Translational pharmacokinetic/pharmacodynamic model for mRNA‐0184, an investigational therapeutic for the treatment of heart failure

Abstract

Heart failure (HF) is a complex, progressive disorder that is associated with substantial morbidity and mortality on a global scale. Relaxin‐2 is a naturally occurring hormone that may have potential therapeutic benefit for patients with HF. To investigate the therapeutic potential of relaxin in the treatment of patients with HF, mRNA‐0184, a novel, investigational, lipid nanoparticle (LNP)–encapsulated mRNA therapy that encodes for human relaxin‐2 fused to variable light chain kappa (Rel2‐vlk) was developed. A translational semi‐mechanistic population pharmacokinetic (PK)/pharmacodynamic (PD) model was developed using data from non‐human primates at dose levels ranging from 0.15 to 1 mg/kg. The PK/PD model was able to describe the PK of Rel2‐vlk mRNA and translated Rel2‐vlk protein in non‐human primates adequately with relatively precise estimates. The preclinical PK/PD model was then scaled allometrically to determine the human mRNA‐0184 dose that would achieve therapeutic levels of Rel2‐vlk protein expression in patients with stable HF with reduced ejection fraction. Model‐based simulations derived from the scaled PK/PD model support the selection of 0.025 mg/kg as an appropriate starting human dose of mRNA‐0184 to achieve average trough relaxin levels between 1 and 2.5 ng/mL, which is the potential exposure for cardioprotective action of relaxin.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

The global burden of heart failure imposes significant health and economic consequences. Relaxin‐2 is a naturally occurring short‐acting hormone that may have potential therapeutic benefit as a heart failure treatment. mRNA technology can be leveraged to enable production and secretion of a long‐acting relaxin protein.

• WHAT QUESTION DID THIS STUDY ADDRESS?

mRNA‐0184 is a novel, investigational, lipid nanoparticle‐encapsulated mRNA therapy that encodes for human relaxin‐2 fused to variable light chain kappa (Rel2‐vlk). This study explored if the application of a translational semi‐mechanistic population pharmacokinetic/pharmacodynamic model, coupled with an efficacy study in non‐human primates, could guide dose selection of mRNA‐0184 for a phase I clinical trial of adult patients with stable heart failure with reduced ejection fraction.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This work further shows how a pharmacokinetic/pharmacodynamic model can be used to quantitatively translate preclinical pharmacology data regarding an investigational mRNA therapeutic agent presumed to be a minimally efficacious dose for a first‐in‐human study in patients with heart failure.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

While this application is specific to the novel mRNA‐0184, the employed principles and assumptions are relevant for the continued translation of pharmacology efforts to evaluate other mRNA therapeutics.

INTRODUCTION

Heart failure (HF) is a complex, progressive disorder caused by structural or functional abnormalities in the heart that result in reduced cardiac output. ¹ , ² Approximately 2% of the worldwide population has HF, with risk increasing with age. ³ , ⁴ Symptoms may include impaired systolic or diastolic dysfunction, dyspnea, fatigue, wheeze, anorexia, and delirium. ⁵ Although many treatment options are available, including angiotensin‐converting enzyme inhibitors, angiotensin receptor blockers, β‐blockers, mineralocorticoid receptor antagonists, and advanced device therapies, substantial morbidity and mortality are associated with HF. ⁶

Relaxin‐2 is a naturally occurring peptide hormone initially known for its muscle relaxant effects during pregnancy; it can also induce cardioprotective effects by influencing hemodynamics, including increasing cardiac output and decreasing systemic vascular resistance, moderating blood pressure and myocardial hypertrophy, and conferring anti‐inflammatory and anti‐fibrotic properties. ⁷ , ⁸ , ⁹ , ¹⁰ Therefore, relaxin‐2 may have potential therapeutic benefit in the treatment of patients with HF. ⁹ Several studies have previously evaluated serelaxin, a highly purified, synthesized, recombinant form of relaxin‐2, as a potential HF therapy. A phase II dose‐finding study and an initial phase III, randomized, placebo‐controlled trial of serelaxin in patients who were hospitalized for acute HF suggested beneficial effects on both dyspnea and clinical outcomes. ¹¹ , ¹² A subsequent phase III trial of serelaxin in patients hospitalized for acute HF demonstrated short‐term clinical improvements but no effect on mortality or overall disease progression, ¹³ which may be attributed to the short 7‐ to 8‐h half‐life of serelaxin. ¹⁴ Subsequent research efforts have investigated the optimization of pharmacokinetic (PK) properties to extend the half‐life and therapeutic potential of exogenous relaxin therapeutics, including fatty acid‐conjugated relaxin analogs, ¹⁵ single‐chain relaxin peptide mimetics, ¹⁶ and a monomeric relaxin analog with a serum albumin–binding heavy chain‐only variable domain. ¹⁵ , ¹⁶ , ¹⁷ While promising preclinical evidence supports investigations of these approaches to extend the half‐life of the resulting protein, clinical evidence has yet to be established.

To investigate the therapeutic potential of long‐acting relaxin in the treatment of patients with HF, mRNA‐0184, a novel, investigational, lipid nanoparticle (LNP)‐encapsulated mRNA therapy that encodes for human relaxin‐2 fused to variable light chain kappa (Rel2‐vlk) is currently under clinical development. In this report, a translational, semi‐mechanistic, population PK/pharmacodynamic (PD) model of Rel2‐vlk mRNA was developed. The objectives of this study were to assess and model the PK of Rel2‐vlk mRNA and PD of translated Rel2‐vlk protein in cynomolgus monkeys and to use the PK/PD model to guide mRNA‐0184 dose selection for a phase I clinical trial of adult patients with stable HF with reduced ejection fraction (NCT05659264).

METHODS

Data used for model development

The concentrations of Rel2‐vlk mRNA and Rel2‐vlk protein, measured at various timepoints up to 337 h after administration of a single dose of mRNA‐0184 (at 0.15, 0.5, and 1 mg/kg) infused over a period of 1 h to male cynomolgus monkeys (N = 4 per dose), were used to develop the model. Sample analysis for Rel2‐vlk mRNA and Rel2‐vlk protein was performed using pre‐qualified bDNA and ELISA assays, respectively. The LLOQ of Rel2‐vlk mRNA and Rel2‐vlk protein was 0.050 ng/mL and 20 pg/mL, respectively.

Cynomolgus monkeys were individually accommodated in stainless steel cages, designed to suit their species and size. The arrangement of cages facilitated visual and auditory interaction among the animals. All work sponsored by Moderna, Inc., was carried out in accordance with the Guide for the Care and Use of Laboratory Animals by the National Institutes of Health and ethical standards to ensure the safety, compliance, and welfare of research animals. Experimental protocols were approved by the Institutional Animal Care and Use Committee and complied with all relevant regulations regarding the use of research animals.

Statistical analysis of preclinical data

Descriptive statistics, including sample size (N), arithmetic mean, median, standard deviation, and/or standard errors, were calculated to quantify expression of Rel2‐vlk mRNA and Rel2‐vlk protein concentration in cynomolgus monkeys following mRNA‐0184 administration.

Model development

Model development was performed in a stepwise manner (Figure 1): (1) exploratory analyses of plasma concentration‐time profiles for Rel2‐vlk mRNA and Rel2‐vlk protein were conducted to identify trends in the experimental data and to guide model selection; (2) concentration‐time profiles of Rel2‐vlk mRNA in cynomolgus monkeys were modeled using a semi‐mechanistic PK model; and (3) Rel2‐vlk protein expression rate was modeled as a linear function of plasma Rel2‐vlk mRNA concentration using a hypothetical effect compartment to account for the delay in effect. To ensure stability of the model fitting, volume of distribution of Rel2‐vlk mRNA in plasma (V1) and volume of distribution of Rel2‐vlk in plasma (central) compartment (Vc) estimates were fixed based on a prior estimation. Subsequently, all other parameters were estimated. Model equations are provided in Table S1. It may be worthwhile to note that Rel2‐vlk protein expression is a result of interplay of multiple processes, with the rate‐limiting step being the cellular uptake of the Rel2‐vlk mRNA loaded LNPs, followed by diffusion of Rel2‐vlk mRNA from the LNPs into the cytoplasm, followed by translation to Rel2‐vlk protein. Due to inherent complexity of this process, the parameters pertaining to diffusion of mRNA from LNPs cannot be derived from the plasma data alone and a more mechanistic modeling (exploring LNPs translocation into the cell and its cellular kinetics) has been explored, ¹⁸ and is beyond the scope of this work.

FIGURE 1.

Semi‐mechanistic PK/PD model schematic of Rel2‐vlk mRNA and Rel2‐vlk protein distribution. The semi‐mechanistic mRNA‐0184 PK model consisted of plasma and tissue compartments, whereas a hypothetical effect compartment was used for the PD model to account for the delay in onset of the plasma concentration‐time profile of Rel2‐vlk mRNA following mRNA‐0184 dosing. IV, intravenous; K₁₂, distribution of mRNA between plasma and tissue compartments; K₂₀, elimination rate of Rel2‐vlk mRNA; K₂₃, K₃₂, intercompartmental distribution rates of Rel2‐vlk mRNA; K₅₀, first‐order elimination rate; K_prot, transfer rate constant of Rel2‐vlk protein to and from central and peripheral compartments, respectively; PD, pharmacodynamic; PK, pharmacokinetic; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa; slope, linear function for Rel2‐vlk protein product; V₁, volume of distribution of Rel2‐vlk mRNA in plasma; V₂, volume of distribution of Rel2‐vlk mRNA in tissue (target site); V _c, volume of distribution of Rel2‐vlk protein in plasma (central) compartment; V _p, volume of distribution of Rel2‐vlk protein in tissue (peripheral) compartment.

Population analyses and simulations were conducted using Phoenix NLME, version 8.3.4.295 (Certara; Princeton, NJ). The first‐order conditional estimation‐extended least squares algorithm method was employed for all model runs. Datasets and graphics were prepared using R Studio, version 2021.09.0 + 351 (Posit; Boston, MA). Assessments of model adequacy and final model selection were guided by goodness‐of‐fit criteria, including (1) visual inspection of diagnostic scatter plots (observed vs. predicted concentration; residual/weighted residual vs. predicted concentration or time); (2) successful convergence of the minimization routine; (3) plausibility of parameter estimates; (4) visual predictive (prediction corrected) checks; and (5) numerical diagnostics (i.e., model identifiability assessment, parameter precision, correlation between parameters, and ε‐ and η‐shrinkage).

Extrapolation of the PK/PD model to humans

The developed preclinical PK/PD model was extrapolated to humans using allometric scaling of volume and clearance parameters; scaled parameter estimates were calculated using Equation 1. A scaling coefficient of 1 was utilized for volume parameters, and 0.75 and 0.85 were used for clearance of Rel2‐vlk mRNA and Rel2‐vlk protein, respectively. The choice of 0.85 as a scaling coefficient for clearance of Rel2‐vlk protein was based on our literature review. Wherein a better correlation between the estimated human clearance based on cynomolgus monkey PK data and the observed human clearance was achieved with an allometric scaling exponent of 0.85 for CL for various therapeutic proteins. ¹⁹ Similar results were reported and acknowledged by others in the field. ²⁰ , ²¹

$$Y=a\times {\mathrm{BW}}^{\beta }$$ (1)

where Y = scaled parameters in humans; a = PK parameters in cynomolgus monkeys; BW = body weight ratio of non‐human primates and humans; and β = scaling coefficient (with varying magnitude for different parameters).

Model‐based simulations of mRNA‐0184 were then performed to guide selection of the dose for the first‐in‐human clinical study. The dose selection was guided by the average plasma concentration at steady state (Cavg_SS), which was calculated using Equation 2 and was based on target efficacious Rel2‐vlk protein exposure area under the effect curve (AUEC) associated with an improved cardiac function without influencing heart rate and blood pressure in aged, high‐fat induced non‐human primates with naturally developed cardiovascular and metabolic disease. The AUEC of Rel2‐vlk protein over a 2‐week interval (AUEC_SS) was estimated to be 486 ng/mL*h after weekly mRNA‐0184 dosing at 0.15 mg/kg in aged and obese cynomolgus monkeys to reflect HF comorbidity that showed cardiovascular improvement from the echocardiogram (data not shown).

$${\text{Cavg}}_{\mathrm{ss}}=\frac{{\text{AUEC}}_{\mathrm{ss}}}{\tau }$$ (2)

where Cavg_SS = average plasma concentration at steady state; AUEC_SS = area under the effect curve at steady state; and $\tau$ = dosing interval.

RESULTS

Pharmacokinetic model for mRNA‐0184

In the exploratory data analysis, a delayed or secondary peak in the plasma concentration‐time profiles was observed after once weekly intravenous administration of mRNA‐0184 in healthy cynomolgus monkeys (Figure S1). We believe that the emergence of second peak can be attributed to distribution and re‐distribution of the LNPs. This phenomenon has also been recognized by others in this field and a similar modeling approach has been utilized. ²² , ²³ Furthermore, this characteristic associated with LNPs has been evaluated using the simulation results from a minimal PBPK‐QSP model, by incorporating LNP recycling to identify conditions necessary for observing a second peak in mRNA pharmacokinetics (PK). Simulations predicted that with a fast recycling and slow tissue re‐uptake rates, a robust second peak is observed in the plasma mRNA concentration curve. ¹⁸ The observed time course of Rel2‐vlk mRNA distribution in plasma informed development of a semi‐mechanistic PK model, which consisted of plasma and tissue compartments (Figure 1). In this model, distribution and redistribution clearances are assumed to indicate tissue uptake and redistribution between plasma and tissue compartments, as well as elimination clearance from the tissue compartment. The model assumed (1) the rate of transfer to and from the tissue compartments are first‐order processes; and (2) that clearance of Rel2‐vlk mRNA from the target tissue occurred. To account for the time delay of the emergence of the second peak in the concentration profiles, the sum of Rel2‐vlk mRNA levels in the plasma compartments was incorporated into the model structure. The PK model converged successfully, with a reasonable percentage of standard error estimates. Parameter estimates are shown in Table 1. Visual predictive checks and goodness‐of‐fit plots are shown in Figures 2 and 3, respectively. Visual diagnostics suggested that the model adequately described the mRNA‐0184 PK in cynomolgus monkeys without systemic bias.

TABLE 1.

Model parameter estimates.

Parameter (unit)	Model estimate	SE (%)	Shrinkage (%)
PK model parameter estimates
tvV1 (mL)	112	Fixed	—
tvCL (mL/h)	258	25.3	—
tvCL2 (mL/h)	42.4	5.85	—
tvV2 (mL)	160	34.1	—
tvCL3 (mL/h)	9.86	1.22	—
ηCL	0.266	0.106	36.8
ηV2	1.06	0.253	23.9
ηCL3	0.171	0.034	16.6
Proportional residual error (%)	0.557	0.080	—
ε‐shrinkage (%)	12.3	—	—
mRNA‐0184‐Rel2‐vlk PK/PD model parameter estimates
tvKe0 (h−1)	0.193	0.050	—
tvVc (mL)	163	Fixed	—
tvVp (mL)	364	42.7	—
tvKprot (h−1)	25.5	5.21	—
tvK50 (h−1)	4.58	0.491	—
tvSlope (dimensionless)	0.540	0.106	—
ηK50	0.073	0.065	14.5
ηSlope	0.417	0.160	2.76
Proportional residual error	0.378	0.027	—
ε‐shrinkage	9.02	—	—

Note: V ₁ and V _c estimates were fixed based on a prior estimation run.

Abbreviations: CL and CL₃ intercompartmental clearance of Rel2‐vlk mRNA; CL₂, systemic clearance of Rel2‐vlk mRNA; h, hour; K₅₀, first‐order elimination rate; K_e0, transfer rate of Rel2‐vlk mRNA to/from hypothetical effect compartment; K_prot, distribution rate constant of Rel2‐vlk protein; PD, pharmacodynamic; PK, pharmacokinetic; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa; slope, linear function for Rel2‐vlk protein production; tv, typical value; V₁, volume of distribution of Rel2‐vlk mRNA in plasma; V₂, volume of distribution of Rel2‐vlk mRNA in tissue (target site); V_c, volume of distribution of Rel2‐vlk in plasma (central) compartment; V_p, volume of distribution of Rel2‐vlk in tissue (peripheral) compartment; η, interindividual variability.

FIGURE 2.

Visual predictive checks stratified by mRNA‐0184 dose (a) 0.15 mg/kg, (b) 0.5 mg/kg, and (c) 1.0 mg/kg confirm the estimated parameters are predictive of the non‐human primate PK observations with good fitting of the individual plots. Dashed green lines are the observed quantiles, 5%, 50%, and 95% (bottom to top). Solid red lines are the predicted quantiles 5%, 50%, and 95% (bottom to top). Blue circles are the observed data. h, hour; PK, pharmacokinetic.

FIGURE 3.

Goodness‐of‐fit plots for the mRNA‐0184 PK model for individual and population predictions adequately demonstrate a lack of systemic bias in the model. Observed Rel2‐vlk mRNA concentration are indicated by (a) individual and (b) population predictions. Conditional weighted residuals are captured by (c) time, (d) population, and (e) individual predictions. (f) Individual weighted residuals are shown by population predictions. Blue line indicates linear fit. Circles are observed data. Red line represents spline. h, hour; PK, pharmacokinetic; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa.

PK/PD model for Rel2‐vlk protein expression

The synthesis rate of Rel2‐vlk protein was modeled as a linear function of plasma Rel2‐vlk mRNA concentration via a hypothetical effect compartment. An effect compartment was added to describe the slightly delayed onset in the plasma concentration‐time profile of Rel2‐vlk mRNA observed following the dosing of mRNA‐0184. The biexponential decline of Rel2‐vlk protein concentrations observed in the exploratory analysis (Figure S1) were best described using a two‐compartment model consisting of plasma and tissue compartments (Figure 1). The PD model was based on the following assumptions: (1) Rel2‐vlk mRNA levels in the hypothetical effect compartment drive Rel2‐vlk protein production; (2) the delayed emergence of Rel2‐vlk protein in circulation (manifested as the maximum serum concentration [C _max]) is an interplay between the distribution and physiological mechanism; (3) the distribution rate constant of Rel2‐vlk protein is the same between the plasma and tissue compartments; and (4) the elimination of Rel2‐vlk protein is first order and is from the plasma compartment.

The PK/PD model converged with a reasonable percentage of standard error estimates, indicating the reliability of the values estimated for the parameters shown in Table 1. Visual diagnostic plots and individual fits for the model suggest that it adequately described the observed concentration‐time profiles of Rel2‐vlk protein expression (Figure 4). Based on the numerical diagnostics, model identifiability for the PK and PK/PD models was confirmed by the lack of sensitivity of model parameters to their respective initial estimates. Precision was confirmed if the asymptotic standard error of the estimated parameter was <50% ²⁴ (Table 1). In addition to the precision of the estimated parameters, model adequacy was confirmed by evaluating the correlation matrix of the estimated parameters. A lower correlation (<0.8) confirmed that the estimated parameters were independent and that there was no redundancy in explaining the data. Finally, ε‐ and η‐shrinkage, which refers to the residual error and inter‐individual variation in the model, respectively, was evaluated. A ε‐and η‐shrinkage of <40% further indicated model adequacy (Table 1). ²⁵

FIGURE 4.

Goodness‐of‐fit plots for the mRNA‐0184 PK/PD model for individual and population predictions adequately demonstrate a lack of systemic bias in the model. Observed Rel2‐vlk protein concentration are presented by (a) individual and (b) population predictions. Conditional weighted residuals are captured by (c) time, (d) population, and (e) individual predictions. (f) Individual weighted residuals are shown by population predictions. Blue line indicates linear fit. Circles are observed data. Red line represents spline. h, hour; PD, pharmacodynamic; PK, pharmacokinetic; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa.

Extrapolation of the PK/PD model to predict first‐in‐human starting dose

Model simulations to predict the first‐in‐human dose for mRNA‐0184 expressing Rel2‐vlk protein were guided by the parameter Cavg_ss which was calculated as 1.45 ng/mL (as described previously; Equation 2). Circulating levels of relaxin‐2 in humans have been reported in the literature and are observed to be elevated during stress to provide cardioprotection. ²⁶ Therefore, maintaining the Cavg_ss at a minimal level and within the reported levels of endogenous relaxin‐2 (i.e., 1–2.5 ng/mL – during various gestational period) during the planned first‐in‐human trial will ensure exposures are sufficient to provide cardioprotective effects (Figure 5). Additionally, an effort was made to include all reported benchmark relaxin‐2 levels (i.e., stabilized at 0.5 ng/mL – after the second trimester and the remainder of pregnancy; between 0.01 and 0.05 ng/mL in healthy men, women, and women during menopause; and finally, between 0.0067 and 0.0091 ng/mL in patients with HF) to highlight the progression of endogenous relaxin‐2 levels in different population. The estimated parameters in non‐human primates, scaling coefficient, and allometrically scaled estimates are shown in Table 2.

FIGURE 5.

Extrapolation of the PK/PD model to humans. (a) Overlay of predicted Rel2‐vlk protein concentrations in humans with observed serelaxin clinical data (30 μg/kg/day) ¹⁴ , ²⁷ and circulating levels of relaxin‐2 in humans reported in the literature are presented. ²⁶ (b) Simulated C _trough levels of Rel2‐vlk protein across various dose levels of mRNA‐0184 in humans are presented. In panel (a), the shaded light blue region represents the peak relaxin‐2 levels (1–2.5 ng/mL) during various gestational periods, the horizontal dashed line represents the stable relaxin‐2 levels after the second trimester and the remainder of pregnancy (0.5 ng/mL), the shaded green region represents the relaxin‐2 levels in healthy men, women, and women during menopause (0.01–0.05 ng/mL), the shaded bright blue region represents the relaxin‐2 levels in patients with HF (0.0067–0.0091 ng/mL), the solid dark blue line represents the median, the gray shaded region represents the 95% prediction interval, ²⁶ the vertical dashed line represents the region between 1008 to 1344 h used for AUCss, the red circles represent observed serelaxin levels in healthy volunteers 24 and 48 h following a single 30 μg/kg per day dose, ¹⁴ , ²⁷ and the black circles represent the observed serelaxin levels in patient volunteers 24 and 48 h following a single 30 μg/kg per day dose. ¹⁴ , ²⁷ In panel (b), the solid blue line represents the median, the shaded region represents the 95% prediction interval, and the red dotted lines represents the peak relaxin‐2 levels (1–2.5 ng/mL) during various gestation periods. ²⁶ AUCss, area under the curve at steady state; C _trough, trough concentration; h, hour; PD, pharmacodynamic; PK, pharmacokinetic; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa.

TABLE 2.

Human‐scaled parameter estimates for mRNA‐0184 PK/PD model simulation.

Parameter (unit)	Model‐estimated (or fixed) parameters (non‐human primates 2.5 kg)	Scaling coefficient	Scaled estimate (human 70 kg)
tvV1 (mL)	112	1	3136
tvCL (mL/h)	258	0.75	3144
tvCL2 (mL/h)	42.4		516
tvV2 (mL)	160	1	4490
tvCL3 (mL)	9.86	0.75	120
ηCL	0.266	—	0.266
ηV2	1.06	—	1.06
ηCL3	0.171	—	0.171
tvKe0 (h−1)	0.193	—	0.193
tvVc (mL)	163	1	4550
tvVp (mL)	364		10,182
tvKprot (h−1)	25.5	0.85	433
tvK50 (h−1)	4.58	0.85	77.7
tvSlope (dimensionless)	0.540	1	15.1
ηK50	0.073	—	0.073
ηSlope	0.417	—	0.417
Proportional residual error (Rel2‐vlk mRNA)	0.557	—	0.557
Proportional residual error (Rel2‐vlk protein)	0.378	—	0.378

Note: V₁ and V_c estimates fixed based on prior estimation run.

Abbreviations: CL and CL₃, intercompartmental clearance of Rel2‐vlk mRNA; CL₂, systemic clearance of Rel2‐vlk mRNA; K₅₀, first‐order elimination rate; K_e0, transfer rate of Rel2‐vlk mRNA to/from hypothetical effect compartment; K_prot, distribution rate constant of Rel2‐vlk protein; Rel2‐vlk, human relaxin‐2 fused to variable light chain kappa; slope, linear function for Rel2‐vlk protein production; tv, typical value; V₁, volume of distribution of Rel2‐vlk mRNA in plasma; V₂, volume of distribution of Rel2‐vlk mRNA in tissue (target site); V_c, volume of distribution of Rel2‐vlk in plasma (central) compartment; V_p, volume of distribution of Rel2‐vlk in tissue (peripheral) compartment; η, interindividual variability.

One of the hypotheses favoring the clinical development of mRNA‐0184 was the longer half‐life of the expressed Rel2‐vlk protein. A priori hypothesis testing was performed by comparing the model‐simulated profile of mRNA‐0184–induced Rel2‐vlk profiles with relaxin‐2 protein levels after the administration of serelaxin (single dose of 30 μg/kg per day) from phase III clinical trial data ¹⁴ , ²⁷ (Figure 5). Although the primary endpoints were not met in the serelaxin phase III trial, the benefits of comparing simulated Rel2‐vlk protein profiles with the serelaxin profile are two‐fold: first, mRNA‐0184–induced Rel2‐vlk protein expression has a longer half‐life (approximately 6–7 days) than serelaxin (6–8 h). Second, the inability of serelaxin to maintain the trough level of relaxin‐2 (e.g., Cavg_ss between 1 and 2.5 ng/mL, the potential exposure for cardioprotective action of relaxin‐2) may have also led to the failure to maintain therapeutic effects.

DISCUSSION

The global burden of HF imposes significant health and economic consequences that are expected to continue to increase over time; however, safe and effective HF treatments remain a significant unmet medical need. ¹ Previous studies have suggested that the peptide hormone relaxin‐2 may have potential therapeutic benefit for the treatment of patients with HF ¹¹ , ¹² ; however, development of a recombinant formulation has been limited by its short half‐life and the need for extended, 24‐ to 48‐h, intravenous administration. ¹⁴ , ²⁷ mRNA‐0184 is a novel investigational LNP‐mRNA therapy consisting of mRNA encoding Rel2‐vlk engineered as a fusion protein with an extended half‐life.

Initial focus was to build a translational PKPD from non‐human primate to extrapolate to human by leveraging both preclinical and literature data to guide dose selection through a model‐based approach to support decision‐making. This is the basis of model‐informed drug development (MIDD) to apply modeling to guide decision‐making and to continuously update the PKPD model with emerging data and/or knowledge through a learn and confirm paradigm. Translational PK/PD modeling and simulation frameworks are crucial and are encouraged by regulatory agencies to support rational dose selection in clinical trials and various phases of clinical development. The purpose of this investigation was to develop a translational PK/PD model to guide dose selection for a first‐in‐human clinical study of mRNA‐0184, an investigational LNP‐encapsulated mRNA therapy encoding Rel2‐vlk protein, for the treatment of patients with stable HF with reduced ejection fraction. We developed a translational semi‐mechanistic PK and PK/PD model to quantify the relationship between mRNA‐0184 dose, mRNA‐0184 PK, and mRNA‐0184–induced Rel2‐vlk protein expression (PD).

A stepwise approach was applied to develop the translational model, initially characterizing the PK of Rel2‐vlk mRNA and PD of Rel2‐vlk protein in non‐human primates. Our results demonstrated that the plasma PK of Rel2‐vlk mRNA was suitably described by a semi‐mechanistic PK model. This model of mRNA‐0184 is in part able to explain the emergence of the delayed second peak (suggesting distribution and redistribution) in the PK profile of Rel2‐vlk mRNA, which is attributed to LNP‐encapsulated modalities, and is consistent with the published literature. ²² , ²³

PD assessments of mRNA‐0184–induced Rel2‐vlk protein expression were adequately described using a linear function of plasma Rel2‐vlk mRNA concentration with a hypothetical effect compartment. The use of a hypothetical effect compartment to account for delays in peak drug plasma concentrations has been previously explored in the literature. ²⁸ , ²⁹ The delay associated with protein expression induced by LNP‐encapsulated mRNA has been acknowledged and modeled previously, ²² , ²³ wherein the hypothetical compartment was presumed to be hepatocytes, the site of Rel2‐vlk protein translation. Briefly, once in circulation, LNPs are rapidly taken up by both circulating monocytes and tissue‐resident macrophages (specifically hepatic macrophages). ³⁰ Modeling such processes is beyond the scope of this manuscript; for that reason, the use of a hypothetical effect compartment model to minimally explain such processes was considered. A subsequent PK/PD model adequately described the observed PK/PD for mRNA‐0184 in non‐human primates. This preclinical PK/PD model was subsequently extrapolated to humans based on allometric scaling of model parameters and was used to simulate PD responses at escalating doses. Allometric scaling predicted the Rel2‐vlk protein half‐life in humans to be 6 to 7 days and supported a recommended starting human dose of 0.025 mg/kg at 2‐week intervals for the subsequent clinical investigation of mRNA‐0184.

The translational PK/PD modeling approach affords more efficient patient enrollment and dose selection strategy for clinical trials, which may prevent potential exposure to suboptimal or supratherapeutic dose levels. ²² , ²³ , ³¹ These findings have informed the starting dose and dose range selection for the first‐in‐human clinical trial of mRNA‐0184 (NCT05659264) in adult patients with HF with reduced ejection fraction. This work further shows how a PK/PD model can be used to quantitatively translate preclinical pharmacology data from an investigational mRNA therapeutic agent to determine a safe starting dose for a first‐in‐human HF study aimed at enabling relaxin expression to mimic physiologic human protein expression levels. While this application is specific to mRNA‐0184, the employed principles and assumptions are relevant for the continued translation of pharmacology efforts to evaluate other mRNA therapeutics.

AUTHOR CONTRIBUTIONS

N.K., H.A., R.S., L.V., and M.L. wrote the manuscript; H.A. and R.S. designed the research; N.K., H.A., and M.J.I. performed the research; N.K., H.A., M.J.I., L.V., and M.L. analyzed the data; N.K., H.A., and M.L. contributed new reagents or analytical tools.

FUNDING INFORMATION

This study was funded by Moderna, Inc.

CONFLICT OF INTEREST STATEMENT

N.K., M.J.I., R.S., L.V., and M.L. are employees of Moderna, Inc., and hold stock/options in the company. H.A. and R.S. were employees of Moderna, Inc., during manuscript development and held stock/stock options in the company.

Supporting information

Data S1.

Kaushal N, Attarwala H, Iqbal MJ, Saini R, Van L, Liang M. Translational pharmacokinetic/pharmacodynamic model for mRNA‐0184, an investigational therapeutic for the treatment of heart failure. Clin Transl Sci. 2024;17:e13894. doi: 10.1111/cts.13894