From Radiocopper to Cold Copper: Mechanistic Modeling and Simulation to Define Clinical Response Criteria and Biomarkers for VTX‐801 in Wilson Disease

ABSTRACT

We developed a comprehensive, mechanistic model of human copper metabolism to support biomarker qualification for VTX‐801, an adeno‐associated vector‐based gene therapy which is being developed to restore the mutated ATP7B copper transporter gene in Wilson disease (WD). The model integrates physiological copper kinetics with pathophysiological features of WD by distinguishing between ceruloplasmin‐bound and non‐ceruloplasmin‐bound copper (NCC), and by explicitly incorporating ATP7B‐dependent processes: biliary excretion and ceruloplasmin loading of copper. Literature‐derived time–activity data from healthy subjects, heterozygous carriers, and WD patients, as well as clinical radiocopper data in plasma and feces from a pilot study in non‐WD subjects, were used for model development and validation. VTX‐801's dose–response was quantified in WD mouse models using ceruloplasmin oxidase activity measurement and ⁶⁴Cu fecal excretion. This enabled derivation of activity factors (AFs) corresponding to restored ATP7B function, with 15% and 40% selected as minimal and optimal efficacy targets. Simulations linked AFs to clinical biomarkers, demonstrating that the 48/2‐h plasma radioactivity ratio can effectively differentiate VTX‐801 responders from non‐responders, providing a decision criterion to safely withdraw standard treatment in participants of a phase 1/2 trial. To broaden applicability beyond radiotracer studies, we simulated “cold” copper kinetics under steady‐state conditions, deriving expected values for plasma copper, NCC, urinary copper excretion, and relative exchangeable copper (REC). These simulations suggest that REC may also serve as a suitable and simpler to implement, non‐radioactive biomarker for ATP7B gene therapy. This model provides a robust quantitative framework to assess copper‐related biomarkers in WD and their response to treatment in silico.

Trial Registration: EudraCT number: 2019‐001157‐13

Study Highlights.

• What is the current knowledge on the topic?

  • ○
    Existing copper models describe physiological kinetics but do not capture the pathophysiology of Wilson disease (WD) or support biomarker‐based assessment of gene therapy.
• What question did this study address?

  • ○
    Can a mechanistic model of copper metabolism be developed to support dose selection and biomarker qualification for VTX‐801 gene therapy in WD?
• What does this study add to our knowledge?

  • ○
    This study presents the most complete physiological model of copper metabolism to date, incorporating ATP7B‐dependent processes and disease‐related dysfunction. It links preclinical dose–response data to human biomarker predictions and defines activity factor thresholds tied to treatment efficacy.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    This modeling and simulation framework supported key decisions in the clinical development of a novel gene therapy for WD. It provides a mechanistic basis for biomarker qualification and trial design and may serve as a platform for future research in copper‐related disorders or therapies targeting similar pathways.

1. Introduction

Wilson disease (WD) is a rare autosomal recessive disorder of copper metabolism caused by loss‐of‐function mutations in the ATP7B copper transporter gene. In healthy individuals, ATP7B in hepatocytes incorporates copper into ceruloplasmin (CP) for export into blood and mediates biliary excretion of excess copper. In WD patients, ATP7B dysfunction leads to impaired biliary copper elimination and deficient incorporation of copper into CP. Consequently, copper accumulates in the liver and other tissues and circulating CP concentrations are abnormally low. Untreated WD results in hepatic failure, neurologic damage, and other systemic complications due to copper toxicity, and ultimately death; notably early death was reported in various countries with up to 20 years of life lost [1]. Current standard treatments (chelators or zinc) facilitate copper removal or prevent its uptake but require patient adherence to lifelong therapy and monitoring. There is a critical need for disease‐modifying therapies that permanently or durably restore physiologic copper homeostasis. VTX‐801 is an investigational adeno‐associated vector‐based (AAV) gene therapy encoding a miniATP7B gene, designed to introduce functional ATP7B into the liver and thereby correct the underlying copper‐handling defect.

A key question for such a gene therapy is how much restoration of ATP7B activity is needed to normalize copper metabolism and how to reliably measure treatment response in the presence of standard of care (SOC) background treatment. Copper levels in plasma and 24‐h urinary copper excretion are commonly used to monitor WD. However, SOC treatment (chelators or zinc) can normalize these “cold copper” biomarkers, potentially confounding the assessment of additional therapeutic effects when gene therapy is introduced before SOC can be safely withdrawn in responders. A more robust method is the radiocopper balance test, in which a trace dose of radioactive copper (e.g., ⁶⁴Cu or ⁶⁷Cu) is administered and radioactivity in plasma as well as fecal samples are tracked over time. In non‐WD subjects, radiocopper is incorporated into CP, leading to a secondary increase in plasma radioactivity a few days after administration. In contrast, WD patients fail to incorporate radiocopper into CP, and this secondary increase is absent [2]. Therefore, the ratio of plasma radioactivity at 24 or 48 h to that at 2 h after radiocopper administration (the “⁶⁴Cu incorporation ratio”) has been shown to discriminate between WD patients on one hand and heterozygous carriers and healthy individuals on the other hand, with exceptional sensitivity and specificity. While heterozygous carriers typically do not exhibit clinical symptoms of WD and are therefore considered clinically healthy, we will refer to non‐carriers as ‘healthy subjects’ or ‘healthy controls’ for clarity.

These radiocopper‐based metrics, along with other traditional measures, were meant to be used in an open‐label first‐in‐human study of VTX‐801 in WD patients to determine whether subjects respond well at a given dose levels or if otherwise the dose should be escalated. Robust responder criteria (e.g., cut‐off value of the 48/2 h‐ratio) needed to be defined before the trial started.

Mechanistic modeling of copper kinetics offers a way to integrate data from diverse sources (preclinical and clinical) and simulate the behavior of different biomarkers under varying degrees of ATP7B function. Here we describe a minimal physiologically based pharmacokinetic (PBPK) model of copper distribution and elimination, developed to characterize the kinetics of radioactive “hot” copper in healthy individuals, heterozygous carriers and in WD patients. The model incorporates disease‐specific impairments in CP incorporation and biliary excretion, using an activity factor (AF) to represent the fraction of normal ATP7B function. Radiocopper data from both literature sources and a dedicated clinical study (‘COMET’) in healthy volunteers and WD heterozygotes calibrate and validate the model. Dose–response data from WD mice treated with VTX‐801 were used to link gene therapy dose to the restoration of copper elimination and ceruloplasmin activity [3].

Simulations of single i.v. doses of radiocopper were performed to investigate the response of the 48/2 h‐ratio to partial restoration of ATP7B activity in humans and thus helped to define the ideal cut‐off value to be used in the first‐in‐human trial.

Simulations of daily oral doses of copper over a long period (i.e., mimicking natural ingestion) allowed the derivation of “cold” copper metrics and supported the in silico comparison of traditional markers (such as plasma and urinary copper) to relative exchangeable copper (REC), a novel biomarker quantifying the fraction of circulating copper not bound to CP.

We present the development and evaluation of this PBPK model and illustrate its application in predicting treatment effects of VTX‐801.

2. Materials and Methods

2.1. Literature Data

Time–activity profiles for plasma, fecal, and urinary excretion of radiolabeled copper were digitized from 18 published human studies dating back to 1954. These included healthy individuals, heterozygous carriers, and WD patients, with radiocopper administered either orally or intravenously. Some WD patients were under zinc or chelator therapy (e.g., penicillamine, trientine). Most studies used ⁶⁴Cu (half‐life ≈12 h), limiting observations to 3–5 days, though some used longer‐lived ⁶⁷Cu (~60 h) or stable ⁶⁵Cu. We also included studies measuring liver radioactivity via positron emission tomography (PET) [4, 5], or scintillation [6].

Data were mostly reported as means or medians, with some individual values. Plasma radioactivity was standardized to % of injected dose using an average plasma volume of 2800 mL [7]. An overview of the 18 references is provided in Table S1.

2.2. Clinical Data

The COMET study (EudraCT: 2019‐001157‐13) was an open‐label, non‐randomized trial assessing fecal, urinary, and plasma ⁶⁴Cu levels after a single intravenous dose of [⁶⁴Cu]CuCl₂ in three healthy adults and three WD carriers [8]. The study was approved by the Ethics Committee of the study center and the competent authority of Austria (BASG). A total of 84 plasma samples were obtained up to 96 h post‐dose, of which 16 were below the limit of detection and hence discarded from the analysis.

For external model qualification, raw data from a retrospective study (partially published in Członkowska et al. [2]) provided ratios of serum radioactivity at 24 and 48 h relative to 2 h post‐dose in WD patients (N = 86), carriers (N = 40), and healthy controls (N = 18).

2.3. Preclinical Data

Dose–response relationships for VTX‐801 were derived from two experiments in mice (partially published in Murillo et al. [3]). In both studies 6‐week‐old male and female WD and wild‐type mice (n = 6 per group) received vehicle, 1.5 × 10¹², 5 × 10¹², or 15 × 10¹² vg/kg of VTX‐801 intravenously. In the first study, Ceruloplasmin oxidase activity was measured at week 4 after VTX‐801 treatment. In the second study 3 months after treatment ⁶⁴Cu was administered to assess radioactivity in plasma, feces, and urine.

2.4. Model Development

2.4.1. Basic Structure and Copper Distribution

Several compartmental models describing human copper kinetics have been developed previously, including those by Scott [9], Buckley [10], and Harvey et al. [11] which served as the starting point for this work. These models represent copper exchange among interconnected compartments such as liver, plasma, gastrointestinal tract, kidneys, and peripheral tissues. An initial version of the model was developed before the COMET data became available solely using literature data of WD and non‐WD subjects (carriers and healthy subjects) receiving i.v. or oral radiocopper. While it captured plasma, urine, and feces profiles well, liver copper predictions were unrealistic. The updated model described in the present article incorporates hepatic sub‐compartments—extracellular, precursor, and storage—based on the cellular framework proposed by Das and Ray [12] and was calibrated using liver‐specific PET and scintillation data. A schematic representation of the model is shown in Figure 1.

FIGURE 1.

Schematic representation of updated copper model. AFbil, activity factor for transporter‐mediated excretion of copper into the bile; AFcp, activity factor for binding of copper to ceruloplasmin; Cu‐CP (ceruloplasmin), copper bound to ceruloplasmin; NCC (non‐ceruloplasmin‐bound copper), non‐ceruloplasmin bound copper. See text and Table 1 for a description of the parameters.

The model explicitly distinguishes between CP‐bound copper (Cu‐CP) and non‐CP‐bound copper (NCC). Although often referred to as “free” copper, NCC may be loosely bound to albumin or other proteins. Both copper incorporation into CP and biliary excretion are ATP7B‐dependent; loss of this transporter function leads to NCC accumulation. Cu‐CP kinetics are governed by the kinetics of the protein itself, with distribution rates taken from studies of labeled protein and elimination based on a reported half‐life of 5.5 days [13, 14].

Another important feature of the model, also described by Buckley [10], is the reabsorption of copper that enters the gastrointestinal (GI) tract via the saliva (and other GI fluids) and, more importantly, the bile. According to these authors, approximately 25% of biliary copper is reabsorbed in the proximal gut; this was implemented by splitting the biliary elimination rate (k_livPre_out) into k_liv_abs and k_liv_gutD in a 25:75 ratio.

Model fitting was performed simultaneously on literature data for plasma (total and CP‐bound), feces, urine, and liver radioactivity, and on COMET plasma and feces data. Only i.v. datasets were used in this update, with oral absorption parameters fixed to values obtained in the earlier model.

Separate residual error terms were estimated for each matrix as well as a separate term for observations in plasma from the COMET study. Since most of the data from literature originated from mean curves or tabulated averages, the residual error was weighted according to the number of subjects contributing to the average as shown in the following equation:

$$y_i=y_i^{\prime }+ϵ$$

where y _i is the ith observation, $y_i^{\prime }$ the corresponding prediction and $ϵ$ (epsilon) is a normally distributed random variable with mean 0 and standard deviation: $\sqrt{\frac{{y_i^{\prime }}^2\times \mathrm{pro}{\mathrm{p}}_{\mathrm{ruv}}^2}{N}}$, prop_ruv being the proportional residual variability and N the number of subjects.

The number of subjects in the COMET study was too small to reliably estimate the inter‐individual variability (IIV) on any parameter and the first‐order estimation method in NONMEM (version 7.4.3) could be used.

2.4.2. Effect of Wilson Disease and Treatments on the Model Parameters

In WD, as mentioned, dysfunction of the hepatic ATP7B transporter impairs both biliary copper excretion and incorporation of copper into CP. To reflect this, the updated model uses separate activity factors (AF) for each process: AFcp for CP binding and AFbil for biliary excretion, both set to 0 in WD patients, 1 in non‐heterozygous controls and freely estimated for heterozygote genotypes.

2.4.3. VTX‐801

The effect of VTX‐801 in this modeling context is to increase the AF of a WD patient to a value > 0. The (minimal) value necessary to achieve a clinical response was derived from the preclinical experiments by fitting a sigmoid Emax model to measurements of CP oxidase activity in plasma and ⁶⁴Cu fecal excretion at 72 h [3], where the hyperbolic term of the Emax equation represents the implicit activity or functioning of the ATP7B transporter at a given dose level $\left(\mathrm{AF}=\frac{{\text{DOSE}}^n}{{\text{DOSE}}^n+\mathrm{ED}{50}^n}\right)$.

2.4.4. Chelators

Chelators like penicillamine, trientine or dimercaprol (a.k.a. BAL), increase the urinary elimination of copper by forming soluble copper complexes. The effect was introduced in the model on the renal elimination rate as follows:

$$\mathrm{k}\_\mathrm{pl}\_\text{urine}=\mathrm{TV}\_\mathrm{k}\_\mathrm{pl}\_\text{urine}\times \left(1+\text{chel}\_\mathrm{eff}\times \text{onChel}\right)$$

where TV_k_pl_urine is the typical value of the renal elimination rate for someone not taking a chelator (onChel = 0), Chel_eff is the fractional increase of the renal elimination rate for a patient taking a chelator (onChel = 1).

2.4.5. Zinc

Zinc reduces copper absorption by inducing enterocyte metallothionein expression, which preferentially binds copper and prevents its entry into portal circulation [15]. This effect was modeled as a reduction in gut absorption rate:

$$\mathrm{k}\_\mathrm{abs}\_\mathrm{liv}=\mathrm{TV}\_\mathrm{k}\_\mathrm{abs}\_\mathrm{liv}\times \left(1+\text{Zinc}\_\mathrm{eff}\_\mathrm{abs}\times \text{onZinc}\right)$$

where Zinc_eff_abs is negative when zinc is present (onZinc = 1). Zinc also alters plasma radioactivity profiles after IV ⁶⁴Cu administration—likely due to changes in tissue binding [16]. This was mimicked by modifying the plasma‐to‐tissue distribution rate:

$$\mathrm{k}\_\mathrm{pl}\_\mathrm{tis}=\mathrm{TV}\_\mathrm{k}\_\mathrm{pl}\_\mathrm{tis}\times \left(1+\text{Zinc}\_\mathrm{eff}\_\text{dist}\times \text{onZinc}\right)$$

2.5. Simulations

2.5.1. VTX‐801 Responder Criteria Based on Radiocopper

Radioactive‐dose‐normalized concentration‐time profiles of 42,000 virtual patients (2000 per AF from 0 to 1 in 0.05 steps) were simulated and the 48/2‐h plasma concentration ratio (as well as cumulative fecal excretion, not discussed here) was derived.

Knowing the true AF, classification accuracy—false positives, false negatives, and correct identifications—was assessed across cut‐off values.

Some adjustments and assumptions were made in the simulations:

• Differences in certain distribution rate constants between WD and non‐WD subjects were not applied, as their behavior under treatment remains uncertain (see results and discussion).

• Inter‐individual variability (IIV) was set to 15% CV and residual variability to 15% CV (a scenario with higher variability was also simulated).

• To account for distinct activity factors for ceruloplasmin binding and biliary excretion, a power function was used to interpolate their combined effects.

2.5.2. Model Qualification

Simulations for visual predictive checks (VPC) for the subjects in COMET were done similarly using the estimated AF for heterozygous WD carriers and one for healthy subjects. The observed profiles were then compared to the prediction interval derived from the 2000 virtual subjects.

External validation of the model against 48/2‐h and 24/2‐h plasma ratios as reported by Członkowska et al. [2] was conducted. Again, 2000 virtual subjects per population (heterozygote, healthy and WD patients) were simulated and the plasma ratios were derived and compared to the observed values.

2.5.3. Simulations of “Cold” Copper Biomarkers

Cold copper simulations assumed radiocopper and endogenous copper behave identically. Daily oral intake (1.4 mg for men [https://ods.od.nih.gov/factsheets/Copper‐HealthProfessional/]) was simulated over 1500 days to ensure steady state in all compartments and all subjects.

We simulated four different populations (N = 2000 each): healthy controls, untreated WD patients, as well as WD patients under chelator or zinc as background therapy. To mimic gene therapy, AF in WD patients was increased from 0 to values between 0.05 and 0.4, representing various levels of effectiveness. Output variables were obtained at baseline and 36 weeks after the hypothetical gene therapy vector injection and included: copper content in the liver, daily urinary copper excretion, total serum copper (CP‐bound + CuEXC), exchangeable copper (CuEXC), relative exchangeable copper (REC), calculated as: REC = 100 × CuEXC (μmol/L)/total plasma copper (μmol/L). Note, that NCC and CuEXC while very differently determined in practice aim to measure the same moiety, which is “free” copper, not bound to ceruloplasmin. In our model there is no distinction between the two.

3. Results

3.1. Model Fitting

The model predictions were in good agreement with the observations as can be judged by the goodness‐of‐fit plots and residual plots (Figures S1 and S2) as well as plots showing the predicted time‐course per matrix (Figure 2).

FIGURE 2.

Observed and predicted radioactivity‐time profiles in different matrices. Thick solid lines are model predictions, dots and thin lines are observations. %ID, percent radioactivity of injected dose, WD, Wilson patients. “Liver” is the sum of amounts in the precursor, storage and extracellular liver compartments in the model, “plasma total” is the sum of copper bound to ceruloplasmin and copper not bound to this protein.

As can be seen in these plots, the model captures well the key features of copper kinetics in WD patients, heterozygote carriers and healthy subjects, namely:

• Accumulation of copper in the liver of WD patients and to a lesser degree also in heterozygotes.

• Higher fecal and urinary excretion of copper in healthy subjects (non‐heterozygotes), as well as faster and more binding to CP.

• Increasing levels of radioactivity in (total) plasma due to CP binding in heterozygotes and healthy subjects, while the levels consistently decrease in WD patients.

The VPC shown in Figure 3 also demonstrates good resemblance of model predictions with the data from the COMET study. An alternative VPC including parameter uncertainty is shown in Figure S7.

FIGURE 3.

Visual predictive check for COMET data—low variability scenario. Circles are the observations in the COMET study. The shaded areas comprise the 5th–95th percentiles of the 2000 simulated profiles, corresponding to the 90% prediction intervals; the solid line represents the median of the simulations. Inter‐individual variability is set to 15% CV on all rate constants. Residual variability is set to 15%.

The AFs informed by data from heterozygotes, were estimated at 0.369 and 0.101 for AFcp and AFbil, respectively. Attempts to estimate a single activity factor for both processes resulted in a significantly worse model fit (∆OFV = 40.657).

The following parameters were significantly different between WD patients and non‐WD subjects:

• The rate from extracellular/vascular space in the liver to the precursor pool (k_liv_livPre) was 15.3% lower in WD patients.

• The rate from the precursor pool back to the extracellular/vascular space in the liver (k_livPre_liv) could only be estimated in WD patients (i.e., value fixed to 0 in non‐WD subjects).

• The rate of distribution from tissue back to plasma (k_tis_pl) and from deep tissue to plasma (k_tis2_pl) was 2.43‐ and 27.6‐fold higher in WD patients, respectively.

The higher tissue‐to‐plasma distribution rates most likely reflect the excess of unbound copper in tissue and the saturation of binding sites, in line with the pathomechanism of WD.

Chelator treatment increased renal elimination 16.8‐fold, though this estimate relied heavily on a single study [17]. Fecal excretion in COMET subjects (for both heterozygotes and healthy individuals) was slower than in the literature, as reflected by a 60.7% reduction in the gut transfer rate (k_gutD_feces).

Model parameters are listed in Table 1.

TABLE 1.

Parameters of the copper model.

Parameter	Description	Estimate a	RSE%	Comment
k_abs_liv (day−1)	Absorption rate from absorption compartment (duodenum/stomach) to the liver	6	Fixed	Fixed to value obtained with initial version of the model
k_abs_gutD (day−1)	Transfer rate from absorption compartment to distal gut compartment	14.0	Fixed	Fixed to value obtained with initial version of the model
k_livPre_out (day−1)	Biliary elimination rate	1.13	20.4
E_bill	Fraction of Cu in bile that is NOT reabsorbed	0.75	Fixed	Fixed to value reported in Buckley [10]
k_liv_livPre (day−1)	Rate from extracellular liver compartment to precursor pool	169	16.4
k_livPre_liv (day−1)	Rate from precursor pool to extracellular liver compartment	4.9	27.6	Only estimated in Wilson patients. Fixed to 0 for the others
k_livPre_plCP (day−1)	Rate from precursor pool to binding to CP	0.516	18.4
k_livPre_livStore (day−1)	Rate from precursor to storage pool	5.26	29.1
k_livStore_livPre (day−1)	Rate from storage to precursor pool	0.711	18.0
k_gutD_feces (day−1)	Elimination rate from distal gut to feces	1.00	13.0
k_pl_liv (day−1)	Plasma → liver distribution rate of NCC	459	Fixed	Fixed to physiological value
k_liv_pl (day−1)	Liver → plasma distribution rate of NCC	1339	16.0
k_pl_tis (day−1)	Plasma → tissue distribution rate of NCC	91.8	13.8
k_tis_pl (day−1)	Tissue → plasma distribution rate of NCC	29.7	18.0
k_pl_tis2 (day−1)	Plasma → deep tissue distribution rate of NCC	42.9	7.0
k_tis2_pl (day−1)	Deep tissue → plasma distribution rate of NCC	0.0863	33.7
k_plCP_tisCP (day−1)	Plasma → tissue distribution rate of CuCP	0.380	Fixed	Fixed to value reported in Waldmann et al. [13]
k_tisCP_plCP (day−1)	Tissue → plasma distribution rate of CuCP	0.800	Fixed	Fixed to value reported in Waldmann et al. [13]
k_plCP_liv (day−1)	Rate of conversion of CuCP to NCC	0.126	Fixed	Fixed to biological half‐life of CP ~5.5 days (Hellman and Gitlin [14])
k_tis_sweat (day−1)	Tissue to sweat excretion rate of NCC	0.00443	Fixed	Fixed to value suggested in Buckley [10]
k_pl_urine (day−1)	Plasma to urine excretion rate of NCC	0.411	8.6
k_tis_abs (day−1)	Elimination rate from tissue to the absorption compartment	0.983	26.1
Chel_eff	Fractional change in urine elimination if treated with chelators	15.8	36.6	16.8‐fold increase
WD_liv_livPre	Fractional change of k_liv_livPre in WD patients	−0.153	64.8	15.3% decrease
WD_tp	Fractional change of k_tis_pl in WD patients	1.43	38.6	2.43‐fold increase
WD_t2p	Fractional change of k_tis2_pl in WD patients	26.6	36.6	27.6‐fold increase
AFcp	Activity factor in heterozygotes for binding of copper to ceruloplasmin	0.369	5.5
AFbil	Activity factor in heterozygotes for biliary elimination	0.101	24.8
COMET_eff_gut	Fractional change of k_gutD_feces in COMET subjects	−0.607	7.5
Zinc_eff_abs	Fractional change in absorption rate if treated with zinc	−0.401	Fixed	Fixed to value obtained with initial version of the model
Zinc_eff_dist	Fractional changing in plasma‐tissue distribution rate with zinc	1.2	Fixed	Fixed to value obtained with initial version of the model
Residual error terms
propRUVpl	Proportional residual error of plasma measurements (in literature)	0.698	6.1
propRUVplCo	Proportional residual error of plasma measurements in COMET study	0.176	9.3
propRUVfe	Proportional residual error of fecal excretion	0.711	10.4
propRUVur	Proportional residual error of urinary excretion	0.611	11.8
propRUVliv	Proportional residual error of liver measurements	0.203	9.5
propRUVcp	Proportional residual error of ceruloplasmin‐bound fraction in plasma	0.441	16.6

Abbreviations: CuCP, ceruloplasmin‐bound copper; NCC, non‐ceruloplasmin‐bound copper; RSE, relative standard error.

^a All rate constants were estimated on the log‐scale and back‐transformed by exponentiation.

The results of the external validation using 48/2‐h and 24/2‐h plasma ratios as reported by Członkowska et al. [2] show an excellent agreement between the simulated and observed values for heterozygotes and healthy controls (Figure 4). While the ratios for WD patients tend to be overpredicted, they are still well below the diagnostic cut‐off value of 0.395 determined empirically and reported by Członkowska et al.

FIGURE 4.

External validation comparing simulated and observed 48/2‐h and 24/2‐h ratios. Dots are individual measurements, boxes encompass the 25th to 75th percentile of the observed or simulated distributions. WD, Wilson patients. Dotted line indicates the diagnostic cut‐off proposed by Członkowska et al. for the 48/2 h‐ratio.

The NONMEM model code is provided in the Supporting Information.

3.2. Dose–Response in Mice

Untreated WD mice showed ~2.5% fecal excretion and low CP activity (< 0.3), while the highest dose of VTX‐801 restored both markers to near or above wild‐type levels. Parameter estimates and Emax model fits are provided in Table S2 and Figures S3 and S4.

Based on overall preclinical findings, including liver enzymes and animal health status, 1.5 and 5.0 × 10¹² vg/kg were considered efficacious doses. The corresponding activity factors were 0.13–0.208 (low dose) and 0.374–0.404 (high dose). For human responder simulations, consensus AF values of 0.15 and 0.4 were selected to represent minimal and optimal target miniATP7B activity.

3.2.1. Simulation of Responder Criteria

Figure 5A illustrates how varying the activity factors influences plasma radioactivity profiles over time. The delayed secondary rise reflects ATP7B‐dependent incorporation of copper into ceruloplasmin, a slower kinetic process. This highlights why the 48/2‐h ratio is a valuable marker for distinguishing WD patients from healthy subjects.

FIGURE 5.

(A) Simulated typical profiles of radiocopper in plasma for different values of the activity factor (AF). (B) Simulated 48/2‐h plasma radioactivity ratio as function of AF, the line is the median, the shaded area encompasses the 5th and 95th percentile of the 2000 virtual subjects per AF. (C) Probability of correct classifications using the 48/2‐h plasma ratio for two different target AF. A correctly classified responder would be when the 48/2‐ratio is above the given cut‐off target (x‐axis in the plot) and the underlying AF of the virtual subject is above 0.15 (or 0.4). P(correct), percentage of correct decision; P(FN), percentage of false‐negative classifications; P(FP), percentage of false‐positive classifications.

The expected 48/2‐h ratios for different values of the AFcp were derived from the virtual population and are shown in Figure 5B. From these simulations, the proportion of correctly classified responders—those with an AFcp above a given target value—was calculated across a range of cut‐off values for the 48/2‐h plasma ratio. Results are presented as probability curves in Figure 5C for both optimistic (AFcp = 0.15) and realistic (AFcp = 0.4) response targets. In this scenario, the optimal 48/2‐h cut‐off values to determine responders (peak of the red curves) would be around 0.3 for the optimistic case and 0.65 for the realistic case. In a high‐variability scenario (see Figure S5), the optimal values were 0.2 and 0.4. Eventually the value of the criterion was set to 0.4 in the ongoing clinical trial in WD patients, as this matched the diagnostic cut‐off determined by Członkowska et al. [2] and, according to our simulations, results in a probability of correct classification > 80% in all scenarios.

3.2.2. Cold Copper Simulations

Simulations of cold copper biomarkers show that our model based on radiotracer data can also predict clinically relevant cold copper endpoints, aligning with known reference values in healthy individuals and WD patients (Figure 6). REC emerged as the most sensitive marker for detecting changes in AF, as it reflects both the decline in CuEXC and the rise in total serum copper due to CP incorporation. In contrast, 24‐h urinary copper—though commonly used—is less responsive to small AF changes, especially in patients receiving ongoing chelator or zinc therapy.

FIGURE 6.

Simulations of cold copper biomarkers. Top row: Raw values. Green areas illustrate areas below or above common diagnostic threshold values (i.e., healthy if within green zone); Healthy subjects have an activity factor (AF) of 1, while WD patients are assumed to have value of 0 at baseline. The potential effect of ATP7B‐restoring gene therapy (36 weeks post‐injection), is shown for different activity factors. Bottom row: Same data represented a change from baseline. Note, in the panel for REC simulations for all three WD populations are overlapping. Standard of care WD treatment was assumed to be at steady state throughout the simulation period.

The amount of copper in liver tissue is not commonly measured in practice given its invasive nature. Values below ~50 μg/g dry tissue can be considered normal [18], while values in (untreated) WD patients often exceed 250 μg/g [19], suggesting that our model tends to underpredict the amount in the liver in this population (~75 μg/g).

4. Discussion

A compartmental model was developed describing the kinetics of copper in the human body. The model attempts to closely reflect human physiology in both the healthy and disease states as it was calibrated with literature data from a variety of different published studies, including patients with WD and individual clinical data from the recent COMET study in healthy heterozygotes. In a recent study Munk et al. [20] proposed a multi‐compartment model based on data from PET scans; however, with the limitation that the observation period was only up to 20 h post‐dose, it therefore didn't capture the increase of radioactivity in blood due to binding to CP. This mechanism, however, was of particular interest in the present work since it is dependent on the ATP7B activity and thus is impaired in WD patients. In healthy individuals more than 95% of the copper in the human plasma is bound to CP [14], and the kinetics of unbound copper and copper bound to CP differ substantially, offering the opportunity to indirectly determine ATP7B activity by analyzing plasma kinetics. Radiocopper levels 24 h and more after injection are determined by the rate of incorporation of copper into CP (k_liv_liv_CP) and is the basis of the 48/2‐h plasma ratio as a useful diagnostic tool to differentiate between WD and non‐WD subjects with a high degree of certainty [2].

In contrast to standard of care treatment (zinc, chelators) gene therapy is expected to correct impaired loading of copper on CP in WD patients; hence the 48/2‐h ratio is also useful as a marker of efficacy in gene therapy clinical trials. Our model was used to find the optimal cut‐off value of this ratio that would allow us to identify, with high probability, patients who respond to the treatment. Importantly, we assumed that full restoration of ATP7B is not required for clinical benefit, supported by the fact that heterozygous carriers—who in theory retain about 50% function—typically show no signs of copper toxicity (e.g., neurological or hepatic symptoms). Mouse dose–response data suggested that restoring 15%–40% function is sufficient for efficacy [3].

However, prediction of the efficacious gene therapy dose in humans based on animal data is notoriously challenging given the profound species differences regarding the immunological response as well as the kinetics of the administered vector [7, 21, 22]. Therefore, our model could not be used for this purpose requiring a dedicated dose‐finding study in humans. However, the model informed prespecified 48/2‐h ratio cut‐offs to guide the safe withdrawal of the standard of care treatment and dose escalation in the first‐in‐human gene therapy trial. Fecal ⁶⁴Cu excretion was also evaluated as a responder criterion but showed higher uncertainty due to an unexplained ~60% slower gut transit time in COMET compared to the literature. Additionally, the limited availability of published fecal data in the literature hindered robust model calibration. Beyond these scientific limitations, the logistical complexity of implementing fecal sampling in multicenter trials further reduces its practical utility anyhow.

While the radiocopper technique and the derived ratio provide a robust and sensitive method to determine treatment effects in a proof‐of‐concept study, the approach is impractical in larger trials. We therefore explored alternative, non‐radioactive biomarkers in silico. Assuming biological equivalence between stable and radiolabeled copper, we utilized our model to simulate (cold) copper concentrations in serum and urine in response to ATP7B altering treatment. We first had to bring the kinetic system to steady state by assuming a constant daily intake of 1.4 mg copper. It is acknowledged that daily copper consumption can vary significantly across geographical regions (typical food composition), gender, and disease (e.g., copper restrictions) [23]. However, for the purpose of comparing the usefulness of different biomarkers we considered it of limited relevance to account for this source of variability.

Our simulations showed that relative exchangeable copper (REC) is highly sensitive to small changes in ATP7B activity, outperforming 24‐h urinary copper, which requires a ≥ 0.2 increase in AF to show meaningful change (e.g., 20%). This is because REC reflects the ratio of exchangeable copper (which decreases with treatment) to total copper (which increases due to ceruloplasmin incorporation). Urinary copper is more variable, less sensitive, and complicated by background treatments like chelators or zinc, which reduce its responsiveness—while REC remains largely unaffected. On the other hand, exploratory simulations indicate that REC is not changing at all when chelators or zinc are given to naïve WD patients (see Figure S8 and campsis simulation code in Supporting Information). This is plausible since these treatments do not change CP binding and therefore total copper and exchangeable copper decrease simultaneously.

The model was externally qualified against an independent dataset of 48/2‐h ratios across WD, heterozygotes, and healthy subjects. Simulated baseline values for cold copper biomarkers were consistent with reference ranges in the literature. However, a limitation of our model is, that given the scarcity of individual profiles, we could not determine between‐subject variability and assumed IIV of 15% CV on all rate constants and 15% CV RUV. Although arbitrary, the prediction intervals of simulations compare better to the observed variability of the external 48/2‐h (and 24/2‐h) data as well as the internal plasma and fecal elimination profiles from COMET, than a scenario with higher variability (30% IIV/20% RUV; see Figure S6).

The model also lacks nonlinear tissue binding. Tissue‐plasma distribution kinetics of radiocopper in WD patients were found to be different compared to healthy controls, most likely due to saturation of binding sites or storage space in tissue leading to an excess of unbound copper in tissue. This was captured in the model by a binary covariate on the relevant parameters. It may possibly impact, to some extent, the simulations of decreasing copper concentration in tissues of a WD patient who ‘converts’ back to normal as ATP7B functioning improves under treatment.

Moreover, the finding that binding of copper to CP and elimination via the bile—both processes governed by ATP7B—required a different activity factor in heterozygous carriers (0.37 and 0.10, respectively), is plausible; however, it requires further investigation. In the current simulations the two processes are linked by an empirical power function anchored between 0 (WD) and 1 (healthy). These anchor points are themselves assumptions; as given biological variability, WD patients may still have some minor residual ATP7B activity (i.e., AF > 0) despite presenting clinical symptoms.

Nevertheless, the presented modeling and simulation approach was an invaluable tool in supporting the design and conduct of a Phase I/II clinical trial investigating VTX‐801, a novel gene therapy, and informed decision making throughout the development program [24]. Furthermore, it may serve as a platform for future research in copper‐related diseases or drug development programs, perhaps with different mechanisms of action.

Author Contributions

A.L. wrote the manuscript; B.B., J.‐P.C., and G.G.A. designed the research; A.L. performed the modeling; all authors analyzed and reviewed the data.

Conflicts of Interest

J.‐P.C. is an employee and shareholder of Vivet. BB, G.G.A., were employees and shareholders at the time of the conduct of the research. A.L. is a paid consultant for Vivet. VTX‐801 is a property of Vivet Therapeutics.

Supporting information

Data S1: psp470153‐sup‐0001‐DataS1.docx.

Data S2: psp470153‐sup‐0002‐DataS2.zip.

Lindauer A., Benichou B., González Aseguinolaza G., and Combal J.‐P., “From Radiocopper to Cold Copper: Mechanistic Modeling and Simulation to Define Clinical Response Criteria and Biomarkers for VTX‐801 in Wilson Disease,” CPT: Pharmacometrics & Systems Pharmacology 15, no. 1 (2026): e70153, 10.1002/psp4.70153.