2′-Ribose-modified nucleotides: A strategy for reducing off-target effects of oligonucleotide drugs

Abstract

Necessary chemical modifications can enhance the metabolic stability and therapeutic efficacy of oligonucleotide-based drugs. Herein, a series of mononucleotides with different functional group modifications at the ribose 2′-position were designed and synthesized for reducing off-target effects of oligonucleotide drugs. After incorporating them into the antisense strand of small interfering RNAs (siRNAs), we assessed their biophysical and biological properties. The results revealed that, following the substitution of the conventional 2′-O-methyl or 2′-O-fluoro ribose modifications with our newly designed functional groups, although the thermal stability of the duplexes decreased, the in vitro silencing activity against proprotein convertase subtilisin/kexin type 9 (PCSK9) gene was significantly enhanced and the off-target effects of specific genes could be effectively mitigated. These data prove that 2′-ribose-modified nucleotides can be effectively utilized in preclinical and clinical disease model studies for gene silencing and reducing off-target effects.

Graphical abstract

Song and colleagues have successfully synthesized a panel of 2′-modified ribonucleotides and embedded them within the seed region of PCSK9-directed siRNAs. These substitutions preserve potent PCSK9 silencing while simultaneously suppressing off-target effects and the attendant hepatotoxicity.

Introduction

In 1998, Andrew Fire and Craig Mello reported that double-stranded RNA (dsRNA) has potent gene-silencing effects in Caenorhabditis elegans and first introduced the concept of RNA interference (RNAi).¹ RNAi is a post-transcriptional gene regulation pathway mediated by small interfering RNAs (siRNAs).^2,3,4 Upon siRNAs loading onto Argonaute 2 (Ago2), the sense strand is discarded and the RNA-induced silencing complex (RISC) is formed, which recognizes and binds to complementary mRNA to cleave the target mRNA using its endonuclease activity, leading to its degradation and thereby reducing the expression of the encoded protein.⁵ However, when serving as effective therapeutic agents, natural RNA duplexes still face numerous challenges to overcome.⁶ For instance, they exhibit poor stability in vivo, as the phosphodiester bonds of oligonucleotides are highly susceptible to rapid degradation by various nucleases that are widely present in the blood and cells.^7,8 Additionally, they have low binding affinity and specificity to target genes, which can easily lead to “off-target” effects, and they can be recognized by Toll-like receptor (TLR) pathways. These factors trigger immune stimulation and toxic side effects.^9,10 Moreover, oligonucleotides are typically large polyanionic molecules, making it difficult for them to enter target organs and tissues and to cross the lipophilic cell membrane to enter cells. Besides, the unmodified RNAs are cleared very quickly in blood. These result in their low bioavailability.^11,12 Thus, the synthetic siRNA must include chemically modified nucleotide-building blocks for use as a therapeutic.

Since the 1970s, numerous strategies for nucleotide structural modifications have been successfully developed through alterations to the phosphate backbone, sugar moieties, and nucleobases. So far, many of these modified nucleotides are used for the clinical treatment of multiple diseases including cardiovascular, endocrine and metabolic disorders, bacterial infections, tumors, neuromuscular diseases, dermatological, gastrointestinal, and ophthalmological conditions.^{11,13,14,15,16,17,18,19,20} All siRNA drugs approved by the Food and Drug Administration (FDA) for clinical use have incorporated chemical modifications and been delivered via methods such as lipid nanoparticles (LNPs) or a trivalent N-acetylgalactosamine (GalNAc) conjugation.^{21,22,23,24,25,26,27,28,29} Common chemical modifications involve substitutions at the C2' position of the ribose ring, where the hydroxyl group (-OH) is replaced with 2′-O-methyl (2′-OMe) or 2′-deoxy-2′-fluoro (2′-F). These modifications can maintain and potentially enhance preorganization of the nucleotide for an RNA-like C3′-endo conformation and are favorable binding to Ago2. Among them, 2′-OMe stabilizes the C3′-endo conformation less effectively than 2′-F due to steric hindrance and electronegativity. Nevertheless, both modifications can enhance binding affinity to RNA and increase resistance to nuclease degradation compared with siRNAs containing the parent ribonucleotides.^30,31,32,33 Besides, off-target toxicity is also one of the key issues and challenges in the development of oligonucleotide drugs. This toxicity primarily arises from the binding of drug molecules to non-target RNAs, which induces toxicological responses. For siRNA, the major off-target effects stem from the pairing of the antisense strand seed region (nucleotides 1–9 at the 5′ end) with non-target mRNAs.^34,35 This interaction leads to aberrant gene expression through a mechanism similar to that of microRNAs (miRNAs), resulting in miRNA-like off-target effects.³⁶

Although thermodynamically destabilizing modifications like unlocked nucleic acid (UNA) or (S)-glycol nucleic acid (GNA) have the ability to mitigate off-target effects,^{28,37,38,39,40,41,42,43,44,45} they simultaneously increase the risk of reducing gene silencing activity due to structural alterations of the ribose ring. Therefore, we continued to explore strategies that preserve the ribose ring structure while introducing modifications at the 2′ position This approach aims to reduce off-target effects while ensuring gene silencing activity. Herein, we first describe the synthesis of several nucleotides modified at the 2′ position (Figure 1). Then, we assessed the impact of these modifications on the thermal stability and conformation of RNA duplexes targeting the human PCSK9 gene (Figure 1), both on-target and off-target activities of siRNA. Finally, our findings suggest that further evaluation of these modifications for reducing off-target effects is warranted in the context of siRNA-based therapeutic approaches.

Figure 1.

siRNA sequence and chemical structures of mononucleotides

(A) Schematic of the parent siRNA duplex targeting human PCSK9. This siRNA was previously characterized.⁴⁶ (B) Structures of 2′-OMe, 2′-F, (S)-GNA, natural phosphate, and phosphorothioate linkages. (C) Structures of 2′-modified mononucleotides used in this study.

Results

Mononucleotide and oligonucleotide synthesis

A common intermediate INT-I was synthesized from a commercial starting material methyl β-D-ribofuranoside via a four-step reaction. Subsequently, modification at 2′ position and glycosidation were applied to produce the corresponding nucleosides. Finally, desired 2′-modified pyrimidine phosphoramidites that were named from YK-NUM-101∼105 were obtained using standard nucleoside protection with a 4,4′-dimethoxytrityl group at the 5′ position and phosphitylation (Figures S1–S7). Among these phosphoramidites, YK-NUM-101 and 102 shared the same modification at the 2′ position, but their bases were N-acetylcytosine and uracil, respectively. In contrast, YK-NUM-102∼105, which all had uracil, contained methoxymethyl, fluoromethyl, N-(2-methoxyethyl)carbamido, and N, N-diethylformamido group at 2′ position, respectively (Figure 1). The characterization details of intermediates and 2′-modified-2′-deoxynucleoside phosphoramidite monomers are provided in the supplemental information (Figures S8–S19).

YK-NUM-101∼105 mononucleotide building blocks were successfully incorporated into oligonucleotides via solid-phase synthesis. Following cleavage and deprotection, the target oligonucleotides were obtained in the acceptable yields (85%–95%) (Table S1) and analytical HPLC revealed high purity (>90%) for all synthesized sequences (Figure S20), and LC-MS analysis confirmed the expected molecular weights (Table S2; Figure S21), consistent with the successful site-specific incorporation of the modified nucleotides.

Thermodynamic stabilities of duplexes containing 2′ modification nucleotides

The melting temperatures (T_ms) of siRNA duplexes containing single 2′-modified-mononucleotide or GNA were determined. The parent siRNA duplex (siRNA-1), a fully modified 21-mer duplex with 2′-OMe and 2′-F, had a two-nucleotide overhang at the 3′ end of the antisense strand siRNA-1. To enhance stability, phosphorothioate (PS) linkages were incorporated at both ends of the antisense strand and at the 5′ end of the sense strand (Table 1; Figure S26). The sequence with high PCSK9 inhibition has been explored in our previous work.⁴⁶ The modified residues were located at position 6 or 7 from the 5′ end of the antisense strand. The effect of the modification on thermal stability was evaluated by calculating the differences in observed T_m values between the modified and the controlled (siRNA-1) RNA duplexes. The ΔT_m relative to siRNA-1 ranged from −18.4°C to −21.5°C, with the specific values depending on the identity of the 2′-modification. Notably, the incorporation of 2′-N,N-diethylformamido-modified nucleotides (siRNA-6) resulted in the most significant thermodynamic destabilization, with a ΔT_m of −21.5°C. In this study, GNA modification was employed as a positive control group. As shown Table 1, the RNA duplex with GNA (siRNA-7) also achieved a significant reduction in T_m, with a decrease of 12.8°C compared to siRNA-1, but the degree of decline was not as pronounced as that observed with siRNA-2 to siRNA-6.

Table 1.

Melting temperatures of RNA duplexes with 2′ modified nucleotides or GNA

siRNA no.	Sequencea	Tm (°C)b	ΔTm (°C)c
siRNA-1 Parent	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUUCUAUAsAs-5′	84.4	–
siRNA-2	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUUC101UAUAsAs-5′	66.0	−18.4
siRNA-3	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUU102CUAUAsAs-5′	65.2	−19.2
siRNA-4	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUU103CUAUAsAs-5′	64.5	−19.9
siRNA-5	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUU104CUAUAsAs-5′	64.5	−19.9
siRNA-6	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUU105CUAUAsAs-5′	62.9	−21.5
siRNA-7	5′-CsUsUUUGUAACUUGAAGAUAUU-3′ 3′-ACsGsAAAACAUUGAACUUGNACUAUAsAs-5′	71.6	−12.8

^aUppercase letters indicate 2′-OMe. Italics indicate 2′-F. Letters with subscript indicate YK-NUM-101∼105 and GNA, respectively. Lowercase “s” indicates phosphorothioate linkage.

^bT_m values were obtained from the maximum of the first derivatives of the melting curves (Abs₂₆₀ vs. temperature) recorded in PBS buffer (137 mM NaCl, 2.7 mM KCl, 8 mM Na₂HPO₄, 8 mM KH₂PO₄, pH 7.4) with 2.0 μM concentrations of each strand. Each value was the average of three independent experiments.

^cΔT_m is the difference in melting temperature between the parent and the modified RNA duplexes.

Circular dichroism spectroscopy studies

The CD spectra revealed that all RNA duplexes, including the control (siRNA-1), 2′-modified nucleotide-containing duplexes, and GNA-modified duplex (siRNA-7), exhibited typical A-form conformational characteristics (Figure S22). Specifically, a strong negative peak was observed near 210 nm, and a pronounced positive peak near 260 nm, resulting from π–π∗ transitions of nucleobases. Compared to the control (siRNA-1) and the GNA-containing duplex (siRNA-7), RNA duplexes containing 2′-modified nucleotides exhibited stronger absorption intensities near 260 nm.

In vitro RNAi activity of siRNAs with 2′-modified nucleotides

The gene silencing activities of siRNAs with 2′-modified nucleotides targeting PCSK9 were evaluated in cell culture. The parent RNA duplex (siRNA-1) mentioned above was employed as a comparison. To systematically evaluate the siRNAs that contained 2′ modified nucleotides on the gene silencing activity, each nucleotide in antisense strands was individually replaced with its modified nucleotide or GNA counterpart. siRNA-2 featured a cytosine, resulting in the replacement at position 6 from the 5′ end of the antisense strand. In contrast, siRNA-3∼siRNA-7 contained uracil as their base, and thus the replacement occurred at position 7. Apart from these differences, all seven siRNAs were identical, as shown in Figure S23.

Human hepatocarcinoma (Hep3B) cell line was applied for the evaluation, and quantification was accomplished by real-time qPCR at 24 h after addition of the siRNA to the cell line at a range of concentrations (15, 3, 0.6, 0.12, 0.024, 0.0048, and 0.00096 nM; siRNA-4 had an additional point at 0.000192 nM to refine the IC₅₀ determination) using the transfection reagent RNAiMAX. The impact of 2′ modification of nucleotide on the gene-silencing activity was quantitatively assessed using the half-maximal inhibitory concentration (IC₅₀) values of siRNA duplexes. The results demonstrated that all tested siRNAs exhibited robust inhibitory effects on the target gene (Figure 2). Moreover, the suppression efficiency of the siRNAs progressively decreased with a reduction in their concentration. The IC₅₀ value of parent siRNA duplex was 0.043 nM. The GNA modification, as expected, exhibited an inferior suppression compared to the parent, and specifically, siRNA-7 had an IC₅₀ of 0.081 nM, almost 2-fold loss of potency relative to siRNA-1. All siRNAs containing 2′ modified nucleotides, siRNA-2∼siRNA-6, exhibited better gene silencing effects than siRNA-7 and had potency similar to or even better than that of parent. Notably, siRNA-4 with YK-NUM-103 exhibited an IC₅₀ value of 0.002 nM, 21.5- and 40.5-fold gain of potency relative to the parent siRNA and GNA-modified siRNA, indicating an exceptionally significant inhibitory effect on PCSK9.

Figure 2.

Fitted dose-response curves for determination of IC₅₀ values of siRNAs targeting PCSK9 mRNA

(A–G) qPCR analysis of the inhibitory efficiency showing log siRNA concentration (horizontal axis) and mRNA expression% of control (vertical axis) from siRNA-1 to siRNA-7 (n = 3 wells). Data represent the mean ± SD.

We also plotted the IC₅₀ values against the difference in T_m between the 2′-modified siRNA and the parent to investigate whether there was a correlation between the thermal stability of duplexes containing 2′-modified nucleotides and in vitro silencing activity.³³ The results revealed no statistically significant linear correlation between the binding affinity of the siRNA duplexes and their silencing efficiency (Figure S24).

2′-Modified nucleotides mitigate off-target effects

A homology search of our parent siRNA sequence against publicly available human RNA sequences was performed, and 13 potential off-target genes were identified. Subsequently, we narrowed down the list to three off-target genes: hDSE, PCYOX1, and TRIM65 by calculating off-target scores.⁴⁷ We then assessed the off-target effects of the seven siRNA duplexes mentioned above on these three genes. The experimental approach was similar to the method used to determine PCSK9 inhibition efficiency, with the target gene replaced accordingly. qPCR was employed to quantify the knockdown efficiency levels of these genes.

Upon investigating the knockdown efficiency of siRNAs targeting hDSE, siRNA-2, siRNA-4, and siRNA-6 had better performance than the parent siRNA-1 and positive control siRNA-7 (Figure 3). Specifically, it was observed that parent siRNA-1, siRNA-3, siRNA-5, and positive control siRNA-7 exhibited a pronounced upward trend with increasing siRNA concentration. At the highest concentration of 15 nM, their inhibition rates reached relatively high levels of 47.56%, 45.61%, 45.35%, and 34.06%, respectively, implying potential off-target risks of these siRNAs for hDSE. Comparatively, although siRNA-2 also showed concentration-dependent inhibition, its inhibition rate was relatively low, standing at 28.92% at 15 nM, suggesting a lower potential off-target risk compared to siRNA-1, siRNA-3, siRNA-5, and siRNA-7. Notably, the knockdown efficiencies of siRNA-4 and siRNA-6 did not increase with rising concentration, and at 15 nM, their inhibition rates were merely 17.83% and 22.25%, respectively, demonstrating their better performance than siRNA-1 and siRNA-7 in mitigating the off-target effects of hDSE.

Figure 3.

Fitted dose-response curves for determination of the off-target effects of different siRNAs targeting hDSE

(A–G) qPCR analysis of the inhibitory efficiency showing log siRNA concentration (horizontal axis) and mRNA expression% of control (vertical axis) from siRNA-1 to siRNA-7 (n = 3 wells). Data represent the mean ± SD.

In the process of detecting the knockdown efficiency to PCYOX1, siRNA-3∼siRNA-6 exhibited off-target resistance comparable to that of GNA-modified siRNA (siRNA-7) (Figure 4). Specifically, it was found that the inhibition rates of siRNA-1 and siRNA-2 had the concentration dependence. When the concentration reached 15 nM, their inhibition rates were measured to be 45.99% and 20.79%, respectively. This finding implies that both siRNA-1 and siRNA-2 pose off-target risks in the context of PCYOX1, while siRNA-2 exhibited relatively lower risk. At 15 nM, siRNA-3∼siRNA-7 demonstrated superior off-target mitigation capabilities for the PCYOX1 gene compared to siRNA-1. Notably, the inhibition rates of siRNA-5 and siRNA-6 exhibited a marked decline relative to siRNA-1, dropping by 41.17% and 39.85%, respectively.

Figure 4.

Fitted dose-response curves for determination of the off-target effects of different siRNAs targeting PCYOX1

(A–G) qPCR analysis of the inhibitory efficiency showing log siRNA concentration (horizontal axis) and mRNA expression% of control (vertical axis) from siRNA-1 to siRNA-7 (n = 3 wells). Data represent the mean ± SD.

Systematic evaluation of TRIM65 targeting revealed distinct pharmacological profiles among the siRNA candidates (Figure 5). Quantitative analysis showed that siRNA-1 exhibited classical concentration-dependent activity, with increasing inhibition rates at higher concentrations and reaching a maximum inhibition rate of 20.64% at 15 nM. In contrast, siRNA-2 to siRNA-7 demonstrated flat response curves across the tested concentration range. These results suggest that siRNA-1 may pose off-target risks in TRIM65 targeting, whereas siRNA-2∼siRNA-7 exhibit superior on-target specificity compared to siRNA-1.

Figure 5.

Fitted dose-response curves for determination of the off-target effects of different siRNAs targeting TRIM65

(A–G) qPCR analysis of the inhibitory efficiency showing log siRNA concentration (horizontal axis) and mRNA expression% of control (vertical axis) from siRNA-1 to siRNA-7 (n = 3 wells). Data represent the mean ± SD.

RNA sequencing analysis of siRNAs with 2′-modified nucleotides

Encouraged by the in vitro RNAi activity and off-target mitigation of siRNAs with 2′-modified nucleotides, we conducted strand-specific RNA sequencing (RNA-seq) to quantify off-target effects of siRNAs bearing 2′-modified nucleotides. Hep 3B cells were transfected with the indicated siRNAs for 48 h, and differentially expressed genes (DEGs) were analyzed (Table 2; Figure S25). Based on the results, the transcriptomic profiles of siRNA-2 and siRNA-3 were comparable to that of the parent siRNA-1, both of which induced a substantial number of treatment-related and sequence-related DEGs. In contrast, siRNA-5 mirrored the GNA-modified siRNA-7, with fewer of both DEGs, whereas siRNA-4 and siRNA-6 exhibited the lowest numbers of treatment- and sequence-related DEGs. These data imply that appropriate 2′-modifications can possibly mitigate siRNA off-target activity while preserving on-target efficacy.

Table 2.

No. of DEGs in Hep 3B

Name	No. of DEGs in human	No. of seed-region-related DEGs in human
siRNA-1	320	18
siRNA-2	254	10
siRNA-3	232	9
siRNA-4	110	2
siRNA-5	219	7
siRNA-6	170	3
siRNA-7	264	6

In vivo activity of siRNAs with 2′-modified nucleotides

Next, we conjugated these siRNAs to GalNAc, which can be administered subcutaneously, then evaluated them in mice. The humanized PCSK9 mice were dosed with a single subcutaneous injection at a dose of 3 mg/kg. Serum PCSK9 levels were determined by ELISA on days 7, 14, 21, 28, and 35 post-dosing (Figure 6A). All siRNAs achieved peak inhibition on day 7, reducing serum PCSK9 by ≥ 80%, and maintained significant suppression for at least 35 days. Throughout the observation period, siRNA-4 and siRNA-6 consistently exhibited better serum PCSK9 suppression relative to the parent siRNA-1, whereas siRNA-2, siRNA-3, siRNA-5, and the GNA-modified siRNA-7 achieved lower knockdown at every time point. An identical trend was observed for serum low-density lipoprotein cholesterol (LDL-C) and total cholesterol (TC) reduction (Figures 6B and 6C).

Figure 6.

In vivo activity and toxicology tests of siRNAs with 2′ modification

(A) Inhibition of serum hPCSK9 in mice at single dose of 3 mg/kg on days 7, 14, 21, 28, and 35 (n = 6). (B) Inhibition of serum LDL-C in mice at single dose of 3 mg/kg on days 7, 14, 21, 28, and 35 (n = 6). (C) Inhibition of serum TC in mice at single dose of 3 mg/kg on days 7, 14, 21, 28, and 35 (n = 6). (D–F) Serum ALT, AST, and GLDH levels of rats administrated with candidate siRNAs (n = 6). (G) Histopathology examinations of rat livers. ∗∗∗∗p < 0.0001; ns means no significant difference. Data represent the mean ± SD.

We also evaluated the safety of 2′-modified nucleotides siRNA against the control at doses considerably higher than the pharmacologically effective dose in rats. The mice were dosed with three weekly 30 mg/kg subcutaneous injections of siRNA-1, siRNA-4, siRNA-6, and siRNA-7. The control group was administered with 0.9% sodium chloride (saline). Liver injury markers alanine aminotransferase (ALT), aspartate aminotransferase (AST), and glutamate dehydrogenase (GLDH) were detected at 24 h after last dose (Figures 6D–6F). The level elevations of siRNA-1 were significantly elevated, whereas those for siRNA-4, siRNA-6, and siRNA-7 did not exhibit significant differences compared to the control group. Rats were then sacrificed at the endpoint, and histopathology examinations were performed on liver (Figure 6G). Compared to the parent siRNA-1, siRNA-4 and siRNA-6, as well as siRNA-7 with GNA modification induced minimal to mild hepatotoxicity, confirming that the 2′-modifications are as effective as GNA modifications in mitigating hepatotoxic effects.

Discussion

The development of siRNA therapeutics represents a breakthrough in precision medicine for chronic conditions mediated by the overexpression of specific mRNAs. Particularly noteworthy is the clinical success of inclisiran, which has demonstrated the therapeutic potential of siRNA technology in managing HeFH—a genetic disorder marked by impaired LDL-C metabolism.^21,22 Despite significant advancements, the development of siRNA pharmaceuticals continues to face substantial challenges, with off-target toxicity representing a critical translational barrier. In this work, we introduced an innovative 2′-ribose-modified strategy to reduce the off-target effect for oligonucleotide drugs.

First, we investigated the stereochemical configurations at the 1′ and 2′ positions of the nucleoside during the synthesis process of YK-NUM-101∼105. Generally, the ribose and nucleobases were used to prepare nucleosides via the Silyl-Hilbert-Johnson reaction, which typically yielded a mixture of α- and β-anomers in varying ratios. After purification, the natural β-anomer could be separated from the undesired α-anomer. The α- and β-anomers were readily distinguished by their ¹H NMR spectral data. For the β-anomer, H1' and H2' exhibit a cis coupling (in the furanose form, J_H1-H2 ≈ 6–10 Hz), whereas for the α-anomer, H1' and H2' show a trans coupling (J_H1-H2 ≈ 2–5 Hz).⁴⁸ In the synthesis of this series of compounds, YK-NUM-101-PM2, YK-NUM-102-PM1, YK-NUM-104-PM3, and YK-NUM-105-PM2 tended to afford exclusively the β-anomer. This stereoselectivity can be mechanistically explained by the formation of a cyclic oxocarbenium ion intermediate, where the oxygen atom of the 2′-substituent interacts with the C1' carbocation of the ribose ring,⁴⁹ favoring the formation of the β-anomer. In contrast, for YK-NUM-103-PM2, which bears a fluoromethyl group (-CH₂F) at the 2′-position, a slight increase in the α-anomer could be observed. Nevertheless, the β-anomer remained the major product and could be effectively separated from the α-anomer by column chromatography (β/α ≈ 3/1). The ¹H NMR spectra (Figure S8) revealed a J_H1-H2 coupling constant of 3.2 Hz for the α-anomer (YK-NUM-103-PM2-α-Anomer) and 7.6 Hz for the β-anomer (YK-NUM-103-PM2).

Regarding the stereochemistry of the 2′-substituent of the ribose, it has been previously reported that a defined chiral configuration can be obtained via specific synthetic protocols.^50,51 The hydroxyl groups at the 3′- and 5′-positions of the D-ribose ring are first protected. Subsequently, the hydroxyl group at the 2-position is converted to an alkenyl group via oxidation followed by a Wittig reaction. A downward-oriented hydroxymethyl group is then installed through a hydroboration-oxidation sequence employing 9-BBN and hydrogen peroxide. In the present work, this synthetic route was adopted. Initially, 9-BBN was also used in the hydroboration-oxidation step from INT-PM3 to INT-I, which afforded the desired product in low yield (∼28%). After reaction optimization, 9-BBN was replaced with borane. Analysis of the ¹H NMR spectra confirmed that the product obtained possessed the same stereochemical structure, and the yield (∼50%) was significantly improved.

Then the siRNA sequences were obtained on an automated synthesizer that incorporated the YK-NUM-101∼105 mononucleotide building blocks into the seed region. All of the mononucleotides are mixtures of two diastereomers (Rp and Sp). Given that the modified nucleotide monomers would not retain their chiral configuration during the solid-phase synthesis, no diastereomeric separation was performed. The seed region refers the positions 2 to 8 from the 5′ end of the siRNA antisense strand, which is the critical region for binding between siRNA and the target mRNA and has a significant impact on the specificity and off-target effects of siRNA. The introduction of GNA into the seed region of the siRNA antisense strand reduces the stability of base pairing between individual nucleotides, thereby lowering the T_m and achieving the goal of mitigating off-target effects.^28,52 Subsequently, we evaluated the siRNA duplexes containing 2′-modified nucleotides and GNA on their thermodynamic stabilities. Compared with the parent siRNA-1, all siRNAs bearing 2′-position modifications exhibited a significant reduction in T_m. Given the critical role of duplex stability in siRNA activity and specificity, a thorough mechanistic exploration of this striking T_m reduction is essential. This observation has been previously reported that the incorporation of certain 2′-modified ribonucleoside monomers into oligonucleotide chains can lead to a decrease in the T_m.^32,53,54,55 The extent of the T_m reduction varies depending on the specific modifying group. For instance, it has been reported that each 2′-formamido modification reduces the T_m by 6.0°C–7.6°C, regardless of the nucleobase type.⁵³ Separately, a T_m decrease of approximately 15°C could be observed for oligonucleotides containing 2′-F/Me modifications compared to their unmodified counterparts.³² In the present work, the incorporation of the five synthesized 2′-modified monomers into oligonucleotide sequences resulted in the T_m reduction about 20°C relative to the unmodified sequence (Table 1; Figure S26).

This significant decrease in melting temperature can be explained by three primary factors. First, the introduction of bulky substituents at the 2′-position of the sugar ring can physically impede the close approach of the two strands, thereby interfering with base stacking and pairing. Second, the modifications can directly participate in or otherwise interfere with the formation of canonical Watson-Crick hydrogen bonds. Third, the modifications can force the sugar-phosphate backbone to adopt a conformation that is incompatible with the standard duplex geometry.⁵⁶ It indicates that the 2′-modified nucleotides induce subtle conformational changes in the RNA duplex and prevent the paired bases from forming stable interactions due to steric hindrance, resulting in a significant reduction of T_m compared to the parent siRNA duplex. Besides, the degree of decline of siRNA-7 with GNA was not as pronounced as that observed with siRNA-2 to siRNA-6. This result implies that the 2′-modified nucleotides we designed might possess the better off-target effects of siRNA.

Next, the circular dichroism spectroscopy was also investigated (Figure S22). The intensity of these peaks was significantly influenced by the type of functional group modification, the position of insertion, and the corresponding base. This suggests that the insertion of single modified nucleotides or GNA into the antisense strand does not alter the overall conformation of the RNA duplex or its ability to form double-stranded structures with complementary strand, while it does affect the regularity of the RNA duplex. The five new monomers possess bulky, flexible side chain that destabilize the helical state through steric clash and conformational entropy penalty.^53,57 Specifically, the large side chain introduces severe steric clashes with residues on the complementary strand, while replacement of the 2′-oxygen with a carbon atom slightly lowers the rotational barrier of the 2′-substituent and, more importantly, increases the number of rotatable bonds in the side chain. These factors impose a conformational entropy penalty and generate local disorder without altering the C3′-endo sugar pucker. Consequently, our designed nucleosides maintain a C3′-endo sugar conformation yet exhibit a reduced melting temperature.

PCSK9 (proprotein convertase subtilisin/kexin type 9) is a serine protease encoded by the PCSK9 gene and predominantly produced by the liver. It binds to the low-density lipoprotein receptor (LDL-R) on the surface of hepatocytes, leading to the degradation of LDL-R. Given that LDL-R is responsible for binding LDL-C in the plasma and facilitating its uptake into the liver, where plasma lipids are processed and excreted via bile, the degradation of LDL-R results in elevated plasma lipid levels.^58,59 siRNA targeting PCSK9 inhibits its expression, thereby increasing the levels of LDL-R on the surface of hepatocytes. This, in turn, reduces plasma LDL-C levels, achieving the goal of plasma lipid-lowering. In this study, while all tested siRNA duplexes demonstrated inhibitory effects on PCSK9 expression, significant potency variations were observed (Figure 2). Notably, siRNA-4 exhibited superior silencing efficacy with an exceptionally low IC₅₀ of 0.002 nM. This pronounced activity differential underscores the critical influence of 2′-position modification sterics on RNA interference efficiency. The comparatively lower IC₅₀ of siRNA-4 appears to be associated with the 2′-CH₂F substitution introduced by monomer YK-NUM-103, suggesting a potential contribution to its enhanced potency. Replacing the 2′-OMe or 2′-CH₂-OMe groups present in siRNA-1∼siRNA-3 with 2′-CH₂F simultaneously increases lipophilicity; this combination may strengthen membrane association and displace water molecules from the sugar edge, accelerating target-RNA recognition. Compared with siRNA-5 and siRNA-6, 2′-CH₂F imposes minimal steric demand, locking the duplex in a near-ideal A-form geometry and thereby enhancing RISC loading efficiency. Subsequent in vivo efficacy studies also demonstrated that siRNA-4 has superior capability to inhibit PCSK9 protein levels in the serum of humanized mice compared to the parent siRNA-1 and other modified siRNAs, with a similar trend observed in the inhibition of LDL-C and TC levels (Figures 6A–6C). This suggests that appropriate 2′ modifications can effectively enhance the targeting efficacy of siRNAs. Furthermore, the relationships between IC₅₀ values and T_m suggests that, aside from its influence on the strand-asymmetry-mediated displacement of the sense strand during the loading of the antisense strand into RISC,⁶⁰ the stability of siRNA duplexes is not a critical determinant of RISC activity.

Safety is crucial for the development of siRNA therapeutics, and off-target effects mediated by RNAi are a major cause of hepatotoxicity associated with siRNAs. Thus, we evaluated the off-target effects. In practical applications, a significant number of non-target mRNAs that exhibit only partial complementarity to the seed region of the antisense strand of siRNA are suppressed. This phenomenon is primarily attributed to off-target effects, which cause gene silencing of unintended targets through the miRNA-type interactions. To address this issue, Alnylam Pharmaceuticals has employed GNA modification at position 7 of the siRNA antisense strand to disrupt the seed region, thereby significantly reducing off-target effects and alleviating hepatocellular toxicity.⁶¹ Such modifications can influence the way siRNA binds to non-designated targets via seed region recognition, thereby inhibiting off-target effects. So, we applied GNA modification (siRNA-7) as positive control. In our cases, all of the siRNA contained 2′ modified nucleotides, and GNA contributed to a lower Tm compared to the unmodified parent siRNA, indicating less thermodynamic stability. Thus, it was reasoned that 2′ribose-modification should reduce off-target effects. Three off-target genes, hDSE, PCYOX1, and TRIM65, were determined to investigate the off-target effect (Figures 3, 4, and 5). Overall, siRNA-2∼siRNA-6 outperformed siRNA-1 in terms of off-target effect mitigation capabilities. Specifically, siRNA-1 exhibited off-target risks across all three tested genes. siRNA-2 presented a slight off-target risk in the context of PCYOX1, while siRNA-3 and siRNA-5 in hDSE exhibited off-target risks. By contrast, siRNA-4 and siRNA-6 were free of off-target risks for any of these three genes and even demonstrated better capabilities than the positive control siRNA-7 in mitigating off-target effects. To further investigate, we employed RNA-seq to comprehensively evaluate the off-target effects of siRNAs for toxicology studies (Table 2; Figure S25). The RNA-seq results were consistent with above observation, where both siRNA-4 and siRNA-6 induced fewer DEGs than siRNA-7, indicating that these two siRNAs potentially confer a lower off-target effect risk. This finding also corroborates the outcomes of the hepatotoxicity assays performed in rats (Figures 6D–6G). In the in vivo toxicity experiment, siRNA-4 and siRNA-6 bearing specific 2′-position modifications, and GNA-modified siRNA-7, which exhibit reduced off-target effects, all demonstrated lower hepatotoxicity in rats compared to the parent siRNA-1, indicating that reducing off-target silencing may mitigate liver injury induced by GalNAc-conjugated siRNAs.

In conclusion, the incorporation of 2′-modified mononucleotides into the antisense strands has no impact on the antisense strands’ capacity to pair with the sense strand and form a complete RNA duplex structure. However, the melting temperature of these modified siRNA duplexes is reduced, indicating decreased thermodynamic stability. In the in vitro gene silencing activity assays, it is observed that siRNA duplexes containing 2′-modified nucleotides exert the inhibitory effect on the expression of the target gene PCSK9, which is more pronounced than that of the parent siRNA. Notably, siRNA-4 demonstrates a 21.5-fold reduction in IC₅₀ value compared to the parent, indicating a highly significant inhibitory effect. Additionally, we evaluated off-target effects and found that siRNA-2∼siRNA-6 collectively exhibited superior off-target mitigation compared the parent siRNA-1. Furthermore, the RNA-seq and hepatotoxicity assays indicate that siRNA-4 and siRNA-6 bearing 2′ modified nucleotides potentially confer a lower off-target effect risk and the attendant hepatotoxicity compared with siRNA-7, which contained GNA modification and has been previously validated. The results suggest that the 2′-modified mononucleotides we designed here increase potency of action of siRNAs and mitigate off-target effects and thus will likely enhance the safety of siRNAs. Our data indicate that the siRNAs containing 2′-modified nucleotides warrant further investigation in preclinical and clinical models of disease for effective gene silencing.

Materials and methods

Chemical synthesis of 2′-ribose-modified nucleotides

Procedures for the synthesis and characterization of the mononucleotides are given in the supplemental information.

Solid-phase oligonucleotide synthesis

Oligonucleotides were synthesized on a Syn-HCY-192P DNA/RNA synthesizer using standard solid-phase synthesis and deprotection protocols. The solid-phase carrier was a universal carrier of cross-linked polystyrene beads, model Primer Support 5G UnyLinker 350 (purchased from Cytiva manufacturer). Synthesis was performed at 1-μmol scale using commercially available (S)-GNA-U-phosphoramidite and phosphoramidite monomers modified with 2′-OMe or 2′-F. These modified phosphoramidite monomers (YK-NUM-101, YK-NUM-102, YK-NUM-103, YK-NUM-104, and YK-NUM-105) were also used as starting materials for oligonucleotides synthesis. GalNAc L96 was conjugated to the 3′ end of the SS during solid-phase synthesis for in vivo experiments. Upon completion of synthesis, oligonucleotides were treated with 0.5 M anhydrous piperidine in acetonitrile for 10 min, then washed thoroughly with anhydrous acetonitrile, and dried using nitrogen. Oligonucleotides were cleaved from solid support and deprotected using ammonium hydroxide at 30°C for 17 h, followed by filtration through a nylon syringe filter (0.45 μm). Oligonucleotides were purified using IEXHPLC using an appropriate gradient of mobile phase (0.15 M NaCl, 10% MeCN and 1.0 M NaBr, 10% MeCN) and desalted using size-exclusion chromatography with water as an eluent. Oligonucleotides were then quantified by measuring the absorbance at 260 nm using the following extinction coefficients: A, 13.86; T/U, 7.92; C, 6.57; and G,10.53. The purified sense strand and antisense strand were mixed in a 1:1 molar ratio and then heated to 95°C and kept for 3 min, then slowly cooled down to room temperature to obtain hybridization.

Determination of thermal denaturation temperatures of hybridization

Thermal denaturation temperatures were measured with equimolar concentrations of RNA duplexes (2.0 μM) in PBS (137 mM NaCl, 2.7 mM KCl, 8 mM Na₂HPO₄, 8 mM KH₂PO₄, pH 7.4) by monitoring absorbance at 260 nm with increasing temperature (1°C/min). Values were reported as the maximum of the first derivative and are the average of at least three experiments.

Circular dichroism of modified siRNAs

The circular dichroism (CD) spectra were obtained on a Jasco J-815 spectropolarimeter equipped with a Julaba F25 circulating bath. The sample was allowed to equilibrate at 15°C in PBS at a final concentration of 2.0 μM duplex. The spectrum was an average from three independent experiments. Spectra were collected at a rate of 500 nm/min, with a bandwidth of 1 nm and sampling wavelength of 0.2 nm using fused quartz cells (Starna 29-Q-10). The CD spectra were recorded from 350 to 200 nm at 15°C. The molar ellipticity was calculated from the equation [θ] = θ/(10·c·l), where θ is the ellipticity (mdeg), c is the molar concentration of oligonucleotides (mM), and l is the path length of the cell (cm). The data were processed on a PC computer using Windows-based software supplied by the manufacturer (JASCO, Inc.) and transferred into Microsoft Excel for presentation.

Cell culture

Hep3B cell lines were maintained in Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific) supplemented with 10% (vol/vol) dialyzed fetal bovine serum (FBS,ExCell Bio-FSP500), 100 U/mL penicillin, and 100 mg/mL streptomycin in a humidified atmosphere at 37°C with 5% CO₂.

Care and use of laboratory animals

All studies were conducted using protocols consistent with applicable local regulations and were approved by the Institutional Animal Care and Use Committee (IACUC, approval number: HLZG-DWLL-2025-1025-06). Samples were diluted to the appropriate concentrations using PBS buffer and administered via subcutaneous injection to C57BL/6 mice (6–8 weeks old) or Sprague-Dawley rats (6–8 weeks old).

mRNA quantification in vitro

Hep3B cells were transfected with oligonucleotide using RNAiMAX reagent (Invitrogen) according to manufacturer′s recommendations. Briefly, cells were thawed just prior to transfection and plated onto 384-well plate with a seed density of ∼5,000 cells/well in Williams Medium E supplemented with 10% fetal bovine serum. Pre-incubated lipid/siRNA complex (0.1 μL RNAiMax, siRNA, in 5 μL Opti-MEM for 15 min; both reagents from Thermo Fisher Scientific) was added to a 384-well collagen-coated plate (BioCoat; Corning). Cells were incubated for 20 h at 37°C in an atmosphere of 5% CO₂. Media was then removed, and the cells were washed and lysed. RNA was extracted, using Dynabeads mRNA isolation kit (Invitrogen) according to manufacturer′s protocol, then reverse transcribed using ABI high capacity cDNA reverse transcription kit. Quantification was accomplished by real-time quantitative PCR, where the cDNA (2 μL) was added to a master mix containing 5 μL SYBR green gel dye, 0.8 μL Primer F/R mix, and 2.2 μL ddH₂O. Amplification was done in an ABI 7900HT-RT-PCR system (Applied Biosystems) using the ΔΔCt (RQ) assay (Table S3; Figures S27 and S28). Each data point was tested with at least three biological replicates. Each well was normalized to GAPDH control, and the mRNA remaining was calculated relative to cells treated with a non-targeting siRNA. IC₅₀ values were calculated from fitted curves using GraphPad Prism.

On- and off-target effects of modified siRNAs

PCSK9 was chosen as on-target mRNA, and hDSE, PCYOX1, and TRIM65 were chosen as off-target mRNA. Following the procedure described for mRNA quantification in vitro, Hep3B cells were transfected with pre-incubated lipid/siRNA complex (0.25 μL RNAiMax [Invitrogen] and 1 μL siRNA in 3.75 μL Opti-MEM for 15 min). Differential gene expression was measured by RT-PCR system.

RNA-seq

Hep 3B cells were thawed in a water bath at 37°C, resuspended in recovery medium, and centrifuged at 50 ×g for 2 min. After removal of the supernatant, cells were gently resuspended in seeding medium and seeded onto collagen-coated 24-well plates for 24 h. siRNA was complexed with transfection reagent, diluted to 50 nM in maintenance medium, and used to replace the seeding medium; incubation was continued for 48 h. Total RNA was extracted with TRIzol (Invitrogen), and cDNA libraries were prepared using the Hieff mRNA Library Prep Kit (Yeasen) following the manufacturer’s instructions and sequenced on a HiSeq sequencer (Illumina). After quality filtering, differentially expressed genes were identified with DESeq2.

Analysis of in vivo activity of modified siRNAs

The male hPCSK9 mice were dosed with 3 mg/kg GalNAc-siRNAs via subcutaneous injection at a dose volume of 5 mL/kg (n = 6 animals per group). The blood samples were collected 2 days before dosing and every 7 days after dosing. Blood samples were centrifuged at 3,000 rpm at 4°C to separate plasma, and hPCSK9 levels in serum were determined using ELISA (Proteintech). Serum LDL-C and TC levels were analyzed using an AU5800 chemistry analyzer (Beckman Coulter), with reagents provided by Beckman Coulter.

Toxicity evaluation

Whole-venous blood was collected into serum separator tubes (BD Microtainer). Serum ALT, AST, and GLDH levels were analyzed using an AU5800 chemistry analyzer (Beckman Coulter), with reagents provided by Beckman Coulter. After sacrificed, the livers were collected for histopathological examination.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

H.Z., Y.L., and J.M. contributed equally to this work. We thank Dr. Shuo Yang for helpful advice and discussion.

Author contributions

G.S. provided the project conceptualization, design, administration, and supervision; H.Z. performed conceptualization, investigation, and writing—review and editing; Y.L. performed data curation, formal analysis, software, and writing—original draft; J.M. performed the formal analysis and writing—original draft; H.H., W.Z., and D.X. performed methodology; T.M., M.Z., and F.Y. performed validation.

Declaration of interests

All authors were employed by Beijing Youcare Kechuang Pharmaceutical Technology Co., Ltd. All authors declare that the research was conducted without any commercial or financial relationships that could be viewed as a potential conflict of interest.