Polycyclic aromatic hydrocarbons (PAHs) in the aquatic ecosystems of Soweto and Lenasia

XXIII Chapter 1: General and thesis introduction 1.1 General introduction 1 1.2 Hypothesis, aims and objectives 3

Unacceptable risk for ingestion is greater than 1 in 10 000 (1 x 10-4) (red line)   Clarias gariepinus and wetland bird eggs, including the limit of detection (LOD) and limit of quantification (LOQ) expressed in μg/kg   and benzo(a)pyrene (Tsai et al., 2004) as well as fish potency factors (Barron et al., 2004) used to calculate toxic equivalence quotients Table 2.7 Diagnostic ratios for source identification of polycyclic aromatic hydrocarbons  Lenasia (2013Lenasia ( & 2014 in terms of the MacDonald et al. (2000) guidelines and the CCME (2012) guidelines.
Shading according to index scale  Lenasia (2013Lenasia ( & 2014, compared to the TEQ guidelines of the CCME (2001). Guidelines exceedance indicated by shading   Table 4.1 Sex ratio, mean mass, mass range, mean total length, and total length range of Clarias gariepinus sampled in Soweto and Lenasia (2013 and. Standard deviation in parenthesis where means are reported   Van der Oost et al., 2003 Chapter 5  Table 5.2 Exposure parameters for human health risk assessment for adults and children (USEPA, 2004;Health Canada, 2004) adapted to represent the List of tables XXI population in the study area Table 5.3 Risk characterisation variables for dermal-and oral exposures to soils, water and fish (USEPA, 2004;Health Canada, 2004)  Chapter 6 Table 6.1 Levels of total polycyclic aromatic hydrocarbons in the sediments (ng/g) of Soweto and Lenasia against the scale by Baumard et al. (1998) pyrene, benzo(a)anthracene, chrysene, benzo(a)pyrene, benzo(b)fluoranthene, benzo(k)fluoranthene, dibenz(a,h)anthracene (USEPA, 2008) [The last 6 compounds in the list are regarded as human carcinogens (NTP, 2005)].
The total global anthropogenic release of the 16 USEPA priority PAHs to the atmosphere was estimated to be at 520 000 tons in 2004, of which 18.8% was emitted from Africa (Zhang & Tao, 2009). An appreciable amount of this is expected to return to the surface through deposition and may wind up in the aquatic system. The direct PAH emissions to soil, water and sediment is not known, and there is little data for South African freshwaters (Das et al., 2008;Quinn et al., 2009;Nieuwoudt et al., 2011;Roos et al 2012;Moja et al., 2013;Nekhavhambe et al., 2014). We therefore know that PAHs are present in the South African environment, specifically in the section of the Vaal River catchment running through the Vaal Triangle (Nieuwoudt et al., 2011;Moja et al., 2013). The total concentration of PAHs in the former study ranged between 44 and 39 000 ng/g, dry mass (dm) and the concentration of carcinogenic PAHs ranged between 19 and 19 000 ng/g, dm (Nieuwoudt et al., 2011). The concentrations of native congeners in the water ranged between 23. 5 and 110.8 µg/L (Moja et al., 2013). Pyrogenic (burning) processes were the most likely sources, with minimal petrogenic (derived from fuels and oils) contributions (Nieuwoudt et al., 2011;Moja et al., 2013). PAH levels were in the same range as levels reported from other countries.
In the study completed for the Water Research Commission (Project no K5/1561) on POPs in freshwater sites throughout the entire country, the PAHs had the highest levels of all of the organic pollutants analysed for. One of the sites with the highest PAH levels, was within the Soweto and Lenasia urban area, with 5 408 ng/g (Roos et al., 2012). The cumulative probability of developing cancer resulting from exposure to benzo(a)pyrene at this site as a result of exposure to fish contaminated with benzo(a)pyrene was calculated to be between 0.181 and 0.859 in 1 000. [This can be rounded off to 2 in 10 000 and 9 in 10 000]. This is much higher than what is considered as an acceptable risk (approximately 6 in 10 000 versus the acceptable risk of 1 in 100 000 of the WHO (2010)].

Hypothesis, aims and objectives
The findings of Roos et al. (2012) led to the need to investigate the Soweto and Lenasia area in more detail as it showed to be experiencing high PAH exposures and therefore lead to this study. Thus, the main aim of this study was to determine the levels of the 16 priority PAHs in the Klip River that flows through the densely populated urban areas of Soweto and Lenasia, where high levels were previously found to have a more detailed view of PAH pollution and exposure in this particularly populated area of the country.

Hypothesis
Humans and wildlife from Soweto and Lenasia dependant on the Klip River are exposed to the 16 priority PAHs.

Aims and objectives
The hypothesis was investigated along the following aims: Aim 1: Determine the levels of the 16 priority PAHs in the Klip River that flows through the densely populated urban areas of Soweto and Lenasia where high levels were previously found.

Objectives:
Measure concentrations of 16 PAHs in sediment at 9 sites over a two year period Measure concentrations of 16 PAHs in fish tissue at 4 sites over a two year period Measure concentrations of 16 PAHs in wetland bird eggs over a two year period Aim 2: Investigate the pollutant profile of 16 PAHs in the sediment

Objectives:
Compare site PAH composition percentages by grouping congeners with the same number of cyclic rings to investigate similar pollution profiles between sites.
Calculation of diagnostic ratios to determine origin of the pollution, i.e. pyrogenic vs.
petrogenic Aim 3: Determine the toxicity posed by the PAHs in the study area

Objectives:
Assessing sediment toxicity to benthic organisms, by comparing to international sediment quality guidelines and calculating sediment quality indices Investigating a very specific form of toxicity: that of the aryl hydrocarbon receptor mediated toxicity, in sediment using the H4IIE-luc reporter gene bioassay Aim 4: Investigating biochemical responses of the fish to the environmental stressors by performing biomarker response assays

Objectives:
Determining the levels biomarkers of exposure (acetylcholinesterase activity and total cytochrome P450s levels), to identify direct effects of xenobiotic stressors on the fish.
Determine the levels of biomarkers of oxidative stress (superoxide dismutase and catalase activity) and oxidative damage (malondialdehyde and protein carbonyl content) to identify potential oxidative stress and associated damage in the fish Determine the energy budget (available carbohydrates, lipids and proteins and available energy) to indicate if the stressors on the fish affect the cellular energy allocation

Aim 5: Investigating individual and community fish health by applying health indices for fish
Objectives: Calculating Fulton's condition factor to describe the overall condition of individual fish and the population Determine the organo-somatic indices for the liver, spleen and gonads, to investigate physiological endpoint deviations with in individuals and the population Applying the fish health assessment index to determine the overall health of individuals and the community Aim 6: Gauging potential risk to human health by conducting a theoretical human health risk assessment

Objectives:
Determining non carcinogenic risk using the Hazard Index for dermal and ingestion exposure to the various matrices Calculating cancerous risk from dermal exposures and ingestion of the various matrices

Study area-Klip River catchment 1.3.1 The Klip River catchment
The Klip River catchment is situated in South Africa's most densely populated province Gauteng, and drains the Witwatersrand region, the southern part of Johannesburg, one of the most developed urban areas in Africa (Kotze, 2002, DWAS 2009

Figure 1.1: Klip River catchment showing the Klip River and its tributary from origin to confluence with the Vaal River
Domestic users of the Klip River mainly include rural settlements along the Klip River and its tributaries. The water utility, Rand Water, supplies potable water from the river to various municipalities in the catchment (Howie & Otto, 1996;Kotze, 2002). Industrial use of the water (Regions 1 & 2) is restricted to the middle reaches of the catchment. Main users are processing industries, such as product packaging, roofing and cladding material production, three waste water treatment plants and mining activity (Kotze, 2002). Industrial water is also supplied by Rand Water.
Mining (gold, base metals and industrial minerals) is the most important activity in the upper catchment of the Klip River (DWAS, 2012). Agricultural activities such as livestock watering and crop irrigation also use water in the catchment (Kotze, 2002).

Because the Klip River flows through the Witwatersrand region it is considered as one of South
Africa's most polluted rivers (McCarthy & Venter, 2006;McCarthy et al., 2007). The mining activities and WWTPs in the catchment act as primary sources of point pollution. The sources of diffuse pollution mainly consist of informal settlements and old mine slime dams/waste dumps (Kotze, 2002).
A summary of the potential pollution in the Klip River was compiled by Kotze (2002)

Site selection
Of all the compounds investigated by Roos and co-authors (2012) in their WRC study, the PAHs were one of the most abundant. Their report recommended specifically that further investigations were needed at areas where the highest potential risks had been calculated. Soweto and Lenasia was one of these areas, which was also one of the areas with the highest PAH levels (Roos et al., 2012). From the literature on the Klip River catchment-indicating the urbanisation, industrialisation and the pollution sources-as well as the recommendations made by Roos et al. (2012), the study area

Figure 1.2: Sampling sites within the greater Soweto and Lenasia area
Sediment samples were collected from nine sites within the study area (Table 1.2). Two of these sites formed part of the Klip River upstream of the Klip Spruit confluence. Six sampling sites were located on the Klip Spruit (Region 2) and its smaller tributaries, and one site in region 3 (Figure 1.1). Due to the geographic nature and availability of fish at the sediment sites, fish was sampled only from the sediment sites able to produce fish (Table 1.2). These fish sampling sites were also chosen to represent the different areas within the study area. The Nancefield weir (Nc) (Figure 1.2) was one of these fish sampling sites-representing the farthest downstream area-however after the first sampling session was unsuccessful an alternative site within the area had to be identified.

Protea Glen [PG]
The sampling site at Protea Glen is the most western site of the project (Figure 1.3). It is located in the Protea Glen/Naledi residential area. This site is located on the Klip River and represents the most upper part of the Klip (upwards to the origin). Only sediment was collected from this site. The Klip River at PG flows through a large wetland. Sediment was collected where a public road transects this wetland (Figure 1.3A). The river (where sediment was collected) has a slow deep flow (Table 1.3) and the bottom consists of soft mud and large rocks. The riparian zones consisted of mainly wetland reeds, trees and shrubs (Figure 1.3B). Pollution at this site noted included residential waste and litter.

Lenasia [Le]
The Lenasia site, together with Protea Glen forms the western sites. It is downstream from PG ( Figure   1.2). At this site sediment, fish, and bird eggs were collected. It was also the only site where eggs were collected. The Lenasia site is located at a dam that forms part of the Lenasia wetland. The Klip River enters this system at the dam and flows through a series of small impoundments and wetland patches before exiting downstream from Lenasia. Fish and sediment were collected from the dam.
This dam is a deep impoundment with large patches of water grass and weeds, and most of the shore is covered with reeds ( Figure 1.4A). The sediment of the dam has a granular muddy consistency and was filled with small stones. The heronry where the eggs were collected was around a small opening of water in a dense reed bed, which is connected to the main dam (Figure 1.4B). Notable pollution seen at the site included: domestic garbage, construction rubble, condoms (suggests sewage leakage) and evidence of burnt tyres.
A B

Fleurhof [Fl]
Fleurhof Dam is the most northern site of the study area ( Figure 1.2), situated in the residential area of Fleurhof, bordering on the industrial areas of Lea Glen, Robertville, and Vogelstruisfontein. The dam is a large impoundment that drains into the Klip Spruit to the south-east. Sediment and fish were collected from this site. The dam had a weed covered bottom and was surrounded by reeds ( Figure   1.5A). The sediment collected was sandy. Pollution seen at the site included an informal waste dump and construction rubble. Numerous developments were under construction around the dam ( Figure   1.5B), and as the dam is situated lower/downhill of these sites, the run-off from these sites drained into the dam. The western banks of the dam are next to old mine dumps in the Vogelstruisfontein area ( Figure 1.5C). Fleurhof Dam is connected to Florida Lake (to the north) by means of a cement channel. The cement was to decrease water seepage into two mine reefs and old mine workings (DWAS, 2010).

Dobsonville [Db]
This site is situated to the north west of the Klip Spruit (Figure 1.2) in the residential area of Dobsonville. This site, where only sediment was collected, is in the Dorothy Nyemba Park (maintained

A B
A B C by Johannesburg City Parks). The small dam receives run-off from the surrounding area and the nursery that is located in the park, and then drains into a small stream that flows into the Klip Spruit.
The dam is shallow and very small relative to the other sampling dams. The shores of the site were lined with reed beds (Figure 1.6A) and the sediment was firm clay mixed with small rocks. The dam was polluted with municipal garbage littering the edges of the shore and the reed beds (Figure 1.6B).

Orlando East [OE]
Orlando East is one of the eastern sites of the project, part of the Klip Spruit tributary (Figure 1.

2).
This site is located to the south of Orlando East and west of Diepkloof at the iconic Orlando Towers

Eldorado Park [ElD]
Eldorado Park is the most downstream sampling site on the Klip Spruit (Figure 1.2) and is located between Eldorado Park and Klipspruit West. The sampling was done below the bridge of Klipspruit

Nancefield and Bushkoppies Waste Water Treatment Plant (WWTP) [Nc]
The sites at Nancefield are the most southern sites of the project (Figure 1.2). This area has two sampling areas because fish could not be found in the weir in the Klip River. Bushkoppies WWTP (operated by Joburg Water) was chosen as a substitute for fish sampling as it is in the same vicinity as the sediment sampling site and it drains into the Klip River just downstream from the weir. The sediment site on the Klip River is situated next to the Golden Highway in the Nancefield area and is downstream from a wetland and the Olifantsvlei WWTP. The river is narrow and deep downstream from the weir ( Figure 11A) and becomes shallow and fast downstream from the bridge (Figure 1.11B).
Here the stream flows at 1.5 m/s and has large boulders in the granular yet muddy sediment. The area where sediment was sampled is directly downstream of a demolished bridge (Figure 1.11B). The rocky bank of the river is lined with reeds ( Figure 1.11A), grass and trees (Figure 1.11B). Although litter was not prominent, this site is used by locals for various religious ceremonies and is also a local fishing spot. The fishing site at the Bushkoppies WWTP was strategically chosen as the most downstream sampling site. The complex is located next to the N1, N12 and Golden Highway, in the Nancefield district. To eliminate the potential influence of the plant, only fish was collected from the last set of ponds on the property that drained into the Klip River. These dams were deep and had rocky shores lined with Sweet thorn trees, grass and patches of reeds (Figure 1.12). The visible pollution at this site was minimal as it is a restricted area.

Use of sediment as abiotic matrix
The use of sediment quality for evaluating the levels of contaminants is necessary as it assists in assessing potential toxicity in a system (Chakravarty & Patgiri, 2009). Numerous studies have investigated environmental health by studying the concentrations of pollutants in the sediment (Angulo, 1996;Atgin et al., 2000;Meybeck et al., 2004;Nieuwoudt et al., 2009). Organic pollutants in aquatic ecosystems generally exist in low levels in water (Morillo et al., 2009) and accumulate in the sediments (Reid et al., 2000;Zhang et al., 2003). Sediment has a long residence time in the aquatic systems (Saha et al., 2001;Varol, 2011). Due to their variable physical and chemical properties, sediments are important sources for organic and inorganic pollutants (Praveena et al., 2008). Thus because sediments are ideal sinks for PAHs and other pollutants discharged into the environments, they can reflect on the input of point-and non-point sources (Hahladakis et al., 2013;Yohannes et al., 2013). During favourable conditions, sediments play a functional role in the mobilization of contaminants in aquatic systems (Eggleton & Thomas, 2004;Li et al., 2016). In riverine communities, the population is directly and indirectly exposed to sediment and so the pollutants, and are at risk of contamination (Miller et al., 2004). Thus it is important to evaluate the pollution status of the sediment.

Sampling and processing of sediment
Surface sediment samples were collected at the nine identified sites (Figure 1.2) during low flow conditions (June/July) during 2013 and 2014. Three sediment subsamples were collected in a 5 m range at each site using pre-cleaned stainless steel hand shovels. A composite sample was prepared by thoroughly mixing the subsamples in a pre-cleaned stainless steel bowl before storing it in high density polyethylene (HDPE) bottles, pre-cleaned (rinsed thrice with first acetone and then hexane) according to USEPA method 1613(USEPA, 1994. The samples were protected against microbialand UV degradation by transporting at 4°C and storing at -20°C in the laboratory. The samples were air dried in the absence of light, ground, and sieved (0.5 mm mesh size) to obtain homogenous samples (Kralik, 1999).
To determine the physical characteristics of the sediments, total organic carbon and grain size was determined. The loss-on-ignition method (Schumacher, 2002) was done on a TruSpec CN analyser to determine the total organic carbon. The different sediment grain sizes was partitioned using a series of sieves (4 000, 2 000, 500, 212, 106, 53 µm) (ISO, 2002).

Fish used as biotic matrix for environmental studies
Fish are an ideal biotic matrix of aquatic environment studies, as they are represented in various trophic levels in aquatic food webs (Kidd et al., 2001). Biota act as bio-indicators-which are groups or individual organisms that are used to describe the quality of an ecosystem, depending on their abundance or well-being (Gerhardt, 2002). Disturbances in the lower levels will affect the apex predators, as they feed on prey in lower levels of the food web (Kidd et al., 2001). Fish have been used in numerous studies as biotic matrix, investigating organic pollutants (Vives & Grimalt, 2002;Weber & Goerke, 2003;McHugh et al., 2011;Wepener et al., 2011). The sharptooth catfish (Clarias gariepinus) is an opportunistic omnivorous bottom-feeder and have also been found to be intentional detritus feeders, but are known as formidable predators, as well as a hardy and resilient fish, surviving harsh conditions (Skelton, 2001).
Clarias gariepinus was chosen as an indicator species because of its abundance in South Africa, their hardiness and because they are an apex predator. Their position on the food web and its preference for bottom-dwelling in the aquatic systems makes it ideal to study exposure to pollutants, as well as the bio-accumulation and bio-magnification of organic chemical pollutants. These fish are also a valued food source, allowing for investigation in possible transfer of pollutants to humans.

Use of bird eggs as biotic matrix for environmental studies
Birds are also popular biotic matrices for environmental studies and have been used in various studies investigating organic pollution (Custer et al.,2001;Barnhoorn et al., 2009;Pereira et al., 2009;Quinn et al., 2013;Khan et al., 2014). Piscivorous birds represent a different trophic level than fish, and because organic pollutants bio-accumulate and bio-magnify within food webs (Zhou et al., 2006;Antoniadou et al., 2007;Herbert et al., 2011), they can show the trophic transfer between different biotic matrices (fish to birds). The use of bird eggs specifically is considered to be a better matrix than the adult organism itself. Eggs are easy to handle and uses a relatively fast and non-invasive method of collection (Medvedev & Markova, 1995). The eggs are representative of the female parent-as contaminants is transferred from the parent bird to her lipophilic eggs (Van den Steen et al., 2006;Verreault et al., 2006)-reflecting the pollutant body burden of that female parent (Braune, 2007).
Wetland birds were chosen as bio-indicators for this study because they are exposed to pollutants from their feeding regimes, their direct habitat selection (aquatic systems) and breeding behaviours.
Various types of wetland birds occur in the study area. The only heronry identified in the study area was at Lenasia. Individual target species were not identified beforehand and the species present at the heronry were sampled.

Clarias gariepinus
Air breathing catfish, from the family Clariidae, form part of one of the world's most abundant fish types (order Siluriformes) (Bruton, 1988;Skelton, 2001). Clarias gariepinus is found throughout Africa and parts of Asia and is the most widely spread freshwater species in the world (longitudinally) (Bruton, 1988;Skelton & Teugels, 1992). Its distribution in Africa ranges from as far north as the Nile River and as far south as the Orange-and Umtamvuna systems. In South Africa it has been translocated to the Eastern Cape (Sundays-, Fish-and Keiskamma Rivers) and the Western Cape (Cape Flats and Clanwillam-Olifants River) (Skelton, 2001).
Clarias gariepinus is a dorso-ventrally flattened fish; the body is compressed towards the tail and  Clarias gariepinus is completely omnivorous, it preys and scavenges on any available organic food source including fish, birds, frogs, small mammals, reptiles, molluscs, crustaceans, seeds, fruit and even plankton. They may hunt in packs, herding and trapping small fish in shallower water. Clarias gariepinus breeds in the summer after rain showers when large sexually mature adults migrate to flooded shallow grassy borders of the dam or river (Skelton, 2001).

Sampling of fish
Clarias gariepinus was sampled during the high flow seasons (October) of both 2013 and 2014. Fish were sampled using gill nets (118 and 150 mm) and, rod and line. The gill nets were checked every two hours and caught fish were kept in an aerated container until field analysis was performed. The fish (n = 20; 1:1 F:M) used for the control were sampled from various other sites in the Vaal River basin (Boskop Dam, Klerkskraal Dam and Vaal River) and were kept at the Water Research Group aquarium at the NWU (Potchefstroom Campus) for 6 months. Fish were kept in 5000 L recirculating tanks with added aeration. They were kept in aquarium standard water and temperature (CCAC, 2005). The water was replaced bi-weekly.

Herons and egrets
Herons and egrets are non-swimming water birds belonging to the Ardeidae family. These birds are described as tall, slender birds with long legs and -necks (Sinclair et al., 2011). They have dagger-like bills and are mostly aquatic hunters, relying on stealth and occasionally by means of active pursuit (Sinclair et al., 2011). Two species from the family Ardeidae was found at the sampling area: the  Ardea melanocephala is common and widespread throughout sub-Saharan Africa. They are found in various habitats such as marshes and floodplains, although they are not depended on them (Hockey et al., 2005;Sinclair et al., 2011). They are solitary hunters, mainly feeding on terrestrial-and aquatic

A B
invertebrates, fish, reptiles, small birds and-mammals (Hockey et al., 2005;Birdlife International, 2015a). Breeding takes place throughout the year with seasonal peaks during the wet months (Hockey et al., 2005;Tarboton, 2011). They breed alongside other species and build their nests to the central, highest points in the heronries, above the smaller species (Hockey et al., 2005;Tarboton, 2011) in large reed beds above water and less often in trees (Tarboton, 2011).
Bubulcus ibis is widely spread throughout South Africa. It is a common resident of open habitats such as grasslands, pastures and agricultural lands (Hockey et al., 2005;Sinclair et al., 2011). They are gregarious birds flocking near watering places before flying to their roosts after sunset. They roost communally with other water birds. The main diet of B. ibis consists of insects, fish, amphibians, reptiles and small rodents and nestlings (Hockey et al., 2005;Birdlife International, 2015b). Bubulcus ibis often associates itself with larger animals such as livestock and antelope standing on their backs (Hockey et al., 2005;Sinclair et al., 2011). The advantage of this is that they can forage in long grass where visibility is poor as well as prey on the ecto-parasites of the mount (Hockey et al., 2005;Sinclair et al., 2011). Bubulcus ibis is a colonial monogamous breeder either nesting in monospecific colonies or mixed species heronries, far outnumbering the other species (Hockey et al., 2005;Tarboton, 2011).
Colonies are often over water where they construct their loosely made nests in reeds or overhanging trees from dry twigs, weeds and reed stems (Tarboton, 2011).

Ibises
Ibises are a group of partly aquatic birds belonging to the family Threskiornithidae. Ibises are fairly large, long-legged birds with down curved bills (Hockey et al., 2005;Sinclair et al., 2011) used to probe mud and soft soil for food (Hockey et al., 2005). Two species from the Threskiornithidae family were found breeding at the bird egg sampling site: the glossy ibis (Plegadis falcinellus) (  Plegadis falcinellus is distributed all over the world. In South Africa their distribution is centred on the

A B
International, 2015c). Plegadis falcinellus forages in flocks in shallow water or on soft ground. Their main diet consists of adult and larval insects, molluscs and crustaceans and less frequently small fish, frogs, lizards and small mammals (Hockey et al., 2005). Plegadis falcinellus is a monogamous breeder that nests in colonies with other species (Hockey et al., 2005). Its nesting is dispersed rather than clustered within these colonies. They construct their nests in large undisturbed reed beds over water, and sometimes in flooded trees from reeds stems to form a platform (Tarboton, 2011).
Threskiornis aethiopicus is found throughout sub-Saharan Africa and in South Africa throughout the country, especially in high densities in the moist eastern high altitude plateau. The provision of permanent water bodies however has facilitated an increase in numbers in the western part of South Africa (Hockey et al., 2005, Birdlife International, 2015d. The sacred ibis prefers margins of inland freshwater wetlands, specifically the shallows, but also inhabit open grasslands and agricultural lands. T. aethiopicus has adapted to the presence of man, by taking advantage of man-made habitats including farm dams, sewage works, abattoirs, refuse tips and farmlands (Hockey et al., 2005).
They eat a great variety of prey, from insects (aquatic and terrestrial), crustaceans, worms and molluscs. They also feed on vertebrates such as fish, frogs, reptiles and small mammals and nestlings. Threskiornis aethiopicus is a monogamous breeder that nests colonially in mixed species heronries (Hockey et al., 2005;Tarboton, 2011). Within these heronries they nest in close proximity of one another, but to the fringes of the colony in discrete groups (Hockey et al., 2005). Sacred ibises will nest in reed beds and trees over water, and often on the ground on isolated islands (Tarboton, 2011).

Sampling of bird eggs
As previously mentioned, the only heronry found in the study area was located in the Lenasia wetland. Water bird eggs (n = 10 per species) were sampled from the nest in the heronries (accessed by boat) during the summer breeding season (November/December) of 2013. The largest egg was collected (assuming it was the first laid) from each nest and marked with a pencil. The eggs were wrapped in pre-cleaned aluminium foil (rinsed thrice with first acetone and then hexane) and also marked on the foil. The eggs were transported in at 4°C and stored at -20°C in the laboratory.

Physical and chemical characteristics of polycyclic aromatic hydrocarbons
As these compounds are part of a very large group, they often differ from one another in physical and chemical characteristics based on their molecular mass as well as the number of aromatic rings they consist of (CCME, 2008). The PAHs all have a high molecular mass with low volatility and the group is regarded as semi-volatile (Ollivon et al., 1999) (  , however their release from anthropogenic activities is one of the most important environmental pollution sources (Van Metre et al., 2000).
The widespread occurrence of PAHs is largely due to their formation and release in all processes of incomplete combustion of organic materials or high pressure processes (Gehle, 2009). This include the production of coke fuel and carbon, coal power plants, petroleum processing, furnaces, fireplaces, gas and oil burners, and automobile sources and petroleum products (Angerer et al., 1997;Maliszewska-Kordybach, 1999;Yunker et al., 2002). Anthropogenic PAHs originate from two distinct processes namely pyrogenic and petrogenic sources.
Pyrogenic PAHs are formed during the combustion of biomass (coal and petroleum, wood, and grass, and industrial waste) (Liu et al., 2009;Chen & Chen, 2011;He et al., 2014)  . Petrogenic PAHs are defined as the congeners that originate from petroleum products, including crude oil, petrol and diesel fuels, lubricants and their derivatives (Angerer et al., 1997;Maliszewska-Kordybach, 1999;Yunker et al., 2002;Saber et al., 2006). The PAH profile of different types of petroleum products vary depending on their production process (Stout et al., 2001), for example, fuel with a lighter mass (jet fuel) contain more LPAHs and the heavier fuels, more HPAHs, due to the distillation temperatures and the PAHs' boiling points and vapour pressures (

Environmental fate
After PAHs are released from their various sources they spread into the environment due to their semi-volatility and they can disperse between air, water, soils and sediments (Nadal et al., 2004). The PAHs that finally deposit into the aquatic environment, bind to the sediment (see section 2.1.2). Here they are subjected to various degradation processes: chemical-, photochemical-and biological. These degradation processes include volatilisation, dissolution, emulsification, microbial breakdown, chemical-and photo-oxidation (Page et al., 1996;Brenner et al., 2002;Warren et al., 2003). Once the degradation processes are active, the physico-chemical properties of the congeners are changed (Kochany & Maguire, 1994;Page et al., 1996). Biological degradation of PAHs seem to be the main pathway of breakdown in sediments and soils (Wilson & Jones, 1993;Lu et al., 2012).

Toxicity
PAH exposure is a danger to wildlife and humans as these compounds are known to be mutagenic and carcinogenic (NTP, 2005). Studies specifically on humans in the workplace where PAH releases are high (industries) or occupations where the products are themselves PAH sources (petroleum-and tar industries) have found that the PAHs enter the body via different exposure pathways and that accumulated PAHs may have detrimental effects (Väänänen et al., 2005;Cirla et al., 2007;McClean et al., 2007). Harmful effects include developmental and reproductive deffects, cytotoxicity (i.e. erythrocyte damage), DNA mutation and other health impacts (Zhang & Tao, 2009;Safe et al., 2010;USEPA, 2011). Apart from the before mention effects some of the PAHs are also known carcinogens (Myers et al., 1994;Savinov et al., 2003;Qiao et al., 2006)

Polycyclic aromatic hydrocarbon sediment quality guidelines
The assessment of a system's pollution status is achieved by means of various methods and indices based on specific guidelines. Indices are sets of aggregated and measured parameters or indicators (OECD, 2003), that are used to compare results indiscriminately between one another. Quality guidelines are sets of values that act as goals for environmental quality. These quality guidelines often have values that are specific for the purpose of the guideline, i.e. aimed at specific compounds and end points (protection of ecosystems, benthic organisms, and aquatic life, amongst others) (Swartz, 1999;MacDonald et al., 2000).
Currently South Africa does not have any sediment quality guidelines, therefore the sediment quality guidelines used to evaluate the PAH contamination in the present study are those of MacDonald et al. quality guidelines calculated on the PAHs' toxicity to sediment dwelling organisms and based on a three step evaluation. The first step was the matching of biological effects and sediment chemistry from published research. Secondly, each of these sediment samples' measured compounds was compared to its corresponding existing guidelines, predicting its toxicity. Finally, accuracy of prediction was evaluated by determining if the sediments were toxic to one or more aquatic organisms, using standardised toxicity tests (MacDonald et al., 2000). If a response was significantly different from a reference or control then the concentration in the sediment responsible for the response was regarded as toxic. Simultaneously, if sediments did not cause a significant response in at least one test endpoint the concentrations in the sediments were considered non-toxic. The threshold effects concentrations (TEC; lower values) were set at values where 75% of the sediments were correctly predicted as non-toxic. Similarly, the probable effects concentration (PEC; higher value) was set where 75% of the sediment samples were correctly predicted to be toxic. Thus, the values set out by the guidelines are the TEC, where concentrations below this value is expected not to have harmful effects to sediment dwelling organisms, and the PEC where concentrations above the value are expected to have harmful effects to benthic organisms at a more frequent interval. When the contaminant concentration is between the PEC and TEC guideline harmful effects are expected to occur, depending on the pollutant. These incidence of harmful effects would be greater than that of the TEC and less than the PEC. In the case when a chemical only exceeds the bottom TEC value and falls between the guidelines, interpretation would be that the toxicity toward benthic organisms would increase as the chemical concentration increases (MacDonald et al., 2000).
The Canadian quality guidelines were implemented to protect, sustain and enhance the environment.
These guidelines were created using environmental and human health risk protocols, thus are used to protect human and animal health (CCME, 2012). In the case of the Canadian sediment quality guidelines, an interim sediment quality guideline (ISQG; lower value) and a probable effects level (PEL; higher value) for PAHs was created, to evaluate the degree to which adverse effects could occur from exposure to PAH containing sediment and so can be a useful ecotoxicological assessment tool (CCME, 1999;2012). This set of guidelines was derived on a similar principle as followed by MacDonald et al. (2000) that include the combination of both a modification of the National Status and Trends Program (NTSP) approach, as well as the spiked-sediment toxicity test (SSTT). The modified NSTP approach associates filed collected sediment concentrations of each compound measured, with any adverse biological effects observed (compiled in the Biological Effects Database for Sediments, BEDS). The SSTT approach is an independent evaluation of information from toxicity tests completed with spiked-sediments. This approach estimates the concentration of a chemical where adverse effects are not expected.

Sediment indices
The ecological toxicity risk that polluted sediments may pose to benthic organisms can be calculated with the sediment quality guideline index (SQG-I) (Fairey et al., 2001). The quality of the sediment can also be calculated in terms of the target compound or a mixture of compounds by using the sediment quality index (SQI) (Marvin et al., 2003). These tools can then be used to compare toxicity and quality of different sites in a study area in order to determine a status (of wellness) for that ecosystem.

Toxic equivalent quotient calculation
The xenobiotics present in the sediment have the potential to activate specific biochemical pathways in organisms, and through these pathways be toxic to the organism. These may include the inhibition or activation of neurotransmitter enzymes (Ellman et al., 1961;Lionetto et al., 2013), anti-oxidant systems (Lau et al., 2004;Ferreira et al., 2007), and endocrine systems (Jobling & Tyler, 2003;Mills & Chichester, 2005). Another example of such a biochemical pathway is the activation of the arylhydrocarbon receptor (AhR), which is a ligand dependant transcription factor that regulates the expression of cytochrome P450 genes, specifically CYP1A1 (Aarts et al., 1995). This enzyme is responsible for the metabolism of the activating xenobiotic. More detail on the mechanism of the transcription of the CYP450s and the AhR will be discussed later (see section 3. Canadian interim sediment quality guidelines (ISQG) for dioxin-like compounds were used in the assessment. This guideline was specifically created for the protection of aquatic life from dioxin-like compounds, using the TEF values for fish (CCME, 2001), and since PAHs use the same mechanism of action, it was used to evaluate PAH toxicity.
Along with the TEF values derived by , which are expressed in terms of 2,3,7,8-TCDD, another set of TEF values exists. These were specifically created to express PAH toxicity in terms of the most toxic PAHs, benzo(a)pyrene (Delistraty, 1997;Tsai et al., 2004). There are, however, no guidelines for BaP-equivalency, but it can be used to compare to BaP equivalencies in published literature to the results of the present study.
In conjunction with the TEF values discussed above, a TEQ can be calculated to express a toxic potency specifically toward fish. Barron et al. (2004), derived fish potency factors (FPF) from published data on CYP1A induction and AhR binding. Calculating a TEQ using these potency factors can be used to specifically evaluate risk towards fish in a system. TEQs calculated with  TEFs represent toxicity towards a broader range of aquatic life. The TEQs calculated using Barron et al. (2004) and  factors are comparable, because 2,3,7,8-TCDD was used as the most potent AhR-ligand, and the FPFs were derived accordingly (Barron et al., 2004).

Polycyclic aromatic hydrocarbon source diagnostic ratios
Possible sources of PAHs can be identified by using diagnostic ratios. These ratios are determined by using the concentrations of specific congeners in a sample relative to one another. Low molecular polycyclic aromatic hydrocarbons (LPAHs) congeners with 2 and 3 rings are mostly released by petrogenic sources, while pyrogenic sources are dominated by 4-6 ring congeners (HPAHs). Thus, the ratio between the LPAHs and HPAHs can be used to identify pyrogenic or petrogenic sources (Scolo et al., 2000). The ratio between anthracene and phenanthrene (Ant/Ant+Phe) can also distinguish between pyrogenic and petrogenic sources (Pies et al., 2008;Chen & Chen, 2011

Sample collection
Sediment samples were collected from the nine sediment sampling sites proposed for this project   was evaporated under a gentle stream of nitrogen gas at 33°C and reconstituted in 10 mL hexane.

Sediment extraction
The final step was solid phase extraction (SPE). Normal phase SPE is a technique used to remove polar compounds from an extract in a non-polar solvent. The highly polar packing material, magnesium silicate, strongly adsorbs polar compounds while non-polar compounds, including PAHs pass through the packing material (USEPA, 1996;2007). For this project commercially available SPE cartridges Supelco 12 mL 2 g/2 g LC-Si/Florisil ® cartridges were used. These cartridges are designed to remove nitrogen compounds from a hydrocarbon sample (Florisil ® component) and also to remove polar substances by forming hydrogen bonds (silica gel component) (Zhang, 2007). The method followed was adapted from USEPA methods 3620C and 3630C (USEPA, 1996;2007): The cartridge was conditioned with 10 mL hexane under gravitational pull. The conditioning solvent was discarded as waste. As soon as the conditioning solvent passed the top of the matrix, but before it ran dry, the 10 mL sample was added and collected in a Turbo-Vap flask positioned under the SPE cartridge.
Once the sample passed through the matrix, 24 mL DCM:hexane (1:1, v/v) was added, 6 mL at a time, ensuring the surface of the matrix never ran dry followed by 8 mL DCM. The bioassay samples were evaporated under nitrogen gas at 33°C and reconstituted in 1 mL hexane. The samples destined for chemical analysis were reconstituted in 1 mL toluene. All samples were stored at -80°C until analysis.

Biota chemical extraction
Polycyclic aromatic hydrocarbons were extracted from fish tissue and wetland bird eggs using a liquid-liquid extraction, coupled with dispersive solid phase extraction (dSPE) known as QuEChERS  The samples were suspended in 5 mL double deionised water (ddH2O), after which 10 mL extraction solvent, acetone:hexane (1:1,v/v) was added. Liquid-liquid extraction was performed by vigorously shaking the samples for 2 hours. The QuEChERS extraction salts (anhydrous magnesium sulphate and sodium acetate) were added to the samples and mixed by vortex for a minimum of two minutes.

Fish muscle tissue extraction
These salts remove the polar fraction and promote partitioning of the analytes to the non-polar organic solvent phase (Lehotay, 2006). After the addition of the salts the samples were spun at 181 g for 5 min at 5°C. The organic phase supernatants were decanted into Turbo-Vap flasks and were evaporated under gentle nitrogen gas flow at 33°C to 1 mL. The flasks were rinsed with extraction solvent and the extract transferred into the 1.5 mL dSPE tube. The dSPE sorbents (primary secondary amine [PSA], C18 and anhydrous magnesium sulphate) removes fatty acids and other interference as well as any remaining water from the extract (Lehotay, 2006). The salts and extracts were mixed by vortex, and centrifuged for 5 min at 181 g (5°C). The supernatants were transferred into gas chromatography (GC) vials, evaporated under gentle nitrogen gas flow at 33°C, and reconstituted in 1 mL toluene. These samples were stored at -80°C until instrumental analysis.

Fish bile sample extraction
Bile was also sampled during fish necropsy (see section 4.2.2). Bile was either withdrawn from the gall bladder (in the case of large volumes) or the gall bladder was collected as a whole (when there was little bile content) and stored in pre-cleaned 4 mL amber vials, transported at 4°C and stored at -20°C in the laboratory. The bile extraction method was adapted from Guo et al. (2013) and Bortey-Sam et al. (2016). Thirty microliters bile was decanted into 15 mL centrifuge tubes. Enzymatic deconjugation was done by adding 10 µL β-glucorunidase (bovine liver, Type B-1, 1 240 U/mL), 10 µL aryl-sulfatase (Patella vulgata Type V 34 U/mL) and 2 mL sodium acetate buffer. The pH was adjusted to between 5.5 and 5.6 with acetic acid and incubated for 6 hours at 37°C. The sample volume was increased with 2 mL ddH2O. The internal standard (OHFlu-13 C6, OHPhe-D9 and OHPyr-13 C6) was added before liquid-liquid extraction commenced with 5 mL pentane for 30 minutes on a rotary shaker. The samples were subsequently centrifuged at 1 600 g for 10 minutes and the organic phase decanted into an evaporation flask. The pentane extraction was repeated and the second organic phase was added to the first. The collected supernatant was evaporated under a gentle flow of nitrogen gas at 37°C to 50-100 µL. The samples were filtered through a 0.2 µm syringe filter and reconstituted in 500 µL methanol for analysis. Samples were stored at 4°C before instrumental analysis.

Wetland bird egg extraction
The sampled eggs (see section 1.4.3) were homogenised by means of sonication. The extraction of the bird eggs followed the same extraction process as the fish samples using the QuEChERS method. Due to higher lipid levels in the egg samples, an additional GPC clean-up step followed the QuEChERS method to remove these excess fats. On the last step of the QuEChERS method, the supernatant of the dSPE was evaporated under nitrogen gas at 33°C and solvent exchanged into 2 mL DCM. The GPC step was the same as with the sediment extraction: The sample passed through both the Envirogel TM GPC clean-up columns-separating analytes based on molecular sizes-and the PAH containing fraction was collected based on the relative PAH retention times. The collected PAH containing fraction was evaporated under nitrogen gas at 33°C and solvent exchanged into 300 μL toluene and stored at -80°C until instrumental analysis.

Instrumental analysis
The extracted samples were analysed at the National Metrology Institute of South Africa (NMISA).
Sediment, fish and egg samples were analysed for the 16 priority PAHs (Table 2.1). The chromatographic system used was a LECO Pegasus IV, gas chromatography coupled to a time of flight-mass-spectrometer (GC-TOFMS) system. Separation was achieved on an Rxi®-PAH (60 m, 0.25 mm ID, 0.1 µm df) priority phase column. The autosampler, GC and TOFMS methods were optimised, in the presence of matrix to achieve optimal separation between the various target analytes and matrix interferences. During analysis compounds were identified based on retention time, elution order and obtained mass spectra (Figure 2.1). Additionally selected sediment analysis were repeated using comprehensive two dimensional chromatography coupled to time of flight mass spectrometry (GCxGC-TOFMS) using a non-polar Rxi®-5SilMS (30 m, 0.25 mm ID, 0.25 µm df) as the primary column and a mid-polar Rxi®-17SilMS (1 m, 0.25 mm ID, 0.25 µm df) as the secondary column. This method was used due to enhanced sensitivity and selectivity for the target PAHs with a smaller molecular mass as chromatographic interferences were more pronounced in the beginning of the chromatogram. [BeP] and 9-OH BaP and separated on an Agilent Eclipse PAH column (150 x 2.1 mm, 3.5 µm dp) with gradient mobile phase (Table 2.3) at a flow rate of 250 µL/min. Target analytes were identified based on retention time, elution order and obtained mass spectra in the negative electrospray ionisation (ESI) mode. Only selected bile samples of 2013 were analysed as part of a screening for OH-PAHs in the Soweto and Lenasia fish. Mean recovery with the bile analysis was 116%±10.7.

Quantification and quality control
Ten and eight point matrix-matched calibration curves were constructed for quantification using extracted blank matrix. Matrix matched calibration curves were selected due to the complexity of the samples analysed. Matrix matching eliminates many of the errors and biases associated with matrix effects, and allows for more accurate quantification. For the sediment quantification a ten point nonmatrix matched calibration curve (0-5 000 ng/mL) was used. The quantification of the bird eggs and the fish samples used an eight point matrix matched calibration curve (0-700 ng/mL). Calibration curves were constructed for each of the 16 PAHs analytes and the 10 OH-PAH metabolites, each with R 2 values greater than 0.9. Quantification was done using peak areas to mass ratios, as described by the following formula:

Sample mass extracted
Where m is the gradient of calibration curve and a is the y-intercept of calibration curve.
The limit of detection (LOD) and limit of quantification (LOQ) were calculated using linear regression analysis of the constructed calibration curves, where Sa was defined as the intercept (Miller & Miller, 2010). The LOD was defined as three times the standard deviation of SY/X and the LOQ was defined as ten times the standard deviation of SY/X.
Recovery was assessed through the analysis of either gravimetrically spiked blank matrix or where available reference material including NIST SRM 1944 (heavily contaminated New York/ New Jersey Waterway sediment) for soil and NIST SRM 2974a (organics in freeze dried mussel tissue) as a representative of the biological material analysed. The recoveries obtained across all matrices analysed are summarised in Table 2.4. Additionally a matrix blank and solvent blank were analysed with every batch of samples and retention times were confirmed throughout the course of the analysis using SRM 2260a (aromatic hydrocarbons in toluene).

Polycyclic aromatic hydrocarbon sediment quality guidelines
The concentrations of polycyclic aromatic hydrocarbons in the sediment of Soweto and Lenasia were gauged against the guidelines of proposed by MacDonald et al. (2000) and the CCME (2001). The congeners were categorised according to their exceedance of the guidelines: below threshold level (lower level), between guidelines (lower and upper level), and above the probable effects level (upper level

Sediment indices
In the present study the sediment quality guideline index (SQG-I) was used to determine the ecological risk the sediment pose to benthic organisms (Fairey et al., 2011). The SQG-I incorporates the more protective guideline (lower) values and the measured concentrations of the target compounds to calculate the index value.
The SQG-I is the arithmetic mean of how many times the measured concentration (CPAH(Sample)) of individual PAHs at a specific site were higher than the lower guideline levels (CPAH(Threshold)) (Fairey et al., 2001).
In addition to the SQG-I, the quality of the sediment can be calculated in terms of the PAH contamination. The sediment quality index (SQI) as described by Marvin et al. (2003), incorporates the percentage of PAHs per site that did not meet the lower guidelines and their magnitude of The calculation of the index takes into account two elements, namely the scope (F1) and amplitude (F3). The scope is the percentage of variables that did not meet the guidelines Amplitude is the magnitude by which the failed variables exceed the guidelines.
Failed test value = amount of samples not meeting guidelines i = Individual guideline p = Total amount of guidelines used

Toxic equivalent quotient calculation
The TEQ was calculated using the following equation: Where: Ci is the concentration of the polycyclic aromatic hydrocarbon congener TEFi is the toxic equivalence of the polycyclic aromatic hydrocarbon congener The TEF values used for the calculation of the TEQTCDD, TEQBaP, as well as the TEQs using the FPFs (TEQFPF) are listed in Table 2.6.

Polycyclic aromatic hydrocarbon source identification and compositions
The diagnostic ratios used to identify the possible sources of the PAHs (see section 2.1.7) are listed in Table 2.7. The respected sources were determined by the ranges of each of the ratios.

Sediment chemical analysis results
The detailed chemical analytical results of both 2013 and 2014 sediment samples are shown in Table 2.8. Samples that showed poor separation of the smaller molecular PAHs, were subsequently reanalysed using the more sensitive GCxGC-MS-TOF. The samples that were below the limit of detection (LOD) and limit of quantification (LOQ) are reported as half LODs and half LOQs respectively. There was no correlation between the PAH concentrations and either total organic carbon content or the sediment grain size (data not shown).
The summary of the results in Table 2  The mean ΣPAHs quantified in the sediments, and their ranges were compared to other South African-, and also African-, European-, North American-and Asian studies (Table 2.9). These ranges and totals are for the 16 priority PAHs only (where available). If authors reported on more than the 16 priority PAHs, new ranges and mean totals were calculated to reflect only the 16 US EPA PAHs if the individual concentrations were available in the paper. In the instances where PAH congeners reported were not for only the 16 priority congeners, the ranges and totals were given as is and it is indicated how many congeners contributed to the totals (e.g. Σ9PAHs).
The study by  in the Soweto and Lenasia section of the Klip River shared many sites with the present study. Therefore the data was used for comparisons in order to determine temporal changes in the study area. The ΣPAHs of  were comparable to this study.
The highest levels of total PAHs from that study was 5 408 μg/kg, measured in the sediment of a site between the Eldorado Park and Nancefield sites of the current study. Overall, the levels measured by  were similar to that measured during the present study's 2013 and 2014 seasons (Table 2.9). In a study by Quinn et al. (2009), where the authors investigated organic pollutants in the central part of South Africa, which also included the present study area, the mean ΣPAHs was lower, however the range between minimum-and maximum concentrations was larger ( The Orange River is considered to be one of the few rivers in South Africa to be less polluted by industry and mining and more by urban development and agriculture (DWAS, 2004;Bucas, 2006) and therefore low PAHs levels can be expected in this system. This was indeed the case in a sample from the Orange River, taken at the agricultural town of Douglas just below the confluence with the Vaal River by Nieuwoudt et al. (2011), which had a mean ΣPAH of half that of the lowest ΣPAHs of this study (Table 2.9). The low prevalence of PAHs in the Orange River was further corroborated by findings of Pieters et al. (2015) in which the majority of the Orange-Senqu River basin was sampled to screen for the levels and occurrences of PAHs. However, the maximum range of the Orange-Senqu Basin ΣPAH levels was much wider. This is because the basin covers approximately 45% of South Africa, and encompasses industrial and heavy populated areas.
In a study on PAH distribution from sources, Okedeyi et al. (2013) investigated PAHs in soils at coalfired power stations and found high levels of PAHs on site (Table 2.9), which declined over distance away from the source. The Lethabo-and Rooiwal (also known as Kelvin) power stations are 60 and 30 km away from Soweto and Lenasia-close enough for air transported PAHs to end up in aquatic systems.
Other African countries which reported on PAHs in sediments include Egypt and Ethiopia. Lake Maryut, in the Nile Delta, is one of the most polluted lakes in Egypt and is situated close to the highly industrialised and populated city of Alexandria (Barakat et al., 2011). The authors analysed Σ39PAHs in the sediments, and the mean and range of the 16 priority PAHs has been calculated for the purposes of this study (Table 2.9). The maximum ΣPAHs in Lake Maryut sediments were higher than the sediments from Soweto and Lenasia. However, the mean concentration for Lake Maryut is ten times lower than our study area's (Table 2.9). This is due to outliers in their samples-3 of the 13 samples was above 1 000 ng/g ΣPAHs, one of these measured 6  Mekonnen et al. (2015) was the first to screen for PAHs. The ΣPAHs quantified in these lake sediments were notably lower than for Soweto and Lenasia ( The high altitude lakes of Norway and Austria had levels of the 16 USEPA PAHs within the range of our results (Table 2.9), evidence that PAHs also travel to remote areas via the atmosphere (Fernández et al., 1999). In heavily industrialised countries such as Germany, and to a lesser extent the Czech Republic, the levels of PAHs in the sediments were immense (Table 2.9). The floodplain soils of the Saar and Mosel Rivers that flow through the coal mining areas of the Rhineland-Palatinate

Fish muscle tissue chemical analysis results
No PAHs (in native form) were detected in the fish samples because PAHs are bio-transformed by vertebrates. (Hylland, 2006). Xenobiotics are bio-transformed via two pathways: hydrophilic xenobiotics go through phase I where a polar conjugate is introduced by means of oxidative, reductive and/or hydrolytic processes (Tuvikene, 1995;Rose & Hodgson, 2004;. Xenobiotics that are already water soluble are directly passed to the phase II pathway which involves the conjugation of xenobiotics or their phase I metabolites. The conjugates that are added to the compounds include acetate, amino acids, glutathione, glucuronic acid, methyl groups and sulphate, to name a few (Tuvikene, 1995;. Once conjugated the xenobiotic is water soluble and excreted from the organism (Tuvikene, 1995;. The main PAH metabolism pathway in fish involves cytochrome P450 (CYP450), monooxygenases, epoxide hydrolase and conjugating enzymes (Tuvikene, 1995, Rose & Hodgson, 2004). The metabolism of PAHs is well studied and the collective process was described in detail by Tuvikene (1995): The subfamily of the CYP450 genes that are activated in fish by PAHs is the CYP1A family.
After the PAH has bound to the Ah-receptor (see section 3.1.2), the CYP450s are induced. Specific forms of P4501A1 are induced after exposure to PAHs: aryl hydrocarbon hydroxylase (AHH), ethoxyresorufin-O-deethylase (EROD) and 7-ethoxycoumarin-O-deethylase (ECOD). The release of these enzymes results in the addition of an oxygen atom to the AhR-ligand (like PAHs) and in most cases this oxygen is reduced to a hydroxyl group by monooxygenases. Following these reactions the metabolites are conjugated by several enzymes and anti-oxidants such as glutathione-S-transferase (GST), uridine 5-diphosphate-glucuronosyltransferase (UDP-GT) and glutathione (GSH). These enzymes complete biotransformation phase II: reducing the toxicity of the compound and making it easier to excrete (Rose & Hodgson, 2004). The deconjugated metabolites or hydroxylated PAHs quantified in the bile samples are presented in Table 2.10. There was no significant differences were observed between the sexes. The site with the fish that had the highest mean of total hydroxylated PAHs concentration was Orlando (East), 947.4 ng/mL (Table 2.10), 2-,3-hydroxyfluorene had the highest levels (mean 708 ng/mL, range 172-1 429 ng/mL) contributing to the total OH-PAHs in Orlando fish. The second highest mean ΣOH-PAHs, was in fish from Fleurhof (371.2 ng/mL) followed by Lenasia (201.7 ng/mL). These levels are notably lower than the Orlando results (Table 2.10). As expected, the control fish had the lowest concentrations of OH-PAHs-these fish were kept in a controlled environment and the depuration period would have allowed the majority of the PAHs to be metabolised and excreted. The Orlando fish had the greatest variety of hydroxyl-PAHs in their bile (7 of the 11, Table 2.10). The metabolite profiles indicate relative trends between each other. The 2,3-OH fluorene was invariably the dominant metabolite, contributing between 75% and 89% to the total in the fish from the study area, and 92% in the control fish  One of the best general indicators of PAH exposure in fish is considered to be 1-hydroxyl pyrene (Van der Oost et al., 1994;Ruddock et al., 2003) and is the main metabolite of pyrene-one of the most abundant pyrogenic PAHs together with fluoranthene and to a lesser extent, phenanthrene (Page et al., 1999;De Luca et al., 2004). The only phenanthrene metabolite present in the Soweto and Lenasia fish was 1-OH Phe (Figure 2.2). The fact that phenanthrene has a relatively long half-life in soil-like matrices (LeBlanc, 2004) and have a high affinity for sediment particles (Yuan et al., 2001), may be the cause for the lower concentrations in biota (Table 2.10), as these characteristics decrease its bioavailability.
To the author's knowledge the data on biliary PAH metabolites for the fish in this study is the first for South Africa. Only a very few international studies did report on biliary PAH metabolites from fish. The majority of the international PAH metabolite studies in fish bile were on estuarine (Richardson et al., 2001;Ruddock et al., 2003;Jonsson et al., 2004) or marine fishes (Escartin & Porte, 1999a;Aas et al., 2000;Richardson et al., 2001;Kammann, 2007). Thus, the results above will be gauged against gariepinus. These high levels suggest that the estuarine environments of the UK were far more polluted. Escartin and Porte (1999b), reported on the biliary levels of hydroxylated fluorene, phenanthrene, and pyrene metabolites in brown trout (Salmo trutta) from Norwegian and Austrian high altitude lakes. These authors reported a mean of 218.3 ng/mL and 154.2 ng/mL of 1-OH pyr for the Norwegian and Austrian lake fish respectively. These levels are notably higher than those found in the present study. However, the levels of 9-OH Flu in S. trutta were lower than that in C. gariepinus of South Africa. This trend, along with the high levels of the other Flu-metabolite suggests that the sources of the two areas differ in origins (pyrogenic or petrogenic).

Wetland bird egg chemical analysis results
The chemical analysis results for the wetland bird eggs were similar to that of the fish results. Most of the PAHs were metabolised and only naphthalene, phenanthrene and to a lesser extent acenaphthylene were detected (Table 2. Herbert et al. (2011) investigated the 16 priority PAHs in the eggs of two tern and three gull species from northern Alberta, Canada. The PAHs results reported are similar to that of this studynaphthalene and phenanthrene (Table 2.11) were present in most of the samples in low levels, however they also measured Ant, BkF, BaP, DBA, and BgP, also in low concentrations and in a minority of the samples.

Polycyclic aromatic hydrocarbon sediment compositions
The ratios of the different size classes (based on number of rings) of PAHs measured in the sediments were calculated to determine the composition percentages at each site. These compositions represent the inputs of PAHs at each site from the surrounding area, and therefore references to site names in this section refer to the area, and not so much particular sites. The use of these compositions allows for interpretation of the chemical concentrations and supplements the source identification ratios. PAHs were the most abundant, followed by the 5-ring PAHs and then the 3-and 6-ring congeners, during their study in the same area.

Polycyclic aromatic hydrocarbon source identification
By calculating the ratios between various PAHs found at a site the original source categories (pyrogenic or petrogenic) can be determined. The results from the source identification are reported in Table 2 InP/(InP+BgP) ratio. This ratio suggests that 11% of the sites had PAHs that originated from petrogenic sources in 2013, and of the remaining 89% pyrogenic sites, 11% of these were due to biomass combustion.
The temporal variation observed for the InP/(InP+BgP) ratio could be attributed to the Lenasia site that seemed to have petrogenic sources for 2013 but indicated a pyrogenic source of petroleum combustion. Although Nancefield stayed in the pyrogenic category for both sampling years, its source nature changed from biomass combustion in 2013 to petroleum combustion in 2014 (Table 2.12).
Although the source distribution between petrogenic and pyrogenic showed the same percentages for the LPAH/HPAH ratio between sampling events, the particular sites that were identified with this ratio, differed between the years. In the 2013 sites, Lenasia and Nancefield had the petrogenic sources, but in 2014, the sites were Lenasia and Orlando West (Table 2.12). Similarly, the percentages shown for the Fla/(Fla+Pyr) ratio were also attributed to different sites: for 2013 Lenasia calculated for petroleum pyrogenic sources and in 2014 it was Fleurhof (Table 2.12). The origins identified by  were similar to what was discovered for the present study. It seems that the predominant source of anthropogenic released PAHs in South Africa is pyrogenic, as was discovered when the source ratios were calculated for the other South African studies mentioned in Table 2.5 (data not shown).  et al., 2015). The HPAHs reported on in the publications of Malik et al. (2004) and Chen et al., (2004) had their origins very similar to our sites which were dominated by biomass combustion. The Ethiopian sites (Mekonnen et al., 2015) had a mixture of both petrogenic (Lake Ziway) and pyrogenic (biomass combustion-Akaki River and Lake Awassa) sources.

Polycyclic aromatic hydrocarbon sediment quality guidelines
The sediment quality guidelines were used to determine ecological risk posed by the sediments from the sample sites, based on their chemical concentrations. The results of the application of the PAH sediment quality guidelines on the sediments of Soweto and Lenasia is reported in Figure 2.5 (for the entire study area) and Table 2.13 & 2.14 (for individual sites).
The CCME guidelines regarding PAHs are more protective than that set up by MacDonald et al. (2000), and are therefore lower. This is visible in Figure 2    When the more sensitive Canadian guidelines (CCME, 2012) were applied, all the sites of 2013 surpassed three ISQG levels. Moroka was joined by two other sites with multiple congeners higher than the ISQG present in their sediments. Fleurhof had 10 out of 12 levels above the lSQG, followed by Moroka and Orlando West with 9/12. Eldorado Park and Orlando East had 6/12 and 5/12 exceedances respectively (Table 2.14). During the study of , only 3 of the 13 sites exceeded the CCME guidelines. The congener that exceeded guidelines by all three these sites was benz(a)anthracene (BaA). Only one site had PAHs other than BaA with levels higher than the ISQG: naphthalene (Nap), phenanthrene (Phe), pyrene (Pyr), and benzo(a)pyrene (BaP) . . ISQG = interim sediment quality guidelines; PEL = probable effects levels An increase in concentration of PAH levels in sediments were noted in 2014. Subsequently, more guidelines were surpassed even to such an extent that some levels were higher than the PEL, e.g. at Moroka, Eldorado Park and Lenasia (Table 2.14). Interestingly, Dobsonville's sediment seemed to have been less polluted in 2014 than in 2013, exceeding only one of the 12 guidelines.

Sediment assessment indices
The potential ecological risk that the sediment of the study area posed to benthic organisms, in terms of PAH exposure is expressed by the SQG-I (Table 2.15). In terms of the MacDonald et al. (2000) guideline, Moroka sediment sampled in 2013 and 2014, posed high probability to be toxic to biota.
The only other site of 2013 that posed a moderate risk was Orlando East whereas, Lenasia and Eldorado Park scored moderate probability for 2014.  SQI scores were calculated for six of the  sites. Of these only one was fair, the rest were all in a poor state, achieving a SQI score below 45% (Scale in Table 2.16). The "poor" sites were comparable to the current study sites as they fell within the same urban areas as what we sampled.
Lenasia had a poor SQI score, which was the same for , but the present study's Nancefield site was scored as marginal SQ, which was better than the poor SQI of the 3 sites in the area. The only difference in sediment quality between the two studies is at Protea Glen, where for this study a marginal SQ was calculated and for  a fair SQI score was assigned.
The toxic equivalent quotient (TEQTCDD) calculated from the measured polycyclic aromatic hydrocarbons concentrations in sediments are reported in Table 2.17.  (2013). These overall results indicate that there are AhR mediated toxic responses to be expected in aquatic organisms.

Figure 2.6: Mean toxic equivalent quotient (TEQ) results calculated from literature A) TEQTCDD compared to the TEQ guidelines of the CCME (2001); B) TEQBaP
When comparing the mean TEQTCDDs of this study to those calculated from literature (Table 2. The risk that PAHs in the sediment have toward fish health as shown by the TEQFPF, using the fish potency factors derived by Barron et al. (2004), were also compared to the Canadian interim sediment quality guideline ( Park and Lenasia (both 2014 sediments) also exceeded the guideline with more than 10 fold, Eldorado Park by 12 times and Lenasia by 11. The remaining sites for both sampling years exceeded the guideline by a range of 1-7 times.

Conclusion
It is clear, from the combined results, that there are sites that were severely affected by the PAHs in the Soweto and Lenasia study area including Lenasia, Moroka, Eldorado Park and Orlando West. The area of most concern is Moroka-the site that had the highest PAH concentrations for both years.
Moroka had pyrogenic sources, mainly dominated by 4-ring congeners and was the site that exceeded the most guidelines (both sample sets). Its toxicity assessment indicated that it is likely toxic to benthic biota (from the guideline scores and the SQG-I). Moroka's sediment posed harmful risk to aquatic organisms and specifically to fish. The risk assessment over all indicates that the sediments of the Soweto and Lenasia area may be harmful to the aquatic organisms residing in the area.
Our results were comparable to that of  who also studied sites within Soweto and Lenasia. The SQI and SQG-I calculated for sites for this study were better than for  where comparable. This means, according to the indices, that the sediment quality has improved and the potential of toxicity to benthos has decreased. The presence of PAHs in the Soweto and Lenasia area (in all matrices investigated) was confirmed by the instrumental analysis. Although the wetland bird eggs did not yield sufficient data, except for some of the LPAHs, the conclusion can be made that the wild birds have often been exposed to PAHs. This was also confirmed with the analysis of the OH-PAH metabolites in the fish bile, indicating exposure to fluorene and pyrene and to an extent phenanthrene.

References:
Aarts JMMJG, Denison   Measuring the levels of these chemicals within an environmental sample is important to determine the level of pollution in that sample (Hilscherova et al., 2001). However, with chemical determination, compounds can only be analysed if applicable analytical methods and standards exist (Garrison et al., 1996). The instrumental analysis of an environmental sample also does not take into account the interactions and synergy of the mixture and provide limited information on their potential biological effects (Hilscherova et al., 2000;Vanderperren et al., 2004). Bioassays address this limitation of instrumental analysis and provide the estimations of the biological effects substances have on living cells and tissues (Hilscherova et al., 2000;Behnisch et al., 2002;. Various types of bioassays exist that investigate different biomarker endpoints. Bioassays were developed to answer the need for rapid and relatively inexpensive methods that detect and estimate relative potencies of complex mixtures (Baston & Denison, 2011) and quantifiably analyse the responses in a biological manner (Behnisch et al., 2002). One of the many types of bioassays is the reporter gene in vitro cell bioassays. Cell bioassays offer a rapid and sensitive solution to the limitations of instrumental analysis, with ability to estimate total biological activity of a mixture of chemicals with the same mode of action (Hilscherova et al., 2000).
In vitro cell bioassays are used to assess different modes of toxicity as endpoints such as genotoxicity, endocrine disruption and activation of the aryl-hydrocarbon (Ah) receptor. The DNArepair-deficient chicken DT40 B-lymphocyte cell line is used to screen and characterise genotoxicity of compounds (Ji et al., 2009). Similarly, the Ames test assesses genotoxic effects like point and frame shift mutations using the Salmonella TA98 and TA100 strains respectively (Mortelmans & Zeiger, 2000). The effect of endocrine disrupting chemicals (EDCs) can be measured with various cell lines, focusing on different sections of the endocrine system. The H295R cell line measures endocrine disrupting activity by modulation of the steroidogenesis pathway (Hecker et al., 2006). Oestrogen activity is quantified using the MVLN oestrogen receptor-mediated luciferase reporter gene bioassay (Demirpence et al., 1993). The androgenic chemical effects are measured similarly by means of the MDA-kb2, androgen receptor-mediated luciferase reporter gene bioassay (Wilson et al., 2002). The Ah-ligand mediated toxic responses are quantified by measuring ethoxyresorufin-O-deethylase (EROD) activity (CYP1A1 activity) using the RTL-W1 cell line (Lee et al, 1993). The H4IIE-luc cell line also measures the CYP1A1 activity as endpoint but quantifies the activity with a receptor-mediated luciferase reporter gene bioassay (Sanderson et al., 1996).

Polycyclic aromatic hydrocarbons and cellular responses
Polycyclic aromatic hydrocarbons (PAHs) are known carcinogens and have adverse effects on human and wildlife health (Balch et al., 1995, Spink et al., 2008, Larsson et al., 2012. Some PAHs are toxic by acting through the aryl hydrocarbon receptor (AhR), which is a ligand-activated transcription factor that mediates many of the biological effects of these compounds (Denison & Heath-Pagiuso, 1998;Baston & Denison, 2011), and a number of PAHs may also interfere with the oestrogen receptor (ER)-mediated signalling (Machala et al., 2001).  (Villeneuve et al., 1999) and are collectively referred to as the carcinogenic PAHs (CPAHs). The AhR-ligands enter the cytoplasm of cells and bind to the AhR complexes-unbound AhRs are complexed with heat shock proteins (HSP) (Denison & Heath-Pagiuso, 1998;Tian et al., 2015) (Figure 3.1). Upon binding, the heat shock proteins dissociate and activates the complex (Hilscherova et al., 2000).

2000)
The activation of the AhR has been reported to exhibit anti-oestrogenic cross-talk with the oestrogen receptor (Chen et al., 2001), blocking the receptor (ER) (Safe, 2001). This cross-talk mechanism between the AhR-ERα is complex, but involves the inhibition of oestradiol responsive genes by DRE structures that bind to the AhR complex and so disrupting the oestrogen action through multiple mechanisms (Navas & Segner, 2000;Safe et al., 2000), which may lead to detrimental effects. Thus the ability of PAHs to bind to DNA is not their only role in carcinogenesis, but can include disruption in hormone systems (Baird et al., 2005).

Polycyclic aromatic hydrocarbons and the H4IIE-luc reporter gene bioassay
Dioxin-like toxicity (AhR mediated toxicity) of PAHs was specifically investigated for the sediment sampled in Soweto and Lenasia. The AhR mediated responses of PAHs can be quantified with the H4IIE-luc reporter gene bioassay. The H4IIE-luc bioassay results represent the total amount of bioactivity due to AhR-ligands present in the environmental sample as a result of gene activation. The H4IIE-luc reporter gene bioassay consists of rat hepatoma cells that had been stably transfected with a firefly luciferase reporter gene. The bioassay indirectly measures cytochrome P450 induction, as mentioned above, which is an endpoint in the AhR mediated response (Hilscherova et al., 2000;Denison et al., 2004). The luciferase gene was inserted downstream of the cytochrome genes and the DRE in the H4IIE-luc cells. In the presence of luciferin (substrate for luciferase), light is produced (Figure 3.2). The amount of light that is released is directly proportional to the amount of AhR agonists present in the sample (Hilscherova et al., 2000). The toxicity of the sample is quantified in terms of the reference compound, 2,3,7,8-tetrachloro dibenzo-p-dioxin (2,3,7,8-TCDD). This quantification is based on the assumption that the investigated sample is a diluted form of the reference material, or a mixture of chemicals behaving like the reference compound, 2,3,7,8-TCDD, which is the most toxic congener of the AhR binding compounds (Yoo et al., 2006). The results are given as relative potency values (REP) or TCDD-equivalence.
The results obtained from using this reporter gene bioassay 1) establish whether there are AhR agonists present in a sample and 2) quantify the toxicity of that sample relative to TCDD. The chemical data obtained from instrumental analysis identify the possible AhR agonists and concentrations of occurrence. The bioassay and chemical analysis complement each other: the relative toxicity quotient (TEQ) can be calculated with the chemical data and compared to the biological toxicity equivalent which is the REP or TCDD-equivalence from the assay. These equivalents can be used to assess the risk the compounds pose to humans and the environment (Yao et al., 2002) and can be compared to environmental guidelines, such as international sediment quality guidelines.

Sample collection
Composite sediment samples were collected from the nine sediment sampling sites proposed for this project (Figure 1.2

Sample extraction
Processed sediment samples (see section 1.4.1), were extracted chemically using the same methods as for instrumental analysis: accelerated solvent extraction, gel permeation chromatography and solid phase extraction (see section 2.2.2). However, no deuterated standards were added prior to extraction because these standards would also bind to the AhR and would elicit a response from the cells indistinguishable from that of the AhR ligands extracted from the sediment producing false positive responses. The final extract was reconstituted into 1 mL hexane.

Maintenance of H4IIE-luc cell culture
During routine maintenance of the cell culture, aseptic conditions were followed. The H4IIE-luc cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) (Sigma-Aldrich) with L-glutamine and 10% glucose, and without phenol red and sodium bicarbonate. The DMEM was supplemented with 10% foetal bovine serum (FBS) (Sigma-Aldrich) and 0.04 M sodium bicarbonate. The cells were kept in tissue culture dishes in an incubator at 37°C in humidified air (5% CO2: 95% air). The cells were rinsed with phosphate buffered saline (PBS) (Sigma-Aldrich) and treated with 1.5 mL trypsin (Highveld Biological) to passage them (Aarts et al., 1995).

H4IIE-luc reporter gene bioassay
The method of the luminescence bioassay is a modified version of that described by Tillitt et al. The digestion of luciferin by luciferase produced light, measured in relative light units (RLUs).

Calculating bioassay equivalence (BEQs)
Dose-response curves were prepared for the samples as well as the positive control by plotting the (The unit of these REPs is mass TCDD-equivalents/volume extract). Reporting all three REPs is necessary as it cannot be assumed that the complete mixture of the environmental samples will respond the same as TCDD (Villeneuve et al., 2000). The REP values were back calculated to represent the TCDD-eq in terms of the mass sediment extracted . The TCDD-eq calculated from bioassay results are commonly known as bioassay equivalents (BEQ) (Baston & Denison, 2011). The limit of detection (LOD) for the H4IIE-luc bioassay was calculated by determining the mean EC0 for the TCDD response curves. The utmost intercept with 95% confidence was calculated and used as the LOD, back calculated to a ngTCDD/g value (Villeneuve et al., 1999;Thomsen et al., 2003).

MTT viability assay
A viability test was performed parallel to the luminescence bioassay and dosed with the same series of samples and controls as in the bioassay. The 3-[4,5-dimethyltiazol-2yl]-2,5-diphenyl tetrazolium bromide (MTT) viability assay was used to prevent false negative results in the luminescence bioassay, where low or below LOD responses in the H4IIE-luc bioassay might not necessarily be due to the absence of AhR agonists, but rather from cytoxicity. The MTT assay mechanism involves the metabolism of yellow MTT solution by the living cells into blue formazan crystals. The viability of the cells was determined by spectrophotometric quantification of formazan formation (Vistica et al., 1991).
The MTT plates were seeded, dosed, and incubated in the same manner as the luminescense plates.
On the fifth day, the MTT plates were rinsed with PBS, but did not receive lysis buffer. Viability was calculated by expressing the OD of the wells that received samples were expressed as a percentage of the OD from the control wells, representing 100% viable cells. Statistical differences between the control cells (100% viability) and the exposed cells were tested using single-tailed Mann-Whitney U tests where p-values lower than 0.05 was considered significant.

Results and discussion
The H4IIE-luc reporter gene bioassay results of the low flow season of 2013 and 2014 are presented and discussed below. Luciferase induction for both years was reproducible with coefficients of variance (CV) less than 11%. The limit of detection for the sediments was 14.8 ng/g TCDDeq/g (95% confidence).
The response elicited by sediment extracts was reported as the maximum response on the doseresponse curve (%TCDDmax) relative to TCDD as the standard (Figure 3.3A). The concentrations of the biological equivalents (BEQ) were quantified by comparing the dose-response relationship of the sediment extract (known mass extracted) to the TCDD standard curve and the BEQs were quantified as REP20, 50 and 80 (Table 3.1).
The cell viability reported in Table 3.1 is for the raw extracts only. The 2013 samples that were cytotoxic to the cells were Protea Glen, Moroka, Eldorado Park, Orlando East and -West, and Nancefield (Table 3.1). Cytotoxicity was again seen in the 2014 sample for Moroka, Eldorado Park and Orlando East. The Lenasia site registered cytotoxicity only for the 2014 sample (Table 3.1).  Table 3.1). This indicates that the extracts contained either AhR-ligands that were cytotoxic at high concentrations and/or non-AhR binding compounds that killed the cells. The consequence of this is that the maximum elicited response might very well have

Evidence of cytotoxicity is also noticeable in
been higher than what was reported here. Schirmer et al. (1998) revealed that the 2-and 3-ring PAHs were directly cytotoxic to a rainbow trout gill epithelial cell line, RTgill-W1, specifically Nap, Acey, Acea, Flu and Phe and others demonstrated that the toxicity of some PAHs are additive (Beach & Harmon, 1992;Muñoz & Tarazona, 1993;Schirmer et al., 1998). Therefore, there is a real likelihood that the observed cytotoxicity could have been due to the PAH contents of the extracts even though the concentrations of individual PAH congeners in the sediment (Table 2.8) were not as high as those Schirmer et al. (1998) observed cytotoxicity for. It is the additive effect that could have contributed to the cytotoxicity observed in our bioassays.
The maximum luciferase activity of viable cells per site was given as %TCDDmax (Table 3.1) and were included in the table to give an indication of the approximate reactivity, but since the mass of the extracted sediment had not been brought into consideration yet (as for the REP values) the %TCDDmax cannot be used to compare between sites and over time. The 2013 sediment bioassay had a %TCDDmax range of 29.7% and 106.2%. The range for 2014's %TCDDmax was lower than 2013 and varied between 30.8% and 76.1%. Lenasia was also the only site to have a value higher than 50% in 2014 (Table 3.1).
As mentioned earlier in this chapter, it is important to report all three REP values ( values were not measured for all the sites, some were extrapolated values, and therefore only the REP20s will be used for comparison between sites (Figure 3.4).
The highest REP20 value was calculated for the Moroka sediments (221 pgBEQ/g) (Table 3.1), followed by Eldorado Park and Orlando East-all greater than 100 pgTCDD-eq/g (Table 3.1).
Sediments from Orlando West and Nancefield had responses below the limit of quantification. Moroka was once again the site with the highest REP20 for 2014, followed by Eldorado Park and Lenasia (Table 3.1). Protea Glen was below the limit of quantification and all the remaining sites calculated a REP20 below 100 pgTCDD-eq/g. It is immediately apparent that the BEQs measured for 2014 sediments were greater than 2013 ( Figure 3.4). Moroka 2014 and Eldorado Park's BEQ are 4 times greater than the previous year's ( Figure 3.4). The only site that had a notable temporal decrease was Protea Glen, from 91.6 pgBEQ/g to below the limit of detection. Protea Glen was also the only site from the 2014 sampling that was below the LOQ (Table 3.1). In addition to Moroka and Eldorado Park, Lenasia showed a great increase over the two sampling years (Figure 3.4). Orlando West and Nancefield also had temporal increases from below the LOQ to quantifiable BEQs, even if they were low (Figure 3.4). The definite increase in 2014 BEQs suggests that there must have been an increase in AhR-ligands between the sampling events. The extraction method followed to isolate the PAHs works very effective because there is a strong correlation between the BEQs and the concentration of the CPAHs ( Table 2.8) of each site (Spearman's correlation r = 0.78, p = 0.0002) (Figure 3.5A). Thus, the luciferase activity seen in the bioassays could be mainly attributed-but not exclusively-to PAHs. Other compounds such as PCBs could have been co-extracted and isolated with the PAHs based on their size during the GPC step (see section 2.2.2), and dl-PCBs definitely bind to the AhR (Hilscherova et al., 2000).

B) dual axis line graph showing the relationship between BEQs and ΣCPAHs
One clear example of the presence of compounds other than PAHs present in the extract is that of the Moroka 2013 sample (Figure 3.5B). The highest levels of ΣCPAHs were measured in this sample (Table 2.8), but its BEQ was not the highest (Figure 3.5). It may be that compounds with a higher affinity for the AhR but less potent than the PAHs prevented many PAHs from binding to the AhR (Brown et al., 1994;Petrulis & Bunce, 2000), causing only a small volume of luminescence. This binding to the AhR may even lead to an inhibitory effect. In the case of Lenasia, Moroka, and Eldorado Park sediments of 2014, the BEQs were greater than the ΣCPAHs (Figure 3.5). It is likely that other AhR-ligands were present at these sites and contributed to a higher BEQ along with the CPAHs.
The toxicity of the sediments has already been gauged against international sediment quality guidelines (see section 2.2.4). The TEQs calculated in chapter 2 were based on instrumental analysis, and therefore reflected the toxicity of the measured PAHs relative to 2,3,7,8-TCDD. The BEQs calculated with the H4IIE-luc bioassay are also in terms of 2,3,7,8-TCDD toxicity, but reflect biological reactions to the entire extract contents and is not related to chemical quantification of the separate congeners. The BEQs were also compared to the Canadian interim sediment quality guidelines (ISQG) for dioxin-like compounds, reported in Table 3.2. Only Dobsonville 2014 had the same toxicity assessment results between the TEQs and BEQs (Table 2.17 & 3.2). All the other sites were well above the upper probable effect level (PEL) guideline (Table 3.

2). Moroka and Eldorado
Park were the two sites that had the greatest BEQs, and subsequently exceeded the guidelines the farthest. Moroka's 2013 BEQ was 260 times higher than the lower guideline, and its 2014 BEQ 960 fold the ISQG (Table 3.2). The upper PEL was exceeded by a factor of 10 and 30 over the two consecutive years respectively. Similarly Eldorado Park surpassed the lower ISQG by 160 and 800 times, and the PEL threshold six and 30 times, for 2013 and 2014. This is of concern as it shows that there is a definite risk to benthic organisms in the aquatic systems of Soweto/Lenasia, exposed to the sediments.   sampled in the same location as the Nancefield site (Figure 1.2), for two consecutive years. The BEQs reported decreased between the two surveys, from 161.24 pgBEQ/g to 86 pgBEQ/g . The higher BEQ is comparable to the levels of this study (Table 3.1).
It is important to note that the  study analysed a different fraction of the extract.
These authors treated their extracts with sulphuric acid, before running the bioassay. This step would have destroyed most AhR-ligands (non-persistent) including the PAHs. The only compounds that would have survived such a treatment would be the very persistent polychlorinated dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs) and dioxin-like polychlorinated biphenyls (dl-PCBs).
Less AhR-ligands would mostly, but not always, lead to lower BEQ levels. If Roos and co-authors were to perform bioassays with the extract before being treated with acid, they probably would have found much higher BEQ levels. Therefore, one would expect a higher BEQ in the  samples if the PAHs were included, seeing that their PAHs levels they found are very similar to the concentrations of this project (Table 2.9).
In another South African study, Edwards et al. (2016) (Table 3.1), the sites with the lowest BEQs in the current study. This comparison shows that the urban area of Soweto and Lenasia was more polluted than the rural Phongolo area, despite of its share of AhR-ligand sources.
In a study by Keiter et al. (2008) following the distinct decline in fish populations in the Danube River, Germany, the AhR-agonists of sediment were tested. The lack in knowledge of the levels of organic pollutants led to the biological investigation using the H4IIE cell line's commercial version, the DR-CALUX (Dioxin Response Chemically Activated Luciferase Expression) assay. Keiter et al. (2008) dosed a raw extract onto the cells (which contained all AhR-ligands) as well as a second extract containing only persistent AhR-ligands. Chemical analysis of their crude extract showed a mean ΣPAH16 of 4 898 ng/g and a mean ΣCPAH of 2 423 ng/ (Keiter et al., 2008). The BEQs calculated from the raw extracts were all above 1 000 pgBEQ/g and the mean BEQs of the extract with persistent ligands was 395 pgBEQ/g (Keiter et al., 2008). Even with the high levels of halogenated aromatic hydrocarbons (HAHs) measured in the Danube River, it seems that the very high PAHs, compared to the sites in the present study, contributed greatly to the BEQs measured in this polluted river (Keiter et al., 2008). Sediments from the Soweto and Lenasia sites have considerably lower levels of both ΣPAHs and BEQs when compared to this European river. the sites least influence by the marine water, were chosen. These inland sites had BEQ values lower than our study. The Umhlatuzana River had a BEQ of 4.26 pgTCDD-eq/g (ΣPAHs: 58.7 μg/kg) and the Umbilo River had BEQ of 7.7 pgTCDD-eq/g (ΣPAHs: 186.6 μg/kg). Closer to the Durban harbour the BEQs were much higher (Vogt, 2013). Louiz et al. (2008) studied the dioxin-like activity, along with endocrine activity, in the sediments of the Bizerta lagoon, one of Tunisia's most polluted lagoons. For comparison, the most inland sites were chosen. These sites were mainly exposed to anthropogenic stressors such as metallurgy industry (Louiz et al., 2008). The authors also determined the concentrations of PAHs in the sediments with instrumental analysis. These authors determined the cytochrome P4501A1 activity by quantifying EROD after exposing PLHC-1 cells. This cell line was derived from hepatoma cells of the desert topminnow (Poeciliopsis lucida). The cytochrome gene activation is mediated by the AhR, the same as the H4IIE cells (Villeneuve et al., 2001). Louiz et al. (2008) calculated BEQs (in terms of 2,3,7,8-TCDD) that was compared to our sites. The BEQs quantified by Louiz et al. (2008) were an order of magnitude higher than ours, ranging from 0.8-12.8 ngBEQ/g. Although these authors measured PAH concentrations, they dosed the cells with the raw sediment extracts, that was not subjected to cleanup or fractioning steps. Thus, the very high BEQs reported by Louiz et al. (2008), cannot wholly be attributed to the PAHs they quantified. BEQs (mean of 9 ngTCDD-eq/g). This is to be expected as the PAHs reported by Hilscherova et al. (2001) were at high concentrations in the sediments ( . The Haihe and Dagu Rivers flow through the Tianjin City and had been exposed to industrial and domestic waste historically (Song et al., 2006). From their eleven sites the four most inland sites were compared to the Soweto and Lenasia sites. The BEQs bioassay ranged between 693.6 pgBEQ/g and 6 834 pgBEQ/g (Song et al., 2006), and the mean of the four chosen sites was 2 313.85 pgBEQ/g. The authors attributed these high BEQs to PAHs, literature reports PAH concentrations in the area between 800 and 1 200 ng/g in soil (Wang et al., 2003). The measuredBEQs in the fractioned extracts were considerably higher than the 2013 BEQs of Soweto and Lenasia (Table 3.1). The BEQs of 2014's Moroka and Eldorado Park were comparable (Table   3.1) to three of the four selected upstream sites (693.6-926.6 pgBEQ/g) (Song et al., 2006).

Conclusion
The H4IIE can be conclude that in terms of the biological responses to PAHs in Soweto and Lenasia, the potential risk of AhR mediated toxicity is great for the aquatic organisms in the system and potentially people exposed to the sediments.

References
Aarts

Health assessment of aquatic environments
The assessment and evaluation of stressor effects on ecosystem health is constantly faced with challenges such as the validation of laboratory experiments (Adams, 2001) that are done as environmental analogues. Standard laboratory toxicology tests are used to investigate single or mixed contaminants for simple but critical endpoints, such as survival, growth and reproduction potential (Adams, 2001). Test organisms usually do not function normally by the time these endpoints are reached (Larsson et al., 1985), and the test conditions seldom reflect the complex conditions of a natural environment (Lagadic et al., 1994). The use of laboratory studies (using environmental samples) are thus ideal for single endpoint studies such as targeted enzymatic (Aas et al., 2000) or physiological changes (Villalobos et al., 2000) in response to contaminants. There are shortcomings associated with laboratory studies in terms of making an interpretation or conclusion regarding environmental studies. Therefore, using organisms from the environment are ideal indicators of ecosystem health. Indicators are parameters or parameter derived values that describe or provide information on the state of the environment (UNstats, 1997). In 1954 Feibleman published a paper on the role of integrated levels in organisations of different disciplines in science: physics, chemistry, biology, psychology and anthropology and he explained that disturbances introduced at any level of organisation create reactions in all the other levels it is associated with (Feibleman, 1954). Thus the assumption can be made that indicators can be used to report on the disturbance in the organisation of the environment where organisation of the environment means: molecular, sub-cellular, cellular, tissue, organ, systems, whole organism, population, communities, and ecosystem. The health of organisms often reflects the ecosystem health and biotic integrity it occupies (Adams et al., 1993). In essence this is why organisms are used in environmental studies as indicators. Biota represent different levels of organisation from molecular-up to community level and even ecosystem level if different organisms are studied together i.e. fauna and flora. Numerous studies have used aquatic biota as indicators, such as molluscs (Wootton et al., 2003;Sureda et al., 2013), amphibians (Bryer et al., 2006;Leney et al., 2006), birds (Triosi et al., 2006;Herbert et al., 2011), and fish (Kayal & Connell, 1995;Vives & Grimalt, 2002).

Using fish for health assessments
The investigation of organisms at higher trophic levels often gives a cumulative indication of the effects that act on the system as it is mediated through the food web . Fish have variable diets, changing as they grow older-they continue to grow throughout their lives-and as their dietary needs change they occupy different levels of the food web (Larkin, 1978). Fish are therefore useful as long term indicators of aquatic system health. Fish cope daily with natural physiological stressors such as changes in water temperature and velocity, and dissolved oxygen concentrations, as well as sediment loading and food availability (Adams et al., 1993). The cumulative effect of these natural factors along with introduced anthropogenic factors may lead to impairment of fish health. In order to deal with these stressors energy is required, often allocating useful energy away from critical functions of growth and reproduction (Adams et al., 1993). The severity of stress can lead to the reduction of growth, limiting physiological systems and impairing reproduction, resulting in an inability to tolerate additional stress (Adams, 1990).
There are various approaches to assess the effects of stressors on fish health and the ones used in this study are discussed below. These are aimed at specific biological endpoints used to describe effects resulting from exposure to a chemical (or any other form of stressor).

Biomarkers and bio-indicators
There are various definitions for the term "biomarker". Strimbu and Tavel (2010) define it as any characteristic-any biological marker-that can be measured or evaluated as an indicator of normal biological or pathogenic processes, or pharmacological responses to therapeutic intervention.
Similarly, the World Health Organisation (WHO) defines it as a measurement of the interaction between biological systems and environmental hazards (WHO, 1993a). The WHO's International Programme on Chemical Safety further defined biomarkers by adding there is a measurement involved that reflects this interaction between a biological system and a potential hazard whether it is chemical, physical, or biological in nature. The measured response may be functional and physiological, biochemical at the cellular level, or a molecular interaction (WHO, 1993b). Furthermore, Adams (2001) characterised biomarkers as the functional measures of exposures to stressors that are expressed at the suborganismal level of biological organisation. Adams (2001) go further and also define bio-indicators as the structural and functional entities that represent biological effects and end points at higher levels of organisation. In this study, the term "biomarkers" will be used to specifically refer to the lower level of organisation responses (up to subcellular), and "bio-indicators" to any responses at higher levels of organisation i.e. cellular responses and higher.
Since biomarkers and -indicators highlight discrepancies on various organisational levels of the individual organism they are often applied as an early warning or proactive tool. They are used to measure the effect of toxicants before serious permanent damage is done in an entire ecosystem because changes in the organism is generally detectable before adverse effects are seen in higher levels in the biological organisation   (Figure 4.1). The technique in which biomarkers or bio-indicator endpoints are measured are very diverse and depend on the level of organisation investigated (Figure 4.1). One such a technique is using bioassays. As previously mentioned (see section 3.1.1): bioassays are investigating or monitoring tools to estimate the biological effects (Hilscherova et al., 2000;Behnisch et al., 2002; in organisms. They are often designed as rapid and relatively inexpensive methods to quantifiably analyse the reaction to stressors in terms of biological responses (Behnisch et al., 2002).
In the previous chapter a very specific bioassay was described the reporter gene bioassay. Bioassays in this chapter also include, biochemical response assays (molecular level, Figure 4.1) and various biometric indices (tissue to community level, Figure 4.1) each investigating a different biomarker or bio-indicator endpoint.

Biomarker responses
Biochemical response bioassays have been developed to measure various biomarkers of interest.
From here on, biomarker responses will refer to the biochemical response bioassays completed in the laboratory to measure each respective biomarker endpoint. Van der Oost et al. (2003) categorised biomarker responses into three: (1) exposure, (2) effect, and (3) susceptibility. In this project: biomarkers of (1) exposure and (2) effect were the focus. Biomarkers of exposure measure the product of the interaction of exogenous substances or their metabolites, and xenobiotics with target molecules or cells within the body ( Van der Oost et al., 2003;USEPA, 2014).
The biomarkers of exposure used in this project were acetylcholinesterase activity (AChE), and cytochrome P450 activity (CYP450). Some authors however consider AChE as a biomarker of effect (Rickwood & Galloway, 2004), but in this thesis the convention suggested by Van der Oost et al. (2003) and others is followed. According to Van der Oost et al. (2003), biomarkers of effect are the measurable biochemical, physiological and other alterations within tissues and body fluids of an organism that are recognisable due to possible compromised health or disease. In this study cellular energy allocation (CEA) as well as biomarkers indicating oxidative stress and oxidative stress damage were used. Biomarkers that show oxidative stress responses are superoxide dismutase (SOD) and catalase activity (CAT), and oxidative stress damage biomarkers are protein carbonyl formation (PC), and lipid peroxidation indicated by malondialdehyde content (MDA).
Acetylcholinesterase (AChE) is a specialized carboxylic ester hydrolase that is responsible for breaking the synaptic connections of nerve firing. The enzyme hydrolyses acetylcholine into acetate and choline (Lionetto et al., 2013). By deactivating acetylcholine AChE prevents constant nerve firing (Solé et al., 2006(Solé et al., , 2010. These enzymes play a crucial role in the signal transmission in animals, controlling functions such as movement, respiration, hormonal function and reproduction (Solé et al., 2010). AChE is found in the brains of fish, but is also present in large quantities in the liver (Van der Oost et al., 2003). AChE can be inhibited when exposed to organophosphate pesticides (Van der Oost et al., 2003) and PAHs (Kang & Fang, 1997;Lau et al., 2004). Payne et al. (1996) suggested that complex mixtures (other than pesticides) can inhibit the AChE system, such as extracts from used engine oils and wood leachate. Petroleum oils and wood preservatives are known sources of PAHs (Aprill et al., 1990).
The cytochrome P450s (CYP450) are a superfamily of haeme containing enzymes that are widely diverse with regards to substrate specificity and catalytic activity (Guengerich, 2008). The P450 enzymes are generally regarded as the enzymes which are the first defence against exogenous compounds (Liska, 1998). When an organism is exposed to a toxicant, the CYP450 enzymes are expressed (Ellero et al., 2010). Various subfamilies are expressed depending on the type of toxicant present.
Polycyclic aromatic hydrocarbon contamination specifically leads to the expression of the CYP1A1 gene (see section 3.1.2). This expression is the endpoint of the aryl-hydrocarbon (AhR) mediated response (Hilscherova et al., 2000;Denison et al., 2004). The Ah-receptors are located inside the cytoplasm of the cells. When AhR activating agents, such as PAHs, enter the cells, they bind onto the AhR complex (Figure 3.1). Upon binding, the AhR is transported into the nucleus, where it attaches onto a specific DNA sequence (called the dioxin response element, DRE), which consequently results in the transcription of the genes, such as the CYP450s (Aarts et al., 1995;Denison et al., 2004;Whyte et al., 2004). The inhibition and activation of the P450s can be used as a biomarker of exposure as it reacts to the presence of toxicants.
Superoxide dismutase (SOD) and catalase serve as the first enzymatic defence against reactive oxygen species. SODs form the first tier of this cellular antioxidant system (Bartosz, 2005). These metalloenzymes catalyse the dismutation of reactive superoxides (O2 − ) into oxygen and hydrogen peroxide. Thus, they are an important antioxidant defence in nearly all cells exposed to oxygen and reactive oxygen species (ROS) (Pandey et al., 2003). Reactive oxygen species and superoxides are formed through natural biochemical processes, but have been found to increase during pollution exposure ( Van der Oost et al., 2003). An increase in SOD levels indicates high levels of ROS that need to be broken down. PAHs have been found to induce SOD levels in common carp (Cyprinus carpio) (Van der Oost et al., 1998) and dab (Limanda limanda) (Livingstone et al., 1993). PAHs that are phototoxic, like fluoranthene, often generate free radicals when photosensitized or photomodified (Delistraty, 1997).
Catalase (CAT) is the second enzyme of the enzymatic defence against ROS (Bartosz, 2005). These enzymes are mainly responsible to counteract the toxicity of peroxide (Lionetto et al., 2003). CAT is produced in response to the increase of ROS, reducing the hydrogen peroxide formed by SOD (Pandey et al., 2003). Hydrogen peroxide is produced through the reaction of superoxides and water, and is important on a cellular level to fight infections. Even though it is an important compound, high levels can cause oxidative damage to cells (Pandey et al., 2003). Catalase activity is indicative of the levels of an oxidative stress compound within the organism and an induction of catalase activity by PAHs has been found in the livers of dab (Limanda limanda) (Livingstone et al., 1993) and in shorthorn sculpin (Myoxocephalus scorpius) (Stephensen et al., 2000).
Malondialdehyde content (MDA) and protein carbonyls (PCs), discussed below, are by-products of oxidative damage and are used to measure the extent of the oxidative stress (Parvez & Raisuddin, 2005;Üner et al., 2006). Quantifying MDA and PCs are meaningful indicators of pollution in freshwater and marine ecosystems ( Van der Oost et al., 1994).
Lipid peroxidation prediction is important in toxicology research pertaining oxidative stress and is quantified by measuring MDA content (Üner et al., 2006). MDA is formed when lipid membranes degrade due to oxidation (Solé et al., 2006(Solé et al., , 2010. Lipid peroxidation is a major contributor to cellular damage, as it can affect the cellular antioxidant system (Ferreira et al., 2007). Malondialdehyde is one of the final products of the peroxidation of polysaturated fatty acids (Gawel et al., 2004). MDA content is therefore used to indicate if lipid damage occurred in an organism due to oxidative stress. It is an important biomarker as the levels reflect the severity of lipid peroxidation in an organism.
The direct damage or oxidation of amino acid in proteins results in the formation of protein carbonyls.
If the PCs increase it can cause damage to cellular systems and tissue and once PCs are formed they cannot be reversed (Parvez & Raisuddin, 2005). Protein carbonyls decrease enzymatic functions and can cause delayed protein regeneration (Ferreira et al., 2007). It is the most commonly used biomarker of protein oxidation (Dalle-Donne et al., 2003).
The energy availability and food intake of fish vary due to seasonality, food availability, flow rate, trophic level and life stages and their energy budgets are often compromised by the increase of stress and pollutant exposure (Gourley & Kennedy, 2009). Assessing the energy budget of organisms at a cellular level is a quick method of determining the effects of stress on that organism. The method entails determining the reserve energy as well as the energy consumption of the organism because the exposures of organisms to contaminants result in a net decrease of energy budget (De Coen & Janssen, 1997).

Bio-indicators
Biological indices can be used as bio-indicators as they explain the condition or state of a biological variable/system (often at levels of organisation higher than cellular level). An index is a formula derived number that is used to describe a condition or status. It represents a single value that is made from a variety of mathematical ratios of different variables, which says something about the sample or site, in terms of the variables, to allow for comparison over time.
The fish health assessment index (FHAI) is a multi-aspect index, which incorporates many variables into one final value, to describe the overall health of individual fish as well as the population sampled.
Other useful indices are the organo-somatic indices which include Fulton's condition factor (CF), the gonado-somatic index (GSI), the hepato-somatic index (HSI), and the spleen somatic index (SSI).
The latter three indices incorporate the value of each index's organ mass such as mass of gonad, liver or spleen, into a percentage in terms of the body mass. The calculated value is used to describe the state of physiological systems in numerical values. These indices are used in temporal and spatial comparisons. In addition to these applications, CF is also gauged against a specific scale, expressing different conditions of fish.
Adams and co-authors (1993) described the FHAI as a rapid and inexpensive quantitative index. It was developed as a field necropsy method, where the results provide a health profile of the fish, based on ratio of abnormalities observed in the tissue and organs of individuals sampled from a population (Goede & Barton, 1990;Adams et al., 1993). The variables assessed are divided into three categories: haematological assessment, external-, and internal evaluation. The haematological assessment consists of the measuring blood protein levels, haematocrit, and leukocrit. The external evaluation is the degree at which abnormalities are present in the eyes, opercula, fins, and skin. The internal evaluation consists of gauging between normal and abnormal internal organs-and if abnormal, what abnormalities or alterations are present in the gills, liver, spleen, kidney, and hindgut.
In order to do quantitative statistical analysis, variables are assigned numerical ranking scores based on the degree of severity or alteration. A FHAI variable score of 30 indicates a severe abnormality or alteration for that variable, where decreasing severity is scored 20, 10 and 0. A score of zero indicates normal state or absence of abnormalities (Adams et al., 1993). The FHAI score for individual fish is calculated by adding all the variable scores whereas the site FHAI is the mean of the FHAI of all the fish sampled at that site. The standard deviation of the mean FHAI score represents the variability of the health of the fish caught at the site and the coefficient of the variance indicates the level or degree of the health of the fish population. The FHAI allows for statistical comparisons of fish health between data sets (Adams et al., 1993).
The general well-being of fish is described in terms of their condition (poor, good or excellent).
Fulton's condition factor (CF) shows the volumetric relationship between the body mass and the total length of the fish and in the formula is represented by the letter K. It expresses the condition (wellbeing, relative robustness or fatness) in numerical terms (Mortuza & Rohman, 2006), where the general condition of the fish is proportionate to the factor value (K). The factor calculated is compared to a range of values that describes the condition of the fish. The condition factor was designed for fusiform fish like salmonids (Barnham & Baxter, 1998) and poses a problem when investigating dorsal ventrally flattened fish such as Clarias gariepinus (see section 1.4.3). Lückhoff (2005) investigated the condition factor for C. gariepinus. His findings showed that catfish in good condition had condition factor values above 0.85 and that the upper limit for C. gariepinus' condition factor range is 1.04 (excellent condition) and the lower limit, 0.6 (poor condition). Studies have shown that PAHs reduce growth and condition in fish (Dutta et al., 2005;Meador et al., 2008).
Fish livers are regarded as the main site of storage, bio-transformation and excretion of pollutants (Hinton & Laurén, 1990;Velmurugan et al., 2007) as well as storing energy reserves in the form of glycogen (Miranda et al., 2008). The hepato-somatic index (HSI) is the relationship between the liver mass and body mass, and indicates the energy reserves of the fish or the effects of xenobiotics on the liver. The HSI may increase following exposure to pollutants due to an increase in hepatocyte size and numbers  in order to increase the liver's detoxification potential (Goede & Barton, 1990). In contrast pollution can also reduce the HSI-with a decrease in cell size and number or even atrophy of hepatocytes (Sanchez et al., 2008;Marchand et al., 2009). However, it is important to keep in mind that the size of the liver is also affected by various other variables such as energy stores and food availability, parasites and seasonal changes (Goede & Barton, 1990;Sanchez et al., 2008). To interpret the HSI results, the previously mentioned factors must be considered. PAHs have previously been demonstrated to have effects on the livers of fish. The livers of Japanese medaka (Oryzias latipes) and guppies (Poecilia reticulata) exposed to waterborne benzo(a)pyrene and 7,12dimethylbenz(a)anthracene (DMBA) had developed hepatocellular neoplasms .
Liver neoplasm in the American catfish, the brown bullhead (Ameiurus nebulosus), was associated with high PAHs recorded in the sediments Smith et al., 1994). Karami and coauthors (2016) found that acute phenanthrene toxicity to Clarias gariepinus juveniles included many histopathological alterations to their livers. These alterations included vacuolisation, bleeding and a decrease in liver cells, together with the cellular degradation and necrosis .
The spleen is a lymphatic organ of which the main function is to produce and store blood (Fänge & Nilsson, 1985). It also plays a role in antigen and erythrocyte degradation, and antibody production (Goede & Barton, 1990;Rohlenová et al., 2011). The spleeno-somatic index (SSI) is the relationship between the spleen mass and body mass, and is used as an indication of immuno-responses (Rohlenová et al., 2011). Swelling or enlargement of the spleen can be indicative of disease in the fish or related to immunological problems (Adams et al, 1992). Fish immune systems are very sensitive to PAHs, which can affect either specific or non-specific immunity (Reynaud & Deschaux, 2006). Nonspecific immunity effects include the decrease of lysozomes and phagocytes (Reynaud & Deschaux, 2006). Examples of specific immunity effects are the reduction of total white blood cells in Nile tilapia (Oreochromis niloticus) (Hart et al., 1998), as well as the depression of lymphocytes in medaka (Oryzias latipes) (Carlson et al., 2002;, spot (Leiostomus xanthurus) (Faisal & Huggett, 1993), and common carp (Cyprinus carpio) (Reynaud et al., 2003) that had been exposed to various PAHs.
The GSI expresses the gonad size relative to the body size to describe sexual maturity or growth. It is used as a popular, simple and instantaneous measure of reproductive effort of a fish (Fouche et al., 2010). It may also indicates irregularities, such as enlargements or tumours caused by contaminants (Stentiford et al., 2003) when individuals are compared with each other within a sample. This index is used to report on reproductive potential between sites and can also possibly indicate endocrine disruption within a site between individuals, based on gonad size (atrophy or hypertrophy). The potential impact that PAHs have on gonadal growth is easily measured in the field using the GSI.
However, more sensitive and specialised methods exist to evaluate these effects such as gamete quality (Casillas et al., 1991;Nagler & Cyr, 1997) Applying the methods described previously in an integrated assessment, where various endpoints are investigated, would give a better representation of the overall health state of these animals, and so indirectly the ecosystem.
It is important to state that although this study is to assess PAHs in the aquatic environs of Soweto and Lenasia, the health effects seen in the fish sampled cannot be wholly attributed to PAHs only.
The fish sampled for this study had been exposed to many different stressors. However, the preceding literature review contains evidence of PAH specific influences on the respective biomarkerand bio-indicator responses.

Sample collection
Fish samples were collected from the four fish sampling sites proposed for this project (Figure 1.2). and were depurated at the Water Research Group aquarium at the NWU Potchefstroom Campus for 6 months. They were kept in aquarium standard water that was replaced every fortnight. They were processed in the same manner as the fish sampled from the study sites.

Fish sampling and field necropsy
Clarias gariepinus was sampled during the high flow season (October) of 2013 and 2014, as described in Chapter 1 (Sampling of fish). Fish were kept in an aerated container until field analysis was performed.
The necropsy protocol followed was a modified version as described by  (adaption of Goede & Barton, 1990) and the scoring system of Adams et al. (1993). Although collecting blood for the haematological assessment was completed, it was not included in our final FHAI score, due to unforeseen circumstances which lead to the loss of sample. The euthanized fish-severed spinal cord-were measured (total length, TL) and weighed, and the values recorded. The external characteristics, which are the eyes, skin, fins, gills, opercula and number of external parasites, were macroscopically evaluated for abnormalities or injuries, and scored accordingly (Adams et al., 1993).

Sample preparations
Three stock batches were prepared from the samples collected. The mass of the tissue samples were noted and the appropriate buffer added. The first batch (labelled A) was for CAT, SOD, PC and CYP450 activity: a mass of 0.1 g liver tissue was added to 1 mL general homogenising buffer (GHB) Protein content of each sample of each batch was determined using the Bradford (1976) method. This method is based on the binding of Coomassie brilliant blue (active ingredient in Bradford's reagent) dye to proteins and then measuring its absorbance (optical density was read at 590 nm). The sample's absorbance was compared to a protein standard curve of bovine serum albumin (BSA) at a concentration series varying between 0 and 2 500 µg/mL. Protein content is determined because the biomarkers are expressed as activity per milligram protein.
All assays (except for CYP450) were done in clear 96 well microtitre plates. The CYP 450 assay was completed using the black half area 96 well plate included in the kit. Optical density was read on a Berthold multi-mode microplate reader (LB941) and fluorescence on a BioTek multi-detection microplate reader (FLx 800).
All reagents used in the biomarker analysis were acquired from Sigma-Aldrich, unless otherwise stated.

Acetylcholinesterase activity assay
The samples used in this assay was prepared according to batch B (see Sample preparations), which was centrifuged at 9 500 g for 10 minutes at 4°C. The procedure of determining AChE activity was adapted from Ellman et al. (1961) was added to the wells and mixed by tapping. The kinetic reaction was recorded immediately by measuring OD at 450 nm every minute for 6 minutes starting at time = 0 minutes (7 intervals). This assay is based on measuring the enzyme activity as a yellow colour is produced when thiocholine reacts with the Ellman's reagent (Ellman et al., 1961).
acetylcholine AChE thiocholine + acetate thiocholine + DTNP yellow colour (Ellman et al., 1961) AChE activity was calculated by determining the mean absorbance (Abs) of the readings at each time interval. The gradient for each sample was calculated using the absorbance (y-values) versus time (xvalues). The tempo (absorbance per minute) was calculated by dividing the gradient by the assay time (6 minutes) and then by the sample's protein content (see Sample preparations). The AChE activity is expressed as Abs/min/mg protein (Ellman et al., 1961).

Cytochrome P450
The cytochrome P450 activity was determined using the DetectX® P450 demethylating fluorescent activity enzyme linked immuno-sorbent assay ( instructions) and incubated (37°C for 30 minutes). The reaction was stopped with the 5 µL glacial acetic acid that was included in the kit. The DetectX® formaldehyde detection reagent was added to each well using repeat pipette (25 µL), gently tapped to ensure mixing, and incubated for 30 minutes at 37°C. The fluorescence of the samples was read at 510 nm after an initial excitation at 450 nm.
Sample concentrations were determined from the standard curve and reported as nM CYP450/mg protein (protein content determination see Sample preparations).

Superoxide dismutase activity
The SOD method was adapted from Del Maestro & McDonald (1989). Samples used for the SOD assay was from batch A. A 50 mM Tris buffer containing 1 nM diethylene triamine penta-acetic acid (DTPA) (49:1 v/v) was prepared and aerated vigorously for 20 minutes at room temperature. The pH was adjusted to 8.2 by adding HCl and aerated again for another 10 minutes. The assay was performed on ice until quantification. The samples and a Tris buffer blank (4 μL) together with 245 μL DTPA/Tris buffer was added to the wells of the 96 well microtitre plate in triplicate. The reaction was started by adding 4 μL pyrogallol (24 nM in 10 mM HCl) and the kinetic reaction recorded by measuring the OD at 560 nm every 30 seconds for 5 minutes starting at time = 0 minutes (11 intervals). This assay was performed in the dark as pyrogallol is light sensitive. This assay is based on the ability of SOD to scavenge superoxide anions (O2 -). The removal of superoxides during the assay reduces the overall autoxidation rate of pyrogallol. One unit of SOD activity is defined as the amount of enzymes that inhibit the rate of pyrogallol's autoxidation by 50% (Del Maestro & McDonald, 1989). The absorbance gradient and tempo was determined. The sample tempos and the blank were compared to calculate the inhibition magnitude (by normalising the tempo into a percentage) and then expressed in terms of 50% inhibition (dividing by 50). This value represents the SOD activity in units.
To convert SOD into a concentration value the unit SOD value is multiplied by 125 ngSOD per unit (125 ng/mL represents one SOD unit). The fraction responsible for enzymatic activity was determined multiplying the dilution factor with the buffer volume. Finally, the SOD was expressed in terms of is protein content (see Sample preparations) as ngSOD/mg protein (Del Maestro & McDonald, 1989) Catalase activity Catalase activity was measured using the method of Cohen et al. (1970). The CAT assay is light sensitive and was performed in the dark, on ice. The sample supernatant (batch A; centrifuged at 10 000 g for 10 minutes at 4°C) and a blank (PPB) added (10 µL) to the 96 well plate in triplicate.
Similarly 102 µL PPB was added into three wells as a standard. Only 10 samples were analysed at a time. The reaction was started by adding hydrogen peroxide [6 mM] (93 μL) to each well (samples and blank only) and mixed gently by tapping. The plates were incubated at room temperature for 3 minute. The reaction was stopped by the addition of 19 μL sulphuric acid (6 N). Using a repeat pipette KMnO4 (130 μL) was added and immediately OD was read at 490 nm. The CAT activity assay is based on the first order kinetics and linear dose response relationships of catalase and its substrate hydrogen peroxide. The colour change that is quantified by OD is from the reaction of KMnO4 and the remaining H2O2 (lowering the OD). The first order kinetics of the decomposition of H2O2 by catalase is as follows: where: k is the first order reaction rate constant; S0 is the substrate concentration at time zero (mean of standard absorbance); S3 is the substrate concentration at 3 minutes (standard absorbance subtracted by the mean sample absorbance); 2.3 a first order kinetic conversion factor; and t is the time interval (3 minutes). After the calculation of k, the value is normalised in terms of the protein content (divided by protein content, see Sample preparations) and adjusted to a µmol unit (multiplied by 1000). The CAT activity is reported as µmol H2O2/min/mg protein.

Malondialdehyde content
The methodology from Ohkawa et al. (1979) as modified by Üner et al. (2006)  to the wells of a microtitre plate in triplicate. The absorbance was read at 540 nm and the MDA content calculated using the TMP standard curve. The MDA content was expressed in terms of its protein content (see Sample preparations) as nmol MDA/mg protein. This is a colorimetric assay that is measures the reaction products of TBA and a degradation product of lipid peroxidation, MDA. This method has to follow very specific conditions, specifically high temperature and acidity, to generate the pink coloured adduct that is quantified.

Protein carbonyl induction
PC content was assayed as originally described by Levine et al. (1990) and modified by Floor & Wetzel (1998) (Parvez & Raisuddin, 2005. Samples from batch A were centrifuged at 10 500 g for 30 minutes at 4°C to prepare a supernatant containing the soluble proteins in each sample. The supernatant of the homogenates, a hydrochloric acid [2 M] blank, and a bovine serum albumin (BSA) standards [0-40 nmol] was added (500 μL) to 2 mL centrifuge tubes. An equal amount (500 µL) of 2,4-dinitrophenylhydrazine (DNPH) [10 mM DNPH in 2 M HCl] was added to the samples and standards (not blank) and allowed to react for an hour. After the reaction time proteins were precipitated using 500 μL trichloro-acetic acid (TCA) [6%] into all the tubes. From here samples, standards and the blank were processed in the same manner. The solution was centrifuged at 10 000 g for 3 minute to form a pellet of precipitated proteins.
The TCA (supernatant) was discarded carefully. The protein pellets were resuspended in 1 mL absolute ethanol/ethyl ether [1:1 v/v] and mixed by vortex. After 10 minutes the samples were centrifuged at 10 000 g for 3 minute. This washing step was repeated three times. At the end of the final wash step the samples were centrifuged for ten minutes (10 000 g) and the supernatant discarded, care taken not to lose the protein pellet. Guanidine hydrochloride [6 M in 50% formic acid] was added (400 µL) to dissolve proteins. The sample incubated for 15 minutes (37°C) and centrifuged again for 5 minutes (16 000 g) to remove any remaining insoluble materials. The supernatant was added (100 µL) to the microtitre plate in triplicate. Absorbance was read at 390 nm and carbonyl content quantified from the BSA standard curve. Protein carbonyls were quantified as nmol carbonyls/mg protein (see Sample preparations). This assay is dependent on the ability of DNPH to react with carbonyl groups in proteins. The 2,4-dinitrophenylhydrazine binds to carbon-oxygen double bond of the carbonyl to form an intermediate compound that then loses a water molecule and ends as 2,4-dinitrophenylhydrozone, which is the product quantified in this assay.

Energy allocation biomarkers
Cellular energy allocation CEA analysis was adapted from De Coen & Janssen (1997, 2003 for determination of total available energy (total protein, -carbohydrates, and -lipids) and energy consumption via the electron transport system (ETS) activity.
The homogenised muscle tissue (batch C) was divided into two sets. The first was used for determination of the available energy (Ea), where 100 μL of the homogenate was further diluted with 400 μL deionised water. The samples used for energy consumption (Ec) consisted of 100 μL of the homogenate mixed with 400 μL ETS buffer.

Available energy (Ea)
The available energy is calculated by determining the total protein contents, carbohydrate contents as well as the lipid contents of each sample (batch C). Samples were mixed by vortex before the total proteins were determined following the method of Bradford (1976)

Energy consumption (Ec)
The energy consumption was quantified using an electron transport system (ETS) activity assay. The water v/v]. The absorbance was read kinetically at 490 nm for 5 minutes (11 intervals). Energy consumption is given as J/g. The kinetic reaction was recorded by measuring OD at 490 nm every 30 seconds for 5 minutes starting at time = 0 minutes (11 intervals). This assay is based on the theoretical stoichiometrical relationship that each 2 µmol formazan crystals formed by the reduction of INT, 1 µmol oxygen is consumed. The amount of oxygen consumed was transformed into energetic equivalents by an oxyenthalpic equivalent (average for a protein, carbohydrate and lipid mixture) (De Coen & Janssen, 2003). The calculation of energy consumption was as follows: the gradient of the samples was calculated using the absorbance (y-values) versus time (x-values) followed by calculating the tempo (absorbance per minute) by dividing the gradient by the assay time (5 minutes).
The tempo was then divided by the amount of formazan formed. This is determined by multiplying the formazan extinction coefficient (ε = 15 900/M cm) with the depth of the light path length of the total volume of the well (0.7 cm) and the volume used in the well (0.075 mL) (De Coen & Janssen, 2003).
As stated above for every 2 µmol formazan formed one µmol oxygen is consumed (halving the formazan content). The energy used to reduce oxygen (energy consumed, mJ) is the product of the amount of oxygen (µmol) and the oxyenthalpic equivalent (484 kJ/mol O2) (De Coen & Janssen, 2003). Finally, the energy consumption in the organism is the product of the energy required to reduce oxygen and the inverse of the sample mass (1/0.2 g), mJ/g adjusted to J/g.

Health assessment indices calculations
The individual fish FHAI scores are calculated by adding all the variable scores. The population or site FHAI score is calculated by determining the mean of the individual FHAI scores (Adams et al., 1993).
FHAI= ∑ Individual FHAI score n n A standard deviation and coefficient of variance for each site was also calculated Adams et al. (1993).
Where n is the number of fish sampled at that site, X is the mean FHAI value at that site and V i is the FHAI value for fish i.
The mass and length of each fish recorded during the necropsy was used to calculate Fulton's condition factor, by using the formula described by Bolger & Connolly (1989): Where K is Fulton's CF, W is mass in gram (g), and L is total length in centimetres (cm). The GSI, HSI and SSI were calculated by expressing the index respective organ mass as a percentage of the gutted body mass (Rohlenová et al., 2011).

OSI=
Organ mass (g) Gutted body mass (g) x 100 The evaluation of fish condition often requires the measurement of the parameters of the same fish over time to reveal a more accurate condition over time-as factors such as climate, food availability, and gonad maturation affect the results (Bolger & Connolly, 1989). For this project however, only single CFs were recorded to indicate the condition of the fish at the time it was sampled. These results were used in conjunction with the remaining indices to report on the health of the fish.

Statistical analysis
The results obtained with the biomarker-and bio-indicator assessments were tested for normality using the D'Agostino & Pearson omnibus normality test. One-way analysis of variance (ANOVA) was performed if the data were distributed normally and Tukey's multiple comparison test as a post-test.
The Kruskal-Wallis test was used together with Dunn's multiple comparison test as post-test for nonparametric data sets. A p<0.05 was considered significant. Significance was indicated by common conscripts on graphs for seasonal differences per site and spatial differences within a survey.
Statistical analysis was done using Graphpad Prism version 5.

Results and discussion
The following results are for the biomarker and bio-indicator assessments done on the fish sampled from Soweto and Lenasia. Biometric information of Clarias gariepinus sampled from the study area is given in Table 4.1 and the biomarker responses summarised in (Figure 4.2-4.5) The bio-indicators such as the fish health assessment index (FHAI) is presented in Figure 4.6-4.7 and the other body indices in Figures 4.8-4.11.

Biometric information of Clarias gariepinus sampled
There were more females caught than males for most of the sampling events in both years except for Fleurhof and Nancefield 2013 (Table 4.1).
There were no significant differences between the sites for the mass of the fish sampled or their total lengths, indicating fish were all similar in size and are comparable when interpreting results. The sex of the fish also had no significant effect on the mass and length of the fish.

Biomarker responses
The stress induced by xenobiotic stressors leads to biochemical irregularities, which results in the uptake, biotransformation and elimination of that compound . Although only PAHs have been quantified in the study area, and their metabolites detected in the fish, other contaminants might have influenced the biomarker responses. Table 4.2 summarises already known increases/decreases of the various biomarker responses as a reaction to specific exposures and effect, aiding the interpretation of the results discussed below. Decrease due to stress compensation (higher energy demand). Increase due to increase in energy sources

Biomarkers of exposure
There was an inhibition of AChE in the fish at all the sampling sites during both years compared with the AChE levels in the control fish, but the 2014 AChE results were lower than those of 2013. Four of the sets of fish had AChE activity significantly lower than the control fish indicating inhibition of AChE activity (Figure 4.2A). The greatest inhibition for the 2013 fish were at Fleurhof (5 x 10 -4 ± 2 x 10 -4 Abs/min/mg protein) and Nancefield (6 x 10 -4 ±2 x 10 -4 Abs/min/mg protein) which had 56% and 43% inhibition of AChE compared to the control. The highest AChE value was at Lenasia (1 x 10 -3 ±4 x 10 -4 Abs/min/mg protein), and only had 14% inhibition compared to the control. All the 2014 AChE results showed that the enzyme had been inhibited more than 50%, ranging between 71 and 80% (2-3 x 10 -4 Abs/min/mg protein) (Figure 4.2A). The inhibition of AChE activity higher than 50% may lead to obvious toxic effects, and an inhibition higher than 80% may lead to the death of those individuals (Connell et al., 1999). AChE activity declined significantly (p<0.05) between the consecutive years for Lenasia and Nancefield only.
PAHs have been found to inhibit acetylcholinesterase (Kang & Fang, 1997;Lau et al., 2004). Kang & Fang (1997) showed that several PAHs had dose dependant inhibitory effects on the AChE activity, in vitro, of the electric eel, Electrophorus electricus. These authors found that anthracene, benzo(a)pyrene and chrysene had the highest inhibitory effect (IC50), in that order. These PAHs congeners were also found to have high competitivity to inhibit AChE, i.e. high Ki values (Kang & Fang, 1997). The AChE activity in the livers of Oreochromis mossambicus exposed to sewage water was lower than their control (27% inhibition) (Al-Ghais, 2013), and although the author did not quantify any contaminants in the sewage PAHs have been found in sewage in other studies (Wild & Jones, 1989;Mo et al., 2001). The Nancefield site was the last pond of a waste water treatment plant (See section 1.3.2) and showed a similar decrease in AChE activity compared to the control fish (43% and 80% (Figure 4.2A) than the Al-Ghais (2013) study. In Tanzania, Mdegela et al. (2010) also investigated sewage exposed fish. These authors assessed various biomarkers in wild C. gariepinus and quantified PAH metabolites in the fish as well. The AChE activity in the brain tissue of the Tanzanian C. gariepinus was 33% inhibited which was higher than the AChE inhibition in the Orlando and Lenasia catfish of 2013 (24% and 13% respectively), but lower than the inhibition of the other Soweto and Lenasia sites.
Clarias gariepinus was sampled from the Phongolo River system to investigate organochlorine pesticides (OCPs) in this malaria vector control area (Edwards et al.,2016). These fish were expected to have been exposed to malaria vector control (MVC) pesticides, which include OCPs (specifically dichlorodiphenyltrichloroethane, DDT), pyrethroids, carbamates and organophosphates (Brooke et al., 2013). Of the preceding list the organophosphates, pyrethroids, and carbamates inhibit AChE acitivity (Lionetto et al., 2013). The AChE results of the Phongolo C. gariepinus showed no significant changes between the two sampling years (2012 and 2013) even though there was a significant increase in ΣOCPs (Edwards et al.,2016).
Clarias gariepinus was sampled downstream from Soweto and Lenasia, from the Lower Klip River   Padros et al. (2000) investigated the hepatic metabolising enzymes in brook trout (Salvelinus fontinalis) that had been exposed to benzo(a)pyrene interperitoneally. The activity of EROD and ethoxycoumarin-O-deethylase (ECOD) were quantified. EROD is predominantly mediated by cytochrome P4501A and ECOD is considered a non-specific enzyme that is catalysed by several forms of cytochrome P450 (Padros et al., 2000). These authors reported significantly up-regulated levels of EROD (170%) and ECOD (187%) compared with their control. Similarly, Levine and Oris (1999) showed that waterborne BaP increased EROD in rainbow trout (Oncorhynchus mykiss) by 38and 88-fold after 24 hour exposure to 0.42 and 1.13 µg/L BaP respectively. Vigano et al. (1995) also investigated liver microsomal enzyme activity in O. mykiss, injected with sediment extracts from sites previously proven to contain PAHs and PCBs. These crude extracts with unknown pollutant concentrations were transferred into corn oil, of which the final volume added depended on the %TOC in each extract. This allowed that the sites had the same nominal concentration of organic carbon and that the exposure range could be expressed in terms of TOC.
The dosages injected into the fish were based on the dry mass of the sediments extracted per kilogram body mass (BM) of the fish which corresponded to 6, 28, 138 & 690 mgTOC/kg BM. Plain corn oil was also injected as a blank. After six days of exposure EROD and AHH activities were measured. The AHH activity of the lowest TOC dose was lower than their control (12.8% inhibited).
The AHH activity increased with the increase of dosage but was inhibited at highest concentration (39% lower than control) (Vigano et al., 1995), suggesting that the PAH containing extracts (as reported as the bile fluorescent metabolites) the inhibited this metabolising enzymes. Likewise, the EROD system also followed the same increasing trend with the final concentration (690 mgTOC/kg BM) a decrease activity was noted. The first dosage elicited a 112% increase in EROD activity followed by a 125% and 267% increase for the next two dosages (28 & 138 mgTOC/kg BM). The decrease of activity at the highest concentration is not considered as inhibition as it still has a significantly higher activity than the control (169%) (Vigano et al., 1995).
A difference in organ specific EROD activity in C. gariepinus was noted by Mdegela et al. (2006).
Depurated fish were exposed to waterborne BaP and after a 4 day exposure EROD activity was determined in the gills and livers of these fish. An 11 and 17 fold EROD increase in response to BaP in the gills of laboratory exposed female and male Clarias gariepinus had been reported, whereas the EROD activity in the liver of both sexes increased by two fold (Mdegela et al., 2006). Naicker et al.
(2007) established a primary cell line of C. gariepinus hepatocytes. They validated the cell line by exposing the cells to BaP and measuring the EROD response. The hepatocytes exposed to 0.01, 0.1 and 1 µM BaP showed significant EROD activity increases: 96%, 129%, and 68% respectively (Naicker et al., 2007).
In environmental studies, increases of cytochrome enzymes in PAH polluted areas have been reported. Hepatic EROD activity and the concentration of bile fluorescent aromatic compounds (FACs, ngBaP-eq/mL) were measured in brown bullhead (Ameiurus nebulosus) sampled from the lower Great Lakes (Arcand-Hoy & Metcalfe, 1999). The authors reported a significant increase in EROD activity (between 120 and 200%) at sites with high levels of biliary PAHs (89-4 649 ngBaPeq/mL). These sites, Hamilton harbour, Black-and Detroit Rivers also had high levels of ΣPAHs in the sediments (Table 2.5).
Hepatic CYP450 induction was also seen in caged channel catfish (Ictalurus punctatus) exposed to river water contaminated with PAHs and PCBs (Haasch et al., 1993). There were slight inhibition of EROD, CYP1A1 and total CYP450 activities in these fish after one day of exposure. However, the expression of hepatic cytochromes increased as the exposure period increased. After one week's exposure EROD and ECOD were 200% and 12% higher than the control. However, the total CYPs were 44% lower than the control. When the fish were exposed for 14 days, the EROD and ECOD results increased significantly to 822% and 233% induction relative to the control (Haasch et al., 1993) but the total cytochrome P450s did not show any significant fluctuation, even with the high induction of EROD and ECOD. In a study on C. gariepinus from sewage ponds in Tanzania there was high EROD activity increases compared to fish from their control site. An increase of more than 350% was noted in these fish (Mdegela et al., 2010).
High levels of PCBs, OCPs, and polybrominated diphenylethers (PBDEs), and detectable levels of metals in the muscle of Labeo umbratus, from the Vaal River, have been reported by Wepener et al., (2011). These fish showed an increase in EROD activity, however the site that had the greatest levels of PCBs and other organic pollutants had the lowest EROD activity , indicating to possible inhibition of this enzyme system in this fish species.
As mentioned earlier, the fish from the Phongolo system is known to be subjected to various pesticides used for MVC. The total CYP450 activity measured in the C. gariepinus of this system increased 3.75 times, as the ΣOCPs quantified increased (from 4.68 ng/g lipid mass to 479 ng/g lipid mass) (Edwards et al.,2016).
Finally, C. gariepinus the Lower Klip River had significantly lower CYP450s (p<0.05) compared to the Soweto and Lenasia sites. The highest CYP450 was 0.012±0.003 nM/mg protein)  which is in the same range as the control of this study (Figure 4.2B). As already mentioned, fish from these sites where exposed to metals and OCPs (320.57 ng/g lipid mass) in the environment .
When comparing the cytochrome results of the two American catfishes, Ameiurus nebulosus (Arcand-Hoy & Metcalfe, 1999), and Ictalurus punctatus (Haasch et al., 1993), as well as C. gariepinus form the Tanzanian study by Mdegela et al. (2010), to this study, keep in mind that the total cytochromes did not seem to have been affected as much as the individual monooxygenase (MOs) enzymes. The total cytochromes generally are a less sensitive assay (Bucheli & Fent, 1995). The reaction to xenobiotics by single isoenzymes may differ considerably, inhibiting one form and up-regulating another, and so not affecting the total CYP450 levels (Miranda et al., 1990;Van der Oost et al., 2003).
The responses of total cytochromes and the specific MOs of freshwater fishes to various pollutants have been summarised by Van der Oost et al. (2003) and is shown in Table 4.3. From literature (Table 4.3) it is evident that the total cytochromes do not fluctuate as regularly as the specific monooxygenases, especially when the fish were exposed to PAHs. In the instances where the total cytochromes were increased, these specific MOs were strongly induced (Table 4.3). In these cases halogenated organic pollutants, such as PCBs, dioxins, and to a lesser extent PAHs, were responsible for the observed responses (Table 3.4).
Although AHH and EROD were not quantified for these fish, PAHs quantified in the sediment and fish bile (Table 2.7 & 2.10), and the activation of the AhR (Table 3.1) suggest that these enzyme systems must have been induced, contributing to the total CYP450 activity. However, as stated earlier the AHH and EROD activity together do not necessarily increase the total CYPs (Table 4.3) and the increased induction seen for the fish of Soweto and Lenasia can therefore not be attributed to PAHs alone. Pollutants such as PCBs and OCPs-previously quantified in the study area )-together with the quantified PAHs can be attributed to the increased total CYP450 activity.

Biomarkers of effect
The biomarkers of effect results are divided into three sections: 1) the oxidative stress biomarkers 2) the biomarkers of oxidative stress damage, and 3) cellular energy allocation. The oxidative stress biomarkers are superoxide dismutase (SOD) and catalase (CAT) and are reported in Figure 4.3 A & B, respectively.

Biomarkers of oxidative stress
Superoxide dismutase forms the first tier of the anti-oxidant system responsible for converting reactive oxygen species (ROS) and superoxides into oxygen and hydrogen peroxide and therefore an increased SOD level may indicate high levels of ROS.
The SOD levels were elevated in the Lenasia fish from 2013 at 0.88±0.36 ngSOD/mg protein; which was a 160% increase compared with the SOD levels of the control. This was the only increases that was statistically significant (p<0.05). The smallest increase of SOD was found for the Fleurhof fish of 2014 at 0.46±0.23 ngSOD/mg protein which amounted to a 39% increase (non-significant) ( Figure   4.3A) Where levels were available for both years, the 2013 levels were always visually greater than that of 2014 (Figure 4.3A).
The increase in the SOD levels at Lenasia (2013) suggests that the organisms experienced oxidative stress and that the enzyme system was up-regulated to decrease these oxidative compounds to prevent cellular damage (Pandey et al., 2003).
Otitoloju and Olagoke (2011) reported that crude oil-a known PAH source-has an inhibitory effect on SOD in C. gariepinus. These authors showed a 83% inhibition, which was significantly lower relative to their control (Otitoloju & Olagoke, 2011). Almroth et al. (2008) reported similar results. A 21% inhibition of SOD was seen in female corkwing wrasse (Symphodus melops) from a PAH polluted site relative to a reference site. In contrast to this, the male S. melops from the same area had an increase of 9.6% SOD (Almroth et al., 2008). This gender difference was not seen for C.
gariepinus in the present study and SOD activity was similar between sexes.
The SOD measured in the C. gariepinus of the Phongolo River and floodplain increased over the two sampling surveys as the OCPs increased, specifically total DDTs and total hexachlorocyclohexanes (ΣHCH) (Edwards et al.,2016). The SOD in the C. gariepinus from the Lower Klip River reported by Wepener et al. (2015), were all significantly lower (p<0.05) than those upstream (present study). The  The crucian carp (Carassius carassius) is one of China's higher trophic level fish. Ji et al. (2010) transplanted laboratory-reared fish into a known PAH polluted site where PCBs and OCPs also had been detected, as part of active biomonitoring. The fish's reaction was measured using an integrated biomarker response assessment. The sites with the highest PAH concentrations also had the highest biomarker responses. Catalase activities at the PAH polluted sites were significantly higher than their control with a 40-66% increased activity (Ji et al., 2010). Oliviera et al. (2008) investigated the organ specific antioxidant responses in golden grey mullet (Liza aurata) when exposed to phenanthrene. These fish were exposed to a concentration range of 17.8-480 ng/mL phenanthrene for 16 hours, after which oxidative stress biomarkers were analysed (Oliviera et al., 2008). These authors reported significant increases of catalase activity (75%) in the gills of phenanthrene exposed fish to a control group and the CAT activity in their livers was significantly higher than in the gills. The highest hepatic CAT activity was 50% higher relative to their control (Oliviera et al., 2008).
Petroleum refineries are not only sources of PAHs but also phenols, oils and metallic pollutants.
Therefore, Avci et al. (2005) sampled wels catfish (Silurus glanis) up and downstream of a petroleum industry to determine the effects of the wastewater on the oxidant/antioxidant system of the fish living near the industry. They found a small increase of catalase activity in the livers of S. glanis between the two sites (11%) but it was not significant (Avci et al., 2005). These authors also measured SOD levels and found that there was a significant increase in fish downstream. Comparing SOD and CAT responses of S. glanis where there was a high response of SOD and low response of CAT, which was opposite to what was seen in the C. gariepinus from of Soweto and Lenasia (Figure 4.3 A & B).
The C. gariepinus of the Phongolo system's CAT levels increased similarly to the SOD levels i.e. an increase in enzymatic response as the measured OCPs increased in vitro (Edwards et al., 2016). The catalase activity measured from the Lower Klip River C. gariepinus , was significantly lower (p<0.05) than the CAT activity of the 2014 fish of Soweto and Lenasia. This shows that the fish upstream are responding to oxidative agents, such as metals and pesticides (Wepener et al., 2015).

Biomarkers of oxidative stress damage
The biomarkers of oxidative stress damage measured in this study was lipid peroxidation by measuring malondialdehyde content (MDA) (  years, but only that of Nancefield showed a significant increase in carbohydrate stores from 2013 to 2014. The total carbohydrate level for the Nancefield fish from 2014 was the highest at 548±23 J/g and was 230% higher than the control (Figure 4.5A). Although not the main source of energy, fish use carbohydrates to maintain growth rates, if carbohydrate reserves are depleted the main energy sources which are lipids and proteins will be catabolised for energy to provide the intermediates for the synthesis of biological important compounds (Wilson, 1994).
Lipid storage play an important role in fish health as it is critical for survival, fitness and reproduction (Adams, 1999).   Energy storages enable the fish to survive for a period of time when environmental resources become scarce (Adams, 1999). When an environment is significantly altered or disrupted, energy allocation in fish change. This disruption in natural energy allocation is to provide energy to maintain homeostasis in battling environmental stressors causing the disruption. This disruption may compromise the success of a species (Adams, 1999). The information in the available energy graph (

Bio-indicator assessment
The fish health assessment index (Figure 4.7 & 4.8), and organo-somatic indices  were determined during the bio-indicator assessment of C. gariepinus.

Fish health assessment index (FHAI)
The FHAI allows for the judgement of the general health and condition of a fish population and was developed as a comparative, but relative, biomonitoring tool to compare the same site over time and/or different sites at the same time . The higher the FHAI score shows that the fish are in poorer health.

Figure 4.7: Observed abnormalities during necropsy: A) liver enlargement and darker discolouration; B) altered testes containing vesicles [arrows]; C) increase of connective tissue and fusion [arrows]; D) liver discolouration; E) increased fatty deposits in liver
A B

D E C
The 2014 fish from Fleurhof had relatively the best health of all the environmental sites for both years as the FHAI was the lowest at 45±14 (Figure 4.8) and can be considered to be in fair health. However, they were still in poorer condition than the control fish whose FHAI score was even lower at 22±20 (Figure 4.8). Of the fish caught at the Lenasia site in 2014, 82% had discoloured and/or nodular livers and 36% presented enlarged spleens. Similarly, the fish from Nancefield in the same year had 75% liver abnormalities, 38% enlarged spleens, and 38% clubbed gills. The FHAI scores reported in literature cannot be directly compared to one another if they are not from the same system or for the same fish species. Comparisons made will be relative to each study's control or reference sites and the final mean FHAI scores. In doing so, the effect of known/measured stressors and pollutants on fish can be interpreted. Watson et al. (2012) (Watson et al., 2012). This score is greater than for the current study ( Nematodes were present in 33% of C. gariepinus sampled by Van Dyk et al. (2009). No parasites were observed in and on the fish collected from Soweto and Lenasia aquatic systems. The Okavango fish had a lower FHAI score than all the sites of the current study including the control (Figure 4.8).
The fish sampled by Wepener et al. (2015) from the Lower Klip River had significantly lower FHAI scores than Soweto and Lenasia of both surveys with 19.5±22.5 for 2013 and 13.3±21.1 for 2014 (p<0.05). The FHAI of these fish was mainly caused by pale or discoloured gills (10%) and discoloured livers (25%), and no other abnormalities were noted in these fish .
From literature we can deduce that the fish sampled from Soweto and Lenasia were in poor health.
The wide gap between the health of the sampled fish and that of the aquarium control, and that the difference margin was greater than for studies in known polluted sites, it is clear that the stressors heavily affected the health of Clarias gariepinus from the study area.

Fulton's condition factor
As mentioned previously, the condition factor expresses the condition of the fish, its well-being, relative robustness or fatness at the moment that they were sampled. Lückhoff (2005)   indicate that the fish were not short on food supply and that a decline in overall health could not be attributed to malnutrition. This is corroborated by the colour and volume of the bile sampled during the necropsy. Whenever bile was found in the gall bladder, it was light and dark straw coloured. This is an indication that the fish fed within a few days (yellow bile) or hours (empty gall bladder) of sampling (Goede & Barton, 1990). The fish sampled from the Lower Klip River during the same sampling years, had comparable (no significant differences) condition factors . Clarias gariepinus from the Loskop Dam and Bronkhorstspruit Dam in the Olifants River system had condition factors similar to our findings with 0.84±0.06 and 0.86±0.1 respectively (Watson et al., 2012). The fish sampled in Flag Boshielo Dam and Return Water Dam had higher condition factors, 1.2±0.6 and 1.4±0.8 respectively (Madanire-Moyo et al., 2012). Lower condition factors were calculated for C.
The before mentioned results of Mdegela et al. (2010) were lower than the CF for the fish from the sewage ponds at Nancefield (0.77±0.08 in 2013 and 0.7±0.05 in 2014). Barnhoorn et al. (2004) investigated intersex in wild sharptooth catfish exposed to endocrine disrupting chemicals, such as p-nonyl phenol. There was a noted difference between the condition factors calculated for both sexes and the intersex specimens from the Marais Dam site, Gauteng South Africa. The intersex catfish had a CF of 1.05, higher than the males and female CFs of 0.9 and 0.88-this trend was however not seen at their other site, Rietvlei Dam (Barnhoorn et al., 2004). The condition factors for the intersex fish were higher than all the fish from the current study (

Hepato-somatic index (HSI)
The HSI indicates the energy reserves of the fish or the effects of xenobiotics on the liver. An increase in HSI may be seen after exposure to pollutants due to hyperplasia or hypertrophy  to increase the liver's potential to metabolise toxicants (Goede & Barton, 1990). Pollution may also lower the HSI due to atrophy of hepatocytes or necrosis Sanchez et al., 2008), but it is important to keep in mind that the size of the liver is also affected by various other variables (Goede & Barton, 1990;Sanchez et al., 2008).  showed that the HSI values of C. gariepinus of Southern Africa-sampled from various aquatic systems-are close to 0.6%.
The HSI values are shown in

Spleeno-somatic index (SSI)
The spleen is a lymphatic organ which main function is to produce and store blood (Fänge & Nilson, 1985). It also plays a role in antigen-and erythrocyte degradation and antibody production (Goede & Barton 1990;Rohlenová et al., 2011). Swelling or enlargement of the spleen can be indicative of necrosis and relate to immune problems such as infection and/or disease in the fish (Adams et al., 1992;Goede & Barton 1990).
The SSI results show the immune responses of the fish sampled in this study (Figure 4.11) and an increase in the size can be ascribed to an increase in immune responses (Rohlenová et al. 2011).
There were no significant differences between the sites and the control (Figure 4.11). There were no parasites observed in or on any of the fish sampled, at any of the sites and thus the SSI values could not have been attributed the presence of parasites (Rohlenová et al., 2011) and toxicants suspected for causing the effect.
The SSI reported by Bester (2013) et al., 2013), were lower than all fish sampled during the current study, as well as our control (Figure 4.10). It is probable, but not statistical, that the stressors in the Soweto and Lenasia study area had an effect on the spleen, and so possibly on the immune system of the fish.

Gonado-somatic index
The gonado-somatic index provides information on gonadal health and developmental stages, in response to changes, such as environmental stressors and/or seasonal changes (McDonald et al, 2000). The sampling took place during the pre-spawn season of C. gariepinus, and a majority of the female fish were in the maturing stage for spawn season, except for 4 fish from Orlando with underdeveloped ovaries (these fish were excluded from the GSI calculations). There is a high variation in the GSI, standard errors ranged from 0.98-8.6 (Figure 4.12A). Fish caught in 2013 at the Nancefield site had the highest GSI value followed by the 2014 values of the same site (11.19±4.29 and 10.25±5.44 respectively), and was the only site of which the GSI was significantly higher than that of the control (p<0.05) (Figure 4.12A). There was no correlation between the total lipids available ( Figure 4.5B), suggesting that the lipids were not utilised as energy for gonadal development. Fish from Nancefield (both surveys) and Orlando had the only GSI significantly higher than the C.
gariepinus from the Lower Klip River . The control fish had the lowest GSI of all the fish sampled ( Figure 4.12A & B). The reason for this is that the fish were processed in April 2014, when they have not yet become gravid with eggs as they would have been during October. Bruton (1979) mentioned that the gonads of C. gariepinus are in a resting period during April (in southern Africa,). Maturation of gonads begins in August and by October (spring and early summer in the southern hemisphere) the bulk of the population are sexually mature (Bruton, 1979). The female fish from all the sites (both surveys) were in the developing and maturing stages (IV and V stages) (Bruton, 1979). The lowest GSI from the sampling area was Fleurhof 2013 (  The female fish from Orlando that had been excluded from the calculations due to underdeveloped ovaries that were not necessarily juveniles. Clarias gariepinus reach maturity at a length of 330-340 mm, or within two years (Bruton, 1979). When referring to their size (510-640 mm) and mass (1 000-2 000 g), there were fish within the population smaller than these individuals that had developed ovaries and that were gravid with eggs (assuming that all individuals in the population would be of approximately the same developmental stage). According to the description of the gonad development by Bruton (1979), these fish's gonads fall in Class I (immature virgin) and II (developing virgin) of the gonad maturity stages. This underdevelopment indicates to possible endocrine disruption in these individuals and warrants further investigation.
The male fish from Lenasia 2013 had the highest GSI (1.22±2.61, Figure 4.12B) and was the only population that showed decrease in GSI over the two years. The second highest GSI was from the 2014 Fleurhof fish, with a 0.92±0.46 value (Figure 4.12B). The Orlando males had the lowest GSI (0.18±0.1), which was significantly lower than that of the control (0.63±0.14) (Figure 4.12B), indicating that these male fish had smaller testes. Potential feminization was observed in one male from Nancefield, which had sac like vesicles growing within a deformed testis (Figure 4.7B)-this needs to be confirmed with histology. Feminization and decrease in testes size is indicative of hormonal imbalance possibly from exposure to endocrine disrupting chemicals (Mills & Chichester, 2005).
Intersex in C. gariepinus was seen in fish from the Rietvlei-and Marais Dams. The intersex was linked to endocrine disruption from anthropogenic pollution such as wastewater treatment plant effluent, agriculture or industry (Barnhoorn et al., 2004). The C. gariepinus exposed to sewage effluent had a significant difference in GSI between ponds and reference site. This may be from exposure to compounds in the sewage causing enlargement of the gonads (Mdegela et al., 2010).

Conclusion
From the fish biomarker data it seems that there were environmental stressors that affected the health of the fish on a biochemical level. An inhibition of the acetylcholinesterase system was noted in the fish from Soweto and Lenasia. Apart from this inhibition, exposure responses also included an increase in cytochrome P450 activity. Both the activation of the detoxification system (CYP450), along with the up-regulation of the SOD-CAT defences is evidence that the fish from the study area had been defending against xenobiotic stressors. The lower SOD (relative to CAT) showed that this enzyme system definitely was working against compounds; however inhibition or the sheer overwhelming of SOD could be seen in the high activation of the CAT defences. The low concentrations of malondialdehyde showed that the antioxidant defences had been successful in stopping lipid peroxidation, which were also corroborated by the higher lipid stores. Conversely, the lower protein stores and the higher protein carbonyls showed that there had been damage to proteins in the fish and the increase of enzymatic defences. The energy budget of the fish showed that there had been an increase in available energy over the two surveys, indicating that although there were stressors activating the fish's defences that it had not either yet affected the energy budget of the individuals, or that the defences were successfully in mitigating the threats these stressors pose.
The overall health, on a systematic level, of the Soweto and Lenasia fish was visibly affected by environmental stressors. The Soweto and Lenasia fish were in poorer health than fish sampled in the Lower Klip River. The overall health of the fish from Soweto and Lenasia was also lower than fish from other heavily polluted or affected aquatic systems, such as the Olifants River system or Rietvlei Dam. The condition factor indicated that the fish were in a fair condition, thus indicating that the health effects observed were not due to malnutrition. The significantly higher HSI at Orlando showed that fish from this site have definitely been exposure to xenobiotics. The livers in the fish from this study were larger than other studies C. gariepinus. The SSI did not vary significantly between sites, but the slight enlargement of the spleen was not attributed to parasites, as none were found during the necropsy. The GSI showed that many of the fish were preparing for spawning; however there were individuals at Orlando that were classified as first time spawners, suggesting that these adult fish had endocrine stressors acting on them.
Although it is difficult to pinpoint the specific causative agents of the stressors it is clear from the results presented in this chapter that the fish in the Soweto and Lenasia at the sites where we sampled were under stress and in poorer health when compared to the control fish. Of all these sites C. gariepinus from the Orlando site proved to be the worst affected, followed by those from Nancefield and Fleurhof.

Introduction
It has been confirmed by numerous studies that polycyclic aromatic hydrocarbons (PAHs) have adverse effects on both wildlife Black & Baumann, 1991; and humans (Obana et al., 1981;Farmer et al., 2003;Binková & Šrám, 2004;Kim et al., 2013). For this reason the USEPA singled out 16 PAHs (USEPA, 2008a) to be monitored in the environment (Srogi, 2007;Pies et al., 2007;Pieters et al., 2015;Sun et al., 2015), foodstuffs (Falcó et al., 2003;Perelló et al., 2009;Rose et al., 2009), and in the human workplace (Zhao et al., 1990;Petry et al., 1996;Väänänen et al., 2005;McClean et al., 2007). When these compounds are present in the environment they pose a threat to human health. In order to determine if the PAHs in a given environment may pose a threat, a human health risk assessment is completed. The goal of a human health risk assessment is to estimate the likelihood of adverse effects in a population at defined exposures to (a) target compound/s.
Calculating the threat based on the concentrations in sediments, water and fish tissue, can provide the ability to estimate the probability of health effects in a specific area. The risk from the exposure to PAHs is predicted by evaluating the toxicity of these chemical contaminants over different exposure times, at different concentrations and exposure routes-which are measured and identified at each site. The World Health Organisation (WHO) suggests four steps for the risk assessment process: 1) hazard identification, 2) hazard characterisation, 3) exposure assessment, and 4) risk characterisation (WHO, 2010).
Hazard identification is the first step in a health risk assessment. It involves the identification of the underlying toxicological properties of chemical, i.e. its ability to cause various health effects to humans (enHealth, 2002;EA, 2009). Chemical characteristics like physicochemical properties of the chemical, metabolic properties, routes of exposure, toxicological effects, chronic-and acute animal exposure studies are used in these identifications (USEPA, 2000).
Hazard characterisation mainly consists of the description of inherent properties of the compounds and its potential to have adverse health effects (WHO, 2010). This involves the understanding of how the mode of action or the mechanism of toxicity of the compound works.

Exposure assessment
The exposure assessment parameters for Soweto and Lenasia are shown in Table 5.2. The application of the human health risk in this study is used from an ecotoxicological perspective, thus the exposure parameters was selected from literature-chosen to represent the population as best possible.

Risk characterisation
During this step of the assessment the risk is calculated for different matrices through different exposure pathways. Both carcinogenic risk (CR) (for the CPAHs) and hazardous risk (for the nCPAHs) were calculated. The risk of PAHs in Soweto and Lenasia was characterised for sediments, water and fish. The risk characterisation variables used in the various calculations are summarised in Table 5.3.

Calculating non-carcinogenic and carcinogenic risk
Hazardous risk was described using the hazard index.

HQ = Dose RfD
where RfD is the reference dose of non-carcinogenic effects  Hazard index (HI): HI = ∑ HQ i n i=n  Cancer risk (CR) for each PAH was calculated by the following formula: Multistage risk model: where CSF is the cancer slope factor  In the case where the calculated risk is greater than 0.01, the one-hit risk model is applied: CR = 1-e (Dose ×CSF)  Total cancer risk: A hazard index value calculated smaller than 0.1 indicates no hazard. A HI value between 0.1 and 1 shows low hazard risk, whereas if a HI is between 1.1 and 10 there is a moderate hazard risk. Finally, if the HI is greater than 10 there is a risk for high hazardous effects (Lemly, 1996).
A cancer risk calculated for ingestion, which is less than 1 x 10 -4 or 1 in 10 000 is considered acceptable risk. Similarly, a risk less than 1 x 10 -6 or 1 in 1 000 000 is consider negligible for dermal exposures. Once these "acceptable risks" are exceeded the population may be more prone to manifest cancers from the exposure to PAHs The hazard index and cancer risk calculations were used to assess the chance of carcinogenic and non-carcinogenic effects to manifest due to exposure to the PAHs quantified from the specific study sites in Soweto and Lenasia. Assessment of the different matrices and exposure routes only differed in the calculation of the dosages (as daily average dose, DAD). Each dose calculation is shown below and uses variables and parameters set out in Tables 5.2 and 5.3.

Risk characterisation from exposure to PAHs in water
Potential risk from PAHs from water was characterised for dermal exposure and oral ingestion. The fact that for this project water was not analysed, a cross media transfer equation was used to extrapolate concentrations. Accepting there is an equilibrium condition between the sediments and water, the concentrations of the PAHs were calculated (Galassi & Migliavacca, 1986;Provini et al., 1989): where Csediment is the concentration of PAHs in the sediment and Kp is the absorption coefficient onto soil (in terms of the organic fraction of the sediment) where Koc is the organic carbon-water partition coefficient and OC the fraction of measured organic carbon per site.
The setting for dermal exposure includes activities such as swimming, playing, and bathing in the water as well as water collection. The parameters were set at one event a day-35 minutes per event-for 350 days of the year. The total skin surface area is the sum of the feet, legs, trunk, arms and hands (USEPA, 2004). Where the DAevent is the absorbed dose per event (mg/cm 3 /event), which is calculated by: DA event = K P × C water × T event (For explanation of abbreviations refer to Tables 5.2 and 5.3) (USEPA, 2004) Kp in this case is the dermal permeability coefficient which is chemical specific, and is derived using the water-octanol partition coefficient (Kow) (USEPA, 2004).
The ingestion of water follows the set parameters in Table 5.3. The daily intake of water ingested is accepted as the water that originated from the contaminated sites.

Risk characterisation from exposure to PAHs in the sediment
A population in the vicinity of rivers and dams often comes into regular contact with contaminated sediments while walking along the river bank, swimming, bathing, or water collection for domestic use. The skin surface area exposed used for the calculations was for the hands, legs and feet. The remaining parameters are set out in Table 5.3 Dose dermal sediment = C sediment × SE × SL × AF skin × EV × D 1 ×D 2 × D 3 BM × LE (For explanation of abbreviations refer to Tables 5.2 and 5.3) (USEPA, 2004) The risk from sediment ingestion was calculated using the soil ingestion formula, the amount of sediments ingested was set at a conservative and worst case amount (Table 5.3)

Risk characterisation from consuming PAH contaminated fish
The ingestion of fish follows the set parameters in Table 5.3. The daily intake of contaminated fish is accepted as the fish caught from the study sites.
The dosage for calculating risk from consuming fish contaminated with PAHs was derived using the formula below:   et al. (2011) were used to calculate cancer risks following the same assessment parameters as for this study. Additionally, their sediment concentrations were also extrapolated to determine the prospective concentrations in the water for which CRs were also calculated. These values were used to compare relative risk between the studies.

Non-carcinogenic risk characterisation (Hazard Index)
Non-carcinogenic, or hazard risk was calculated for Soweto and Lenasia using the hazard index (HI).
The HI was calculated for the ingestion of water (adults and children) ( The hazard risks from ingesting PAH contaminated water calculated for adults from the study area of 2013, had all low hazard scores, except for Protea Glen and Nancefield. The greatest HI was at Fleurhof, followed by Eldorado Park (Figure 5.1A). The fact that Moroka did not have the greatest HI-thus far the most responsive site (see Chapters 2 & 3) may be attributed to its overall lower concentrations of nCPAHs (Table 5.4). The reference doses of acenaphthene and fluorene calculates the majority of the hazard index score-the RfD is the denominator in calculating the HQ, and when the denominator is low, the calculated quotient will be high. Thus, the low concentrations of fluorene  The hazard risks calculated for water ingestion was dominated by the 2-and 3-ring PAHs, naphthalene, acenaphthene and fluorene. These congeners have been found to have different health effects including reproductive and developmental as well as aplastic anaemia. The Agency for Toxic Substances and Disease Registry (ATSDR) reviewed the health effects of PAHs and according to their findings naphthalene decreases metabolism in rat livers, apart from being cytotoxic.
Acenaphthene increases liver cholesterol levels and fluorene decreases red blood cells and haemoglobin, and increases hemosiderin deposits in the spleen. All three these PAHs also increased the liver mass of exposed rats (Mumtez et al., 1996).

Cancer risk from water exposure
The characterisation of cancer risk was implemented for exposure through water, sediment and fish consumption. The exposure pathways selected were dermal and ingestion. Oral cancer slope factors were available for all the CPAHs as well as anthracene, fluoranthene and benzo(g,h,i)perylene, whereas only CPAHs had dermal slope factors (Table 5.1). The results for these assessments are reported in . A total cancer risk (sum of all CRs calculated) was also calculated and is reported in Figure 5.6.
The PAH concentrations in the water that had been extrapolated from the sediment levels, (Table 5. 4) were used for the water dermal and ingestion risk assessment. The dermal cancer risk from water exposure is reported in Figure 5.2. All the 2013 sites had dermal CR values higher than the acceptable risk level (1 in 1 000 000). The greatest risk was at Moroka (16 in 1 000 000), then Fleurhof (13 in 1 000 000) and Orlando West (11 in 1 000 000) ( Figure 5.2A). The children CR determined from the 2013 data were similar to that of the adult assessment. These sites-Moroka, Fleurhof and Orlando West-were the only sites exceeding the acceptable risk level (1.6, 1.3, and 1.1 in a 1 000 000 CR, respectively) ( Figure 5.2B). A temporal decrease in the adult CR was noted for Fleurhof, Moroka and Orlando West, but all the other sites had increases. Moroka still had the greatest risk (13 in 1 000 000), followed by Eldorado Park (9.7 in 1 000 000) ( Figure 5.2C).
The CR calculated for the children (2014)   study were lower for both adults and children compared to our study. The highest CR calculated using the  data was a risk of 5 in 100 000.
The CPAHs responsible for Moroka's CR barely exceeding the acceptable risk level was fluoranthene, anthracene, benzo(a)anthracene and chrysene.

Cancer risk posed by exposures to sediments
The cancer risks associated with dermal exposure to the sediments of the study sites are reported in Cancer risks from dermal exposure to the sediment sampled by  were higher at two sites, S/L 8 and 7 (3 250 and 1 290 in 1 000 000, respectively). These sites are in the same area as our Nancefield site, indicating a great decrease in cancer risk in this area between the two studies.
The PAHs that drove the dermal exposure assessment was dibenz(a,h)anthrance and to a lesser extent benzo(a)pyrene.

Cancer risk from consuming fish from a PAH polluted system
Since fish metabolise PAHs, the level of the PAH metabolites quantified represent the fish's environmental exposure to these pollutants. The metabolite data was used as the value for the worst case scenario when consuming the fish from the area. The only biliary PAH that was quantified in this study that has a cancer slope factor is benzo(a)pyrene BaP. Using the BaP concentrations ( Figure   2.2) a cancer risk for adults and children was calculated for the 2013 survey. The mean CRs for Soweto and Lenasia were well below the 1 in 10 000 risk level. The mean CR for adults was 2±0.4 in 100 000 000. It is clear from the results that the fish does not contribute the most risk towards human health (because fish metabolise PAHs). PAHs, however, pose a greater threat to humans through other exposure pathways, as shown in the results earlier in this chapter.

Total cancer risk from the PAH pollution in Soweto and Lenasia
The total cancer risk posed to the population of Soweto and Lenasia is shown in Figure 5.6. This risk level was calculated as the sum of the CRs for the various pathways. Unacceptable risk was set at the lower level-1 in 10 000-as a sensitive threshold. As seen in the previous results, temporal variations at the sites were all the same for each assessment, because they were calculated using the same chemical data set. In Figure 5.6 these trends are clearly visible. These seasonal changes, however do not affect the total CR for the adult population in Soweto and Lenasia, exposed to the water and sediments of our study area ( Figure 5.6A). Once more, the site that posed the greatest risk was Moroka for both survey years The PAHs that were the main drivers of the carcinogenic risk associated within the Klip River system of Soweto and Lenasia were benzo(a)anthrancene, benzo(a)pyrene, dibenz(a,h)anthrancene, and to a lesser extent fluoranthene, anthracene and chrysene. In laboratory studies these congeners have been found to have various carcinogenic effects at different doses and exposure pathways (Mumtaz et al., 1996;IRIS, 2016 From all the exposure routes investigated in this assessment, dermal exposure to the sediments posed the greatest cancer risks of which the highest CR was 1 100 in 1 000 000 at Moroka (2013) ( Figure 5.4A). The dermal exposure cancer risks calculated for our study were similar to those quantified by , also for PAH exposure. However, the CRs of two sites from the Roos et al. (2011) study, which were very close to each other, were considerably greater than our study's results. Overall, the CR determined for the current study is slightly lower than the  study. In contrast to this, the cancer risks due to dermal exposure to water showed an increase between the two studies, and the risk from ingestion of the PAH contaminated water (extrapolated) was the greatest for our study.
From all the different assessments Moroka has shown to be the site with the greatest risk, followed by Eldorado Park and Orlando East ). These results are supported by the quantified PAH concentrations in the environment (Table 2.8 & Table 3.1).

Conclusion
The PAHs quantified in the Soweto and Lenasia study area pose unacceptable risks to the population that are exposed to the Klip River system. The human health risks were modelled using instrumentally determined levels in the sediment and extrapolated concentrations in water, along with parameters and exposure routes applicable to the area and human population. It is important to note that the human health risk modelled here is dependent on the various exposure factors, thus a site with high concentrations of certain PAHs may not necessarily have high risks. This shows the importance of completing these risk assessments, to identify risk in areas where pollutant concentrations alone would have not indicated risk. In the case of this study, the sites with the greatest concentrations had the greatest risks. The hazardous risk expressed by the hazard index of ingesting PAH-contaminated water posed the greatest non-carcinogenic risk to both the adult-and child population in the study area. Dermal exposures were the most potent pathways in the cancerous risk assessment, where the exposure to water and fish were negligible. A small decrease in risk was noted between the two surveying years of 2013 and 2014. This study also had relatively lower cancer risk compared to the risk calculated from 's PAH data-data obtained in the same study area but from an earlier period. The CR to children was lower than the adults'. The site with the greatest CR for both years was Moroka, exceeding the acceptable risk level for each of the assessments. The most notable CRs of this site were for dermal exposure to sediments (1 100 in 1 000 000) and total cancer risk (277 in 10 000). The overall risk assessment of the Soweto and Lenasia area has shown that there are both carcinogenic and non-carcinogenic risks to the human population exposed to the Klip River flowing through the area.

5.5
Chapter 6: Statistical integration of results, the study conclusion, and recommendations

Introduction
In this chapter the findings discussed in the previous chapters will be compared to each other to provide an overview of the PAHs in the aquatic ecosystem of the Klip River system in Soweto and Lenasia. The comparative integration of results for the discussion presented here was achieved using descriptive and multivariate statistics. Following the general comparative discussion are the thesis' conclusion and final recommendations for future studies.

Multivariate statistics
The multivariate statistical techniques used in this chapter were principle component analysis (PCA) and redundancy analysis (RDA). Principle component analysis is based on a linear response model that explains the variation between species data and environmental variables (Scott & Clarke, 2000;Van den Brink et al., 2003). In doing so new combinations of variables that explain the greatest variation in the data set are created (principal components, which are linear combinations of the original variables) (Fowler et al., 1998). The first principal component (Factor 1) explains the greatest amount of information in the data set. The following principal component (Factor 2) is as different as possible from the first and explains the second largest portion of the data (Fowler et al., 1998). PCAs are used in ecotoxicology to determine differences and similarities between environmental datasets looking for underlying variables that might explain the patterns observed in the PCA (Quinn et al., 2009;Malherbe et al., 2015;Gerber et al., 2015).
Redundancy analysis is a multivariate analysis technique that uses two sets of variables-an explanatory and dependant variable set (Israëls, 1992). Although an RDA is based on similar principles as a PCA, an RDA allows the driving or explanatory variables to be selected, which then allow focus on the part of the variance that is explained by these selected external explanatory variables ( Van den Brink et al., 2003). The RDA chooses principle components (factors) for only the explanatory variables selected for maximal association with the dependant variables (Scott & Clarke, 2000). RDAs are used to determine similarities or differences between datasets based on specific variables.
The biplots (PCA) and triplots (RDA) were interpreted as described by Šmilauer & Lepš (2014). The angle between the vector arrows indicate the correlation between the individual environmental variables and/or species data arrows: an angle close to 0° indicates a positive correlation between the variables; and angle closer to 90° shows no correlation; and an angle approaching 180° indicates negative correlation between variables. The perpendicular line between a sample symbol and a particular species arrow can be used to estimate the value of that sample in terms of the variable it is perpendicular to.

PAHs in the sediment
The concentrations of the PAHs in the sediments of the study area were discussed in chapter 2. The greatest ΣPAHs for both years was at Moroka, followed by Protea Glen in 2013 and Eldorado Park in 2014 (Table 2.8). These levels were comparable to levels reported in literature (Table 2.9). According to a scale set by Baumard et al. (1998) the degree of PAH contamination in the sediment of Soweto and Lenasia ranged between 'moderate' and 'high' with the exception of Moroka 2014 that was 'very high' (Table 6.1).
In 2013 the sites with a 'high' rating were those situated in the centre of the study area (Figure 6.1A) and a decrease in concentrations was seen downstream from these sites: Protea Glen to Lenasia; Orlando West, Orlando East, and Moroka to Eldorado Park ( Figure 6.1A) Table 6.1: Levels of total polycyclic aromatic hydrocarbons in the sediments (ng/g) of Soweto and Lenasia against the scale by Baumard et al. (1998) In 2014 the situation between Protea Glen and Lenasia was reversed from that in 2013 with an increase in concentration downstream from Protea Glen (Figure 6.1B). The 2014 relational distribution of PAHs between Moroka and Eldorado Park was the same as in 2013, with a downstream decrease relative to Moroka, despite these two sites' overall higher levels in 2014 than in 2013 (Figure 6.1B).
The levels at Nancefield were always lower than the two upstream sites, Lenasia and Eldorado Park, for both sampling years after the confluence of the Klip River and Klip Spruit. The most probable reason for this large decrease in PAH concentrations may be due to the filtering capability of the large wetland system (Liu et al., 2008;Wang et al., 2012) that stretches from both Lenasia and Eldorado Park to Nancefield (Figure 1.2 & 6.1) of which the impact was greater in 2014 than in 2013 (Table   6.1).
Other spatial and temporal variations of the PAHs in the sediments of Soweto and Lenasia were investigated using a PCA. In this analysis, the concentrations of the individual congeners, LPAHs, HPAHS, CPAHs and ΣPAHs were included (from Table 2.8). Values of samples that were below the limit of detection were replaced with half LOD values to avoid using zero values. The data set was log transformed during the analysis.
Factors 1 and 2 describe 92% of the variation in the data (Figure 6.2). Factor 1 (74.46%) distinguishes between sites with 'high' and 'very high' PAH contamination (see Table 6.1) on the positive side of the factor and those with 'moderate' PAH levels on its negative side Factor 2 (17.5%) contrasts the sites with the higher HPAHs (BkF and BbF) on its positive side to those with higher LPAHs (Acea and Nap) on its negative side ( Figure 6.2). The co-linearity between CPAHs and HPAHs can be explained by the fact that the carcinogenic PAHs form part of the heavier congeners.

Relationship between biliary PAHs in Clarias gariepinus and the PAHs in the sediments
The deconjugated PAH metabolites were quantified in Clarias gariepinus from three sites of 2013 and the control (Table 2.10). A redundancy analysis was completed to establish whether the PAH metabolites and the native compounds found in the sediments were associated with one another (Figure 6.3). Of all of the hydroxyl PAHs that were analysed, only those that had quantifiable levels in most of the fish were included in the RDA as explanatory variables. These were: 2-OH Nap, 2-,3-OH Flu, 9-OH Flu (as ΣOH Flu), 2-OH Phe, 4-OH Phe (as ΣOH Phe), 1-OH Pyr, and 9-OH BaP. Half LOQ values were assigned to those OH-PAHs below the LOQ where applicable. Those metabolites that were all below limit of quantification for all samples were not included in the analysis to prevent skewed results. The corresponding native PAHs in the sediment (Nap, Flu, Phe, Pyr and BaP) were used as the species variables. The control fish were not included in this analysis as there were no corresponding sediment values for them.
The explanatory variables account for 57.47% of the variation in (Figure 6.3). The fish from the three sites were grouped separate from each other on the graph. This pattern shows that the OH-PAHs present in the individual fish corresponds sufficiently enough for them to ordinate according to their sites. Factor 1 (43.74%) distinguished between the fish with OH-PAHs (positive side) and those with little or no OH-PAHs (negative side). Factor 2 (13.73%) separated fish with more 'lighter' OH-PAHs (positive side) from those with the 'heavier' metabolites (negative side) (Figure 6.3). The angles the vector lines of the metabolites and the native PAHs are all less than 90° (Figure 6.3). The strong correlation between the OH-PAHs and their native forms indicates that the C. gariepinus must have been exposed to the native PAHs in the sediment. This can be largely attributed to the fact that C. gariepinus are bottom dwelling fish and are associated with sediments (Bruton, 1988). This confirms that the PAHs quantified in the sediments of Soweto and Lenasia transfer to C. gariepinus.

Sediment toxicity in terms of instrumental data and sediment indices
The potential toxicity posed by the PAHs in the sediment was determined by using various toxicological assessment indices (see Chapter 2). These included the sediment quality guideline index (SQGI), sediment quality index (SQI) and toxic equivalent quotients (TEQs) (see section 2.2.4).
Both the SQI and SQG-I use existing guidelines to express different toxicological endpoints. The SQI indicates the quality of the sediment as calculated by magnitude of exceedance of each guideline per congener. The SQG-I calculates probable toxicity to benthic organisms based on the lower and upper levels of the guidelines. Thus, one would expect that these indices would report similar results. The Canadian SQGs were used for comparisons because they are the more protective guideline set (Table 2.  Another method where toxicity is predicted using instrumentally analysed concentrations is by means of toxic equivalent quotients. As described in section 2.2.4 (Toxic equivalent quotient calculation) this method incorporates a toxic equivalent factor that helps rank the toxicity of a sample in terms of 2,3,7,8-TCDD. Correlations were drawn between the SQG-I and the TEQs to compare the toxicity of the PAHs in the sediment to different organisms, and whether these showed similar results. There was a strong correlation (r = 0.8, p<0.0001) between the SQG-I and the TEQs based on the fish potency factors (TEQFPF) (Figure 6.5 A). The TEQs based on mammalian TEFs (TEQTCDD) correlated less than those for fish (moderate correlation, r = 0.56, p = 0.0147) ( Figure 6.5 B). The TEQTCDD and the TEQFPF showed strong correlation (r = 0.81, p<0.0001) ( Figure 6.5 C).

Figure 6.5: Spearman correlation scatterplot for: A) the sediment quality guideline index (SQGI) vs toxic equivalence using the fish potency factors (TEQFPF); B) the SQG-I vs toxic equivalence using toxic equivalent factors derived by Villeneuve et al. (2002)(TEQTCDD); C) the TEQTCDD vs TEQFPF
The correlations seen above show that the overall prediction of toxicity using different organismal endpoints (benthos, fish and mammalian) are constant. It was expected that the predictions of benthos toxicity (which is based on overall toxicity; see section 2.1.6) and the toxicity using the mammalian system (specifically AhR-mediated toxicity; see section 2.1.6) would not correlate strongly, as they are derived from separate toxicity endpoints. Analogously, both TEQ data sets were derived from the same mode of action and corresponded highly. The fish potency factors used to calculate the TEQFPF were derived by using the same mode of action-AhR mediated toxicity-but specifically for fish (see section 2.1.6, Toxic equivalent quotient calculation) and both (TEQTCDD and TEQFPF) are expressed in terms of TCDD.

Sediment toxicity in terms of biological responses
The toxicity of the PAHs in the sediment was also tested using the H4IIE-luc reporter gene bioassay (see Chapter 3). This method uses the Ah-receptor to express PAH toxicity of a sample (see section 3.1.3). The results obtained are therefore biological responses to the PAH containing fraction of the sediment extract and is expressed in terms of TCDD equivalency (see section 3.2.5). The maximum elicited responses and BEQs for both years were reported in Table 3.1. The viability assay showed that the low or below LOD responses in the H4IIE-luc bio-assay were due to cell death and not the absence of AhR agonists (Table 3.1). This cytotoxicity was attributed to the LPAHs in the samplesthose PAHs that do not bind to the AhR-as shown by Schirmer et al. (1998). Only the ΣLPAHs and acenaphthylene (Acey) of 2013 correlated significantly with the MTT viability test. The ΣLPAHs showed moderate negative correlation with the cell viability (r = -0.61, p = 0.0429), and Acey showed strong negative correlation (r = -0.7, p = 0.0216). The cytotoxicity of the LPAHs seems to be linked to lower concentrations. The correlating levels of LPAHs and Acey were lower in 2013 than 2014.
The AhR-mediated toxicity of the sediments from Soweto and Lenasia was from the CPAHs in the extracts, as shown by the strong correlation (r = 0.77, p = 0.0002) (Figure 3.5). Comparison between the BEQs and the TEQs (calculated with TEF values derived from the H4IIE-luc cell line ) illustrated that the AhR-mediated toxicity quantified by both the instrumental and biological analysis were similar. These results correlated strongly (r = 0.78, p = 0.0001) (Figure 6.6A). Similarly, the BEQs correlated strongly with the TEQFPF (r = 0.77, p = 0.0002) (Figure 6.6B), signify that the fish potency factors (Barron et al., 2004) are equally sensitive to predict AhR mediated toxicity of PAHs using instrumental data. Finally, the prediction of sediment toxicity to benthic organism (SQG-I) was compared to the BEQs ( Figure 6.7). The result was that the BEQ and the SQG-I strongly correlated (r = 0.74, p = 0.0005), proving that the toxicity predicted for benthic organisms and the toxicity quantified by the H4IIE-luc bio-assay posed similar threats, each in terms of their own modes of toxicity.

Biomarker and bio-indicator responses to polycyclic aromatic hydrocarbons 6.5.1 Effects of PAHs on the biomarkers and fish health indices
As previously stated, the PAHs in the sediment associate with the biliary PAHs ( Figure 6.3), together with the fact that bile was only sampled in 2013, the PAHs in the sediments were ordinated against the biological responses in the fish for both 2013 and 2014 (Figure 6.8). The PAHs in the sediment were used as the explanatory variables in the redundancy analysis. Half LOQ values were assigned where applicable. All the biomarker responses and selected fish health indices (CF, FHAI, HSI and SSI) were used as the species variables. In order to prevent data skewing (because Canoco treats zeros as "missing values" consequently treating the data point differently from one which was measured but resulted in no values) species values which had zeros were replaced with a value equal to a tenth of the blank (biomarkers) or a tenth of the lowest sample value (FHAI). Control fish were included in this analysis, however, because they do not have corresponding sediment values, zero values were assigned as the control fish were not exposed to sediments (thus "missing values"). The inclusion of "missing values" did not affect the ordinations of explanatory and species data compared to a similar PCA but without the congener dataset (graph not shown).
Due to strong co-linearity of the environmental data (PAHs in the sediment) the forward selection option was chosen in the RDA, to reduce overestimation of the amount of explained variance.
Forward selection is a method where a parsimonious subset of explanatory variables is objectively selected. This results in a data set with fewer variables that explain the same amount of variance as the original dataset (Blanchet et al., 2008).
The explanatory variables explain 44.53% of the variation in Figure 6.8. Six of the 16 PAHs were chosen during the forward selection, these were: BaP (explains 29.3% of the 44.53%, p = 0.002), DBA (12.8%, p = 0.002), Acey (7.8%, p = 0.002), Acea (4%, p = 0.002), Ant (2%, p = 0.008) and BaA (1.5%, 0.028). Factor 1 (36.98%) distinguishes between the control fish on the positive side and the environmental fish on the negative side based on the ordination of the biological response in the fish: Decrease in AChE, but increase in the remainder biomarkers and health indices, arranging them on the opposite sides of Factor 1. The inhibition of AChE relative to the control showed in Figure 4.2A is supported in Figure 6.8. AChE strongly negatively correlated (180° angle) with DBA and Acey. Factor 2 explains very little variance (7.5%) and is based on the presence of protein carbonyls on the positive side and the absence thereof on the negative side. Orlando was the only site that ordinated with PCs and this is supported by the results in Figure 4.4B. The protein carbonyls did not correlate with the PAHs (90° angle), suggesting that other xenobiotics may be responsible for the PC formation in the study area.
This suggests that Orlando's biomarkers and health indices did not associate strongly with the PAHs.
The sites that were best explained by environmental and species data were those from 2014 ( Figure   6.8). The SSI, SOD, MDA and the condition factor did not contribute to the explained variation in the data set, as shown by the short vector lines. CAT strongly correlated with anthracene and to a lesser extent benz(a)anthracene (<0°). The strong association of the CAT (long vector) to the sites compared to low association of the SOD (short vector), showed that the SOD-CAT system is out of balance: the SOD is supressed rather than inhibited, and CAT is over expressed to compensate for the decreased SOD and to regain homeostasis. The HSI correlate mostly with BaA and Ant, while FHAI mostly with DBA. Effects on the lower levels of organisation (such as the molecular level) suggest brief contaminant influences. Higher in the levels of organisation the effects seen are caused by multiple stressors over a longer time period (Munkittrick & McCarty, 1995). Although the organ/systematic level (represented by the health indices) are also on a lower level of the biological organisation, it seems that the PAHs in the study area do not have such a strong effect on the overall fish health (cellular, organs and individuals) as they have on the molecular level (Figure 6.8).

Conclusion
In a previous study  PAHs were identified as one of the most widely spread and abundant pollutant class in South Africa. One region that specifically had a high PAH burden was Soweto and Lenasia. The findings of  were the main motivation for the current project: a further, in-depth investigation of the PAHs and their potential effects in the area.
The previous chapters of this thesis aimed to determine the levels of the 16 priority PAHs in the sediment and biota from the study area; in addition to the measured PAH levels, the pollutant profile was determined using composition percentages and source appointment ratios; and finally, the toxicity posed to fish and human health by the PAHs in the study area were assessed (see section

1.2.2).
It is evident that PAHs are ubiquitous in the study area. Levels present in the sediments, of which the dominant 3-and 4-ring congeners were mainly from biomass combustion, were comparable to international studies. Evidence of PAH presence in the biota was seen: low levels of Nap, Acea and Phe were found in the wetland bird eggs; and PAHs metabolites in fish bile.
The central region of the study area had the most PAHs (not exclusively): Moroka had the greatest ΣPAH levels; similarly, Eldorado Park and Orlando East had high levels of PAHs in the sediments; and the fish with the greatest ΣOH-PAHs were at Orlando (Orlando East).
The toxicity predicted for benthic organisms, fish and mammalian systems, based on the instrumentally derived PAH sediment concentrations (International SQGs, sediment indices, TEQs) were accurate compared to the biological responses generated by the H4IIE-luc reporter gene bioassay (BEQs).
The inhibition of AChE together with the activation of the detoxification system (CYP450), and upregulation of the SOD-CAT defences, is evidence that the fish from the study area had been resisting against xenobiotic stressors. Although the fish might have been exposed to a much broader variety of xenobiotics there was strong indication that the biomarker responses were due to PAHs: like acenaphthylene, acenaphthene, anthracene, benz(a)anthracene, benzo(a)pyrene and dibenz(a,h)anthracene (Figure 6.8). Literature supports that biomarkers respond to PAH exposures.
Even though the biomarker response were up-regulated, the fish could cope with the stress as was evident from the low levels of MDA and PC and the fact that the energy budget was not depleted.
The overall health, on a systematic level, of the Soweto and Lenasia fish was visibly affected by environmental stressors. The fish from Soweto and Lenasia proved to be in poorer health (relative to control) than other South African studies on C. gariepinus. The fish were in a fair condition, showing no malnutrition. The HSI of the fish from this study were greater than other studies investigating Clarias gariepinus. The SSI and GSI showed no significant difference from the control. The underdeveloped ovaries of three adult fish from Orlando suggested endocrine stressors. Regardless of what was causing the stress in the fish, the fish from Soweto and Lenasia were clearly under stress compared to the control fish. Of all these sites Orlando proved to be the worst affected, followed by Nancefield and Fleurhof.
The potential of PAHs to harm human health was shown with a human health risk assessment applied to different matrices and exposure pathways. Of these exposure pathways the intentional ingestion (geophagia) and dermal exposure of sediments posed the highest cancer risk in the study area. Moroka, Eldorado Park and Orlando East proved to be the sites with the greatest cancer risk. A total cancer risk of 277 in 10 000 was calculated for Moroka, which was the highest for the whole project.
From all the results obtained it can be concluded that the areas of most concern are Moroka and Orlando East, based on the fact that Moroka had the greatest ΣPAH and highest human health risk; and Orlando East because of the biomarker and fish health results that were the highest for 2013, and that here were no fish available to sample in 2014, most likely due to high pollution because of power station collapse. By completing the objectives, and meeting the aims of this project, the hypothesis that the humans and wildlife of Soweto and Lenasia, dependant on the Klip River, are exposed to the 16 priority PAHs is therefore accepted.

Recommendations
 The target compounds of this study were the 16 priority PAHs. The effective metabolism of these compounds by biota resulted in very low levels of native PAHs in the biota. The quantification of PAH metabolites has shown very effective in establishing the exposure of these chemicals in biota. Therefore, an in-depth chemical analysis of PAH-metabolites is suggested for future studies regarding biotic matrices for example mussel tissue, bird eggs and fish bile.
 The presence of other pollutants is seen in the results of this study and that of the precursor study by . Thus for future studies a broad spectrum screening for a much larger variety of organic and inorganic chemical contaminants/pollutants is encouraged (if finances permit). Organic compounds that can be considered include: polychlorinated biphenyls, brominated flame retardants, organochlorine and -phosphate pesticides, plasticisers, bisphenol-A/B, pharmaceuticals and personal care products and perfluorinated compounds. Inorganic compounds that may be involved are metals such as mercury and chromium.
 The biochemical responses were sufficient to show that the fish in the study area was under stress. It is recommended that a broader array of biomarkers be used in the future to further investigate the effects of the environmental stressors on the target biota. These may include: more specific cytochrome activity biomarker assays such as ethoxyresorufin-O-deethylase (EROD) and aryl-hydrocarbon hydrolase (AHH); phase II biotransformation biomarkers such as glutathione S-transferase (GST); additional oxidative stress biomarkers like reduced glutathione (GSH) and glutathione peroxidase (GPOX); metallothioneins for metal exposure; vitellogenin content assays for endocrine disruption in male fish; DNA adducts and the comet assay for DNA damage.
 On a systematic and organism level, the FHAI and OSIs indicated deviation in fish health in the study area. A histopathological assessment is recommended to link the molecular level changes with the higher level changes determined by the health indices, and determine changes on a cellular level.
 It is also suggested that macro-invertebrates, such as molluscs and aquatic insects be included in future studies. The inclusion of invertebrates would fill the gap of biotic sampling where fish cannot be sampled. The evaluation of invertebrate and fish species composition and numbers can further describe pollution effects in the system and would further the ecological assessment of the area.
 A wide variety of bio-assays that are able to detect a variety of mechanisms of actions through which biota can be harmed can be employed in future research. The effectiveness of these bio-assays has been proven in this study. The array of bio-assays can be broadened to include assays capable of detecting various endocrine disruptive effects, as well as genotoxicity.
 A more frequent sampling regime for abiotic matrices (include water sampling to sediments) is proposed. This allows for a higher frequency of testing for bio-assay responses and possibly screening for contaminants.
 The last aim of this study, the theoretical human health risk assessment indicated to real carcinogenic risks. Due to this result, if would be advisable to investigate exactly how much fish from the aquatic system is consumed by the local inhabitants, as well as the rate of water and sediment exposure. This can be addressed by implementing questionnaires and interviewing the local people, as well as sampling hair and blood for chemical analysis.