Inferring Half-Lives at the Effect Site of Oligonucleotide Drugs

Abstract

Knowledge of the kinetics of the active drug in biophase, that is, at the effect site, is fundamental to select dose and to reason about safety. Unfortunately, the kinetics is cumbersome to measure in vivo. We describe how dose–response–time (DRT) analysis estimates the biophase and the target–response half-lives from data of the circulating protein of the encoded messenger RNA for seven antisense oligonucleotides (ASOs) and four small interfering RNA (siRNA) drugs. The biophase half-lives were estimated with acceptable precision (relative standard error <26%). For ASOs, the estimates were similar to, or slightly longer than, the reported terminal plasma half-lives. Terminal plasma half-life was reported for only one siRNA, precluding any general comparison. The estimated half-lives of response were 0.5–12 days cross drugs and shorter than the biophase half-lives. We recommend DRT analysis when limited plasma pharmacokinetic data are available, or when the biophase half-life differs from the terminal plasma half-life.

Introduction

Oligonucleotide drugs, such as antisense oligonucleotides (ASOs) and small interfering RNAs (siRNAs), are normally administered by parenteral injection, either intravenous infusion or subcutaneous injection. For ASOs, which are single stranded, the plasma pharmacokinetic (PK) profile is commonly characterized by a marked distribution phase during the first hours to day(s), followed by an extended terminal phase with a half-life in the order of days to several weeks [1,2]. siRNAs are double stranded containing a sense and an antisense strand, of which the latter elicits the pharmacological effect [3,4]. The plasma half-life of intact siRNA is highly dependent on formulation and chemical stability, and reported values range from hours to days [5,6].

Oligonucleotide drugs are mainly distributed to liver and kidney but also to a lower extent to tissues such as muscle, lung, adipose, adrenal gland, and peripheral nerves [6 –10]. The terminal plasma half-life of oligonucleotides depends on the sum of redistribution of drug from tissues [1]. For ASOs that act in tissue, the plasma terminal half-life can constitute a surrogate measure for target tissue half-life. For siRNAs, mouse data indicate that the concentration of Ago2-loaded antisense siRNA is strongly associated with, and likely drives, functional activity [3,4]. The kinetics (eg, half-life) of Ago2-loaded siRNA may differ from that of the total siRNA in target tissue [5].

In this work, we aim to estimate the kinetics of the active drug in biophase, that is, at the effect site [11,12]. The biophase kinetics guides choice of dose and dosing schedule for clinical studies and is fundamental for reasoning of safety-related aspects. In vivo, it is usually infeasible to measure the drug biophase half-life. However, it is often feasible to measure the protein of the encoded messenger RNA (mRNA), especially if the protein is circulating in the blood. The protein level reflects drug–target engagement.

It is well known that temporal pharmacodynamic (PD) response data alone, without measured drug concentrations, contain information about the drug's biophase kinetics (half-life), the turnover characteristics of response (turnover rate, half-life of response), and the PD characteristics (potency) [11]. Therefore, so called dose–response–time (DRT) analysis constitutes an alternative to obtain an estimate of biophase half-life when exposure data are sparse or missing. Of particular interest for oligonucleotide drugs is to infer the biophase half-life from drug–target engagement data. To our knowledge, DRT methodologies are not routinely applied in the oligonucleotide area, and there is no peer-reviewed publication assessing DRT modeling for oligonucleotides.

The estimated biophase half-lives cannot be strictly validated from currently available data. However, for ASOs acting in tissue, the estimated biophase half-life can be compared with the terminal plasma half-life, which reflects the tissue half-life. For siRNAs, the estimated biophase half-life represents the antisense half-life at target site. Mouse data indicate that the half-life of Ago2-loaded antisense siRNA reflects antisense half-life at target site [5]. The relationship between antisense half-life at target site and half-life of intact siRNA in plasma in higher species remains to be established.

The objective of this work is to assess the use of DRT modeling to predict biophase kinetics and PDs of oligonucleotides. The article first describes the modeling, and then demonstrates its use on data from oligonucleotide phase 1 studies. We conclude with a discussion of the main results obtained and the methods used.

Materials and Methods

Data from 11 clinical phase 1 studies were used in this assessment: IONIS ANGPTL3-LRx, IONIS APO(a)Rx, IONIS-APO(a)-LRx, Mipomersen (ISIS 301012), Volanesorsen (ISIS 304801, IONIS-APOCIIIRx), Inotersen (ISIS 420915)/IONIS-TTRRx, ISIS-FXI RX (ISIS 416858), all from Ionis Pharmaceuticals [13 –21], and ALN-PCS02, Inclisiran (ALN-PCSsc), Patisiran (ALN-TTR02), and ALN-TTRSC02, all from Alnylam Pharmaceuticals [22 –25]. First, public target–engagement biomarker data measured over time, from multiple dose levels, were collected and digitized. Then, a mathematical dose–response model was fitted to each data set. The model structure is depicted in Fig. 1A, and has been used before on a wide range of data sets (see, eg, Gabrielsson et al. [11], and Gabrielsson and Peletier [12]).

FIG. 1.

(A) Schematic overview of dose–response time model used in the assessment. The administered dose (1) rapidly enters the drug biophase (2). The drug is eliminated by first-order kinetics, and the biophase half-life is one of the key estimated parameters. The drug in biophase inhibits the build-up of the target biomarker (3) through an inhibitory Hill function. The target biomarker degrades by first-order kinetics. (B) Data (crosses) and model fits (solid lines) for the 11 studies considered in this assessment. The y-axes represent relative protein levels and the x-axes represent time in days. Dose groups of each study are separated by color and the level (mg) of each dose is indicated by digits. The dosing schedules differ (from one acute dose to multiple doses with various schedules) and we refer to the original references for full detail. For PCSK9 siRNA sc, the doses 125 and 250 mg were assigned the colors of 100 mg (pink) and 300 mg (purple), respectively, to improve readability. siRNA, small interfering RNA. Color images available online at www.liebertpub.com/nat

The model consists of an apparent PK part with linear uptake to, and linear elimination from, the drug biophase (D) as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} D \prime ( t ) = A \times {k_a} \times {e^{ - {k_a} \times t}} - {k_e} \times D ( t ) , \tag{1} \end{align*} \end{document}

where t denotes time, k_a and k_e are the absorption and elimination rates of the drug in biophase, and A is the administered dose. Parameter k_a was assumed rapid in comparison with other rates in the system [26], and was set to 10 day⁻¹.

The response (R) level was represented by an indirect-response model where the drug inhibits the production rate (k_in) of the protein, and where the response was assumed to be cleared by first-order kinetics (with rate k_out) as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} R \prime ( t ) = {k_{in}} \times ( 1 - I ( D ) ) - {k_{out}} \times R ( t ). \tag{2} \end{align*} \end{document}

Inhibition I(D) was modeled by an inhibitory Hill function as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} I ( D ) = { I_ { \max } } { \frac { { D^n } } { E { D_ { 50 } } ^n + { D^n } } } , \tag { 3 } \end{align*} \end{document}

where I_max represents the maximum inhibitory potential, ED₅₀ (mg) corresponds to the biophase drug level at 50% inhibition, and n is the Hill exponent. We note that the DRT model can account for potential time delay between drug in biophase and response. For sufficiently rich time series data, ideally from several dose groups, one may identify the elimination rate from the drug biophase, the turnover rate of the target, and the potency (ED₅₀). In the literature, the ED₅₀ has often been reported as amount per dosing frequency. This measure is easily derived by multiplying ED₅₀ by k_e and by the dosing frequency, for example to weekly dosing as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \[E{D_{50}}(mg/week) = E{D_{50}}(mg) \times {k_e}(1/day) \times 7{\mkern 1mu} day/week.{\text{ }} \tag{4}\] \end{document}

Throughout this analysis we use the time-independent ED₅₀.

The biophase half-life (t_1/2,biophase) and the PD half-life (t_1/2,PD) were estimated as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {t_{1 / 2 , biophase}} = \ln 2 / {k_e} , \tag{5} \end{align*} \end{document}

and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {t_{1 / 2{ ,\; {PD}}}} = \ln 2 / {k_{out}}. \tag{6} \end{align*} \end{document}

The model was fitted to experimental data using Phoenix WinNonlin 7.0 (Pharsight Certara, Princeton, NJ) using the naive-pooled method. A multiplicative error model was used ( \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${E_{obs}} = E \times ( 1 + res )$$ \end{document} , where E_obs is the observed response, E is the model prediction, and res is the residual error). For each data set, we systematically tested several alternative model instances by fixing either the Hill factor or the I_max parameter, or both. Model discrimination was based on precision of parameter estimates, visual diagnostics, correlation of parameter estimates, and the log-likelihood ratio as implemented in Phoenix WinNonlin. For I_max, if the estimated value was >1, the parameter was fixed to 1 and the model was re-estimated. Model simulations were done in Matlab (R2014b; The MathWorks, Inc., Natick, MA).

For IONIS ANGPTL3-LRx, Inotersen IONIS-TTRRx, ALN-PCS02, Inclisiran (ALN-PCSsc), and Patisiran (ALNTTR02), all discrimination criteria consistently selected the same model. For IONIS Apo(a)Rx, IONIS-APO(a)-LRx, Mipomersen (ISIS 301012), Volanersen (IONISI-APOCIIIRx), and ISIS-FXI (ISIS 416858), the biophase half-life was occasionally highly correlated with one of the other parameters. The correlation indicates that the model may be unreliable and, therefore, we selected the best model that had an acceptable low (<0.88) such correlation. For studies wherein clinical PK data were publicly available, the reported terminal plasma PK half-lives were collated to allow comparison with the estimated biophase half-lives from the DRT analysis.

Results and Discussion

The proposed model structure was sufficiently flexible to fit data from the 11 studies (Fig. 1B; Table 1). In most cases, the best model was found after fixing either the Hill factor to 1 or the I_max parameter to 1, or both (Table 1). Parameters were generally estimated with acceptable precision (relative standard error [RSE] <25%).

Table 1.

Dose–Response Parameter Estimates (% Relative Standard Error) Based on Phase 1 Clinical Data for the 11 Studies Considered in This Assessment

Compound	Target	Chemistry	Elimination rate k_e (1/day) (% RSE)	Potency ED₅₀ (mg) (% RSE)	Hill factor n (% RSE)	I_max (% RSE)	k_out (1/day) (% RSE)	Estimated biophase t_1/2 (days) (% RSE)	Estimated target t_1/2 (days) (% RSE)	Reported terminal PK t_1/2 (days)	Reference
IONIS ANGPTL3-LRx	ANGPTL3	ASO	0.0288 (18)	25.4 (32)	1.13 (21)	0.933 (7.3)	0.211 (58)	24.0 (18)	3.28 (58)	29^a	Graham et al. [13]
IONIS APO(a)Rx	Apo(a)	ASO	0.0193 (11)	433 (21)	1 (fix)	1 (fix)	0.108 (4.6)	35.9 (11)	6.41 (4.6)	23	Tsimikas et al. [14]
IONIS-APO(a)-LRx	Apo(a)	ASO	0.0179 (16)	19.2 (14)	1 (fix)	1 (fix)	0.126 (10)	38.9 (16)	5.48 (10)	Not found	Viney et al. [16]
Mipomersen (ISIS 301012)	ApoB apolipoprotein (apo) B-100	ASO	0.0197 (23)	1185 (25)	1 (fix)	1 (fix)	0.0597 (13)	35.2 (23)	11.6 (13)	31	Geary et al. [9,15]
Volanesorsen (ISIS 304801, IONIS-APOCIIIRx)	ApoCIII	ASO	0.0237 (4.3)	547 (11)	1 (fix)	1 (fix)	0.0909 (16)	29.2 (4.3)	7.61 (16)	19	Gaudet et al. [19]
Inotersen (ISIS 420915)/IONIS-TTRRx	TTR	ASO	0.0184 (12)	651 (20)	1.99 (15)	0.900 (7.9)	0.209 (16)	37.7 (12)	3.30 (16)	19	Ackermann et al. [32]
ISIS-FXI RX (ISIS 416858)	FXI	ASO	0.0220 (14)	562 (11)	1.50 (9.0)	1 (fix)	0.131 (14)	31.5 (14)	5.30 (14)	Not found	Büller et al. [17,18]; Liu et al. [20]
ALN-PCS02	PCSK9	siRNA	0.128 (20)	0.0238 (24)^b	1 (fix)	0.679 (6.0)	1.34 (14)	5.39 (20)	0.518 (14)	Not found^c	Fitzgerald et al. [22]
Inclisiran (ALN-PCSsc)	PCSK9	siRNA	0.00869 (26)	54 (42)	1 (fix)	0.858 (4.4)	0.135 (14)	77.5 (26)	5.1 (14)	Not found^d	Fitzgerald et al. [23]
Patisiran (ALN-TTR02)	TTR	siRNA	0.0624 (12)	0.0527 (26)^b	1.67 (13)	1 (fix)	0.361 (4.8)	11.1 (12)	1.92 (4.8)	1.6–2.5^e	Coelho et al. [24]
ALN-TTRSC02	TTR	siRNA	0.00781 (25)	4.61 (77)	0.866 (30)	1 (fix)	0.0857 (16)	88.6 (25)	8.08 (16)	Not found	Gollob [25]

The ED₅₀ expressed as amount per dosing frequency can be derived by Equation (4).

PK t_1/2 reported as 3–5 weeks.

ED₅₀ as mg/kg all other in mg.

Description of half-life measurements in appendix, but no data reported [22].

Specified on clinicaltrial.gov, but not reported.

Based on 14–28 days of sampling after last dose [27].

ASO, antisense oligonucleotide (Ionis Pharmaceuticals); fix, value fixed in the model, and not estimated; PK, pharmacokinetic; RSE, relative standard error; siRNA, small interfering RNAs (Alnylam Pharmaceuticals).

One exception was the ED₅₀ of ALN-TTRSC02 with an RSE of 77%. One reason for the uncertain estimate might have been that the number of free model parameters was too high. We investigated this by re-estimating the parameters using a fixed Hill factor of 1. The RSE fell <20% for all parameters, although goodness-of-fit decreased. Importantly, however, the estimated biophase and target half-lives were robustly estimated, as they changed by <8% when fixing the Hill coefficient. Another exception was k_out of IONIS ANGPTL3-LRx with an RSE of 58%, mainly due to high variability in data during the initial protein decline as well as during washout. Fixing other parameters in the model fitting did not improve the precision of target turnover (ie, k_out).

In addition to these two exceptions, there are five parameter estimates given in Table 1 with RSE >25%. However, these are primarily ED₅₀ values and are restricted to the range 26%−42%. Consequently, the secondary half-life parameters could be estimated with acceptable precision (RSE <25%) for most studies. For the biophase half-life, Inclisiran (ALN-PCSsc) was borderline with RSE of 26%. For target half-life, IONIS ANGPTL3-LRx was an outlier with RSE of 58%, due to the uncertain estimate of k_out discussed previously.

For ASOs that act in tissue, it is relevant to compare the reported terminal plasma half-lives with the estimated biophase half-lives, since the first traditionally have been used as a surrogate measure for tissue half-life, and since the biophase half-life likely reflects tissue half-life. Interestingly, we observed a reasonable agreement of the half-lives, but with a tendency that the biophase half-lives were slightly longer (but not more than twofold) than the reported terminal plasma half-lives.

For siRNAs, one cannot generally assume similar plasma half-lives and biophase half-lives. Among the four analyzed drugs, we only found the reported PK plasma half-life for ALN-TTR02. In this study, the biophase half-life was estimated to 11 days, based on data up to day 70 after dose, whereas the reported plasma half-life was between 1.6 and 2.5 days [24,27], based on 14–28 days of plasma sampling after last dose. There is some support in the literature that the biophase concentration that we estimate reflects the Ago2-loaded antisense siRNA [4,5]. We investigated this further by analyzing digitized single-dose mice data from Nair et al. [5]. We estimated the Ago2-loaded antisense siRNA half-life to 6–8 days, and DRT analysis of mRNA data from the same study estimated the biophase half-life to 4–5 days. Our analysis hence supports that Ago2-loaded antisense can be the driver for functional activity.

The estimated half-lives of response had a median of 5.9 days (range 0.52−12 days). The estimated biophase half-lives had a median of 35 days (range 5.4–89). For each drug, the estimated biophase half-life was longer than the estimated half-life of response (Table 1). Thus, the duration of early clinical studies is primarily governed by the biophase half-life.

The analysis is based on reported aggregate data digitized from publications. To get the most out of this kind of analysis, individual data and a population modeling approach are preferred. In particular, this is important if the group sizes vary within a study. In our case, the number of individuals in each dosing group varied between 3 and 16 across all studies, whereas the number of individuals did not vary more than twofold between groups within each study. We also note that all drugs used herein target the liver. The DRT methodology is applicable regardless of target tissue, but the results presented regarding parameter precision and half-lives may not necessarily be representative for other tissues.

The choice of a one-compartment representation of the drug in biophase may seem simplistic as the plasma kinetics are well known to follow multiple-compartment kinetics. However, a comprehensive PK model of data from monkey exposed to a triantennary N-acetyl galactosamine (GalNAc3)-conjugated oligonucleotide compound indicates approximate linear kinetics from the two tissue compartments [26]. Linear decline of drug exposure in tissue was also observed for non-GalNAc ASOs in kidney cortex, spleen, and liver in monkey [1] and the liver in human [9]. Consistent with this, a PK model of liver data from monkeys exposed to a nucleotide compound also indicates linear elimination in tissue [28].

A valid objection to these model predictions is that tissues were not sampled in the first hours after dose administration. However, densely sampled liver exposure data from mouse and rat indicate linear kinetics for ASOs (29). For siRNAs, there are similar trends in the mouse although data are less dense [5,6,10]. Naturally, one cannot exclude a concentration peak in human tissue immediately upon dosing. The height of such a potential peak would likely be lower than the corresponding peak height in plasma (due to the time delay from circulation to tissue), and, would be further reduced by tissue accumulation in case sufficiently frequent dosing is applied. A complicating fact is that distribution within a tissue can be nonhomogenous, something that is not revealed by analysis of whole-tissue homogenate [29]. Taken together, the assumption of monophasic tissue exposure for ASOs is reasonable given current data.

The chosen PD model structure, that is, turnover model, is one of the most commonly used empirical models for describing PDs. Ideally, the turnover rate should be drug independent. However, in practice and in particular for data sets lacking detailed PK data, the turnover rate is a composite of, for example, drug binding including recruitment by the RNA-induced silencing complex, turnover rate of RNA, and turnover rate of protein, and, therefore, drug dependent. This can explain differences in reported half-lives of response for drugs acting on the same target. The k_out parameter will mainly be determined by the slowest process of the composite system.

For the data sets used in this publication, we observed that the turnover rates of PCSK9 (for the drugs ALN-PCS02 and Inclisiran (ALN-PCSsc)) and TTR (for the drugs Inotersen, Patisiran, and ALN-TTRSC02) differed between the drugs. For both PCSK9 and TTR, we evaluated the impact of estimating a common k_out instead of having drug-specific k_out parameters. We obtained significantly inferior model fits with a common k_out, indicating that the turnover rates likely have a drug-specific component.

An alternative to the turnover model could be to use an effect-compartment model, where the concentration in the effect compartment is coupled to the biophase compartment by a first-order equilibrium rate constant, and where the PD depends on the effect-compartment concentration. We tried this model structure on our data sets. For the ASO data, the effect-compartment approach generally led to inferior model fits but similar estimates of tissue and response half-lives as those reported by the indirect-response model. For the siRNA data, the effect-compartment approach encountered numerical instabilities leading to unreliable parameter estimates (eg, highly correlated parameters or negative half-lives).

Based on visual assessment, the data sets did not have the characteristic pattern of a receptor-occupancy model (ie, the maximum response did not occur earlier at high doses compared with low doses); therefore, this type of model was not evaluated. Another alternative is to use mechanistic models should there be sufficient prior knowledge, and amount and quality of data (see, eg, Peng et al. [8] and Attarwala et al. [30,31]).

It is known that the PK and biodistribution of oligonucleotides are largely sequence independent within a chemical class [9]. This implies that the cumulative data from previous studies, and meta-analyses of their data, can be highly useful when designing new studies and when interpreting new data. We believe that the present analysis can contribute to this knowledge building.

We conclude that

DRT analysis of data from oligonucleotide drugs can be effectively used to assess the biophase half-life with acceptable precision when temporal target–engagement biomarker data are available.

The estimated biophase half-lives of ASOs were generally similar to, or potentially slightly longer than, the reported terminal plasma half-lives.

For siRNAs, DRT can estimate the biophase half-life of the active drug at target site, a measure that is not accessible from plasma PK.

The estimated half-lives of response were 0.5–12 days cross drugs and shorter than the biophase half-lives days, and shorter than the PK half-lives.

Footnotes

Acknowledgment

The authors thank Viktor Sokolov at M&S Decision for providing the digitized data for ALN-PCS02.

Author Disclosure Statement

The authors are employed by AstraZeneca.

References

Geary

. (2009). Antisense oligonucleotide pharmacokinetics and metabolism. Expert Opin Drug Metab Toxicol, 5:381–391.

, Grundy

, Henry

, Kim

, Norris

, Burkey

, Wang

, Vick

and Geary

. (2015). Predictive dose-based estimation of systemic exposure multiples in mouse and monkey relative to human for antisense oligonucleotides with 2′-o-(2-methoxyethyl) modifications. Mol Ther Nucleic Acids, 4:e218.

Crooke

, Witztum

, Bennett

and Baker

. (2018). RNA-targeted therapeutics. Cell Metab, 27:714–739.

Foster

, Brown

, Shaikh

, Trapp

, Schlegel

, Qian

, Sehgal

, Rajeev

, Jadhav

, et al. (2018). Advanced siRNA designs further improve in vivo performance of GalNAc-siRNA conjugates. Mol Ther, 26:708–717.

Nair

, Attarwala

, Sehgal

, Wang

, Aluri

, Zhang

, Gao

, Liu

, Indrakanti

, et al. (2017). Impact of enhanced metabolic stability on pharmacokinetics and pharmacodynamics of GalNAc-siRNA conjugates. Nucleic Acids Res, 45:10969–10977.

Park

, Park

, Pei

, Xu

and Yeo

. (2016). Pharmacokinetics and biodistribution of recently-developed siRNA nanomedicines. Adv Drug Deliv Rev, 104:93–109.

Hung

, Xiao

, Peralta

, Bhattacharjee

, Murray

, Norris

, Guo

and Monia

. (2013). Characterization of target mRNA reduction through in situ RNA hybridization in multiple organ systems following systemic antisense treatment in animals. Nucleic Acid Ther, 23:369–378.

Peng

, Andrews

, Nestorov

, Brennan

, Nicklin

and Rowland

. (2001). Tissue distribution and physiologically based pharmacokinetics of antisense phosphorothioate oligonucleotide ISIS 1082 in rat. Antisense Nucleic Acid Drug Dev, 11:15–27.

Geary

, Norris

, Yu

and Bennett

. (2015). Pharmacokinetics, biodistribution and cell uptake of antisense oligonucleotides. Adv Drug Deliv Rev, 87:46–51.

10.

Gupta

. (2015). Drug metabolism and pharmacokinetic (DMPK) properties of siRNA-GalNAc conjugates. DIA/FDA Oligonucleotide Based Therapeutic Conference. http://www.alnylam.com/web/assets/DIA_DMPK_09092015.pdf Accessed August 8, 2018.

11.

Gabrielsson

, Jusko

and Alari

. (2000). Modeling of dose–response-time data: four examples of estimating the turnover parameters and generating kinetic functions from response profiles. Biopharm Drug Dispos, 21:41–52.

12.

Gabrielsson

and Peletier

. (2014). Dose–response-time data analysis involving nonlinear dynamics, feedback and delay. Eur J Pharm Sci, 59:36–48.

13.

Graham

, Lee

, Brandt

, Tai

, Fu

, Peralta

, Yu

, Hurh

, Paz

, et al. (2017). Cardiovascular and metabolic effects of ANGPTL3 antisense oligonucleotides. N Engl J Med, 377:222–232.

14.

Tsimikas

, Viney

, Hughes

, Singleton

, Graham

, Baker

, Burkey

, Yang

, Marcovina

, et al. (2015). Antisense therapy targeting apolipoprotein(a): a randomised, double-blind, placebo-controlled phase 1 study. Lancet, 386:1472–1483.

15.

Geary

, Baker

and Crooke

. (2015). Clinical and preclinical pharmacokinetics and pharmacodynamics of mipomersen (kynamro((R))): a second-generation antisense oligonucleotide inhibitor of apolipoprotein B. Clin Pharmacokinet, 54:133–146.

16.

Viney

, van Capelleveen

, Geary

, Xia

, Tami

, Yu

, Marcovina

, Hughes

, Graham

, et al. (2016). Antisense oligonucleotides targeting apolipoprotein(a) in people with raised lipoprotein(a): two randomised, double-blind, placebo-controlled, dose-ranging trials. Lancet, 388:2239–2253.

17.

Büller

, Bethune

, Bhanot

, Gailani

, Monia

, Raskob

, Segers

, Verhamme

, Weitz

and FXI-ASO TKA Investigators. (2015). Factor XI antisense oligonucleotide for prevention of venous thrombosis. N Engl J Med, 372:232–240.

18.

Büller

, Gailani

and Weitz

. (2015). Factor XI antisense oligonucleotide for venous thrombosis. N Engl J Med, 372:1672.

19.

Gaudet

, Alexander

, Baker

, Brisson

, Tremblay

, Singleton

, Geary

, Hughes

, Viney

, et al. (2015). Antisense inhibition of apolipoprotein C-III in patients with hypertriglyceridemia. N Engl J Med, 373:438–447.

20.

Liu

, Dessouki

, E, J Grundy, Monia

and Bhanot

. (2011). ISIS-FXIRx, a novel and specific antisense inhibitor of factor XI, caused significant reduction in FXI antigen and activity and increased aPTT without causing bleeding in healthy volunteers. Blood, 118:209.

21.

Crooke

. (2014). IONIS PHARMACEUTICALS ISIS-FXIRx Program Update Webcast. http://slideplayer.com/slide/3856572/ Accessed August 8, 2018.

22.

Fitzgerald

, Frank-Kamenetsky

, Shulga-Morskaya

, Liebow

, Bettencourt

, Sutherland

, Hutabarat

, Clausen

, Karsten

, et al. (2014). Effect of an RNA interference drug on the synthesis of proprotein convertase subtilisin/kexin type 9 (PCSK9) and the concentration of serum LDL cholesterol in healthy volunteers: a randomised, single-blind, placebo-controlled, phase 1 trial. Lancet, 383:60–68.

23.

Fitzgerald

, White

, Borodovsky

, Bettencourt

, Strahs

, Clausen

, Wijngaard

, Horton

, Taubel

, et al. (2017). A highly durable RNAi therapeutic inhibitor of PCSK9. N Engl J Med, 376:41–51.

24.

Coelho

, Adams

, Silva

, Lozeron

, Hawkins

, Mant

, Perez

, Chiesa

, Warrington

, et al. (2013). Safety and efficacy of RNAi therapy for transthyretin amyloidosis. N Engl J Med, 369:819–829.

25.

Gollob

. (2017). Patisiran, an investigational RNAi therapeutic for the treatment of hereditary transthyretin-mediated (hATTR) amyloidosis. http://www.alnylam.com/wp-content/uploads/2017/08/2017-0803-TTR-RNAi-Roundtable-FINAL.pdf Accessed August 8, 2018.

26.

, Gunawan

, Post

, Zanardi

, Hall

, Burkey

, Kim

, Graham

, Prakash

, et al. (2016). Disposition and pharmacokinetics of a GalNAc3-conjugated antisense oligonucleotide targeting human lipoprotein (a) in monkeys. Nucleic Acid Ther, 26:372–380.

27.

Suhr

, Coelho

, Buades

, Pouget

, Conceicao

, Berk

, Schmidt

, Waddington-Cruz

, Campistol

, et al. (2015). Efficacy and safety of patisiran for familial amyloidotic polyneuropathy: a phase II multi-dose study. Orphanet J Rare Dis, 10:109.

28.

Geary

, Yu

, Watanabe

, Henry

, Hardee

, Chappell

, Matson

, Sasmor

, Cummins

and Levin

. (2003). Pharmacokinetics of a tumor necrosis factor-alpha phosphorothioate 2′-O-(2-methoxyethyl) modified antisense oligonucleotide: comparison across species. Drug Metab Dispos, 31:1419–1428.

29.

, Zhang

, Geary

, Graham

, Masarjian

, Lemonidis

, Crooke

, Dean

and Levin

. (2001). Pharmacokinetics and pharmacodynamics of an antisense phosphorothioate oligonucleotide targeting Fas mRNA in mice. J Pharmacol Exp Ther, 296:388–395.

30.

Attarwala

, Goel

, Vest

, Karsten

, Green

and Robbie

. (2018) Development of a pharmacokinetic-pharmacodynamic model of ALN-TTRSC02, an investigational RNAi therapeutic for the treatment of hereditary TTR amyloidosis. Clin Pharmacol Ther, 103:S58–SS58.

31.

Attarwala

, Goel

, Madigan

, Akinc

and Robbie

. (2017). Development of a pharmacokinetic-pharmacodynamic (PK-PD) model of fitusiran, an investigational RNAi therapeutic targeting antithrombin for the treatment of hemophilia in patients with and without inhibitors. http://www.alnylam.com/wp-content/uploads/2017/07/Attarwala_Fitusiran-PKPD_ISTH-2017_09July2017.pdf Accessed August 8, 2018.

32.

Ackermann

, Guo

, Benson

, Booten

, Freier

, Hughes

, Kim

, Jesse Kwoh

, Matson

, et al. (2016). Suppressing transthyretin production in mice, monkeys and humans using 2nd-Generation antisense oligonucleotides. Amyloid, 23:148–157.