverige (FASS) (https://www.fass.se/LIF/startpage) which is a Swedish pharmaceutical data base. The FASS data base does not have PNEC values for deregistered drugs. Both the EUCAST and the FASS databases were accessed between 25th March and sixth May 2021. The lowest value between the two parameters (PNEC-resistance and PNEC-ecotoxicity) and the highest measured environmental concentration of the antibiotic were therefore used as worst-case scenarios in estimation of the Risk Quotient (HQ) using Eq. 1. The criterion for RQ values is as follows; RQ ≤ 0.1 (environmental risk is low), 0.1 < RQ ≤ 1 (environmental risk is moderate) and RQ > 1 (environmental risk is high) (Rodriguez-mozaz et al., 2020). R Q =   M E C P N E C (1) where R Q is the Risk Quotient, M E C is the maximum environmental concentration and P N E C is the predicted non-effect concentration.

Matrix effects are known to affect instrument response towards target analytes (Rivera-Jaimes et al., 2018). In the current study, all the analytes experienced signal suppression effects with the flumequine most affected while the sulfonamides (sulfamethizole, sulfamethazine, sulfamethoxazole) and the diaminopyrimidine (trimethoprim) were less affected with signal suppression ranging between 4 and 19% (Table 1). However, of the sulfonamides, only sulfamethizole was least affected recording 4% suppression in signal response which was within the ±15% range considered acceptable in this study. In this regard, matrix-matched calibrations were performed for all the targeted antibiotics. For those calibration curves where the standard deviation of the model was greater than the y-intercept, the origin (point 0; 0) was set as the intercept. This was done for the calibration for trimethoprim. The linearity of the models was good ranging between 0.984 and 0.997. The method validation parameters are given in Table 1. Recoveries were recorded as averages for three spiking levels of 1, 10 and 100 µg L⁻¹ which equated to about 20, 200 and 2,000 times the limit of quantitation of most analytes, respectively. The recovery values were within the acceptable range (89–97%) and also had good repeatability of RSD ≤12%.

There were 15 antibiotics detected in the stream carrying wastewater effluents and dumpsite leachates of which 5 compounds (3 sulfonamides, 1 fluoroquinolone and 1 diaminopyrimidine) were quantified (Table 2). These numbers are comparable with other studies that did extensive screening elsewhere. For example, comprehensive monitoring of 53 antibiotics across Europe (Cyprus, Finland, Germany, Ireland, Norway, Portugal and Spain) in 2015/6 led to detection of 17 antibiotics (Rodriguez-mozaz et al., 2020), in China there were 17 antibiotics detected in WWTP effluents from a list of 50 targeted antibiotics (Zhou et al., 2013) while a study in the Netherlands investigated 52 antibiotics (Sabri et al., 2020). The quantified sulfonamides (sulfamethizole, sulfamethazine, sulfamethoxazole) in the current study ranged from not detected to 0.133 µg L⁻¹, the fluoroquinolone (flumequine) from 0.222 to 0.686 µg L⁻¹ while the diaminopyrimidine, trimethoprim was up to 0.0618 µg L⁻¹. Most of the antibiotics were only detected at the point of exit of effluents into the stream (Site 1) which was indicative of either dilution effects or possibly degradation by environmental conditions. Antibiotic reduction downstream has been reported elsewhere (Zhang et al., 2013). Wastewater dilution factors for South Africa are predicted at 10–40 (Keller et al., 2014).

While there are no guidelines on the release of antibiotics at the point of exit of effluents from wastewater treatment works, a previous study on species sensitivity and antimicrobial resistance has recommended 100 ng L⁻¹ as the maximum concentration for any antibiotic (Le Page et al., 2017). The concentrations reported for the selected antibiotics in the current study show that only trimethoprim and sulfamethoxazole were within the suggested value (Table 2). The implication is that the wastewater effluents and dumpsite leachates are releasing elevated concentrations of antibiotics into the surface water. Since the concentration of the other antibiotics detected in the effluents was not ascertained, there is a likelihood that more of these were also above the 100 ng L⁻¹ set point. It should be noted that 100 ng L⁻¹ set value is less than the PNEC values (Section 3.4) for most antibiotics detected in the current study except for Penicillin V whose PNEC value is 64 ng L⁻¹.

Sulfamethoxazole and trimethoprim are the frequently detected antibiotics in Africa and Asia (Faleye et al., 2018; Madikizela et al., 2020; Wang et al., 2020). Sulfamethoxazole has been detected in African wastewater effluents ranging between 0.15 and 10 μg L⁻¹ (Hendricks and Pool, 2012; K’oreje et al., 2016; Ngumba et al., 2016; Mhuka et al., 2020). In the current study, the sulfamethoxazole concentration was 0.0868 μg L⁻¹ which is lower than any reported concentration in Africa. Trimethoprim and flumequine in the current study also had lower concentrations compared to wastewater effluents reported in other African studies (Tahrani et al., 2017; Abou-Elwafa Abdallah et al., 2019; Mhuka et al., 2020). The wastewater effluents and the dumpsite leachates are first held in an artificial dam and release into the sampled stream mostly due to overflow. This might allow ample time for pollutants to undergo further degradation before release into the stream. While recent African reviews have observed that most environmental antibiotic concentrations are higher in Africa than anywhere in the world (Faleye et al., 2018; Madikizela et al., 2020), the reported concentrations for sulfamethoxazole (0.0868 μg L⁻¹) and trimethoprim (0.0618 μg L⁻¹) in the current study are comparable with those reported elsewhere. For example, in one study in China their concentrations in WWTP effluents were up to 0.106 and 0.064 μg L⁻¹, respectively (Zhou et al., 2013) while in Europe they ranged from not detected to 0.123 μg L⁻¹ (sulfamethoxazole) and 0.0152–0.190 μg L⁻¹ (trimethoprim) (Rodriguez-mozaz et al., 2020). Sulfamethazine is lower than that reported in effluents from a Japanese WWTP (Behera et al., 2011).

Our study also included 12 unregistered antibiotics (flumequine, sarafloxacin, ceftazidine, difloxacin, sparfloxacin, oleandomycin, oxacillin, oxolinic acid, sulfamerazine, sulfamethoxypyridazine, sulfanilamide, sulfapyridine). Antibiotic consumption in South Africa is regulated by the South African Health Products Regulatory Authority website; https://www.sahpra.org.za/registered-health-products/. Interestingly, flumequine and oxacillin were detected in the stream which indicates a possibility of illegal prescriptions or entrance into the country carried by tourists. The concentration of flumequine was almost constant in all sampling sites ranging between 0.222 and 0.686 µg L⁻¹. Flumequine is a broad-spectrum antibiotic and its detection in surface water maybe from veterinary applications rather than human consumption. Previous studies have shown that flumequine is persistent in the environment (Jara et al., 2021).

On a positive note, none of the derivatives of selected antibiotics were detected in the stream. The detection of the antibiotics downstream was limited with most being undetected in sites before the stream-river confluence. An exception was observed for the stable flumequine. A spike in the concentration of 5 antibiotics (flumequine, oxacillin, sulfadiazine, azithromycin, roxithromycin) was also observed at sampling site 4 where the study stream meets another stream from a small WWTP (Figure 1). Of these, azithromycin, roxithromycin were first detected at this point implying that their source was the small WWTP. In addition, roxithromycin and sulfadiazine already existed in the Klip River before receiving the effluents (Site 5) implying another source upstream. The Klip River passes through a dense township where proper sanitation is limited.

The health risk assessment data is summarized in Table 2. Of the 5 quantified antibiotics, only three (trimethoprim, sulfamethoxazole and flumequine) had PNEC values. Their calculated RQ values were 0.12, 0.15 and 2.7, respectively. These estimations imply that there is moderate risk for the environment due to trimethoprim and sulfamethoxazole while the risk is high for flumequine. This raises further concern regarding the other antibiotics that were detected but not quantified. Across Europe, for example, most of the reported concentrations of antibiotics are within PNECs (Carvalho and Santos, 2016) while only a few such as trimethoprim, azithromycin, clarithromycin and amoxicillin and erythromycin have been identified as antibiotics of interest with RQ ≥0.1 (Riva et al., 2019; Sörengård et al., 2019; Rodriguez-mozaz et al., 2020). Elsewhere in Asia, some studies have reported very high risks due to antibiotics notably sulfamethoxazole with an RQ value of 31.3 in surface water (Li et al., 2012; Hu et al., 2018).