A novel bispecific siRNA concept: Efficient dual knockdown of YAP1 and WWTR1 with a single guide strand

Abstract

Yes-associated protein (YAP, encoded by YAP1) and transcriptional coactivator with PDZ-binding motif (TAZ, encoded by WWTR1) are master transcriptional regulators in the Hippo pathway. Because of their complementary roles, simultaneously inhibiting these proteins can efficiently attenuate downstream signaling. Their aberrant activation contributes to serious diseases, including cancer and fibrosis, making them key therapeutic targets. However, no approved therapeutic agents directly target YAP/TAZ because of technical challenges in designing small-molecule inhibitors for transcription regulators. Here, we present a novel approach using a bispecific small interfering RNA (siRNA) that can simultaneously suppress both YAP1 and WWTR1 with a single guide strand by targeting the consensus sequence of the two genes. Our lead bispecific siRNA, bsYW-61, can potently knock down YAP1/WWTR1 across human, cynomolgus monkey, mouse, and rat cell lines. Dual knockdown of YAP1/WWTR1 synergistically suppressed downstream gene expression patterns and cancer cell proliferation. Favorable target specificity and off-target profile were confirmed by luciferase reporter assays and RNA sequencing. Additionally, a chemically modified bispecific siRNA showed effective knockdown in vivo. Our data demonstrate the potential for using a bispecific siRNA targeting YAP1/WWTR1 as a therapeutic agent for liver diseases, presenting a novel approach for dual targeting with a simple conventional siRNA structure.

Graphical abstract

Bispecific siRNA technology enables simultaneous knockdown of YAP1 and WWTR1 by targeting their consensus sequence. The lead candidate efficiently suppressed downstream signaling and cancer cell proliferation across species, with favorable specificity and in vivo efficacy. This work highlights a novel strategy for dual targeting based on a simple siRNA architecture.

Introduction

Paralogs are ancestrally duplicated gene pairs with overlapping functions, which constitute approximately two-thirds of the human genome.^1,2 Because these genes can functionally compensate for each other, the loss of an individual paralog is often well tolerated. Paralogs frequently exhibit synthetic lethality, with the simultaneous impairment of both genes leading to cell death, while the impairment of either gene alone does not affect cell viability. Therefore, dual inhibition of paralogs may be an effective approach to overcoming the limited number of therapeutic targets in traditional drug discovery.^3,4 However, designing inhibitors with sufficient potency and specificity remains a challenge.

Small interfering RNAs (siRNAs) are short, double-stranded non-coding RNAs, offering highly specific and potent suppression of gene expression through RNA interference (RNAi).^5,6 After cellular internalization, the guide (antisense) strand of the siRNA associates with the Argonaute 2 protein to form the RNA-induced silencing complex (RISC), which degrades target mRNA in a sequence-specific manner.⁷ Additionally, the in vivo stability and delivery of an siRNA are determined by its scaffold, such as sequence chemical modification patterns, ligand conjugation, or encapsulation by lipid nanoparticles (LNPs).^8,9 After the scaffold for efficient and specific delivery has been validated in tissues, any gene can theoretically be targeted by altering the siRNA sequence. This high programmability enhances the feasibility of developing new therapeutic intervention methods. Furthermore, the long-acting nature of siRNAs is another significant advantage. For example, inclisiran, an siRNA drug for lowering low-density lipoprotein cholesterol (LDL-C), only requires a biannual dosing regimen.¹⁰ This greatly enhances patient compliance compared with monoclonal antibodies, which can also reduce LDL-C but necessitate administration every 2–4 weeks. Successful siRNAs in clinical applications have targeted genetically defined diseases with a single, well-validated pathway to target. A prime example is patisiran, the first approved siRNA drug, which targets transthyretin (TTR) to treat hereditary TTR-mediated amyloidosis. By reducing the production of unstable TTR proteins, patisiran prevents the formation of harmful amyloid deposits in organs.¹¹ Additionally, for diseases with more complex etiologies,^12,13 methods have been developed for the simultaneous knockdown of multiple targets. One practical approach is to administer a mixture of two siRNAs for different targets,¹⁴ which has already progressed to clinical trials. For example, CBP-4888, which was developed for treating pre-eclampsia, contains two siRNAs that target different mRNA isoforms of soluble fms-like tyrosine kinase-1, a key factor in pre-eclampsia pathogenesis.¹⁵ It is currently being evaluated in a phase 1 clinical trial (NCT05881993). Furthermore, various designs of unimolecular dual-targeting siRNAs have emerged, connecting siRNAs through a nucleotide linker or non-nucleotide linker via chemical crosslinking.^16,17,18 These unimolecular strategies are highly desirable for streamlining the clinical development path compared with mixing separate siRNAs. However, limitations related to chemical synthesis and delivery to targeted tissues may still exist relative to conventional siRNAs. The manufacturing complexity and cost can rise because of the increased nucleotide length with the nucleotide linkers or additional synthetic steps required for incorporating non-nucleotide chemical linkers. The increased molecular size or negatively charged nucleic acid can impact the efficiency of delivery to targeted tissues and endosomal escape.^16,19 Thus, additional optimization of the siRNA scaffold may be needed, as the delivery methods established for conventional single siRNAs may not be directly applicable, potentially requiring further efforts. Here, we report the development of a novel bispecific siRNA designed to simultaneously knock down two genes of a paralog pair with a conventional siRNA structure by targeting the consensus sequence of both genes. This novel approach for dual targeting using a typical single siRNA can potentially simplify the manufacturing and development processes, ensuring consistent pharmacokinetics for two target genes. The programmability of this siRNA design could enhance the ability to target paralog genes and broaden their use as therapeutic targets.

Yes-associated protein (YAP, encoded by Yes1-associated transcriptional regulator [YAP1]) and transcriptional coactivator with PDZ-binding motif (TAZ, encoded by WW domain containing transcription regulator 1 [WWTR1]) are key paralog transcriptional regulators of the Hippo pathway that critically regulate organ size, tissue homeostasis, and disease progression.^20,21 Because of the lack of a DNA-binding domain, YAP and TAZ interact with transcription factors, such as TEA domain (TEAD) family members, SMAD family members, and RUNX family members, to modulate gene transcription.^22,23 YAP and TAZ orchestrate organ growth, as well as cell proliferation, renewal, and repair, through interactions with key signaling pathways, such as the Notch, Wnt/β-catenin, and transforming growth factor β/Smad pathways.²⁴ Although activation of these pathways is essential for tissue repair, uncontrolled or excessive activation can eventually lead to cancer, fibrosis, and other injuries.²⁵ YAP and TAZ have common and distinctive structural features, as they have about 40% amino acid conservation,²⁶ reflecting partially overlapping regulatory mechanisms. Thus, some reports have suggested that dual inhibition of YAP and TAZ can efficiently suppress their signaling.^27,28,29 For example, double knockout of YAP and TAZ was found to strongly inhibit tumor cell proliferation and significantly suppress Akt/NRas-driven hepatocarcinogenesis compared with each single knockout.³⁰ Despite the therapeutic need to target YAP and TAZ, direct pharmacological inhibition remains challenging because of their lack of enzymatic active sites, structural plasticity, and broad protein-protein interaction interfaces.^31,32 Consequently, therapeutic efforts have focused on the transcriptional cofactors that mediate YAP/TAZ activity. Among these, only TEAD family inhibitors have advanced to clinical trials.^31,33 VT3989, which targets TEAD palmitoylation, showed clinical efficacy against advanced malignant mesothelioma and solid tumors with NF2 mutations in a phase 1 trial (NCT04665206).³⁴ However, safety concerns have emerged for TEAD inhibitors, as exemplified by K-975, which induces reversible, but significant, renal toxicity and proteinuria.³⁵ These limitations underscore the need for alternative approaches. RNAi enables the direct suppression of YAP/TAZ by leveraging the sequence programmability of siRNAs, which may allow for effective inhibition of YAP/TAZ signaling. Furthermore, advances in delivery technologies facilitate tissue-specific siRNA administration, potentially minimizing off-target toxicity. Several studies have explored siRNA approaches against YAP/TAZ, for example, reducing liver fibrosis and modulating inflammation,^36,37 typically by targeting either YAP1 or WWTR1 alone or combining separate siRNAs.

In this study, bispecific siRNAs were engineered to simultaneously target the YAP1 and WWTR1 genes, both of which play pivotal roles in pathologies, such as fibrosis and cancer. To maximize the translational potential, we systematically screened the siRNA sequences capable of dual gene suppression across multiple species. Additionally, the significance of YAP1/WWTR1 dual suppression was validated in vitro, with the specificity for target genes investigated through both in silico and in vitro approaches. To further enhance the therapeutic potential, we introduced chemical modifications and a targeting moiety to the bispecific siRNA, subsequently evaluating the suppression of Yap1 and Wwtr1 expression patterns in mouse liver tissues. This study provides a novel approach for nucleic acid-based dual targeting and introduces siRNA-mediated targeting of YAP1/WWTR1 as a potential therapeutic method for liver diseases.

Results

Design of bispecific siRNAs targeting YAP1 and WWTR1 with a single guide strand

Using our strategy, a bispecific siRNA was designed to target the consensus sequences of two gene transcripts, thereby knocking down the expression of both genes using a single guide strand (Figure 1A). Following the workflow outlined in Figure 1B, we first performed bioinformatics analysis to design a novel bispecific siRNA targeting both YAP1 and WWTR1. This siRNA consists of a conventional 21-nucleotide (nt) duplex with 2-nt 3′ overhangs on each strand. As indicated in previous reports, the 5′ ends of the guide and passenger (sense) strands of effective siRNAs should be adenine (A)/uracil (U) and guanine (G)/cytosine (C), respectively.^38,39,40 This terminal base asymmetry is important because an unstable 5′ end is preferable for RISC formation.^41,42 We thus initially designed the central 17-nt sequences to be complementary to all YAP1 and WWTR1 transcript variants and then added A/U and G/C to the 5′ end of the guide and passenger strands, respectively, with the counter strands being G/C and A/U. Despite YAP1/WWTR1 being a paralog pair with approximately 40% homology at the amino acid level,²⁶ continuous 17-nt matches were not frequently found. Therefore, in this study, the “consensus sequence” is defined as a stretch of 17 consecutive nt in which at least 13 bases are identical across all YAP1 and WWTR1 transcript variants, thereby allowing up to four mismatches. After selecting 17-nt consensus sequences, we refined them as passenger strands, ensuring no more than two mismatches for both YAP1 and WWTR1. Subsequently, the 17-nt passenger strand sequences were extended to 19 nt by adding one base at each end according to predefined selection rules. For the 5′ end, G was added when at least one of the corresponding YAP1/WWTR1 nucleotide was G and none was C; in all other cases, C was used. For the 3′ end, U was added when at least one of the corresponding nucleotide was U and none was A; otherwise, A was used. Finally, fully complementary double strands were generated with 2-nt thymine (T) overhangs at the 3′ ends of both strands. Our bispecific siRNA design retained the typical siRNA architecture, consisting of fully complementary 21-nt double strands with 2-nt 3′ overhangs on each strand.

Figure 1.

Design of bispecific siRNAs targeting YAP1 and WWTR1 with a single guide strand

(A) Schematic diagram of a novel bispecific siRNA design in which each single guide strand can bind to two target genes. The mRNA sequences colored in purple represent the consensus sequence shared by the two target genes. (B) Summary of the workflow for designing the bispecific siRNA sequences targeting Yes1-associated transcriptional regulator (YAP1) and WW domain containing transcription regulator 1 (WWTR1). The mRNA sequences colored in purple indicate the identical sequence shared by YAP1 and WWTR1. Red highlights the sequence unique to YAP1, while blue denotes the sequence unique to WWTR1. Illustrations in (A) were created using BioRender.com.

In vitro screening of bispecific siRNAs exhibiting efficient dual knockdown

Using our proprietary siRNA design algorithm, we generated bispecific siRNA sequences that are specific to the coding sequence or 3′ untranslated region (UTR) of all YAP1 and WWTR1 mRNA transcripts. These included sequences targeting only humans and those cross-targeting humans and one or more species among cynomolgus monkeys, mice, or rats to streamline the path toward clinical development. A total of 76 siRNA candidates across 36 sites were synthesized for in vitro screening using native RNA (2′-OH) with DNA overhangs (Figure 2A). All synthesized sequences are listed in Table S1. Primary screening was performed in human HepG2 cells, which identified 28 siRNAs across 20 sites that resulted in a ≥50% reduction of both YAP1 and WWTR1 mRNA expression levels relative to the mock-transfected control (Figure 2B). In contrast, the negative control siRNA (siNC) showed no knockdown effects. For siRNAs perfectly matched to both genes, the expression levels of the two genes strongly correlated (R = 0.91), whereas the correlation was moderate when considering all siRNAs (R = 0.50) (Figure 2C). Additionally, siRNAs with mismatches still exhibited clear knockdown. To investigate the position-dependent impact of mismatches on siRNA potency, we compared the knockdown activity of bispecific siRNAs containing a single or two mismatches in the seed region (2–8 nt from the 5′ end) and/or the 3′ supplementary region (12–18 nt from the 5′ end). The results showed no significant difference in activity depending on the mismatch position (Figure S1). The siRNA sites with high activity for both genes were distributed across two regions of the YAP1 and WWTR1 mRNAs, specifically at positions 1538–1546/1142–1150 and 1657–1667/1261–1271, respectively, when counted based on transcript variant 1 (YAP1 transcript variant 1: NM_001130145.3, WWTR1 transcript variant 1: NM_015472.6). For the 28-siRNA panel, which demonstrated 50% knockdown efficiency for both target genes, we investigated the complementarity to off-target genes. From our in silico analysis results, 17 siRNAs with a 17-nt targeting region that recognized off-target genes with zero or one mismatch were removed (Table S2). Next, we considered the knockdown activity in mice. A total of three siRNAs (bsYW-29, 30, and 31) with more than three mismatches for mouse Yap1 or Wwtr1 were excluded from the candidate list. The remaining eight siRNAs (bsYW-32, 33, 55, 59, 61, 63, 64, and 66) were subsequently evaluated in human HepG2 and mouse Hepa1-6 cells at three different concentrations. In HepG2 cells, bsYW-61 reduced YAP1 and WWTR1 expression levels by ≥50% even at 0.01 nM (Figure 2D). In Hepa1-6 cells, only bsYW-60, 61, and 62 achieved a ≥50% reduction in Yap1 and Wwtr1 expression levels (Figure 2E). Among these, bsYW-61 was the most potent at all concentrations, leading us to select it as the lead siRNA sequence. We further evaluated 61 mouse-specific bispecific siRNAs to assess the cross-species constraint, with bsYW-61 still demonstrating the most potent activity among them (Figure S2A; Table S3).

Figure 2.

In vitro screening of bispecific siRNAs exhibiting efficient dual knockdown

(A) Schematic representation of the bispecific siRNA. All siRNAs consist of native RNA with 3′ DNA overhangs. (B) The mean Yes1-associated transcriptional regulator (YAP1) and WW domain containing transcription regulator 1 (WWTR1) mRNA expression levels relative to the mock-transfected human HepG2 cells. Twenty-four hours post reverse transfection with 1 nM siRNA, negative control siRNA (siNC), or mock transfection, the mRNA expression levels were quantified by quantitative reverse-transcription polymerase chain reaction (RT-qPCR) (n = 3, mean ± standard deviation [SD]). The dotted line below indicates siRNAs with ≥50% relative repression. The numbers below represent the targeted nucleotide position on variant 1 of the human YAP1 and WWTR1 mRNA transcripts. The diagram above represents the target region in the YAP1 and WWTR1 mRNAs per siRNA. The colored area represents the hotspots where highly active siRNAs were obtained. (C) Correlations between the suppression of YAP1 and WWTR1 expression levels using 76 bispecific siRNAs. The dot types indicate the number of mismatches against YAP1 and WWTR1 in the 17-nt targeting region of the bispecific siRNAs (n = 3, mean). (D and E) The lead siRNA candidates were evaluated at three different concentrations in (D) human HepG2 cells and (E) mouse Hepa1-6 cells (n = 3, mean ± SD). The dotted line below indicates siRNAs with ≥50% relative repression.

Cross-species dual targeting by bsYW-61

The target mRNA sequences of bsYW-61 are well conserved across humans, non-human primates (NHPs), mice, and rats (Figure 3A). bsYW-61 does not have a mismatch to the human YAP1/WWTR1 and cynomolgus monkey YAP1 sequences. There is one mismatch to cynomolgus monkey WWTR1, mouse Yap1/Wwtr1, and rat Wwtr1, and two mismatches to rat Yap1. We thus determined the half-maximal inhibitory concentration (IC₅₀) values in human HepG2, cynomolgus monkey MK.P3(F), mouse Hepa1-6, and rat H-4-II-E cells. bsYW-61 exhibited robust knockdown activity against YAP1/WWTR1 across all four species, despite the presence of mismatches (Figure 3B). Notably, even with two mismatches against rat Yap1, bsYW-61 still demonstrated clear knockdown activity (IC₅₀ = 382 pM).

Figure 3.

Cross-species dual targeting by bsYW-61

(A) bsYW-61 targets Yes1-associated transcriptional regulator (YAP1) and WW domain containing transcription regulator 1 (WWTR1) in humans, cynomolgus monkeys, mice, and rats with up to two mismatches. Gray areas represent the targeted regions, and red letters represent the mismatched bases aligned to each sequence. (B) In vitro dose-response curves for bsYW-61 in human HepG2 cells, cynomolgus monkey MK.P3(F) cells, mouse Hepa1-6 cells, and rat H-4-II-E cells. Twenty-four hours post reverse transfection with the small interfering RNA (siRNA) or control, the mRNA expression levels were quantified by quantitative reverse-transcription polymerase chain reaction (RT-qPCR). The data are shown as a percentage of the mock control values; highest dose: 10 or 1 nM, 10-fold serial dilutions (n = 3, mean ± standard deviation from three experiments).

Synergistic inhibition of YAP/TAZ downstream gene expression and cell proliferation

To verify the necessity of inhibiting both YAP1 and WWTR1, we compared the expression changes of genes downstream of YAP/TAZ, including cellular communication network factor 1 (CCN1), cellular communication network factor 2 (CCN2), and ankyrin repeat domain 1 (ANKRD1), with the individual inhibition of YAP1 and WWTR1. Human HepG2 cells were transfected with bsYW-61, single-targeting siRNA for human YAP1 (siYAP1), single-targeting siRNA for human WWTR1 (siWWTR1), a combination of siYAP1 and siWWTR1 (siYAP1 + siWWTR1), or siNC. Notably, the single-targeting siRNAs used here exhibited robust knockdown activity at similar levels to or exceeding those of 10 additionally designed fully complementary siRNAs and two commercially available siRNAs for each target gene (Figure S2; Tables S4 and S5). bsYW-61, siYAP1, siWWTR1, and siYAP1 + siWWTR1 demonstrated similar knockdown levels of their respective target genes (approximately 20% YAP1 remaining and 30% WWTR1 remaining) compared with the mock-transfected control (Figure 4A). In contrast, downregulation of CCN1, CCN2, and ANKRD1, which are YAP1/WWTR1 downstream genes, was observed only when both YAP1 and WWTR1 were simultaneously knocked down by bsYW-61 or siYAP1 + siWWTR1. Consistent with these findings, dual targeting of Yap1 and Wwtr1 in mouse Hepa1-6 cells also effectively reduced YAP/TAZ downstream gene expression patterns (Figure S3A). We next evaluated the impact on cell proliferation. In human HepG2 cells, proliferation was assessed 7 days after transfection using a cell viability assay and cell confluence analysis. The data indicated that cell proliferation rates were significantly reduced by bsYW-61 and siYAP1 + siWWTR1 compared with siYAP1 or siWWTR1 (Figures 4B and 4C). A similar trend was observed in mouse Hepa1-6 cells (Figure S3B). Collectively, these results demonstrate that simultaneous suppression of YAP1 and WWTR1 is essential for efficient inhibition of YAP/TAZ signaling, providing an effective strategy for suppressing cancer cell proliferation.

Figure 4.

Synergistic inhibition of YAP/TAZ downstream gene expression and cell proliferation

(A) Expression levels of YAP/TAZ downstream genes in human HepG2 cells following knockdown of Yes1-associated transcriptional regulator (YAP1) and WW domain containing transcription regulator 1 (WWTR1) using 1 nM small interfering RNA (siRNA). Twenty-four hours post reverse transfection, the mRNA expression levels were quantified by quantitative reverse-transcription polymerase chain reaction (RT-qPCR). (B and C) HepG2 cell proliferation following YAP1 and WWTR1 knockdown. (B) Cell viability was quantified using the CellTiter-Glo 2.0 Assay, and (C) cell confluence was monitored using the Incucyte system, 7 days after reverse transfection with 10 nM siRNA or control. (A–C) The data are shown as n = 3, mean ± standard deviation, one-way ANOVA with Tukey’s test; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

In vitro safety evaluation of Yap1 and WWTR1 knockdown

We then evaluated the safety profile of inhibiting YAP1 and WWTR1 by bsYW-61 to assess its potential as a therapeutic agent. Cellular caspase activity, a well-established indicator of apoptosis that reportedly correlates with in vivo hepatotoxicity, was measured to assess apoptosis induction.⁴³ We selected two antisense oligonucleotides (ASOs) as controls to compare with bsYW-61: PTEN-ASO, known to be hepatotoxic,⁴⁴ and APOB-ASO (commercialized as mipomersen), known to be tolerated.⁴⁵ The two control ASOs and bsYW-61 were transfected into human HepG2 cells, and then caspase activity was assessed after 24 h. Hepatotoxic PTEN-ASO treatment resulted in a dramatic increase in caspase activity, exceeding 1,000% at 30 nM compared with the untreated cells. Mipomersen also caused increased caspase activity, reaching approximately 250% at 300 nM. In contrast, bsYW-61 did not induce any significant increase in caspase activity, even at 300 nM (Figure 5A).

Figure 5.

In vitro safety evaluation of YAP1 and WWTR1 knockdown

(A) Human HepG2 cells were transfected with PTEN-ASO, mipomersen (APOB-ASO), or bsYW-61. Caspase-3/7 activity was measured 24 h post transfection and normalized to the activity in the untreated control cells (n = 2, mean). (B) The oxygen consumption rate (OCR) in HepG2 cells was measured using a Flux Analyzer XF96e to evaluate mitochondrial toxicity. Steady-state oxygen consumption was measured 27.5 h post transfection with bsYW-61. Oligomycin was injected to inhibit ATP synthase, and carbonyl cyanide-4-[trifluoromethoxy] phenylhydrazone (FCCP) was injected to uncouple the mitochondria and determine the maximal OCR. Final OCR values were normalized to the mock-transfected control (n = 3–12, mean ± standard deviation [SD]). (C and D) HepaSH cells (human hepatocytes) were transfected with 10 nM of siYAP1 + siWWTR1 or bsYW-61. (C) Cell viability was measured 72 h after transfection using the CellTiter-Glo 2.0 Max assay. (D) Dead cells were detected using the Cytotox Green Dye and represented as relative fluorescence intensity compared with the mock samples (n = 3, mean ± SD, one-way ANOVA with Dunnett’s test; ns, not significant [p ≥ 0.05]).

YAP and TAZ are known to influence mitochondrial fusion and fission, which are critical for maintaining mitochondrial function.⁴⁶ They also regulate metabolic pathways, such as glutaminolysis, supporting the tricarboxylic acid cycle and oxidative phosphorylation.⁴⁷ To evaluate the impact of bsYW-61-mediated inhibition of YAP/TAZ signaling on mitochondrial activity and bioenergetics, we measured the cellular oxygen consumption rate (OCR) using the Flux Analyzer XF96e. Human HepG2 cells were transfected with bsYW-61, and then the OCR in steady state was measured 27.5 h after transfection. Subsequently, oligomycin was injected to inhibit ATP synthase, followed by carbonyl cyanide-4-[trifluoromethoxy] phenylhydrazone (FCCP) to uncouple mitochondria and measure the maximal OCR. Our results showed that bsYW-61 did not decrease the steady-state OCR nor the maximal OCR, even at the highest concentration of 300 nM (Figure 5B).

Next, we investigated the cytotoxicity to normal cells using HepaSH cells, which are human hepatocytes isolated from chimeric TK-NOG mice.⁴⁸ Working with these cells reduces the interindividual variability seen in primary human hepatocytes and improves reproducibility. Cell viability was measured 72 h after transfection with siYAP1 + siWWTR1 or bsYW-61. Compared with the mock-transfected controls, no significant reduction in cell viability was observed with either siYAP1 + siWWTR1 or bsYW-61 transfection (Figure 5C), while the expression levels of YAP1/WWTR1 and their downstream genes were clearly suppressed (Figure S4). Additionally, we quantified cell death by detecting cell membrane integrity disruption using the Incucyte Cytotox dye. Neither siYAP1 + siWWTR1 nor bsYW-61 increased the fluorescence intensity compared with the mock-transfected controls (Figure 5D). Taken together, bsYW-61 did not affect the viability of human hepatocyte HepaSH cells.

Characterization of off-target activities of bsYW-61

Recent advances in RNAi therapeutics have established chemical modifications as essential for enhancing siRNA stability while minimizing off-target activity and immune activation.^10,49 The unmodified siRNA bsYW-61 was therefore redesigned as bsYW-61m1, incorporating 2′-O-methyl (2′-OMe)/2′-deoxy-2′-fluoro (2′-F) ribosugars, phosphorothioate (PS) linkages, and a 5′ phosphate (P) (Figure 6A). The IC₅₀ values were determined in human HepG2, cynomolgus monkey MK.P3(F), mouse Hepa1-6, and rat H-4-II-E cells. The chemically modified bsYW-61m1 exhibited knockdown activity against YAP1/WWTR1 across all four species, including against rat Yap1 with two mismatches (IC₅₀ = 4,671 pM; Figure 6B). Moreover, eight additional candidates with chemical modifications were evaluated in both HepG2 and Hepa1-6 cells to confirm that the relative activity ranking was preserved after being modified (Figure S5). Considering the activity across both species, bsYW-61 consistently exhibited the most favorable knockdown levels.

Figure 6.

Characterization of off-target activities of bsYW-61

(A) Chemical modification patterns of bsYW-61m1. (B) In vitro dose-response curves for bsYW-61m1 in human HepG2 cells, cynomolgus monkey MK.P3(F) cells, mouse Hepa1-6 cells, and rat H-4-II-E cells. (C) Transcriptional regulation by RNA sequencing (RNA-seq) analysis in human HepG2 cells 24 h after transfection with 10 nM bsYW-61, relative to the mock control (n = 3, mean). The volcano plot depicts log₂-fold change (FC) and p values: gray = not significant (n.s.); green = |FC| ≥ 2; blue = p < 0.05; red = |FC| ≥ 2 and p < 0.05. (D) Off-target genes with a half-maximal inhibitory concentration (IC₅₀) value within 100-fold of those for on-target genes. (E) Number of in vitro probable off-target candidates, in silico-predicted genes with complementarity to the 17-nt targeting region (including up to two mismatches), and in silico-predicted 3′ UTR seed-matched genes. (F) Design of the reporter plasmids to evaluate the full-length hybridization-dependent RNA interference (RNAi) on- and off-target activities mediated by either the guide strand (GS-CM) or passenger strand (PS-CM). (G) RNAi on-target activities of bsYW-61 and bsYW-61m1, and full-length hybridization-dependent off-target activity of the bsYW-61 passenger strand. (H) Design of the reporter plasmids to evaluate the seed-mediated off-target activity of the guide strand (GS-SM4) or passenger strand (PS-SM4). (I) Guide strand seed-mediated off-target activity of bsYW-61 and bsYW-61m1, and passenger strand seed-mediated off-target activity of bsYW-61. (G and I) Luciferase activity was measured 24 h after co-transfection of human HeLa cells with the siRNA and plasmids. Relative luciferase (luc) activity was calculated as the Renilla/firefly ratio relative to siNC (n = 3, mean ± standard deviation).

To investigate off-target activities, RNA sequencing (RNA-seq) was performed in human HepG2 cells treated with 10 nM bsYW-61, which is approximately 1,000 times higher than the IC₅₀ of YAP1 and WWTR1. Total RNA was isolated 24 h after transfection, and then the global gene expression changes relative to the mock-transfected controls were visualized using a volcano plot (Figure 6C). Because YAP1 and WWTR1 regulate various transcription factors, their suppression can affect the expression patterns of numerous genes. Indeed, we identified 21 additional significantly differentially expressed genes (DEGs; |fold change| ≥ 2, p < 0.05) following 10 nM bsYW-61 transfection. To distinguish on- and off-target activities, the DEGs were validated by quantitative reverse-transcription polymerase chain reaction (RT-qPCR) after transfection with either 10 nM bsYW-61 or siYAP1 + siWWTR1. The genes suppressed by both treatments (AMOTL2, CHST9, and KRT23) were considered on-target effects of YAP/TAZ inhibition, with five genes showing no significant changes by RT-qPCR (Figure S6). Two genes (AMPD3 and HAS3) were not sufficiently expressed in HepG2 cells for analysis (data not shown). The remaining 12 genes were suppressed by bsYW-61 but not by siYAP1 + siWWTR1, identifying them as potential off-targets. Further analysis revealed that only four genes (PSME2, NAPG, TUBG1, and PLS3) had IC₅₀ values within 100-fold of YAP1 or WWTR1 (Figure S7). Importantly, no toxicity concerns have been reported for suppression of these genes in the liver. The mRNA sequences of PSME2 and NAPG are complementary to the 17-nt targeting region of bsYW-61 with two mismatches, whereas those of TUBG1 and PLS3 are not, even with up to four mismatches (Figure 6D). Comprehensive in silico analysis of the full-length hybridization-dependent off-target activities of the guide strand revealed that bsYW-61 has no perfect-match or single-mismatch off-targets, but 62 genes with two mismatches were identified (Figure 6E). Among these, only PSME2 and NAPG were downregulated in HepG2 cells. For seed-mediated (microRNA-like) off-target activities, 1,500 genes contained sequences matching the bsYW-61 seed region in their mRNA 3′ UTRs, yet none exhibited significant expression pattern changes in HepG2 cells.

We next assessed the off-target profile of the chemically modified bsYW-61m1. In HepG2 cells, 10 potential off-target genes were identified through RNA-seq and RT-qPCR analyses. Further analysis of the IC₅₀ values revealed that the four candidate genes identified with unmodified bsYW-61 (PSME2, NAPG, TUBG1, and PLS3) were consistently suppressed by bsYW-61m1. Moreover, an additional gene, LAMC1, was newly affected, exhibiting an IC₅₀ value within a 100-fold range of those for YAP1 or WWTR1 (Figure S8). To distinguish sequence-dependent effects from modification-dependent effects, two other bispecific siRNAs (bsYW-32 and bsYW-59) and their modified counterparts (bsYW-32m1 and bsYW-59m1) were tested against the five candidate genes. Here, RT-qPCR analysis showed inconsistent expression changes across sequences or modification states, suggesting that the off-target activities are likely influenced by multiple factors beyond a single determinant of sequence or modification (Figure S9).

The specificity of bsYW-61 was evaluated by luciferase reporter assays using psiCHECK-2 vectors. The full-length hybridization-dependent on- and off-target RNAi activities were measured using reporter constructs containing a complete-match sequence in the 3′ UTR for either the guide strand (GS-CM) or passenger strand (PS-CM; Figure 6F). Both bsYW-61 and modified bsYW-61m1 efficiently suppressed guide strand-dependent luciferase activity by over 90% in a dose-dependent manner (Figure 6G). In contrast, no reduction was observed for the passenger strand, indicating strong strand selectivity and minimal passenger strand-mediated off-target RNAi activity. Seed-mediated off-target activities were then assessed using reporter constructs containing four tandem repeats complementary to the seed region of the guide strand (GS-SM4) or passenger strand (PS-SM4) in the 3′ UTR (Figure 6H). bsYW-61 reduced luciferase activity in the guide strand seed reporter, confirming seed-mediated off-target effects (Figure 6I). In contrast, modified bsYW-61m1 showed no such activity, suggesting that the chemical modification effectively abolished these interactions. The passenger strand showed no seed-mediated activity, confirming minimal contribution to off-target effects. These results indicate that bsYW-61 exhibits strong guide strand bias with potent on-target activity, with seed modification in bsYW-61m1 effectively minimizing any seed-mediated off-target activities.

In vivo dual targeting of Yap1 and Wwtr1

Next, we evaluated the in vivo activity of bsYW-61. To deliver siRNAs to the liver, we formulated them into LNPs containing the ionizable lipid ALC-0315⁵⁰ (Figure 7A). Mice received a single intravenous dose of siYap1/LNP (1 mg/kg), siWwtr1/LNP (1 mg/kg), siYap1 + siWwtr1/LNP (1 mg/kg each; total 2 mg/kg), or bsYW-61m1/LNP (1 mg/kg). The liver tissues were collected 2 days later for mRNA analysis. siYap1/LNP, siYap1 + siWwtr1/LNP, and bsYW-61m1/LNP achieved comparable Yap1 knockdown levels (approximately 80%–85%) (Figure 7B). Similarly, siWwtr1/LNP, siYap1 + siWwtr1/LNP, and bsYW-61m1/LNP showed equivalent Wwtr1 knockdown levels (approximately 85%–90%). No significant changes in the liver weight or serum aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels were observed in any group (Figures 7C and 7D). Notably, the single-targeting siRNAs used in this in vivo study exhibited knockdown activity comparable with or higher than that of 10 additionally designed candidates in mouse Hepa1-6 cells, all fully complementary to their intended targets (Figures S2F–S2I; Tables S4 and S5).

Figure 7.

In vivo knockdown of Yap1 and Wwtr1

(A) Schematic of lipid nanoparticle (LNP) formulation. (B) In vivo suppression of Yes1-associated transcriptional regulator (Yap1) and WW domain containing transcription regulator 1 (Wwtr1) 2 days after single intravenous (i.v.) administration of siRNA/LNPs (1 mg/kg for single siRNA or 2 mg/kg for siYap1 + siWwtr1 mixture). The expression levels are relative to the HEPES-treated controls. (C) Liver weight and (D) plasma aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels were examined. (E) Schematic of the chemical modification patterns of bsYW-61m2 and bsYW-61m3. GalNAc was conjugated to the passenger strand 3′ end. (F) In vivo suppression of Yap1 and Wwtr1 7 days after single 10 mg/kg subcutaneous (s.c.) administration of the GalNAc-siRNA conjugates. The expression levels are relative to the phosphate-buffered saline-treated controls. (G) Body weight and liver weight, as well as the (H) plasma AST and ALT levels, were examined. (B–D and F–H) The data are shown as n = 4, mean ± standard deviation, one-way ANOVA with Dunnett’s test; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant (p ≥ 0.05).

We next investigated conjugating N-acetylgalactosamine (GalNAc) to bsYW-61 to expand its application (Figure 7E). Two variants, bsYW-61m2 and bsYW-61m3, which contained additional PS linkages and differed in 5′ guide phosphate modifications, were first tested in HepG2 and Hepa1-6 cells. Both variants showed similar knockdown activity (Figures 7F and 7G). Mice were given a single subcutaneous dose of 10 mg/kg of either conjugate. Seven days later, Yap1/Wwtr1 knockdown levels were approximately 50%–60% with bsYW-61m2/GalNAc and approximately 70%–80% with bsYW-61m3/GalNAc compared with the phosphate-buffered saline (PBS) group (Figure 7H). These findings confirmed that the use of 5′-vinyl phosphonate (VP) could effectively achieve increased in vivo activity, largely from enhanced siRNA stability as described in previous studies.⁵¹ No significant change in body or liver weight was observed (Figure 7I), nor did the plasma AST or ALT levels increase (Figure 7J). These results show that bsYW-61 can achieve dual gene targeting in vivo when delivered by LNP or GalNAc conjugation, supporting further evaluation of its therapeutic potential.

Bispecific siRNAs achieving efficient dual knockdown of CREBBP and EP300

To evaluate the general applicability of our bispecific approach, we designed siRNAs targeting additional gene pairs. We selected CREB-binding protein (CREBBP) and E1A-binding protein p300 (EP300), histone acetyltransferases that function as transcriptional coactivators and are implicated in various diseases, including cancer.^3,52,53 Following the same workflow used for the YAP1/WWTR1 bispecific siRNAs, we employed our proprietary design algorithm to generate sequences specific to all CREBBP/EP300 transcripts. A total of 94 candidates across 69 sites were synthesized for in vitro screening (Figure 8A; Table S6). Screening in HepG2 cells identified 19 siRNAs across 17 sites that reduced both CREBBP and EP300 mRNA levels by ≥50% relative to the mock-transfected control. Among them, one potent sequence, bsCE-55, showed robust dual knockdown activity in HepG2 cells (Figure 8B). We next compared CREBBP, EP300, and MYC expression levels following transfection with bsCE-55, validated single-targeting siRNAs against CREBBP (siCREBBP) or EP300 (siEP300), or a combination of both (siCREBBP + siEP300) (Figure 8C; Figure S10; Table S5). Each siRNA efficiently reduced its respective target expression pattern to a similar extent, while MYC expression decreased only with dual knockdown by bsCE-55 or siCREBBP + siEP300 (Figure 8C). At the protein level, c-Myc expression levels were reduced by approximately 90% with dual knockdown, whereas individual knockdown by siEP300 or siCREBBP resulted in moderate reductions of approximately 50% and 40%, respectively (Figure 8D). Consistent with these molecular effects, dual knockdown suppressed HepG2 cell proliferation rates more effectively than single knockdown (Figures 8E and 8F). These results indicate that the unimolecular bispecific siRNA approach can be extended to additional gene pairs beyond YAP1/WWTR1.

Figure 8.

Bispecific siRNAs achieving efficient dual knockdown of CREBBP and EP300

(A) The mean of CREB-binding protein (CREBBP) and E1A-binding protein p300 (EP300) mRNA expression levels relative to the mock-transfected human HepG2 cells. Twenty-four hours post reverse transfection with 1 nM siRNA, negative control siRNA (siNC), or mock transfection, the mRNA levels were quantified by quantitative reverse-transcription polymerase chain reaction (RT-qPCR) (n = 3, mean ± standard deviation [SD]). The dotted line below indicates siRNAs with ≥50% relative repression. The numbers below represent the targeted nucleotide position on variant 1 of the human CREBBP and EP300 mRNA transcripts. (B) In vitro dose-response curves for bsCE-55 in human HepG2 cells; highest dose: 10 nM, 10-fold serial dilutions (n = 3, mean ± SD). (C) Expression levels of CREBBP, EP300, and MYC mRNA after reverse transfection of 10 nM siCREBBP, siEP300, siCREBBP + siEP300, or bsCE-55 in HepG2 cells. (D) c-Myc protein expression levels following CREBBP and EP300 knockdown in HepG2 cells, quantified by Jess 5 days post transfection (n = 1). (E and F) HepG2 cell proliferation rates following CREBBP and EP300 knockdown. (E) Cell viability was quantified using the CellTiter-Glo 2.0 assay, with (F) cell confluence monitored using the Incucyte system, 5 days after reverse transfection with 10 nM siRNA or control. (C and E–F) The data are shown as n = 3, mean ± SD, one-way ANOVA with Tukey’s test; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Discussion

In this study, we illustrate the concept of a novel bispecific siRNA that can simultaneously target YAP1 and WWTR1 through a single guide strand. Aberrations in the Hippo signaling pathway are associated with cancer progression and organ fibrosis, making YAP/TAZ attractive therapeutic targets.^20,21,25 Although TEAD inhibitors have advanced to clinical trials, direct inhibition of YAP/TAZ remains challenging because of structural constraints.^31,32,33 Previous dual targeting siRNAs rely on mixing two distinct siRNAs or linking them together.^17,18 In contrast, our approach achieves dual targeting with a single conventional siRNA structure (21-nt duplex with 2-nt 3′ overhangs), which streamlines manufacturing and optimization of delivery methods to targeted tissues, as well as ensures consistent pharmacokinetics for both targets. We demonstrated that dual knockdown of YAP1 and WWTR1 can synergistically suppress YAP/TAZ downstream gene expression and cancer cell proliferation. The lead compound, bsYW-61, showed minimal cellular toxicity and off-target activities, with in vivo dual targeting confirmed using both LNP and GalNAc ligand. Finally, we demonstrated that this strategy can also be applied to target CREBBP and EP300, highlighting its applicability.

Designing a bispecific siRNA is challenging because continuous 17-nt sequences perfectly matching both targets are rare, despite approximately 40% homology between YAP and TAZ at the amino acid level.²⁶ To accommodate both targets, we allowed up to two mismatches per gene. The guide strand 5′ and 3′ end bases were fixed as A/U and G/C, respectively. This adjustment potentially reduced the influence of strand selection bias during RISC incorporation, allowing us to focus on mRNA accessibility. Knockdown activity comparisons for bispecific siRNAs containing a single or two mismatches in the seed region and/or the 3′ supplementary region revealed no position-dependent impact on siRNA potency (Figure S1). These observations suggest that the knockdown efficiency is determined not only by mismatch position or complementarity but also by the secondary structure of YAP1 and WWTR1 mRNAs and their protein binding profiles. Additionally, we identified two hotspots that yielded potent siRNAs against both YAP1 and WWTR1. These regions likely reflect the particularly high homology between the mRNA sequences and relatively low complexity of their higher-order structures.⁵⁴ Although recent advancements have indicated the importance of chemical modifications,^8,55 we initially screened unmodified sequences to validate the dual-gene targeting concept under simplified conditions, avoiding sequence-dependent loss of activity caused by chemical modifications.⁵⁶ After the lead siRNA candidate was identified, chemical modifications were applied, with bsYW-61 retaining its superior activity. This demonstrated the robustness of our design strategy. Broader evaluation across additional sequences and targets will be needed to generalize this approach.

For clinical development, researchers evaluate the efficacy and safety of siRNA drugs in animal models that exhibit tissue characteristics and disease mechanisms like those in humans. Identifying siRNAs with cross-species reactivity has accelerated the development by enabling testing in various experimental models. Obtaining siRNAs with cross-species reactivity can be challenging because of significant differences in gene sequences across species. Moreover, even with high sequence complementarity, knockdown activity is not always guaranteed. For example, of the eight siRNAs in this study that showed high activity in human HepG2 cells (≥50% suppression at 1 nM) and complementarity to mouse transcripts, only two retained comparable activity in mouse Hepa1-6 cells. This inconsistency is likely caused by differences in mRNA distribution, structural conformation, and protein binding dynamics between species or cell lines.⁵⁷ In this context, bsYW-61, which can potently knock down YAP1 and WWTR1 expression in humans, mice, cynomolgus monkeys, and rats, is groundbreaking as a potential therapeutic candidate.

To investigate the significance of dual inhibition, we evaluated the expression patterns of CCN1, CCN2, and ANKRD1, representative transcriptional targets of YAP1/WWTR1. In both human HepG2 and mouse Hepa1-6 cells, targeting both YAP1 and WWTR1, either by a mix of validated single-targeting siRNAs or bsYW-61, resulted in a synergistic suppression of CCN1, CCN2, and ANKRD1 expression, despite equivalent knockdown of YAP1/WWTR1. This indicates that simultaneous inhibition of YAP1 and WWTR1, which function partially complementarily, can effectively block the Hippo pathway. Furthermore, dual inhibition markedly reduced cancer cell proliferation rates. Consistent with this, a previous study reported that the double knockout of YAP and TAZ strongly inhibits hepatocellular carcinoma cell proliferation and significantly suppresses Akt/NRas-driven hepatocarcinogenesis,³⁰ likely through coordinated regulation of the cell cycle and DNA replication. Combining these findings, bsYW-61 represents a reasonable strategy as a therapeutic agent for diseases such as liver cancer.

Previous studies have reported that YAP/TAZ knockout in the liver leads to injury, including embryonic lethality and impaired liver regeneration with deletion in adult bile ducts.^58,59 In contrast, our assessment of caspase activity and OCR in HepG2 cells, along with viability analysis in human hepatocyte HepaSH cells, revealed no toxic phenotype following YAP1 and WWTR1 knockdown by bsYW-61. These results indicate that partial suppression of YAP/TAZ in adult hepatocytes is tolerated under the tested conditions. Further studies using siRNAs with known or suspected hepatotoxic potential as reference controls could be informative. While a few siRNAs have been reported to induce hepatotoxicity in preclinical studies,^60,61 there appear to be no well-established or widely accepted positive siNCs for hepatotoxicity to date. Moreover, although the chemical modifications used in this study, such as 2′-OMe and 2′-F, have been widely reported to mitigate immune stimulation,^62,63,64 it would be valuable to further assess the potential siRNA-induced innate immune responses using primary human cells, such as peripheral blood mononuclear cells.

Bispecific siRNAs intentionally target two on-target genes, which may be considered off-targets in conventional siRNA design, raising questions about specificity. Therefore, we systematically examined the off-target profile of bsYW-61.⁶⁵ In silico predictions for full-length hybridization-dependent off-targets in humans identified no transcripts with a perfect match or single mismatch to the 17-nt targeting region, while 62 transcripts were found with two mismatches. Because even fully complementary siRNA sequences do not always result in effective knockdown, these in silico-predicted genes likely include many false positives. Consistent with this, RNA-seq analysis in HepG2 cells detected only 12 off-target DEGs, of which, PSME2 and NAPG were identified as genuine hybridization-dependent off-targets. With chemical modification (bsYW-61m1), suppression of PSME2 and NAPG was retained, indicating consistent target recognition between the unmodified and modified siRNAs. In addition, LAMC1 expression was newly affected, showing an approximately 10-fold higher IC₅₀ value compared with the on-target genes. However, two additional siRNAs with the same chemical modification pattern as bsYW-61m1 did not significantly reduce LAMC1 expression levels, suggesting that this off-target effect was unlikely modification dependent. Importantly, no evidence suggests that inhibition of these genes results in severe toxicity. For seed-mediated, microRNA-like off-targeting, although in silico prediction identified 1,500 transcripts with seed matches in their mRNA 3′ UTRs, none showed significant expression changes by RNA-seq analysis. In contrast, luciferase reporter assays, which capture both translational repression and mRNA destabilization, detected seed-mediated off-target activity for the unmodified bsYW-61. This activity was largely abolished with the chemically modified bsYW-61m1. These findings are consistent with those of previous reports demonstrating that chemical modification of the seed region can effectively suppress off-target activities.⁶⁶ Our luciferase reporter assays also showed that the passenger strand of bsYW-61 had no activity, indicating inefficient RISC loading. This strand-specific selection bias is likely influenced by the relative thermodynamic asymmetry⁴¹ that is further reinforced in bsYW-61m1 by the 2′-OMe at position 14 of the passenger strand, which disrupts its incorporation into the RISC.⁶⁷ Collectively, although further in vitro evaluation across multiple cell types and in vivo toxicity studies are needed, our preliminary evidence suggests that bsYW-61 does not induce toxic concerns from off-target suppression.

We also investigated the in vivo applicability of bsYW-61 using LNPs, a clinically validated platform for nucleic acid delivery as demonstrated by the approval of siRNA (Onpattro, 2018) and mRNA (COVID-19 vaccines, 2020) products.⁵⁰ ALC-0315, which is a synthetic ionizable lipid and key component of the Pfizer/BioNTech COVID-19 vaccine, was used to formulate liver-targeted LNPs.⁶⁸ In mice, bsYW-61m1 achieves dual knockdown in the liver at similar levels to the mixture of validated single-target siRNAs, demonstrating the versatility of the bispecific design. We further explored GalNAc conjugation, the most clinically advanced platform for liver-specific siRNA delivery.⁶⁹ To enhance the metabolic stability, we incorporated additional PS linkages and 5′-VP modifications.^70,71 In vivo experiments showed that 5′-VP conferred a clear advantage for knockdown efficacy in the liver. Although some reports have suggested that 5′-VP may not be required for tissues with high siRNA accumulation,^55,72 our data demonstrate a clear benefit for bispecific siRNAs. Increasing the 2′-OMe content is expected to further enhance the metabolic stability, suggesting potential for future improvements.^73,74

Beyond YAP1/WWTR1 targeting, we designed a bispecific siRNA against CREBBP and EP300. Through both enzymatic and scaffold activities, these targets can regulate c-Myc signaling and other transcriptional programs.^52,75 Previous studies have indicated that targeting CREBBP/EP300 can suppress c-Myc protein more effectively than direct MYC knockdown.⁵³ In our study, dual targeting of CREBBP and EP300 reduced both the MYC mRNA and c-Myc protein expression levels more effectively than individual knockdown, with c-Myc protein levels showing the most pronounced decrease. This stronger effect likely reflects the functional redundancy and cooperative action of CREBBP and EP300, as dual targeting can impair enhancer- and promoter-dependent MYC transcription and additionally reduce c-Myc protein stability via loss of CBP/p300-mediated acetylation.⁷⁶ The effect of dual knockdown was further reflected by significantly lower HepG2 cell proliferation rates. This siRNA approach can degrade target mRNAs and inhibit multiple protein functions, providing potential advantages over small-molecule inhibitors. Although further studies are needed to investigate the chemical modifications, in vivo efficacy, and cross-species activity, our findings support the broader applicability of bispecific siRNA strategies.

In conclusion, our bispecific siRNA technology offers a powerful strategy for targeting paralogs, which are promising therapeutic candidates, especially in the context of synthetic lethality. Compared with existing dual-targeting siRNA strategies, our design simplifies the manufacturing process, facilitates the optimization of delivery to targeted tissues, and streamlines the clinical development pathway. Additionally, bsYW-61 has shown cross-species activity and in vivo knockdown with a favorable safety profile regarding off-target activities and toxicity. While this approach is limited to homologous gene pairs, future refinements in sequence design and chemical modification may broaden its applicability to additional disease-relevant targets. Overall, our study introduces an innovative approach for simultaneously inhibiting two genes using a single siRNA and suggests a promising path toward new treatment methods for liver diseases.

Materials and methods

Statistics and reproducibility

Data analyses were performed using GraphPad Prism 10 software (GraphPad Software, Boston, MA, USA). One-way ANOVA with Dunnett’s or Tukey’s test was used for multiple comparisons, as specified in the figure legends. Differences in all datasets were considered statistically significant at p < 0.05.

Oligonucleotide synthesis

For in vitro siRNA sequence screening, we obtained oligonucleotides from NIPPON GENE Co., Ltd. (Tokyo, Japan) in desalting grade. For further in vitro and in vivo experiments, we synthesized the siRNAs using standard phosphoramidite, solid-phase synthesis conditions on a 0.2–1 μmol scale using a MerMade 12 (LGC Biosearch Technologies, Teddington, UK) and NTS T-60-HT (NIHON TECHNO SERVICE, Ibaraki, Japan). Oligonucleotides were synthesized on polystyrene beads or controlled pore glass, which were modified with DNA, 2′-OMe RNA, Unylinker (Kinovate Life Sciences [Oceanside, CA, USA] or LGC Biosearch Technologies), or GalNAc (Primetech, Minsk, Belarus). Phosphoramidites (Hongene Biotech [Shanghai, China], Thermo Fisher Scientific [Waltham, MA, USA], or LGC Biosearch Technologies) were prepared at 0.1 M in anhydrous acetonitrile (ACN), with added dry 10% dimethylformamide in the 2′-OMe-U amidite. 5-(Benzylthio)-1H-tetrazole was used as the activator at 0.3 M (Sigma-Aldrich, St. Louis, MO, USA). Detritylation was performed in 3% trichloroacetic acid in dichloromethane (FUJIFILM Wako, Osaka, Japan). Capping was done with CAP A, tetrahydrofuran:acetic anhydride:pyridine (8:1:1, v/v/v), and 10% n-methylimidazole in tetrahydrofuran (FUJIFILM Wako). Oxidation was performed using 0.05 mol/L iodine in pyridine:water (9:1, v/v) (FUJIFILM Wako). Sulfurization was performed with 0.05 M 3-((dimethylamino-methylidene)amino)-3H-1,2,4-dithiazole-3-thione in pyridine and acetonitrile (Glen Research, Sterling, VA, USA). Phosphoramidite coupling times were 3–6 min for all amidites used. For the chemically modified siRNAs, 5′-P or VP was introduced on the guide strand using solid chemical phosphorylating reagent (solidCPR, LK2127, LGC Biosearch Technologies) or 5′-POM-vinyl phosphonate, 2′-OMe-U CE-phosphoramidite (LK2579, LGC Biosearch Technologies).

Oligonucleotides were cleaved and deprotected with 40% methylamine aqueous for 1 h at 45°C. The solid-phase support was subsequently filtered, and then solutions were evaporated for 3 h using a centrifugal vacuum concentrator. Additionally, oligonucleotides with 2′-TBDMS-protecting groups were desilylated with triethylamine trihydrofluoride for 1 h at 65°C. Subsequently, the samples were washed with n-butanol and centrifuged for 10 min at 5,000×g and 4°C, and then the supernatant was discarded. This process was repeated three times.

Purifications were performed by anion exchange chromatography on an Agilent 1290 Infinity HPLC system (Agilent Technologies, Santa Clara, CA, USA) using Tricorn 10/100 columns with Source 30Q resin (Cytiva, Uppsala, Sweden). The loading solution used was 20 mM Na₂HPO₄ in 5% ACN in water (pH 7.0), and the elution solution was 1 M NaBr, 20 mM Na₂HPO₄ in 5% ACN in water (pH 7.0). A linear gradient was used from 25% to 50% for 29 min at a flow rate of 8 mL/min. The column temperature was set to 65°C. Peaks were monitored at 260 nm. Pure fractions (>85% HPLC purity) were combined and further desalted by size-exclusion chromatography on AKTA pure (Cytiva) using HiPrep 26/10 Desalting columns (Cytiva) at a flow rate of 10 mL/min. Single oligonucleotide strands were lyophilized from water, resuspended at the appropriate concentrations, and mixed in equimolar ratios to be annealed to form siRNA duplexes.

The purity and identity of fractions were confirmed by LC/ESI-MS on the 1260 Infinity LC/MSD G6125B system (Agilent Technologies) using an ACQUITY UPLC oligonucleotide BEH C18 column (1.7 μm 2.1 × 50 mm) at 65°C. Buffer A consisted of 10% 1,1,1,3,3,3-hexafluoro-2-propanol and 1% triethylamine in water, while buffer B was methanol: acetonitrile (1:1, v/v). A gradient from 2% to 30% of buffer B was used over 6 min at a flow rate of 0.5 mL/min. The column temperature was set to 65°C for single-stranded oligonucleotides and 75°C for double-stranded oligonucleotide analysis. Peaks were detected by measuring UV at 260 nm. The data are shown in Table S7.

Cell culture

Human HepG2 cells (Japanese Collection of Research Bioresources [JCRB], Osaka, Japan, JCRB1054) were maintained in low-glucose Dulbecco’s modified Eagle’s medium (DMEM, Thermo Fisher Scientific, 11885084). Human HeLa cells (American Type Culture Collection [ATCC], Manassas, VA, USA, CCL-2) and rat H-4-II-E cells (ATCC, CRL-1548) were maintained in Eagle’s minimum essential medium (ATCC, 30-2003). Mouse Hepa1-6 cells (ATCC, CRL-1830) were maintained in high-glucose DMEM (Thermo Fisher Scientific, 11995065). NHP MK.P3(F) cells (JCRB, JCRB0607) were maintained in DMEM/Ham’s F-12 (FUJIFILM Wako, 048-29785). All media were supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, 2437704) and 1% penicillin-streptomycin (Thermo Fisher Scientific, 15140122). All cells were grown at 37°C with a 5% CO₂ supply and were split every 2–7 days. For the initial broad screening across multiple species, HepG2, MK.P3(F), Hepa1-6, and H-4-II-E cells were used. Among these, HepG2, Hepa1-6, and H-4-II-E cells are well-established liver-derived models widely applied in liver disease research with extensive use for siRNA evaluation,^77,78,79,80 although kidney-derived MK.P3(F) was selected as a practical cynomolgus monkey cell line because of its stable culture properties and reported use in siRNA studies.⁸¹

Quantification of mRNA knockdown in cell lines

The siRNAs were introduced into cells by reverse transfection using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific, 13778-150) according to the manufacturer’s instructions. Cells were seeded at 25,000 cells/well or 12,500 cells/well (for IC₅₀ calculation) in 96-well plates. Twenty-four hours after transfection, the culture medium was removed, and total RNA was isolated using the RNeasy 96 Kit (QIAGEN, Venlo, the Netherlands, 74182). Total RNA was reverse transcribed into cDNA using the PrimeScript RT Reagent Kit (Takara Bio, Shiga, Japan, RR037A) and SimpliAmp Thermal Cycler (Thermo Fisher Scientific). Then, RT-qPCR was performed using the THUNDERBIRD Next SYBR qPCR Mix (TOYOBO, Osaka, Japan, QPX-201) and QuantStudio 5 PCR System (Thermo Fisher Scientific). The qPCR cycling conditions were as follows: an initial denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 5 s and extension at 60°C for 30 s. This was concluded with a melting/dissociation curve analysis. The relative expression levels of each target gene were determined by comparison with levels of the internal reference gene, RPLP0 for humans and cynomolgus monkeys, Atp5f1 for mice, and Gapdh for rats, using the ΔΔCt method. The data are represented as relative to the mock-transfected controls. The primer sets used for RT-qPCR are shown in Table S8. IC₅₀ curves were fitted using log(inhibitor) vs. normalized response-variable slope in GraphPad Prism (GraphPad Software, Inc., Boston, MA, USA).

Quantification of mRNA knockdown in human hepatocytes

Human hepatocyte HepaSH cells (Central Institute for Experimental Medicine and Life Science, Kanagawa, Japan, FSSF001, 25-2423) were seeded at 1 × 10⁶ cells/well in 96-well collagen I-coated plates (Corning, Corning, NY, USA, 356407) in culture medium (MIL222) according to the manufacturer’s instructions. After 24 h, the medium was replaced with Cellartis Power Primary HEP Medium (Takara Bio, Y20020), and then the siRNAs were transfected using Lipofectamine RNAiMAX according to the manufacturer’s instructions. After 72 h of transfection, RT-qPCR was performed as described previously.

Cell viability assay

Human HepG2 cells were seeded at 12,500 cells/well in 96-well plates (for CellTiter-Glo 2.0 Assay) or 48-well plates (for confluence analysis by Incucyte) and reverse transfected with siRNAs using Lipofectamine RNAiMAX, as described previously. At 5 or 7 days post transfection, cell viability was assessed using the CellTiter-Glo 2.0 Assay (Promega, Madison, WI, USA, G9241), with luminescence measured on an EnSight plate reader (PerkinElmer, Waltham, MA, USA). Cell confluence was quantified with the Incucyte live-cell imaging system (Sartorius, Göttingen, Germany). Human HepaSH cells were seeded at 1 × 10⁶ cells/well in 96-well collagen I-coated plates and transfected using Lipofectamine RNAiMAX 24 h later. At 72 h post transfection, cell viability was measured using the CellTiter-Glo 2.0 assay, with dead cells quantified with Cytotox Green Dye (Sartorius, 4633) using the Incucyte system. All results are presented as values compared with the mock controls.

Apoptosis assay

Apoptosis assays were performed by Axcelead Inc. (Tokyo, Japan). Human HepG2 cells were seeded at 3,000 cells/well in a 384-well plate. Oligonucleotides were transfected into the cells using Lipofectamine RNAiMAX and incubated for 24 h. Caspase-3/7 activity was determined using the Caspase-Glo 3/7 assay (Promega, G8090) according to the manufacturer’s instructions on an EnVision plate reader (PerkinElmer). Caspase activity was represented as relative to the untreated controls. The control ASO sequences are listed in Table S9.

Oxygen consumption measurement

The OCR was measured using a Flux Analyzer XFe96 (Agilent Technologies) by Axcelead Inc. Human HepG2 cells (ATCC, HB8065) cultured in a 96-well plate were transfected with siRNAs using Lipofectamine RNAiMAX. After 24 h, the medium was replaced with Seahorse XF DMEM medium (Agilent, 103575-100) containing glucose (Agilent, 103577-100) and pyruvate (Agilent, 103578-100). The steady-state OCR was first measured, and then oligomycin (2 μM), an ATP synthase inhibitor, and FCCP (0.5 μM), a mitochondrial oxidative phosphorylation uncoupler, were injected to assess the maximal OCR. Each OCR value was represented as relative to the mock controls.

Luciferase reporter assay

Human HeLa cells were seeded at 20,000 cells/well in 96-well plates. Twenty-four hours later, the cells were co-transfected with 10 ng luciferase reporter plasmid and siRNAs at 6-fold dilutions using 0.2 μL Lipofectamine 2000 (Thermo Fisher Scientific, 11668027). The cells were harvested 24 h post transfection for the dual-luciferase assay (Promega, E1960) according to the manufacturer’s instructions. Luminescence was measured using an EnSight Plate Reader (PerkinElmer). Reporter plasmids were generated by cloning specific sequences into the psiCHECK-2 vector (Promega, C8021) between the Xho1 and Not1 restriction sites.⁸² GS-CM and PS-CM each contained a full-length bsYW-61 complementarity site (guide or passenger strand, respectively) in the Renilla luciferase 3′ UTR. GS-SM4 and PS-SM4 each contained four tandem seed-complementary sites (guide or passenger, respectively) in the Renilla luciferase 3′ UTR. Both plasmids co-expressed Firefly luciferase as a transfection control. Relative luciferase activity was calculated by normalizing the Renilla luciferase activity to Firefly luciferase activity relative to the siNC data. The insert sequences are displayed in Table S10.

RNA-seq and bioinformatics analyses

Human HepG2 cells were seeded at 125,000 cells/well in 24-well plates. The siRNAs were introduced into the cells by reverse transfection using Lipofectamine RNAiMAX, as described previously. After 24 h, total RNA was isolated as described previously. The library preparation, sequencing, and reference genome mapping were performed by Takara Bio Inc. The DNA libraries were prepared using the Nextera XT DNA Library Prep Kit (Illumina, San Diego, CA, USA) and sequenced on a NovaSeq 6000 sequencer (Illumina). RNA-seq reads were aligned to the Homo sapiens reference genome (GRCh38.primary_assembly.genome.fa.gz). Differential gene expression analysis was performed using the R package DESeq2, with log₂ fold-change estimates for low-count genes shrunken using lfcShrink (type = “apeglm”).

The siRNA sequence complementarity to mRNA sequences was examined in silico using GGGenome (https://gggenome.dbcls.jp/en/). Transcripts with 3′ UTR sequences complementary to the siRNA seed region were identified using the R package SeedMatchR.⁸³

siRNA/LNP formulation

LNPs were constituted with ALC-0315 (HY-138170, MedChemExpress, Monmouth Junction, NJ, USA), DSPC (MC-8080, NOF, Tokyo, Japan), cholesterol (08721-62, Nacalai Tesque, Kyoto, Japan), and DMG-PEG2K (GM-020, NOF) at a molar ratio of 50:10:38.5:1.5. siRNA/LNPs were formulated using the NanoAssemblr System (Precision NanoSystems, Vancouver, BC, Canada). Briefly, the lipid mixture in ethanol and siRNA solution (50 mM citrate buffer, pH 3.5) were mixed through microfluidic cartridges, with the aqueous and organic phases combined at a 65:35 ratio (buffer to ethanol). The nitrogen/phosphate ratio was 3.5. The resulting LNP solution was diluted with 20 mM HEPES containing 9% sucrose (pH 7.45) and then concentrated and purified by ultrafiltration (Amicon Ultra-15, MWCO 100 kDa; Millipore, Billerica, MA, USA). LNPs were filtered through a 0.2 μm polyethersulfone filter and stored in liquid form at 4°C.

Animal studies

All animal studies were approved by the Nitto Denko Corporation Animal Care and Use Committee (Approval No. C-23O) and followed the guidelines for the care and use of laboratory animals. Male wild-type C57BL/6J mice were purchased from Jackson Laboratory Japan (Kanagawa, Japan). The mice were housed under conditions of 24°C ± 1°C, with a 12/12-h light/dark cycle and free access to standard solid feed (CRF1, Oriental Yeast, Tokyo, Japan) and tap water. The mice were 5 weeks old at the time of the experiments.

Four mice per group were intravenously administered a single dose of HEPES buffer, 1 mg/kg siRNA/LNPs (single siRNA), or 2 mg/kg siRNA/LNPs (siYap1 + siWwtr1 mixture) and sacrificed on day 2 post dose. Separately, four mice per group were subcutaneously administered a single dose of PBS or 10 mg/kg GalNAc-siRNA conjugates and sacrificed on day 7 post dose. Blood was collected from the inferior vena cava using a heparin-treated 26G syringe (Japanese Pharmacopoeia, Tokyo, Japan). Plasma was separated by centrifugation for 5 min at 20,400 × g and 4°C. Liver tissues were harvested and weighed, and then portions (<50 mg) were immersed in Allprotect Tissue Reagent (QIAGEN, 76405) for subsequent mRNA quantification. These samples were stored at −80°C until analysis. Plasma AST and ALT levels were measured using the Transaminase CII-Test Wako (FUJIFILM Wako, 38556000) according to the manufacturer’s instructions. Absorbance values were measured using an EnSight multimode plate reader (PerkinElmer).

Quantification of mRNA knockdown in tissues

Liver tissue samples were transferred to tubes containing QIAzol Lysis Reagent (QIAGEN, 79306) and homogenized using a bead homogenizer (Multi-Beads Shocker, Yasui Kikai, Osaka, Japan). Total RNA was isolated from the tissue samples using the RNeasy Plus Universal Mini Kit (QIAGEN, 73404). cDNA synthesis and RT-qPCR were performed following the same procedure as for the in vitro experiments. Relative expression levels were determined by comparison with Atp5f1 and represented as relative to the vehicle group.

Quantification of c-Myc protein by western blot analysis

Human HepG2 cells were seeded at 45,000 cells/well in 6-well plates (Corning Life Sciences, 3516). After 24 h, the cells were transfected with siRNAs using Lipofectamine RNAiMAX, as described previously. Five days post transfection, the culture medium was removed, and the cells were lysed in RIPA buffer (Thermo Fisher Scientific, 89901) supplemented with phosphatase inhibitor tablets (PhosSTOP, Roche, Basel, Switzerland, 4906845001) and protease inhibitor cocktail (cOmplete, Roche, 11836153001). The protein concentrations of the lysates were determined using the BCA Protein Assay Kit (Thermo Fisher Scientific, 23225). c-Myc protein expression levels were analyzed using an anti-c-Myc antibody (Cell Signaling Technology, Danvers, MA, USA, 18583, 1:50 dilution) on the Simple Western System Jess (ProteinSimple, San Jose, CA, USA).

Data and code availability

The authors confirm that the data supporting the findings of this study are available within the article and/or its supplemental information.

Acknowledgments

The authors thank Junko Yamanaka, Kanako Ishiguro, and Akiko Sakotani for their excellent support in the experiments. All work in this study is sponsored by Nitto Denko Corporation with no additional funding or grants. The authors would also like to thank J. Iacona, Ph.D., from Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript. The graphical abstract was created with BioRender.com.

Author contributions

K.U., H.T., and J.S. designed the siRNAs. K.U. and J.S. synthesized and purified the siRNAs. K.U. and Y.Y. conducted the in vitro sequence screening. K.U. carried out the knockdown evaluations in multiple cell lines, cell viability assays, off-target evaluations, and bioinformatic analyses. T.S. performed the in vivo studies. S.S. planned and managed the outsourced toxicity studies. M.S., K.M., and H.Y. supervised the overall studies. The manuscript was written by K.U. and reviewed by M.S. and H.T. All authors read and approved the final manuscript.

Declaration of interests

All authors are employees of and have stock options in Nitto Denko Corporation. A patent application has been submitted for the initial findings from this work: Japanese patent application no. 2024-171245.