Application of improved GalNAc conjugation in development of cost-effective siRNA therapies targeting cardiovascular diseases

Abstract

N-Acetylgalactosamine (GalNAc)-conjugated small interfering RNA (siRNA) therapies have received approval for treating both orphan and prevalent diseases. To improve in vivo efficacy and streamline the chemical synthesis process for efficient and cost-effective manufacturing, we conducted this study to identify better designs of GalNAc-siRNA conjugates for therapeutic development. Here, we present data on redesigned GalNAc-based ligands conjugated with siRNAs against angiopoietin-like 3 (ANGPTL3) and lipoprotein (a) (Lp(a)), two target molecules with the potential to address large unmet medical needs in atherosclerotic cardiovascular diseases. By attaching a novel pyran-derived scaffold to serial monovalent GalNAc units before solid-phase oligonucleotide synthesis, we achieved increased GalNAc-siRNA production efficiency with fewer synthesis steps compared to the standard triantennary GalNAc construct L96. The improved GalNAc-siRNA conjugates demonstrated equivalent or superior in vivo efficacy compared to triantennary GalNAc-conjugated siRNAs.

Graphical abstract

Li and colleagues aimed to identify better designs of GalNAc-siRNA conjugates with improved efficacy and simplified chemical synthesis process for cost-effective manufacturing. They conducted studies on trivalent GalNAc conjugated with siRNAs against ANGPTL3 and Lp(a), and demonstrated equivalent or better in vivo efficacy compared to triantennary GalNAc-conjugated siRNAs.

Introduction

Despite the recent development of proprotein convertase subtilisin/kexin 9 (PCSK9) inhibitors to lower low-density lipoprotein cholesterol (LDL-C) either in conjunction with statins or as monotherapy, atherosclerotic cardiovascular diseases (ASCVDs) mainly including coronary artery disease, stroke, and peripheral vascular disease remain the leading global cause of death, affecting more than 500 million individuals and causing 19 million annual deaths.¹ There is a clear need for novel and cost-effective therapeutic agents.

It is widely recognized that dyslipidemia is a strong causal factor in ASCVD development, and lipid-lowering therapy is the most powerful strategy. Nucleic acid–based drugs, such as antisense oligonucleotides (ASOs) and small interfering RNAs (siRNAs), have emerged as a new class of precision medicines by targeting hepatic transcripts to modulate lipoprotein production or metabolism.² Inclisiran, an N-acetylgalactosamine (GalNAc)-conjugated siRNA against PCSK9, has demonstrated robust and durable reductions of PCSK9 and LDL-C levels with a dosing schedule of once every 6 months after initial loading doses.^3,4 Clinically, this favorable dosing schedule offers improved patient adherence to therapy compared to monoclonal antibodies targeting circulating PCSK9. It is estimated that inclisiran could become the first targeted siRNA therapy to help 1 million patients with primary hypercholesterolemia or mixed dyslipidemia.

Key interventions for ASCVDs are focused on lowering plasma LDL-C levels, as demonstrated to reduce major vascular events in clinical trials.⁵ However, approved lipid-lowering drugs showed limited impact on lipoprotein (a) (Lp(a)) or plasma triglycerides (TG), two other risk factors for ASCVDs.^6,7 Angiopoietin-like 3 (ANGPTL3), an inhibitor of lipoprotein lipase, has emerged as a novel regulator of TG and LDL-C levels. ANGPTL3 ASOs and siRNA silencing approaches have shown significant reductions in LDL-C and TG in phase I and II clinical trials.^8,9,10 Separately, genetic and population studies implicate Lp(a) as an independent risk factor for ASCVDs^11,12 and aortic valve stenosis.^13,14 In clinical trials, three oligonucleotide-based therapeutic agents targeting liver Lp(a) mRNA markedly reduced the concentration of Lp(a) in a dose-dependent manner, resulting in ∼20% LDL-C reduction with favorable safety profiles.^15,16,17

The development of targeted delivery of oligonucleotides to liver hepatocytes using GalNAc conjugates is critical for the successful approval and clinical advancement of the new generation of lipid-lowering drugs. The approach resulted in enhanced liver specificity, greater potency, and reduced systemic toxicity.^18,19,20 To maximize oligonucleotide delivery efficiency, structural improvements have been made to GalNAc design, optimizing ligand interaction with asialoglycoprotein receptor (ASGPR).^20,21,22 Common approaches include synthesizing a triantennary GalNAc (GalNAc3, L96) complex and linking it to oligonucleotides siRNA by coupling the complex to a solid support via a specific linker before oligonucleotide synthesis^18,23 (A in Figure 1). Alternatively, monomeric GalNAc can be attached to individual linkers sequentially before oligonucleotide synthesis.^24,25,26 We took advantage of the simplicity and flexibility of the latter approach and produced the cost-effective Lp(a)-lowering therapeutic candidate GNX107 with excellent preclinical in vivo efficacy.

Figure 1.

Representative GalNAc-siRNA conjugate designs

The triantennary GalNAc moiety in design A is attached to the oligonucleotide through a (3R,5S)-3-hydroxy-5-hydroxymethylpyrrolidine linker (yellow). Design B differs from design A in the unique linker (purple) that it shares with design C. The trivalent GalNAc ligand of design C comprises serial assembly of monovalent GalNAc with a 5-hydroxypentanoic acid tether (brown) between the sugar moiety (green) and the linker (purple) of (2R,3S,6R)-6-aminomethyl-2-hydroxymethyl-tetrahydro-pyran-3-ol. The conjugates are linked to the 3′ end of the SS of the siRNA by a phosphodiester linkage between the linkers and the SS in these designs.

Results

Synthesis of cost-effective siRNAs with improved GalNAc conjugation

To develop better liver-targeted siRNA therapies with superior efficacy and a streamlined, cost-effective synthesis process, we made improvements on how the conventional GalNAc clusters were linked and the solid-phase syntheses of siRNAs were carried out using L96 as our benchmark. We introduced pyran as a linker to create GalNAc clusters in a trivalent conjugation motif and used an amino resin with a higher loading capacity of 250 μmol/g for our solid-phase synthesis. These modifications reduced the synthesis steps of TrisGal-6 to 13, compared to 25 steps for L96 (Figure 2). Higher production efficiency was achieved by using fewer raw materials, resulting in at least a 30% cost savings in manufacturing compared to conventional L96 production.

Figure 2.

Comparison in synthesis steps between triantennary GalNAc L96 and trivalent TrisGal-6

Difference in synthesis step numbers between triantennary GalNAc (L96) and trivalent TrisGal-6 siRNA conjugates from respective starting materials of solid support resins, hydroxymethyl pyrrolidinol HCl, trometamol, diacetyloxy dihydropyran methyl acetate, and GalNAc.

With this newly developed proprietary synthesis process, we designed and produced a panel of siRNAs against mouse Angptl3 and LPA with different GalNAc conjugation motifs and oligonucleotide modifications through solid-phase synthesis. Each sense strand (SS) or antisense strand (AS) RNA oligo was synthesized, purified, characterized, and annealed according to published standard procedures.^18,25 Sequences, modifications, and GalNAc conjugations of all relevant siRNAs used in our studies are presented in Table 1.

Table 1.

GalNAc-siRNA conjugates used in this study

siRNA duplex ID	Abbreviation of siRNA duplex	Strand	Sequence (5′ to 3′)a,b	Target
Geno-2-102M	102M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm	ANGPTL3
		301SS	AmsCmsAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUm[L96]
Geno-2-104M	104M	303AS	UmsAfsUm CmGmAf CmGmUm GmUmCm CmAfGm CfUmAm GmsUmsUm	Non targeting control
		304SS	AmsAmsCm UmAmGm CfUmGf GfAfCm AmCmGm UmCmGm AmUmAm[L96]
Geno-2-123M	123M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm	ANGPTL3
		334SS	AmsCmsAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUm[Gal-5]
Geno-2-151M	151M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		364SS	AmsCmsAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUms[Gal-5]
Geno-2-124M	124M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		335SS	[Gal-6]sAmsCmAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUm[Gal-6]
Geno-2-125M	125M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		336SS	[Gal-6]sAmsCmAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUms[Gal-6]
Geno-2-127M	127M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		338SS	[Gal-6]s[Gal-6]sAmCmAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUm[Gal-6][Gal-6]
Geno-2-128M	128M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		339SS	[Gal-6]s[Gal-6]sAmCmAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUms[Gal-6]s[Gal-6]
Geno-2-148M	148M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		361SS	AmsCmsAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUm[Gal-6][Gal-6][Gal-6]
Geno-2-149M	149M	300AS	AmsAfsAm AmAmGf AmCmUm GmAmUm CmAfAm AfUmAm UmGmUms UmsGm
		362SS	AmsCmsAm UmAmUm UfUmGf AfUfCm AmGmUm CmUmUm UmUmUms[Gal-6]s[Gal-6]s[Gal-6]
Geno-1-107M	GNX107	359AS	AmsUfsAm AmCfUmCmUmGm UmCmCm AmUfUm AfCmCm AmUmGms GmsUm	LPA
		370SS	CmsAmsUm GmGmUm AmAmUf GfGfAm CmAmGm AmGmUm UmAmUms[Gal-6]s[Gal-6]s[Gal-6]

^aSS and AS indicates sense and antisense strand.

^bAm, Cm, Gm, and Um indicate 2′-O-methyl sugar modifications, respectively, to adenosine (A), cytidine (C), guanosine (G) and uridine (U); Af, Cf, Gf, and Uf indicate 2′-deoxy-2′-fluoro sugar modifications, respectively, to adenosine (A), cytidine (C), guanosine (G) and uridine (U); and s indicate PS linkage.

siRNAs with improved GalNAc conjugation induced robust gene silencing in mice

To evaluate the delivery efficiency of novel designs of the GalNAc conjugation, in vivo screening of these GalNAc-siRNA conjugates targeting Angptl3 was conducted in wild-type C57 BL/6J mice with single subcutaneous (s.c.) dosing. The circulating mouse ANGPTL3 protein levels in serum were measured at specified time points. Several trivalent GalNAc-siRNAs demonstrated improved in vivo efficacy compared to the commonly used L96-siRNA conjugate (Figure 3A). Specifically, as demonstrated in Figure 3B, at day 14 post injection, the three sequential Gal-6 units (TriGal-6) conjugated siRNA (148M) with a phosphodiester (PO) linkage and TrisGal-6 conjugated siRNA (149M) with a phosphorothioate (PS) linkage reduced serum ANGPTL3 levels by 83.0% and 83.7%, respectively, compared to 60.7% shown by the triantennary GalNAc ligand (Gal-5) siRNA conjugate (123M). The 149M was the most potent among the evaluated conjugations that succeeded in maintaining mRNA knockdown at 41.0% on day 28, whereas 123M and 148M showed the remaining mRNA level increasing to 71.8% and 60.9% (Figure 3B). Furthermore, 149M displayed enhanced dose-dependent reduction in serum ANGPTL3 protein, with an average of 40.1% and 76.6% at day 21 post single dose at 1 and 3 mg/kg, respectively. In the meantime, for the triantennary L96-siRNA conjugate (102M), only 25.4% and 51.5% target protein reductions were achieved (Figure 3C). At 42 days post administration, 3 mg/kg 149M was still able to maintain a 46.7% reduction in the ANGPTL3 protein, whereas its level in 102M treated animals returned to the baseline (Figure 3D). TrisGal-6 was therefore selected as the final candidate ligand for conjugation with siRNAs for therapeutic molecule development.

Figure 3.

Compare in vivo efficacy of various GalNAc-siRNAs targeting ANGPTL3 in mice

Results are presented as percentage of mouse ANGPTL3 protein remaining in serum after single s.c. administration of siRNA, relative to PBS-treated mice. Blood samples were drawn at specific time points post dose for mouse ANGPTL3 protein evaluation by ELISA. Serum mouse ANGPTL3 protein levels from individual animals were normalized to PBS-treated control group and are expressed as mean ± SEM. Data for (A) and (B) came from one study; data for (C) and (D) came from another study. (A) Gene silencing of siRNA conjugates 102M–151M (3 mg/kg) in wild-type C57BL/6 mice (n = 3). (B) Gene silencing of siRNA conjugates 123M, 148M, and 149M (3 mg/kg) in wild-type C57 BL/6J mice. Blood samples were drawn at 14 and 28 days post dose. (C) Gene silencing of siRNA conjugates 102M and 149M in wild-type C57BL/6 mice (n = 3). Mice were s.c. administered single 0.3, 1, and 3 mg/kg siRNA. Blood samples were drawn at 21 days post dose. (D) Gene silencing of siRNA conjugates 102M and 149M in wild-type C57BL/6 mice, Mice were s.c. administered single 3 mg/kg siRNA dose.

TrisGal-6 conjugated Lp(a) siRNA demonstrated robust and durable in vivo efficacy in humanized mouse model and nonhuman primates (NHPs)

We applied the TrisGal-6 conjugation to Lp(a) mRNA targeting siRNAs for the development of therapeutics for ASCVDs. Extensive studies were conducted in vitro and in vivo to screen for the best Lp(a) siRNA triggers. The final candidate (GNX107) was evaluated in human LPA plasmid hydrodynamic injection (HDI) mouse model. In these humanized mice carrying human Apo(a), a single dose of 0.3, 1, 3, or 9 mg/kg of GNX107 exhibited a dose-dependent reduction in serum Apo(a) concentrations. On day 7, Apo(a) protein reduction was 44.6%, 70.7%, 96.1%, and 98.9%, respectively, at doses of 0.3, 1, 3, and 9 mg/kg. The target protein reduction lasted longer than one month (Figure 4A).

Figure 4.

In vivo efficacy studies of TrisGal-6 conjugate siRNA GNX107 targeting LPA in mice and cynomolgus monkeys

(A) Dose response of Lp(a) siRNA with conjugates TrisGal-6 (GNX107) in hLPA HDI NOD SCID mice model (n = 5). Results are presented as percentage of human Apo(a) protein remaining in serum after single s.c. administration at 0.3, 1, 3, and 9 mg/kg doses, relative to PBS-treated mice. Blood samples were drawn at 4, 7, 14, 21, 28, and 35 days postdose for human Apo(a) protein evaluation by ELISA. Serum human Apo(a) protein levels from individual animals were normalized to a PBS-treated control group and are expressed as mean ± SEM. (B) GNX107 pharmacodynamics in cynomolgus monkeys (n = 3). Results are presented as mean percentage change from the baseline level in Lp(a) serum concentrations after a single dose of GNX107 in male cynomolgus monkeys. Blood samples were drawn once weekly post dose for human Lp(a) protein evaluation. Data are expressed as mean ± SEM.

GNX107 was further tested in male cynomolgus monkeys. A single s.c. injection of 3 mg/kg GNX107 in the animals led to more than 80% reduction in serum Lp(a) concentration for up to 8 weeks. The maximum reduction of 88.4% was achieved at day 35. At day 84 post dosing, a 52.8% Lp(a) reduction was still maintained (Figure 4B).

TrisGal-6 conjugation facilitated efficient and specific delivery of siRNAs to the liver

The pharmacokinetic parameters of GNX107 were evaluated in rats. Plasma samples were collected at different time points after a 10 mg/kg dosing. GNX107 demonstrated comparable T_max (h), C_max (ng/mL), AUC_last (h × nmol/L), and t_1/2 (h) in male and female rats (Table 2). Plasma C_max between 1 and 2 h were 56.5 and 70.8 nmol/L in female and male rats, respectively, and the plasma half life was around 2 h. We also measured the AS concentration of GNX107 in rat livers after once-weekly dosing for 2 weeks. At 48 h after the second dose, GNX107 concentrations in female and male rat livers were 59.3 and 102.3 μg/g, respectively, which were reduced to 7.8 and 18.7 μg/g at 168 h (Table 3). All of these observations were consistent with other reported GalNAc-conjugated siRNAs,²⁷ indicating the efficient uptake of the TrisGal-6-siRNAs by the liver via ASGPRs.²⁷

Table 2.

GNX107 AS plasma pharmacokinetic parameters in rats after a single s.c. dose

	Female	Male
tmax (h)	1.0 ± 0.0	1.3 ± 0.6
Cmax (nmol/L)	56.5 ± 5.6	70.8 ± 38.4
AUClast (h × nmol/L)	217.0 ± 38.9	251.0 ± 81.2
t1/2 (h)	2.1 ± 0.6	2.1 ± 0.6

Male and female Sprague-Dawley rats were administered (s.c.) a single dose of 10 mg/kg siRNA GNX107 with TrisGal-6 conjugates (n = 3). Plasma samples were collected at specific time points and kept frozen until analysis. siRNA GNX107 AS concentration was determined via LC-MS/MS. Plasma pharmacokinetic parameters were analyzed by WinNonlin. Data are presented as mean ± SD.

Table 3.

GNX107 AS liver concentration in rats after two s.c. doses

liver concentration (μg/g)	Female	Male
48 h	59.3 ± 13.2	102.3 ± 21.9
168 h	7.8 ± 5.9	18.7 ± 7.9

Male and female Sprague-Dawley rats were administered (s.c.) once-weekly doses of 10 mg/kg siRNA GNX107 with TrisGal-6 conjugates (n = 3) twice. Liver tissue samples were collected at 2 and 7 days post second dose and kept frozen until analysis. siRNA GNX107 AS concentration in liver homogenate was determined via LC-MS/MS. Data are presented as mean ± SD.

To evaluate GNX107 distribution, we measured drug concentration in mouse liver and kidney, with the former as the target organ and the latter for drug excretion.²⁸ Mice were administrated single s.c. doses of GNX107 at 10 mg/kg. Tissue samples were collected at 1 and 24 h post dosing. As demonstrated in Table 4, the liver-to-kidney ratio of GNX107 is 16.4 at 1-h post dosing. This ratio increased to 91.3 at the 24-h time point, indicating preferential hepatic accumulation of siRNA. These observations were consistent with the specific hepatic delivery of GalNAc-siRNA and its main renal route for elimination.²⁸

Table 4.

GNX107 concentration in mouse liver or kidney homogenate

	Liver concentration (ug/g)	Kidney concentration (ug/g)	Liver-to-kidney ratio
1 h	25.3 ± 4.7	1.5 ± 0.3	16.4
24 h	77.0 ± 5.7	0.8 ± 0.2	91.3

The ratio of siRNA concentration in wild-type mouse liver and kidney. Wild-type C57BL/6J mice were administered (s.c.) 10 mg/kg siRNA GNX107 with TrisGal-6 conjugates (n = 3). Liver and kidney samples were collected at 1 and 24 h postdose, respectively. The siRNA concentration was determined, and the ratio of siRNA concentration in liver and kidney was calculated.

Discussion

A number of solutions for chemical synthesis have been developed to facilitate multivalent GalNAc conjugation to siRNA for targeted liver delivery. Thus far, the synthesis approach used in all approved siRNA therapeutics required preassembled triantennary GalNAc building blocks, followed by postsynthesis conjugation to siRNA.¹⁸

We took an alternative approach by conjugating scaffolds of serial GalNAc for oligonucleotide synthesis. In a head-to-head comparison study, the siRNA conjugated with trivalent TrisGal-6 demonstrated better in vivo efficacy than did the conventional triantennary L96 (Figure 3D). These results differed from previous work in which several groups showed similar in vitro and in vivo efficacies when comparing monovalent and triantennary GalNAc-conjugated oligonucleotides.^24,25,26

The variance in outcomes may be attributed to the differences in scaffold structures among various designs, as illustrated in Table 1. When comparing the in vivo efficacy of siRNAs with our pyran-based scaffold in different GalNAc conjugation designs, such as trivalent 148M (TriGal-6 conjugated) and triantennary 123M (Gal-5 conjugated), 148M showed better Angptl3 knockdown compared to 123M on day 14 and day 28 post dosing. Meanwhile, 149M (TrisGal-6 conjugated) displayed better efficacy than 148M (TriGal-6 conjugated) on days 14 and 28 (Figure 3B). This indicated that the pyran-based trivalent (1 + 1 + 1) GalNAc conjugates were more efficacious than the triantennary design, and the PS linkage seemed to perform better than the PO linkage in this design. One possible explanation is that PS modification improves the stability of the SS and the corresponding SS/AS duplex.²⁹ Also, the trivalent structure was considered to present more resistance to the activity of phosphatases and nucleases than the triantennary structure.

However, results from our ex vivo stability assays conducted using plasma, cytosol, and tritosome samples were unable to support this hypothesis due to assay sensitivity limitations (data not shown). Recent publications suggest that there is no correlation between binding affinity to ASGPR and in vivo efficacy, where trivalent (1 + 1 + 1) GalNAc conjugates showed less binding affinity to ASGPR than triantennary ones and displayed similar in vitro efficacy in both transfection and free uptake assays. Nevertheless, trivalent (1 + 1 + 1) GalNAc conjugates functioned better than triantennary ones in in vivo study, considering both efficacy and durability,³⁰ which was also observed in our study. Due to the improved in vivo efficacy from siRNAs conjugated with TrisGal-6, it was chosen as the delivery ligand for our pipeline development.

We applied the TrisGal-6 conjugation to Lp(a)-targeting siRNAs. To ensure that our newly designed GalNAc moiety retains the pharmacokinetic properties as a liver-targeting ligand, we investigated the tissue distribution of GNX107, a TrisGal-6 conjugated Lp(a) siRNA, in mice. GNX107 demonstrated a favorable liver/kidney distribution ratio of more than 50-fold at 24 h post dosing of 10 mg/kg (Table 4), with a PK profile consistent with reported L96-conjugated siRNAs.^31,32

In the cynomolgus monkey study, a single subcutaneous dose of GNX107 at 3 mg/kg exhibited 80% reduction in serum Lp(a) concentration for up to 56 days (Figure 4B). The in vivo efficacy in monkeys at the same dose is better than or equivalent to what has been reported for olpasiran, the lead Lp(a) siRNA therapy progressing in clinical trials.

Furthermore, from a manufacturing perspective, the repeated use of smaller building blocks in TrisGal-6 with a new conjugation linker, reduced steps for synthesis, and an amino resin with high loading capacity for solid phase synthesis can boost commercial production efficiency by simplifying the synthesis process and reducing the cost of starting materials. This will, in turn, benefit patients tremendously with lower-cost drugs in the future.

In summary, we have developed a unique GalNAc design of TrisGal-6 and applied the platform to generate therapeutic agents for the treatment of ASCVDs. The improved design has the potential to produce alternative efficacious, long-lasting, and cost-effective therapies.

Materials and methods

Animals

All of the animal study protocols were approved by Institutional Animal Care and Use Committees of local contract research organizations in accordance with their guidelines. Each study was conducted ethically in compliance with regulations for experimental animals of the People’s Republic of China.

Synthesis of siRNA with improved GalNAc designs

To synthesize and evaluate siRNA-GalNAc conjugates containing new ligand designs of B and C (Figure 1), building blocks 5 with a monovalent GalNAc moiety and 12 with triantennary GalNAc display and their corresponding solid supports 7 and 13 were synthesized as shown in Figure S1. Gal-5 and Gal-6 were synthesized first and then conjugated with SSs of siRNA under standard conditions for solid-phase synthesis and deprotection, followed by annealing, purification, and characterization of the double-stranded siRNAs.

Preparation of compound 5

Compound 3 (1.45 g, 3.24 mmol) was dissolved in anhydrous dimethylformamide (DMF) (10 mL) and stirred with hexafluorophosphate azabenzotriazole tetramethyl uronium (1.64 g, 4.32 mmol), molecular sieve (1 g, 3 Å), and N,N-diisopropylethylamine (DIEA) (1.07 mL, 6.47 mmol) at room temperature for 30 min. Compound 4 (1 g, 2.16 mmol) was dissolved in DMF (10 mL) and added to the above reaction solution. The mixture was stirred at room temperature overnight under argon protection. The reaction solution was filtered and purified through a reverse column C18 to obtain compound 5 (1.8 g). ¹H nuclear magnetic resonance (NMR): (400 MHz, DMSO-d6)δ 7.80 (d, J = 9.2 Hz, ¹H), 7.71–7.73 (m, ¹H), 7.39–7.41 (m, ²H), 7.24–7.30 (m, ⁶H), 7.18–7.21 (m, ²H), 6.87 (d, J = 8.8 Hz, ⁴H), 5.21 (d, J = 3.6 Hz, ¹H), 4.95–4.98 (m, ¹H), 4.59 (d, J = 6.4 Hz, ¹H), 4.47 (d, J = 8.4 Hz, ¹H), 4.00–4.05 (m, ²H), 3.83–3.90 (m, ¹H), 3.64–3.73 (m, ⁷H), 3.23–3.38 (m, ⁸H), 2.97–3.14 (m, ³H), 2.04–2.14 (m, ⁵H), 1.98 (s, ³H), 1.89 (s, ³H), 1.77 (s, ³H), 1.64–4.66 (m, ¹H), 1.42–1.50 (m, ⁴H).

Preparation of compound 6

Compound 5 (2.1 g, 2.35 mmol) was dissolved in dichloromethane (DCM) (30 mL) and stirred with tetrazole (33 mg, 0.47 mmol), N-methylimidazole (77.2 mg, 0.94 mmol), and molecular sieve (2 g) at room temperature for 20 min under argon. Phosphorus (920.81 mg, 3.06 mmol) was dissolved in a small amount of methylene chloride and stirred into the above reaction at room temperature for 1 h. The reaction solution was washed twice with saturated NaHCO₃ solution, once with water and once with brine. It was then concentrated at room temperature and purified through reverse column C18 to obtain compound 6 (1.5 g). ¹H NMR: (400 MHz, CD₃CN) δ 7.48–7.50 (m, ²H), 7.20–7.36 (m, ⁷H), 6.83–6.87 (m, ⁴H), 6.45–6.51 (m, ¹H), 5.28 (d, J = 3.2 Hz, ¹H), 4.98–5.02 (m, ¹H), 4.49 (d, J = 8.4 Hz, ¹H), 3.90–4.12 (m, ⁴H), 3.24–3.76 (m, ²⁰H), 3.02–3.11 (m, ¹H), 2.49–2.59 (m, ¹H), 2.36–2.39 (m, ¹H), 2.08–2.25 (m, ⁹H), 1.97 (s, ³H), 1.91 (s, ³H), 1.83 (s, ³H), 1.71–1.74 (m, ¹H), 1.46–1.63 (m, ⁵H), 0.85–1.39 (m, ²⁰H).

Preparation of compounds 5B and 7

Compound 5 (420 mg, 0.47 mmol), succinic anhydride (117.74 mg, 1.18 mmol), molecular sieve (0.5 g, 3 Å), and triethylamine (0.2 mL, 0.38 mmol) were dissolved in DCM (6 mL) and stirred at room temperature overnight. The reaction solution was filtered, washed with 5% NaCl solution, and concentrated at room temperature. The product was purified through reverse column C18 to obtain compound 5B (150 mg). ¹H NMR: (400 MHz, DMSO) δ 7.88–7.69 (m, ²H), 7.42 (d, ²H), 7.34–7.15 (m, ⁷H), 6.87 (d, ⁴H), 5.21 (d, ¹H), 4.97 (dd, 3.4 Hz, ¹H), 4.61 (d, J = 6.1 Hz, ¹H), 4.47 (d, J = 8.5 Hz, ¹H), 4.01 (d, J = 7.2 Hz, ²H), 3.93–3.81 (m, ¹H), 3.73 (s, ⁶H), 3.67 (dd, ¹H), 3.38 (dd, ²H), 3.31–3.20 (m, ³H), 3.06 (m, ³H), 2.41–2.15 (m, ⁴H), 2.17–2.05 (m, ⁵H), 1.99–1.96 (m, ³H), 1.92–1.88 (m, ³H), 1.77 (s, ³H), 1.66 (d, ¹H), 1.56–1.18 (m, ⁷H).

Compound 5B (150 mg, 0.151 mmol) was suspended in acetonitrile (4 mL) and DMF (2 mL), and DIEA (0.062 mL, 0.378 mmol) and O-Benzotriazole-N,N,N',N'-tetramethyl-uronium-hexafluorophosphate (HBTU) (85.93 mg, 0.129 mmol) were added dropwise with swirling at room temperature for 5 min. Amino resin (409.43 mg, 200–400 mesh, amino content: 250 μmol/g) was added to the reaction mixture and shaken at 25°C with a speed of 220 rpm for 16 h. The support was filtered and washed 3 times with DCM, acetonitrile, and n-hexane, each 30 mL. The product was dried under vacuum for 2 h. The capping reaction was conducted with CapB1C, 4-dimethylaminopyridine, N-methylimidazole, and acetonitrile on a shaker at 25°C with a speed of 220 rpm for 16 h. The mixture was filtered and washed with 30 mL acetonitrile 3 times. The solid support was dried overnight under vacuum to obtain compound 7 (loading 250 μmol/g, 500 mg).

Preparation of compound 13

Compound 13 was synthesized according to published procedures in which the amine scaffold was replaced by pyran.²⁵

Synthesis of siRNA

Sense and antisense strands (specific sequences in Table 1) were synthesized with commercially available nucleoside phosphoramidite monomers according to published procedures under standard conditions for solid-phase synthesis and deprotection, followed by annealing, purification, and characterization of the double-stranded siRNAs.^18,25

In vivo efficacy of GalNAc-conjugated siRNAs in mice

Male C57BL/6 mice were purchased from Shanghai Lingchang Lab Animal Company. Mice 7–9 weeks old were maintained on a verified irradiated diet according to the vendor’s recommendations. All of the mice received an s.c. administration of 3 mg/kg GalNAc-conjugated siRNA. For the dose titration study, animals received an s.c. injection of GalNAc-conjugated siRNA at 0.3, 1, or 3 mg/kg with a dose volume of 5 mL/kg. Blood samples were collected via retro-orbital bleeding under anesthesia before and after dosing. Serum was prepared by centrifugation at 2,500 × g for 10 min at 4°C. The mouse ANGPTL3 level was determined by mouse ANGPTL3 ELISA kits (mANGPTL3, catalog nos. DY136 and DY008, R&D Systems) following the manufacturer’s instructions. Serum ANGPTL3 levels from individual animals were normalized to PBS-treated control group and expressed as the mean ± SEM.

Measurement of GNX107 silencing effect in plasmid hLPA HDI mouse model

Nonobese diabetic severe combined immunodeficiency (NOD SCID) mice were purchased from Vital River (Jiaxing). Male mice 6–8 weeks old were maintained on a verified irradiated diet according to the vendor’s recommendations. The gene-silencing effect of Lp(a) targeting siRNA conjugated with TrisGal-6 (GNX107) was evaluated in vivo using transgenic mice transiently expressing LPA. Ten days before GNX107 administration, plasmids containing human LPA copy DNA were injected into NOD SCID mice via tail vein HDI. These hLPA-plasmid HDI mice were then dosed s.c. with GNX107 at 5 mL/kg at doses of 0.3, 1, 3, or 9 mg/kg. Control serum samples were taken from mice in the pretreatment phase on days −3 and 0. Post-treatment serum samples were taken from the mice on days 4, 7, 14, 21, 28, and 35. Serum was prepared by centrifugation at 2,500 × g for 10 min at 4°C. The Apo(a) level was determined using a human Lp(a) ELISA kit (Abcam, catalog no. ab212165).

Measurement of silencing effect of GNX107 in NHPs

Six male cynomolgus monkeys were selected and assigned to each of the two dose groups (three animals per group) based on their Lp(a) levels. A single dose of 3 mg/kg GNX107 was administered s.c. to the animals on day 0. The whole-blood samples were collected from peripheral veins after overnight fasting at the specified time points (days 0, 7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77, and 84) in serum separation tubes. Serum was prepared by centrifugation at 2,500 × g for 10 min at 4°C. The Lp(a) level was analyzed using the Roche C311 or C501 biochemical analyzer.

Measurement of TrisGal-6 conjugated siRNA GNX107 levels in rat plasma and liver

Male and female Sprague-Dawley rats received once-weekly s.c. injections on days 1 and 8 of GNX107 at 10 mg/kg. Plasma samples were collected at specific time points postfirst dose and kept frozen until analysis. Liver samples were collected at 2 and 7 days post second dose and snap-frozen for further analysis. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was used to quantitate the concentrations of target siRNA GNX107 in plasma and liver.

Measurement of TrisGal-6 conjugated siRNA GNX107 levels in mice liver and kidney

A single dose of 10 mg/kg GNX107 was administered s.c. to male C57 BL/6J mice 6–7 weeks old (n = 3 per group) at a volume of 5 mL/kg. Animals were sacrificed at 1 or 24 h post dosing. Liver and kidney were snap-frozen for further analysis. Hybridization ELISA was used to quantitate the concentrations of target siRNAs in animal liver and kidney samples for the final liver-to-kidney ratio. cDNA probes with biotin and digoxigenin labels were synthesized to specifically bind to target siRNAs. Annealing of the probes and siRNAs was conducted by mixing 100 μL target siRNAs and probes. The complexes were then transferred to a neutravidin-coated plate, and the standard hybridization ELISA protocol was followed. Results were measured by a Tecan plate reader in fluorescence mode at the standard gain setting. All of the samples were analyzed in duplicate with quality controls and calibration standards.

Statistical analysis

Statistical analysis was carried out in GraphPad Prism 9 with one-way ANOVA followed by Tukey’s multiple comparison tests. Significance levels were indicated as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. siRNA conjugate treatments were analyzed against the vehicle control of similar conditions, unless otherwise stated in the figure legends.

Data availability

All of the data included in this study are available upon request by contacting corresponding author Wendy Wei Gu (wendy.gu@genovaltx.com).