Advertisement
Research ArticleDrug Discovery and Translational Medicine

Minimal Physiologically Based Pharmacokinetic-Pharmacodynamic (mPBPK-PD) Model of N-Acetylgalactosamine–Conjugated Small Interfering RNA Disposition and Gene Silencing in Preclinical Species and Humans

Vivaswath S. Ayyar, Dawei Song, Songmao Zheng, Thomas Carpenter and Donald L. Heald
Journal of Pharmacology and Experimental Therapeutics November 2021, 379 (2) 134-146; DOI: https://doi.org/10.1124/jpet.121.000805

Abstract

Conjugation of small interfering RNA (siRNA) to tris N-acetylgalactosamine [(GalNAc)3] can enable highly selective, potent, and durable knockdown of targeted proteins in the liver. However, potential knowledge gaps between in vitro experiments, preclinical species, and clinical scenarios remain. A minimal physiologically based pharmacokinetic-pharmacodynamic model for GalNAc-conjugated siRNA (GalNAc-siRNA) was developed using published data for fitusiran (ALN-AT3), an investigational compound targeting liver antithrombin (AT), to delineate putative determinants governing the whole-body–to-cellular pharmacokinetic (PK) and pharmacodynamic (PD) properties of GalNAc-siRNA and facilitate preclinical-to-clinical translation. The model mathematically linked relevant mechanisms: 1) hepatic biodistribution, 2) tris-GalNAc binding to asialoglycoprotein receptors (ASGPRs) on hepatocytes, 3) ASGPR endocytosis and recycling, 4) endosomal transport and escape of siRNA, 5) cytoplasmic RNA-induced silencing complex (RISC) loading, 6) degradation of target mRNA by bound RISC, and 7) knockdown of protein. Physiologic values for 36 out of 48 model parameters were obtained from the literature. Kinetic parameters governing (GalNAc)3-ASGPR binding and internalization were derived from published studies of uptake in hepatocytes. The proposed model well characterized reported pharmacokinetics, RISC dynamics, and knockdown of AT mRNA and protein by ALN-AT3 in mice. The model bridged multiple PK-PD data sets in preclinical species (mice, rat, monkey) and successfully captured reported plasma pharmacokinetics and AT knockdown in a phase I ascending-dose study. Estimates of in vivo potency were similar (∼2-fold) across species. Subcutaneous absorption and serum AT degradation rate constants scaled across species by body weight with allometric exponents of −0.29 and −0.22. The proposed mechanistic modeling framework characterizes the unique PK-PD properties of GalNAc-siRNA.

SIGNIFICANCE STATEMENT Tris N-acetylgalactosamine (GalNAc)3–conjugated small interfering RNA (siRNA) therapeutics enable liver-targeted gene therapy and precision medicine. Using a translational and systems-based minimal physiologically based pharmacokinetic-pharmacodynamic (mPBPK-PD) modeling approach, putative determinants influencing GalNAc-conjugated siRNA (GalNAc-siRNA) functionality in three preclinical species and humans were investigated. The developed model successfully integrated and characterized relevant published in vitro–derived biomeasures, mechanistic PK-PD profiles in animals, and observed clinical PK-PD responses for an investigational GalNAc-siRNA (fitusiran). This modeling effort delineates the disposition and liver-targeted pharmacodynamics of GalNAc-siRNA.

Introduction

Recent advances in RNA-based gene therapy for both rare and prevalent diseases are transforming precision medicine. By harnessing the natural biologic process of RNA interference (RNAi) (Hannon, 2002), an important novel class of RNA-based therapeutics is being developed. One such modality is small interfering RNA (siRNA), a therapy that mediates RNAi by acting upstream of most drugs, potently silencing mRNA—the genetic precursors—that encode for targeted proteins, thus preventing their synthesis.

Within cells, unbound cytoplasmic concentrations of siRNA act by engaging the RNA-induced silencing complex (RISC), inducing the RISC-mediated cleavage of the targeted mRNA (Hannon, 2002; Huang, 2017). Historically, efficient in vivo delivery of siRNA has been hampered by multiple factors such as rapid elimination (e.g., nuclease degradation, renal clearance), unfavorable physicochemical properties for cellular uptake (e.g., high molecular weight, large size, negative charge), potential for immunogenicity, and significant endo-lysosomal degradation (Wang et al., 2010; Dowdy, 2017). Recent breakthrough innovations to improve the pharmacokinetic (PK) and pharmacodynamic (PD) properties of siRNA include the use of enhanced stabilization chemistry for oligonucleotide synthesis, 2′-F and 2′-O-methyl nucleoside modifications to one strand of the siRNA duplex to lower immunogenicity, lipid nanoparticle formulations, and the development of tris N-acetylgalactosamine (GalNAc)3-conjugated siRNA (GalNAc-siRNA hereafter) (Layzer et al., 2004; Nair et al., 2014; Khvorova and Watts, 2017; Springer and Dowdy, 2018). With the regulatory approvals of two GalNAc-siRNA therapies, OXLUMO (lumasiran), and GIVLAARi (givosiran), and recent clinical successes of many other siRNA compounds, the clinical pipeline for GalNAc-siRNA therapy has increased substantially, with over a dozen late-stage trials to treat diverse liver-related diseases now underway (Khvorova and Watts, 2017; Springer and Dowdy, 2018; Setten et al., 2019; Debacker et al., 2020; Hu et al., 2020).

Administered via the subcutaneous route, siRNA conjugated to tris-GalNAc maintains potent and durable gene silencing activity in the clinic (e.g., weeks to months), which occurs through a series of kinetic steps involving 1) rapid and extensive distribution into the liver, 2) binding to the asialoglycoprotein receptor (ASGPR) that is expressed abundantly and exclusively on hepatocytes, 3) internalization via ASGPR-mediated endocytosis, 4) endosomal transport and receptor recycling, 5) escape from the endosomal compartment into cytoplasm (via mechanisms not fully understood yet), 6) binding to cytoplasmic RISC, 7) stimulation of the natural degradation rate of the targeted mRNA by the siRNA-loaded RISC, and 8) consequent inhibition of production of the targeted protein (Wei et al., 2011; Gilleron et al., 2013; Nair et al., 2017; Humphreys et al., 2019; Brown et al., 2020; Debacker et al., 2020). An interesting feature of the PK-PD relationship for siRNA is an apparent disconnect between their transient plasma exposure and prolonged pharmacologic effects. This unique PK-PD characteristic enables monthly, quarterly, or even twice-yearly dosing schedules to maintain therapeutic efficacy.

Although there has been a general understanding of a RISC-dependent mechanism for siRNA (Wei et al., 2011; Nair et al., 2017), quantitative information on determinants influencing the extent and duration of GalNAc-siRNA activity remains incomplete. As such, systematic and quantitative model-based interspecies assessments of GalNAc-siRNA disposition and gene silencing activity are warranted. Mechanism-based characterization of PK-PD relationships of siRNA present some challenges and unique opportunities due to the complex, cascading, and potentially nonlinear processes involved in their disposition and pharmacology. This necessitates the use of integrative and multiscale modeling approaches to delineate and quantify specific determinants controlling their dose-exposure-response relationships in vivo. Beyond enabling the separation of drug- and system-specific determinants, properly assembled translational, physiologically based PK-PD (PBPK-PD) models provide a mechanistic platform to extrapolate and predict drug effects from in vitro and in vivo preclinical systems to clinical responses for new and existing therapies (Mager and Jusko, 2008; Shah et al., 2012; Singh et al., 2020). Based on a recent review article originating from the US Food and Drug Administration (Fairman et al., 2021), and to our knowledge, a PBPK-PD framework for characterizing RNAi mediated by GalNAc-siRNA has not been reported thus far.

Here, we employed a stepwise approach to 1) develop a minimal PBPK-PD (mPBPK-PD) model to comprehensively characterize the whole-body–to-intracellular PK-PD actions of GalNAc-siRNA, 2) physiologically scale the platform model across three preclinical species (mouse, rat, and monkey), 3) characterize clinical response in a first-in-human study for a specific GalNAc-siRNA, and 4) demonstrate the utility of translational systems modeling to quantitate the PK-PD properties of this novel therapeutic platform. The model was developed by leveraging published data comprising in vitro experiments, comprehensive in vivo PK-PD studies conducted in preclinical species, and reported clinical phase I data for fitusiran (ALN-AT3), an investigational GalNAc-siRNA conjugate targeting endogenous antithrombin (AT) to treat hemophilia A and B (Pasi et al., 2017). Global sensitivity analysis (GSA) was further performed to understand the impact of selected model parameters on gene silencing. The proposed model provides a mechanistic framework that further assesses the preclinical and clinical PK-PD behavior of GalNAc-siRNA therapeutics.

Methods

Minimal PBPK-PD Model

Based upon the relevant physiologic and pharmacological mechanisms of GalNAc-siRNA disposition and actions described in the Introduction and found in the literature (Sarin, 2010; Bon et al., 2017; Nair et al., 2017; Humphreys et al., 2019; Debacker et al., 2020), a multiscale PK-PD model was developed (Fig. 1). The model was assembled in a stepwise manner. The definitions for all model parameters are provided within selected main and supplemental tables⇓. Definitions for state variables in the model are listed in Supplemental Table 1 in Supplemental Appendix 2. The system of ordinary differential equations describing the model is provided in Supplemental Appendix 3.

Fig. 1.
Fig. 1.

Schematic of the mPBPK-PD model for GalNAc-siRNA–mediated target gene knockdown. (A) An mPBPK model depicting the whole-body disposition of GalNAc-siRNA. All organs are represented by a rectangular compartment and connected in an anatomic manner with blood flow (solid arrows). Arrows represent the direction of the flow. Each tissue compartment within this model, except plasma and remainder, is divided into subcompartments. (B) The structure of the organ-level mPBPK-PD model for GalNAc-siRNA disposition and pharmacodynamics is shown. The liver tissue is divided into vascular, extracellular, and cellular subcompartments. Cellular (hepatocyte) uptake processes as relevant to ASGPR dynamics, GalNAc-binding, internalization, endosomal degradation, and cytoplasmic entry are shown (enlarged within dashed box). For a detailed description of the processes please refer to the Minimal PBPK-PD Model subsection in the Methods.

Minimal Physiologically Based Pharmacokinetic Model to Characterize Pharmacokinetics and Biodistribution of GalNAc-siRNA

The whole-body disposition of GalNAc-siRNA was characterized using a mPBPK modeling approach (Cao and Jusko, 2012) (Fig. 1A). The model incorporated a distinct kidney compartment subdivided into vascular and tissue spaces. A bound depot was included to account for drug sequestration in the extracellular and/or tissue space(s). Renal filtration (i.e., elimination) of the free (unbound) drug was assumed to occur from the kidney vascular space. The main organ of pharmacologic relevance, liver, comprised a vascular space, an extracellular compartment, a bound pool, and a cellular space accounting for drug entry into hepatocytes via specific, ASGPR-mediated intracellular uptake.

Liver mPBPK-PD Model to Characterize Target Gene Silencing by GalNAc-siRNA

Free (unbound) GalNAc-siRNA in the plasma entered from the liver vasculature into the extracellular space, controlled by the liver uptake clearance. Unbound siRNA either bound within a pool or interacted specifically with the ASGPR on hepatocytes. Tris-GalNAc binding to the ASGPR caused internalization of GalNAc-siRNA into endosomes followed by rapid cleavage of the GalNAc moiety, along with rapid and extensive recycling of a majority of free ASGPR (90%) back to the cell surface (Schwartz et al., 1982; Weigel and Oka, 1983). Furthermore, it was assumed that a very small fraction of siRNA (1%) underwent endosomal escape into the cell cytoplasm (Gilleron et al., 2013), bound to free cytoplasmic RISC, and consequently caused gene silencing by stimulating the natural degradation rate of target mRNA (Debacker et al., 2020) (Fig. 1B).

Cellular Model of Target-Mediated Uptake and Intracellular Delivery of GalNAc-siRNA

The kinetics of drug binding and uptake via the ASGPR was described by adapting the general framework of cell-level target-mediated drug disposition (Krzyzanski et al., 2016). The model incorporated the turnover of ASGPR on the cell surface (de novo synthesis and degradation), kinetics of tris-GalNac-siRNA binding with ASGPR, internalization of the bound complex into endosome, subsequent GalNAc cleavage from conjugate, endosomal degradation and/or escape of siRNA, and rapid recycling of free ASGPR from the endosome to the cell surface (Fig. 1B). Upon binding to ASGPR, GalNAc-siRNA was assumed to distribute within the total volume of hepatocytes, which comprise 80% of the cellular volume of liver (Zhou et al., 2016). Since adult hepatocytes display a long average life span of 200–300 days (Duncan et al., 2009), constant numbers of hepatocytes were assumed over time.

Each “scale” described within the mPBPK-PD model is related mathematically based on the underlying cellular and organ physiology. Physiologic parameter values describing organ volumes, plasma flows, filtration rates, liver fractional volumes, and hepatocellularity in mice, rats, monkeys, and humans were obtained from the literature (Davies and Morris, 1993; Sohlenius-Sternbeck, 2006; Duncan et al., 2009; Delanaye et al., 2012; Shah and Betts, 2012; Iwama et al., 2014; Sasaki et al., 2014). Liver concentrations of ASGPR in each species was calculated by accounting for species-specific cellular expression, hepatocellularity, and liver weight (Supplemental Appendix 4).

Model Assumptions

Certain additional assumptions were made during the modeling. These include that 1) elimination of GalNAc-siRNA in plasma is negligible, which is supported by in vitro stability assessments (Nair et al., 2017); 2) presence of functional ASGPR (i.e., expression of protein subunits 1 and 2) is confined to the hepatocytes (https://www.proteinatlas.org/ENSG00000141505-ASGR1/celltype; https://www.proteinatlas.org/ENSG00000161944-ASGR2/celltype); 3) drug distribution is confined within the vascular and extracellular volumes of “remainder” organs (Dowdy, 2017); 4) only “free” ASGPR (cleaved from conjugate) recycles to the cell surface; 5) only free siRNA escaping the endosome is able to interact with the RISC in the cytoplasm (Sen and Blau, 2005; Gilleron et al., 2013); 6) once bound, dissociation of siRNA from RISC is negligible (Bartlett and Davis, 2006); 7) RISC-loaded siRNA concentrations stimulate the natural degradation rate of target mRNA; and 8) knockdown of mRNA is translated to decreased protein expression.

Data Sets and a Stepwise Modeling Approach

All data used in this study were obtained from the literature. The PK data sets (plasma, total liver, RISC-bound siRNA in liver, and kidney concentration-time profiles) examined in this analysis include GalNAc-siRNA targeting AT, transthyretin (TTR), and delta-aminolevulinate synthase 1 that were collected across four species (mice, rats, cynomolgus monkeys, and humans), whereas mRNA and protein time courses were examined only for ALN-AT3, an investigational GalNAc-siRNA targeting AT, in mice, monkeys, and humans (Akinc et al., 2012; Sehgal et al., 2015; Pasi et al., 2017). Relevant information (including target, dose, species, measurements) regarding each preclinical and clinical study obtained from the literature is summarized in Table 1 and briefly stated below. A stepwise approach was applied to develop and qualify the platform model using experimental data obtained from in vitro (hepatocyte) and in vivo preclinical studies as well as clinical responses in humans. Due to availability of rich data sets describing both PK and PD properties, the mouse, a relatively well established preclinical animal to study GalNAc-siRNA pharmacology (Sehgal et al., 2015; Nair et al., 2017), was chosen as the starting species for model development and calibration. The stepwise model-building approach involved:

View this table:
TABLE 1

Summary of published preclinical and clinical in vivo data sets used to develop the proposed translational mPBPK-PD model.

Step 1 (In Vitro Modeling of ASGPR-Mediated GalNAc Uptake)

Two independent in vitro experiments examining kinetics of (GalNAc)3–interleukin 2 (IL -2) uptake in primary hepatocytes from mice (Sato et al., 2002) were modeled simultaneously using a basic model of receptor-mediated endocytosis (Supplemental Fig. 1A in Supplemental Appendix 1) to obtain select kinetic rate constants for GalNAc endocytosis by ASGPR.

Step 2 (mPBPK Model Development in Mice)

An mPBPK model incorporating distinct kidney and liver compartments was developed to characterize the available plasma and tissue PK profiles in mice. For PK model development, plasma and total tissue (liver and kidney) concentrations of GalNAc-siRNAs targeting AT (ALN-AT3) or TTR (siTTR-2) in mice were obtained from the literature at doses ranging from 1 to 25 mg/kg i.v. or s.c. (Akinc et al., 2012; Sehgal et al., 2015; Nair et al., 2017) (Table 1). Specifically, concentration-time profiles of siTTR-2 in plasma and kidney after 10 mg/kg i.v. and s.c. doses were obtained. In addition, both plasma and total liver concentration-time profiles for ALN-AT3 after 1, 2.5, 5, and 25 mg/kg single i.v. and/or s.c. doses were included.

Step 3 (mPBPK-PD Model Development and Validation in Mice)

Based upon availability of multiple, comprehensive published PK-PD data sets for ALN-AT3 (fitusiran), this compound was chosen as a relevant case example for full PK-PD characterization. In mice, the time courses of RISC occupancy (liver Ago2-loaded siRNA concentrations), AT mRNA in liver, and serum AT protein after 1, 2.5, and 5 mg/kg s.c. were used for further model development and characterization of the PK-PD relationship (Sehgal et al., 2015; Nair et al., 2017). In addition, a data set profiling the knockdown of serum AT protein after repeated 0.75, 1.5, and 3 mg/kg [once weekly (QW) × 5] s.c. dosing of ALN-AT3 in mice was used for model qualification (Sehgal et al., 2015).

Step 4 (mPBPK Model–Based Scale Up to Rats)

The proposed mPBPK-PD model was scaled based on physiologically based PK principles (Kagan et al., 2011; Ayyar and Jusko, 2020) to predict the pharmacokinetics of GalNAc-siRNA in higher-order species, including the rat. Model-based simulations in the rat, based on the model developed in Steps 1–3, were compared with available data (total liver PK at 10 mg/kg s.c. and RISC-bound siRNA at 1 and 5 mg/kg s.c.; Table 1) for givosiran (Lee et al., 2019). The RISC association rate constant (kon,RISC) and turnover constant (kDR) were assumed to be drug- and sequence-specific and were therefore estimated. All other model parameters were either fixed or scaled based on the model developed in Steps 1–3.

Step 5 (Modeling Knockdown in Nonhuman Primates)

Next, the platform model was applied to characterize the in vivo time profiles of plasma AT knockdown by ALN-AT3 in nonhuman primates. Relative AT activity in plasma after a single subcutaneous injection of ALN-AT3 at 1, 3, 10, and 30 mg/kg in cynomolgus monkeys were obtained from Sehgal et al. (2015). Two model parameters relating to pharmacodynamics, i.e., the in vivo drug potency (SC50) and the AT protein degradation rate constant, were estimated. All other model parameters were either fixed, scaled, or adjusted to species-specific values based on the model developed in Steps 1–3.

Step 6 (Modeling Plasma Pharmacokinetics and Knockdown in Humans)

Plasma pharmacokinetics and protein knockdown by fitusiran (ALN-AT3SC) were obtained from a published phase I dose-escalation study in four healthy subjects and 25 patients with moderate or severe hemophilia A or B (Pasi et al., 2017). Mean plasma concentrations of fitusiran over time after a single subcutaneous injection across a dose range of 0.015–1.8 mg/kg were investigated. In addition, the time course of plasma AT knockdown was characterized at 0.03 mg/kg s.c. (single dose) in healthy subjects and at 0.015 mg/kg QW × 3 s.c., 0.225 once monthly (Q4W) × 3 s.c., and 80 mg Q4W × 3 s.c. in subjects with hemophilia (Pasi et al., 2017). As a final model qualification step, AT knockdown profiles were simulated under 0.45 and 1.8 mg/kg Q4W × 3 s.c. regimens and compared with observed data. Two model parameters, namely, the subcutaneous absorption rate constant (ka) and SC50, were estimated, whereas the AT protein degradation rate constant was fixed to a reported value (Menache et al., 1990). The structural model remained unchanged across species, and all other model parameters were either fixed or scaled based on the model developed in Steps 1–3.

Parameter GSA

Considering the complexity of the underlying system, a GSA (Marino et al., 2008; Alden et al., 2013) was performed in which multiple drug- and system-specific parameters describing the model (Fig. 1) were simultaneously varied over time to deduce the overall uncertainty described in the final model output, i.e., overall protein knockdown. The partial rank correlation coefficient (PRCC) method was implemented in the R software environment (Singh et al., 2020), in which a total of 30 model parameters were simultaneously perturbed with a sampling size of 5000. The limits for lower and upper bounds varied for each parameter and were based on physiologically plausible values. PRCC-based sensitivity indexes were simulated as a time course for up to 40 days. PRCC describes the relative importance of a parameter along with its positive or negative correlation on the desirable model output.

Data Collection and Model Implementation

The data used in this report were obtained from the literature. The data were extracted as either mean values or individual observations (where available) from published graphs by computer digitization (WebPlotDigitizer, version 4.1; https://automeris.io/WebPlotDigitizer). Thus, estimated PK and PD parameters should be considered approximate and fit-for-purpose (i.e., proof of-concept). The PK-PD analysis was performed in ADAPT 5 (D'Argenio et al., 2009). Data were pooled and fitted using the maximum likelihood algorithm. The residual error variance model was defined as follows: Embedded Image where Vi is the variance of the ith observation, Yi is the ith model prediction, and σintercept and σslope are additive and proportional variance model parameters representing a linear relationship between the standard deviation of the model output and Yi. The goodness-of-fit was assessed by visual inspection of the fitted curves, improvement in the log-likelihood objective function, examination of residual plots (observed versus predicted and predicted versus residual), and precision (CV%) of the estimated parameters.

R (http://www.r-project.org), R Studio, and GraphPad Prism 8 (GraphPad Software, Inc., San Diego, CA) were used for GSA and graphical assessments, respectively.

Results

Modeling In Vitro Kinetics of GalNAc Endocytosis in Mice Hepatocytes (Step 1)

A basic model of receptor-mediated endocytosis (Supplemental Fig. 1A in Supplemental Appendix 1) simultaneously captured observed in vitro cellular disposition data from two separate experiments in mouse primary hepatocytes (Sato et al., 2002) reasonably well (Supplemental Fig. 1, B and C in Supplemental Appendix 1). The dissociation rate constant (koff) and internalization rate constant for ASGPR-mediated endocytosis were estimated as 0.022 minute−1 (23.9% CV) and 0.04 minute−1 (7.7% CV), which were subsequently fixed during the in vivo modeling (Table 2). To derive the association rate constant (kon) for (GalNAc)3 with ASGPR (Table 2), a reported value of 2.48 nM for the equilibrium dissociation constant of this interaction was used (Kanasty et al., 2013). An inherent assumption in utilizing these data of tris-GalNAc–IL-2 uptake within the current framework is that the entity conjugated to tris-GalNAc does not significantly influence the binding and internalization rates of the conjugate, which is instead driven primarily by the tris-GalNAc moiety itself.

View this table:
TABLE 2

mPBPK-PD model parameter values in mice.

mPBPK Model Development in Mice (Step 2)

The developed model (Fig. 1) was fitted simultaneously to multiple PK data sets obtained in mice, which included measurements of drug in plasma, liver (total and RISC-bound), and/or kidney for two compounds (ALN-AT3 and siTTR-2; Table 1). As depicted in Fig. 2 and in Supplemental Figs. 2 and 3 in Supplemental Appendix 1, the model characterized the concentration versus time profiles of both siRNA in plasma, liver, and kidney reasonably well across the broad dose range examined (1 mg/kg to 25 mg/kg i.v./s.c.). Several model parameters were fixed a priori to physiologically tenable estimates, whereas selected parameters (e.g., liver uptake clearance, endosomal degradation constant, and siRNA-RISC association constant), which were unknown or best identified from in vivo profiles, were estimated with acceptable precision (Table 2). The model captured well the initial steep, rapid drop in plasma siRNA concentrations after intravenous and subcutaneous dosing (Supplemental Figs. 2, A–C and 3, A and B in Supplemental Appendix 1), consistent with the short “apparent” plasma half-life typical for this class of compounds (Fairman et al., 2021). In addition, the model fits produced profiles which characterized a shallower, prolonged phase in plasma appearing at later times, which was evident particularly at the higher doses of 10 and 25 mg/kg i.v./s.c.. This phase may be attributable to some exchange of drug between plasma and kidney, which displayed extensive uptake and prolonged concentration-time profiles (though to a lesser extent than liver). The model-estimated apparent subcutaneous bioavailability in mice was ∼33% (5.0% CV) with a first-order ka value of 0.7 hour−1 (6.5% CV) (Table 2). The tissue-plasma partition coefficient for the remainder compartment was set as 0.1. This value was fixed because 1) owing to its larger size and net negative charge, GalNAc-siRNA do not freely permeate into the cell space within non–ASGPR-expressing tissues and hence is confined to extracellular fluid spaces, and 2) biodistribution studies indicate very low concentrations (relative to plasma exposure) in all tissues, except the liver and kidney (McDougall et al., 2021).

Fig. 2.
Fig. 2.

Model-based characterization of the concentration-time profiles of total liver pharmacokinetics (A) and RISC-loaded siRNA in liver (B) as well as time courses of liver AT mRNA (% baseline) (C) and serum AT (% baseline) (D) after a single dose of ALN-AT3 in mice at 1, 2.5, or 5 mg/kg s.c. Model simulation predicting serum AT after ALN-AT3 QW × 5 at 0.75, 1 and 3 mg/kg s.c. injections are shown (E), and observed data (Sehgal et al., 2015) are overlaid (F). Solid lines represent model-fitted or simulated profiles, whereas symbols indicate observed data points. Use of unique colors and symbols is described within the figure.

In general, liver concentration-time profiles of ALN-AT3 were captured reasonably well across 1–25 mg/kg s.c. doses (Fig. 2A and Supplemental Figure 2D in Supplemental Appendix 1), with slight exception for the liver Cmax at the highest dose. The rapid disappearance of drug in plasma (Supplemental Figs. 2, A–C and 3, A and B in Supplemental Appendix 1) was associated with efficient uptake into liver interstitium, with a net uptake clearance of 207 ml/h (8.5% CV). As expected, based on the physicochemical properties of GalNAc-siRNA, the model-estimated value of drug movement from liver tissue back into vascuature was extremely small (0.0025 ml/h; 24% CV), indicating minor to negligible drug re-entry into circulation after leaving the liver vasculature into tissue. Several parameters relating to ASGPR-mediated endocytosis of GalNAc-siRNA were fixed or calculated a priori based on published values (Supplemental Fig. 1 in Supplemental Appendix 1 and Table 1) (Schwartz et al., 1982; Sato et al., 2002; Prakash et al., 2014; Bon et al., 2017). Furthermore, the rate constant of cleavage of tris-GalNAc from the siRNA upon internalization was fixed as 1.32 hour−1, calculated from a reported 30-minute half-life associated with this step (Prakash et al., 2014). Of mechanistic relevance, simulation of the concentration-time profile of free surface ASGPR indicated that 1 mg/kg s.c. did not result in full occupancy of available ASGPR at any time, whereas 2.5 and 5 mg/kg s.c. doses produced a very transient ASGPR saturation in a dose-dependent manner (data not shown). On the other hand, a 25 mg/kg s.c. dose fully reduced free ASGPR over time (>48 hours after dosing), which may explain the less than dose-proportional increase in total liver Cmax and area under the curve at this high dose (Supplemental Fig. 2D in Supplemental Appendix 1).

The prolonged half-life of drug in liver was captured well, which was mostly governed by the degradation constant of siRNA in the endosomal compartment (0.01 hour−1; 3.0% CV), resulting in a prolonged profile for the escaped fraction of siRNA into the cytosol. To prevent overparameterization, two additional rate constants describing endosomal escape and cytoplasmic degradation of siRNA were fixed based on an imaging study tracking endosomal release of siRNA (Gilleron et al., 2013) and a local parameter sensitivity analysis, respectively (Table 2). As depicted in Fig. 2B, the concentration-time profile of RISC-bound ALN-AT3 in liver after 2.5 mg/kg s.c. (Nair et al., 2017) was well described. Consistent with reports of a prolonged occupancy profile for RISC (Wang et al., 2009; Deerberg et al., 2013), the bound complex, which comprised only ∼0.1–1% of total liver siRNA at any time, was characterized by a slow degradation rate constant (kDR; 0.005 hour−1; 21% CV).

Concentration-time profiles of drug in kidney were described using permeability-limited entry, incorporated a sequestered pool accounting for prolonged drug accumulation, and assumed renal filtration of free GalNAc-siRNA from the vascular space (Fig. 1A), which captured observed data well (Supplemental Fig. 3, C and D in Supplemental Appendix 1). The model-estimated fraction unbound in kidney of 0.07 (15.7% CV) was small in comparison with that in plasma (0.15) (Humphreys et al., 2019), indicating greater distribution into the kidney. The permeability flow rate of free drug into kidney was smaller than renal plasma flow, consistent with permeability rate–limited entry (Table 2).

mPBPK-PD Model Development and Validation in Mice (Step 3)

In addition to GalNAc-siRNA pharmacokinetics and RISC occupancy in liver, gene knockdown by ALN-AT3, as assessed by liver AT mRNA and serum AT protein (Sehgal et al., 2015; Nair et al., 2017), was integrated within the model framework and well characterized. Notably, modest onset delays (5–8 hours) in both mRNA and protein responses were accounted for by the mPBPK-PD model. Hysteresis among RISC-loaded siRNA, liver AT mRNA, and serum AT was captured well using tandem, precursor-like indirect response models (Ayyar et al., 2018), which accounted for turnover of AT mRNA [0.06 hour−1 (41% CV)] and protein [0.05 hour−1 (25% CV); Table 2]. The maximal stimulation of target mRNA degradation (Smax) was calculated based on the maximal knockdown observed and fixed, whereas the SC50 was estimated as 2.4 nM (6.5% CV; Table 2). Serum AT protein was reduced to a greater extent compared with its associated mRNA, which was captured empirically using an estimated γ value of 1.5 (5.6% CV).

The developed mPBPK-PD model was used to simulate the time course of AT protein knockdown by ALN-AT3 after multiple weekly doses in mice (0.75, 1.5, and 3 mg/kg QW × 5). Based on the established mPBPK-PD relationship, the simulations predicted a dose-dependent knockdown in AT ranging from 70% to 90% relative to baseline (Fig. 1E). These simulations were qualified upon comparison with mean observations from a published study in mice (Sehgal et al., 2015), which assessed AT knockdown under the same dosing regimen (Fig. 1F). Both the extent of knockdown and time to achieve steady-state reductions were predicted well, providing further confidence in the model's predictiveness.

Characterization of Givosiran Liver and RISC Pharmacokinetics in Rats (Step 4)

The developed mPBPK model was physiologically scaled from mice to rats. Available profiles of liver pharmacokinetics and RISC occupancy of givosiran in rats (Lee et al., 2019) were used for model evaluation. All parameters were either fixed or scaled [based on body weight (BW )–based allometric exponents] based on Step 2 (Supplemental Tables 2 and 3 in Supplemental Appendix 2), except kon,RISC and kDR, which are considered drug- and sequence-specific. As depicted in Fig. 3, the model successfully captured the concentration-time profile of givosiran in liver at 10 mg/kg s.c. (Fig. 3A) as well as the bound siRNA-RISC profile after 1 and 5 mg/kg s.c. (Fig. 3B). Consistent with the RISC profile for ALN-AT3 in mice, the peak of RISC-bound siRNA for givosiran was delayed compared with the peak in total liver pharmacokinetics, and this profile was captured reasonably well by the model. The model-estimated value for kon,RISC of givosiran (0.00016 nM−1·h−1; 15.1% CV) varied by 1.5-fold compared with that estimated for ALN-AT3 in mice (Table 2). This difference is plausible, and the results suggest that the current model may be extended to describe the kinetics of different GalNAc-siRNA compounds.

Fig. 3.
Fig. 3.

Model-based characterization of the concentration-time profiles of total liver pharmacokinetics at 10 mg/kg s.c. (A) and RISC-loaded siRNA in liver at 1 and 5 mg/kg s.c. (B) givosiran in rats. Solid lines represent model-fitted or simulated profiles, whereas symbols indicate observed data points. Use of unique colors and symbols is described within the figure.

Characterization of Serum AT Dynamics in Monkeys (Step 5)

Next, the platform model was scaled to characterize pharmacodynamic effects of ALN-AT3 in nonhuman primates (cynomolgus monkeys). All model parameters related to plasma, tissue, and intracellular pharmacokinetics were either fixed or scaled based on BW-based allometric exponents (Supplemental Tables 2 and 3 in Supplemental Appendix 2). Two parameters related to pharmacodynamics, namely, the SC50 and AT degradation rate constant, were estimated. As depicted in Fig. 4, the model characterized well the full reported time course and extent of reductions of approximately 50%, 70%, 80%, and >90% in AT protein levels after single subcutaneous doses of ALN-AT3 at 1, 3, 10, and 30 mg/kg (Sehgal et al., 2015). Of importance, the model inherently captured the delayed onset in response, evidenced at the first measured time point ∼48 hours after dosing, which is likely attributable to the cascade of relevant PK and PD processes, including liver uptake, slow cytoplasmic release, RISC loading, and natural turnover of mRNA and protein. Notably, the SC50 in cynomolgus monkeys was estimated to be reasonably similar compared with mice (i.e., ∼2-fold, 2.4 versus 5.3 nM). The AT degradation rate constant estimate was 5-fold lower in monkeys as compared with that in mice (see Fig. 7 and Discussion). Both parameters were estimated with good precision (Supplemental Table 3 in Supplemental Appendix 2).

Fig. 4.
Fig. 4.

Model-based characterization of plasma AT (% baseline) after single doses of ALN-AT3 in cynomolgus monkeys at 1, 3, 10, and 30 mg/kg s.c. Solid lines represent model-fitted profiles, whereas symbols indicate observed data (Sehgal et al., 2015). Use of unique colors and symbols is described within the figure.

PK and PD Characterization of ALN-AT3 (Fitusiran) in Humans (Step 6)

As a final step, published first-in-human study data for fitusiran (Pasi et al., 2017) were evaluated using the model developed through preclinical data. Similar to the approach taken in Step 4, nearly all model parameters related to plasma, tissue, and intracellular pharmacokinetics were either fixed or scaled (based on BW-based allometric exponents; Supplemental Tables 2 and 3 in Supplemental Appendix 2); only the ka for subcutaneous absorption was estimated based on the observed plasma profiles. Furthermore, the SC50 parameter relating to PD effect was estimated. As shown in Fig. 5A, the model was able to characterize the observed plasma pharmacokinetics of fitusiran across a broad range (0.015–1.8 mg/kg s.c.). In addition, selected PD data sets of serum AT reduction in healthy volunteers or in subjects with hemophilia at 0.03 mg/kg s.c. (single dose), 0.015 mg/kg QW × 3 s.c., 0.225 Q4W × 3 s.c., and 80 mg Q4W × 3 s.c. were characterized reasonably well by the model (Fig. 5B). Notably, there was some bias in capturing the exaggerated response at very low doses (reason currently unknown). To qualify the model in humans, AT knockdown profiles were simulated under 0.45 and 1.8 mg/kg Q4W × 3 s.c. regimens of fitusiran and compared with observed data (Pasi et al., 2017). The model simulations successfully predicted the mean AT knockdown response under these clinically relevant dosing regimens and over an extended duration (months) (Fig. 5C). Of translational relevance, estimated drug potencies were reasonably similar (i.e., within a 2-fold range) across mice, monkeys, and humans (Supplemental Table 3 in Supplemental Appendix 2).

Fig. 5.
Fig. 5.

Model-based characterization of the plasma concentration-time profiles of fitusiran (A) and the time course of plasma AT (% baseline) (B) after various subcutaneous dosing regimens in humans. Model simulations predicting plasma AT (% baseline) after fitusiran given Q4W × 3 at 0.45 and 1.8 mg/kg s.c., along with observed data (Pasi et al., 2017) overlaid, are shown (C). Use of unique colors and symbols is described within the figure.

Parameter GSA

Figure 6 describes a time-variant GSA on the mPBPK-PD model developed for ALN-AT3 (Fig. 1; Table 2) on percent protein knockdown. At early time points (e.g., <5 days), it was observed that parameters related to drug absorption, liver uptake, and endosomal degradation were sensitive to initial protein knockdown. As expected, parameters such as apparent subcutaneous bioavailability and liver uptake clearance were positively correlated with protein knockdown, whereas rate constants for intracellular degradation in liver, including endosomal and bound RISC, were negatively correlated. Furthermore, the GSA indicated that steps controlling siRNA degradation in liver (after entry into hepatocytes via ASGPR) remained sensitive at early times before gradually decreasing (e.g., endosomal degradation). This gradual decrease corresponded with a time-dependent increase in model sensitivity to the degradation rate constant for loaded RISC. This parameter, along with PD effect parameters (Smax, SC50, and γ), governed percent protein knockdown over time (see SC50; others not shown). These findings are consistent with the mPBPK-PD modeling results (see Fig. 2).

Fig. 6.
Fig. 6.

Results from the GSA on the developed mPBPK-PD model for ALN-AT3: PRCC-based sensitivity indexes of liver uptake clearance, subcutaneous bioavailability (F), SC50 (bound RISC), endosomal degradation, bound RISC degradation, and target protein degradation on target protein knockdown.

Discussion

Owing to their high selectivity for targeted RNA, siRNA-based therapeutics have witnessed an explosion of interest across academia and industry over the past decade. These compounds offer the ability to target previously undruggable human and viral genomes to selectively knock down disease-causing proteins (Dowdy, 2017). To date, three RNAi therapeutics, ONPATTRO, GIVLAARi, and OXLUMO, the latter two being GalNAc-siRNAs, are approved for the treatment of hereditary amyloidogenic transthyretin amyloidosis, acute hepatic porphyria, and primary hyperoxaluria type 1, respectively. Another GalNAc-siRNA, inclisiran, has been submitted for new drug application to the US Food and Drug Administration (Hu et al., 2020). As of 2020, seven siRNAs were undergoing phase III trials, and more candidates are under early development (Hu et al., 2020).

Despite tremendous successes, the quantitative impact of putative determinants of GalNAc-siRNA activity is only partially understood. Given its unique PK-PD characteristics in comparison with small molecule and protein-based therapeutics, it has been challenging to establish mechanistic PK-PD models for GalNAc-siRNA (Fairman et al., 2021), and there are, as yet, no established paradigms or defined guidelines that leverage preclinical data to predict safe and efficacious doses for GalNAc-siRNA in humans (Fairman et al., 2021). Transient plasma exposure accompanied by extensive uptake and prolonged residence within the liver is typically observed for this class. Rather than free (unbound) plasma concentrations, cytoplasmic concentrations of siRNA bound to RISC relate more closely with the time course of gene knockdown in target cells (Wei et al., 2011; Nair et al., 2017). As such, the development of mechanistic and translational systems PK-PD models could be a highly beneficial first step to integrate key factors associated with the whole-body–to-intracellular kinetics and activities of these agents.

Efforts in the mathematical modeling of RNAi have been published, providing in-depth insights into the complicated interplay of different components (e.g., Dicer, Ago2) and better understanding of RISC assembly and molecular biology (Wang et al., 2009; Wang et al., 2012; Deerberg et al., 2013). The PK-PD relationships of some antisense oligonucleotides have been adequately characterized using simpler models. For instance, a mechanism-based PK-PD model for an antisense oligonucleotide targeting hepatic apolipoprotein B mRNA in mice has been reported (Shimizu et al., 2015). Although the model characterized well preclinical PK-PD profiles in mice, its scope for translation (viz. PK-PD prediction in humans) seems limited. Recently, a compartment-based allometric PK modeling approach employing rodent liver pharmacokinetics (liver and RISC-bound profiles) was used to predict human RISC pharmacokinetics (Lee et al., 2019).

Here, we describe a stepwise approach to develop a first-in-kind mechanistic platform model for GalNAc-siRNA, which demonstrated preclinical to clinical translatability. The first step of our modeling approach leveraged reported data characterizing the kinetics of IL-2–tris-GalNAc conjugate binding and uptake to ASGPR in primary hepatocytes from mice (Sato et al., 2002). By modeling these profiles with a simple, fit-for-purpose model (Supplemental Fig. 1 in Supplemental Appendix 1), estimates for the first-order dissociation and internalization rate constants were derived. These data did not provide information to distinctly identify both the association rate constant as well as ASGPR density. Therefore, upon transitioning to the second step (i.e., mPBPK model development in mice), published values for system-specific parameters related to ASGPR (e.g., receptors per cell, degradation half-life, and recycled fraction) derived from in vivo assessments in mice were obtained and fixed a priori (Table 2). This approach not only conferred mechanistic confidence in the cellular- and tissue-related processes described but also minimized the issue of over-parameterization frequently encountered with mechanistic systems models. In Steps 2 and 3, the proposed model, in totality, well characterized multiple PK (plasma and tissue) and PD (mRNA and protein) data sets in mice (Fig. 2; Supplemental Figs. 2 and 3 in Supplemental Appendix 1) and successfully predicted PD response under different ALN-AT3 dosing regimens (Fig. 2, E and F). It is acknowledged that the model overpredicted plasma ALN-AT3 concentrations at two later time points across the 1–5 mg/kg range.

In Steps 4 and 5, the platform model was physiologically scaled from mice to rats and a higher-order preclinical species (i.e., cynomolgus monkeys; Supplemental Tables 2 and 3 in Supplemental Appendix 2). Key mechanistic aspects considered during interspecies extrapolation were physiologic flows and volumes, hepatocellularity, ASGPR expression and kinetics, intracellular elimination/turnover constants (endosomal, RISC, mRNA, and protein), intracellular binding constants, and the RISC-PD effect relationship (Smax and SC50; Supplemental Tables 2 and 3 in Supplemental Appendix 2). Since values for ASGPR expression were not found in rats and nonhuman primates, reported estimates in mice (1.8 million receptors per cell) (Bon et al., 2017) and in humans (1.1 million receptors per cell) (Miki et al., 2001) were employed for rats and monkeys, respectively. The affinity constants for (GalNAc)3 binding to ASGPR were assumed to be unchanged across species. In addition, fraction unbound in plasma as well as binding constants in both kidney and liver were fixed across species (Humphreys et al., 2019). Due to unavailability of suitable published PK and PD data sets for ALN-AT3 in rats, reported liver PK and RISC occupancy profiles of givosiran, an approved GalNAc-siRNA therapeutic, were used. Using this approach, the model captured both the liver and RISC PK profiles reasonably well (Fig. 3). Furthermore, the model characterized well the time course and extent of serum AT reduction in monkeys across 1–30 mg/kg s.c. (Fig. 4).

Finally (Step 6), the model was applied across a broad range of doses to describe observed plasma pharmacokinetics of fitusiran as well as associated serum AT reduction in humans. The model described the entire data set reasonably well (Fig. 5, A and B). There was some bias (underprediction) of PD response at very low doses. The exact reason(s) for misprediction at very low doses is unclear. It is speculated that the efficiency for hepatocyte uptake and/or endosomal escape could be higher (than predicted) at very low doses, which needs further evaluation through experiments. This might also possibly be a consequence of underestimating the true subcutaneous bioavailability (discussed below). Nonetheless, as shown in Fig. 5C, simulations of AT reduction at pharmacologically relevant doses successfully predicted the observed clinical response. Of translational relevance, drug potency was found to be similar (within ∼2-fold) across mice, monkey, and humans (Table 2; Supplemental Table 3 in Supplemental Appendix 2).

Interspecies comparison of two parameters, which were estimated during model fitting, was performed. The analysis revealed BW-based allometric relationships in ka and serum AT removal constant across the species examined (Fig. 7, A and B). The ka scaled with an exponent of −0.29 (R2 = 0.82), and the serum AT removal constant scaled with an exponent of −0.22 (R2 = 0.95). In both cases, these estimates scaled across body weight with exponents close to −0.25, consistent with allometric expectation (West and Brown, 2005). Model-fitted estimates for ka in rat and cynomolgus monkey were obtained based on plasma pharmacokinetics for other GalNAc-siRNA (data not shown). A similar exponent value of −0.35 has been reported for the subcutaneous absorption constant of human recombinant erythropoietin, a peptide of ∼30 kDa (Woo and Jusko, 2007). In addition, similar trends in plasma subcutaneous absorption profiles across rats, monkeys, and humans were shown for at least two other GalNAc-siRNA (McDougall et al., 2021). Whether interspecies differences in lymphatic physiology could explain the observed trends in ka after subcutaneous dosing requires further exploration.

Fig. 7.
Fig. 7.

Interspecies comparisons of parameter values for ka and plasma AT degradation constant (kdeg,p).

Three mechanistic hypotheses have been proposed to explain the prolonged liver half-life of GalNAc-siRNA (Humphreys et al., 2019). These include 1) existence of a “protein bound” depot within liver, 2) that the depot is a subcellular organelle such as the endosome, and 3) binding of siRNA to RISC in the cytosol and consequent “protection” of free siRNA from intracellular degradation (Humphreys et al., 2019). The processes considered within our mPBPK-PD model generally reflect such mechanisms. For example, the “bound pool” in liver, which is currently implemented in an empirical manner, may be rationalized by recent experimental findings (Brown et al., 2020), which show that acidic intracellular (endo-lysosomal) compartments can serve as a long-term depot for GalNAc-siRNA conjugates and contribute to the extended durability in tissue. Our model-based results and GSA suggest that intracellular (endosomal) degradation is a major determinant of liver half-life. Our findings also point to the importance of bound RISC kinetics, which also controls PD durability. However, since the RISC-bound siRNA comprises <1% of total liver siRNA at any given time after dosing, it does not significantly influence the overall PK half-life of total siRNA in liver.

A few theoretical and experimental considerations on the pharmacokinetics of GalNAc-siRNA require further discussion within the context of the present mPBPK modeling. Experimental considerations and challenges relating to transient saturation of ASGPR after intravenous dosing have been described (McDougall et al., 2021), which can impact the estimation of the true subcutaneous bioavailability of GalNAc-siRNA using conventional fitting to plasma pharmacokinetics after intravenous and subcutaneous dosing. In this work, estimation of bioavailability relied on simultaneous fitting of multiple plasma PK data sets after dosing in mice and/or monkeys. Included intravenous data were after relatively high doses of siRNA (10 and 25 mg/kg) in mice. Therefore, it is likely that our reported “apparent” subcutaneous bioavailability estimates are affected (i.e., reflect an underestimate of true subcutaneous bioavailability) resulting from an early, transient saturation of ASGPR after higher intravenous doses. Proper computation of the true absolute bioavailability may be achieved by conducting low-dose intravenous infusion studies, as an approach to reduce ASGPR saturation occurring with intravenous bolus dosing. Based on their size (∼15 kDa) and hydrophilic nature, renal clearance of GalNAc-siRNA was assumed to occur as a function of the plasma fraction unbound, glomerular filtration rate, and concentrations within the kidney vascular space (Embedded Image. Under these assumptions, the current model predicts that renal clearance, at pharmacologic doses, accounts for a relatively minor pathway of elimination at ∼11% of total clearance of GalNAc-siRNA in humans. Interestingly, the rates of renal clearance of inclisiran and lumasiran were found to approximate the total glomerular filtration rate without accounting for plasma fraction unbound (Wright et al., 2020). These findings possibly suggest that nonspecific binding of GalNAc-siRNA to plasma proteins may not effectively restrict glomerular filtration; however, more studies are needed.

Despite such considerations, the developed mPBPK model that integrated in vitro biomeasures, preclinical PK-PD data in animals, and clinical responses successfully characterized the unique PK-PD properties of GalNAc-siRNA therapeutics. It is particularly encouraging that our mathematical model–based demonstration of feasibility in translating the PK-PD properties of GalNAc-siRNA across species (based on allometric/physiologically based PK principles) was in general agreement with extensive absorption, distribution, metabolism, and excretion and PK-PD experiments that showed good predictability in preclinical to clinical translatability for this drug class (McDougall et al., 2021). Although the current model offers a mechanistic framework for understanding the clinical behavior of GalNAc-siRNA, additional mechanistic studies (Humphreys et al., 2019; Agarwal et al., 2021) are needed to fully decipher the binding and cellular disposition properties of this drug class, which contribute to their unique PK-PD properties. In vitro quantification of the cellular and subcellular disposition kinetics of GalNAc-siRNA (of varying chemistries) in viable experimental cell systems will enable further in vitro–in vivo bridging using more evolved mathematical models. Similarly, there remains considerable opportunity to employ in vitro stability and gene knockdown assays in hepatocytes from different species to inform in vivo potency a priori toward prediction of activity in humans (Basiri et al., 2020).

In summary, the approaches taken in this work demonstrate how several basic, well established processes governing pharmacokinetics (e.g., species physiology, binding, transport, target-mediated drug disposition), target engagement, and pharmacodynamics (e.g., mechanism of action, biomarkers, and turnover processes) can be assembled (Ayyar and Jusko, 2020) to establish integrated PK-PD models for novel therapeutics such as GalNAc-siRNA. The proposed framework is expected to promote mechanistic and quantitative reasoning to guide experimental designs during the preclinical and early clinical development of GalNAc-siRNA. This mechanistic systems model may also form the basis for characterizing the PK-PD properties of other ligand-conjugated oligonucleotide therapeutics.

Acknowledgments

The authors thank Drs. Honghui Zhou and Weirong Wang for critical review of this work. Drs. Hideo Makimura and Elisa Fabbrini are acknowledged for insightful discussions on translational and clinical aspects. V.S.A. and D.S. gratefully acknowledge Dr. William J. Jusko (University at Buffalo, SUNY) for introducing them to concepts of minimal PBPK and PD systems modeling of indirect gene-mediated effects, during their graduate research. These concepts were applicable and inspirational to this work.

Authorship Contributions

Participated in research design: Ayyar, Heald.

Conducted experiments: Ayyar, Song, Carpenter.

Contributed new reagents or analytic tools: Ayyar, Song, Zheng, Carpenter, Heald.

Performed data analysis: Ayyar, Song, Zheng, Carpenter.

Wrote or contributed to the writing of the manuscript: Ayyar, Song, Zheng, Carpenter, Heald.

Footnotes

    • Received June 22, 2021.
    • Accepted August 16, 2021.

  • ↵1 Current affiliation: Adagene Inc., San Diego, California, USA.

  • ↵2 Current affiliation: Tenacia CP BIO Consulting, Exton, Pennsylvania, USA.

  • ↵* V.S.A. and D.S. contributed equally to this work.

  • This work was supported by Janssen Research and Development, LLC.

  • https://doi.org/10.1124/jpet.121.000805.

  • ↵Embedded ImageThis article has supplemental material available at jpet.aspetjournals.org.

Abbreviations

ASGPR
asialoglycoprotein receptor
AT
antithrombin
GalNAc
N-acetylgalactosamine
BW
body weight
GalNAc-siRNA
GalNAc-conjugated siRNA
GSA
global sensitivity analysis
IL-2
interleukin 2
mPBPK-PD
minimal physiologically based pharmacokinetic-pharmacodynamic
PD
pharmacodynamic
PK
pharmacokinetic
PRCC
partial rank correlation coefficient
QW
once weekly
Q4W
once monthly
RISC
RNA-induced silencing complex
RNAi
RNA interference
siRNA
small interfering RNA
TTR
transthyretin

References

  1. ↵
    1. Agarwal S
    2. Allard R
    3. Darcy J
    4. Chigas S
    5. Gu Y
    6. Nguyen T
    7. Bond S
    8. Chong S
    9. Wu JT
    10. Janas MM
    (2021) Impact of serum proteins on the uptake and RNAi activity of GalNAc-conjugated siRNAs. Nucleic Acid Ther 31:309–315.
    OpenUrl
  2. ↵
    1. Akinc A
    2. Sehgal A
    3. Brodsky J
    4. Qin J
    5. Racie T
    6. Foster D
    7. Milstein S
    8. Barros S
    9. Hettinger J
    10. Panesar S et al.
    (2012) RNAi-mediated inhibition of a natural anticoagulant for the treatment of hemophilia, in Proceedings of the 8th Annual Meeting of the Oligonucleotide Therapeutics Society, pp A9–A10, Oligonucleotide Therapeutics Society, San Diego, CA.
  3. ↵
    1. Alden K
    2. Read M
    3. Timmis J
    4. Andrews PS
    5. Veiga-Fernandes H
    6. Coles M
    (2013) Spartan: a comprehensive tool for understanding uncertainty in simulations of biological systems. PLOS Comput Biol 9:e1002916.
    OpenUrl
  4. ↵
    1. Ayyar VS
    2. Jusko WJ
    (2020) Transitioning from basic toward systems pharmacodynamic models: lessons from corticosteroids. Pharmacol Rev 72:414–438.
    OpenUrlAbstract/FREE Full Text
  5. ↵
    1. Ayyar VS
    2. Sukumaran S
    3. DuBois DC
    4. Almon RR
    5. Jusko WJ
    (2018) Modeling corticosteroid pharmacogenomics and proteomics in rat liver. J Pharmacol Exp Ther 367:168–183.
    OpenUrlAbstract/FREE Full Text
  6. ↵
    1. Bartlett DW
    2. Davis ME
    (2006) Insights into the kinetics of siRNA-mediated gene silencing from live-cell and live-animal bioluminescent imaging. Nucleic Acids Res 34:322–333.
    OpenUrlCrossRefPubMed
  7. ↵
    1. Basiri B
    2. Xie F
    3. Wu B
    4. Humphreys SC
    5. Lade JM
    6. Thayer MB
    7. Yamaguchi P
    8. Florio M
    9. Rock BM
    (2020) Introducing an in vitro liver stability assay capable of predicting the in vivo pharmacodynamic efficacy of siRNAs for IVIVC. Mol Ther Nucleic Acids 21:725–736.
    OpenUrl
  8. ↵
    1. Bon C
    2. Hofer T
    3. Bousquet-Mélou A
    4. Davies MR
    5. Krippendorff BF
    (2017) Capacity limits of asialoglycoprotein receptor-mediated liver targeting. MAbs 9:1360–1369.
    OpenUrl
  9. ↵
    1. Brown CR
    2. Gupta S
    3. Qin J
    4. Racie T
    5. He G
    6. Lentini S
    7. Malone R
    8. Yu M
    9. Matsuda S
    10. Shulga-Morskaya S et al.
    (2020) Investigating the pharmacodynamic durability of GalNAc-siRNA conjugates. Nucleic Acids Res 48:11827–11844.
    OpenUrl
  10. ↵
    1. Cao Y
    2. Jusko WJ
    (2012) Applications of minimal physiologically-based pharmacokinetic models. J Pharmacokinet Pharmacodyn 39:711–723.
    OpenUrlCrossRefPubMed
    1. D’Argenio D
    2. Schumitzky A
    3. Wang X
    (2009) ADAPT 5 User’s Guide: Pharmacokinetic/Pharmacodynamic Systems Analysis Software, BioMedical Simulations Resource, Los Angeles, CA.
  12. ↵
    1. Davies B
    2. Morris T
    (1993) Physiological parameters in laboratory animals and humans. Pharm Res 10:1093–1095.
    OpenUrlCrossRefPubMed
  13. ↵
    1. Debacker AJ
    2. Voutila J
    3. Catley M
    4. Blakey D
    5. Habib N
    (2020) Delivery of oligonucleotides to the liver with GalNAc: from research to registered therapeutic drug. Mol Ther 28:1759–1771.
    OpenUrl
  14. ↵
    1. Deerberg A
    2. Willkomm S
    3. Restle T
    (2013) Minimal mechanistic model of siRNA-dependent target RNA slicing by recombinant human Argonaute 2 protein. Proc Natl Acad Sci USA 110:17850–17855.
    OpenUrlAbstract/FREE Full Text
  15. ↵
    1. Delanaye P
    2. Schaeffner E
    3. Ebert N
    4. Cavalier E
    5. Mariat C
    6. Krzesinski JM
    7. Moranne O
    (2012) Normal reference values for glomerular filtration rate: what do we really know? Nephrol Dial Transplant 27:2664–2672.
    OpenUrlCrossRefPubMed
  16. ↵
    1. Dowdy SF
    (2017) Overcoming cellular barriers for RNA therapeutics. Nat Biotechnol 35:222–229.
    OpenUrlCrossRefPubMed
  17. ↵
    1. Duncan AW
    2. Dorrell C
    3. Grompe M
    (2009) Stem cells and liver regeneration. Gastroenterology 137:466–481.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Fairman K
    2. Li M
    3. Ning B
    4. Lumen A
    (2021) Physiologically based pharmacokinetic (PBPK) modeling of RNAi therapeutics: opportunities and challenges. Biochem Pharmacol 189:114468.
    OpenUrl
  19. ↵
    1. Gilleron J
    2. Querbes W
    3. Zeigerer A
    4. Borodovsky A
    5. Marsico G
    6. Schubert U
    7. Manygoats K
    8. Seifert S
    9. Andree C
    10. Stöter M et al.
    (2013) Image-based analysis of lipid nanoparticle-mediated siRNA delivery, intracellular trafficking and endosomal escape. Nat Biotechnol 31:638–646.
    OpenUrlCrossRefPubMed
  20. ↵
    1. Hannon GJ
    (2002) RNA interference. Nature 418:244–251.
    OpenUrlCrossRefPubMed
  21. ↵
    1. Hu B
    2. Zhong L
    3. Weng Y
    4. Peng L
    5. Huang Y
    6. Zhao Y
    7. Liang X-J
    (2020) Therapeutic siRNA: state of the art. Signal Transduct Target Ther 5:101.
    OpenUrl
  22. ↵
    1. Huang Y
    (2017) Preclinical and clinical advances of GalNAc-decorated nucleic acid therapeutics. Mol Ther Nucleic Acids 6:116–132.
    OpenUrlCrossRefPubMed
  23. ↵
    1. Humphreys SC
    2. Thayer MB
    3. Lade JM
    4. Wu B
    5. Sham K
    6. Basiri B
    7. Hao Y
    8. Huang X
    9. Smith R
    10. Rock BM
    (2019) Plasma and liver protein binding of N-acetylgalactosamine-conjugated small interfering RNA. Drug Metab Dispos 47:1174–1182.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    1. Iwama R
    2. Sato T
    3. Sakurai K
    4. Takasuna K
    5. Ichijo T
    6. Furuhama K
    7. Satoh H
    (2014) Estimation of glomerular filtration rate in cynomolgus monkeys (Macaca fascicularis). J Vet Med Sci 76:1423–1426.
    OpenUrl
  25. ↵
    1. Kagan L
    2. Gershkovich P
    3. Wasan KM
    4. Mager DE
    (2011) Physiologically based pharmacokinetic model of amphotericin B disposition in rats following administration of deoxycholate formulation (Fungizone®): pooled analysis of published data. AAPS J 13:255–264.
    OpenUrlPubMed
  26. ↵
    1. Kanasty R
    2. Dorkin JR
    3. Vegas A
    4. Anderson D
    (2013) Delivery materials for siRNA therapeutics. Nat Mater 12:967–977.
    OpenUrlCrossRefPubMed
  27. ↵
    1. Khvorova A
    2. Watts JK
    (2017) The chemical evolution of oligonucleotide therapies of clinical utility. Nat Biotechnol 35:238–248.
    OpenUrlCrossRefPubMed
  28. ↵
    1. Krzyzanski W
    2. Harrold JM
    3. Wu LS
    4. Perez-Ruixo JJ
    (2016) A cell-level model of pharmacodynamics-mediated drug disposition. J Pharmacokinet Pharmacodyn 43:513–527.
    OpenUrl
  29. ↵
    1. Layzer JM
    2. McCaffrey AP
    3. Tanner AK
    4. Huang Z
    5. Kay MA
    6. Sullenger BA
    (2004) In vivo activity of nuclease-resistant siRNAs. RNA 10:766–771.
    OpenUrlAbstract/FREE Full Text
  30. ↵
    1. Lee J
    2. Goel V
    3. Agarwal S
    4. Melch M
    5. Guan A
    6. Li J
    7. Dennin S
    8. Arciprete M
    9. Robbie GJ
    (2019) Pharmacokinetic/pharmacodynamic analysis of givosiran in rats, in Proceedings of the Tenth American Conference on Pharmacometrics, pp. 2688–3953, International Society of Pharmacometrics, Bridgewater, NJ.
  31. ↵
    1. Mager DE
    2. Jusko WJ
    (2008) Development of translational pharmacokinetic-pharmacodynamic models. Clin Pharmacol Ther 83:909–912.
    OpenUrlCrossRefPubMed
  32. ↵
    1. Marino S
    2. Hogue IB
    3. Ray CJ
    4. Kirschner DE
    (2008) A methodology for performing global uncertainty and sensitivity analysis in systems biology. J Theor Biol 254:178–196.
    OpenUrlCrossRefPubMed
  33. ↵
    1. McDougall R
    2. Ramsden D
    3. Agarwal S
    4. Agarwal S
    5. Aluri K
    6. Arciprete M
    7. Brown CR
    8. Castellanos-Rizaldos E
    9. Charisse K
    10. Chong S et al.
    (2021) The nonclinical disposition and PK/PD properties of GalNAc-conjugated siRNa are highly predictable and build confidence in translation to man. Drug Metab Dispos DOI: 10.1124/dmd.121.000428 [published ahead of print].
    1. Menache D
    2. O’Malley JP
    3. Schorr JB
    4. Wagner B
    5. Williams C
    6. Alving BM
    7. Ballard JO
    8. Goodnight SH
    9. Hathaway WE
    10. Hultin MB et al
    ; Cooperative Study Group (1990) Evaluation of the safety, recovery, half-life, and clinical efficacy of antithrombin III (human) in patients with hereditary antithrombin III deficiency. Blood 75:33–39.
    OpenUrlAbstract/FREE Full Text
  35. ↵
    1. Miki K
    2. Kubota K
    3. Inoue Y
    4. Vera DR
    5. Makuuchi M
    (2001) Receptor measurements via Tc-GSA kinetic modeling are proportional to functional hepatocellular mass. J Nucl Med 42:733–737.
    OpenUrlAbstract/FREE Full Text
  36. ↵
    1. Nair JK
    2. Attarwala H
    3. Sehgal A
    4. Wang Q
    5. Aluri K
    6. Zhang X
    7. Gao M
    8. Liu J
    9. Indrakanti R
    10. Schofield S et al.
    (2017) Impact of enhanced metabolic stability on pharmacokinetics and pharmacodynamics of GalNAc-siRNA conjugates. Nucleic Acids Res 45:10969–10977.
    OpenUrl
  37. ↵
    1. Nair JK
    2. Willoughby JL
    3. Chan A
    4. Charisse K
    5. Alam MR
    6. Wang Q
    7. Hoekstra M
    8. Kandasamy P
    9. Kel’in AV
    10. Milstein S et al.
    (2014) Multivalent N-acetylgalactosamine-conjugated siRNA localizes in hepatocytes and elicits robust RNAi-mediated gene silencing. J Am Chem Soc 136:16958–16961.
    OpenUrlCrossRefPubMed
  38. ↵
    1. Pasi KJ
    2. Rangarajan S
    3. Georgiev P
    4. Mant T
    5. Creagh MD
    6. Lissitchkov T
    7. Bevan D
    8. Austin S
    9. Hay CR
    10. Hegemann I et al.
    (2017) Targeting of antithrombin in hemophilia A or B with RNAi therapy. N Engl J Med 377:819–828.
    OpenUrlCrossRef
  39. ↵
    1. Prakash TP
    2. Graham MJ
    3. Yu J
    4. Carty R
    5. Low A
    6. Chappell A
    7. Schmidt K
    8. Zhao C
    9. Aghajan M
    10. Murray HF et al.
    (2014) Targeted delivery of antisense oligonucleotides to hepatocytes using triantennary N-acetyl galactosamine improves potency 10-fold in mice. Nucleic Acids Res 42:8796–8807.
    OpenUrlCrossRefPubMed
  40. ↵
    1. Sarin H
    (2010) Physiologic upper limits of pore size of different blood capillary types and another perspective on the dual pore theory of microvascular permeability. J Angiogenes Res 2:14.
    OpenUrlCrossRefPubMed
  41. ↵
    1. Sasaki Y
    2. Iwama R
    3. Sato T
    4. Heishima K
    5. Shimamura S
    6. Ichijo T
    7. Satoh H
    8. Furuhama K
    (2014) Estimation of glomerular filtration rate in conscious mice using a simplified equation. Physiol Rep 2:e12135.
    OpenUrlAbstract/FREE Full Text
  42. ↵
    1. Sato H
    2. Kato Y
    3. Hayasi E
    4. Tabata T
    5. Suzuki M
    6. Takahara Y
    7. Sugiyama Y
    (2002) A novel hepatic-targeting system for therapeutic cytokines that delivers to the hepatic asialoglycoprotein receptor, but avoids receptor-mediated endocytosis. Pharm Res 19:1736–1744.
    OpenUrlPubMed
  43. ↵
    1. Schwartz AL
    2. Fridovich SE
    3. Lodish HF
    (1982) Kinetics of internalization and recycling of the asialoglycoprotein receptor in a hepatoma cell line. J Biol Chem 257:4230–4237.
    OpenUrlFREE Full Text
  44. ↵
    1. Sehgal A
    2. Barros S
    3. Ivanciu L
    4. Cooley B
    5. Qin J
    6. Racie T
    7. Hettinger J
    8. Carioto M
    9. Jiang Y
    10. Brodsky J et al.
    (2015) An RNAi therapeutic targeting antithrombin to rebalance the coagulation system and promote hemostasis in hemophilia. Nat Med 21:492–497.
    OpenUrlCrossRefPubMed
  45. ↵
    1. Sen GL
    2. Blau HM
    (2005) Argonaute 2/RISC resides in sites of mammalian mRNA decay known as cytoplasmic bodies. Nat Cell Biol 7:633–636.
    OpenUrlCrossRefPubMed
  46. ↵
    1. Setten RL
    2. Rossi JJ
    3. Han SP
    (2019) The current state and future directions of RNAi-based therapeutics. Nat Rev Drug Discov 18:421–446.
    OpenUrlCrossRefPubMed
  47. ↵
    1. Shah DK
    2. Betts AM
    (2012) Towards a platform PBPK model to characterize the plasma and tissue disposition of monoclonal antibodies in preclinical species and human. J Pharmacokinet Pharmacodyn 39:67–86.
    OpenUrlCrossRefPubMed
  48. ↵
    1. Shah DK
    2. Haddish-Berhane N
    3. Betts A
    (2012) Bench to bedside translation of antibody drug conjugates using a multiscale mechanistic PK/PD model: a case study with brentuximab-vedotin. J Pharmacokinet Pharmacodyn 39:643–659.
    OpenUrlPubMed
  49. ↵
    1. Shimizu R
    2. Kitade M
    3. Kobayashi T
    4. Hori S
    5. Watanabe A
    (2015) Pharmacokinetic-pharmacodynamic modeling for reduction of hepatic apolipoprotein B mRNA and plasma total cholesterol after administration of antisense oligonucleotide in mice. J Pharmacokinet Pharmacodyn 42:67–77.
    OpenUrl
  50. ↵
    1. Singh AP
    2. Zheng X
    3. Lin-Schmidt X
    4. Chen W
    5. Carpenter TJ
    6. Zong A
    7. Wang W
    8. Heald DL
    (2020) Development of a quantitative relationship between CAR-affinity, antigen abundance, tumor cell depletion and CAR-T cell expansion using a multiscale systems PK-PD model. MAbs 12:1688616.
    OpenUrlCrossRef
  51. ↵
    1. Sohlenius-Sternbeck AK
    (2006) Determination of the hepatocellularity number for human, dog, rabbit, rat and mouse livers from protein concentration measurements. Toxicol In Vitro 20:1582–1586.
    OpenUrlCrossRefPubMed
  52. ↵
    1. Springer AD
    2. Dowdy SF
    (2018) GalNAc-siRNA conjugates: leading the way for delivery of RNAi therapeutics. Nucleic Acid Ther 28:109–118.
    OpenUrlCrossRefPubMed
  53. ↵
    1. Wang D
    2. Zhang Z
    3. O’Loughlin E
    4. Lee T
    5. Houel S
    6. O’Carroll D
    7. Tarakhovsky A
    8. Ahn NG
    9. Yi R
    (2012) Quantitative functions of Argonaute proteins in mammalian development. Genes Dev 26:693–704.
    OpenUrlAbstract/FREE Full Text
  54. ↵
    1. Wang HW
    2. Noland C
    3. Siridechadilok B
    4. Taylor DW
    5. Ma E
    6. Felderer K
    7. Doudna JA
    8. Nogales E
    (2009) Structural insights into RNA processing by the human RISC-loading complex. Nat Struct Mol Biol 16:1148–1153.
    OpenUrlCrossRefPubMed
  55. ↵
    1. Wang J
    2. Lu Z
    3. Wientjes MG
    4. Au JLS
    (2010) Delivery of siRNA therapeutics: barriers and carriers. AAPS J 12:492–503.
    OpenUrlCrossRefPubMed
  56. ↵
    1. Wei J
    2. Jones J
    3. Kang J
    4. Card A
    5. Krimm M
    6. Hancock P
    7. Pei Y
    8. Ason B
    9. Payson E
    10. Dubinina N et al.
    (2011) RNA-induced silencing complex-bound small interfering RNA is a determinant of RNA interference-mediated gene silencing in mice. Mol Pharmacol 79:953–963.
    OpenUrlAbstract/FREE Full Text
  57. ↵
    1. Weigel PH
    2. Oka JA
    (1983) The large intracellular pool of asialoglycoprotein receptors functions during the endocytosis of asialoglycoproteins by isolated rat hepatocytes. J Biol Chem 258:5095–5102.
    OpenUrlAbstract/FREE Full Text
  58. ↵
    1. West GB
    2. Brown JH
    (2005) The origin of allometric scaling laws in biology from genomes to ecosystems: towards a quantitative unifying theory of biological structure and organization. J Exp Biol 208:1575–1592.
    OpenUrlAbstract/FREE Full Text
  59. ↵
    1. Woo S
    2. Jusko WJ
    (2007) Interspecies comparisons of pharmacokinetics and pharmacodynamics of recombinant human erythropoietin. Drug Metab Dispos 35:1672–1678.
    OpenUrlAbstract/FREE Full Text
  60. ↵
    1. Wright RS
    2. Collins MG
    3. Stoekenbroek RM
    4. Robson R
    5. Wijngaard PLJ
    6. Landmesser U
    7. Leiter LA
    8. Kastelein JJP
    9. Ray KK
    10. Kallend D
    (2020) Effects of renal impairment on the pharmacokinetics, efficacy, and safety of inclisiran: an analysis of the ORION-7 and ORION-1 studies. Mayo Clin Proc 95:77–89.
    OpenUrl
  61. ↵
    1. Zhou Z
    2. Xu M-J
    3. Gao B
    (2016) Hepatocytes: a key cell type for innate immunity. Cell Mol Immunol 13:301–315.
    OpenUrlCrossRefPubMed
Back to top

In this issue

Download PDF
Article Alerts
Email Article
Citation Tools

Research ArticleDrug Discovery and Translational Medicine

Ayyar et al.

Vivaswath S. Ayyar, Dawei Song, Songmao Zheng, Thomas Carpenter and Donald L. Heald
Journal of Pharmacology and Experimental Therapeutics November 1, 2021, 379 (2) 134-146; DOI: https://doi.org/10.1124/jpet.121.000805
Reddit logo Twitter logo Facebook logo Mendeley logo

Jump to section

Advertisement