Galactosylated LNP-mediated hepatic Sam68 silencing improves glucose homeostasis in diabetes by suppressing gluconeogenesis

Abstract

Excessive hepatic glucose production is a key driver of the progression of type 2 diabetes (T2DM), a highly prevalent global metabolic disorder. We previously reported that Sam68 expression is upregulated in the livers of both diabetic patients and mouse models, and hepatocyte-specific knockdown of Sam68 greatly alleviates hyperglycemia and improves insulin sensitivity in diabetic mice. Here, we engineered a series of ligand-functionalized lipid nanoparticles (LNPs) and identified a galactose-decorated formulation (LNP-Gal) that enables efficient hepatocyte-selective delivery of Sam68 siRNA, thereby achieving robust Sam68 silencing and suppressing hepatic gluconeogenesis in both cellular and animal models. In both genetic and diet-induced diabetic mouse models, systemic administration of siSam68/LNP-Gal improved glycemic control and insulin responsiveness and attenuated hepatic gluconeogenic output, accompanied by suppression of the hepatic gluconeogenic program. Thus, we establish siSam68/LNP-Gal as a hepatocyte-selective siRNA delivery system that elicits a potent antihyperglycemic effect with a favorable safety profile in vitro and in vivo, providing a promising siRNA-based strategy for the treatment of T2DM and related metabolic disorders.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04511-1.

Introduction

Diabetes mellitus is a globally prevalent chronic metabolic disorder characterized by elevated blood glucose levels [1–3]. In 2021, approximately 529 million people worldwide were living with diabetes (about 90% to 95% of them have T2DM), and this number is projected to exceed 1.3 billion by 2050 [4, 5]. Owing to its insidious onset, multifactorial pathogenesis, and frequent association with cardiovascular disease, chronic kidney disease, and other complications, T2DM has emerged as one of the leading causes of morbidity and mortality, posing a major global public health challenge [6–8]. Although the pathophysiological mechanisms of T2DM are complex, disruption of the finely tuned balance between insulin, which lowers blood glucose, and counter-regulatory hormones such as glucagon is recognized as a major driver [9]. In T2DM, upregulated glucagon signaling-mediated sustained activation of the cAMP-response element binding protein (CREB)/CREB-regulated transcription coactivator 2 (CRTC2) pathway drives excessive hepatic gluconeogenesis and chronic hyperglycemia, highlighting the CREB/CRTC2 complex and its upstream regulatory components as potential therapeutic targets [10].

Src-associated in mitosis 68 kDa (Sam68), a member of the signal transduction and activation of RNA family of RNA-binding proteins, orchestrates a wide spectrum of cellular processes, including RNA splicing, transcription, signal transduction, and metabolic regulation [11–17]. We previously demonstrated that global Sam68 knockout mice are resistant to high-fat diet (HFD)–induced obesity and exhibit enhanced systemic energy expenditure [18]. Moreover, our subsequent findings indicated that Sam68 mRNA and protein expression are elevated in the livers of both diabetic patients and mouse models, whereas hepatocyte-specific silencing of Sam68 in diabetic mice markedly attenuates hepatic glucagon signaling by suppressing the CRTC2-CREB transcriptional complex-mediated gluconeogenesis, thereby mitigating hyperglycemia and improving insulin sensitivity [19, 20]. Collectively, these findings establish hepatic Sam68 as a pivotal regulator of glucose and lipid homeostasis and highlight it as a promising molecular target for the treatment of T2DM-associated metabolic disorders.

RNA interference (RNAi) using small interfering RNA (siRNA) is a broadly applicable therapeutic strategy that enables sequence-specific post-transcriptional silencing of disease-relevant genes [21, 22]. However, compared with small molecules or biologics, naked siRNA suffers from rapid nuclease degradation, inefficient cellular uptake, and poor tissue-selective biodistribution, which together limit in vivo efficacy [23, 24]. These delivery barriers have driven the development of carrier-based systems for systemic siRNA delivery. Among these, lipid nanoparticles (LNPs) have emerged as the most clinically advanced non-viral platform for nucleic acid delivery, providing efficient encapsulation, protection in circulation, and enhanced intracellular delivery [25–27]. The platform’s transformative clinical potential is exemplified by patisiran (Onpattro), the first FDA-approved LNP-formulated siRNA, which targets hepatic transthyretin (TTR) to suppress TTR production and improve polyneuropathy in patients with hereditary TTR-mediated amyloidosis [28, 29]. Nevertheless, despite the apparent liver tropism of systemically administered LNPs, a substantial fraction is preferentially sequestered by hepatic non-parenchymal cells, particularly liver sinusoidal endothelial cells and Kupffer cells, via the fenestrated sinusoidal architecture and reticuloendothelial system (RES)-mediated clearance, thereby limiting the amount that reaches hepatocytes across the space of Disse [30–33]. Consequently, effective modulation of hepatocyte-enriched targets such as Sam68 requires delivery strategies that enrich functional siRNA delivery to hepatocytes while reducing RES sequestration.

Ligand-decorated LNPs have been widely exploited for cell type-selective delivery in vivo by engaging receptors enriched on target cells, thereby reducing non-specific uptake and enhancing functional payload delivery [34, 35]. In the liver, several ligands such as glycyrrhetinic acid (Gly) [36], folic acid (Fa) [37], mannose (Man) [38], galactose (Gal) [39], cholic acid (Ca) [40] and retinoic acid (Ret) [41] have been reported to enhance liver cell targeting by leveraging receptor-mediated uptake. Building on this concept, we established a modular ligand-functionalization and screening platform in which engineered ligand-polymer conjugates are installed onto LNP surfaces via click chemistry, enabling rapid, standardized ligand exchange and efficient head-to-head comparison of hepatocyte delivery performance. Using this workflow, we identified galactose-decorated LNPs (LNP-Gal) as providing the highest hepatocyte delivery efficiency, and subsequently encapsulated Sam68 siRNA to generate siSam68/LNP-Gal. We systematically evaluated this formulation in mouse primary hepatocytes, an HFD/streptozotocin (HFD/STZ)-induced T2DM model, and in db/db mice. Here we show that siSam68/LNP-Gal efficiently silences hepatic Sam68 and suppresses gluconeogenic gene expression both in vitro and in vivo, thereby mitigating hyperglycemia and improving systemic glucose homeostasis. Thus, our results establish siSam68/LNP-Gal as a promising therapeutic modality with strong translational potential for the treatment of T2DM.

Materials and methods

LNP formulation reagents

DLin-MC3-DMA, DSPC, and DMG-PEG were purchased from A.V.T. Pharmaceutical Co., Ltd. DMG-PEG-Mal was obtained from Weihua Biotechnology Co., Ltd. Thiol/amino double-terminated PEG1000 (HS-PEG-NH₂) and thiol/carboxyl double-terminated PEG 1000 (HS-PEG-COOH) were obtained from Ponsure Biotechnology Co., Ltd. Mannopyranosylamine and galactopyranosylamine were purchased from Biosynth, and other chemical reagents were sourced from Macklin.

Synthesis of thiol-terminated targeting ligands

Targeted ligands were synthesized as illustrated in Fig. S1C. For Gly-SH, Fa-SH, and Ca-SH, the corresponding acids (glycyrrhetinic, folic, or cholic acid) were dissolved in anhydrous dichloromethane (DCM), followed by the addition of 1.1 equivalents of N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDCI) and N-hydroxysuccinimide (NHS). After activation, 0.9 equivalents of HS-PEG-NH₂ were added, and the reaction was stirred overnight at room temperature. For Man-SH and Gal-SH, HS-PEG-COOH was dissolved in anhydrous DCM, followed by 1.1 equivalents of EDCI and NHS, then 1.1 equivalents of mannosamine or galactosamine were added and stirred overnight. For Ret-SH, retinol was first reacted with 1.1 equivalents of succinic anhydride in anhydrous DCM for 1 h, followed by aqueous sodium bicarbonate extraction, drying, weighing, and redissolving in anhydrous DCM. Then, 1.1 equivalents of EDCI and NHS and 0.9 equivalents of HS-PEG-NH₂ were sequentially added and stirred overnight. All reaction mixtures were purified by silica gel column chromatography using DCM: methanol (15:1, v/v) as the mobile phase. Product structures were confirmed by ¹H NMR spectroscopy:

Gly-SH: ¹H NMR (500 MHz, MeOD) δ 0.77–0.79 (m, 3 H, CH3), 0.82–0.84 (m, 3 H, CH3), 0.96–1.08 ( m, 6 H, 2CH3), 1.13–1.17 (m, 9 H, 3CH3), 1.42–1.51 (m, 9 H, CH2CH2C, CCH2CH2, CH), 1.72–1.80 (m, 2 H, CH2), 1.89–1.99 (m, 4 H, CH2CH2), 2.18–2.21 (m, 2 H, CH2 ), 2.51–2.54 (m, 3 H, CH, CH2 ), 2.73–2.79 (m, 4 H, CH2CCH2 ), 2.89–2.91 (m, 1H, OH), 3.16–3.20 (m, 2 H, CH2 ), 3.38–3.40 (m, 3 H, CH2, OCH ), 3.56–3.58 (m, 4 H, CH2N, OCH2), 3.63–3.66 (m, 162 H, nOCH2CH2), 5.67 (s, 1H, NH).

Fa-SH: ¹H NMR (500 MHz, DMSO) δ 1.02–1.05 (m, 1H, CH),1.12–1.15 (m, 1H, CH), 2.36–2.37 (m, 2 H, OCCH2), 2.64–2.65 (m, 2 H, SCH2), 3.02–3.03 (m, 2 H, CH2N), 3.13–3.17 (m, 4 H, CH2O, OCH2), 3.31–3.33 (m, 4 H, CCHN, NCH2C, CHC), 4.07 (s, 142 H, nOCH2CH2), 6.62–6.82 (m, 1H, NH).

Man-SH: ¹H NMR (500 MHz, MeOD) δ 1.31–1.34 (m, 1H, SH), 2.51–2.54 (m, 2 H, SCH2), 2.75–2.79 (m, 2 H, CH2O), 2.89–2.93 (m, 1H, CCH), 3.13–3.19 (m, 1H, CCH), 3.32–3.40 (m, 3 H, OCH, OCH2), 3.54–3.59 (m, 3 H, CCHCHCHC) 3.65 (s, 148 H, nOCH2CH2), 3.77–3.80 (m, 2 H, OCH2C), 4.22–4.29 (m, 3 H, CCOHCOHCOH), 5.51 (s, 1H, COH).

Gal-SH: ¹H NMR (500 MHz, MeOD) δ 1.31–1.33 (m, 1H, SH), 2.42–2.46 (m, 4 H, SCH2, CH2O), 2.52–2.56 (m, 1H, CCH), 2.59–2.64 (m, 3 H, CCH, OCH2), 2.89–2.93 (m, 1H, OCH), 3.02–3.23 (m, 7 H, nOCH2CH2, CH), 3.35–3.40 (m, 4 H, nOCH2CH2), 3.52–3.55 (m, 4 H, OCH2C, CH2), 3.56–3.59 (m, 5 H, CCHCHCHC, OHCOH), 3.62–3.69 (m, 146 H, nOCH2CH2), 3.78–3.80 (m, 1H, COH), 4.13–4.16 (m, 1H, COH).

Ca-SH: ¹H NMR (500 MHz, MeOD) δ 0.73–1.05 (m, 9 H, 3CH3), 1.29–1.48 (m, 7 H, CH2, CH2, CH2, CH), 1.54–1.68 (m, 6 H, CH2C, CH2, CH2 ), 1.81–1.89 (m, 5 H, CCH2,CH2,CH ), 1.98–2.01 (m, 2 H, CH2), 2.26–2.34 (m, 2 H, CH2), 2.51–2.56 (m, 1H, CH), 2.60–2.63 (m, 2 H, SCH2), 2.74–2.80 (m, 2 H, OCCH2), 2.88–2.91 (m, 2 H, CH2N ), 3.10–3.19 (m, 4 H, CH2O, OCH2 ), 3.34–3.41 (m, 3 H, CH, OH, OH), 3.53–3.58 (m, 3 H, CH, CH, OH ), 3.65 (s, 72 H, nOCH2CH2), 5.51 (s, 1H, NH).

Ret-SH: ¹H NMR (500 MHz, MeOD) δ 1.27–1.30 (m, 6 H, CH3, CH3), 2.48–2.77 (m, 4 H, CCCH2CH2), 2.88–2.88 (m, 4 H, CH3, CH), 3.01 (m, 4 H, CH2, CH2), 3.11–3.14 (m, 4 H, CH3, CH), 3.18–3.19 (m, 1H, CH), 3.45–3.47 (m, 1H, CH), 3.55–3.56 (m, 4 H, CH3, CH), 3.62–3.66 (m, 20 H, CH2, nOCH2CH2), 4.60–4.63 (m, 3 H, OCH2CH), 7.99 (s, 1H, NH).

Preparation of LNPs

Non-targeted LNPs were synthesized using a rapid microfluidic mixing method. DLin-MC3-DMA, DSPC, cholesterol, and DMG-PEG were dissolved in ethanol at a molar ratio of 50:10:38.5:1.5. siRNA was dissolved in 25 mM sodium acetate buffer (pH 4.0). The two phases were combined using herringbone-patterned microfluidic chips at an aqueous-to-ethanol volumetric ratio of 3:1 with a total flow rate of 12 mL/min. The resulting LNPs were dialyzed against PBS (1×, pH 7.4) using Slide-A-Lyzer MINI dialysis devices (MWCO 3.5 kDa, Thermo Fisher) for 2 h. For targeted LNPs, DMG-PEG was partially replaced by DMG-PEG-Mal to obtain LNP-Mal. Ligands (Gly-SH, Ca-SH, Man-SH, Gal-SH, Fa-SH, or Ret-SH) were added at a 1:1 molar ratio relative to maleimide groups and incubated for 8 h at 4 °C to generate targeted LNPs (LNP-Gly, LNP-Ca, LNP-Man, LNP-Gal, LNP-Fa, and LNP-Ret). The resulting targeted LNPs were purified and washed using ultrafiltration tubes (MWCO 100 kDa, Merck Millipore) to remove unmodified free ligands. All formulations were filtered through a 0.22 μm sterilizing membrane before in vitro or in vivo use.

Characterization of LNPs

Hydrodynamic size, polydispersity index, and zeta potential were measured using a Dynamic Light Scattering (Malvern, Zetasizer Pro). siRNA encapsulation efficiency (EE) was quantified using the RiboGreen assay kit (Invitrogen, R11490). Free siRNA was detected by fluorescence (Ex/Em = 480/520 nm) on a microplate reader (Agilent, BioTek Synergy H1). Total siRNA was measured after LNP disruption with 1% Triton X-100, and EE was then calculated. LNP morphology was observed by negative-stain transmission electron microscopy (JEOL, HT7800). Cryo-transmission electron microscope (Cryo-TEM) imaging was performed on a 300 kV Cryo-TEM microscope (Thermo Fisher, Krios G4) with a direct electron detector. Samples were prepared using a FEI Vitrobot Mark IV and imaged at 92,000× magnification (1.566 Å/pixel). The efficiency of targeted ligand modification was assessed by ultrafiltration of the modified LNP system using ultrafiltration tubes (MWCO 100 kDa, Merck Millipore). The filtrate was then reacted with 0.25 mg/mL 5,5’-Dithiobis-(2-nitrobenzoic acid) (DTNB; Bidepharm, BD15685), and the absorbance at 412 nm was measured to determine the free thiol content, which was used to estimate the ligand modification level on the LNPs. Agarose gel electrophoresis was performed using 1% agarose in 0.1 M histidine/0.1 M MES buffer (pH 6.1) at 100 V for 30 min.

siRNA design and validation

To identify effective siRNA sequences targeting mouse Sam68, we designed three candidate siRNAs, siSam68-1 (sense: 5′-GAAAGAACGCGUGCUGAUAdTdT-3′, antisense: 5′-UAUCAGCACGCGUUCUUUCdTdT-3′), siSam68-2 (sense: 5′-GAGGAGAAUUAUUUGGAUUdTdT-3′, antisense: 5′-AAUCCAAAUAAUUCUCCUCdTdT-3′), and siSam68-3 (sense: 5′-UUACGAAGCCUACGGACAAdTdT-3′, antisense: 5′-UUGUCCGUAGGCUUCGUAAdTdT-3′), each targeting different regions of the mouse Sam68 mRNA. A non-targeting scrambled siRNA (sense: 5′-UUCUCCGAACGUGUCACGUdTdT-3′, antisense: 5′-ACGUGACACGUUCGGAGAAdTdT-3′) with no known homology to the mouse genome was used as a control. These siRNAs, including the FAM-labeled siRNA and Cy5-labeled siRNA, were commercially synthesized and purchased from RiboBio (Guangzhou, China). Total RNA and protein in mouse primary hepatocytes were extracted 24 and 48 h after Lipofectamine 3000-mediated siRNA transfection, respectively, and Sam68 knockdown efficiency was assessed by qRT-PCR and immunoblotting. The siRNA that most effectively reduced Sam68 expression at both the mRNA and protein levels was selected for subsequent experiments.

Cell lines

AML12, RAW264.7, and mHSC-SV40 cell lines were obtained from Fenghui Biotechnology Co., Ltd. AML12 cells were cultured in DMEM (Gibco, C11995500BT) supplemented with 10% fetal bovine serum (FBS; Gibco, A5669701), 1% penicillin/streptomycin (Gibco, 15140122), 5 µg/mL insulin (Medchemexpress, HY-P0035), 5 µg/mL transferrin (Procell, PB180429), 5 ng/mL selenium (Medchemexpress), and 40 ng/mL dexamethasone (Medchemexpress, HY-14648). RAW264.7 and mHSC-SV40 cells were cultured in DMEM with 10% FBS and 1% penicillin/streptomycin.

Cellular uptake

RAW264.7, AML12, and mHSC-SV40 cells were seeded into glass-bottom dishes and cultured for 24 h. FAM-labeled siRNA-loaded LNPs (0.2 nmol) were added and incubated with the cells for 8 h. Nuclei were stained with Hoechst 33342 for 15 min, followed by PBS washing. Cellular uptake was visualized using a confocal laser scanning microscope (OLYMPUS, FV3000). For flow cytometry analysis, cells were seeded into 6-well plates and cultured for 12 h, then treated with FAM-siRNA-loaded LNPs (0.2 nmol) for 8 h. After incubation, cells were collected, centrifuged, and re-suspended in FACS buffer. Fluorescence intensity was measured using a flow cytometer (Agilent, NovoCyte Quanteon). Data were analyzed using FlowJo software.

Endosomal escape

Endosome-mimicking liposomes with FRET functionality were prepared by thin-film hydration using a lipid mixture of DPPG: DOPC: DOPE: NBD-PE: Rho-DHPE (22:22:41:1:1, w/w). In a black 96-well plate, 100 µL of PBS (pH 5.5 or 7.4) was added to each well, followed by 2 µL each of LNPs and FRET-liposomes. After 5 min incubation at 37 °C, fluorescence spectra (Ex = 465 nm, Em = 550–650 nm) were recorded using a microplate reader at designated time points. For confocal microscopy analysis, AML12 cells were seeded into glass-bottom dishes and treated with FAM-siRNA-loaded LNPs (0.2 nmol) for 1–4 h. Nuclei and lysosomes were stained with Hoechst 33342 and LysoTracker Red DND-99, respectively, according to the manufacturer’s protocols. Images were acquired using a confocal laser scanning microscope, and fluorescence distribution was analyzed using ImageJ software.

In vitro cytotoxicity

AML12 cells were seeded into 96-well plates and cultured for 12 h. Cells were then incubated with siSam68/LNP or siSam68/LNP-Gal at siRNA concentrations of 10, 30, or 100 nM for 24 h. After treatment, CCK-8 (Glpbio, GK10001) reagent (10 µL/well) was added to the culture medium and incubated at 37 °C for 2 h. Absorbance at 450 nm was measured using a microplate reader, and cell viability was calculated relative to untreated controls.

Proteomic analysis of protein corona

Plasma was obtained from wild-type (WT) (C57BL/6J; SM-001) mice and incubated with LNPs (1:1, v/v) at 37 °C for 1 h. LNP–protein complexes were collected by centrifugation and washed three times with PBS. Total protein was quantified using the BCA assay. For each sample, 15 µg of protein was reduced with dithiothreitol, alkylated with iodoacetamide, and digested with trypsin overnight at 37 °C. Peptides were desalted, vacuum-concentrated, and reconstituted in Buffer A (H₂O containing 0.1% formic acid) before LC-MS/MS analysis. Samples were analyzed using a mass spectrometer (Thermo Fisher Scientific Orbitrap Ascend Tribrid) operated in data-dependent acquisition mode. Raw LC-MS/MS data were processed using Spectronaut or MaxQuant software and searched against the UniProt mouse proteome database. The false discovery rate (FDR) was set to 1% at the protein, peptide, and site levels.

Animal studies

All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the Zhongshan Institute for Drug Discovery, Chinese Academy of Sciences, and were conducted in strict accordance with the Regulations for the Administration of Laboratory Animals and the Measures for the Quality Management of Laboratory Animals of the People’s Republic of China, as well as other relevant ethical guidelines. WT mice and db/db (Lepr^db/J) mice were purchased from Shanghai Model Organisms Center, Inc. and Jiangsu Huachuang Sino Pharma Tech Co., Ltd., respectively. Male mice aged 2–3 months were used for all experiments unless specified otherwise. Animals were housed in specific pathogen-free facilities under controlled temperature (22 °C) and humidity (~ 43%) with a 12-h light/dark cycle, and provided with standard chow and water ad libitum unless specified otherwise. All procedures followed approved experimental guidelines including randomization, appropriate anesthesia and analgesia, and humane endpoints. Mice were euthanized using approved methods to ensure animal welfare.

In vivo imaging and liver distribution

Cy5-labeled siRNA-loaded LNPs were intravenously injected into C57BL/6 mice at a dose of 37.5 nmol/kg. Whole-body fluorescence imaging was performed at various time points post-injection using an IVIS imaging system (PerkinElmer, Lumina III). The fluorescent dyes and organelle staining reagents were obtained from Yeasen. After 24 h, mice were euthanized, and major organs were harvested for ex vivo imaging. Fluorescence intensity was analyzed using Living Image software. For tissue distribution studies, FAM-labeled siRNA-loaded LNPs (1 mg/kg) were administered intravenously into the mice. At 4 h post-injection, livers were harvested, fixed in 4% paraformaldehyde, cryoprotected in 30% sucrose, and embedded in OCT. Cryosections (8–10 μm) were prepared using a cryostat (Thermo Fisher, CryoStar NX50), stained with DAPI, and scanned using a digital slide scanner (Olympus, VS200).

Establishment of HFD/STZ–induced type 2 diabetic mouse model

4-week-old WT male mice were fed HFD (58% of total calories from fat, D12492, Research Diets) for 12 weeks. At the end of week 12, following 6 h of fasting, mice were given a single intraperitoneal injection of freshly prepared STZ (100 mg/kg in 0.1 M citrate buffer, pH 4.5; LEAGENE, R00522). Following the injection, mice were maintained on HFD for an additional 4 weeks. To confirm the establishment of T2DM model, fasting blood glucose (after 3 h fasting) was measured on day 5 post-injection and at the end of the 4-week HFD maintenance period. Mice with fasting blood glucose ≥ 250 mg/dL at either time point were considered diabetic and included in subsequent experiments.

siRNA/LNP treatment

HFD/STZ diabetic or db/db mice were randomly assigned to receive siNC/LNP-Gal, siSam68/LNP-Gal, or metformin. siRNA/LNP-Gal was administered at a dose of 1 mg/kg via tail-vein injection every six days, throughout the 4-week intervention period. Metformin was administered by oral gavage once daily at either 50 mg/kg (low dose) or 150 mg/kg (high dose). Body weight and general health status were monitored throughout the treatment period. Blood glucose measurements and other metabolic tolerance tests were performed at the indicated time points.

Glucose, pyruvate, glucagon, and insulin tolerance tests (GTT, PTT, GcTT, and ITT)

Prior to metabolic tolerance tests, mice were fasted for different durations according to the experimental protocol; 16 h for GTT and PTT, and 6 h for GcTT and ITT. Following fasting, mice received intraperitoneal injections of the indicated agents at doses adjusted for each diabetic model used. For HFD/STZ diabetic mice, glucose and pyruvate were administered at 1.0 g/kg, glucagon at 10 µg/kg, and insulin at 2 U/kg. For db/db mice, glucose and pyruvate were administered at 0.5 g/kg, glucagon at 5 µg/kg, and insulin at 2 U/kg. Blood glucose was measured from the tail vein at 0, 15, 30, 45, 60, 90, and 120 min after agent injection using a handheld glucometer (Yuwell, Model 590/Bayer Contour Blood Glucose Meter). To minimize experimental variability, all injections and blood sampling were performed by the same investigator in a quiet, temperature-controlled environment (22 ± 2 °C).

In vivo insulin signaling analysis

Mice were fasted for 16 h, and then received intraperitoneal injections of either saline or insulin (1 U/kg for HFD/STZ diabetic mice and 2 U/kg for db/db mice). Twenty minutes after injection, mice were euthanized, and liver tissues were rapidly excised, snap-frozen in liquid nitrogen, and stored at − 80 °C until analysis. Protein extraction and immunoblotting were performed as described in the “Protein Extraction and Immunoblotting” section. Total protein lysates were subjected to SDS-PAGE and immunoblotting using specific antibodies against phosphorylated AKT (Ser473 and Thr308) and total AKT (Supplementary Table 1).

Serum biochemical analysis

To evaluate the potential impact of siRNAs/LNP-Gal on hepatic and renal function, blood samples were collected from the orbital sinus on day 6 after the last administration. Serum was obtained by centrifugation at 3,000 × g for 30 min at 4 °C. ALT (alanine aminotransferase), AST (aspartate aminotransferase), BUN (blood urea nitrogen), and Cr (creatinine) levels were measured using commercial automated biochemical assay kits (Nanjing Jiancheng Bioengineering Institute, China) in accordance with the manufacturer’s protocols. Colorimetric readings were recorded using a BioTek multifunctional microplate reader (BioTek Instruments, USA). All measurements were performed in technical replicates.

Hematoxylin and Eosin (H&E) staining

Tissue samples were fixed in 4% paraformaldehyde (Servicebio, G1101) for 24–48 h, dehydrated through a graded ethanol series (70%, 80%, 95%, and 100%), cleared in xylene, and embedded in paraffin. Paraffin blocks were then sectioned into 4 μm slices using a microtome, mounted on glass slides, and baked at 60 °C for 1 h to promote adhesion and remove residual paraffin. Subsequently, sections were deparaffinized in xylene, rehydrated through a descending ethanol series (100%, 95%, 80%, 70%), and stained with hematoxylin for 5–10 min. After rinsing in running water, sections were differentiated in 1% hydrochloric acid-ethanol for 10–30 s and blued under alkaline conditions. Eosin solution (LEAGENE, DH0006) was then applied for 1–3 min, followed by dehydration through graded ethanol, clearing in xylene, and mounting with neutral resin. The stained sections were examined and imaged under a light microscope (OLYMPUS, VS200) for histological and morphological evaluation.

Mouse primary hepatocytes isolation and culture

Primary mouse hepatocytes were isolated using a two-step liver perfusion method as previously described [42]. Briefly, mice were anesthetized and the portal vein was exposed for cannulation. The liver was perfused with pre-warmed (37 °C) liver perfusion buffer followed by digest buffer. The liver was excised, minced under sterile conditions, and gently pipetted to release hepatocytes, which were re-suspended in DMEM medium supplemented with 10% FBS, and 1% penicillin-streptomycin (Gibco, 15140122). Cells were seeded onto collagen (R&D Systems, 3440-100-01)-coated 6 or 12 well plates at a density of 1.0–2.0 × 10⁶ cells/well. Cultures were maintained at 37 °C in a humidified 5% CO₂ incubator. After 6 h, the medium was replaced to remove unattached cells, and hepatocytes were cultured overnight before being treated with siRNAs or siRNAs/LNP-Gal, as specified in the analyses.

Hepatic glucose production assay

Mouse primary hepatocytes were seeded into six-well plates at a density of 1 × 10⁶ cells/well. Cells were then treated with siNC/LNP-Gal or siSam68/LNP-Gal for 24 h. After attachment, cells were then washed three times with PBS and incubated for 4 h in glucose production buffer containing phenol red-free, glucose-free DMEM (Gibco, A14430-01) supplemented with 20 mM sodium lactate, 2 mM sodium pyruvate, and 1 mM glycerol. Subsequently, cells were treated with 100 nM glucagon, or 10 µM forskolin, or 100 µM Bt2-cAMP. At the end of the incubation, 0.5 mL of culture medium was collected, and glucose concentration was measured using a glucose colorimetric assay kit (Eton Bioscience, Inc.) according to the manufacturer’s protocol. Glucose levels were normalized to total protein content.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from tissues or cells using TRIzol reagent (Invitrogen, 15596026CN) according to the manufacturer’s instructions. RNA purity and concentration were assessed using a BioTek microplate reader (BioTek Instruments, USA). cDNA was synthesized from 0.5 µg of total RNA using a reverse transcription kit (Vazyme, R433-01). qRT-PCR was performed on an ABI 3000 Real-Time PCR system (Applied Biosystems) with SYBR Green Master Mix (Vazyme, Q712-02/03). Each sample was analyzed in triplicate, and relative gene expression levels were calculated using the 2^-ΔΔCt method, with β-actin as the internal control. Primer sequences for target genes are listed in Supplementary Table 2.

Protein extraction and immunoblotting

Cells or frozen liver tissues were lysed in RIPA buffer for protein analysis. Approximately 1 × 10⁷ cells or 100 mg of liver tissue were homogenized in 0.5 mL RIPA buffer (50 mM Tris-HCl, pH 8.0; 150 mM NaCl; 1% Triton X-100; 0.1% SDS; 1 mM EDTA) supplemented with protease inhibitor cocktail (Aqlabtech, AQ551) and phosphatase inhibitor cocktail (Aqlabtech, AQ552). Samples were processed using a tissue homogenizer at 6 m/s for three cycles of 10 s with cooling intervals to ensure complete disruption, followed by incubation on ice for 90 min. Lysates were centrifuged at 13,000 × g for 10 min at 4 °C, and the supernatants were collected for protein quantification and immunoblotting. Protein concentrations were determined using the BCA assay (Sevenbio, SW201-02).

Equal amounts of protein were denatured at 95 °C for 7 min, separated by SDS-PAGE, and transferred onto PVDF membranes (Bio-Rad, 1620177). Membranes were blocked with 5% non-fat milk in Tris-Buffered Saline (TBS) for 1 h at room temperature, and then incubated overnight at 4 °C with primary antibodies diluted in 0.1% Tween-20 in TBS (TBST) containing 3% BSA. After washing three times with TBST, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h at room temperature, followed by three washes. Protein signals were visualized using enhanced chemiluminescence (Thermo Fisher Scientific, A38555) and captured with a Bio-Rad ChemiDoc imaging system. Details of all primary and secondary antibodies used in experiments are provided in Supplementary Table 1.

Statistical analysis

Data are shown as mean ± SD. Differences between two groups were assessed using unpaired two-tailed Student’s t-tests, whereas comparisons involving three or more groups were analyzed by one-way or two-way ANOVA, depending on the number of independent variables. Statistical analyses were performed using GraphPad Prism 8. A p-value < 0.05 was considered statistically significant.

Results

Construction and characterization of targeted LNPs

LNPs were assembled using a microfluidic mixing technique, with a classical ionizable lipid composition of 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC), (6Z,9Z,28Z,31Z)-Heptatriaconta-6,9,28,31-tetraen-19-yl 4-(dimethylamino) butanoate (DLin-MC3-DMA), cholesterol, and 1,2-Dimyristoyl-sn-glycerol–polyethylene glycol (DMG-PEG), encapsulating scrambled siRNA [43]. The effect of different lipid-to-siRNA ratios was evaluated using the nitrogen-to-phosphate (N/P) molar ratio between DLin-MC3-DMA and siRNA. Complete siRNA encapsulation was achieved when N/P was ≥ 8, whereas higher ratios did not further improve encapsulation efficiency but instead affected particle size and zeta potential (Fig. 1A, S1A). Therefore, an N/P ratio of 8 was selected for subsequent LNP formulations. The resulting LNPs had an average hydrodynamic diameter of ~ 100 nm and exhibited a uniform, spherical morphology under transmission electron microscopy (TEM) (Fig. 1B). To enable surface ligand conjugation, DMG-PEG was partially replaced with maleimide-presenting PEG-lipid (DMG-PEG-Mal) at varying molar proportions to produce LNP-Mal. When the replacement ratio was less than 30%, no significant changes in particle size, zeta potential, or siRNA leakage were detected (Fig. 1C, S1B). To covalently install targeting ligands onto the surface PEG–maleimide, we synthesized six thiol-terminated ligands with reported hepatic targeting potential including Gly-SH, Fa-SH, Man-SH, Gal-SH, Ca-SH, and Ret-SH (Fig. S1C). Their structures were confirmed by ¹H-NMR spectroscopy.

Fig. 1.

Construction and characterization of targeted LNPs. (A) Particle diameters and zeta potentials of LNP with different N/P ratios (n = 3). (B) DLS and TEM (inset) characterization of LNP (Scale bar, 100 nm). (C) Particle diameters and zeta potentials of LNP which DMG-PEG was replaced by different ratio of DMG-PEG-Mal (n = 3). (D) Modification ratio of different targeting ligands in LNP-Mal (n = 3). (E) Zeta potentials of LNP and targeted LNP in buffer with different pH (n = 3). (F) Encapsulation efficiency of siRNA in LNP and targeted LNP (n = 3). (G-L) DLS and TEM (inset) characterization of LNP-Gly (G), LNP-Fa (H), LNP-Man (I), LNP-Gal (J), LNP-Ca (K), LNP-Ret (L) (Scale bars, 100 nm). Data are presented as mean ± SD. “n” denotes independent LNP samples

The targeted ligands were conjugated to LNP-Mal via a thiol–maleimide Michael addition reaction between the maleimide and thiol groups. The resulting LNP-Gly, LNP-Fa, LNP-Man, and LNP-Gal formulations showed conjugation efficiencies of ~ 80%, whereas LNP-Ca and LNP-Ret exhibited lower efficiencies of ~ 50% and ~ 30%, respectively (Fig. 1D). Zeta potential analysis across pH values indicated that all targeted LNPs retained acidic pH-triggered charge reversal, suggesting that ligand decoration did not compromise the pH-responsive behavior associated with endosomal escape (Fig. 1E). Moreover, ligand conjugation did not cause appreciable siRNA leakage, and all formulations maintained encapsulation efficiencies of ~ 90% (Fig. 1F, S1D). Targeted LNPs displayed a moderate size increase (~ 20–30 nm) relative to unmodified LNPs while preserving spherical morphology, as confirmed by dynamic light scattering (DLS) and TEM (Fig. 1G-L, S1E). All formulations remained stable in size over an 8-day storage period, indicating favorable colloidal stability (Fig. S1F).

Delivery efficiency and mechanistic evaluation of targeted LNPs

The delivery efficiency of different LNP formulations was first evaluated in vitro using AML12 (hepatocytes), RAW264.7 (macrophages), and mouse hepatic stellate cells (mHSCs) as model cell lines. LNPs and targeted LNPs encapsulating 6-carboxyfluorescein (FAM)-labeled scrambled siRNA were incubated with these cells, and their uptake was assessed using confocal microscopy and flow cytometry. After 8 h, all targeted LNPs showed significantly enhanced uptake in AML12 cells compared to unmodified LNPs, with LNP-Gal exhibiting the highest cellular uptake (Fig. 2A-B). In contrast, RAW264.7 and mHSCs displayed substantially lower uptake of the targeted LNPs than that observed in hepatocytes, indicating that ligand decoration preferentially increased hepatocyte uptake relative to non-parenchymal liver cell types (Fig. 2A, C-D). Next, Cy5-labeled siRNA-loaded LNPs and targeted LNPs were intravenously injected into C57BL/6 mice. In vivo imaging showed that fluorescence intensity in the LNP-Man and LNP-Gal groups was significantly higher than in other groups from 1 to 24 h post-injection (Fig. 2E). At 24 h, ex vivo imaging of major organs revealed that all LNP formulations were predominantly distributed in the liver, with the LNP-Gal group exhibiting the strongest hepatic fluorescence signal (Fig. 2F-G). Analysis of the parenchymal region in liver sections from mice treated with FAM-siRNA-loaded LNPs further confirmed maximal hepatocyte-associated fluorescence for LNP-Gal (Fig. 2H). Together, these data identify LNP-Gal as the most efficient formulation for hepatocyte-selective siRNA delivery among the ligands tested.

Fig. 2.

Delivery efficiency and mechanistic evaluation of targeted LNPs. (A) Representative confocal image of AML12, RAW264.7 and mHSC treated with LNPs and targeted LNPs loaded with FAM-labeled siRNA (Scale bar, 50 μm). (B-D) Cellular uptake analysis of LNPs and targeted LNPs in AML12 (B), RAW264.7 (C) and mHSC (D) by flow cytometry analysis (n = 3). (E) Representative in vivo image of the mice intravenously injected with Cy5-labeled siRNA-loaded LNPs and targeted LNPs over time. (F) Representative ex vivo image of major organs collected from the mice at 24 h post-injection of LNPs and targeted LNPs (H, heart; Lu, lung; Li, liver; K, kidney; S, spleen). (G) Cy5 fluorescence intensity in major organs collected from the mice at 24 h post-injection of LNPs and targeted LNPs (n = 3). (H) Representative confocal image of parenchymal region in liver section from the mice at 24 h post-injection of LNPs and targeted LNPs loaded with FAM-labeled siRNA (100 μm, 20×). (I) Isoelectric point distributions of protein corona around LNPs or LNP-Gal with an abundance above 0.1%. (J) Proteomic analysis of protein corona around LNP or LNP-Gal with an abundance above 0.1% were categorized based on their biological functions. (K) Representative protein abundance and fold-change of corona proteins around LNP-Gal compared with LNP. (L) Representative confocal image of AML12 cells treated by LNP-Gal loaded with FAM-labeled siRNA (10 μm, 63×). (M-N) Profiles of the fluorescence intensity distribution of siRNA (green) and endosomes (red) along the cyan indicator lines in the magnified images (L) in 1 h (M) or 4 h (N). (O) Representative cryo-TEM image of siSam68/LNP-Gal (100 nm). Data are presented as mean ± SD. Significance levels are indicated as ^ns P ≥ 0.05, *P < 0.05, ***P < 0.001, ****P < 0.0001 (One-way ANOVA followed by Tukey’s multiple comparisons test). “n” denotes biologically independent cell samples

In addition to asialoglycoprotein receptor (ASGPR)-mediated hepatocyte targeting, we performed proteomic analyses of the plasma-derived protein corona formed on LNPs and targeted LNPs. Molecular weight profiling showed significantly increased adsorption of approximately 60–80 kDa proteins on LNP-Man and LNP-Gal, the formulations with higher hepatocyte targeting (Fig. S2A). Isoelectric focusing analysis indicated that LNP-Gal preferentially adsorbed more acidic proteins (pI < 5), including albumin, transferrin, and α1-acid glycoprotein, a corona composition that may attenuate opsonization and thereby reduce RES-mediated clearance (Fig. 2I and Fig. S2B). Functional analysis showed that, compared to unmodified LNPs, the protein corona of LNP-Man, LNP-Gal, and LNP-Ret exhibited a significantly reduced relative abundance of apolipoproteins. Additionally, LNP-Gal showed decreased levels of immunoglobulins and complement proteins on its surface. These findings suggest that hepatocyte delivery of LNP-Gal is less dependent on endogenous apolipoprotein-mediated pathways than that of unmodified LNPs (Fig. 2J and Fig. S2C). Overall, the corona signature of LNP-Gal suggests a compositional shift that may reduce phagocytic recognition and favor hepatocyte-biased delivery (Fig. 2K).

To evaluate endosomal escape, we performed Förster resonance energy transfer (FRET)-based liposome fusion assays. At neutral pH, FRET signal remained unchanged, indicating minimal membrane fusion. By contrast, under acidic conditions that mimic the endosomal environment, we observed a marked reduction in FRET signal, consistent with pH-dependent membrane fusion between LNPs and vesicle membranes (Fig. S2D-E). Confocal microscopy of AML12 cells treated with FAM-siRNA-loaded LNP-Gal showed that siRNA signals initially co-localized with endosomal markers at 1 h, but dispersed into the cytoplasm by 4 h, indicating that LNP-Gal retained the endosomal escape competence of unmodified LNPs (Fig. 2L-N). Finally, siRNA targeting Sam68 (siSam68) was loaded into LNPs and LNP-Gal. Cryo-electron microscopy confirmed the structural integrity of the siSam68/LNP-Gal particles, revealing a spherical and monodisperse morphology (Fig. 2O). At the cellular level, both siSam68/LNP and siSam68/LNP-Gal exhibited no cytotoxicity to hepatocytes across a range of concentrations (Fig. S2F-G).

Selective hepatic Sam68 silencing by siSam68/LNP-Gal treatment

To screen the most effective Sam68-targeting sequence, the silencing efficiency of three candidate siSam68 constructs was first evaluated in mouse primary hepatocytes after Lipofectamine 3000-mediated transfection, with scrambled siRNA (siNC) serving as the control. Compared with siNC group, transfection with siSam68-1, siSam68-2, and siSam68-3 reduced Sam68 mRNA expression by 48%, 74%, and 70%, respectively, at 24 h (Fig. 3A), and Sam68 protein expression was reduced by 21%, 32%, and 67% at 48 h (Fig. 3B), respectively. Thus, siSam68-3 was selected for subsequent experiments owing to its superior overall silencing efficiency.

Fig. 3.

Validation of efficient and selective silencing of hepatocyte Sam68 by siSam68/LNP in vitro and in vivo. (A–D) Sam68 silencing in mouse primary hepatocytes. siNC or siSam68-1, -2, and − 3 were transfected into primary hepatocytes, respectively. Sam68 mRNA levels were quantified at 24 h by qRT-PCR (A, n = 3), and protein levels were assessed at 48 h by WB (B, left, image; right, quantification; n = 3). Primary hepatocytes were treated with siNC/LNP-Gal or siSam68/LNP-Gal (100 nM). Sam68 mRNA levels were measured at day 1, 3, 5, and 7 by qRT-PCR (C, n = 4–6), and protein levels were determined at day 2, 4, and 6 by WB (D, left, image; right, quantification; n = 4). (E–F) siNC/LNP-Gal or siSam68/LNP-Gal (1 mg/kg) were administered to WT mice by tail vein injection. Hepatic Sam68 mRNA and protein levels were measured at 2, 6, and 12 days post-injection by qRT-PCR (E, n = 3) and WB (F, left, image; right, quantification; n = 4). (G) Serum ALT, AST, BUN, Cr levels were analyzed at day 6 post-injection (n = 5). Data are presented as mean ± SD. Significance levels are indicated as *P < 0.05, ***P < 0.001, and ****P < 0.0001 (A, right panel of B, C and G: One-way ANOVA followed by Tukey’s multiple comparisons test; E, right panel of D, F: Unpaired two-tailed Student’s t-tests). “n” denotes biologically independent primary hepatocyte samples or liver tissues

Next, we encapsulated siSam68-3 and siNC separately into LNP-Gal and knockdown efficiency was validated in mouse primary hepatocytes. Treatment with siSam68/LNP-Gal reduced Sam68 mRNA levels by 49%, 39%, 32%, and 0% on day 1, 3, 5, and 7 respectively (Fig. 3C), and protein levels declined by 50%, 47%, and 19% on day 2, 4, and 6, respectively (Fig. 3D), indicating effective Sam68 silencing for up to six days. Subsequently, siSam68/LNP-Gal and scrambled siRNA-loaded lipid nanoparticles (siNC/LNP-Gal) were administered to WT mice by tail vein injection. On days 2 and 6 after a single administration of siSam68/LNP-Gal, hepatic Sam68 mRNA levels decreased by approximately 47% and 44%, while protein levels reduced by 26% and 53%, respectively, compared with the control group. The gene-silencing effect persisted for nearly one week before returning to baseline by day 12, indicating durable in vivo efficacy (Fig. 3E-F). As expected, Sam68 mRNA and protein expression remained unchanged in heart, lung, spleen, kidney, skeletal muscle, and adipose tissues following siSam68/LNP-Gal delivery at day 6 (Fig. S3A-B). Furthermore, delivery of either siSam68/LNP-Gal or siNC/LNP-Gal at day 6 did not alter serum ALT, AST, BUN or Cr levels (Fig. 3G), or cause histological abnormalities in the liver and kidney tissues, as shown by H&E staining (Fig. S3C-D). Altogether, these findings demonstrate that siSam68/LNP-Gal achieves efficient and selective silencing of hepatocyte Sam68 with an excellent safety profile both in vitro and in vivo.

siSam68/LNP-Gal-mediated inhibition of gluconeogenesis suppresses glucose production in hepatocytes

Next, we determined the effect of siSam68/LNP-Gal treatment on gluconeogenic gene expression and glucose production in primary mouse hepatocytes. In line with our previous observations [20], siSam68/LNP-Gal treatment was accompanied by a significant reduction in the expression of key gluconeogenic genes, including PGC-1α, PEPCK, and G6Pase, and markedly suppressed glucose production in response to glucagon (Fig. 4A-B), forskolin (Fig. S4A-B) or Bt2-cAMP stimulation (Fig. S4C-D). Western blot analysis further confirmed the reduced protein levels of PGC-1α, PEPCK, G6Pase, FOXO1, and HNF4α in the siSam68/LNP-Gal group compared to the siNC/LNP-Gal group (Fig. 4C). Notably, FOXO1 and HNF4α are well-established transcription factors that cooperate synergistically with PGC-1α to drive gluconeogenic gene transcription, forming a critical regulatory network essential for hepatic glucose homeostasis [44]. Consequently, their simultaneous downregulation by siSam68/LNP-Gal treatment disrupts this core transcriptional machinery, impairing the coordinated activation required for efficient gluconeogenesis. Both total CREB and phosphorylated CREB protein levels were comparable between siSam68/LNP-Gal and siNC/LNP-Gal groups (Fig. 4C), indicating that the effects of siSam68/LNP-Gal are independent of the canonical CREB upstream pathway. Remarkably, the efficacy of siSam68/LNP-Gal in suppressing gluconeogenic gene expression and glucose production was greater than that of low-dose metformin and comparable to high-dose metformin (Fig. 4A-C, S4A-D). This suggests that liver-specific silencing of Sam68 could rival traditional drug interventions in regulating hepatic glucose output. Collectively, these results demonstrate that siSam68/LNP-Gal treatment efficiently silences Sam68 and suppresses gluconeogenesis in mouse primary hepatocytes. Moreover, the coordinated suppression of key transcriptional regulators, gluconeogenic enzymes, and functional glucose production provides compelling evidence that Sam68 is a critical node in hepatic gluconeogenic regulation and a promising therapeutic target for metabolic disorders associated with excessive hepatic glucose production.

Fig. 4.

siSam68/LNP-Gal treatment suppresses glucose production by inhibiting gluconeogenesis in primary hepatocytes. (A–C) Mouse primary hepatocytes were treated with siNC/LNP-Gal (100 nM), siSam68/LNP-Gal (100 nM), or metformin (50 or 500 µM). The mRNA levels of PGC-1α, PEPCK, and G6Pase were quantified by qRT-PCR after 24 h of treatment and subsequent glucagon stimulation for 1, 2, 3 h (A, n = 3). Glucose production was measured at 24 h treatment followed by glucagon stimulation with 4 h (B, n = 3). Protein levels of Sam68, p-CREB (phosphorylated), CREB, PGC-1α, PEPCK, G6Pase, FOXO1, and HNF4α were assessed by WB after 48 h treatment (C, left, image; right, quantification; n = 3). Data are presented as mean ± SD. Significance levels are indicated as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (A: Two-way ANOVA followed by Tukey’s multiple comparisons test; B and right panel of C: One-way ANOVA followed by Tukey’s multiple comparisons test). “n” denotes biologically independent primary hepatocyte samples

siSam68/LNP-Gal improved glycemia by suppressing hepatic gluconeogenesis in HFD/STZ-induced diabetic mice

Subsequently, the therapeutic efficacy of siSam68/LNP-Gal in mitigating hyperglycemia was evaluated in diabetic animal models. HFD/STZ diabetic mice received either siNC/LNP-Gal, siSam68/LNP-Gal, low-dose (50 mg/kg) or high-dose metformin (150 mg/kg). Blood glucose levels were measured before (Fig. S5A) and after 4 weeks of treatment across all groups. While metformin treatment greatly improved hyperglycemia in diabetic mice, the siSam68/LNP-Gal group showed much lower blood glucose than siNC/LNP-Gal group under both fed and fasting conditions (Fig. 5A-B). Compared with the siNC/LNP-Gal group, mice administered siSam68/LNP-Gal demonstrated a significant reduction in hepatic Sam68 mRNA and protein expression accompanied by a marked decrease in CRTC2 protein level without altering its mRNA level (Fig. 5C and E). Consistently, siSam68/LNP-Gal treatment significantly downregulated both mRNA and protein levels of hepatic gluconeogenic genes (PGC-1α, PEPCK, G6Pase, FOXO1, and HNF4α) under both fed and fasting conditions (Fig. 5D-E, S5B). Remarkably, both siSam68/LNP-Gal and high-dose metformin treatments led to a dramatic reduction in CRTC2 protein expression (Fig. 5E). In a positive control group, mice orally administered metformin also displayed downregulated hepatic gluconeogenic gene expression as well (Fig. 5D-E, S5B). Pyruvate tolerance tests (PTT) and glucagon challenge tests (GcTT) revealed effective reduction of substrate- and hormone-induced hepatic gluconeogenesis and glucose production in siSam68/LNP-Gal group, leading to significantly lower blood glucose levels (Fig. 5F-G). Moreover, compared to siNC/LNP-Gal group, siSam68/LNP-Gal group showed enhanced glucose clearance and insulin sensitivity, as determined by glucose tolerance tests (GTT) (Fig. 5H) and insulin tolerance tests (ITT) (Fig. S5C). Consistently, the levels of phosphorylated AKT (at Ser-473 and Thr-308) in the liver were higher after the insulin administration in the siSam68/LNP-Gal group than in siNC/LNP-Gal group (Fig. S5D). Remarkably, siSam68/LNP-Gal treatment exhibited significant efficacy in improving glucose homeostasis compared to low and high dosage of metformin treatment, respectively (Fig. 5B and F-H, S5C). Taken together, siSam68/LNP-Gal administered in HFD/STZ diabetic mice effectively silenced hepatic Sam68 expression and mitigated hyperglycemia via suppressing hepatic gluconeogenesis and enhancing insulin signaling.

Fig. 5.

siSam68/LNP-Gal administration in HFD/STZ diabetic mice significantly improves blood glucose homeostasis via suppressing hepatic gluconeogenesis. Diabetic mice were intravenously injected with siNC/LNP-Gal (1 mg/kg) or siSam68/LNP-Gal (1 mg/kg) every 6 days, or received metformin (50 or 150 mg/kg) via daily oral administration. All treatments lasted for 4 weeks. (A–B) Blood glucose levels were determined under fed (A) or 16 h fasting (B) condition (n = 8). (C) Hepatic mRNA expression of Sam68 and CRTC2 was assessed by qRT-PCR under fed or 16 h fasting condition (n = 6). (D) Hepatic mRNA expression of PGC-1α, PEPCK, and G6Pase was assessed by qRT-PCR under fed or 16 h fasting condition (n = 6). (E) Hepatic protein expression of Sam68, CRTC2, PGC1α, G6Pase, PEPCK, FOXO1, and HNF4α was assessed by WB under fed or 16 h fasting condition. (F–H) Blood glucose was measured at 0, 15, 30, 45, 60, 90, and 120 min after injection of pyruvate (F, 16 h fasting, n = 8), glucagon (G, 6 h fasting, n = 8) or glucose (H, 16 h fasting, n = 8). AUC were analyzed from the corresponding curves (right panel of F-H). Data are presented as mean ± SD. Significance levels are indicated as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (A-D and right panel of F-H: One-way ANOVA followed by Dunnett’s multiple comparisons test). “n” denotes biologically independent blood samples or liver tissues

siSam68/LNP-Gal mitigated hyperglycemia via hepatic gluconeogenesis suppression in db/db mice

Next, we evaluated the effect of siSam68/LNP-Gal administration on glucose metabolism in db/db mice. After four weeks of treatment, siSam68/LNP-Gal markedly reduced Sam68 mRNA levels specifically in the liver under both fed and fasting conditions, with no significant effects observed in other organs under the fed state (Fig. 6A, Fig. S6B). Meanwhile, metformin treatment, at either low or high dose, did not affect Sam68 expression (Fig. 6A).

Fig. 6.

siSam68/LNP-Gal reduces blood glucose levels and suppresses gluconeogenic gene expression in db/db mice. db/db diabetic mice received intravenous injections of siNC/LNP-Gal (1 mg/kg) or siSam68/LNP-Gal (1 mg/kg) once every 6 days, or received metformin (50 or 150 mg/kg) via daily oral administration. All treatments lasted for 4 weeks. (A) Hepatic mRNA expression of Sam68 under fed or 16 h fasting conditions, quantified by qRT-PCR (n = 6). (B) Blood glucose levels were determined under fed (Left) or 16 h fasting (Right) conditions (n = 8). (C–E) Glucose homeostasis tests: pyruvate tolerance test (C, PTT, 16 h fasting; n = 8), glucagon challenge test (D, GcTT, 6 h fasting; n = 8), and glucose tolerance test (E, GTT, 16 h fasting; n = 8). Blood glucose was measured at 0, 15, 30, 60, 90, and 120 min after injection of the respective agents, and AUC values were calculated accordingly. (F–G) The mRNA (F, n = 6) and protein levels (G) of Sam68, CRTC2, PGC1α, G6Pase, PEPCK, FOXO1, and HNF4α were determined by qRT-PCR and WB analysis in liver tissues collected under the same conditions. Data are presented as mean ± SD. Significance levels are indicated as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (A, B, F and right panel of C-E: One-way ANOVA followed by Dunnett’s multiple comparisons test). “n” denotes biologically independent blood samples or liver tissues

Blood glucose levels were comparable among all groups before treatment (Fig. S6A). siSam68/LNP-Gal treatment in db/db mice led to marked improvements in systemic glucose homeostasis, as reflected by reduced blood glucose levels under both fed and fasting conditions (Fig. 6B), improved pyruvate (Fig. 6C), glucagon (Fig. 6D), and glucose (Fig. 6E) tolerances. Concomitantly, both mRNA and protein levels of key gluconeogenic regulators, including PGC-1α, PEPCK, G6Pase, FOXO1 and HNF4α, were significantly decreased under both fed and fasting conditions following Sam68 silencing (Fig. 6F-G; Fig. S6C). Both siSam68/LNP-Gal and high-dose metformin treatment substantially reduced CRTC2 protein abundance without affecting its mRNA expression (Fig. 6F-G). Accompanied by improved glucose homeostasis, siSam68/LNP-Gal treatment significantly enhanced systemic and hepatic insulin sensitivity, as indicated by improved insulin tolerance (Fig. S6D) and increased protein expression of phosphorylated AKT (at Ser-473 and Thr-308) in the liver after insulin administration (Fig. S6E).

Collectively, these data reveal that siSam68/LNP-Gal treatment in diabetic mice greatly improves systemic glucose homeostasis by suppressing the expression of hepatic gluconeogenic genes, an effect attributable specifically to hepatic Sam68 silencing. Notably, its therapeutic efficacy against hyperglycemia is equivalent to high-dose metformin treatment (Fig. 6B-E; Fig. S6D).

Discussion

Dysregulated hepatic gluconeogenesis and impaired insulin sensitivity are central pathophysiological drivers of hyperglycemia in T2DM. We previously demonstrated that liver-specific knockdown of Sam68 in diabetic mice ameliorates hyperglycemia by suppressing hepatic glucose production and enhancing insulin sensitivity [19, 20], suggesting that Sam68 silencing may represent a potential therapeutic strategy for T2DM. However, despite its therapeutic promise, siRNA-based therapies for diabetes remain challenging due to limitations in delivery specificity, stability, cellular uptake, and biosafety.

LNPs have emerged as an efficient and clinically validated platform for nucleic acid delivery [45]. Nevertheless, systemically administered LNPs do not inherently achieve hepatocyte-selective delivery, because a substantial fraction is preferentially captured by hepatic non-parenchymal cells and cleared by the RES. To address this gap, we employed a mild and efficient ligand-conjugation strategy to generate six ligand-decorated LNP formulations through head-to-head comparisons in vitro and in vivo, and identified LNP-Gal as the most effective platform for hepatocyte-biased siRNA delivery. We aimed to develop a hepatocyte-targeted LNP-based siRNA delivery system for Sam68 silencing. To this end, galactose decoration offered a practical strategy to enhance hepatocyte uptake while maintaining the flexibility and optimization potential of the LNP platform [46, 47]. This improved performance is likely attributable to galactose decoration which promotes ASGPR-mediated hepatocyte uptake [45], and also reshapes the plasma protein corona in a manner that may reduce opsonization and phagocytic clearance by Kupffer cells and other RES components [48], thereby favoring hepatocyte-biased intrahepatic distribution. Importantly, LNP-Gal retained key functional properties required for RNA delivery, including pH-responsive behavior and endosomal escape competence, supporting its suitability for therapeutic siRNA delivery.

Subsequent evaluation in mouse primary hepatocytes confirmed the functional efficacy of siSam68/LNP-Gal, demonstrating significant downregulation of both Sam68 mRNA and protein levels without associated cytotoxicity. Consistently, administration of siSam68/LNP-Gal effectively reduced Sam68 expression in the livers of HFD/STZ-induced diabetic mice and db/db mice, without causing hepatotoxicity, nephrotoxicity, systemic inflammation, or histological abnormalities. These findings demonstrate the in vivo safety and tolerability of this delivery platform.

Suppressing excessive hepatic glucose production is a cornerstone of T2DM management, as exemplified by standard-of-care therapies like metformin and GLP-1 receptor agonists, which modulate gluconeogenesis and insulin sensitivity [49–51]. In line with this therapeutic framework, hepatocyte-directed delivery of siSam68 via LNP-Gal directly suppressed the core gluconeogenic program, markedly reducing the expression of PGC-1α, PEPCK, and G6Pase at both the mRNA and protein levels. Concurrently, siSam68/LNP-Gal treatment enhanced hepatic insulin signaling, as evidenced by increased AKT phosphorylation. Thus, these findings consistently support hepatocyte-directed Sam68 silencing as a potential therapeutic strategy grounded in a clear mechanistic basis for T2DM.

Interestingly, additional gluconeogenic transcription factors, HNF4α and FOXO1, which remained largely unchanged in liver-specific Sam68 knockout mice, were also downregulated following siSam68/LNP-Gal administration [20, 52, 53]. This suggests that LNP-based siRNA delivery may engage additional regulatory pathways beyond the direct Sam68-CRTC2 interaction. One possible explanation is that hepatic Sam68 silencing triggers immediate transcriptional changes that are compensated for during development in constitutive knockout models [54]. However, further studies are necessary to clarify how siSam68/LNP-Gal reshapes the broader transcriptional and post-translational landscape of hepatic metabolism.

Our data show that siSam68/LNP-Gal exceeded the efficacy of low-dose metformin and rivaled that of a high-dose, both in suppressing gluconeogenesis in primary hepatocytes and improving blood glucose homeostasis in diabetic mouse models. Although metformin did not affect Sam68 expression, both siSam68/LNP-Gal and high-dose metformin treatments substantially reduced CRTC2 protein levels, suggesting parallel pathways that converge on CRTC2 suppression.

In addition, we observed that hepatic Sam68 silencing persisted for approximately six days after a single administration of siSam68/LNP-Gal duration that is consistent with the transient pharmacodynamic profile reported for many LNP-mediated siRNA therapeutics and sufficient to induce measurable metabolic effects [55]. Clinically, intravenous insulin infusion is routinely used for acute glycemic control, demonstrating that this route can be effective in specific settings [31, 56]. Further optimization of formulation design and delivery route may enhance the practicality and translational potential of this RNAi-based therapy. Nevertheless, clinical translation will require rigorous safety and toxicology evaluation in large-animal models, comprehensive immunogenicity and pharmacokinetic profiling, and optimization of dosing regimens to achieve sustained therapeutic efficacy.

Conclusion

This study further validates hepatic Sam68 as a key driver of pathological gluconeogenesis in T2DM and establishes a hepatocyte-targeted lipid nanoparticle platform, siSam68/LNP-Gal, as an effective strategy for its silencing in vitro and in vivo. By leveraging galactose-decorated LNPs to achieve hepatocyte-selective delivery, we overcame the limitations of conventional LNP formulations, including off-target sequestration by non-parenchymal liver cells. The resulting formulation achieved potent and specific Sam68 knockdown in hepatocytes, leading to suppression of the CREB/CRTC2 transcriptional axis, reduced gluconeogenic gene expression, and alleviated hyperglycemia effect and insulin resistance in two independent preclinical T2DM models. Collectively, these results support the broader potential of hepatocyte-targeted RNAi via siSam68/LNP-Gal in T2DM and its associated metabolic diseases, and highlight the versatility of this platform for future therapeutic optimization.

Author contributions

N. W., X. C. and J. L. performed most of experiments, acquired and analyzed the data, generated the figures, and made a draft. S. A. performed experiments and analyzed the data. Y. G., A. Y., Y. G., and J. Y., C. F. assisted in experiments and data analysis. E. E. N. revised the manuscript. J. X., L. T. and H. P. made intellectual contributions and assisted in data interpretation. A. Q. and S. S. conceptualized the study, interpreted data, wrote, and edited the manuscript.

Funding

This work was supported by research grants from the National Natural Science Foundation of China (82270925, 82470892 to A.Q., 82460821 to L.T.), Innovative Drug Research and Development National Science and Technology Major Project (2025ZD1802602 to S.S.), State Key Laboratory of Drug Research (SKLDR-2025-KF-07 to A.Q.), Lingang Laboratory (LGL-2612-22 to A.Q.), Guangdong Basic and Applied Basic Research Foundation, China (2025A1515010594 to S.S.), the High-level New R&D Institute of the Department of Science and Technology of Guangdong Province (2019B090904008 to A.Q.), the High-level Innovative Research Institute of the Department of Science and Technology of Guangdong Province (2021B0909050003 to A.Q. and S.S.), the Zhongshan Science and Technology Bureau (CXTD2023009 to A.Q.), the Traditional Chinese medicine inheritance innovation development research project of Zhongshan City (2024B3002 to A.Q.).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Competing interests

The authors declare no competing interests.