Occurrence and levels of persistent organic pollutants (POPs) in wild and farmed tilapia (Oreochromis niloticus) from Lake Kariba, Zambia: possible impact on sh health

The current study was carried out to determine occurrence and levels of persistent organic 18 pollutants (POPs) in wild and farmed tilapia (Oreochromis niloticus) in Lake Kariba, Zambia, 19 and possible implications for fish health. Concentrations of organochlorine pesticides (OCPs), 20 polychlorinated biphenyls (PCBs), polybrominated diphenyls (PBDEs), 21 hexabromocyclododecane (HBCDD) and perfluoroalkyl substances (PFASs) were determined in liver samples of tilapia. Concentrations of POPs in wild tilapia were in general higher than in farmed tilapia, however, concentrations of DDTs and PCBs in wild tilapia foraging near the 24 fish farms were more in the range of the farmed fish. The highest median ∑ DDTs (93 and 81 25 ng/g lw) were found in wild tilapia from sites 1 and 2, respectively 165 km and 100 km west 26 of the farms. DDE/DDT ratios seem to indicate recent exposure to DDT. The highest median 27 of ∑ 17 PCBs (3.2 ng/g lw) and ∑ 10 PBDEs (8.1 ng/g lw) were found in wild tilapia from site 1 28 and 2, respectively. The dominating PCB congeners were PCB-118, -138, -153 and -180 and 29 for PBDEs, BDE-47, -154, and -209. Of PFASs, only PFOS, PFDA and PFNA were detected 30 in wild fish with highest median PFOS levels in site 1 (0.66 ng/g ww). The PCB and BDE 31 concentrations in wild and farmed fish were above EQS biota limits set by the EU. This may 32 suggest a risk to the fish species and threaten biodiversity.

in liver samples of tilapia. Concentrations of POPs in wild tilapia were in general higher than 23 in farmed tilapia, however, concentrations of DDTs and PCBs in wild tilapia foraging near the 24 fish farms were more in the range of the farmed fish. The highest median ∑DDTs (93 and 81 25 ng/g lw) were found in wild tilapia from sites 1 and 2, respectively 165 km and 100 km west 26 of the farms. DDE/DDT ratios seem to indicate recent exposure to DDT. The highest median 27 of ∑17PCBs (3.2 ng/g lw) and ∑10PBDEs (8.1 ng/g lw) were found in wild tilapia from site 1 28 and 2, respectively. The dominating PCB congeners were PCB-118, -138, -153 and -180 and 29 for PBDEs, BDE-47, -154, and -209. Of PFASs, only PFOS, PFDA and PFNA were detected 30 in wild fish with highest median PFOS levels in site 1 (0.66 ng/g ww). The PCB and BDE 31 concentrations in wild and farmed fish were above EQSbiota limits set by the EU. This may 32 suggest a risk to the fish species and threaten biodiversity. 33 34

Introduction 35
Fish serves as a major source of proteins to most people in the world and is essential for food 36 security and sustainability (FAO 2020). The growing human population has led to increased 37 demand for fish, leading to overexploitation of wild fisheries and depletion of some fish stocks. The growing aquaculture industry in Africa may be threatened by the presence of POPs and 69 other contaminants in the water and fish. Therefore, to enable sustainable aquaculture 70 development, it is of key importance to gain knowledge on toxicological risk factors and the 71 potential adverse effects of pollutants and other environmental factors on fish health. 72 Contaminant residues in fish may also represent a food safety risk. Environmental stressors 73 including harmful chemical contaminants and biotoxins and other water quality parameters 74 such as pH, oxygenation and eutrophication, have impact on health in wild stocks and farmed 75 fish. Lack of knowledge about levels, sources, environmental behaviour and toxicity, hampers 76 evidence-based decision-making regarding implementation of protective measures. The 77 current study was carried out to establish the occurrence and concentrations of POPs in wild 78 and farmed tilapia from Lake Kariba, Zambia, with emphasis on fish health. 79

Materials and methods 80
Description of sampling area and species 81 Nile tilapia (Oreochromis niloticus) were collected from Lake Kariba, located on the southern 82 border of Zambia with Zimbabwe (16° 28′ to 18° 04′S; 26° 40′ to 29° 03′ E). Lake Kariba is a 83

Sample collection 114
A total of 142 wild and farmed tilapia samples were collected from June to July 2017. 115 Physicochemical parameters (pH, temperature, conductivity, and total dissolved solutes) were 116 measured in all sites (not shown). Live wild tilapia were bought from fishermen as they pulled 117 in their catch from the water. The fish were then placed in a container containing ice water and 118 transported to the shore for dissection. Farmed tilapia were sampled by dip netting and placed 119 in containers containing water. The length and sex of the fish were recorded (Table 1A, 1B). 120 The size of the fish was sometimes considerably different within the study sites. The scale used 121 in the field could only weigh fish up to 1kg. Therefore, only length was used in statistical 122 analyses. Using forceps and scalpel blades, the fish was dissected on a board and liver tissue 123  (Table 1B). Compounds marked with * were 153 included in ∑3PFAS. Other PFAS components were not detected in levels >LOD and were not 154 used in any data analyses further. 155

Chemical analyses of OCPs, PCB, BFRs 156
The analytical method for analysing of OCs was first described by Brevik (1978)  For PFASs, every analytical series included three blanks of solvent, two samples of spiked 217 Atlantic cod (Gadus morhua) for recoveries and one blind sample of non-spiked Atlantic cod. 218 LOD was calculated as 3 times the noise in the chromatogram. The LOD for PFAS ranged 219 between 0.093 ng/g ww to 0.706 ng/g ww. Matrix-matched calibration curves ranged from 0 -220 50 ng/ml and were Linear with R 2 >0,99, except for PFTrDA. The analytical quality of the 221 method was assessed by including an inter-laboratory test (AMAP) in the analysis of samples. 222

Statistical data analysis 223
Detection rate was defined as percentage of samples with a detectable value, i.e., above LOD. 224 The compounds with detection rate above 50% were reported with descriptive statistics and 225 further included in the statistical analyses. Levels below LOD were replaced with 1/2 LOD. 226 Compounds with a detection rate lower than 50% were reported with range and were not 227 included in further statistical analyses. For this latter group, the levels below LOD were 228 replaced with a value of 0.0001 when calculating the sum of the compound group for all results 229 presented, Stata SE/16 (Stata Corp., College Station, TX, USA) was used for statistical 230 analysis. Normality of the data was tested using Shapiro-Wilk. If data from one of the locations 231 failed the Shapiro-Wilk test, the data of all locations were log-transformed. The nonparametric 232 Kruskal-Wallis test were used as the present data failed Shapiro-Wilk after being log-233 transformed. Dunn`s post-hoc test was applied for pairwise comparisons between the locations, 234 with and without Bonferroni corrections for multiple comparison. Spearman rank correlation 235 was used to assess the correlation between variables. The statistical significance level was set 236 at p<0.05. 237

Results 238
Fish characteristics 239 Fish weight and length correlated strongly for fish below 1 kg for both wild fish (r=0.93) and 240 farmed fish (r=0.92). Since fish weight above 1 kg was not specified, fish length was used as 241 indicator of fish size. Fish from farm 1 were significantly longer (mean 36 cm), (p<0.05) than 242 the other locations (Table 1A, Fig. 2). The median liver lipid contents (%) of farmed fish 243 from farm 1 (8.8 %) and 2 (7.5 %) were significantly higher than for wild fish from site 2 (4.9 244 %), and farm 1 was significantly higher than site 1 (4.1 %). There was also a significant 245 difference in liver lipid content between the wild fish at site 2 and 3 (7.2 %). Length of the 246 individual fish for PFAS analyses were all in the same range (Table 1B). 247 Occurrence and levels of OCPs, PCBs, BFRs 248 HCB and p,p'-DDE were the OCPs detected in 100% of the liver samples. Median 249 concentrations of HCB were significantly higher in wild fish from site 3 (Table 2B) compared  250 to site 2 and in farm 1 and 2 (Fig. 2). Of the HCHs, γ-HCH (lindane) and α-HCH were detected 251 in 77% and 50 % of the samples, respectively. The highest median concentration of ∑HCHs 252 was 0.05 ng/g lw in site 3. The γ-HCH was the dominant HCH, contributing 56 %, 69 % and 253 65 % to ∑HCHs in wild fish from site 1, 2 and 3, and 79 % and 58 % in fish from farm 1 and 254 2 (Fig. 3). α-HCH contributed between 21-34 % to the ∑HCHs and β-HCH between 1-14%. 255 The median ∑HCHs was significantly higher in site 3 compared to site 1,2 and farm 2 (Fig. 2). 256 DDTs were the most abundant OCPs in all the locations with the highest median concentrations 257 of ∑DDTs in wild fish from site 1 and 2 (93 and 81 ng/g lw) (Table 2B) and 2 compared to farm 1 and 2, while site 3 was only significantly higher than farm 2 (Fig.2). 261 The contribution of p,p'-DDE to the ∑DDTs was highest in wild fish from site 1 and 2 (48 % 262 and 61 %), but lower in wild fish from site 3, and farmed fish from farm 1 and 2 (46 %, 30 % 263 and 31 %) (Fig. 3) . In farm 1 and 2, the contribution of p,p'-DDD to the ∑DDTs was higher 264 than that of p,p'-DDE with 63 % and 68 %, respectively (Fig. 3). The ratio of p,p'-DDE/p,p'-265 DDT was highest in wild fish from site 2 and farmed fish from farm 2 and lowest in wild fish 266 from site 3. Trans-nonachlor was detected in 73%, and cis-nonachlor, cis-chlordane and trans-267 chlordane in only 17%, 13 % and 7% of the samples, respectively. Trans-Nonachlor 268 contributed 66-87% to ∑CHLs. ∑CHLs were not significantly different between any of the 269 locations. Mirex and heptachlor were not detected in any of the samples. average 8 %, 14 %, 25 % and 16 % to ∑17PCBs respectively (Fig. 3). The highest median 273 concentration of ∑17PCBs was found in site 1 at 3.3 ng/g lw <site 3<site 2<farm 1<farm 2 274 (Table 2B, Fig. S1). Median ∑17PCBs was significantly higher in site 1 compared to 2 and farm 275 1 and 2. Site 3 was significantly higher compared to site 2 and farm 2 (Fig. S1). the PBDE pattern and contributed average 84 % to ∑10PBDEs (Fig. 3). The highest median 279 concentration of ∑10PBDEs (including BDE-209) was 8.1 ng/g lw in site 2. Median ∑9PBDEs 280 (excluding BDE-209) were not significantly different between the locations, however the 281 median ∑9PBDEs was significantly lower in farm 2 compared to site 1,2,3 and farm 1 (Fig. 2). 282 HBCDD was only detected in one sample from farm 1, with a concentration of 0.27 ng/g lw. 283 Occurrence and levels of PFASs 284 PFOS, PFDA and PFNA were the only PFASs detected in levels >LOD in individual wild fish 285 (Table 2B). No PFASs were detected in site 3, farm 1 or farm 2. PFOS and PFNA were detected 286 in 100 % and 40 % and in 100 % and 40 % in wild fish from site 1 and 2, respectively and 287 PFNA was detected in 20 % in site 1 and 2. The highest median level of PFOS (0.66 ng/g ww) 288 was found in wild fish from site 1 while highest median level of PFDA (0.37 ng/g ww) was 289 found in site 2 (  Fish from site 3 forages around the fish farms and shows significantly higher fat content than 316 in wild fish from site 2, confirming that wild fish in site 3 makes use of nutrient spill from the 317 fish farms (Table 2). There is also a possibility that farmed tilapia escape from the cages and 318 thus being caught as wild fish, thus influence the fat content of fish in site 3 (Azevedo-Santos 319 et al. 2011). 320

Levels and congener profile of OCPs, PCBs and BFRs 321
DDTs were the dominant OCPs in both wild and farmed fish liver tissues from Lake Kariba. 322 Wild fish had significantly (p<0.05) higher levels of median ∑DDTs compared to the farmed 323 fish (Table 2, Fig. 2). This was similar to findings in other studies (Berg et al. 1992 control operations in addition to agriculture (Berg et al. 1995). Lake Kariba was filled with 327 water in 1958-1963. Because the flooded areas were earlier treated with DDT, the sediments 328 of the Lake will still be a reservoir of DDT residues. Due to long half-life of DDT and different 329 metabolism under anaerobic conditions, DDT and its metabolites originating from the time 330 before the Lake was filled, may still contribute to exposure of living organisms in Lake Kariba 331 today (Berg et al. 1995;Brevik 1996). In addition, DDT may have entered the Lake Kariba by ∑DDTs was highest in wild fish from site 1 and 2 but decreased eastwards in wild fish from 339 site 3 (Fig. 3). In the farmed fish, p,p'-DDD was contributing most to ∑DDTs (Table 2 (Table 5). Farmed tilapia in the present study had similar levels of mean ∑DDTs to 352 those reported from Lake Kariba, Zimbabwean side, by Berg et al. (1992) (Table 4). 353 The second dominant OCP, HCB, was detected in low levels below the EQS set by the EU 354 (Table 4). Median levels of HCB were significantly higher in wild fish from site 3 compared 355 to site 2 and farm 1 and 2 (Table 2B, Fig. 2). Previously, Berg et al (1992) Table 5). 364 Although levels of HCHs generally were low in all study sites, the median ∑HCHs was 365 significantly higher in wild fish from site 3 than in site 1 and 2 and in farm 2 (p<0.05) (Fig. 2). 366 Median ∑HCHs in farmed fish from farm 1 was not significantly different from site 3, 367 suggesting that HCH in site 3 and farm 1 have a common source. Lindane Heptachlor are also banned substances and were below detection limit in all samples (Table S  383  Victoria and one from Lake Babati in Tanzania. It seems thus, that the environment in Lake 403 Kariba is exposed to a different historic PCB mixture than in other studies in the region. The 404 finding of significantly higher ∑17PCBs in fish from site 1 and 3 (0.92 ng/g lw) than in the 405 nearby farms 1 (0.27 ng/g lw) and 2 (0.19 ng/g lw), may be related to higher age of the wild 406 fish, rather than higher exposure to unknown sources. In the present study, levels of PBDEs exceeded the European standard (EQS) for these 463 contaminants in fish and may harm fish health (Table 4). Follow up studies are needed to ensure 464 that international regulations result in decrease of these and other contaminants that threaten 465 the aquatic environment. Dioxin-like (DL) PCB-118 TEQ levels (pg/g ww) were below EQS 466 for DL-PCPs (Table 4). Liver tissues were used in this study because lipophilic POPs would 467 show highest levels in the lipid rich liver. Results from this study can, therefore, not directly 468 be used in a risk assessment for humans since humans consume fish muscle. However, the 469 percentage of lipid in tilapia liver (range 3-19 %) (present study) is higher than in its muscle, 470

Conclusion 480
The present study shows that levels of OCPs, PCBs, BFRs and PFAS are in general, lower in 481 farmed fish compared to wild fish within the same lake. This indicates less risk for fish health 482 in farmed than in wild fish. Farmed fish from this study is also considered safe for human 483 consumption. There was a geographical trend with higher levels of DDTs, PCBs, PBDEs and 484 PFASs from west to east of Lake Kariba. However, levels of HCB and HCHs were also higher 485 in fish that had foraged near the farms, and this need further investigation to elucidate possible 486 sources. The contribution of p,p'-DDD to ΣDDTs increased eastwards, possibly due to higher 487 environmental impact of anaerobic processes. PCB levels were low, and Fish length, liver lipid content and contaminant concentrations in livers from wild sh (site 1-3) and farmed sh (farm 1-2) in Lake Kariba, Zambia. Fish length is given in cm, liver lipid content in % and liver concentrations are presented as ng/g lipid weight. Box plots show median (line), IQR (box) and minimum to maximum (whiskers). Statistical differences were determined using Kruskal Wallis with Dunn`s post hoc test with and without Bonferroni's corrections for multiple testing. Letters (a-e) indicate statistically signi cant difference (p<0.05) between the sites and farms. Asterisk (*) indicates statistical signi cance (p<0.05) after Bonferroni's corrections for multiple testing. Tables15.pdf SupplementaltablesS1.11.2and guresS1.pdf