Discovery of long-acting APOC3 siRNA for treating patients with hypertriglyceridemia

Abstract

Apolipoprotein C-III (APOC3) plays a crucial role in triglyceride metabolism, and its high expression leads to hypertriglyceridemia, which can contribute to an increased risk of cardiovascular disease and, when severely increased, can lead to acute pancreatitis. Loss-of-function variants in APOC3 are linked to lower triglyceride levels and reduced incidence of coronary artery disease. APOC3 mRNA, primarily synthesized by hepatocytes, is an ideal target for GalNAc-conjugated RNA-targeted therapies such as the antisense oligonucleotide (ASO) oleszarsen and small-interference RNA (siRNA) plozasiran. Herein, we systematically evaluate siRNA chemical modifications or multiple siRNAs to identify a long-acting APOC3 siRNA with a minimal number of 2’-F nucleotides. Using a series of structure-activity relationship (SAR) studies, we explored the effects of various oligonucleotide chemical modification scaffolds on siRNA potency, efficacy, and durability. These efforts led to the identification of an APOC3 targeting siRNA containing a novel chemical scaffold with robust activity and an extended duration of action in preclinical models. Additionally, selectivity and tolerability assessments in human cells, rodents, and nonhuman primates showed excellent safety and tolerability. A comparative analysis of the lead APOC3 siRNA with a surrogate of a clinical-stage APOC3 siRNA drug suggests the potential for similar or better potency and efficacy combined with less frequent dosing, potentially reducing the treatment burden on patients with hypertriglyceridemia.

Graphical Abstract

Graphical Abstract.

Introduction

RNA-targeted therapeutics, defined as drugs that target mRNA and contain DNA, RNA or their modified nucleosides as their active component, offer several potential advantages over more traditional drug discovery and development approaches. This includes: [1] significantly improved specificity allowing discrimination between closely related gene products; [2] predictability due to common pharmacokinetics (PK), pharmacodynamics, and toxicological properties, as well as standardized manufacturing processes and methods; and [3] versatility allowing inhibition of “non-druggable” targets, such as transcription factors, adapter proteins, and other proteins with no intrinsic enzymatic activity or proteins with unknown function [1]. RNA-targeted therapeutics have been evaluated clinically for treating various diseases, including viral, cancer, cardiovascular, metabolic, inflammatory, and neurological diseases. Clinical studies have documented the therapeutic utility of RNA-targeted therapeutic medicines as reflected by the FDA- and/or EMA approval of multiple drugs of this class [2–4].

A major class of RNA-targeted therapeutics are siRNAs, which are double-stranded oligonucleotides with an antisense (guide) strand (As) hybridized to a sense (passenger) strand (Ss) that reduce the expression of a gene of interest via the RNA interference (RNAi) mechanism [5, 6]. The siRNA antisense strand is loaded into Argonaute (Ago) proteins in the RNA-induced silencing complex (RISC), where it selectively binds to the complementary target mRNA, resulting in its degradation and subsequent reduction of protein levels [6]. The activity of siRNA drugs is exquisitely sensitive to their chemical design. Early development efforts focused on increasing stability and reducing immune activation. Initial studies demonstrated that selective incorporation of 2′-O-methyl (2′-OMe) uridine or guanosine into siRNA strands completely abrogates immune stimulation while preserving potent gene silencing activity [7]. Rational, site-specific modifications of the phosphate backbone, sugar ring, and nucleobases have proven critical for advancing siRNA drugs [8–15]. The first fully functional, fully modified siRNA incorporating 2′-fluoro (2′-F) and 2′-OMe chemistries was reported in 2005 [15]. Subsequent optimization of siRNA chemical modifications to further enhance metabolic stability, selectivity, and tolerability led to designs combining 2′-F, 2′-OMe, and phosphorothioate linkages [16–19]. These advances established the chemical foundation for clinically viable siRNA therapeutics. Selective siRNA delivery to hepatocytes is accomplished via conjugation to a triantennary N-acetylgalactosamine (GalNAc) ligand, that binds the asialoglycoprotein receptor (ASGPR) which is highly expressed on hepatocytes [20, 21]. The GalNAc ligand is cleaved upon cellular internalization, followed by sustained intracellular release of the siRNA [22]. Such improvements in design and delivery have contributed to substantial improvements in siRNA safety, tolerability, efficacy, and dosing frequency which can make treatment easier for patients [23].

One area where RNA-targeted therapies have shown benefit is hypertriglyceridemia, which is a significant risk factor for cardiovascular diseases and acute pancreatitis [24–26]. Apolipoprotein C-III (APOC3) is a crucial regulator of triglyceride metabolism [27, 28]. Mechanistic studies have demonstrated that elevated plasma apoC3 impacts plasma triglyceride metabolism by two distinct mechanisms: inhibition of lipoprotein lipase (LPL) and reduced hepatocyte uptake of triglyceride-rich lipoproteins (TRL) [29, 30]. Insufficient LPL-dependent lipolysis of TRLs and poor hepatocyte clearance of TRLs and their remnant lipoproteins contribute to elevated plasma triglyceride levels. Reducing apoC3 facilitates more efficient lipolysis and clearance of TRLs and their remnant lipoproteins, thereby reducing plasma triglycerides. This effect has been shown in human clinical trials by RNA-targeted therapies such as olezarsen (TRYNGOLZA™), which recently gained FDA approval to reduce plasma triglycerides (TG) along with diet in adults with familial chylomicronemia syndrome (FCS) [31].

siRNA inhibitors of APOC3 expression have demonstrated similar effects on triglyceride lowering. In clinical trials, plozasiran has demonstrated sustained and robust plasma apoC3 protein and triglyceride reductions with quarterly (Q3M) dosing [32, 33]. Herein, we report the identification of an APOC3 targeting siRNA containing a novel chemical scaffold with robust activity and an extended duration of action in preclinical models. Selectivity and tolerability assessments in human cells, rodents, and nonhuman primates showed excellent safety and tolerability. A comparative analysis of the lead APOC3 siRNA with a surrogate of a clinical-stage APOC3 siRNA drug suggests the potential for similar or better potency and efficacy combined with less frequent dosing.

Materials and methods

General method for the synthesis of antisense and sense strands

All reagents and solutions used for oligonucleotide synthesis were purchased from commercial sources. The standard phosphoramidites and solid supports were used for incorporation of dA, dU, dG, dC and T residues. A 0.1 M solution of 2’-F, 2’-O-Me, 2’-O-MOE modified nucleoside (A, G, C, U) 3’-phosphoramidites in 50% anhydrous acetonitrile in toluene (V/V) was used for the synthesis. A solution of 5’-deoxy-(E)-5′-(di[bis(pivaloyloxymethyl)-vinylphosphonate]-2’-O-(2-methoxyethyl)-thymidine-3’-phosphoramidite (S7, Scheme S1, Supplementary Figs S8-S12) and 5’-deoxy-5′-[bis(pivaloyloxymethyl)-methylenephosphonate]-2’-O-(2-methoxyethyl)-thymidine-3’-phosphoramidite (S9, Scheme S2, Supplementary Figs S13-S17) [34, 35] (0.12 M solution in acetonitrile/toluene 1:1, V/V) were used for incorporation of 5’-deoxy-(E)-5′-(vinylphosphonate-2’-O-(2-methoxyethyl)-thymidine-(5’-VP-MOE-T) and 5’-deoxy-5′-methylenephosphonate-2’-O-(2-methoxyethyl)-thymidine-(5’-MP-MOE-T) respectively. The modified oligonucleotides were synthesized on VIMAD UnyLinker^TM solid support [36], and the appropriate amounts of solid supports were packed in the column for synthesis. Dichloroacetic acid (15%) in toluene was used as a detritylating reagent. 4,5-Dicyanoimidazole (1 M) in the presence of N-methylimidazole (0.1 M) in anhydrous acetonitrile was used as an activator during the coupling step. Modified oligonucleotides were synthesized on an ÄKTA Oligopilot synthesizer (GE Healthcare Bioscience) or an ABI 394 synthesizer on a 2-200 µmol scale. A solid support preloaded with the Unylinker^TM was loaded into a synthesis column after closing the column bottom outlet, and acetonitrile (CH₃CN) was added to form a slurry. The swelled support-bound Unylinker^TM was treated with a detritylating reagent containing 15% dichloroacetic acid in toluene to provide the free hydroxyl groups. During the coupling step, four to eight equivalents of phosphoramidite solutions were delivered and the coupling was allowed to carry out for 10 min. All other steps in the protocol supplied by the manufacturer were used without modification. Phosphorothioate linkages were introduced by sulfurization with a 0.1 M xanthane hydride in 50% pyridine in acetonitrile (V/V), with a contact time of 3 min. Phosphate diester linkages were incorporated via oxidation of phosphite triesters using a solution of 0.05 M iodine in pyridine/water (9:1 V/V) for a contact time of 3 min. Unreacted sites were blocked using a standard capping reagent (Cap A: acetic anhydride in acetonitrile, 2:8, V/V, Cap B: 1-methylimidazole, pyridine in acetonitrile, 2:3:5 V/V/V, Capping time 2 min). After the desired sequence was assembled, the solid-support-bound oligonucleotide was washed with CH₃CN and dried under high vacuum. Dried solid support bearing oligonucleotides were treated with aqueous NH₄OH (28-30 wt%) containing diethylamine (10% V/V) and ethanol (10% V/V) and heated at 55 °C for 2 h and subsequently allowed to age at room temperature for additional 8-12 h to cleave oligonucleotides from support, remove protecting groups, and hydrolyze the UnyLinker™ moiety. The unbound oligonucleotide was then filtered, and the support was rinsed with water and filtered. Ammonia was boiled off from the combined filtrate and washing by bubbling nitrogen. Oligonucleotides were purified by HPLC on a strong anion exchange column (Waters, Source 30Q, 30 µm, 2.54 × 8 cm, A = 100 mM ammonium acetate in 30% aqueous CH₃CN, B = 1.5 M NaBr in A, 0-60% of B in 60 min, flow 14 mL min⁻¹). The fractions containing full-length oligonucleotide (determined by LC-MS analysis) were pooled together and diluted with water (3 fold volume) and desalted by HPLC on the reverse-phase column to yield the oligonucleotides in an isolated yield of 25-30% based on solid-support loading. The GalNAc conjugated oligonucleotides were synthesized according to the reported procedure [37, 38]. The oligonucleotides were characterized by ion-pair-HPLC-MS analysis with Agilent 1100 MSD system (Supplementary Data, Supplementary Table S1).

General method for the preparation of siRNA Duplexes

An equimolar amount of antisense and sense strands (Supplementary Table S1) was annealed to provide siRNA. All the duplexes were analyzed by size exclusion chromatography and met the purity cut-off of at least 85%.

Mouse studies

Animal experiments were conducted in accordance with the American Association for the Accreditation of Laboratory Animal Care guidelines and were approved by the Animal Welfare Committee. The animals were housed in micro-isolator cages on a 12-h light-dark cycle with a constant temperature and humidity and were given access to food and water ad libitum.

APOC3 siRNA Mouse Protocol

siRNAs were subcutaneously administered to APOC3 transgenic (huAPOC3) mice (B6; CBA-Tg(APOC3)3707Bres/J; JAX stock no. 006907) for siRNA activity assessments or to CD-1 mice (Charles River) for tolerability assessments. In vivo dose response and duration of action siRNA activity assessments were performed in 6- to 8-week-old chow-fed male hAPOC3 transgenic mice with plasma triglycerides (TG) and plasma apoC3 protein greater than ∼5,000 mg dL^–1 and ∼200 mg dL^–1, respectively. In vivo, tolerability assessments were performed in 6- to 8-week-old male CD-1 mice.

Nonhuman primate study

Animal experiments were conducted in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care International (AAAL AC International), and the study was reviewed and assessed by the Institutional Animal Care and Use Committee (IACUC) of Korea Institute of Toxicology (KIT). All procedures in compliance with the Animal Welfare Act and Guide for the Care and Use of Laboratory Animals (ILAR). Cynomolgus monkeys (2–4 years old) were used to evaluate the APOC3 siRNA 26. Approximately 60 g of a normal diet (Certified Primate Diet number 5048, PMI Nutrition International, Inc.) was provided to each monkey twice daily, supplemented with fresh fruit, cereal, or other treats. The animals had ad libitum access to filtered, ultraviolet light-irradiated water. Male and female animals were treated with PBS, 4 or 50 mg kg⁻¹ of 26 (n = 2/group/sex) on study days 1 and 28, and liver APOC3 expression was evaluated on day 31. Test articles were dissolved in sterile saline and administered subcutaneously at a volume of 0.33 mL/kg body weight.

Plasma analysis

Blood was collected with EDTA via retro-orbital plexus bleeding under isoflurane anesthesia (mice) or taken from the cephalic vein, saphenous, or femoral vein (NHP). Plasma was separated by centrifugation at 10,000 rcf for 4 min at 4°C. Plasma transaminases, glutamate dehydrogenase (GLDH), triglycerides and apoC3 were measured using a Beckman Coulter AU480 analyzer using the following assays: ALT (alanine aminotransferase), OSR6107 (Beckman Coulter); AST (aspartate aminotransferase), OSR6209 (Beckman Coulter); GLDH (Randox); triglycerides, OSR61118 (Beckman Coulter); apoC3, LP3865 (Randox).

RT-qPCR

Tissue RNA was extracted for real-time quantitative PCR (RT-qPCR) analysis to measure liver APOC3 mRNA using primer probe set Mf02794312_m1 (ThermoFisher Scientific). This is a TaqMan primer probe set that utilizes FAM-MGB to allow for real-time PCR detection of Macaca fascicularis (cynomolgus monkey) APOC3 mRNA. Results are presented as percent inhibition of APOC3 RNA relative to saline control, normalized to total RNA content, as measured by RIBOGREEN®. Briefly, a piece (∼50 mg) of fresh liver was homogenized in 1 mL of lysis buffer, with the resulting lysate applied to silica membranes for RNA purification using the PureLink® RNA Mini Kit (Invitrogen). Following mRNA purification and quantitation, samples were subjected to RT-qPCR analysis using the Thermo Fisher StepOnePlus Real-Time PCR System. RNA transcripts were normalized to total RNA levels as measured by RIBOGREEN®.

Cell culture experiments

Primary Human Hepatocytes (BioIVT, M0095-P, lot ZFW) were thawed and resuspended in Lonza HCM Hepatocyte Culture Medium. Cells were counted and diluted to 300,000 cells per mL in room temperature growth medium before adding 100 μL of cell suspension to the collagen I coated 96-well culture plate wells. Immediately after plating the cells, 11 μL of 10X oligonucleotide in water was added to the appropriate wells, and the cells were treated at 10 doses with a top concentration of 5 µM and 1:5 dilutions. The culture plate was then incubated at 37°C and 5% CO₂. After 96 h, the cells were lysed for RNA isolation and analysis. Primary monkey hepatocytes (Sekisui, PPCH2000, lot 2110098) were thawed and resuspended in Williams Medium E. Cells were counted and diluted to 200,000 cells per mL and plated according to the methods described for primary human hepatocytes. After 72 h, the cells were lysed for RNA isolation and analysis. The RNA was purified with a glass fiber filter plate (Pall No. 5072). The human APOC3 mRNA levels were quantitated with RT-qPCR on the QS7 instrument (Thermo Fisher). Briefly, 5 μL RT-qPCR reactions containing 1 μL of RNA were run with AgPath-ID One-Step PCR reagents (Thermo Fisher) and the primer-probe sets (Integrated DNA Technologies) listed in the materials section following the manufacturer’s instructions. Total RNA levels were quantitated by measuring human GAPDH mRNA levels (Integrated DNA Technologies) and used to normalize the APOC3 mRNA data. Results were analyzed in Microsoft Excel, Office 365 Version 1705. Percent untreated control values were generated using the following formula.

%UTC = ((Sample QuantityTarget/Sample Quantitynormalization signal)/Average (UTC QuantityTarget/UTC Quantitynormalization signal))*100.

Half maximal inhibitory concentration (IC₅₀) values were calculated with the “log(inhibitor) vs. normalized response (variable slope)” formula using Prism software (v10; GraphPad Software, San Diego, CA).

Sequencing and data processing

Gene counts were tabulated using Salmon version 0.7.2 with pseudo alignment mode, default parameters, and gene models from the ensemble human transcriptome build 109.

Digital gene expression (DGE) and bioinformatic analysis

Significance of overlap between in-silico (seed matched) off-targets and differentially expressed genes was performed using hypergeometric testing, (where total number of genes (N) = 20000). An in-silico list of seed-matched off-targets was determined by searching 3’UTRs for the presence of at least one perfect seed-matched site. Differentially expressed genes analysis was performed using DESeq2 with standard parameters. Differentially expressed genes with adjusted p-value of significance < 0.005 were selected for comparisons.

Volcano and efficacy/potency plots

Differentially expressed genes were identified for each treatment condition and concentration by comparing gene expression against a control dataset of 20 untreated controls. Genes expressed above 2.5 transcripts per million (TPM) and with p-value < 0.05 were called differentially expressed. Concentration-dependent knockdown data (in TPM) for all genes were fitted to a standard hill curve, and IC₅₀ and maximum knockdown were evaluated using a nonlinear least squares fit (Python SciPy optimize.curvefit function). Fit error was measured by computing the 75th percentile of the magnitude of linear error over all 10 concentrations tested. Fit bounds for IC₅₀ were set from 0.0001 to 20 and maximal knockdown from 0.1 to 200. Genes were classified as concentration-responsive (or ‘responders’) if they met all of the following criteria: 1) maximum knockdown achieved was > 50% control 2) linear fit errors were below 25% 3) gene was considered differentially expressed in at least 2 of 3 highest doses tested (p.adj < 0.01).

Results

Identification of Lead APOC3 siRNA Sequences

We previously identified siRNA targeting human apolipoprotein C3 (APOC3) mRNA by synthesizing and screening multiple siRNAs targeting different regions of the APOC3 transcript, and reported activity of lead APOC3 siRNA with alternating 2’-F/2’-OMe chemistries in transgenic mouse models [19]. It also shares seed sequence (2 to 8 nucleotides from the 5’-end of As) identity with plozasiran. We therefore chose to utilize this sequence to identify long-acting APOC3 siRNAs using novel chemical scaffolds in this study.

siRNA containing 2’-O-MOE 5’-deoxy-(E)-5’-vinylphosphonate modified antisense strand enhances durability

Our goal in this study is to enhance the metabolic stability and, therefore, durability of a lead siRNA relative to known APOC3 targeting siRNAs and ASOs. We first set out to establish how benchmarks with known clinical profiles perform in our human APOC3 transgenic mouse model. We first profiled olezarsen (TRYNGOLZA™), a 2’-MOE gapmer ASO which is administered monthly in humans, with 80 mg, the FDA approved dose for FCS [31]. Olezarsen showed maximal observed plasma apoC3 protein and triglyceride reductions (19%) at day 7 post-single dose, with 50% recovery between 14 and 28 days, and a return to baseline by day 42 [Supplementary Fig. S1]. This finding is consistent with the known faster metabolism of mice compared to humans and establishes a mouse benchmark for a monthly administered drug in humans.

Plozasiran is a GalNAc siRNA that has demonstrated robust clinical activity with quarterly dosing [32, 33]. To establish a benchmark in our mice for a quarterly (q3m) administered drug in humans, we synthesized a surrogate of plozasiran, siRNA 1 (Fig. 1, 2), with the identical oligonucleotide chemistry modification pattern [39, 40]. For ease of chemical synthesis, we employed a different hepatocyte targeting ligand, THA-GN3 (GalNAc, Fig. 1), which in previous experiments had been shown to result in essentially identical potency to multiple trivalent GalNAc clusters, including lysine derived clusters such as used in plozasiran [37].

Figure 2.

Duration of activity of hAPOC3 siRNA plozasiran surrogate 1 and 2 to reduce plasma apoC3 protein (A) and triglyceride (B) in huAPOC3 transgenic mice (n = 3) after a single dose of 1 mg kg⁻¹ for up to 105 days. Dose-response comparison of siRNA 1 and 2 following a single dose of 1.0, 0.3, and 0.1 mg kg⁻¹ with plasma apoC3 protein (C) and triglycerides (D) evaluated on day 14. Error bars represent standard deviation.

Figure 1.

A. Structure of the modified nucleic acids and 5’-N-acetylgalactosamine (GalNAc) cluster conjugated sense strand used for developing long-acting APOC3 siRNAs; B. Numbers indicating nucleoside and backbone positions from the 5’-end of the antisense strand (As) and the sense strand (Ss) of siRNA.

Reports suggest that oligonucleotides containing high numbers of 2’-F nucleotides are less stable to nucleases than 2’-O-Me and 2’-O-MOE nucleotides, and furthermore that high 2’-F nucleotide content could contribute to adverse safety and tolerability [41, 42]. siRNA 1 contains a total of eleven 2’-F nucleotides; eight in the antisense strand and three in the sense strand. This prompted us to consider whether previously published 2’-F/2’-O-Me modification patterns having fewer 2’-F nucleotides [16] could maintain potency and increase durability. Because of the known increased stability of 2’-O-MOE relative to 2’-O-Me modifications, we introduced 2’-O-MOE modified phosphate mimics at the 5’-end of the antisense strand. For efficient RNA-induced silencing complex loading (RISC) and slicer activity of an siRNA, it is essential to have a 5’-phosphorylated antisense strand [43–45]. To overcome potential kinase limitations of highly stabilizing nucleotide modifications such as MOE, we designed siRNA 2 (Fig. 2) with a metabolically stable 5’-vinyl phosphonate (5’-VP-2’-O-MOE T) at the 5’-end of the antisense strand [35]. We employed the same 5’-GalNAc modification strategy used for olezarsen on the sense strand [37] (GalNAc, Fig. 1), and introduced two phosphorothioate modifications at the 3’ and 5’-ends of the sense and antisense strands to enhance metabolic stability (Figs 1B and 2). Synthesis of sense and antisense strands was accomplished using the reported methods and characterized by LC-MS analysis (Supplementary Data, Supplementary Table S1) [19, 38].

Next, we examined the effect of chemical modification on the efficacy and duration of action (DOA) of the stability-enhanced 5’-VP-MOE siRNA 2 as compared to the benchmark 1 in transgenic mouse models to assess the utility of this chemical scaffold. Human APOC3 transgenic mice (n = 3/group) were injected subcutaneously with 1 mg kg⁻¹ of siRNAs 1-2. The efficacy and DOA of APOC3 siRNAs 1-2 to inhibit plasma apoC3 protein and triglyceride levels were evaluated in huAPOC3 transgenic mice for up to 105 days (Fig. 2A and B). siRNA 1 showed maximal plasma apoC3 protein reduction (7.3%) at 7 days, with recovery to 50% target reduction at around 42 days, and return to baseline at 56 days, with a similar effect on triglycerides. This is roughly equivalent to plasma triglyceride and apoC3 lowering to that previously published for plazosiran in huApoC3 transgenic mice at 1 mg kg⁻¹, further supporting that differences in the GalNAc cluster of 1 versus plazosiran had no impact on siRNA activity [39, 46], and establishing a mouse benchmark for a quarterly administered human siRNA drug.

In contrast, the same dose of siRNA 2 maintained target suppression dramatically longer than siRNA 1, with ∼90% or better target suppression out to day 42, and maintained a ∼70% reduction to day 105. In addition, in a 2-week dose response experiment at 0.1, 0.3, and 1 mg kg⁻¹, siRNA 2 exhibited ∼3-fold enhanced potency compared to siRNA 1 in reducing plasma apoC3 and triglycerides (Fig. 2C and D). This combined data suggests that reducing 2’-F nucleotide content in combination with a more metabolically stable 5’-VP-MOE T modification can enhance the potency and durability of an APOC3 siRNA, with the potential to extend dosing frequency well beyond quarterly administration in humans.

To investigate the role of 5′-VP in enhancing potency and DOA, we compared the activity of siRNA 2 with and without 5′-VP (Supplementary Data Fig. S2). siRNA 2 lacking 5′-VP showed a faster recovery, returning to baseline levels of apoC3 and triglycerides by day 84, whereas siRNA 2 maintained approximately 70% reduction through day 105. In addition, in a 4-week dose–response study, siRNA 2 demonstrated ∼5-fold greater potency than siRNA 2 without a VP counterpart in reducing plasma apoC3 and triglycerides (Supplementary Data Figs S2C and D). Collectively, these findings suggest that the 5′-VP modification significantly contributes to both the enhanced potency and extended durability of siRNA 2.

siRNA containing 5’-deoxy-5’-methylene phosphonate is less durable in mice

Since 5’-Deoxy-5’-methylenephosphonate (5’-MP, Fig. 1) is also a metabolically stable 5’-phosphate analog for ss-siRNA and siRNA [35], we compared the activity and durability of 5’-MP (3, Fig. 3) and 5’-VP (2, Fig. 3) siRNAs in transgenic mice. Except for the phosphate mimic, the chemical scaffolds of siRNAs 2 and 3 are identical. Synthesis of sense and antisense strands of siRNA 3 was accomplished as reported [35, 38]. The siRNA 3 was well characterized by LC-MS analysis (Supplementary Data, Supplementary Table S1). Potency of hAPOC3 siRNA with VP [2] and MP [3] in primary human hepatocytes was evaluated and found to be similar (Supplementary Data Fig. S3, IC₅₀ siRNA 2 0.63 nM, siRNA 3 0.93 nM).

Figure 3.

(A). Duration of activity of hApoC3 siRNA 2 containing 5’-deoxy-(E)-5’-vinyl phosphonate (VP) and 3 containing 5’-deoxy-5’-methylene phosphonate (MP) to reduce plasma apoC3 protein in huAPOC3 transgenic mice (n = 3) after a single dose of 0.1, 0.3, 1 mg kg⁻¹ for up to 98 days. (B). Dose response comparison of siRNA 2 and 3 of plasma apoC3 reductions at day 14. Error bars represent standard deviation.

The relative dose-responsive DOA of 5’-VP APOC3 siRNA 2 and 5’-MP siRNA 3 was examined in human APOC3 transgenic mice. Mice (n = 3/group) were injected subcutaneously with siRNAs 2 and 3 (0.1, 0.3, and 1 mg kg⁻¹). The efficacy and DOA of siRNAs 2-3 to inhibit plasma apoC3 protein were evaluated in transgenic mouse models at the indicated time intervals (Fig. 3). At all doses, the plasma apoC3 level of animals treated with 5’-MP APOC3 siRNA 3 (Fig. 3A) recovered faster than 5’-VP APOC3 siRNA 2 (Fig. 3A). The potency of siRNAs 2 and 3 was determined in human APOC3 transgenic mice. The human APOC3 transgenic mice (n = 3/group) were injected subcutaneously with a single dose of siRNA 2 and 3 (0.1, 0.3, and 1 mg kg⁻¹). Mice were sacrificed after 14 days and analyzed for reductions in human apoC3 protein in mouse plasma (Fig. 3B). The 5’-VP APOC3 siRNA 2 showed approximately 2-fold enhancement in potency (Fig. 3B, ED₅₀ 0.1 mg kg-1) for reducing human apoC3 protein relative to 5’-MP APOC3 siRNA 3 (Fig. 3B, ED₅₀ 0.2 mg kg^-1). These data suggest that 5’-VP modification is an ideal metabolically stable phosphate analog for enhancing the durability of siRNAs, and we therefore used this chemistry for all our further SAR studies in this report.

FHNA is a viable surrogate 2’-F nucleotide modification

3′-Fluoro hexitol nucleic acid (FHNA, Fig. 1) was developed as a substitute for 2’-F RNA and was examined in ASOs [47], but the value of this chemistry for enhancing siRNA properties is unknown. We designed siRNA 4 (Fig. 4), where the 2’-F nucleotide at position 14 of the antisense strand was replaced with FHNA and siRNA 5 with FHNA at sense strand positions 7 and 11. To examine the effect of FHNA modification on the antisense and sense strands on efficacy, DOA siRNA 6 was generated by duplexing the antisense strand of siRNA 4 and the sense strand of siRNA 5. In addition, we also used 2’-O-Me to minimize 2’-F nucleotide residues in the siRNA. siRNA 7 antisense strand contains FHNA at position 2 and 2'-O-Me at 6, 16. In siRNA 8, all 2’-F residues in the antisense strand were either replaced with FHNA (at positions 2, 14) or 2’-O-Me (at positions 6, 16). The antisense strand of siRNA 8 was duplexed with the sense strand containing only two 2’-F nucleotides at positions 10-11. The FHNA nucleoside monomers were synthesized using the reported procedure [47, 48]. All the sense and antisense strands of APOC3 siRNAs 4-8 were synthesized using the reported method (Supplementary Data) [35, 38, 47, 48].

Figure 4.

Duration of activity of hAPOC3 siRNA 2, 4-8 to reduce plasma apoC3 protein (A and C) and triglyceride (B and D) in huAPOC3 transgenic mice (n = 3) after a single dose of 1 mg kg⁻¹ for up to 28 to 105 days. Error bars represent standard deviation.

The effect of FHNA modification on APOC3 siRNA (4-8, Fig. 4) efficacy and DOA were examined in human APOC3 transgenic mice and compared with the lead siRNA 2. The effectiveness and DOA to inhibit plasma apoC3 protein and triglyceride were evaluated in transgenic mouse models after 1 mg kg⁻¹ subcutaneous injection of siRNAs 4-8 at indicated time intervals (Fig. 4). The activity of FHNA APOC3 siRNA 4-6 (A-B, Fig. 4) at 14 and 28 days was comparable to that of lead siRNA 2. Interestingly, ApoC3 siRNAs 7-8 lacking 2’-F residues at positions 2, 6, and 16 in the antisense strand were significantly less active (Fig. 4C-D) than siRNA 2. The plasma apoC3 and triglyceride levels of animals treated with FHNA APOC3 siRNA 4-6 (Fig. 4A and B) recovered faster than 5’-VP APOC3 siRNA 2 (Fig. 4A and B). A single FHNA modification at position 14 of the antisense strand, as in siRNA 4, was sufficient to reduce the durability of APOC3 siRNA. Although a comprehensive SAR analysis of FHNA modifications was not performed, our findings indicate that multiple FHNA incorporations within an siRNA are generally not well accommodated. Nevertheless, strategic placement of FHNA modifications can preserve silencing potency while extending the duration of action (DOA).

Partial replacement of 2’-F with 2’-O-Me maintains the efficacy and durability

Next, we explored the potential of replacing 2’-F modification in the antisense and sense strands to minimize the number of 2’-F content in APOC3 siRNA. In siRNA 9 (Fig. 5), the 2’-F nucleotides at positions 6 and 16 of the antisense strand and positions 7 and 9 of the sense strand were substituted with 2’-O-Me [16]. To enhance the stability, the internucleotide linkage between 2’-F residues at positions 10-11 was converted to phosphorothioate for siRNA 10. In addition, we also designed APOC3 siRNA 11 (Fig. 5), where all the 2’-F nucleotides in the sense strand were replaced with 2’-O-Me to identify the optimum 2’-O-Me-enriched chemical scaffold for this sequence. Activity and DOA of siRNAs 9-11 were assessed in human APOC3 transgenic mice. Mice (n = 3/group) were injected subcutaneously with 1 mg kg⁻¹ of siRNAs 9-11, and the activity and DOA to inhibit plasma apoC3 protein and triglyceride levels were evaluated at the indicated time intervals. APOC3 siRNA 9-10 showed activity and durability comparable to lead APOC3 siRNA 2 (Fig. 5A and B), whereas APOC3 siRNA 11 was significantly less active. These data demonstrate that APOC3 siRNA containing only four 2’-F nucleotides, as in APOC3 siRNAs 9-10, is as efficacious and durable as siRNA 2 with eight 2’-F nucleotides.

Figure 5.

Duration of activity hAPOC3 siRNA 2, 9-14 to reduce plasma apoC3 protein (A and C) and triglyceride (B and D) in huAPOC3 transgenic mice (n = 3) after a single dose 1 mg kg⁻¹ for up to 28–84 days. Error bars represent standard deviation.

Replacing the limited phosphorothioate backbone with mesyl phosphoramidate chemistry (MsPA) maintains the efficacy and durability

The mesyl phosphoramidate (MsPA, Fig. 1) linkage was recently reported as an alternative to phosphorothioate backbone chemistry of oligonucleotide with improved nuclease stability [49, 50]. We examined the MsPA modification to enhance the nuclease stability of 2’-O-Me enriched APOC3 siRNA 9. siRNA 12 has MsPA modification at backbone position 14 of the antisense strand, siRNA 13 at positions 9 and 11 of the sense strand, and siRNA 14 at both antisense strand position 14 and sense strand positions 9 and 11 (Fig. 5). The MsPA containing sense and antisense strand oligonucleotides was synthesized according to the reported procedure [49]. While siRNA 12 with a single MsPA modification exhibited similar activity to 2 and the direct comparator 9, siRNAs 13 and 14 containing multiple MsPA modifications showed markedly reduced activity (Fig. 5C and D). These data indicate limitations in using several MsPA linkages in siRNAs adjacent to 2’-fluoro nucleotides. Considering the reduced activity observed for 13 and 14, we did not further evaluate the MsPA chemistry scaffold for enhancing APOC3 siRNA durability.

APOC3 siRNA containing 2’-O-MOE and DNA phosphorothioate have enhanced durability

The 2’-O-methoxyethyl (MOE, Fig. 1) modification is currently one of the most widely used 2’-sugar modifications due in part to unique structural features which provide enhanced metabolic stability to oligonucleotides [1, 51]. While MOE is not broadly tolerated in siRNAs, reports suggest that judicious placement of MOE modification in the antisense and sense strand can maintain activity and improve the selectivity of siRNA [52, 53]. We therefore examined the effect of MOE modification when placed at positions 9 and 10 of the antisense strand of APOC3 siRNA 2. siRNAs 15 containing MOE modification at positions 9 and 16 containing MOE at positions 9 and 10 were synthesized (Fig. 6). Interestingly, the activity and DOA of siRNA 15 and 16 were similar to that of the lead siRNA 2 (Fig. 6A and B). Next, we paired antisense strands containing MOE from siRNA 15 and 16 with 2’-O-Me-enriched sense strands to get siRNA 17 and 18. To further enhance the metabolic stability, we introduced two MOE nucleotides at the 3’ and 5’ ends of the sense strand of siRNAs 17 and 18 (Fig. 6). The siRNAs 17 and 18 maintained the activity and improved the durability of both plasma apoC3 protein and triglyceride reduction (Fig. 6A and B). A dose-response duration of action experiment for siRNA 17 (Fig. 6C) and 18 (Fig. 6D) showed improved duration of activity in reducing human plasma apoC3 levels in transgenic mice following treatment of 1, 2, and 3 mg kg⁻¹. This observation of improved duration with MOE substitution at the 5’-end of the antisense strand, as well as both ends of the sense strand, suggests additional use of MOE at the 3’-end of the antisense strand. It has been reported that siRNA containing a dinucleotide overhang with purine residues is preferred over pyrimidines [52, 54, 55], which prompted replacement of the 2’-O-Me UU 3’-overhang with MOE AA in subsequent optimization experiments.

Figure 6.

Duration of activity of hAPOC3 siRNA 2, 15-18 to reduce plasma apoC3 protein (A) and triglyceride (B) in huAPOC3 transgenic mice (n = 3) after a single dose of 1 mg kg⁻¹ for up to 84 days. Dose-responsive duration of activity of hAPOC3 siRNA 17 (C) and 18 (D) to reduce plasma apoC3 protein in transgenic mice after a single dose of 1, 2, and 3 mg kg⁻¹. Error bars represent standard deviation.

We next examined whether the 2’-F nucleotides in APOC3 siRNA 17 and 18 could be replaced with deoxyribonucleotides. While DNA can exist in multiple helical conformations [56, 57], the crystal structure of RNA/DNA hybrid has been shown to assume the A-form conformation [58], and we envisaged that introducing DNA in siRNA duplexes would preserve the A-form helical structure required for the RNAi mechanism [5]. The siRNA 19 and 20 (Fig. 7) with antisense strand containing deoxy nucleotides at positions 6, 14, 16, MOE at position 9, and MOE AA at 22, 23, and 2’-O-Me enriched sense strand with PO or PS linkage at backbone position 10 were synthesized. A similar chemical scaffold containing DNA and MOE was applied to siRNA 18 (Fig. 6) to yield APOC3 siRNA 21 and 22 (Fig. 7). The APOC3 siRNAs 19, 20-22 were shown to have similar or slightly reduced activity and duration in reducing the plasma apoC3 protein and triglyceride in human transgenic mice relative to lead siRNA 2 (Fig. 7A and B). It is conceivable that the duplex breathing of siRNA regions containing DNA could expose these regions to endonucleases. To protect the DNA region from potential endonuclease activity, PO linkages at the DNA modification of 19 and 20 were replaced with PS linkages to provide siRNAs 23 and 24 [59]. The activity and DOA of siRNA 23 and 24 were comparable to that of siRNA 2. It is remarkable that the activity and DOA of siRNA 23 and 24, with only three 2’-F nucleotides, demonstrated similar activity and DOA to siRNA 2, containing eight 2’-F nucleotides. DNA T is widely used in RNA therapeutic drugs compared to DNA U nucleotide [1], so we replaced DNA U nucleotides in siRNA 23 and 24 with T to yield siRNAs 25 and 26, respectively. Surprisingly, this substitution pattern in siRNA 25 and 26 showed subtly better activity and durability than any siRNA tested, including siRNA 2 (Fig. 7A and B).

Figure 7.

Duration of activity of hAPOC3 siRNA 2, 19-26 to reduce plasma apoC3 protein (A) and triglyceride (B) in huAPOC3 transgenic mice (n = 3) after a single dose of 1 mg kg⁻¹ for up to 84 days. Error bars represent standard deviation.

To extend the generalizability of the chemical scaffold of siRNA 26 to different sequences, we evaluated this scaffold on two different sequences targeting a different liver-expressed target and observed robust efficacy and durability (Supplementary Data Fig. S4).

siRNAs containing novel chemical scaffolds identified in this report demonstrate superior duration of activity compared to benchmarks in huAPOC3 transgenic mice

To assess the benefit of novel APOC3 siRNAs identified from our studies in enhancing the durability of siRNA, we compared the duration of action of these siRNAs to our plozasiran surrogate (siRNA 1, Fig. 8). The efficacy and duration of activity of APOC3 siRNAs 1, 6, 17, 18, and 26 in reducing plasma apoC3 protein in huAPOC3 transgenic mice following a single dose of 1 mg kg⁻¹ was examined (Fig. 8A). The recovery curves showed that the novel APOC3 siRNAs [6, 17, 18, 26] significantly improved DOA relative to siRNA 1 (Fig. 8A). Plasma apoC3 levels following a single dose of 1 recovered 50% in approximately 6 weeks, whereas 12 weeks were required to reach 50% for animals treated with APOC3-siRNAs 6, 17, 18, and 26. The APOC3 siRNA 26 with a novel DNA/MOE chemical scaffold with trans vinyl phosphonate at the 5’-end of As had the most extended duration of activity, with 80% plasma apoC3 reductions maintained for up to ∼8 weeks (Fig. 8A).

Figure 8.

(A). Duration of activity of siRNA 1 compared with lead hAPOC3 siRNAs 6, 17-18, and 26 after a single dose of 1 mg kg⁻¹, with plasma apoC3 protein evaluated in huAPOC3 transgenic mice for up to 12 weeks. Composite data from multiple studies representing n = 3–6 for each siRNA. B–D. Dose-responsive activity of hAPOC3 siRNA 26 (0.01, 0.03, 0.1, and 1 mg kg⁻¹ single dose) to reduce plasma liver mRNA (B), plasma ApoC3 protein (C), and plasma triglyceride (D) in transgenic mice. (E). ED₅₀ values for liver apoC3 mRNA, plasma apoC3, and plasma triglyceride reductions following siRNA 26 treatment. Error bars represent standard deviation.

We also evaluated the potency of siRNA 26 in human transgenic mice. Mice were injected with a single dose of 0.01, 0.03, 0.1, and 1 mg kg⁻¹ siRNA 26, and APOC3 mRNA expression (liver), apoC3 protein, and triglyceride levels in plasma were evaluated at 14 days. siRNA 26 showed robust potency in reducing liver APOC3 mRNA (ED₅₀ 0.15 ± 0.04, Fig. 8B and E), plasma apoC3 protein (ED₅₀ 0.12 ± 0.01, Fig. 8C and E), and plasma triglyceride (ED₅₀ 0.15 ± 0.03, Fig. 8D and E).

Lead hAPOC3 siRNA 26 with novel DNA/MOE chemical scaffold exhibited an improved off-target selectivity profile compared to siRNA 2 in primary human hepatocytes

To assess the off-target gene regulation, APOC3 siRNA 2 and 26 were delivered by free uptake to human primary hepatocytes using concentrations 100 000-fold greater than the on-target APOC3 IC₅₀ after 96 h. The majority of downregulated or upregulated genes did not exceed 50% reductions (Fig. 9A, Supplementary Data Figs S5 and S6). The plot of IC₅₀ for downregulated transcripts of APOC3 siRNA 2 and 26 showed more than 100-fold differences between on-target and off-target IC₅₀ values for the majority of responders (Fig. 9B, Supplementary Data Fig. S5). Interestingly, the number of downregulated genes was reduced with siRNA 26 containing the novel DNA/MOE chemical scaffold relative to siRNA 2.

Figure 9.

(A). Volcano plot depicting global gene expression changes in primary human hepatocytes at 96 h after free uptake of 200, 1000, and 5000 nM of hAPOC3 siRNAs 2 and 26. Enrichment of the seed region complementary in 3’-UTRs is tabulated N: 2 technical replicates; (B). Plot of relative IC₅₀ values of down-regulated genes relative to on-target IC₅₀ values after free uptake of hAPOC3 siRNA 2 and 26.

Lead hAPOC3 siRNA 26 is well tolerated in mice

Subsequently, we assessed the non-clinical safety of siRNA 26 in mice, confirming its safety. Mice (Male CD-1, n = 4/group) were injected subcutaneously with 300 mg kg⁻¹ once a week for 3 weeks (3 total administrations). At the termination of the study, the various endpoints were assessed 48 h after the last dose. siRNA 26 was administered at doses with a substantial margin of safety when considering the typical therapeutic doses of GalNAc siRNA drugs. The weight of the liver relative to the body weight of animals treated with siRNA 26 was not significantly changed compared to PBS-treated animals (Supplementary Data Table S2). siRNA 26 also did not cause any significant changes in liver function as measured at three weeks, 48 h post-dose (ALT, AST, and GLDH, Supplementary Data Table S2). Histopathological assessment of the liver, kidneys, spleen, and heart further demonstrated that siRNA 26 was well tolerated (Supplementary Data Fig. S7).

Lead hAPOC3 siRNA 26 demonstrates excellent potency in primary human hepatocytes, monkey hepatocytes, and efficacy in non-human primates

Next, we determined the potency of hAPOC3 siRNA 26 in primary human hepatocytes (Fig. 10A) and monkey hepatocytes (Fig. 10B) with free uptake, whereby the GalNAc siRNA activity is evaluated without the use of lipid transfection or electroporation. siRNA 26 showed excellent potency for reducing APOC3 mRNA expression in primary hepatocytes (IC₅₀ 0.04 nM, Fig. 10A) and in monkey hepatocytes (IC₅₀ 6.7 nM, Fig. 10B) as determined by RT-qPCR.

Figure 10.

(A) Potency of hAPOC3 siRNA 26 in primary human hepatocytes. Error bars represent standard deviation; (B) Potency of hAPOC3 siRNA 26 in monkey hepatocytes. Error bars represent standard deviation; (C) Activity of hAPOC3 siRNA 26 in monkey liver dosed subcutaneously at 4 mg kg⁻¹ and 50 mg kg⁻¹ (n = 4). Error bars represent standard deviation.

We also determined the efficacy of siRNA 26 in monkeys. Male and female cynomolgus monkeys were subcutaneously injected with 4 and 50 mg kg⁻¹ of siRNA 26 on days 1 and 28. Animals were sacrificed three days after the last dose, and organs were harvested to assess APOC3 mRNA reduction by RT-qPCR. hAPOC3 siRNA 26 showed a dose-dependent reduction of APOC3 mRNA in the liver with up to ∼90% APOC3 reduction (Fig. 10C). Consistent with mice tolerability data, siRNA was well tolerated with no obvious toxicities or alterations in blood chemistries or organ weights (Supplementary Data Table S3).

Discussion

One of the main drivers of the success of siRNA therapeutics is the development of design features to enhance metabolic stability while maintaining the ability to elicit the RNAi mechanism for efficient target reduction [2]. The most commonly employed siRNA drug designs incorporate chemical modifications in the sugar and phosphate backbones, combined with a tissue- or cell-specific delivery ligand [2]. The standard chemical scaffold used in siRNA drugs contains the complete substitution of RNA nucleosides with 2’-F and 2’-OMe sugar modifications, as well as limited phosphorothioate backbone modification [18]. The present work aimed to identify a novel APOC3 targeting siRNA employing a chemical scaffold capable of providing increased metabolic stability compared to siRNA containing a typical 2’F/2’-OMe scaffold while maintaining high potency, efficacy, and specificity and an excellent predicted margin of safety and tolerability. We envisaged that an APOC3 targeted siRNA containing such a chemical scaffold would demonstrate a long duration of activity with an appropriate therapeutic index, allowing for reduced dosing frequency (≥ Q6M) while maintaining robust target reduction in the clinic.

Plozasiran is a GalNAc siRNA that has demonstrated robust clinical activity with quarterly dosing [32, 33]. Plozasiran contains a total of 11 2’-F nucleotides; 8 in the antisense strand and 3 in the sense strand. Reports suggest that oligonucleotides containing high numbers of 2’-F nucleotides are less stable to nucleases than 2’-O-methyl and 2’-O-MOE nucleotides, and that high 2’-F nucleotide content could contribute to adverse safety and tolerability [41, 42]. Our goal in this study is to enhance the metabolic stability and, therefore, durability of a lead siRNA relative to a known APOC3 targeting siRNA. We thus examined whether reduced 2’-F nucleotide content in siRNAs could maintain potency and increase durability relative to standard siRNA designs. In addition, we introduced 2’-O-MOE modified phosphate mimics at the 5’-end of the antisense strand. For efficient RISC and slicer activity of an siRNA, it is essential to have a 5’-phosphorylated antisense strand [43–45]. To overcome potential kinase limitations of highly stabilizing nucleotide modifications such as MOE, we designed siRNA 2 (Fig. 2) with a metabolically stable 5’-vinyl phosphonate (5’-VP-2’-O-MOE T) at the 5’-end of the antisense strand [35]. The siRNA 2 (Fig. 2), containing a new chemical scaffold, exhibited three-fold enhanced potency compared to the plozasiran surrogate 1 in reducing plasma apoC3 and triglycerides. In addition, siRNA 2 contains a new chemical scaffold that exhibited a two-fold longer duration of effect than plozasiran surrogate 1 (Fig. 2) in transgenic mice. The data suggest that reducing 2’-F nucleotide content in combination with our metabolically stable 5’-VP-MOE T modification can enhance the potency and durability of an APOC3 siRNA. This finding should have broader implications for the design of more potent and durable siRNA drugs.

In part, the potency enhancement observed for siRNA 2 could be attributed to the presence of a metabolically stable 5’-VP analog, which prompted us to evaluate a 5’-methylene phosphonate (MP) analog, another metabolically stable phosphate mimic reported [35]. While the potency of the siRNA containing 5’-MP-2’-O-MOE T (3, Fig. 3) was found to be comparable, the recovery of the plasma apoC3 level was faster (50% in 70 days) than that of 5’-VP-2’-O-MOE T APOC3 siRNA 2 (50% in more than 90 days), clearly establishing the 5’-VP modification as a superior phosphate mimic. The double-bond character of the 5′-VP group restricts rotation around the C5′−C6′ bond, allowing the majority of phosphate contacts to Ago2 to be retained [60], whereas in 5’-MP, free rotation around the C5′-C6' can occur. It is conceivable that the restricted rotation of the 5’-VP siRNA may increase stability to endonucleases. Alternatively, as siRNA stability and duration are known to be due to an endosomal ‘reservoir’ of duplex siRNA [22], the 5’-MP may promote end fraying, which exposes single-stranded regions to nucleases, potentially explaining the decreased durability compared to the 5’-VP siRNA.

Biophysical and structural characterization of FHNA (Fig. 1) reveals its potential as a surrogate for 2’-F RNA [47]. FHNA modification of a single-stranded oligonucleotide exhibited much higher resistance to exonuclease digestion than widely used oligonucleotide modifications such as 2’-O-MOE and LNA [47]. These observations prompted us to evaluate FHNA (Fig. 1) as a surrogate for 2’-F ribonucleotides, since hexitol (HNA) and altritol (ANA) modifications enhance the nuclease stability of siRNA [61, 62]. We hypothesized that FHNA modification, with its unique structural characteristics [47], could enhance potency and metabolic stability when strategically placed in the sense and antisense strands of APOC3 siRNA. Interestingly, we observed that FHNA residues were well accommodated at positions 14 of the antisense strand and 7 and 11 of the sense strands (5-6, Fig. 4). Consistent with the reported [62] positional preference for ANA and HNA, FHNA was not well accommodated at position 2 of siRNA (7, 8, Fig. 4). However, the ability to use it at other commonly 2’-F nucleotide-modified positions demonstrates that FHNA is a viable surrogate for 2’-F ribonucleotides that can maintain activity and durability, especially when placed at positions 7 and 11 of the sense strand or position 14 of the antisense strand. To our knowledge, this is the first report of the in vivo activity of siRNA-containing FHNA chemistry. A systematic structure-activity study is needed to understand the positional preference and ultimate utility of the FHNA modification.

In addition to the reduction of 2’-F nucleotide content being able to extend the duration of action, it has been reported that extensive use of 2’-F residues in siRNA could lead to potential toxicity concerns [41], hence we pursued minimal 2’-F-containing designs by replacing 2’-F nucleotides with 2’-O-Me in both the sense and antisense strands. Our data demonstrate that siRNAs containing only four 2’-F nucleotide residues maintain efficacy and durability relative to our lead siRNA 2 with eight 2’-F modifications (Fig. 5). Mesyl-phosphoramidite (MsPA, Fig. 1) linkages are a new backbone modification reported for nucleic acid therapeutics capable of enhancing metabolic stability relative to PS modification [49, 50, 63]. As this modification is highly stable, we hypothesized that the metabolism of 2’-F-containing siRNAs could be reduced with adjacent MsPA linkages. Furthermore, complete digestion of the siRNA into monomeric 2’-F nucleosides could potentially be prevented with MsPA adjacent to 2’-F nucleotides in the siRNA, as this would presumably result in 2’-F nucleotide MsPA dimers. As these would be unlikely to serve as substrates for nucleotide salvage and polymerization enzymes in vivo, this could provide a theoretical long-term safety benefit by reducing the likelihood of incorporation of 2’-F nucleosides into RNA and DNA. We therefore studied the positional and number preference of MsPA linkages placed adjacent to 2’-F nucleotides to enhance the metabolic stability of 2’-F adjacent linkages. We discovered that single MsPA placement maintained the efficacy (12, Fig. 5) and DOA in some cases with limited substitutions. However, multiple MsPA modifications significantly reduced the efficacy of APOC3 siRNA (13, 14, Fig. 5). This data suggests that, unlike antisense oligonucleotides [49] the utility of MsPA modification for stabilizing linkages adjacent to 2’-F nucleotides in an siRNA is limited. Further studies are needed to fully understand the positional preferences of MsPA and to determine the ultimate value for enhancing the metabolic stability of siRNAs, a benefit which has clearly been demonstrated for ASOs.

MOE (Fig. 1) modification of oligonucleotides is one of the most widely used nucleic acid modifications in approved antisense drugs and those in development [64, 65]. The MOE-modified oligonucleotides show favorable target binding affinity, specificity, and nuclease stability properties [65, 66]. Earlier, the positional effect of MOE modification on the activity of siRNA in cell culture was reported [8]. Recent reports showed that placing single MOE chemistry at the seed region enhanced the selectivity of siRNA [12, 53]. However, this chemistry is not widely used in siRNAs. We decided to utilize MOE chemistry to improve the properties of the APOC3 siRNA sequence. Consistent with the previous reports, MOE modification was well accommodated at the 9, 10 of the antisense strand and the 3’ and 5’-end of the sense strand (Fig. 6–7). In the single-stranded siRNA chemical scaffold, we have shown that MOE AA overhang in this APOC3 sequence showed good activity in mice [19, 54], which translated to this work, resulting in lead siRNAs with MOE nucleosides at the termini of both sense and antisense strands.

In order to maximally reduce 2’-F nucleotides, we used DNA PS chemistry to replace 2’-F residues at positions 6, 14, and 16 of the antisense strand. A comparative duration of activity studies of APOC3 siRNA containing multiple chemical scaffolds identified in our studies demonstrated dramatically extended durability compared to siRNA 1, which was shown to be an acceptable surrogate for the clinical drug plozasiran (Fig. 8A). Unexpectedly, we discovered that siRNA 26 (Fig. 7) composite having an antisense strand containing MOE (at positions 9, 22, and 23) and DNA PS (at positions 6,14, and 16) and a sense strand containing only two 2’-F (at positions 9, 10, and PS at backbone position 10) demonstrated superior efficacy and durability in APOC3 human transgenic mice (Fig. 8A). Our systematic chemical SAR provided a siRNA scaffold having two-fold enhanced durability, maintaining 80% plasma apoC3 suppression out to 8 weeks following a single dose of 1 mg kg⁻¹ in human APOC3 transgenic mice compared to the siRNA 1, which exhibited rapid recovery after 4 weeks at the same dose (Fig. 8A). Robust efficacy and durability were demonstrated across 2 sequences with this novel scaffold. Additional sequence-specific optimization of this scaffold may be required.

Interestingly, APOC3 siRNA 26 showed improved selectivity compared to siRNA 2, which has a traditional chemical scaffold with 8 2’-F nucleotides, in the analysis of global gene expression changes in primary human hepatocytes (Fig. 9). These data suggest that our novel chemical scaffold employing only three 2’-F nucleotides is able to improve both the durability and selectivity of this APOC3 siRNA sequence.

siRNA 26, containing our novel chemical scaffold, showed excellent potency in inhibiting APOC3 mRNA in the liver (Fig. 8B), plasma apoC3 protein (Fig. 8C), and triglyceride (Fig. 8D) in human APOC3 transgenic mice. It is difficult to precisely determine the dose and frequency required for robust (e.g. sustained apoC3 reductions of > 80%) clinical responses. Attempts to determine the cross-species pharmacodynamic responses suggest a relatively wide range of biophase half-lives of various GalNAc siRNA therapeutics with mouse and human data [39]. For example, Boianelli et al. evaluated 7 GalNAc siRNAs with mouse and human target engagement biomarker data and concluded a 1- to 8-fold longer biophase half-life in humans compared to mice [39]. Such a range could be a result of many factors, such as differences in siRNA potencies and stability that may not directly translate from mouse to human. Additionally, since transgenics with different target expression levels were used for these various targets, it is unknown how much this might impact mouse-to-human scaling. For example, in these studies, we used APOC3 transgenics with extremely high APOC3 expression (>∼200 mg dL⁻¹ versus ∼10 mg dL⁻¹ plasma levels typically observed in normal triglyceridemic humans), which might influence siRNA potency and durability. Because we used a surrogate of plozasiran, siRNA 1, as a comparator in our transgenic mouse studies, these data do indicate a substantial increase in siRNA durability that would be predicted to improve the durability of plasma apoC3 and triglyceride lowering in humans compared to current APOC3 RNA therapeutics in development. Given that the improved duration of 26 (>50% reduction observed at 12 weeks) as compared to 1 (∼50% reduction at 6 weeks, with return to baseline at 8 weeks) is at least 2-fold, we would predict a dosing interval of ≥6 months in humans, given that siRNA 1 is a surrogate of a quarterly dose drug in human clinical trials.

siRNA 26 also demonstrated excellent tolerability in mice with no elevated liver enzymes or weight on repeated dosing of 300 mg kg⁻¹ for three weeks (Supplementary Table S2, Supplementary Data), which exceeds pharmacologically relevant exposure by several logs. APOC3 siRNA 26 potently reduces APOC3 mRNA in primary human and monkey hepatocytes (Fig. 10A-B). In addition, siRNA 26 showed excellent activity (Fig. 10C) and tolerability (Supplementary Table S3, Supplementary Data) in nonhuman primates (Fig. 10). In 13-week GLP toxicology assessments, no adverse siRNA 26-related findings were observed up to the highest dose tested in rodents and NHP. Thus, the NOAEL was determined to be the highest dose tested, 300 mg/kg/Q4W and 100 mg/kg/Q4W, in rodents and NHP, respectively.

In summary, we have identified a novel chemical scaffold that enhances the potency, selectivity, tolerability, and extended durability of an APOC3 siRNA, potentially surpassing the activity and durability of plozasiran. Clinical data suggest plozasiran can provide meaningful clinical benefit with quarterly dosing [32, 33]. Our APOC3 siRNA showed two-fold enhanced duration of action and enhanced potency relative to the plozasiran surrogate used in our transgenic mouse studies. Studies in non-human primates demonstrate that our lead APOC3 siRNA [26] can potentially reduce plasma apoC3 protein and triglycerides by ≥ 75%. Given these data, our APOC3 siRNA may provide clinical benefits with less frequent administration (≥ 6 months) and with robust efficacy, potentially benefiting patients with hypertriglyceridemia.

Supplementary data

Supplementary Data is available at NAR online.

Conflict of interest

All authors are employees of Ionis Pharmaceuticals Inc.

Funding

Funding to pay the Open Access publication charges for this article was provided by Ionis Pharmaceuticals Inc.

Data availability

All DGE datasets are deposited to Gene Expression Omnibus (GEO accession: GSE278574).