In vitro screening for endocrine disruptive activity in selected South African harbours and river mouths

Various waterborne anthropogenic contaminants disrupt the endocrine systems of wildlife and humans, targeting reproductive pathways, among others. Very little is known, however, regarding the occurrence of endocrine disruptive activity in South African freshwater ecosystems, and coastal ecosystems have not been studied in this regard. In a first attempt to investigate endocrine disruptive activity in South African coastal waters, surface water samples collected from harbours, river mouths and estuaries in three metropolitan municipalities, eThekwini (which includes Durban), Nelson Mandela (specifically Port Elizabeth Harbour) and City of Cape Town, were screened for (anti) oestrogenicity and (anti)androgenicity using recombinant yeast bioassays. Moreover, levels of the female hormone 17β-(o)estradiol (E2) were determined in all samples, as well as a selection of hydrocarbons in the eThekwini samples. A high proportion of samples collected from eThekwini were oestrogenic, whereas none from Port Elizabeth Harbour and only a single river mouth sampled in the City of Cape Town were oestrogenic. E2 was detected in all the samples tested, but at higher concentrations at the eThekwini and City of Cape Town localities than Port Elizabeth Harbour. In addition, the recombinant yeast assays revealed that anti-androgenicity was widespread, being detected in the majority of samples screened apart from those representing Port Elizabeth Harbour. Conversely, no anti-oestrogenic or androgenic activity was detected. Anti-androgenicity did not associate with hydrocarbon loads, providing evidence that other anti-androgens were responsible for the observed activity. The present data suggest potential reproductive disruption in marine and estuarine fauna inhabiting the eThekwini and City of Cape Town regions.

in South African coastal waters. Studies on EDCs in the South African coastal environment are extremely limited, and apart from Marshall and Rajkumar (2003), who observed imposex in the mollusc Nassarius kraussianus, report on contaminant loads only (Fatoki and Awofolu 2004;Bollmohr et al. 2007;Ogata et al. 2009;Ryan et al. 2012;Wepener and Degger 2012;La Guardia et al. 2013). It is difficult, however, to accurately predict endocrine disruptive (biological) activity based on chemical data due to mixture interactions (Kortenkamp 2007); bioassays, in vivo exposures or animal tissue are more reliable predictors.
The aim of this study was to screen surface water from selected South African rivers, harbours and estuarine environments in three metropolitan municipalities on the South African coastline, namely eThekwini (which includes Durban), Nelson Mandela (specifically Port Elizabeth Harbour) and the City of Cape Town, for endocrine modulation effects (hormone receptor interaction) and female reproductive hormone contamination. The specific objectives were to measure (anti)oestrogenic and (anti)androgenic activity in surface water samples using recombin ant yeast bioassays (interaction with human steroid receptors), and environmental concentrations of the female hormone 17b-(o)estradiol (E2).
A further aim was to explore the association between hydrocarbon loads and endocrine disruptive activity in surface water samples. For this purpose the concentrations of a selection of hydrocarbons were determined in Durban Bay and selected rivers and estuaries in the greater eThekwini metropolitan area.

Sample collection and extraction
Surface water was collected at 10 localities in the eThekwini region in August 2012, and three and six locations were sampled in the Nelson Mandela (Port Elizabeth Harbour) and City of Cape Town regions, respectively, during October 2012 (Figure 1, Supplementary Table S1, available online).The samples were collected in 500 ml acid-cleaned, amber glass bottles with PTFE-lined caps, kept on ice and shipped to the laboratory within 24 h. The samples were subsequently filtered through 0.5 µm glassfibre filters (MN 85/90 Macherey-Nagel, Germany), and the pH adjusted to 3 using H 2 SO 4 . Non-polar and slightly polar compounds were extracted from 250 ml of water within 96 h of collection using 500 mg DSC-18 columns and a Visiprep ® manifold system (~10 ml min -1 ) (Sigma, South Africa). The columns were flushed with 50 ml of Milli-Q water directly after environmental samples were passed through, in order to remove salts. A Milli-Q negative control was included during each extraction event. The SPE columns were subsequently air-dried overnight, after which compounds were eluted from the column using a solvent mixture (40% hexane, 45% methanol and 15% 2-propanol), air-dried, reconstituted in absolute ethanol and stored at −20 °C.

In vitro recombinant yeast assay
Oestrogenic, anti-oestrogenic, androgenic and anti-androgenic activity of surface water C18 extracts (representing non-polar and slightly polar compounds) were evaluated using recombinant yeast bioassays (Routledge and Sumpter 1996;Sohoni and Sumpter 1998). The yeast culture and exposure was performed as described by Sohoni and Sumpter (1998) Supplementary Table S1 (available online) plate contained a 12-point serial dilution of E2 (≥98% pure), tamoxifen (TAM) (≥99% pure), dihydrotestosterone (DHT) (≥97.5% pure) or flutamide (FLU) (Sigma, South Africa) as standards, and a solvent control. Oestrogen and androgen receptor antagonism were evaluated in the presence of 1.43 nM E2 and 7.13 nM DHT, respectively. Oestradiol-, dihydrotestosterone-, tamoxifen-and flutamide equivalents were calculated for environmental samples using standard curves generated during experiments (see supplementary material for more information). These equivalents express biological activity (i.e. hormone receptor agonism [E2 and DHT] or hormone receptor antagonism [FLU and TAM]) equivalent to the activity associated with the particular hormone or pharmaceutical standard represented. Cell densities were determined spectrophotometrically at 620 nm absorbance (Abs). Sample-containing wells with Abs 620nm below a specific cell density/viability benchmark (i.e. solvent control mean Abs 620 (solvent control) − 3[standard deviation Abs 620 (solvent control) ]) were not included in the analyses due to potential inhibitory effects.

Enzyme-linked immunosorbent assays (ELISAs)
17b-oestradiol levels were measured in the C18 SPE extracts of water collected from the 19 localities using a commercially available ELISA kit (DRG International Inc., USA) according to the manufacturer's instructions. The extracted samples in ethanol (2 000× concentrated) were diluted 1/20 in a 0.1% w/v human serum albumin and 0.9% NaCl solution, and assayed (Swart and Pool 2007).

Statistical analyses
The correlation between E2 concentration and oestrogenicity was evaluated using the Spearman Rank test (Statistica 12, Statsoft, USA). Potential associations between hydrocarbons and anti-androgenic activity among sampling locations were evaluated using a principal component analysis bi-plot (Canoco 5, Microcomputer Power, USA). A p-value < 0.05 was considered significant.

Results
Oestradiol was detected in all samples analysed, although at concentrations <1 ng l -1 in Port Elizabeth and Cape Town harbours and in samples from the Lourens (LOU) and Palmiet (PAL) estuaries ( Figure 2). The highest E2 concentration was detected in the mouth of the Salt River (SLT) in Cape Town (20.96 ng l -1 ), followed by the Isipingo Estuary (ISI) in the eThekwini region (17.41 ng l -1 ). Oestrogenicity (binding to the human (o)estrogen receptor [ER]) was detected in seven of the 19 samples analysed (Figure 3). The eThekwini region had the highest proportion of oestrogenic samples, with activities ranging between 1.33 and 8.01 ng l -1 E2 equivalents (EEQs) (Figure 3). Conversely, no activity was detected in the Port Elizabeth Harbour samples. The sample collected at the mouth of Salt River in Cape Town was the most oestrogenically active, with an EEQ of 12.82 ng l -1 . Although the EEQs ( Figure 3) followed a similar trend as E2 concentrations (Figure 2), for example, being highest in the mouth of the Salt River, EEQs were generally lower than E2 concentrations measured using ELISA. There was a significant correlation between E2 concentrations and oestrogenic activity when the three study regions were evaluated collectively (Spearman rank r = 0.82, p < 0.05).
Anti-androgenicity (inhibition of androgen binding to the human androgen receptor [AR]) was more widespread than oestrogenicity and detected in 14 of the 19 samples analysed (Figure 4). All samples collected in the eThekwini region were anti-androgenic, with flutamide equivalents (FEQ) ranging between 89.10 and 604.44 µg l -1 . None of the Port Elizabeth Harbour samples were anti-androgenic. Similarly, no inhibition of androgen binding to AR was observed in the Cape Town Harbour samples, whereas all four estuarine and river mouth samples collected in the City of Cape Town region were anti-androgenic, ranging between 59.60 and 196.86 µg FEQ l -1 (Figure 4).
No anti-oestrogenic or androgenic activity was detected in any of the samples analysed.
A bi-plot of a principal component analysis of hydrocarbons and anti-androgenicity indicates three major groupings for sampling locations in the eThekwini region ( Figure 5). In particular, locations AMA and ISI1 (see Figure 1) grouped together and were associated with the majority of TPHs analysed, PAH isomers pyrene, fluoranthene, anthracene and phenanthrene, and the sum of all PAH isomers. In addition, locations DBAY1, DBAY2, IVC, UML2, UMB and ISI2 grouped together and were more correlated with anti-androgenicity than other samples analysed. Anti-androgenicity was not closely correlated to any of the hydrocarbons analysed ( Figure 5). Hydrocarbon concentrations are presented in the supplementary material (Tables  S2 and S3). All BTEX components in the eThekwini study area were below the method detection limit (Table S2). Conversely, the total petroleum hydrocarbon (TPH) analyses show the presence of TPHs at all the locations sampled, apart from UMH and UML1 (Table S2). The highest diversity of PAHs was detected at the IVC locality in Durban Bay (total PAH: 0.22 µg l -1 ), followed by AMA, which had the highest abundance of PAHs in the study area (total PAH: 0.28 µg l -1 ) (Table S3). However, virtually no PAHs were detected at the other locations sampled (Table S3).

Discussion
Published records providing concentrations of organic pollutants such as pesticides and industrial chemicals in the South African coastal environment are extremely limited (Wepener and Degger 2012). Moreover, to our knowledge, no study of the South African coastal environment has described biological endpoints of endocrine disruptive activity, apart from Marshall and Rajkumar (2003), who observed imposex in a mollusc species.
The present study confirms oestrogenic and antiandrogenic activity in surface waters collected in Durban Bay and a number of rivers, river mouths and estuaries in the eThekwini and Cape Town regions. Conversely, no such activity was detected in the Cape Town and Port Elizabeth harbours, indicating good quality in terms of EDCs targeting  Figure 1 for sampling locations). Error bars indicate the standard deviation among assay plates. The horizontal dashed line indicates the predicted no-effect concentration (PNEC) for reproductive impairment in fish (Caldwell et al. 2012) oestrogen or androgen receptors. Durban Bay is a highly transformed estuarine embayment that receives inflows from three rivers and surface runoff from a multitude of stormwater outfalls and canals. Conversely, Port Elizabeth Harbour receives inflow from a single small river, historically an estuary, transformed to built-up land (Veldkornet et al. 2015), and subject to urban surface runoff. There is no riverine inflow to Cape Town Harbour. The higher oestrogenic and anti-androgenic activity observed in Durban Bay may, therefore, be explained by the sheer number of anthropogenic sources of contaminants entering the bay via its large catchment, compared to the Cape Town and Port Elizabeth harbours. The oestrogenic activity detected in the eThekwini and City of Cape Town regions (1.33-12.82 ng EEQ l -1 ) is within the range reported for Chinese rivers and estuaries (Zhao et al. 2011;Rao et al. 2013). Studies in other countries have typically reported lower activity, such as in the Rhine River, Germany (1.1-1.3 ng EEQ l -1 ) (Pawlowski et al. 2003), the Shannon River basin, Ireland (0.53-2.67 ng EEQ l -1 ) (Kelly et al. 2010), the Seine and Oise rivers, France (0.30-4.52 ng EEQ l -1 ) (Cargouet et al. 2004), and in a nationwide study in the Netherlands (0-0.17 ng EEQ l -1 , n = 90 samples which included estuaries) (Vethaak et al. 2005). The E2 levels detected throughout the three study areas correlated well with oestrogenic activity, suggesting that the observed oestrogenicity was largely due to female hormone contamination and therefore likely linked to waste water treatment plant (WWTP) discharges.
The anti-androgenic activity detected in the eThekwini and City of Cape Town regions exceeded 100 µg FEQ l -1 at 12 of the locations sampled, and although relatively high, they were similar to activities reported for surface waters elsewhere. Some examples include the Zhujiang River estuary, China (135 µg FEQ l -1 ; Zhao et al. 2011), the Lambro River, Italy (438.15 µg FEQ l -1 ; Urbatzka et al. 2007), and the Ray River, England (5-250 µg FEQ l -1 ; Grover et al. 2011). In a modelling study, FEQs of 0-100.12 µg l -1 were predicted for 30 rivers in the United Kingdom (Jobling et al. 2009).
Both oestrogenic and anti-androgenic chemicals may affect reproduction by altering sex organ development and function (Arukwe 2001), which may have severe consequences for fish populations (Kidd et al. 2007;Jobling et al. 2009). Oestradiol concentrations as low as 4 ng l -1 induce the development of ovarian tissue in testes in Japanese medaka Oryzias latipes (Metcalfe et al. 2001), and the predicted no-effect concentration (PNEC) for this compound in terms of reproductive impairment in fish has been estimated as 2 ng l -1 (Caldwell et al. 2012). Oestradiol levels exceeding the PNEC for fish were detected in 10 of the 19 samples analysed in the present study. The pharmaceutical anti-androgen, flutamide, has been shown to disrupt the reproductive system of Asian catfish Clarias batrachus at 33 µg l -1 Rajakumar et al. 2012), Murray rainbowfish Melanotaenia fluviatilis at 125 µg l -1 (Bhatia et al. 2014), and fathead minnow Pimephales promelas at 320 µg l -1 (Filby et al. 2007). The fish populations inhabiting a large proportion of the systems evaluated in the present study, including Durban Bay, are therefore at risk of endocrine disruption, potentially resulting in reproductive disorders such as intersex and impaired reproduction, due to the combined action of oestrogens and anti-androgens. Future studies evaluating phenotypic endpoints such as male plasma vitellogenin levels (Vethaak et al. 2005), the expression of marker genes integral to endocrine signalling (Truter et al. 2014), and histopathology of the gonads (Allen et al. 1999) are necessary, however, to confirm the presence of endocrine disruption. Moreover, seasonal investigations of potential endocrine disruptive activity will be of great value as baseline data and to aid conservation efforts.
Petrochemical pollution is an important component of contamination in harbours (Mestres et al. 2010). PAHs have been detected in sediment and mussels in the ports of Cape Town, Port Elizabeth and Durban, and in sediment in rivers and estuaries in the greater eThekwini and City of Cape Town areas (BKN, unpublished data; see also Degger et al. 2011;Kampire et al. 2015). The anti-androgenic activity observed in the present study may have been caused by petrochemicals, which are known to be potent androgen receptor antagonists (Kizu et al. 2003;Vrabie et al. 2010). The weak correlation observed between anti-androgenicity and hydrocarbon concentrations in the samples from the eThekwini region ( Figure 5), however, provides some evidence that androgen receptor antagonists other than hydrocarbons were responsible. Environmental anti-androgens include pesticides, pharmaceuticals and industrial chemicals (Korner et al. 2004). Detailed chemical analyses are required to identify more potential anti-androgens in Durban Bay and the surrounding rivers that were sampled.
The present study is the first on endocrine disruptive activity in the South African coastal environment. It indicated that waters in Durban Harbour, at the Umlazi Canal mouth (UML) and in the Isipingo River estuary, and at the mouth of the Salt River in Cape Town, were oestrogenic. Moreover, widespread anti-androgenicity was observed in Durban Bay and selected rivers in the eThekwini region, and in three estuaries and a river mouth in the City of Cape Town region, although at a lower potency than in the eThekwini region. Conversely, no endocrine disruptive activity was detected in Port Elizabeth Harbour. The levels of oestrogenic and anti-androgenic activity observed are high enough to suggest an effect on the reproduction of certain fish species, and further investigation in this context is required.