A Hepatic Compartmental Model of Herbicide Bioconcentration in Nile Tilapia (Oreochromis niloticus)

Technical Summary: A Hepatic Compartmental Model of Herbicide Bioconcentration in Nile Tilapia

Problem Statement
The bioconcentration factor (BCF) is traditionally treated as an aggregate quantity, conflating branchial uptake, systemic distribution, and hepatic biotransformation into a single steady-state ratio. While useful for regulatory screening, this aggregate measure cannot be cross-checked against independent physiological data or decomposed to reveal underlying mechanisms, such as metabolic interactions in chemical mixtures. This limitation is particularly acute in sugarcane agriculture, where Nile tilapia (Oreochromis niloticus) are routinely co-exposed to mixtures of herbicides (ametryn, tebuthiuron, hexazinone, and diuron). The authors seek to determine how far an aggregate BCF, derived from muscle tissue sampling alone, can be resolved into physiologically interpretable rate constants, and to evaluate whether this decomposition reveals mixture-dependent effects invisible to standard partition-coefficient-only models.

Methodology
The study reanalyzes published bioconcentration data for four sugarcane herbicides in Nile tilapia (Jonsson et al., 2019) using a three-compartment kinetic model: Water $\rightarrow$ Liver $\rightarrow$ Muscle. Unlike the original two-compartment (Water $\rightarrow$ Muscle) formulation, this model explicitly isolates the liver as the site of biotransformation.

1. Model Structure: The system is governed by differential equations for liver and muscle concentrations. Under a hepatic pseudo-steady-state assumption (liver turnover $\ll$ muscle dynamics), the system reduces to a single second-order ODE for muscle concentration with two identifiable effective parameters: effective uptake ($K_1$) and effective elimination ($K_2$).
2. Calibration: $K_1$ and $K_2$ were fitted via nonlinear least squares to the steady-state muscle concentrations (plateau means) of the original study across nine compound $\times$ mixture combinations. The authors explicitly note that for seven of these combinations, the number of data points equals the number of fitted parameters, making this a calibration rather than an independent validation.
3. Hepatic Decomposition: To decompose the effective parameters into physiological rates ($k_{up}$, $k_{fm}$, $k_{mf}$, $k_{elim}$), the model was anchored to external, independent sources:
   • Elimination ($k_{elim}$): Derived from the US EPA EPI Suite (BCFBAF module) using QSARs, with a specific measured value used for diuron.
   • Uptake ($k_{up}$): A consensus value was established by taking the geometric mean of two independent estimates: a molecular QSAR (Kuo & Di Toro) and a physiological estimate based on measured gill extraction efficiency (McKim et al.) scaled to tilapia physiology.
   • Distribution: The remaining rate constants ($k_{fm}$, $k_{mf}$) were solved algebraically.
4. Risk Assessment: The deterministic risk assessment from the original study was reformulated as a Monte Carlo simulation (Latin Hypercube Sampling) to propagate uncertainty in BCF and consumer body weight, calculating the probability of exceeding the Acceptable Daily Intake (ADI).
5. Verification: The model was tested against 214 individual fish determinations (not just means) to assess parameter identifiability and to check for structural limitations in the experimental design (e.g., non-monotonic concentration declines).

Key Results

• Model Fit and Identifiability: The reduced model reproduced the empirical BCFs with negligible residuals. However, the day-resolved data confirmed that $K_1$ and $K_2$ are individually unidentifiable in seven of the nine cases because sampling occurred only on the plateau phase. The model's ability to separate these constants relied entirely on the external anchoring described above.
• Physiological Decomposition: The decomposition revealed that branchial uptake ($k_{up}$) is significantly faster than hepatic clearance ($k_{elim}$), which in turn is faster than liver-to-muscle distribution ($k_{fm}$). This supports a "strong hepatic first-pass" mechanism where the liver limits the fraction of chemical reaching the muscle.
• External Plausibility Checks:
  • The implied liver-to-muscle partitioning ratios for tebuthiuron, hexazinone, and diuron agreed within an order of magnitude with independent measurements for atrazine in tilapia.
  • The model failed for ametryn, predicting a liver-to-muscle ratio ~10 times higher than observed benchmarks. The authors identify the liver-to-muscle distribution of ametryn as the least constrained element of the model, likely due to its slow hepatic elimination.
• Mixture Effects: The effective elimination rate ($K_2$) for tebuthiuron decreased monotonically as the number of co-exposed herbicides increased (a 34% reduction from binary to quaternary mixtures). This pattern is consistent with competitive inhibition of hepatic enzymes, a mechanism not detectable by the original aggregate BCF.
• Risk Assessment: The Monte Carlo simulation confirmed the original study's conclusion: the probability of exceeding the ADI is negligible (0% in 1,000 simulations) even under extreme consumption scenarios. BCF was the dominant driver of uncertainty in the Maximum Daily Intake (MDI) calculation.

Significance and Claims
The paper claims that reformulating aggregate BCF data into a mechanistic, hepatic-compartmental model allows for:

1. Decomposition: Breaking down the BCF into physiologically interpretable rate constants ($k_{up}$, $k_{elim}$, etc.) by anchoring them to independent databases, rather than leaving them as opaque fitted numbers.
2. Mixture Sensitivity: Detecting that BCF is not a fixed property of a compound but varies with mixture composition, a dependency that partition-coefficient-only QSARs fail to capture.
3. Hypothesis Generation: Providing a structural basis for hypothesizing metabolic interactions (e.g., competitive inhibition) in co-exposure scenarios, which the original empirical approach could only speculate upon.
4. Methodological Transparency: Highlighting the structural limitations of standard two-phase bioconcentration protocols (sampling only at plateau and after transfer to clean water), which render uptake and elimination rates unidentifiable without external constraints and may mask mixture interactions by removing all co-exposures during the depuration phase.

The authors maintain a modest stance, emphasizing that the model is a calibration to existing data rather than a validation of first-order kinetics, and that the decomposition relies on the convergence of independent estimation methods rather than direct measurement of liver residues. The work serves as a proof-of-concept for resolving aggregate bioconcentration data into mechanistic components when only edible tissue (muscle) is sampled.