Introduction

Organochlorine pesticides

Due to their persistent nature and global distribution, organochlorine pesticides (OCPs) have been banned in most countries around the world (Ritter et al. 1995). While they were still in use these OCPs played a large role in agriculture for their use as insecticides (Ritter et al. 1995; UNEP 2010). South Africa is one of the few African countries where the use of dichlorodiphenyltrichloroethane (DDT) as malaria vector control (MVC) agent through controlled indoor residual spraying (IRS) is still permitted (Bouwman et al. 2011). The IRS method limits the quantity of pesticides used in MVC through only allowing application to the interior walls and ceilings, and exterior eaves of homesteads in malaria risk regions (Gaspar et al. 2016). There is however increasing evidence that DDT has still been entering the environment long after its conventional use was banned. Van Dyk et al. (2010) showed the presence of DDT in chickens and other matrices in and directly around homesteads in the Limpopo Province of South Africa. Gerber et al. (2016) and Viljoen et al. (2016) assessed DDT bioaccumulation in aquatic biota from rivers in the same region and recorded levels of 4000–32,000 (ng/g lipid weight) ∑DDTs in tigerfish (Hydrocynus vittatus) and 5–650 (ng/g wet weight) ∑DDTs in anurans (Xenopus spp.) respectively. Recently Verhaert et al. (2017) investigated trophic transfer of various persistent organic pollutants including OCPs in the aquatic food web of the Olifants River Basin in South Africa. They detected DDTs in most of the food web components analysed and showed trophic magnification of DDTs to be higher in this sub-tropical region (during winter specifically) compared to data from a tropical region. Thompson et al. (2017) and Yohannes et al. (2017) reported DDT exposure in domestic poultry and wildlife from in and around homesteads in the Phongolo floodplain region in northern KwaZulu-Natal Province of South Africa.

This region is regarded as a biodiversity hotspot for amphibians (Netherlands et al. 2015) and the risk of DDT exposure to these and other aquatic organisms is still not fully understood (Dube et al. 2016). Its use is crucial in the prevention of malaria (Bouwman et al. 2011). Despite their environmental persistence, knowledge regarding levels and effects of DDT and other OCPs in anurans are limited with most of the available toxicity data being generated in the 1970s (Pauli et al. 2000). The majority of studies on DDT bioaccumulation in anurans only report on body burdens and not concomitant biological responses (Jofré et al. 2008; Viljoen et al. 2016). According to Monastersky (2014) amphibians are currently amongst the most threatened animals on earth, that is why it is essential to evaluate exposure scenarios that could influence their populations health and fitness (Melvin et al. 2016).

Another organochlorine pesticide group of concern in the Phongolo region are the hexachlorocyclohexanes (HCHs). Buah-Kwofie and Humphries (2017) attributed the high levels of ΣHCHs in sediments from water bodies in northern KwaZulu-Natal (83–187 ng/g dry weight) to the recent agriculture application in the region. These HCHs were historically mainly used in the form of technical HCH (consisting of 55–80% α-, 5–14% β-, 8–15% γ-, 6–10% δ-, and 1–5% ε-HCH isomers) and in later years lindane (γ-HCH) as broad spectrum insecticides in agriculture (UNEP 2010). Main uses included seed treatment, ecto-parasite control on livestock, wood, and tree covering (UNEP 2010). These historical uses have relevance to the study area because of catchment land use.

Study area

The Phongolo floodplain in north eastern South Africa is a biodiversity hotspot with over 400 bird species, more than 40 fish species and 45 frog species (Mallory 2002; Du Preez and Carruthers 2017). Since the mid-1970s this large floodplain has been subjected to artificial flooding due to the construction of the Pongolapoort Dam (Wepener et al. 2012). This regulation, together with below average rain falls for the past decade, has resulted in a decline in the water quantity and quality of the Phongolo River and associated floodplain wetlands (Dube et al. 2016). Dube et al (2016) indicate that the stress of decreased water and increased levels of OCPs such as DDT and other agricultural pesticides pose a threat to the aquatic biodiversity of the floodplain. Land use in the Phongolo River catchment upstream of the floodplain mainly consists of forestry and sugarcane farming, which as stated in section 1.1 has been historically associated to HCH usage. In the floodplain itself maize farming (mostly as subsistence farming) is practiced on a smaller scale (Personal observation). Other OCPs that used to be important in these agricultural activities are Chlordane and Aldrin (Ritter et al. 1995).

Indicator species

The use of Xenopus as a model organism is well established in ecotoxicology with embryos commonly used to study teratogenic effects using the Frog Embryo Teratogenesis Assay – Xenopus (FETAX). Adult Xenopus are not in particular known to have high susceptibility to toxic effects of anthropogenic pollutants (see Pauli et al. 2000), but this can make them good indicators of bioaccumulation and good models for sub-lethal effect. Their central placement in the aquatic food web (Lindholm et al. 2007) means early response effects measured in Xenopus can be monitored and managed before effects reach top predator level at which point irreversible effects may have occurred throughout the rest of the food web. Their fully aquatic nature and bottom dwelling habits also make them more prone to contact with pollutants that accumulate in sediments such as OCPs and uptake is facilitated through their highly permeable skin. Xenopus laevis is the most commonly used model anuran from Southern Africa. The study area marks a slight overlap in the natural distribution of X. laevis and its sub-tropical relative X. muelleri (Du Preez and Carruthers 2017), but X. laevis is in reality quite scarce in this region (Netherlands et al. 2015) resulting X. muelleri as being the obvious choice for this study. Positive identification was easily made through the long (longer than the radius of the eye: distinct from X. laevis) sub-ocular tentacles found on X. muelleri (Du Preez and Carruthers 2017).

Biomarker responses

When attempting to link environmental exposure of pollutants to biological effects, biomarker responses are commonly used. An integrated biomarker response approach has recently been shown to be useful in discriminating among environmentally relevant exposure scenarios in aquatic ecosystems by Schoenaers et al. (2016). The integrated biomarker suite for this study were selected to indicate, biotransformation of xenobiotic compounds through cytochrome P450 (CYP450) activity, exposure to neurotoxic compounds through acetylcholinesterase (AChE) activity, oxidative stress through means of both antioxidant systems—superoxide dismutase (SOD) and catalase (CAT)—and oxidative damage indicators—protein carbonyls (PC) and malondialdehyde (MDA)—, and energetics though cellular energy allocation. This suite of biomarkers is considered to be indicative of anthropogenic stressors according to Gerber et al. (2018), who showed that these biomarker responses can be indicative of OCP and metal accumulation in tigerfish (Hydrocynus vittatus). Oxidative stress biomarkers were used by several authors (Borković-Mitić et al. 2016; Prokić et al. 2016) to determine effects of metal exposure in the marsh frog, Pelophylax ridibundus. The same biomarkers were applied to evaluate the effects of carbamate pesticides (Falfushinska et al. 2008) and organophosphates (Kanter and Celik 2011) to the same species. Biomarkers have also been applied in laboratory bioassays to evaluate the anti-oxidant responses of X. laevis to crude oil extracts (Eriyamremu et al. 2008). These biomarker responses are therefore also regularly used in studies on frogs and may be applied to contaminant groups such as OCPs for investigating the interactions between chemical pollution and biochemical changes. Apart from direct toxicity mechanisms expected to be activated through exposure to OCPs, the persistence of these compounds at detectable levels in an organism may decrease the cellular energy budget and increase radical oxygen species formation (Gerber et al. 2018), which could be detected through means of the measured biomarker responses in this study.

Study aims

The aims of this study were therefore to determine the extent to which amphibians from this system are exposed to DDT and other OCPs, along with the temporal variation in chemical loads, using the anuran Müller’s clawed frog (Xenopus muelleri) as indicator organism. To evaluate the impact of continued DDT inputs through isomer ratio analysis. To compare DDT bioaccumulation in X. muelleri to bioaccumulation of other OCPs expected from historical use in the system. Finally, to relate the OCP exposure to biological effects in X. muelleri through utilisation of a suite of biomarker responses. This includes biomarkers of exposure, oxidative stress and cellular energy allocation.

Materials and methods

Ethical clearance and permits

Frogs were collected under Ezemvelo-KZN Wildlife permit numbers: OP 5139/2012 & OP 5261/2014. The study was approved by the North-West University AnimCare Ethics Committee (NWU-00095-12-A4).

Field methods: sample collection and handling

Frog sample collection was conducted during four separate surveys over a two-year period. IRS application occurs in the summer months (November to February). The November 2012 and 2013 surveys were therefore during the MVC period and are labelled D (Table 1). The two surveys after the MVC period were undertaken in April 2013 and 2014 and are labelled A (Table 1). Sampling was conducted in floodplain ponds within the Ndumo game reserve to determine the chemical loads of the floodplain frogs in the area (Fig. 1).

Table 1 Survey labels and descriptions
Full size table
Fig. 1
figure 1

Map of the study area indicating sampling area. Xenopus muelleri were collected from within the Ndumo Game reserve for chemical and biomarker analysis

Full size image

Xenopus muelleri collection was done through both active (night frogging by hand) and passive (commercial chicken liver baited bucket traps) sampling techniques. The animals were euthanized through double pithing as described by Amitrano and Tortora (2012) as chemical euthanasia could possibly compromise the biomarker results (Velisek et al. 2011). The liver mass and total body mass of each animal was recorded upon dissection for calculation of the hepatosomatic index (HSI), which is defined as the ratio between the two values. For biomarker analyses the liver and muscle from the right thigh was removed and placed in Hendrikson’s buffer (40 mM tris-HCl, 10 nM 2-Mercaptoethanol, 1 mM 0.04% bovine serum albumin [BSA], 1 nM ethylene-diamine-tetraacetic acid [EDTA]) and flash frozen in liquid N2. These samples were later transferred from liquid N2 to an −80 °C freezer in the laboratory until biomarker analysis was undertaken. The remaining carcass, to be used for whole body OCP bioaccumulation analysis, was wrapped in aluminium foil and frozen at −20 °C until analysis. The absence of the entire liver and a piece of muscle for the whole frog analysis was consistent throughout the study, making results from this study comparable. However, the lack of these pieces could affect comparisons with literature, as the liver is considered a major bioaccumulation organ for xenobiotics. Fagotti et al. (2005) found that DDT tissue accumulation patterns in Rana esculenta were that the brain contained the highest concentration, with the liver accumulating the third largest percentage of the total DDT in the body. Therefore it should be borne in mind that the absence of liver in the analysed whole specimens from this study probably result in underestimation of OCPs when compared to other studies using whole organisms in the literature.

Organochlorine pesticide analyses

The method used for chemical residue extraction and analysis was adapted from Yohannes et al. (2013). Automated hot-Soxhlet extraction using 1:3 v/v acetone:hexane was done on anhydrous NaSO4 desiccated whole frog samples (5–10 g). Lipid percentage was determined gravimetrically on a dried 20% aliquot of the extract. Samples containing more than 5% lipids were subjected to gel permeation chromatography for size exclusion separation of lipids. For final clean-up, samples were filtered through 6 g of 5% deactivated Florisil and eluted with 3:7 v/v dichloromethane:hexane (100 mL). Extracts were then evaporated to dryness and reconstituted in 100 μL n-Decane.

Chemical analysis was performed on a gas-chromatograph with a Ni electron capture detector (GC-ECD: Shimadzu GC-2014) with Nitrogen as make-up gas (45 mL/min flow rate). An ENV-8MS capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness) was used for separation of target compounds with He as carrier gas (1 mL/min flow rate). Splitless injection was used (1 μL injection volume; 250 °C inlet temperature). The following GC oven program was set: initial temperature of 100 °C held for 1 min, followed by 12 °C/min ramp to 180 °C, followed by 4 °C/min ramp to 240 °C, followed by 10 °C/min ramp to 270 °C held for 5 min. The detector temperature was set at 320 °C. A mixture containing 22 persistent organic pollutant (POP) pesticides (Dr Ehrenstorfer, GmbH) was used as external standard. Calibrations were done using 5-point standard curves from concentrations ranging between 10 µg/L and 500 µg/L (R2 ranged between 0.997 and 0.999 for all compounds). Recoveries were calibrated using PCB# 77 as surrogate marker and were between 60 and 95% (mean 75.6%). The data reported were adjusted according to recovery. Standard reference material SRM 1947 (Lake Michigan Fish Tissue) analysed using the same method produced recoveries ranging from 75 to 110% with the residual standard deviation less than 12%. Limits of detection based on 3:1 signal to noise ratio (S/N) were below 0.5 ng/g for all OCPs.

Biomarker analyses

Samples were thawed and preparation was conducted at 4 °C to preserve protein stability. Three sample batches (A, B, and C) were prepared. Due to the small size of frog livers, some specimen samples were pooled to obtain sufficient sample mass for biomarker analyses. Sample pooling was done per collection site and samples below the required minimum liver mass were pooled together in equal mass at random (within each site and survey) until all samples had reached the required liver mass. Batch A consisted of 0.05 g of liver tissue in 250 µL of Tris-sucrose buffer (0.05 M Tris-HCl [pH 7.4], 0.2 M sucrose). The samples were homogenised and centrifuged at 9500 × g for 10 min. The supernatant was used for acetylcholinesterase (AChE) and malondialdehyde (MDA) analysis. For Batch B 0.05 g of liver tissue in 1 mL potassium phosphate buffer (0.09 M K2HPO4 + KH2PO4 [pH7.4]) was homogenised and centrifuged at 10,000 × g for 30 min. The supernatant from this batch was used for catalase (CAT), superoxide dismutase (SOD), cytochrome P450 (CYP450) and protein carbonyl (PC) analyses. The third batch (C) consisted 0.2 g muscle tissue in 400 µL of electron transport system (ETS) homogenising buffer (0.1 M Tris-HCl [pH 8.5], 0.2% v/v Triton-X, 15% m/v polyvinyl pyrolidone, 153 µM MgSO4), which was then homogenised and used for the cellular energy allocation (CEA) analyses.

Acetylcholinesterase activity was determined using the assay described by Ellman et al. (1961) where thiocholine, a breakdown product of acetylcholine formed through hydrolysis by AChE, reacts with dithiobisnitro-benzoate (DTNB) to generate a 5-thio-2-nitrobenzoate anion. The product has a yellow colour which is quantified by its absorbance at 405 nm. Cytochrome P450 demethylating activity was measured using an assay kit based on formaldehyde formation (DetectX Demethylating P450 fluorescent activity kit, Arbor Assays). These two biomarkers were performed as biomarkers of exposure indicating both direct toxicant effect and xenobiotic metabolism activation. To measure oxidative stress an array of enzymatic and non-enzymatic biomarkers regarding the cellular antioxidant system were analysed. The SOD activity was determined using the assay described by Del Meastro and McDonald (1989). In this assay the inhibition of pyrogallol autoxidation is measured as a function of the rate of free radical Oxygen conversion to H2O2 by SOD. The CAT activity was determined by a static timed reaction between H2O2 and CAT enzymes. KMnO4 is added after the reaction as photometric agent which in turn reacts with residual H2O2. The excess KMnO4 is then measured as a function of the amount of H2O2 breakdown by CAT in the reaction time period as described by Cohen et al. (1970). The MDA content was determined using a colorimetric assay based on thiobarbituric acid’s reaction with MDA as a measurement of lipid peroxidation as described by Üner et al. (2006). Protein carbonyl content was measured through a 2,4-Dinitrophenylhydrazine (DNPH) reaction with carbonyl groups in proteins and provides a measurement of protein oxidation based on the method of Parves and Riasuddin (2005). The activity and concentrations of all the biomarkers assayed in the liver tissue are expressed in terms of mg protein which was determined using the method of Bradford (1976) for each sample after pooling of batches A and B for the assays performed within the batch.

The cellular energy allocation assay performed is based on the protocols of De Coen and Janssen (1997) and De Coen and Janssen (2003) and reflects the difference between the total energy available (in terms of carbohydrate, lipid and protein reserves) and the energy consumed (as reflected in electron transport system (ETS) activity. Total carbohydrates were determined using the GOD-PAP 1 448 668 Roche test kit. This method Lipid content was assayed according to Bligh and Dyer (1959) while the protein content was once again determined according to the method described by Bradford (1976). The above concentrations were converted into energy equivalents using the combustion enthalpy of glycogen (17,500 mJ/mg), protein (24,000 mJ/mg), and lipids (36,500 mJ/mg) to calculate the total available energy (Ea) (De Coen and Janssen 1997). The consumed energy (Ec) was determined by means of the stoichiometric relationship between formazan formation and oxygen consumption (1:2) of the ETS. The oxygen consumption results were then converted into energy equivalents using the combustion enthalpy of O2 (484 kJ/mol) (De Coen and Janssen 1997). An energy budget was calculated using the equation: CEA = Ea − Ec with Ea being the sum of the protein, carbohydrate and lipid energy equivalents and Ec being the ETS energy equivalents.

Statistical analyses

Prior to analysis the data were checked for normality (D’Agostino-Pearson normality test) and homogeneity of variance (Levene’s test). The temporal differences in OCP bioaccumulation and biomarker responses were measured using the non-parametric Kruskal–Wallis test coupled with Dunn’s post hoc test as none of the datasets conformed to normality. Significance was regarded at p < 0.05. For all statistical analyses non-detect chemical data were replaced with LOQ × F per compound per survey, where LOQ is the limit of quantification of the compound and F is the frequency of detection within the survey (Wepener et al. 2012; Verhaert et al. 2013). Non-detect values are however shown as ND in tables and zero values in graphs for simplified visual interpretation. Temporal correlations in OCP concentrations as well as correlations with abiotic factors were measured through means of Spearman correlations of individual frog chemical data as all data in this instance did not have Gaussian distribution. Pearson’s correlation, and Log (dose) vs. response variable slope nonlinear regression analyses were performed on hepatosomatic index (HSI) values vs. individual OCP concentrations to investigate possible relationships. Generalised linear model (GLM) multiple regressions were performed in conjunction with a redundancy analysis (RDA) between individual biomarker responses and OCP accumulation data to assess relationships while accounting for the composition of the data. Forward selection of explanatory variables in the RDA was used to select significant response variables on which GLM regression was performed.

Results

Organochlorine pesticide bioaccumulation

Chemical analysis revealed the presence of 10 of the 22 analysed OCPs in X. muelleri tissue from the Phongolo floodplain (Table S1). All of the 51 (total number of frogs from all surveys) analysed frog samples contained detectable OCPs. Specific detection frequencies are provided in supplementary material (Table S2). Two compound groups detected with highest prevalence and at distinguishably higher concentrations than others were DDTs and HCHs, making up cumulatively, more than 90% of detected OCPs for all surveys (Fig. 2). Individual compounds detected at highest concentrations per single frog (before pooling) were γ-HCH, p,p-DDD, and p,p-DDE at 19,364, 7219 and 7057 ng/g lipid respectively. γ-Hexachlorocyclohexane made up the majority of the total HCH concentrations detected. Other OCPs detected at lower levels included Aldrin, cis-Chlordane, trans-Heptachlor epoxide, α-HCH, δ-HCH and o,p-DDT (Fig. 3a–e,Table S1).

Fig. 2
figure 2

Bar graph indicating the percentage composition of organochlorine pesticide (OCP) groups contributing towards total OCP accumulation for each survey (DDTs dichlorodiphenyltrichloroethane isomers and their metabolites, HCHs hexachlorocyclohexane isomers, HTPCHLs heptachlor isomers of which only trans-heptachlor epoxide was detected, CHLs chlordane isomers of which only cis-chlordane was detected, Drins Aldrin was the only detected compound from the drin group)

Full size image
Fig. 3
figure 3

Bar graphs indicating the mean (±standard error) concentrations (ng/g lipid) of the various OCP groups in Xenopus muelleri tissue across the four surveys for a DDTs (dichlorodiphenyltrichloroethane isomers and their metabolites), b HCHs (hexachlorocyclohexane isomers), c HTPCHLs (heptachlor isomers and metabolites) of which only trans-heptachlor epoxide was detected, d CHLs (chlordane isomers) of which only cis-chlordane was detected, and e Drins of which only Aldrin was detected. Pooled sample sizes are indicated for each survey with * marking the original number of frogs used

Full size image

During vs after spraying period

The highest total OCPs mean (±SE) concentration as well as the highest maximum value per frog was recorded as 8689 ± 2037 ng/g lipid and 21,399 ng/g lipid respectively, both from the 2014A survey. For the same survey γ-HCH had the highest mean for any single compound detected at 6350 ± 1803 ng/g lipid. OCP compositional changes were observed for sequential surveys per spraying season, where a decrease in total DDTs from during (D) to after (A) spraying surveys and an HCH increase over the same periods was seen (Fig. 3b). There were no significant differences in total OCP levels between this timeframe from during (D) to after (A) spraying per season. However significant differences were recorded in total OCPs between the two active spraying period surveys (2013A and 2013D; p < 0.05) (Fig. 4). A temporal increase in total OCPs was observed over the total period of the study (Fig. 4; Spearman r = 0.84 p < 0.0001). Significant negative correlation (Spearman r = −0.531, p value = 0.0004) was found between total OCP accumulation and average rainfall data from the area during survey months (Rainfall data obtained from Worldweatheronline (2016); Fig. 4, Table S3). No significant correlation was found between rainfall and animal body mass nor total OCPs and animal body mass. Furthermore there was significant negative correlation between total HCH accumulation and rainfall, but not between total DDT accumulation and rainfall (Table S3). The ratio between p,p-DDT and its metabolites (sum of p,p-DDD and p,p-DDE) was used to determine recentness of DDT exposure as described by Strandberg and Hites (2001). These ratios of the individual frogs analysed only indicated samples with recent DDT exposure (DDT/∑metabolites > 1) at the end of the 2012–2013 spraying season (2013A). There was an absence of p,p-DDT for the 2014A survey (Fig. 5). The HCH composition (Fig. 6) indicated γ-HCH as the main contributing isomer in all surveys making up more than 99% of the total HCHs except during the 2012 spraying season where δ-HCH was the main contributing isomer.

Fig. 4
figure 4

Scatter plot of individual frog data points (black diamonds) and mean values (black lines) for total organochlorine pesticide (OCP) accumulation (left y-axis) in Xenopus muelleri for all of the surveys conducted. Overlaid on the right y-axis is the average rainfall in the sampling area during the month of each survey (grey circles and connecting lines). Spearman correlation indicated significant unimodal increase over time (Spearman r = 0.84) for total OCP concentrations. No significant correlation was found between rainfall and OCP data

Full size image
Fig. 5
figure 5

Normalised bar graph indicating the percentage contribution of the different dichlorodiphenyltrichloroethane (DDT) isomers and metabolites towards the total DDTs measured during each survey

Full size image
Fig. 6
figure 6

Normalized bar graph indicating the percentage composition of different Hexachlorocyclohexane (HCH) isomers contributing towards the total HCH concentration for each survey

Full size image

Biomarker responses

During vs after spraying period

All biomarker response results are given in Table 2. The CYP450 assay results indicated no significant temporal changes. Results from the AChE, PC and MDA assays indicated a significant decrease between spraying seasons from 2013A to 2013D, while SOD and CAT activity increased significantly over the same period. The CEA results for total available energy indicated no significant temporal change, however, the composition of available energy sources changed over time, from being mainly protein based to being mainly lipid based, with a subsequent decrease in carbohydrate energy storage. The energy consumption indicated a slight temporal increase. This variation in the CEA data correspond to the changes observed in the oxidative stress biomarkers. The total CEA budget also showed no significant differences on a temporal scale.

Table 2 Biomarker response assay results for Xenopus muelleri indicating the mean value and standard error of the mean (±SEM) for all the surveys conducted
Full size table

Exposure vs biomarkers

A variable slope logarithm of dose vs response curve (Y = Bottom + (Top − Bottom)/(1 + 10^((LogIC50 − X) * HillSlope))) was fitted to bioaccumulation and corresponding HSI data. The relation between HSI values and γ-HCH was found to have a moderate fit with the regression line Y = 0,01047 + (0,02607–0,01047)/(1 + 10^((2683−X)*−2045)) (R2 = 0.54; p < 0.0001), showing a decrease in HSI as the concentration of γ-HCH in the body increases. None of the other OCPs measured fitted dose response regressions towards HSI. In terms of correlation however, the Log values of the total OCP concentrations showed moderate negative correlation with HSI (Pearson’s r = −0.53; p < 0.0001).

A redundancy analysis (RDA) was performed on pooled samples chemical data with their corresponding biomarker response. Only 49.6% of the total variation in chemical data was accounted by the biomarker responses with no clear separation of surveys through this analysis (Figure S2). Only 36.91% of the variation is described across the first two axes. Monte-Carlo permutation test indicated that explanatory variables significantly accounted for variation in the response data (p = 0.006). Individual axis statistics regarding the RDA are given in Table 3.

Table 3 Results of the redundancy analysis (RDA) performed on biomarker response and OCP bioaccumulation data (bi-plot of first two axes shown in supplementary Figure S2)
Full size table

Using forward selection of explanatory variables the GLM regression of normalized responses vs individual OCP concentrations indicated some significant relationships (Table 4). The AChE, MDA, Ea-carbohydrates, and Ea-proteins values had significantly non-zero negative relationships towards p,p-DDE accumulation. The Ec and Ea-lipids showed significant positive relationships towards p,p-DDT. All of these responses were only significantly explained by a single variable. The CYP450 results showed a negative relationship towards α-HCH as the first selected variable and in conjunction with α-HCH a positive relationship towards δ-HCH as the second selected variable explaining CYP450 variation.

Table 4 Generalised linear model regression between normalized biomarker responses and organochlorine pesticide concentrations (ng/g lipid) in Xenopus muelleri from all surveys conducted
Full size table

Discussion

Organochlorine pesticide bioaccumulation

Most compounds were only found in low ng/g lipid amounts as would be expected from legacy pesticides that have already been phased out of use for a few decades. All OCPs detected were well below known effect concentrations for amphibians reported in literature. It is noteworthy however that almost all available toxicity literature with regards to OCPs do not concern African species. A body burden as high as 110,000 ng/g lipid was reported for p,p-DDT with no observable effects for Rana temporaria tadpoles (Cooke 1979). The recent study by Viljoen et al. (2016) in another DDT-sprayed area of South Africa indicated a mean concentration of 402.1 ng/g wet mass total DDTs from lipid bodies of Xenopus spp. sampled in 2009. The current study analysed whole frogs with the highest wet mass mean (per survey) concentration at 16.1 ng/g and the highest lipid mass mean concentration at 2062.1 ng/g. Lipid mass concentration of the whole body would have a closer relation to wet mass concentration of lipid bodies, but an outright comparison of concentrations from the two studies cannot be made to the differences in methods used. Lambert (2001) lists a table of DDT residues in amphibians in sub-Saharan Africa, all of which are in the μg/g range in terms of wet mass, 100 times higher than values from this study. Most of these studies were during the active unregulated use of DDT. Differences in chemical levels between studies in Lambert (2001) could indicate the positive impact of strict regulations and proper IRS implementation by governments in reducing environmental DDT levels, compared to times in history when DDT use was unregulated. The recentness of exposure analysis results indicate the uptake of new DDT from the environment by the end of the first spraying season, but not the second. The term recent in this analysis refers to the lack of time for the DDT to break down into its metabolites. This length of time varies due to a number of environmental factors that apply to the breakdown of these compounds in the environment. Recent introduction of DDT is assumed when the concentration of the parent compound is higher than that of the sum of the breakdown products. This means that there was not enough time for breakdown before uptake, as well as not enough time for biological breakdown after uptake. Even though the DDT metabolites p,p-DDD and p,p-DDE are also well documented to have both primary toxic effects as well as secondary endocrine disruption effects (Bouwman et al. 2011) this ratio is important in understanding the dynamics of the compounds within these organisms and this system as well as interpreting any possible effects observed as the different compounds have different toxicity signatures (ATSDR 2002).

The most abundant isomer of HCH was γ-HCH, which is also the most toxic isomer (ATSDR 2005). Fagotti et al. (2005) measured 2.38 × 10−3 ng/g wet mass of γ-HCH in Rana esculenta (now Pelophylax esculentus) from central Italy compared to 6.59 ng/g wet mass mean concentration in X. muelleri for the 2014 after spraying survey of this study. Toxicity test results on amphibians vary greatly between sources, most likely due to species differences. The lowest LC50 reported was for Rana limnocharis (now Fejervarya limnocharis), with the LC50 being 0.94 mg/L after 48 h exposure (Pan and Liang 1993). Marchal-Ségault and Ramade 1981 indicated reduced time to metamorphosis in X. laevis tadpoles exposed to 2 mg/L of γ-HCH. None of these studies however report body burdens in comparison to toxicity, making interpretation of field results complex. The composition of total HCHs by different HCH isomers can indicate the type of HCH used in the system so some extent. Technical grade HCH has a more balanced composition while lindane consists almost solely of γ-HCH. In this system only the first survey samples contained isomers other than γ-HCH. What is interesting is that neither α-HCH (main isomer in technical grade HCH) nor β-HCH (most persistent isomer) were the most prevalent isomers for that survey which alludes to a unique uptake or metabolic scenario for the frogs during this survey. The consistency of lindane accumulation detected throughout the other surveys of this study corresponds with historic lindane use in the area. This possibility of historic use is based on forestry land use in the upper catchment of the river and livestock dipping for ecto-parasites in the floodplain as possible sources.

The increase in total OCPs over the study period suggests recent availability of the compounds in the floodplain aquatic system. Composition changes irregularly along with the increase. There is a general decreasing trend in rainfall over the study period that shows negative correlation with total OCP accumulation. A similar correlation was found between total HCHs and rainfall, which makes sense as this compound group made up a large proportion of the total OCPs in most samples. This evidence leans toward indicating that accumulation of these compounds is driven by the rainfall dilution effect where increased rainfall (i.e. increased water volume) decreases the effective concentration of compounds in a specific water body. This is expected of historical persistent lipophilic compounds in aquatic systems because of changes in the binding to dissolved organic matter as described for b-HCH by Kalbitz et al. (1997). There was however no such correlation between total DDTs and rainfall. This lack of any relationship shows that the pattern of DDT accumulation was not significantly influenced by rainfall. If DDT in this system were mostly from historical use, one would expect negative correlation similar to that found between HCHs and rainfall. If DDT was largely from new input in surrounding areas one would on the other hand expect a positive correlation with rainfall as runoff would be the main input into a water body. The fact that there is no correlation combined with the presence of recent DDT input in some, but not all surveys, could indicate that DDT fluctuations are from specific non-rainfall related input events and that run-off is not such a major pathway for DDT, to significantly increase accumulation in the frogs. However, four surveys are not enough to make conclusive arguments regarding rainfall patterns and correlations. The question is however raised by this data is if run-off is not a major input factor for DDT in this ecosystem, what is? Illegal use or improper handling of DDT in the area has not been documented, but cannot be ruled out.

Biomarker responses

There is a reduction in oxidative damage (MDA and PC) observed in this study, as the anti-oxidant enzyme activity (SOD and CAT) increases. This indicates that the initial oxidative damage was not necessarily caused by overloading of the anti-oxidant system. This introduces the possibility that the anti-oxidant enzyme activity during the first spraying season could have been inhibited. Antioxidant enzyme inhibition has been observed in Orechromis niloticus in rivers associated with metal contamination by Carvalho et al. (2012), however none of the OCP concentrations measured in this study have been reported to result in such a response. Banerjee et al. (1999) reported increased SOD and CAT activity in humans with high lindane exposure. In this study however the increase in OCP accumulation between the two seasons where the oxidative stress changes were observed consisted mostly of DDTs. It is also possible that the specific mixture of contaminants had additive or opposing effects resulting in the observed change. Mixture exposure effects on oxidative stress responses have previously been documented (Kanter and Celik 2011; Melvin et al. 2016). The changes in Ea composition observed could indicate either a dietary change, or a response towards increased stress on the organism. The increased lipid energy storage is generally considered a positive response, however as it is coupled with a decrease in both, protein and carbohydrate storage, as well as a slight increase in ETS activity, it is more likely to be considered a stress response in this instance. There is increasing evidence that persistent pollutants such as OCPs can act as obsesogens. La Merrill et al. (2013) discusses the effects and influences that POPs accumulation may have on adipose tissue and vice versa. The authors state that both positive and negative correlations may exist between POPs and adipose tissue mass, but that these correlations are compound specific. La Merrill et al. (2013) also states that the obesogenic effects of contaminants may follow non-monotonic dose-response patterns. Increased body mass or adipose tissue mass might be seen at low doses of a certain compound while higher doses result in cachexia. The lipid energy response seen in the current study may well be due to obesogenic effects, but more research in this area would be needed to fully understand the mechanisms in anurans, and at these levels of accumulation. Melvin et al. (2016) indicated a decrease in glycogen levels of the striped marsh frog (Limnodynastes peronii) when exposed to increasing concentrations of wastewater effluent. The change from protein based energy storage to lipid based storage in the current study coincides with an increase in both protein and lipid oxidative damage. An increase in protein damage could lead to reduced protein storage. As this change occurs over the same period as the dramatic increase in lipophilic OCP accumulation, it is possible that the body could increase lipid storage in reaction to exposure to OCPs. The increase in lipid content due to persistent organic pollutant exposure has been documented in fish (Hodson et al. 1992). Even though the specific effects in anurans are not known the evidence does seem to point towards obesogenic effects as stated by La Merrill et al. (2013). All of the observed bioenergetics changes in the present study were only compositional and did not drastically affect the total CEA of X. muelleri. The frogs therefore were able to compensate for minor changes in their energy storage and consumption in such a way that their net energy budget remained fairly unchanged leading to the conclusion that their total fitness (as defined by their bioenergetics) was not influenced by the presence of OCPs.

The negative correlation of HSI with γ-HCH accumulation is contradictory towards the results obtained by Melvin et al. (2016) who found a positive correlation towards wastewater effluent exposure, although they also mention several references where both positive and negative trends have been found in fish exposed to wastewater treatment. An increase in HSI would be expected if there is increased xenobiotic metabolism taking place in the liver. This increase in liver mass has been associated with hypertrophy and increased detoxification enzyme activity in fish exposed to various contaminants (Andersson et al. 1987; Banaee et al. 2013) OCPs are however persistent and accumulative (Ritter et al. 1995) and high enough levels could hypothetically cause liver damage rather than being metabolised. The results observed in this study are most likely associated with liver atrophy similar to that reported in Banded Gourami (Colisa fasciatus, now Trichogaster fasciatus) exposed to γ-HCH (Verma et al. 1975). This atrophy can be the result of oxidative damage evidenced through the PC and MDA analyses results, however histological analysis would be necessary for determination of specific causes. Mchugh et al. 2011 reported histological signs of liver damage such as granular degeneration in tigerfish (Hydrocynus vitatus) from the Pongolapoort dam, with a maximum accumulation of 5537.41 ng/g lipid.

As this study was field based the composition of OCPs between individual frogs showed high variation along with their measured responses. The use of multivariate statistics was incorporated in order to determine at which extent each OCP individually and related to each other was correlated with the observed effects. This takes into account the possibility of multiple OCPs having similar or opposite effects within a mixture. There was no clear separation or grouping of samples based on survey through RDA analysis (Figure S2) indicating that observed relationships between any biomarker response and OCP accumulation observed were not driven by survey temporal differences, but rather by individual sample chemical load and composition. The quantification of significant correlations within the data matrix were done through GLM multiple regression (Table 4). Decreased AChE can be a result of neurotoxicity since it is known to be inhibited by pesticides such as organophosphates. The toxic effects may be enhanced for p,p-DDE due to its persistence in fatty tissue (Van der Oost et al. 2003), even though the toxicity is generally lower for this compound than for p,p-DDT. The decreased MDA response is expected to coincide with increased SOD and/or CAT, however the data suggests that the enzyme activity is not driven by a specific factor, but a combination of compounds regulate the anti-oxidant enzymes with greater variability. The resulting decrease in lipid peroxidation then being a more stable response shows relation to p,p-DDE levels, but is not necessarily directly decreased by p,p-DDE. Lipid peroxidation is not generally associated with DDE exposure in fish, but has in some instances shown decreases related to the presence of OCPs, however it has been indicated as an important response to low level environmental toxicants in general (Van der Oost et al. 2003). As fundamental data on anurans are lacking in this regard, all possible associated responses still require investigation. The CEA components also showed relation to both p,p-DDT and p,p-DDE, which brings slight plausibility to the hypothesis that the changes observed within the CEA data itself were due to stress response. However, without solid evidence this is simply a possibility and cannot be proven. Both protein and carbohydrate energy storage showed negative relation to p,p-DDE, meanwhile the components, Ea-lipids and Ec, showed positive responses to p,p-DDT levels. Spacie and Hamelink (1982) suggested increased rate of hydrophobic chemical uptake in fish having higher lipid content. Following the logic of this idea available lipid energy storage could have led to higher p,p-DDT uptake rates. However, the increased Ec with higher p,p-DDT does rather suggest increased physiological stress on the animals due the presence of the parent DDT compound. The conclusions of Spacie and Hamelink (1982) do not account for these Ec changes and is therefore not likely to be applicable to the present study scenario.

The regression analysis results mostly concern DDT, but two low level compounds δ-HCH and α-HCH showed combined influence on the CYP450 responses. As one response was negative and one positive these results show the need for mixture exposure scenario research. Kortenkamp (2007) states that the effect of pesticides that show individual effects can be reduced when these pesticides are combined. The two opposing responses found in the present study, might very well be the case, but it would have to be confirmed through testing. Both an increase and decrease in CYP450 has been related to OCP exposure in different fish (Van der Oost et al. 2003). The δ-HCH–CYP450 response had the best fit of all regressions in the matrix and was the only variable to show significant secondary influence on a biomarker response. The overall composition of chemicals and the absence of these HCH isomers in some surveys however decrease the likelihood of causality being an accurate interpretation without more knowledge on the behaviour of these compounds in mixtures. In general deriving causality from field based observations is an arduous task due to the large number of variables in play, but the distinct correlations found within the matrix of chemical and response data from this study do suggest some form of interconnectivity in the instances described. More data would be required in order to factually attribute these responses to OCP accumulation specifically, but this study does open a line of scientific questioning to be pursued in future research.

Conclusion

There is clear evidence that anurans from the Phongolo River floodplain in South Africa do accumulate OCPs from the aquatic environment. The level of total OCP accumulation has shown an increase over the period of this study and did not correlate to seasonal trends in terms of introduction of Malaria vector control pesticides in the region. This increase did however show a negative correlation with rainfall. A 2 year study is not enough to draw conclusions regarding such a correlation, but is important for future investigations as drought has been a major issue in the region since 2015. It was shown that DDT still enters the aquatic environment and readily accumulates in Xenopus muelleri, however DDT accumulation was not seen to relate to rainfall. The DDT levels in the frogs also did not follow a distinct pattern relating to MVC spraying seasons in the area. There is also a strong possibility of biochemical responses towards the presence of these pesticides at sub-lethal concentrations in anurans. The results of this study show that not only the concentrations, but also the specific composition of these pesticide mixtures in the environment should be taken into account in assessing possible biochemical effects they may have on aquatic organisms such as frogs. The biochemical response effects of OCPs in anurans warrants further investigation before any attributions can be made. There is also a need for toxicity testing of these compounds on African anuran species with regards to body burdens and mixture effects. With such data available the use of biomarker responses in biomonitoring studies could prove useful as early warning indications for preventative management of chemical exposure.