Organ-selective delivery is the result of a series of complex interactions between LNPs and in vivo environments17. The selective adsorption of plasma proteins onto LNPs has been believed to play a key role in determining the organ targeting of LNPs17,30,31,32,33. In this work, both in vitro and in vivo experiments demonstrated that the Vtn corona and its cognate receptor, αvβ3 integrin, are both essential for the lung-selective delivery of 6R-LNP, indicating the importance to identify the corona proteins and receptors involved in the organ-selective LNP delivery. Further, using Vtn as an example corona protein with lung-targeting potential, we established a theoretical framework to analyse and understand how amino acid sequences might be used to modulate the peptide-protein affinity and, thus, the selective adsorption of key corona proteins, to enable organ-targeting delivery. On the basis of the molecular-mechanics-mediated mechanism, our AI-driven framework predicted peptide sequences with higher affinity to Vtn, which have been validated experimentally with higher efficacy of lung tropism in vivo. This suggests the potential to enable the mechanics-guided rational design of the POST code to direct the organ-selective delivery of POST-LNPs. The POST platform developed in this study, in combination with recent developments of in vivo screening methodology71, might inspire the integration of AI-based information technologies in large-scale design and screening of POST codes for a wide variety of organ- and cell-selective RNA delivery in the future.

All animal experiments were approved by the Institution Animal Care and Use Committees of Tsinghua University (approval numbers 21-SY1 and 22-SY8). Male C57BL/6J mice (6-8 weeks) and E16.5 pregnant C57BL/6J mice were purchased from Beijing Vitalstar Biotechnology. Male ICR-Tg (CAG-EGFP) mice (6-8 weeks) were obtained from Cyagen Biosciences. Male and female B6.129(Cg)-Gt(ROSA)26Sor and Ai14D (B6.Cg-Gt(ROSA)26Sor, 007914) mice (6-8 weeks) were obtained from the Jackson Laboratory and housed in the Tsinghua University animal facility.

Raw materials for the synthesis of peptides, including the Wang resin and Fmoc-linked amino acid derivatives, were purchased from Tanshtech. MC3, SM-102, ALC-0315, DOPC and DOPE were purchased from SINOPEG. LP01, HSPC and DSPE-PEG2000 were purchased from Tanshtech. DSPC and ESM were purchased from Avanti Polar Lipids. Cholesterol and citrate buffer were purchased from Sigma. Cilengitide, Vtn and collagen were purchased from MedChemExpress. Luc mRNA, Itgav and Itgb3 mRNA were purchased from APExBIO (Supplementary Data 2). Cre mRNA, Cas9 mRNA and EGFP mRNA were purchased from TriLink Bio Technologies (Supplementary Data 2). EGFP-ASO (Supplementary Table 16), Cas9 protein and modified sgRNA (Supplementary Table 17) were purchased from GenScript. PBS, LPS, Cy5 and Cy7 dyes, and dialysis tubes (10 kDa) were purchased from Beijing Solarbio Science & Technology. The Quant-iT RiboGreen RNA assay kit and 4',6-diamidino-2-phenylindole dihydrochloride were purchased from Invitrogen. The Quanti-iT OliGreen ssDNA Assay Kit was purchased from Thermo Fisher Scientific. The Cas9 enzyme-linked immunosorbent assay (ELISA) kit was purchased from Cell Signaling. The Vtn ELISA kit was purchased from Abcam. Organ dissociation kits were purchased from Miltenyi Biotec. Type IV collagenase and DNase were purchased from Sigma. d-Luciferin (sodium salt) was purchased from Beijing LABLEAD. The ONE-Glo + Tox Luc Reporter assay kit was purchased from Promega. All the primary and secondary antibodies used in this study are listed in Supplementary Tables 18 and 19.

U-87 MG (HTB-14), A-498 (HTB-44) and NIH/NIH/3T3 (CRL-1658) cell lines were obtained from ATCC. The U-87 MG and A-498 cell lines were authenticated by short tandem repeat profiling by the manufacturer. NIH/3T3 cell line was authenticated via morphology by the manufacturer. The in-house authentication process was performed by cell morphology monthly. U-87 MG and A-498 cell lines were cultured in Dulbecco's modified Eagle's medium (Thermo Fisher Scientific) supplemented with 10% foetal bovine serum (FBS; Thermo Fisher Scientific) and 1% penicillin/streptomycin. The NIH/3T3 cell line was cultured in Dulbecco's modified Eagle's medium supplemented with 10% FBS, 1% penicillin/streptomycin, 1% minimum essential medium non-essential amino acids, 1% of sodium pyruvate and 1% of GlutaMAX-1 (glutamine). All cell lines were cultured at 37 °C and 5% CO.

Artificially designed peptides were synthesized by the solid-phase synthesis method on Wang resin with Fmoc-protected amino groups. The detailed synthesis procedures are described as follows. (1) Adding the first Fmoc-amino acid-OH (1.3 g), 1-hydroxybenzotriazole (1.0 g) and N,N-diisopropylcarbodiimide (1.0 ml) into Wang resin with the protection of nitrogen and continuing the reaction for more than 6 h. (2) Pumping the reaction liquid until ninhydrin detection makes the resin colourless. Wash four times alternately with dimethylformamide and dichloromethane and add 20 ml of l-piperidine to react 15 min before removing the Fmoc protection. (3) Adding the second Fmoc-amino acid-OH (1.3 g), 1-hydroxybenzotriazole (1.0 g) and N,N-diisopropylcarbodiimide (1.0 ml) into the above solution and reacting for 2 h. Remove Fmoc protection by piperidine addition, and proceed to the next step. (4) Adding the Fmoc-amino acid-OH in turn and completing the synthesis of the last amino acid using Fmoc-Cys (Trt)-OH in the peptide sequence (as the cysteine linker to conjugate the peptide to DSPE-PEG). (5) Finally, the resin is washed, drained and cracked with trifluoroacetic acid. The collected solids were purified by high-performance liquid chromatography and the molecular weight was confirmed by MS (Shimadzu GC/MS-QP2010). (6) DSPE-PEG-maleimide (1 mol) and peptides (1 mol) were dissolved in 10 ml of tetrahydrofuran under the condition of distilled acid agent (10 μl) and argon gas, and the reaction was continued for 24 h at room temperature. Finally, the above reaction solution was placed in a dialysis bag and then dialysed with methanol and ultrapure water for 3 days, before being lyophilized to obtain the final DSPE-PEG-peptide. Nuclear magnetic resonance spectroscopy (Bruker AVANCE III) was carried out to analyse the chemical structure of synthesized DSPE-PEG-peptide.

mRNA-encapsulating LNPs were fabricated by chaotic mixing in a microfluidic device, as reported before. Using such microfluidic mixers, the LNPs were formulated using the following recipe: ionizable lipids (MC3/SM-102/ALC-0315/LP01), cholesterol, phospholipids (DSPC/DOPC/DOPE/HSPC), DSPE-PEG-peptide and DSPE-PEG were dissolved in ethanol at a molar ratio of 50:38.5:10:6:0.75. The same molar ratio was used when formulating LNPs with different ionized lipids or phospholipids (Supplementary Table 7). Unless otherwise stated, the mRNA was dissolved in 10 mM of citrate buffer (pH 4) at a nitrogen-to-phosphate (N/P) molar ratio of 5.67. This index describes the ratio between the amino group (N) in the ionizable amino-lipid to the phosphate (P) groups in the backbone of mRNA. The solutions prepared above were mixed at a flow ratio of 1:3 via the microfluidic chaotic mixer system. The final solution was incubated at room temperature for 10 min before being dialysed in 1× PBS solution overnight at 4 °C. The LNPs were stored at 4 °C until use.

The mRNA-loaded, ASO-loaded and RNP-loaded LNPs were characterized by a zeta potential and particle analyser (Otsuka Electronics; ELSZ-2000) for the assessment of zeta potential, size and polydispersity index. The RiboGreen assay was conducted to measure the efficiency of Luc mRNA, EGFP mRNA, Cre mRNA, Cas9 mRNA and sgRNA encapsulation by LNPs. The concentrations of unencapsulated mRNA (C1) and total mRNA (C2) were calculated according to the correlation of the standard curve. The encapsulation efficiency (EE) is defined as EE = 1 - C1/C2.

In the encapsulation of ASO, the Quanti-iT OliGreen ssDNA Assay Kit was used for the quantification of encapsulation efficiency. The experimental procedure is similar of that the RiboGreen assay. To assess the encapsulation efficiency of the Cas9 protein, the Cas9 ELISA kit was utilized to ascertain both total and free Cas9 protein quantities. Briefly, the calculations of unencapsulated Cas9 protein concentration (D1) and total Cas9 protein concentration (D2) were performed based on the standard curve correlation. The encapsulation efficiency of Cas9 protein (EE) is defined as EE = 1 - D1/D2.

The morphology of LNPs was observed using transmission electron microscopy. First, 5-10 μl of freshly synthesized LNPs were dropped on the surface of an ultrathin carbon film. When the carbon film was dried, the samples were imaged by a transmission electron microscope (5 kV, LVEM5, Quantum Design).

LNPs that encapsulate Luc mRNA were intravenously injected into male C57BL/6J mice (6-8 weeks) at a Luc mRNA dose of 0.1 mg kg. After 3 h, d-luciferin was intraperitoneally injected at a dose of 100 mg kg. The main organs such as the liver, lung, spleen, heart and kidney were isolated 5 min after the injection of d-luciferin, before conducting ex vivo organ imaging using an IVIS Lumina system (PerkinElmer).

To evaluate the biodistribution of LNPs, we added Cy5-labelled DSPE (Cy5-DSPE:total lipids, 1:100 molar ratio) into LNPs, and the prepared LNPs were intravenously injected into male C57BL/6J mice (6-8 weeks). After 6 h, the main organs were isolated for ex vivo imaging using an IVIS Lumina system (PerkinElmer). For detecting the biodistribution of mRNA and peptides, LNPs containing Cy5-labelled mRNA and Cy7-labelled peptides were synthesized using the same N/P ratio as described above. These LNPs were intravenously injected into C57BL/6 wild-type mice. After 6 h, organs were isolated and imaged using both Cy5 and Cy7 fluorescence channels using an IVIS Lumina system (PerkinElmer).

Isolated mouse organs were embedded in an optimal cutting temperature compound and immediately frozen at -80 °C until use. Frozen tissue samples were cryosectioned at a thickness of 10 μm, before fixation in 4% paraformaldehyde for 10 min at room temperature, followed by three washes with 1× PBS. Subsequently, the sections were permeabilized with 0.3% Triton X-100 for 10 min and washed three times with 1× PBS. To block non-specific binding, the sections were incubated with 5% donkey serum for 60 min at room temperature. The blocked sections were then incubated overnight at 4 °C with a primary-antibody-targeting CD31 (Supplementary Tables 18 and 19). After washing three times with 1× PBS, the sections were incubated with a secondary antibody (donkey anti-goat IgG (H + L)), cross-adsorbed secondary antibody and Alexa Fluor 594 (1:1,000, Invitrogen) for 1.5 h at room temperature. Then, the cell nuclei were counterstained with 4',6-diamidino-2-phenylindole dihydrochloride for 20 min following three additional 1× PBS washes. Finally, the stained sections were imaged using confocal microscopy (ZEISS LSM710) and analysed by ZEN 3.3 (Blue edition, ZEISS) software.

Briefly, mice were euthanized by CO asphyxiation before the major organs were isolated, fixed with 4% paraformaldehyde solution overnight at 4 °C and transferred to 70% ethanol. Fixed and dehydrated organ samples were embedded in paraffin and sectioned into 3-μm-thick slices. Such paraffin sections were then dewaxed, rehydrated and subjected to antigen retrieval. The samples were stained with 1:100 anti-PTEN antibody (Supplementary Tables 18 and 19) overnight at 4 °C and washed with 1× PBS. Then, the samples were incubated with a goat anti-rabbit horseradish peroxidase (HRP)-conjugated secondary antibody (Jackson, 111-035-003) at 1:200 dilution for 1 h at room temperature. After that, the samples were stained using diaminobenzidine and visualized with the digital pathology section scanner (3D HISTECH, Pannoramic DESK, P-MIDI, P250). Images were visualized via the Pannoramic Scanner software.

LNPs (1 g l) were incubated with mouse plasma (1:1, v/v) at 37 °C for 30 min. 0.7-M sucrose solution was then slightly dropped into the mixed solution of LNPs and plasma at a volume ratio of 1:1. This mixture was centrifuged (15,300g at 4 °C for 1 h), and the precipitated plasma-protein-adsorbed LNPs were washed with cold PBS three times before being lyophilized for 2 days and stored at -20 °C until used for further characterization. The protein-adsorbed LNP complexes were treated with cold acetone to dissolve the lipid components. The proteins were subsequently collected and prepared for SDS-PAGE analysis. The lyophilized plasma-protein-adsorbed LNPs were dissolved in cold acetone to remove lipids, before the adsorbed proteins were denatured using 8 M of urea at 65 °C for 15 min. Such prepared protein corona samples were further processed on 4%-20% SDS-PAGE gel and the blots were visualized by Coomassie Blue R-250 staining.

Isolated corona proteins were digested using trypsin and the digested peptides of each sample were desalted on C18 cartridges (Empore SPE Cartridges C18 (standard density); bed inner diameter, 7 mm; volume, 3 ml; Sigma), concentrated by vacuum centrifugation and reconstituted in 40 µl of 0.1% (v/v) formic acid. Finally, the peptide mixture was analysed on a timsTOF Pro MS (Bruker, 1.5 kV).

The MS raw data for each sample were combined and searched using the MaxQuant v. 1.6.14 software for identification and quantitation analysis. The function of the identified proteins was defined based on the information from the UniProt database (https://www.uniprot.org). The pI value of the identified proteins was calculated in terms of the bioinformatic resource ExPASy.

Differentially enriched proteins were identified according to the fold change of protein abundance and statistical difference (P value). The sequences of differently enriched proteins were locally searched by the NCBI BLAST+ client software and InterProScan, before the GO terms were mapped and sequences were annotated using the software program Blast2GO. The GO annotation results were plotted by R scripts.

All the MD simulations were carried out in the NAMD 2.14 package in CHARMM 36 force field with a TIP3P water model. Unless specified, the temperature applied in the simulations is 300 K, the pressure of which is 1 atm. We use 0.8 nm as the switching distance, 1 nm as the cut-off and 1.2 nm as the maximum distance for the neighbour list calculations. The time step is set as 2 fs, changing to 1 fs in some simulations in the relaxation steps to stabilize the simulation. We prepare the full-atom Vtn protein by homology modelling using MODELLER based on the protein 7RJ9 (ref. ), and relax the structure by the following steps: (1) freeze each atom, except the atoms guessed by MODELLER, and relax the structure for 1 ns; (2) freeze the complex structure and relax it in a neutralized periodic water box with a water thickness of 0.7 nm in the NVT ensemble for 500 ps; (3) do the same simulation with step (2) as the ensemble type is changed to NPT; (4) unfreeze all the atoms and relax the molecular system for 10 ns. We use the program PepNN to predict the binding sites of 6R peptide to the initial Vtn structure, requiring a binding probability of >0.7, followed by manually fitting 6R to the binding sites inside the pocket rich in the β-sheet structure, creating as many binding sites as possible (water molecules and neutralizing ions are removed in this step). The term 'binding site' is defined as the residues in the protein (that is, Vtn) with non-hydrogen atoms, giving contact distances of <3.8 Å to any non-hydrogen atoms in the protein (that is, 6R). Then, the complex is relaxed in a sequence of simulations similar to that for the Vtn protein: (1) freeze the Vtn protein and relax the complex structure for 500 ps; (2) freeze the complex structure and relax it in a neutralized periodic water box with a water thickness of 1.2 nm and salt (Na and Cl ions to neutralize the system) in the NVT ensemble for 500 ps; (3) do the same simulation with step (2) as the the ensemble type is changed to NPT; (4) unfreeze all the atoms and relax the molecular system for 60 ns. Of note, the cysteine linker between 6R and DSPE-PEG was included in this model, as an extra cystine block at the end of the 6R peptide that is far from the 6R-Vtn binding pocket. External tensile force is applied to this cysteine linker block to simulate the fracture of 6R-Vtn assembly under a mechanical loading force.

The resulting structure obtained from the above molecular modelling is used as the initial structure for the SMD simulations. The SMD simulations are carried out by fixing the α-carbon of the 66th residue in the Vtn as the α-carbon of the cysteine is pulled along the direction of the 6R backbone. The bias potential given by the SMD simulation is

where k is the stiffness (in some literature, it is called the force constant), is the coordinate of the atom on which the bias potential is acting on at time t, v is the pulling velocity and is the pulling direction. In this case, k is set as 1 nN Å and v and is set as 1 nm ns.

To conduct the SMD simulations for the 4R-Vtn and 5R-Vtn complexes, the initial structures were obtained as follows: one or two arginine residues in the tail of the 6R peptide are removed, whereas the rest of the peptide structure remains unchanged, to obtain the initial binding structure between 4R or 5R peptides and Vtn. In addition, the initial binding structures between Vtn and 7R or 8R were generated by extending the backbone of the 6R peptide in the 6R-Vtn assembly, with a straight secondary structure and autogenerating the side chains of the added one or two arginine residues. The pulling rate and direction of SMD simulations for the polyR-Vtn (number of arginines is four, five, seven and eight) assembly are the same as that for the 6R-Vtn assembly.

To obtain the energy release rate in the SMD pulling tests, the area under the force-extension curve within the range of extension from 0 nm to 4 nm is taken as the mechanical work input (W), converting to no strain energy in the Vtn protein. The nominal contact area between polyR and Vtn is estimated by the product of the total number of residues in the polyR peptide (n), the average length of an amino acid residue (L = 3.65 Å) and the average width of an amino acid residue (b = 3.2 Å). The energy release rate G is given by G = W/nLb.

The initial structure of the complex of Vtn-6R was obtained by AlphaFold 3 (ref. ), followed by a 10-ns equilibration process by MD simulation, resulting in a stable complex structure. A hybrid autoregressive protein language model for complex systems that utilizes the unsupervised evolution of structural information was subsequently used. The model integrates amino acid values and backbone structural information to evaluate the joint likelihood of all positions in the sequence, enabling the mutation of specified chain sequences for optimal protein design. The Vtn-6R complex structure was used as an input into the structure-informed protein language model, and several mutation sequences of the peptide were generated by mutating the 6R chain. According to the ranking given by the model score, we chose the two top-ranked mutation sequences, RRRYRR and RRERRR, for subsequent screening and validation. The assembly structures of Vtn-RRERRR and Vtn-RRRYRR were obtained using the AlphaFold 3 and MD simulation methods as mentioned above, and the binding free energies were calculated for using the gmx_MMPBSA tool.

The Vtn protein was incubated with POST-LNPs and TLNPs (all LNPs labelled with Cy5) for 2 h at 0.5 μg of Vtn protein per sample. Subsequently, POST-LNP- and TLNP-encapsulating Luc mRNA, either preincubated with or without Vtn, were added to A-498 or U-87 MG cells cultured in 12-well plates (500 ng of mRNA per well) for 2 h at 37 °C. Following the above incubation, the medium was aspirated, and the cells were washed three times with 1× PBS. Subsequently, cells were stained with Hoechst 33342 and washed twice with 1× PBS. Confocal microscopy (ZEISS LSM710) was used for fluorescence imaging, and Cy5 cells were quantified based on the data from five wells. For the Luc expression analysis, LNPs preincubated with or without Vtn were added to A-498 or U-87 MG cells in a 96-well plate (50 ng Luc mRNA per well) overnight. Following this, 20 μl of 1× cell lysis reagent was added per well and incubated for 10 min at room temperature. Subsequently, 100 μl of Luc assay reagent was added per well, and luminescence levels were measured using a microplate reader (TECAN, Spark). The experimental procedure adhered closely to the instructions provided with the Luc assay kit (E1500).

To examine the role of serum-derived Vtn protein in mediating the lung-targeting performance of 6R-LNP, we used immunoprecipitation to specifically deplete Vtn protein from the mouse serum. The efficiency of Vtn depletion was characterized using ELISA. Cy5-labelled 6R-LNP-encapsulating Luc mRNA was incubated in vitro with either untreated serum or Vtn-depleted serum for 30 min. The Vtn protein adsorbed onto 6R-LNP complexes was then quantitatively analysed. Next, the 6R-LNP coated with Vtn was incubated with A-498 and U-87 MG cells, followed by fluorescence imaging and Luc protein expression analysis, using the same procedures as described earlier.

To examine whether cell receptors for Vtn are essential for the cellular uptake of 6R-LNP, a competitive binding assay was conducted by preincubating Vtn or collagen protein with A-498 and U-87 MG cells for 2 h. Subsequently, Vtn-coated 6R-LNP (labelled by Cy5) was added to these cells, before Cy5-LNP fluorescence and Luc protein expression were analysed.

To investigate the role of αβ integrin, A-498 and U-87 MG cells were pretreated with the small-molecule inhibitor cilengitide (10 μm) of the αβ integrin. Vtn-coated 6R-LNP (labelled by Cy5) was then added to these cells, before Cy5-LNP fluorescence and Luc protein expression were evaluated. Additionally, αβ integrin receptor expression was knocked down in A-498 and U-87 MG cells and overexpressed in NIH/3T3 cells. The intracellular delivery efficiency of Vtn-coated 6R-LNP was analysed in these cell lines using the same method above.

Stable Itgav and/or Itgb3 knockdown in A-498 and U-87 MG cells was done using pLKO.1 lentiviral constructs expressing shRNA against human Itgav and Itgb3 (provided by the Center of Biomedical Analysis, Tsinghua University).

Stable overexpression in NIH/3T3 cells was done using the FUGW lentiviral construct with UbiC promoter expressing human Itgav and/or Itgb3 (cDNAs were provided by the Center of Biomedical Analysis, Tsinghua University). For lentivirus packaging, 6 μg of lentivirual plasmid and packaging plasmids (pΔ8.9 and Vsv-G) were co-transfected into NIH/3T3 cells with Lipofectamine 3000 reagent. Virus was collected at 48 h and 72 h post-transfection. To promote the transduction efficiency, 10 μg ml of polybrene was applied. The stably transduced cells were further selected by puromycin (3 μg ml) for 3 days.

To knock out the αβ integrin receptor in the lung, Cas9 mRNA and sgRNAs targeting Itgav or Itgb3 (4/1, wt/wt) were encapsulated in 6R-LNP and administered intravenously to male C57BL/6J mice (6-8 weeks) at an RNA dose of 4 mg kg on day 0 and day 2. On day 7, the animals were euthanized, and lung tissues were collected for subsequent quantitative polymerase chain reaction and flow cytometry analyses.

For the liver-specific overexpression of the αβ integrin receptor, Itgav or Itgb3 mRNA was encapsulated in 3R-LNP and administered intravenously to male C57BL/6J mice (6-8 weeks) at an mRNA dose of 0.3 mg kg at 0 h, 1 h and 3 h. At 8 h post-administration, liver tissues were harvested for further quantitative polymerase chain reaction and flow cytometry analyses.

Total RNA was extracted from biological samples using the EASYSPIN Plus RNA Rapid Extraction Kit (catalogue number RN28, Aidlab). The concentration of the extracted RNA was measured using a Thermo Scientific NanoDrop 2000/2000C spectrophotometer. Reverse transcription of the RNA was performed using the RevertAid First Strand cDNA Synthesis Kit, following the manufacturer's instructions. The reverse transcription reaction was carried out at 42 °C for 60 min, followed by heat inactivation at 70 °C for 5 min. The synthesized cDNA was subsequently amplified using an SYBR Green I detection system. The reaction mixture (20 μl) was prepared according to the manufacturer's instructions. The amplification conditions were as follows: initial denaturation at 95 °C for 30 s, followed by 40 cycles of denaturation at 95 °C for 5 s, annealing at 55 °C for 10 s, and extension at 72 °C for 15 s. Relative gene expression levels were analysed using the 2 method.

Collected organs were dissociated into single cells before cells were incubated with flow cytometry antibodies targeting Itgav and Itgb3 (Supplementary Table 18), for 20 min at room temperature in the dark. Following incubation, cells were washed three times with PBS. The expression levels of Itgav and Itgb3 were then quantitatively analysed using flow cytometry.

To analyse the consistency of POST-LNP performance over time, male C57BL/6J mice (6-8 weeks) were intravenously injected with Luc-mRNA-loaded POST-LNPs before the main organs were isolated and imaged (IVIS Lumina system) at 3 h, 6 h, 8 h and 12 h after Luc-mRNA administration. For the stability assessment, freshly synthesized POST-LNPs were stored at 4 °C for 1 month before being used for Luc mRNA delivery as described above.

To evaluate the tolerance of POST-LNPs in vivo, male C57BL/6J mice (6-8 weeks) were intravenously injected with a higher dose of Luc mRNA (1 mg kg) encapsulated within POST-LNPs. As controls, PBS (i.v.) and LPS (5 mg kg, i.p.) were used. At 24 h, the mice were euthanized and the main organs were isolated for H&E staining and imaged by the digital pathology section scanner (KF-PRO-120).

Male C57BL/6J mice (6-8 weeks) were intravenously injected with TLNP and POST-LNPs loaded with Luc mRNA and EGFP mRNA at a total dose of 0.4 mg kg (Luc mRNA, 0.1 mg kg; EGFP mRNA, 0.3 mg kg). Animals were euthanized at 6 h after POST-LNP administration, and organs were isolated and imaged for luminescence and EGFP fluorescence using the IVIS Lumina system and flow cytometry analysis.

For the encapsulation of ASO, the ratio of ionizable lipid and ASO was fixed at 15/1 (wt/wt). For liver-targeting and spleen-targeting ASO deliveries, the ASO-loaded TLNP, 3R-LNP and 6D-LNP were intravenously injected into male ICR-Tg (CAG-EGFP) mice (6-8 weeks) three times (on days 0, 1 and 2, respectively) at a dose of 1.5 mg kg. For lung-targeting ASO delivery, 6R-LNP was intravenously injected into EGFP mice three times (on days 0, 1 and 2) at a dose of 2 mg kg. After that, mice were euthanized and organs isolated on day 3 for downstream analysis by ex vivo fluorescence imaging (IVIS Lumina system), flow cytometry analysis, or tissue cryosection and confocal fluorescence imaging (ZEISS LSM710).

To quantitatively analyse the Cy5 or EGFP intensity in cells of different organs, isolated organs were dissociated into a single-cell suspension according to the protocols of Miltenyi kits and further proceeded with flow cytometry (Sony ID7000). Specifically, for the liver cell analysis, the liver was isolated from mice, washed with cold PBS, cut into small pieces and put into the liver dissociation solution (Miltenyi Biotec, 130-105-807). The small pieces were incubated at 37 °C for 30 min under shaking. Then, the samples were filtered with a 100-μm mesh. After that, all the isolation procedures were carried out at 4 °C. A two-step separation process was used to collect the hepatocyte and other types of cell. Briefly, the samples were centrifuged at 50g for 2 min to separate the hepatocyte from the rest of samples, and the collected hepatocyte was incubated in a red cell lysis solution (Solarbio, R1010) for 5 min. Meanwhile, the rest of the samples were subjected to centrifugation at 400g for 10 min, and the supernatant was removed and resuspended with 2.5 ml of red blood cell lysis buffer. Finally, the obtained hepatocyte and other types of cell were mixed and washed with cold PBS for three times. Then, a fluorescence-activated cell sorting full-spectrum flow cytometry analyser machine (Sony ID7000) was used to analyse the Cy5 or EGFP signal. The Cy5 or EGFP signal was analysed and visualized using the FlowJo software.

For the flow cytometry analysis of the spleen, lung and adipose tissue, the organ was isolated and washed with cold PBS, before cutting into small pieces and resuspending in a dissociation solution according to the protocols of Miltenyi kits. After incubation for 15 min at 37 °C, the samples were filtered with a 70-μm mesh and centrifuged at 400g for 10 min. Then, the samples were resuspended with 2.5 ml of red blood cell lysis buffer. Finally, the samples were washed with 1× PBS (containing 5% FBS) to harvest cells. The rest of the procedure was similar to that for the liver.

For the placenta and testis, the isolated tissue was cut into small pieces. Then, the sample was resuspended in a dissociation solution (Dulbecco's modified Eagle's medium + 15% FBS + 2 mg ml of type IV collagenase + 7 mg ml of DNase) for 20 min. After that, the samples were filtered with a 70-μm mesh and centrifuged at 400g for 10 min. Then, the samples were resuspended with 1 ml of red blood cell lysis buffer. Finally, the samples were washed with 1× PBS (containing 5% FBS) to harvest cells for flow cytometry analysis.

For the bone marrow, the femur and tibia bones were cut and the cells were collected by centrifugation at 1,000g for 5 min. Then, the isolated cells were filtered with a 100-μm mesh and centrifuged at 400g for 5 min. Thereafter, the samples were resuspended with 2 ml of red blood cell lysis buffer. Finally, the samples were washed with 1× PBS (containing 5% FBS) for flow cytometry analysis.

For the placenta-targeting LNP, termed as 6K-3-LNP, a formulation comprising ionizable lipid:cholesterol:phospholipids (DSPC):DSPE-PEG-6K:DSPE-PEG at a ratio of 50:38.5:10:3:0.75 was used. The encapsulation of EGFP mRNA within 6K-3-LNP was achieved at a molar ratio (N/P) of 5.67. E16.5 pregnant C57BL/6J mice received i.v. injections of 6K-3-LNP-encapsulating EGFP mRNA (0.3 mg kg). As control, similarly formulated TLNPs were used. Subsequently, at 6 h post-injection, ex vivo organ imaging was performed on the isolated placenta, liver, lung, spleen, heart and kidney.

For bone-marrow-targeting LNP, termed ESM-6H-LNP, a formulation comprising ionizable lipid:cholesterol:phospholipids (ESM):DSPE-PEG-6H:DSPE-PEG at a ratio of 50:38.5:35:3:0.75 was used. Following the encapsulation of EGFP mRNA at a molar ratio (N/P) of 5.67, ESM-6K-LNP was applied intravenously to male C57BL/6J mice (6-8 weeks) at a dose of 0.3 mg kg of EGFP mRNA. As control, similarly formulated LNPs without peptide modification (ESM-TLNP) were used. At 6 h post-injection, ex vivo organ imaging was conducted on the isolated bone marrow, liver, lung, spleen, heart and kidney.

For adipose-tissue-targeting delivery, 8K-LNP-encapsulating EGFP mRNA was utilized on i.p. administration. As control, TLNPs were used. At 6 h after i.p. administration (EGFP mRNA, 0.4 mg kg) using male C57BL/6J mice (6-8 weeks), ex vivo organ imaging was performed on the isolated adipose tissue, liver, lung, spleen, heart and kidney.

For testis-tissue-targeting delivery, 6K-LNP-encapsulating EGFP mRNA was utilized on i.p. administration. As control, TLNPs were used. At 6 h after i.p. administration (EGFP mRNA, 0.4 mg kg) using male C57BL/6J mice (6-8 weeks), ex vivo organ imaging was performed on the isolated testis, liver, lung, spleen, heart and kidney.

Cre mRNA was encapsulated within TLNP and POST-LNPs at a molar ratio (N/P) of 5.67. TLNP and POST-LNPs were then intravenously injected into male B6.129(Cg)-Gt(ROSA)26Sor (mTmG) mice (6-8 weeks) on days 0 and 1. For the liver-targeting and lung-targeting LNPs, the injection dose of Cre mRNA was 0.3 mg kg, and for spleen targeting, the injection dose of Cre mRNA was 0.5 mg kg due to relatively lower spleen delivery efficiency. Animals were euthanized on day 2, with the major organs isolated for downstream analysis by ex vivo fluorescence imaging (IVIS Lumina system), flow cytometry analysis, or tissue cryosection and confocal fluorescence imaging (ZEISS LSM710).

Cas9 mRNA and sgSTOP (2/1, wt/wt) were encapsulated within TLNP and POST-LNPs, and then co-delivered intravenously into male Ai14 mice (6-8 weeks) at a total RNA dose of 2 mg kg for three times (on days 0, 1 and 2). On day 7, animals were euthanized and the main organs were isolated for downstream analysis by ex vivo fluorescence imaging (IVIS Lumina system) or tissue cryosection and confocal fluorescence imaging (ZEISS LSM710).

Cas9 mRNA and sgPten (4/1, wt/wt) were encapsulated in TLNPs and POST-LNPs, and then co-delivered intravenously into male C57BL/6J mice (6-8 weeks) every day at an RNA dose of 4 mg kg for a total of 3 days. On day 10, the animals were euthanized and the liver, lung and spleen were collected for further immunohistochemistry (IHC) analysis. For the delivery of Cas9/sgPten RNPs, the molar ratio of Cas9 protein and sgPten was fixed at 1/3. To encapsulate RNPs into LNPs, the Cas9/sgPten RNPs were first dissolved in 1× PBS and incubated for 15 min, before being rapidly mixed with other LNP ingredients dissolved in an ethanol solution, using the microfluidic chaotic mixer. RNPs were delivered intravenously into male C57BL/6J mice (6-8 weeks) at a total RNA dose of 1.5 mg kg every day for 3 days (days 0, 1 and 2). On day 7, the liver, lung and spleen were isolated for IHC analysis.

Cas9 mRNA and sgPcsk9 (1/1, wt/wt) were encapsulated within 3R-LNP and co-delivered into male C57BL/6J mice (6-8 weeks) at a total RNA dose of 2 mg kg every day for 3 days (days 0, 2 and 4). On day 7, the animals were euthanized and the liver was collected for further analysis by western blot. To analyse the plasma level of Pcsk9 and cholesterol, the blood was collected on days 10 and 15, respectively.

For western blot, the liver was ground and all the proteins were extracted according to the protocol of the extraction kit (Invent, SD001). The moderated protein was loaded and isolated by SDS-PAGE, before being transferred onto an NC nitrocellulose membrane and blocked with 5% milk. Then, it was incubated with an anti- Pcsk9 antibody (Supplementary Tables 18 and 19) for 22 h at 4 °C. The membrane was washed using PBS with Tween-20 (PBST) for three times, before a donkey anti-goat IgG H&L (HRP) secondary antibody was applied for 1 h at room temperature. After rinsing with PBST for three times, the membrane was visualized using a gel imaging scanner (GE). To further examine Gapdh on the same blot, the Pcsk9 antibody was stripped from the membrane. The membrane was then incubated with the primary antibody against Gapdh for 1 h at room temperature. Following three washes with PBST, the membrane was incubated with a rabbit anti-mouse IgG H&L (HRP) secondary antibody for 1 h at room temperature. After another three washes with PBST, the Gapdh was visualized using a gel imaging scanner (GE). For the quantitative analyses of plasma levels of Pcsk9 and cholesterol, the collected blood was centrifuged (3,000g at 4 °C for 15 min) to acquire the serum. ELISA kit (Ruixin Biotech, RX203246M) and biochemical kit (Nanjing Jiancheng Biotech, A111-1-1) were used for the detection of Pcsk9 protein and cholesterol, respectively.

To investigate the cellular distribution of LNPs, Luc mRNA-encapsulating lung-targeting 6R-LNPs and spleen-targeting 6D-LNPs were administered to male C57BL/6J mice (6-8 weeks) via a tail-vein injection. At 3 h post-injection, the mice were euthanized, and the lung and spleen tissues were collected. Single-cell dissociation was performed, followed by single-cell transcriptome sequencing. The encapsulated exogenous Luc mRNA served as the sequencing target, enabling the reconstruction of its distribution across different cell types. This approach allowed for the characterization of cell-type-specific delivery profiles of 6R-LNP and 6D-LNP.

The raw scRNA-seq data (FASTQ files) were mapped to the GRCm39/mm39 mouse reference genome using the CellRanger software with the default parameters. The gene expression matrices generated by CellRanger were imported into the Seurat R package for downstream analysis. Quality control was performed on each dataset to exclude low-quality and contaminated cells. Cells with mitochondrial gene expression exceeding 5% were removed. Dataset-specific quality control thresholds were applied: for lung samples, cells with nFeature_RNA > 200 and nFeature_RNA < 3,000 were retained; for spleen samples, cells with nFeature_RNA > 200 and nFeature_RNA < 5,000 were selected.

Data from individual organ samples were processed and integrated using the standard Seurat workflow to assess potential batch effects. Normalization was performed using the SCTransform method from the glmGamPoi package. Pearson residuals were then subjected to principal component analysis, nearest-neighbour graph construction and uniform manifold approximation and projection for dimensionality reduction. Clusters were identified using the shared-nearest-neighbour modularity algorithm in Seurat. Cell types were annotated based on the expression of known cell-type-specific marker genes. The delivery efficiency of LNPs was assessed by evaluating the expression of Luc mRNA. Delivery efficiency was quantified as the proportion of cells expressing these two genes within the dataset.

EGFP mRNA encapsulated within POST-LNPs was intravenously administered to male C57BL/6J mice (6-8 weeks) at a dose of 0.4 mg kg of mRNA. At 6 h post-injection, the liver (3R-LNP group), lung (6R-LNP group), spleen (6D-LNP group), placenta (6K-3-LNP), bone marrow (ESM-6H-LNP), adipose tissue (8K-LNP) and testis (6K-LNP) were individually isolated. Subsequently, organs were dissociated into single cells. The resulting cell suspensions were then incubated with the corresponding antibodies for 20 min at 4 °C in the dark (Supplementary Table 18). The quantitative analysis of cell types was conducted using a full-spectrum flow cytometry analyser machine (Sony ID7000).

Data were analysed and plotted as the mean ± standard deviation (s.d.) using the GraphPad Prism v. 9.5 software, unless otherwise noted. One-way analysis of variance (ANOVA) and Dunnett's multiple comparisons test or unpaired Student's t-test were performed to evaluate the statistical significance, as specified in individual figure captions. The GraphPad Prism v. 9.5 software was used to annotate the statistical difference. No statistical methods were used to predetermine the sample sizes, but our sample sizes are similar to those reported in previous publications. Data distribution was assumed to be normal, but this was not formally tested. Animals and samples were randomly assigned to different experimental groups. The data collection was randomized. The data collection and data analysis were performed blind to the conditions of the experiments. No animals or data points were excluded from the analyses.

To quantify the luminescence intensity in ex vivo organ luminescence images, the average luminescence radiance was measured for each organ using Living Imaging software. The organ selectivity of LNPs was characterized by the proportion of average luminescence radiance from different organs. For ex vivo organ fluorescence (EGFP and tdTomato) imaging and analysis, the absolute fluorescence was quantified by Living Imaging software. Because of notable auto-fluorescence background from tissues, fluorescence radiance change (that is, the difference in average fluorescence between LNP-treated group and PBS-treated group) to reflect the effect of LNP delivery and, thus, the percentage of such radiance change is calculated in different organs to comparatively illustrate the performance of organ-selective delivery. To quantify the level of EGFP or tdTomato fluorescence in the tissue sections, the mean fluorescence intensity was used. Image-Pro Plus software was used to perform quantitative analysis on ten tissue sections from each organ. To quantify the relative expression level of Pten in the IHC images, the mean optical density (that is, integrated optical density of a region divided by the area of the region) of Pten staining was measured for each experimental condition using Image-Pro Plus software, before normalizing to that of the PBS-treated group.

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.