Genetic drugs’ real-life potential is critically dependent on delivery technology that stabilizes the nucleic acid payload and prevents its premature clearance and degradation, while actively delivering the payload to the cells of interest1. In the past decade, lipid nanoparticle (LNP) technology emerged that enabled the first small interfering RNA (siRNA) therapeutic’s clinical translation, involving gene silencing in hepatocytes for the treatment of hereditary transthyretin amyloidosis2,3. Further, LNP technology was essential for the COVID-19 messenger RNA (mRNA) vaccines’ development and successful deployment4,5. While for most vaccination purposes intramuscular administration is adequate, following intravenous administration, LNPs preferentially accumulate in the liver’s hepatocytes6. To enable targeting of tissues and cells beyond the liver, several LNP screening and modification strategies have been developed and evaluated preclinically. These include incorporating charged phospholipids7,8 in, and antibody surface modification of LNPs9–14, as well as DNA barcoding strategies15,16. However, to unlock RNA therapeutics’ full potential for systemic delivery, novel technologies must be developed that possess superb biocompatibility profiles and tunable biodistribution features.
Here, we introduce a nanodelivery strategy based on lipoprotein trafficking. Lipoproteins are endogenous, nanosized transport systems composed of apolipoproteins and aggregates of fatty molecules17. They inherently interact with a variety of cells and exhibit compelling features for RNA delivery18. The widely studied apolipoprotein A1 (apoA1) is high-density lipoprotein’s (HDL) main protein constituent19. Capitalizing on apoA1’s natural function, we developed designer nanotechnology20,21 to deliver siRNA to myeloid cells and hematopoietic stem and progenitor cells (HSPCs) in the bone marrow22–24 (Fig. 1a). After establishing a prototype apolipoprotein lipid nanoparticle (aNP) that stably incorporates siRNA in its core, we built a comprehensive siRNA-aNP library and physicochemically characterized and screened individual formulations’ in vitro properties. We subsequently selected eight siRNA-aNPs that are representative of the library’s diversity, and in vivo screened their capacity to silence lysosomal-associated membrane protein 1 (LAMP1) in diverse immune cell subsets in mice. Finally, we studied the lead siRNA-aNP candidate’s physicochemical properties, tissue biodistribution, and in vivo silencing in HSPCs.
Prototyping apolipoprotein lipid nanoparticles for siRNA immune cell delivery
First, we developed a two-step flow manufacturing process for an aNP prototype containing siRNA, the phospholipid 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), the ionizable lipid (6Z,9Z,28Z,31Z)-heptatriacont-6,9,28,31-tetraene-19-yl-4-(dimethylamino)butanoate (MC3), tricaprylin, cholesterol, and apoA1. Our methodology entails producing a siRNA-lipid assembly intermediate through rapid mixing of the components at pH 4 using a T-junction mixer, followed by dialysis against PBS at pH 7.4, and forming stable nanoparticles through apoA1 incorporation using a second rapid mixing step (Supplementary Fig. 1). The prototype aNP input components depicted in Fig. 1b consist of 46 wt% tricaprylin, a triglyceride that makes up the core matrix for MC3-complexed siRNA, DMPC and cholesterol, and 5 wt% apoA1. Importantly, we found siRNA recovery, entrapment, and retention to be higher than 80% (Fig. 1c), indicative of efficient and stable siRNA incorporation. The prototype aNP contained consistent quantities of apoA1 (Fig. 1d) and had a mean hydrodynamic diameter of approximately 80 nm (Fig. 1e), with a favorably low dispersity of 0.1, as measured by dynamic light scattering (DLS). We qualitatively corroborated both size and size distribution by cryogenic electron microscopy (cryo-EM, Fig. 1f), which revealed spherical core-shell structures. Using a gold nanoparticle antibody labeling method, we visualized apoA1’s integration in the aNP by negative staining transmission electron microscopy (TEM, Fig. 1g).
In an in vitro silencing assay using RAW 264.7 murine macrophages that stably express firefly- and renilla luciferase, we found the IC50 value of our aNP prototype to be 14.5 nM (Fig. 1h), which was considerably better than 154.3 nM for an MC3-based control lipid nanoparticle (LNP) formulation. Encouraged by these findings, we developed a unique siRNA radiolabeling strategy (Fig. 1i and Supplementary Fig. 2) to quantitively measure siRNA biodistribution by in vivo positron emission tomography (PET) and ex vivo gamma counting. Using azide-alkyne click chemistry, we functionalized siRNA with desferrioxamine B (DFO) to enable radiolabeling with zirconium-89 (89Zr, Fig. 1i and Supplementary Fig. 2). While DLS disclosed a slightly increased hydrodynamic diameter of approximately 95 nm with a uniform size distribution (Fig. 1j), cryo-EM (Fig. 1k) showed the same spherical core-shell structure as the nonlabelled aNP (Fig. 1f). In addition to bare 89Zr-labeled siRNA (89Zr-siRNA), we intravenously administered 89Zr-siRNA-LNP and 89Zr-siRNA-aNP to C57BL/6 mice (Fig. 1l, m, at a dose of 8 µCi/mouse, n = 3–5 mice/group). At 24 hours, we subjected mice to in vivo PET imaging, after which they were sacrificed for ex vivo gamma counting of 89Zr in various tissues and organs. PET images displayed clear differences between bare 89Zr-siRNA’s biodistribution (Fig. 1l), showing renal clearance, and LNP or aNP formulated 89Zr-siRNA. Importantly, PET imaging revealed enhanced 89Zr-siRNA uptake in the spleen and bone marrow for aNP as compared to LNP delivery. Indeed, quantitively measuring 89Zr-siRNA using gamma counting (Fig. 1m) corroborated our in vivo PET imaging findings. Bare 89Zr-siRNA was primarily found in the kidneys, while aNPs delivered 89Zr-siRNA to the bone marrow (5.5-fold, p < 0.0001) and spleen (2.5-fold, p < 0.0001) significantly better than control LNPs.
After observing the favorable accumulation in hematopoietic organs, we evaluated the prototype aNP’s silencing potential in leukocytes. Towards that goal, we first ex vivo screened eight siRNA constructs against the lysosomal-associated membrane protein 1 (LAMP1, Supplementary Fig. 3). LAMP1 is a compelling target for screening purposes as it is expressed on the cell surface and therefore can be quantified by flow cytometry, as has been demonstrated by Da Silva Sanchez et al16. We subsequently formulated the most potent anti-LAMP1 siRNA (siRNALAMP1) into the prototype aNP formulation. In a proof-of-concept in vivo study, we administered four intravenous siRNALAMP1-aNP doses (0.5 mg/kg siRNALAMP1, injected 36 hours apart) to six C57/BL6 mice. As compared to an aNP that incorporates scrambled siRNA, we observed significant LAMP1 knockdown in leukocytes in the spleen and bone marrow, which is mostly attributed to LAMP1 silencing in myeloid cells (Fig. 1n). The in vivo siRNALAMP1-aNP silencing potency was on par with the positive control siRNALAMP1-LNP. Encouraged by our prototype siRNA-aNP’s in vivo performance, we set out to optimize the delivery platform through aNP library screening.
Establishing and characterizing an apolipoprotein lipid nanoparticle library
After prototyping the siRNA-aNP delivery concept, we designed an aNP optimization strategy, involving an iterative design process that ultimately resulted in a library of 72 compositionally distinct (Supplementary Table 1) and fully characterized siRNA-aNPs. To achieve this, we formulated and trialed more than 300 aNPs. Within the library of siRNA-aNP formulations, we kept siRNA and apoA1 quantity constant while varying the cholesterol, tricaprylin, and ionizable lipid MC3 content (Fig. 2a, b). MC3 is a lipid that complexes negatively charged siRNA to allow its integration into an aNP’s lipophilic core, cholesterol and tricaprylin provide structural features that are essential for an aNP’s supramolecular organization and structural stability. Amphiphilic phospholipids are required to disperse the lipophilic supramolecular nanoaggregates of siRNA, MC3, tricaprylin, and cholesterol in an aqueous environment and together with apoA1 stabilize the aNP. We selected 3 phospholipids that are known to properly interact with apoA125, namely DMPC, 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC). After formulating more than two dozen DPPC-based aNPs, we did not identify stable formulations with adequate in vitro silencing features (Supplementary Fig. 4) and decided to exclude these formulations from our library and further testing.
The different aNPs’ compositions and quality features are schematically depicted in Fig. 2c. Although we observed increased tricaprylin to consistently improve DMPC-based formulations, this did not apply to aNPs containing POPC. In general, cryo-EM (Fig. 2d) revealed that the inclusion of tricaprylin promoted the formation of spherical structures, but we also found other morphologies such as aNP39. In the absence of tricaprylin (aNP1 to 24), cryo-EM revealed an array of different morphologies, ranging from spherical (e.g., aNP4, 7, 10, 11), to multilamellar for some POPC-based formulations (e.g., aNP3, 6, 15), and elliptical (e.g., aNP30), but also shapes that were less well defined, such as aNP18 and aNP21. In Supplementary Fig. 5, a detailed overview of cryo-EM images and size analyses (Supplementary Table 2) is provided.
As a critical quality feature, we measured aggregation of individual aNPs directly after production by semi-quantitative assessment of cryo-EM images (Supplementary Fig. 6) and found 55 formulations to pass our predefined threshold of < 15%. Additionally, we determined the aNPs’ stability at 4 °C by DLS. Specifically, we measured size dispersity over a period of 28 days. Interestingly, we observed little correlation between initial aggregation (Fig. 2e) and shelf-life stability (Fig. 2f). Finally, we measured all 72 aNPs’ gene silencing capacity in vitro (Fig. 2g). Similar to other aNP features, we observed diverse gene knockdown capabilities at a fixed concentration of 100 nM siRNA, ranging from no silencing to nearly complete gene silencing. In vitro silencing capacity is a poor predictor of in vivo silencing following intravenous administration26. Therefore, we primarily used this silencing metric as a quality control. Overall, the DMPC-based aNPs were prone to aggregate less but did not appear to be more stable. By setting the quality control limits to 15 %for aggregation, a maximum of 40 %increase in PDI at 28 days, and at least 30 %silencing, we identified 30 aNPs with potentially favorable in vivo features. From those 30 formulations, we selected eight representative aNPs (Fig. 2c, in bold) to be included in ensuing in vivo gene silencing experiments in mice. Our selection was based on identifying aNPs that are representative of the library’s diversity. We included three POPC-based formulations, one of which did not contain tricaprylin, and five DMPC-based formulations, with varying quantities of tricaprylin. Throughout this selection of eight aNPs, the formulations’ morphologies, compositions and nitrogen-to-phosphate ratios vary.
Functionally screening apolipoprotein lipid nanoparticles in vivo
Following aNP library characterization and in vitro screening, we selected eight representative aNPs for in vivo testing in mice, including the aNP prototype (aNP72). Similar to the experiments we conducted with the aNP prototype, we encapsulated siRNALAMP1 in the selected aNPs. We assessed these eight siRNALAMP1-aNPs’ knockdown functionality using bone marrow-derived macrophages (BMDM) from mice ex vivo. At 100 nM siRNALAMP1, reverse transcription quantitative polymerase chain reaction (RT-qPCR) measurements disclosed that all eight aNPs substantially reduced LAMP1 expression (Fig. 3a). After establishing their functionality, we intravenously administered the eight different siRNALAMP1-aNPs four times, with 36-hour intervals at 0.5 mg/kg siRNALAMP1, in groups of four to six mice per formulation (Fig. 3b). Following this siRNALAMP1-aNP treatment regimen, we sacrificed the mice and collected bone marrow cells. These cells were stained with flow cytometry antibody panels that we optimized for the detection of hematopoietic stem cells, myeloid progenitors, and mature myeloid cells (Fig. 3c and Supplementary Fig. 7). Within the aNP selection, we found a remarkable diversity in LAMP1 gene silencing features in immune cells (Fig. 3d, e and Supplementary Fig. 8, 9, 10). Foremost, we observed significant LAMP1 silencing in hematopoietic stem cells using aNP8, aNP18 and aNP67. While aNP8 and aNP18 were also capable of significantly silencing LAMP1 in myeloid progenitor cells, we did not observe significant LAMP1 silencing in this progenitor subset using aNP67. We did, however, measure significant LAMP1 silencing in myeloid progenitors using aNP72, a formulation that did not exert any effects on hematopoietic stem cells. A comprehensive overview of the different siRNALAMP1-aNPs’ in vivo silencing features is shown in Fig. 3f. It discloses aNP8’s and aNP18’s potential for broad gene silencing of stem and progenitor populations, while aNP67 silencing displays a bias towards stem cells. For gene silencing specifically in myeloid progenitor cells, aNP72 is most suitable. In light of aNP18’s broad silencing features, we selected this formulation for further studies that focused on reproducibility, pharmacokinetics, safety and pharmacodynamics.
In-depth profiling of lead siRNA apolipoprotein lipid nanoparticle
Starting with the siRNA-aNP prototype (Fig. 1), following comprehensive library screening, and identifying broad gene silencing features (Fig. 2, 3), we selected aNP18 for an in-depth analysis. We incorporated siRNA in aNP18, a DMPC-based formulation that does not contain tricaprylin (Fig. 4a). As holds true for all the formulations that met the predefined quality thresholds, siRNA-aNP18 has siRNA recovery, entrapment, and retention values of more than 80% (Fig. 4b), with consistent apoA1 quantities (Fig. 4c), for multiple batches. In the absence of tricaprylin, the aNP’s hydrodynamic diameter was approximately 50 nm, with a narrow size distribution (PDI < 0.2, Fig. 4d). Encouraged by these strong reproducibility data, we subjected six independently produced siRNA-aNP18 batches to cryo-EM. Cryo-EM disclosed the preparations to predominantly consist of multilamellar, elliptical structures and this was consistent across the six aNP18 batches that we analyzed (Fig. 4e).
After establishing satisfactory reproducibility, we studied the in vivo features of a single siRNA-aNP18 dose in mice. Using the same method that we applied for the prototype aNP, we radiolabeled a siRNA with 89Zr, followed by integration into aNP18 (Fig. 1i). 24 hours following intravenous administration into nine mice (at a dose of 8 µCi/mouse), we sacrificed the mice and subjected different tissues to gamma counting (Fig. 4f). For this specific aNP formulation, we observed a relatively strong bone marrow accumulation and an approximately 3-fold higher uptake in the spleen. As many myeloid cells, including myeloid progenitor cells, migrate from the bone marrow to the spleen, we deem the high splenic uptake to be beneficial. The comparably high liver uptake, which we found to be lower than for control LNPs (Fig. 1m), and kidney accumulation, must be taken into consideration and carefully monitored. We therefore measured the liver enzymes alanine transaminase (ALAT) and aspartate aminotransferase (ASAT), as well as the kidney toxicity markers creatinine and blood urea nitrogen (BUN) (Fig. 4g). In line with the liver and kidney toxicity data, ELISA did not uncover immune activation following a single siRNA-aNP administration (at a dose of 0.5 mg/kg) (Fig. 4h). We measured low tumor necrosis factor alpha (TNFα) and interleukin-6 (IL-6) concentrations comparable to those observed in mice that received intravenous PBS, and significantly lower than the TNFα and IL-6 concentrations following intraperitoneal LPS (0.5 mg/kg) administration, used as positive control.
Finally, we evaluated aNP18’s potential for functional gene silencing of Src homology region 2 domain-containing phosphatase-2 (SHP2). SHP2 is a widely studied therapeutic target in oncology, for example in metastatic breast cancer27. Relevant to the current study and using conditional knockout mouse models, Christofides et al. implicated SHP2 in myeloid cell-driven immunosuppression of the tumor microenvironment28. We established an aNP18 formulation containing siRNA against SHP2, which we refer to as siRNASHP2-aNP18. We first evaluated siRNASHP2-aNP18’s functionality on murine bone marrow-derived macrophages ex vivo and found dose-dependent silencing effects (Fig. 4i). Finally, we conducted an in vivo silencing study and applied siRNASHP2-aNP18 intravenously in C57/BL6 mice (n = 5 per group) inoculated with B16F10 cells according to the regimen depicted in Fig. 4j. Although SHP2 gene silencing varied between individual mice, RT-qPCR disclosed a trend towards a 3-fold reduction in SHP2 mRNA content (Fig. 4k), indicative of the aNP platform’s therapeutic potential.