Implementing Transporter Fluxes in a Hepatic ODE Model

From Flux Laws to ODEs

A rate law becomes useful in a PBPK model only when it is attached to a compartment balance. Each flux must have a defined direction, a rate unit, and a corresponding negative term in its source compartment and positive term in its destination. For hepatic transport, this bookkeeping connects drug in sinusoidal blood, drug inside hepatocytes, and drug entering bile. Getting the signs right is as important as choosing a plausible transporter equation.

The model below describes parent drug in a well-mixed sinusoidal compartment, a hepatocyte compartment, and a bile compartment. It is a local hepatic model: hepatic inflow and outflow are represented explicitly, while whole-body distribution is outside its boundary. The introductory material on hepatocyte mass balance provides the underlying control-volume idea; here the emphasis is implementing several different rate laws together without losing units or mass.

The flow rate \(Q\) is in L/min, and \(C_{\mathrm{in}}\) is the total drug concentration entering the sinusoidal compartment, in nmol/L. Thus \(Q C_{\mathrm{in}}\) and \(Q C_s\) are amounts per time. This simple flow representation assumes a well-mixed sinusoidal compartment and uses the same concentration basis for inflow and outflow. Transporter and passive membrane fluxes are driven by unbound concentrations.

Define Each Flux Before Writing the Balance

Use positive directions to make the bookkeeping explicit. Active uptake \(U_{\mathrm{act}}\) and passive diffusion \(D\) are positive from sinusoid to hepatocyte. Basolateral efflux \(E\) is positive from hepatocyte back to sinusoid. Biliary efflux \(B\) is positive from hepatocyte to bile. Metabolic loss \(M\) removes parent drug from the modeled parent-drug system.

Here \(V_{\max,u}\) is the maximum active-uptake rate in nmol/min, \(K_{m,u}\) is in nmol/L, and \(PS\) is a permeability-surface-area clearance in L/min. The passive term is signed: if \(C_{h,u}>C_{s,u}\), then \(D<0\) and passive movement is from hepatocyte to sinusoid. \(U\) is therefore a net sinusoid-to-hepatocyte flux and need not always be positive.

The efflux and metabolism maximum rates are in nmol/min, and their \(K_m\) values are in nmol/L. These expressions represent saturable, one-direction rate laws over the modeled concentration range. If metabolism is adequately described as first order, its rate may instead be \(M=CL_{\mathrm{int},u}C_{h,u}\), where \(CL_{\mathrm{int},u}\) is in L/min. Choose one metabolic law for a given model; do not add both unless they represent distinct pathways.

Bile is treated as a compartment rather than an immediate loss. If bile drains from the modeled region with first-order rate constant \(k_b\) in min\(^{-1}\), the complete amount balances are:

Every term has units of nmol/min. In the sinusoidal equation, uptake removes drug and basolateral efflux returns it. In the hepatocyte equation, the same uptake adds drug, while efflux, biliary transport, and metabolism remove it. Biliary transport adds drug to bile, and drainage removes it from the modeled bile compartment. The flow terms \(Q(C_{\mathrm{in}}-C_s)\) account for inflow and outflow together.

Why the Signs Conserve Mass

The balance equations follow by applying “rate of change = rates entering − rates leaving” to each compartment. Their internal fluxes should cancel when the compartment equations are added. This cancellation is a direct implementation check, not an extra assumption about the values of the transporter parameters.

The uptake flux \(U\), basolateral return \(E\), and biliary transfer \(B\) each appear once with a minus sign in the source and once with a plus sign in the destination, so they cancel exactly. The remaining terms are external flow, metabolic removal of parent drug, and drainage from the modeled bile compartment. This identity assumes no unmodeled parent-drug source or sink within the three compartments. If metabolites are added as states, metabolic formation must appear as a corresponding input to those states, subject to the chosen stoichiometry.

Implementation Choices and Diagnostics

An explicit ODE model is most useful when transient concentrations, compartment amounts, or separately observed routes of elimination matter. A reduced clearance representation can be convenient when the goal is a compact steady-state relationship, but it does not retain the same compartment time courses. Conversely, adding separate compartments and transporter processes increases the number of parameters and assumptions. Use the level of detail needed to answer the modeling question, and preserve an explicit definition of every modeled state and boundary flux.

For a zonated liver, the same bookkeeping can be applied to each hepatocyte region \(z\): use local \(A_{h,z}\), \(V_{h,z}\), local concentrations, and zone-specific fluxes. Each uptake from a sinusoidal region must be removed from that region’s sinusoidal balance, and each basolateral return must be added back to its destination. Summing over zones should still cancel all internal transfers. The equations do not justify assuming identical transporter activity, binding, or exposure across zones; those are model inputs or hypotheses.

Useful diagnostics can be applied before fitting or simulating the model. Evaluate every rate at a plausible state and confirm its units. Check that \(K_m+C_u\) combines concentrations on the same unbound basis, that transporter rates approach their \(V_{\max}\) values at high substrate concentration, and that passive diffusion changes sign when the concentration gradient reverses. Inspect derivatives at zero amounts: a one-direction efflux term should not remove drug from an empty hepatocyte, and a bile drainage term should be zero when \(A_b=0\). For a closed system with no external flow, metabolism, or bile drainage, the sum \(A_s+A_h+A_b\) should remain constant up to numerical solver error.

A numerical solver can also create apparent problems. Large time steps may yield negative amounts even when the continuous model has nonnegative states; that is a numerical warning, not evidence that negative drug amounts are meaningful. Reduce the step size or use an appropriate solver and inspect tolerances. For saturable kinetics, solver steps must be fine enough to resolve rapid changes when concentrations vary quickly. Monitor both compartment amounts and the summed mass-balance residual rather than relying only on whether a concentration-time curve looks smooth.

Common Mistakes and Limits

• Converting amount to concentration incorrectly. Divide amount by the compartment volume in matching units, then apply the unbound fraction. For example, \(0.020\) nmol in \(0.0020\) L is \(10\) nmol/L total; applying \(f_u=0.20\) gives \(2.0\) nmol/L unbound.
• Giving a transfer flux the same sign in both compartments. A sinusoid-to-cell flux subtracts from sinusoid and adds to hepatocyte. Summing the equations exposes this error because the internal terms fail to cancel.
• Treating passive diffusion as always inward. The signed gradient term \(PS(C_{s,u}-C_{h,u})\) reverses direction when the intracellular unbound concentration is greater.
• Mixing rate and clearance units. A \(V_{\max}\) is an amount per time; a clearance multiplied by concentration is also an amount per time. They are not interchangeable parameter units.
• Counting the same loss twice. If bile is a state, biliary efflux \(B\) moves drug into it; it is not also an immediate loss from the total modeled amount. Bile drainage is represented separately by \(k_bA_b\).

This model assumes each compartment is well mixed, binding is represented by a fixed instantaneous unbound fraction, and the specified rate laws adequately describe the modeled concentration range. It does not by itself describe transporter regulation, pH gradients, active transport energetics, spatial gradients within a cell, or formation and disposition of metabolites. Those details require additional states, mechanisms, or data. A mathematically balanced ODE system is necessary, but it does not establish that its parameter values or structural assumptions are biologically correct.