Per- and Polyfluoroalkyl Substances (PFAS) in the Rusizi River System, Burundi: A Multi-Compartment Assessment from Tributaries to Lake Tanganyika

TL;DR

This study assesses PFAS pollution in the Rusizi River and Lake Tanganyika, finding low levels in water and sediment but higher diversity in fish, with potential health risks for some groups.

Contribution

First multi-compartment PFAS assessment in a tropical freshwater system, highlighting biomonitors and human health risks.

Findings

PFOA was detected in water with uniform spatial distribution, while sediment concentrations were mostly below detection.

Macrophytes showed dominance of short-chain PFAS like PFBS, suggesting potential as biomonitors.

Fish exhibited high PFAS diversity, with PFBS as the dominant compound and emerging PFAS frequently detected.

Abstract

Per- and polyfluoroalkyl substances (PFAS) are global pollutants, yet data from tropical freshwater ecosystems remain scarce. This study provides the first assessment of PFAS occurrence in the Rusizi delta (Burundi), from tributaries to Lake Tanganyika, by analyzing water, sediment, macrophytes, and fish, and by evaluating human health risks from fish consumption. In water, only PFOA (<0.60–7.80 ng/L) was detected and showed a uniform spatial distribution. Sediment concentrations were largely below quantification limits, likely reflecting unfavorable sorption conditions. Macrophytes were dominated by short-chain PFAS, particularly PFBS, without consistent species- or site-specific patterns, supporting their potential as biomonitors of cumulative PFAS exposure. Fish exhibited the highest PFAS diversity, with more diverse profiles in liver than muscle, although tissue-specific patterns were often absent. PFBS was dominant across fish species, and emerging PFAS (e.g., PFBS and NaDONA) were frequently detected. Human health risks from fish consumption were, except for children, mostly below EFSA tolerable weekly intake values for regulated PFAS, but potential concern for adolescents and adults emerged when PFAS were expressed as PFOA equivalents. This study provides essential baseline data for tropical freshwater systems and highlights the need for expanded PFAS monitoring and risk assessment in data-poor regions.

Figures

Full text

1. Introduction

A large variety of chemical pollutants have been discharged and emitted into the environment following global anthropogenic activities [1]. One group of such pollutants is per- and polyfluoroalkyl substances (PFAS), which are synthetic chemicals containing one or more fluorinated methylene or methyl groups [2]. The strong carbon-fluorine bond imparts unique properties including lipo- and hydrophobicity, and providing PFAS with high thermal and chemical stability, making them suitable for a large number of applications [3]. Following their production and use in consumer products, many PFAS were discharged or emitted into the environment, where their high mobility facilitated their global distribution across environmental compartments and biota, in which PFAS may persist for long periods [4,5,6]. The continuous exposure of organisms to PFAS may lead to their bioaccumulation and even biomagnification in food webs [7,8,9].

The fate of PFAS in aquatic ecosystems is determined by partitioning processes between the abiotic and biotic compartments [10]. The length of the carbon-chain determines their mobility, with PFAS becoming less mobile and more prone to bioaccumulation with increasing carbon-chain length [11,12]. Their persistence and ability to partition across compartments facilitate their widespread contamination and chronic exposure of aquatic organisms, which could lead to adverse effects on development, emergence rate, survival, growth, reproduction, and other life-history traits [13]. As their fate depends on the physicochemical characteristics of the PFAS molecules [12], no single matrix fully reflects their distribution or risks. Therefore, multi-compartment monitoring is essential to capture the complexity of PFAS behavior in aquatic ecosystems.

Despite growing global concern about PFAS contamination, tropical river-lake systems remain underrepresented in current research, particularly in Africa [14]. These ecosystems support biodiversity, fisheries, and drinking water sources, and provide many other ecosystem services, and are therefore of ecological and socio-economic importance [15,16]. Recent studies reported that tropical wetlands can remove large proportions of nutrients and organic pollutants: for example, the Lubigi wetland in Kampala retained ~50–60% of orthophosphate and ~20–40% of ammonium nitrogen despite being subject to both stormwater and sewage overflows [17]. Natural riverine wetlands in Ethiopia have demonstrated reductions on the order of 70–80% in biochemical oxygen demand, total phosphorus, and dissolved inorganic nitrogen, while also supporting higher biodiversity downstream [18]. However, they may be vulnerable to pollution due to increasing urbanization and industrialization. In Burundi, as in many other African countries, limited wastewater treatment, rapid urbanization, and industrial growth further increase the risk of PFAS contamination in aquatic ecosystems [19,20]. Untreated industrial and municipal wastewater, alongside emissions from industries such as the textile industry and waste disposal, represent potential sources of PFAS contamination. As local communities rely heavily on these ecosystems for essential services, these pressures highlight the need to investigate PFAS occurrence and fate in such systems, where monitoring data remain scarce. This is critical to assess potential risks and safeguard ecosystem services.

Despite these pressures, knowledge of PFAS distribution in African freshwater ecosystems remains very limited and scattered across the continent [14]. Existing studies in Sub-Saharan Africa are often restricted to single compartments [21,22,23,24,25,26,27,28,29], which do not provide a full overview of environmental exposure and risks. Multi-compartment approaches have been applied sporadically in Nigeria [30,31], Uganda [32], South Africa [33,34,35,36,37], Tanzania [38], and Ethiopia [39,40], but data remain scarce.

This study investigated, for the first time, PFAS occurrence and distribution in a river-lake system in Burundi, with the objectives to assess their presence across multiple environmental and biological compartments, identify spatial patterns within the system, and evaluate potential human health risks through fish consumption. By generating baseline data for the region, this work contributes novel insights into an underrepresented region and continent, and advances the global understanding of PFAS behavior in tropical freshwater ecosystems.

We hypothesize that PFAS concentrations and profiles will differ among compartments due to their physicochemical properties and resulting partitioning. Specifically, we expect short-chain PFAS to dominate in water and plants, because of their higher mobility and solubility, whereas long-chain PFAS will dominate sediments and fish due to their stronger sorption and bioaccumulation potential. Among biota, fish are anticipated to contain the highest PFAS concentrations due to biomagnification. Despite these patterns, human health risks are expected to be low, reflecting generally low PFAS concentrations in African ecosystems [14] and the absence of known major sources in the study region.

2. Materials and Methods

2.1. Study Area

The Rusizi National Park (Figure 1) is one of the largest protected areas in Burundi, covering about 10,673 hectares north of Lake Tanganyika. The park takes its name from the Rusizi River, which flows across the Rusizi Plain and carries waters from Lake Kivu. At an altitude of around 1000 m, the Rusizi River meanders through a broad plain between Burundi and the Democratic Republic of the Congo (DRC). These meanders give rise to swamps and lagoons that are characteristic of the region [41]. The river is fed by several important tributaries, including the Ruhwa, Nyakagunda, Nyamagana, Muhira, Kaburantwa, Kagunuzi, Kajeke, and Mpanda Rivers.

2.2. Sampling

In April 2025, surface water, sediment, above-ground vegetative parts of plants (Phragmites mauritianus Kunth, Typha domingensis Pers., and Vossia cuspidata (Roxb.) Griffith) (N = 5 per species; all mature plants, determined by presence of flowers) and fish (Auchenoglanis occidentalis, Oreochromis niloticus, and Protopterus aethiopicus) (N = 5 per species) were collected from the Rusizi River delta in Burundi. Sampling was carried out in the Rusizi National Park in collaboration with staff from the Burundian office for the protection of the environment (OBPE). The sampling authorization was granted at the launch of the project.

All three plant species are highly productive, rhizomatous wetland grasses common in tropical and warm regions. They are the dominant plant species among Rusizi delta vegetation. They can exhibit invasive behavior due to their rapid growth, wide environmental tolerance, and efficient vegetative reproduction. Nonetheless, they all play important roles by stabilizing soils, protecting banks, providing habitat for wetland birds and fish, storing carbon, and improving water quality through accumulation and filtration of nutrients and pollutants [42,43,44,45].

P. aethiopicus is a benthic-dwelling carnivorous-to-omnivorous fish that is ubiquitously present in central and east Africa. Juveniles feed primarily on insects, whereas adults shift their behaviour to molluscs, fish, insects, and occasionally plants [46]. O. niloticus is native to central and north Africa but has been introduced to many aquatic ecosystems due to its suitability for aquaculture [47]. They occupy a flexible and omnivorous dietary niche shifting from zooplankton and insects as juveniles to a primarily plant-based (phytoplankton, macrophytes, detritus) as adults. Their dietary range is site-specific and may vary among seasons [47]. A. occidentalis is present in many shallow lakes and large rivers across Africa. It occupies a predominantly omnivorous dietary niche and is an adaptive generalist feeder with strong insectivorous preference. However, as is also true for the other species, its diet and feeding habit can vary substantially depending on local food availability [48].

Nine sampling sites along the Rusizi River, its tributaries, and Lake Tanganyika were selected for the abiotic samples (Figure 1; Table S1). Plants were collected by hand (±15 cm from the topsoil) within the Rusizi delta wetland and in the Bujumbura region, whereas fish were caught in Lake Tanganyika by local fishermen from the town Gatumba.

Water was sampled from a boat (lake samples) or from a bridge (river samples). At each site, triplicate unfiltered water samples were collected and stored at −18 °C in 50 mL polypropylene (PP) tubes (Greiner Bio One, Vilvoorde, Belgium), which are typically used in PFAS studies on water due to limited adsorption of PFAS compared to some other types of containers [49,50,51], for PFAS analysis. A fourth water sample was filtered (0.45 µm nitrocellulose Chromafil syringe filters (A-45/25)) (Macherey-Nagel GmbH, Düren, Germany) into new bottles. After determining multiple physicochemical characteristics in situ (see Section 2.3), the water was acidified with 20 µL of hydrochloric acid (HCl; target pH < 2), and stored at 4 °C for subsequent physicochemical analyses (see Section 2.3).

Sediment was sampled from a boat (lake samples) or from a bridge (river samples) using a Van Veen grab sampler (Eijkelkamp Royal, Giesbeek, The Netherlands) up to a depth of approximately 10 cm. At each site, three replicate samples were collected (except at site 2, where two samples were lost) and stored at 5 °C in 50 mL tubes for grain size and organic matter analyses. Afterwards, they were stored at −18 °C for PFAS analysis.

After removal of flowers and rinsing with tap water, the plant material (i.e., stem and leaves combined) was oven-dried at 70 °C and subsequently ground mechanically to a fine powder. Fish were euthanized by freezing (24 h). After thawing, fish were dissected and their muscle and liver tissues were stored in 3 mL Eppendorf tubes (Eppendorf Belgium N.V.-S.A., Aarschot, Belgium) at −18 °C for PFAS analysis.

2.3. Determination of Water Physicochemical Characteristics

In situ measurements of temperature (°C), electrical conductivity (µS/cm), salinity (mg/L), and oxygen (mg/L and %) were taken using a WTW MultiLine^®^ Multi 3620 IDS SET KS1 portable meter (Fisher Scientific, Merelbeke-Melle, Belgium). A HI98713 ISO Portable Turbidity meter (Hanna Instruments B.V., Temse, Belgium) was used to determine the turbidity (FNU).

Concentrations of total dissolved nitrogen (NH_4_^+^ + NO_3_^−^ + NO_2_^−^), phosphate (PO_4_^3−^), dissolved organic carbon (DOC), and dissolved silicon (DSi) were determined with a SAN++ segmented flow analyzer (Skalar Holding B.V., Breda, The Netherlands) [52]. Concentrations of suspended particulate matter (SPM, mg/L) were quantified through filtration (nitrocellulose 0.45 µm; Porafil membrane filter, Macherey-Nagel, Düren, Germany) of water samples [53]. Filters were pre-weighed, oven-dried for 72 h at 70 °C, and reweighed. They were subsequently analyzed for biogenic Si concentration (expressed as mg/L water and mg/g SPM) using alkaline extraction in 0.5 M NaOH [54]. Results are presented in Table S2.

2.4. Sediment Physicochemical Characteristics

2.4.1. Grain Size Distribution

Sediment grain size distribution was determined using a laser diffraction protocol previously described [55], which reflects our laboratory’s standard procedure. Approximately 1 g of homogenized sediment was added to a glass flask, followed by 10 mL HCl and 15 mL hydrogen peroxide (H_2_O_2_) to remove organic compounds and disperse aggregates during overnight incubation. Subsequently, 25 mL of H_2_O_2_ was transferred into the flask, which was placed on a hot plate to boil the samples and enhance the reaction. Following cooling down to room temperature and sieving (1 mm), the samples were analyzed using a Mastersizer 2000 and Hydro 2000 G (Malvern Panalytical B.V., Brussels, Belgium) unit to determine the proportions of sand (>63 μm), silt (<2–63 μm), and clay (<2 μm) (Table S3).

2.4.2. Organic Carbon Content (Corg)

The loss on ignition method [56] was used to determined organic carbon contents. Oven-dried (105 °C, 2 h) and pre-weighed samples were incinerated (550 °C, 5 h) and weighed. The organic carbon content was then calculated through the difference between both measurements.

2.5. PFAS Extraction

Macrophytes and sediment samples were oven-dried (60 °C), and fish tissues were crushed using a Tissuelyser LT and stainless-steel beads (Qiagen, Hilden, Germany). The water samples were extracted without any pretreatment.

Abiotic and biotic samples were extracted using different protocols, both used as standard protocols in our laboratory, described by Groffen et al. [57] and Powley et al. [58], respectively. The initial steps are the same for both protocols. After weighing the samples (337 ± 110 mg of dried sediment, 237 ± 61 mg of dried plants, 199 ± 67 mg of liver, 237 ± 93 mg of muscle, and 10 mL of water), 10 ng of internal standard (ISTD, MPFAC-MXA, Wellington Laboratories, Guelph, ON, Canada) was added, followed by 10 mL of HPLC gradient grade acetonitrile (ACN; Acros Organics BVBA, Geel, Belgium). No ACN was added to the water samples. Subsequently, the samples were mixed, sonicated in an ultrasonic bath (3 × 10 min; Branson 2510, VWR International, Leuven, Belgium) and shaken overnight (GFL 3020, VWR International, Leuven, Belgium) at 135 rpm and room temperature. After centrifugation (4 °C, 10 min, 1037× g; Eppendorf centrifuge 5804R, Eppendorf Belgium N.V.-S.A., Aarschot, Belgium) the biotic and abiotic extracts were treated differently, with abiotic extracts being cleaned up using weak anion exchange solid phase extraction [57] and biota samples being cleaned up using graphitized carbon powder [58].

Abiotic extracts were loaded onto preconditioned (5 mL of ACN and 5 mL of Milli-Q) Chromabond HR-XAW SPE cartridges (Macherey-Nagel, Düren, Germany). The cartridges were rinsed with 5 mL of a 25 mM ammonium acetate solution (dissolved in Milli-Q) and 2 mL of ACN, and PFAS were eluted using 2 × 1 mL of a 2% ammonium hydroxide solution (Thermo Scientific, Merelbeke-Melle, Belgium; dissolved in ACN). The eluent was dried completely under vacuum [57].

Biotic extracts were dried to approximately 0.5 mL under vacuum (30 °C, Martin Christ RVC 2-25) and transferred to Eppendorf tubes containing 0.1 g of graphitized carbon powder (Supelclean ENVI-Carb, Sigma-Aldrich, Belgium) soaked for at least 15 min in 50 μL of glacial acetic acid (Fisher Scientific, Belgium). The tubes that contained the original samples were flushed twice with 250 μL of ACN, which was also added to the carbon powder. Following vortex mixing (1 min) and centrifugation (4 °C, 10 min, 9279.4× g; Eppendorf centrifuge 5415R, Eppendorf Belgium N.V.-S.A., Aarschot, Belgium), supernatants were dried completely under vacuum [58].

The samples were reconstituted with 200 μL of a 2% ammonium hydroxide solution (diluted in ACN), mixed, and filtered (Ion Chromatography 13 mm syringe filter with 0.2 μm Supor polyethersulfone membrane; VWR International, Leuven, Belgium) into PP auto-injector vials [57,58]. Procedural blanks (10 mL of Milli-Q for water, 10 mL of ACN for the other matrices) followed the same procedures.

2.6. Instrumental Analysis (UPLC-MS/MS)

PFAS were analyzed using ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS, ACQUITY TQD, Waters, Milford, MA, USA) operated in negative electrospray ionization mode. Separation was achieved on an ACQUITY BEH C18 column (2.1 × 50 mm; 1.7 μm, Waters, Milford, MA, USA), preceded by an ACQUITY BEH C18 pre-column (2.1 × 30 mm; 1.7 μm, Waters, Milford, MA, USA), inserted between the solvent mixer and injector, to minimize system-derived PFAS contamination. The mobile phase consisted of solvent A (0.1% formic acid in water (HPLC grade, VWR International, Leuven, Belgium)) and solvent B (0.1% formic acid in ACN (LC/MS grade, Fisher Chemical, Merelbeke-Melle, Belgium)). A flow rate of 450 μL/min and an injection volume of 6 μL were applied. The solvent gradient started at 65% A, went to 100% B in 3.4 min, and back to 65% A at 4.7 min. Quantification and confirmation of the 29 target PFAS were performed using multiple reaction monitoring (MRM) (Table S4 [57,59]). Detailed MS/MS conditions (cone voltages, collision energies, and ISTD used for quantification) are provided in Table S4. Retention time (RT) was used as criterion for compound confirmation, with a tolerance of 0.1 min. Ion ratios were not applied for confirmation in this study, as unextracted standards were not analyzed to allow comparison between sample and standard ion ratios.

Quality assurance procedures included the analysis of one procedural blank for every set of 15 samples to evaluate potential contamination introduced during extraction or measurement. None of the target PFAS were detected in these blanks. Instrumental blanks consisted of 100% ACN and were run regularly to minimize carry-over between injections. No field or trip blanks were collected. Therefore, potential contamination during sampling cannot be completely excluded. Limits of quantification (LOQ) for each compound were determined in the sample matrix as the lowest concentration producing a signal-to-noise ratio of 10 and are summarized in Table S5. Calibration was performed using isotopically labelled internal standards as described in Groffen et al. [57], with the response expressed as the log of the analyte-to-ISTD area ratio versus the log of the analyte-to-ISTD concentration ratio. Linearity was verified in the current study for all target PFAS (R^2^ > 0.98). Recoveries of the isotopically labelled internal standards were within an acceptable range of 84% (PFHxS) to 99% (PFOS). Recoveries of the native analytes were not directly determined. Compound identity was confirmed based on retention time matching to internal standards (where available), or compared to unextracted native standards. Ion ratio acceptance criteria were not applied, but blanks and retention time matching provided sufficient confidence in compound identification.

2.7. Human Health Risks Associated with Consumption of PFAS-Contaminated Fish

The tolerable weekly intake (TWI) established by the European Food Safety Authority (EFSA) was used to calculate the maximum edible amount (MEA) of fish (Equation (1)) that can be consumed per person per week, without posing a health risk. This TWI is expressed for the intake of the sum of four PFAS (PFOA, PFNA, PFHxS, and PFOS) and equals 4.4 ng/kg bw/week [60].

$$[eqn]$$

where MEA = maximum edible amount of fish muscle (kg/week; fresh weight), TWI = tolerable weekly intake of PFAS (4.4 ng/kg bw/week), BW = average body weight, and C = PFAS concentration in fish muscle (ng/kg).

These calculations were based on a per capita national average fish consumption of 0.04 kg/week (data from 2019) [61,62]. This value is considerably lower than the reported African average of 0.18 kg/week and may underestimate exposure in the study area, as fisheries from Lake Tanganyika contribute 25–40% of the local population’s protein intake [62], suggesting higher fish consumption than the national average. However, no specific data were available for local communities along Lake Tanganyika.

To address these uncertainties, a scenario-based approach was applied. Since data on per capita minimum and maximum fish consumption are absent, a lower- and upper-bound approach was applied by dividing or multiplying the average fish consumption of 0.04 kg/week by a factor of 3. Body weight was varied continuously from 0 to 100 kg to cover a broad range of potential consumer profiles. For each of the three consumption scenarios (i.e., lower-bound, average scenario, and upper-bound), maximum edible amounts (MEA, g/week) were calculated for each fish species.

To account for the combined toxicity of multiple PFAS, concentrations were also converted into PFOA-equivalents (PEQ) using relative potency factors (RPF) derived from liver toxicity data [63]. Concentrations in PEQ were then applied in Equation (1) to determine the MEA. This approach allows the evaluation of mixture toxicity by expressing PFAS effects relative to PFOA. However, results must be interpreted with caution, since the EFSA TWI is based on impaired immune response in humans [60], while the RPFs are based on liver toxicity in rats [63]. Therefore, the comparison between EFSA TWI and PEQ-based risk estimates should be regarded as exploratory rather than definitive.

All calculations were performed under two scenarios: (1) using average concentrations in fish tissue, with values below the LOQ imputed by maximum likelihood estimation [63], and (2) a scenario using maximum concentrations (corresponding to a worst-case scenario), with values below the LOQ replaced by the LOQ itself. FBSA, 11Cl-PF3OUdS, 9Cl-PF3ONS, PFEESA, PF4OPeA, PF5OHxA, and 3,6-OPFHpA were omitted from these calculations because of the absence of RPFs for these PFAS.

2.8. Statistical Analysis

All statistical analyses were carried out in R Studio (version 2024.12.1+563) using R version 4.3.3. A significance threshold of p ≤ 0.05 was applied. Table S6 reports the frequency of detection of the targeted PFAS in each of the matrices. Prior to analysis, non-detects were replaced by estimated values using the maximum likelihood estimation method described by Villanueva [64]. Unlike simple substitution approaches such as LOQ/2, which assign arbitrary fixed values to censored data, MLE estimates censored values based on the statistical distribution of detected measurements, providing unbiased estimates of mean, variance, and other statistics. This approach is considered more robust and appropriate for datasets with censored values [64].

For PFAS compositional profiles only, non-detects were assigned a value of zero. These profiles were calculated as the percentage contribution of each individual compound to their sum concentration, corrected for molecular weight, and subsequently averaged across replicate samples within each matrix and/or species. Substitution methods, including MLE, rely on the LOQ of each compound and can artificially inflate the apparent contribution of compounds with high LOQs. Furthermore, substituting PFAS that were never detected in any of the samples would artificially introduce them into the compositional profiles, suggesting contributions where there was no evidence of their presence. To assess the sensitivity of this approach, the compositional profiles of the macrophytes, determined with values below the LOQ substituted by zero, were compared to profiles with values below the LOQ substituted by the MLE method. In both cases, PFAS that were never detected were omitted. Both profiles (Figure S1 shows them side by side) exhibit a high degree of resemblance, confirming that the main conclusions of the compositional profiles based on substitution by zero are robust.

Differences in PFAS concentrations among sampling locations or species were assessed using ANOVA, with post hoc comparisons performed using Tukey’s honest significant difference test. In case of non-normality, non-parametric alternatives were used. Spearman rank correlation analysis was used to examine relationships between PFAS concentrations in liver and muscle, as well as between accumulated levels in the fish tissues and each fish’s length and weight. Only compounds detected in at least 30% of the samples were included, either in both tissues when correlating tissues, or in the respective tissue when correlating with fish size and weight.

The high proportion of non-detects in water and sediment (see Section 3.1) prevented us from correlating water and sediment physicochemical properties with PFAS concentrations in the abiotic environment.

3. Results

3.1. Spatial Distribution and Concentrations in the Abiotic Environment

PFOA was the sole compound detected in water (Figure 2; Table S2 for descriptive statistics (i.e., median, minimum, and maximum concentrations)). It was present in all water samples, except at site 9, where it was detected in two out of three samples. No significant differences in PFOA concentrations were observed among the sampling sites (p = 0.315).

Some sampling sites showed high variability among parallel samples, reflecting heterogeneity in PFAS distribution in both water and sediment. All replicates were retained and no analytical errors were detected.

In sediment, PFOA was detected in only one of three samples from sites 4, 5, and 7, with average concentrations remaining below the LOQ. PFDoDA was also detected at site 2 (0.514 ng/g dw; N = 1) and in one of three samples from sites 7 and 9, though average concentrations at these sites were also <LOQ. Median, minimum, and maximum concentrations are provided in Table S3. The large amount of non-detects in sediment limits statistical interpretation but indicates generally low environmental PFAS burdens in sediments.

3.2. Accumulation in Plants

Although plants were collected from different sampling sites, the descriptive statistics are listed in Table S7 irrespective of the sampling site. PFBA, PFPeA, PFOA, PFDoDA, PFBS, PFOS, 4:2 FTS, and 6:2 FTS were each detected in at least one plant sample, with PFBS showing the highest concentrations (Figure 3A). While detection frequencies—and consequently PFAS accumulation profiles (Figure 3B)—varied notably between plant species, Kruskal–Wallis tests (and ANOVA for PFBS) indicated only a significantly higher PFPeA concentration in T. domingensis compared to the other plant species (p = 0.009).

Table S8 differentiates between the Rusizi delta wetland (sites C–E; Figure 1) and the Bujumbura region (sites A and B; Figure 1). A descriptive comparison of mean concentrations indicates slightly higher concentrations of PFPeA in T. domingensis, PFBS in P. mauritianus and T. domingensis, and 4:2 FTS concentrations in P. mauritianus and V. cuspidata from the Bujumbura region. However, higher PFBS concentrations in V. cuspidata, higher PFOS concentrations in P. mauritianus, and higher 4:2 FTS concentrations in T. domingensis were observed in the wetland.

3.3. Accumulation in Fish

Compared to plants, a greater variety of PFAS was detected in fish tissues, with 12 compounds found in muscle and 14 in liver (Figure 4; Table S9 lists descriptive statistics). PFNA and PFTrDA were detected exclusively in liver. PFHpA concentrations were significantly higher in the livers of O. niloticus than in other species (p < 0.05; Figure 4C). The livers of O. niloticus also contained higher NaDONA concentrations than those of P. aethiopicus (p = 0.031; Figure 4C), with a similar trend relative to A. occidentalis (p = 0.080; Figure 4C). Finally, a trend was also observed for PFOS, with higher concentrations in the livers of P. aethiopicus compared to A. occidentalis (p = 0.096; Figure 4C). No other interspecific differences were significant.

For tissue-specific comparisons, the only significant differences were higher PFHpA concentrations in the liver than in the muscle of O. niloticus (p = 0.044), and higher PFOS concentrations in the liver than in the muscle of P. aethiopicus (p = 0.012). No other significant differences between liver and muscle were observed. PFAS concentrations were not correlated between muscle and liver (p > 0.05; Table S10). The length and weight (Table S11) of the fish were negatively correlated (p < 0.05; Table S12) with concentrations of PFHxA, and 6:2 FTS in the liver, as well as PFOS and NaDONA in the muscle.

The composition profiles for muscle (Figure 4B) and liver (Figure 4D) were consistently dominated by PFBS across all species. Although minor interspecific differences were observed, the overall profiles were highly similar both among species and between the two tissue types.

3.4. Human Health Risks Associated with Consumption of PFAS-Contaminated Fish

Maximum edible amounts (MEA) of fish muscle tissue were calculated for three fish species from Lake Tanganyika, considering both the average and maximum PFAS concentration to account for typical and worst-case exposure scenarios. The MEA was assessed using two approaches: (1) the EFSA TWI for four PFAS (i.e., PFOA, PFNA, PFHxS, and PFOS), and (2) an RPF method to express most PFAS as PEQ.

Results (Figure 5) indicate that particularly young children are at risk through fish consumption. Risk patterns differed substantially between the EFSA- and PEQ-based approaches. Under the EFSA-based approaches, exceedances of health-based guidance values occurred primarily at low body weights and was therefore most relevant for children, largely independent of the assumed weekly fish consumption rate. Using the upper-bound consumption scenario of 120 g/week, risks are expected for individuals weighing approximately 20 kg or less based on average PFAS concentrations, and for individuals weighing 39 kg or less based on maximum PFAS concentrations (Figure 5).

In contrast, the PEQ-based approaches showed substantially more risks for heavier individuals. Under both average and worst-case PEQ scenarios, risks were expected for all individuals across the evaluated body weight range (0–100 kg) when applying the national average or upper-bound consumption scenarios. One exception was observed for A. occidentalis under the PEQ average scenario, for which a risk was only expected for individuals weighing 63 kg or less. Even under lower-bound approaches, PEQ-based results indicated several risks (Figure 5).

Among the species, the lowest MEA values were observed for O. niloticus and the highest for A. occidentalis. The reduced MEA based on the PEQ approach is primarily driven by 6:2 FTS, PFDA, and PFDoDA (Figure S2), which are all not included in the EFSA TWI, highlighting additional risk contributions from these compounds. Additionally, the worst-case scenario is substantially influenced by some PFAS that have relatively high LOQ values compared to others (e.g., PFHxS in the EFSA method, and 8:2 FTS in the PEQ method).

4. Discussion

4.1. PFAS Detection and Concentrations in the Abiotic Environment

In water samples, only PFOA was detected, showing a relatively uniform distribution across the study area. This pattern suggests the absence of distinct point sources and indicates more diffuse sources of PFOA. Potential contributors include the atmospheric transport and deposition, as well as the degradation of PFOA and its precursors [65,66]. Furthermore, the lack of significant differences among sites indicates that the presence of a wetland between sites 1 and 5 does not appear to influence PFOA concentrations compared with a similar riverine transect between sites 3 and 4. This suggests limited localized retention or removal of PFOA in the studied hydrological contexts. However, given the snapshot sampling design and high proportion of non-detects, any conclusions regarding wetland filtering capacity remain uncertain and require further investigation. Although wetlands have the potential to mitigate PFOA concentrations through increased sorption to sediment and bioaccumulation in wetland plants [55,67,68,69], their efficiency depends on specific environmental conditions and wetland characteristics [55,67,68]. It is, however, important to note that PFOA has one of the lowest LOQs among PFAS in water, meaning that its detection could in part reflect analytical sensitivity.

The PFOA concentrations detected in the Rusizi River system surface waters (average range 2.15–4.16 ng/L) fall within the lower range of values reported across Africa. For example, PFOA concentrations reached up to 11.7 ng/L in Lake Victoria [27] and spanned between 1.78 ng/L and 321 ng/L in Ghanaian rivers [23]. In South Africa, reported values include 12.8–62.6 ng/L in the Plankenburg River [33], 0.6–4.6 ng/L in the Vaal River [35], 1.7–314 ng/L in the Diep River [36], 0.7–390 ng/L in the Salt River [36], and 3.4–146 ng/L in the Eerste River [36]. Estuarine sites in South Africa contained higher levels of PFOA, including 416–1089 ng/L in the uMvoti estuary and 142–310 ng/L in the aMatikulu estuary [34]. Globally, surface water PFOA concentrations vary by several orders of magnitude depending on local sources of contamination. For instance, a large-scale study across the northern hemisphere reported PFOA concentrations ranging from 0.15 to 52.8 ng/L in aquatic ecosystems of China, the United States, the United Kingdom, Germany, the Netherlands, South Korea, and Sweden [70]. These comparisons indicate that PFOA concentrations in the Rusizi River system are relatively low and do not suggest the influence of a major point source.

Our initial hypothesis that long-chain PFAS would dominate sediments was only partially supported. Although some long-chain PFAS (e.g., PFOA and PFDoDA) were detected, most sediment samples showed a predominance of non-detects, which were distributed relatively uniformly across the study area and did not correspond to specific sampling locations. While long-chain PFAS are known to preferentially sorb to organic matter and fine mineral fractions, the consistently low clay content and variable, often low, organic carbon contents across all sampling sites likely limit the formation of organo-mineral complexes that typically retain long-chain PFAS [71,72,73,74] Consequently, both electrostatic interactions with mineral surfaces and hydrophobic interactions with organo-mineral complexes are restricted, leading to lower PFAS sorption to sediments [72,73,74]. The low adsorption potential may favor PFAS partitioning to the water column, where concentrations may be reduced by dilution or transport, which could, together with the analytical sensitivity in water, explain why they were not detected in water. This might explain why PFDoDA, a hydrophobic PFAS that strongly partitions to organic carbon and mineral surfaces and has a much higher LOQ in water than in sediment (Table S5), was only detected in a few sediment samples while remaining <LOQ in water.

Additionally, low adsorption to sediments could also promote leaching to deeper sediment layers. However, as our study relied on snapshot sampling of both water and sediment, we cannot draw conclusions about temporal fluxes or partitioning. Future work should include temporal sampling, sediment core sampling to assess vertical distribution of PFAS, and more detailed characterization of organic carbon composition and mineralogy to better understand PFAS retention mechanisms.

4.2. PFAS Accumulation in Macrophytes

A range of PFAS were detected in all macrophyte species, with PFBS reaching the highest concentrations. The dominance of short-chain PFAS such as PFBS, PFBA, and PFPeA is consistent with their higher solubility and mobility, which facilitates their passage through root cell walls and membranes into the xylem via water transpiration [75]. In contrast, long-chain PFAS are more strongly adsorbed to sediments, restricting their movement mainly to the root rhizosphere. This enhanced sorption to sediments generally reduces their bioavailability to plants [76]. However, as discussed in Section 4.1, sediment concentrations were low and the sediment properties were not favorable for PFAS sorption, suggesting that even long-chain PFAS may remain bioavailable to plants in the study area.

The overall absence of species-specific differences suggests that PFAS uptake is driven primarily by PFAS properties and environmental availability rather than plant traits. Nonetheless, minor differences in morphology or physiology (e.g., protein and lipid content) may still influence the compositional profiles [75]. Regional comparison between the Bujumbura region and the Rusizi delta wetland showed minor, but inconsistent, differences between regions. Although they may reflect distinct contamination sources, with urban inputs affecting the plants at Bujumbura, the limited sample sizes precluded robust statistical analysis, and the results should therefore be interpreted with caution.

Literature on PFAS bioaccumulation in macrophytes in Africa is scarce and shows considerable variation depending on region and sources. In South Africa, PFOA concentrations in riparian wetland plants ranged from 11.7 to 38 ng/g dw [37]. In Uganda’s Nakivubo wetland, sum PFAS (26 analytes) concentrations were reported at 0.36 ng/g dw in yam root, 0.35 ng/g dw in sugarcane stem, and 0.20 ng/g dw in maize cobs [32]. Similar to surface waters, PFAS concentrations in macrophytes from the present study (on average <LOQ—3.50 ng/g dw) are relatively low compared to those in the northern hemisphere, where much higher concentrations have been documented, especially near hotspots. For instance, macrophytes collected around a contaminated site in Belgium contained PFAS ranging from 0.133 ng/g dw to 3411 ng/g dw depending on the plant species and type of PFAS [77].

Overall, the consistent detection of PFAS across species and sites, even where environmental concentrations were relatively low, demonstrates the use of macrophytes as biomonitors. Macrophytes provide insights into cumulative contamination that may not be apparent from snapshot environmental monitoring. Although hypothetical and requiring further investigation, wetlands may act as natural filters that reduce PFAS concentrations in surface water and pore water, thereby lowering bioavailability and uptake by fish in Lake Tanganyika and ultimately reducing potential human health risks through fish consumption.

4.3. PFAS Accumulation in Fish

The greatest diversity of PFAS was found in fish, followed by plants and then the environmental samples, confirming the role of fish as important bioaccumulators of PFAS within aquatic food webs. This pattern shows that fish integrate exposure from multiple sources and can store both legacy and emerging PFAS in their tissues. Within fish, the liver contained a greater diversity of PFAS than muscle, possibly due to its role in xenobiotic metabolism and storage. Multiple studies reported that liver is one of the primary tissues for PFAS accumulation [35,78,79,80,81]. PFAS, particularly long-chain compounds, are proteinophilic and show a higher affinity to liver fatty acid-binding proteins and albumin [82], which may explain their greater diversity in the liver. Nonetheless, this apparent preference was not reflected in the concentrations and compositional profiles, as only few significant tissue-specific differences were observed, and liver and muscle concentrations were often uncorrelated. This suggests a relatively uniform PFAS distribution between liver and muscle, which may reflect homogeneity in environmental exposure.

In contrast to our initial hypothesis that long-chain PFAS would dominate compositional profiles in fish, PFBS was the dominant PFAS in both fish tissues. Although long-chain PFAS generally dominate fish PFAS profiles [35,40,77,83], short-chain PFAS such as PFBS can become dominant under chronic exposure conditions when uptake rates exceed elimination rates [84].

Bioaccumulation data for PFBS are limited and highly variable, with bioaccumulation factors ranging up to 1736 L/kg for freshwater fish [85]. The increasing detection of short-chain PFAS, such as PFBS, is consistent with the global shift from legacy long-chain PFAS toward short-chain alternatives. The dominance of PFBS in the present study may therefore be the result of both its more recent use and its high environmental mobility, suggesting continuous environmental input and exposure. Although short-chain PFAS are generally considered less bioaccumulative [79,86], the persistence, mobility, and high detection frequency of PFBS highlight the potential for chronic risks to aquatic ecosystems and human consumers of fish, underscoring the importance of including short-chain PFAS in risk assessments.

Species-specific differences in PFAS concentrations were limited, with broadly similar compositional profiles observed across species. The minor differences observed may be related to interspecific differences in feeding ecology, habitat use, or physiology. Due to trophic transfer, carnivorous species typically exhibit higher PFAS concentrations than omnivores [87]. Contrary to expectations, our results indicated higher concentrations of NaDONA and PFHpA in O. niloticus, which is, like A. occidentalis, an omnivorous species, whereas P. aethiopicus is carnivorous. However, O. niloticus is known for its dietary plasticity, and a more carnivorous diet has been reported in juveniles [47,88]. Although speculative, it could be possible that the specimen used in the present study also exhibited a carnivorous diet. In terms of habitat use, O. niloticus is primarily pelagic, whereas the other species are more benthic. Regardless of the trophic position, benthic species typically exhibit higher PFAS concentrations due to sediment exposure [89,90,91], although lower levels have also been reported [92]. Differences between benthic and pelagic species in the present study may be related to water-sediment partitioning and the low detection of PFAS in sediments in the Rusizi River system may help explain why PFAS concentrations were typically lower in the benthic species. Nonetheless, the broadly similar PFAS profiles across species suggest that local exposure conditions are the dominant drivers of bioaccumulation in the Rusizi River system.

Finally, negative correlations between some PFAS concentrations and fish length or weight were observed, suggesting possible growth dilution effects or differences in life history traits (e.g., age or sex). Larger individuals may have reduced concentrations due to growth dilution [93], or alternatively, they may occupy ecological niches with lower PFAS exposure. For example, as was mentioned previously, the diet of O. niloticus may shift during maturation from carnivorous to omnivorous [47,88], which may affect their exposure and thus accumulation. Furthermore, size- or age-related differences in metabolism are known to occur in fish [94,95], which may also help explain the observed correlations. The observed size-related patterns may complicate biomonitoring approaches and exposure assessments. Additionally, these findings imply that smaller fish could pose higher health risks per unit weight consumed if they carry relatively higher PFAS burdens.

PFAS occurrence in fish across the African continent generally ranges from low to moderate levels, although concentrations are highly variable depending on the type of PFAS, fish species, and country [20]. Regardless of the fish species, tissue, and type of PFAS, the concentrations observed in the present study are overall comparable to those of other African countries such as Tanzania [38], Ethiopia [39,40], Tunisia [96], Zambia [97], and Uganda [98]. Compared to fish from South Africa, however, concentrations in the present study are often lower [22,26,29,34,35]. On a global scale, concentrations are also highly variable depending on the region and fish species. For example, PFAS concentrations in the present study were comparable to those reported in Tampa Bay, USA [99], and muscle concentrations were similar to those reported in fish from Taihu Lake, China [100]. Liver concentrations, on the other hand, were lower than those reported in Taihu Lake [100], and concentrations of both muscle and liver were lower than those observed in fish from the Jiulong River, China [101]. These findings show that meaningful cross-country comparisons are challenging due to differences in fish species, trophic levels, tissues analyzed, and local emission sources that strongly influence reported concentrations. Despite this variability, our findings contribute important baseline data from Burundi, filling a critical geographic gap and providing context for understanding PFAS contamination in freshwater ecosystems at regional and global scales.

Overall, our findings illustrate that fish are key accumulators of multiple PFAS, with PFBS emerging as a dominant compound of concern. The detection of emerging PFAS such as NaDONA highlights the need to include both legacy and emerging PFAS in monitoring and risk assessment frameworks, particularly given their differences in bioaccumulation and toxicity.

4.4. Human Health Risks

When applying the EFSA TWI for the four PFAS, estimated risks were strongly dependent on body weight. Under this regulatory framework, risks are primarily expected at low body weight, indicating that children represent the most vulnerable group, while most adults remained below the risk thresholds even at high fish consumption rates. In contrast, exceedances were consistently observed across all body weights when concentrations were expressed in PEQ, both for average and maximum PFAS concentrations, suggesting a potential health risk under this alternative, exploratory approach. These effects on humans may include altered immune and thyroid function, kidney and liver disease, cancer, and developmental and reproductive effects [102].

These findings suggest that limiting health risk assessments to the four EFSA-regulated PFAS may underestimate total exposure, since other PFAS with relatively high potency, such as PFDA and PFDoDA, are not included in the current TWI value. However, the comparison between PEQ-based results and the EFSA TWI should be interpreted cautiously, because the RPF values are derived from rat studies focusing on hepatotoxicity [63], whereas the EFSA TWI value is based on immunological effects [60]. Since the relative potency of PFAS may differ depending on the endpoint considered, a comparison of PEQ values with the EFSA TWI is not entirely appropriate. Accordingly, the comparison between both approaches is best viewed as exploratory. This raises questions about the consistency of toxicity assessments across different health outcomes and highlights the need for further research to refine and expand the use of RPF values. In addition, our results emphasize the importance of broadening regulatory frameworks, such as TWI values, to include a wider range of PFAS.

5. Conclusions

This study provides the first comprehensive investigation of PFAS occurrence and distribution in a river-lake system in Burundi, covering both environmental and biological compartments in the Rusizi delta. By establishing baseline data, it provides novel insights from an underrepresented region of Africa and contributes to a broader global understanding of PFAS fate and behavior in tropical freshwater ecosystems. Collectively, our findings highlight the value of biomonitoring compared to environmental monitoring, which detected fewer PFAS. Future research should incorporate longitudinal sampling to capture temporal variability in PFAS transport and fate under changing hydrological conditions (e.g., due to rainfalls or extreme weather events). Moreover, the potential filtering role of wetlands, which could mitigate PFAS exposure to higher trophic levels and reduce human health risks, warrants further investigation. Overall, our study highlights the need and relevance of biomonitoring and ecological research on PFAS in tropical freshwater ecosystems, where data remain scarce but the potential risks to ecosystems and human health can be substantial.