Megakaryocyte membrane‐wrapped nanoparticles for targeted cargo delivery to hematopoietic stem and progenitor cells

Abstract Hematopoietic stem and progenitor cells (HSPCs) are desirable targets for gene therapy but are notoriously difficult to target and transfect. Existing viral vector‐based delivery methods are not effective in HSPCs due to their cytotoxicity, limited HSPC uptake and lack of target specificity (tropism). Poly(lactic‐co‐glycolic acid) (PLGA) nanoparticles (NPs) are attractive, nontoxic carriers that can encapsulate various cargo and enable its controlled release. To engineer PLGA NP tropism for HSPCs, megakaryocyte (Mk) membranes, which possess HSPC‐targeting moieties, were extracted and wrapped around PLGA NPs, producing MkNPs. In vitro, fluorophore‐labeled MkNPs are internalized by HSPCs within 24 h and were selectively taken up by HSPCs versus other physiologically related cell types. Using membranes from megakaryoblastic CHRF‐288 cells containing the same HSPC‐targeting moieties as Mks, CHRF‐wrapped NPs (CHNPs) loaded with small interfering RNA facilitated efficient RNA interference upon delivery to HSPCs in vitro. HSPC targeting was conserved in vivo, as poly(ethylene glycol)–PLGA NPs wrapped in CHRF membranes specifically targeted and were taken up by murine bone marrow HSPCs following intravenous administration. These findings suggest that MkNPs and CHNPs are effective and promising vehicles for targeted cargo delivery to HSPCs.


| INTRODUCTION
Hematopoietic stem and progenitor cells (HSPCs) are predominantly localized in the bone marrow (BM) and possess the ability to selfrenew or differentiate into cells of all blood lineages. 1 Their ability to differentiate into all blood-related cells make HSPCs ideal candidates for gene regulation, but enabling effective delivery of nucleic acid cargo to HSPCs is a long-standing problem whose solution has "formidable promise… that may transform medical practice." 2 Thus, by facilitating HSPC-specific delivery of nucleic acid cargo and other therapeutics, a host of genetic hematological diseases, such as sickle cell anemia and thrombocytopenia, can be ameliorated as diseased HSPCs can be repaired and ultimately differentiate to various lineages of healthy blood cells. [3][4][5][6] Naked nucleic acids (plasmids, small interfering RNAs [siRNAs], microRNAs [miRNAs]) cannot effectively enter into cells without a carrier or be used clinically because they are rapidly degraded by serum nucleases in vivo, thus resulting in a short blood circulation half-life. 7 Current delivery methods using viral vectors (e.g., based on lentivirus or adeno-associated virus) are hampered by limited loading capacity, poor DNA insertion, and considerable cytotoxicity. 8 Further, these materials are not well suited for use in HSPCs because the cells lack expression of appropriate adenoviral receptors. 2 To successfully deliver nucleic acid cargo to HSPCs, a delivery system must specifically target HSPCs while also protecting the cargo. Here, we demonstrate that poly(lactic-co-glycolic acid) (PLGA) nanoparticles (NPs) can be loaded with nucleic acids and "cloaked" or "wrapped" with megakaryocytic (Mk) membranes that facilitate HSPC-specific cargo delivery. Membrane-wrapped NPs (MWNPs) were pioneered by Hu et al. and Parodi et al., who wrapped red blood cell and leukocyte membranes around NPs for improved circulation time and immune evasion, and the concept was expanded upon by Fang et al. who showed that cancer cell MWNPs could facilitate homotypic tumor targeting. [9][10][11] Here, we exploit this concept to enable targeted cargo delivery to HSPCs.
Human Mk microparticles (MkMPs) are 100-1000 nm naturally occurring extracellular vesicles 12 that bud off the cytoplasmic membrane of Mk cells (derived from cultured HSPCs). 13 They have been shown to specifically target and deliver their cargo to HSPCs, both in vitro 14,15 and in vivo, 16 through receptor-meditated endocytosis and cytoplasmic-membrane fusion. 14 Microparticles (MPs) from CHRF cells (termed CMPs) also effectively target and deliver cargo to HSPCs with similar if not better effectiveness. 15 CHRF cells are a wellcharacterized model human megakaryoblastic cell line used to study megakaryopoiesis. 17 Thus, CHRF membranes are similar to normal Mk membranes, expressing similar membrane glycoproteins. Given the natural targeting ability of MkMPs and CMPs for HSPCs, we hypothesized that membranes extracted from Mk and CHRF cells could be used to produce MWNPs and produce a semisynthetic carrier that can specifically target and deliver cargo to HSPCs (Figure 1). Here, we gen-  Through TEM verification of the membrane shells surrounding the entire particle ( Figure 2h) and the size and charge changes between bare and wrapped NPs, the NPs were considered successfully wrapped.

| MkNPs are readily taken up by HSPCs
We next investigated the interaction between MkNPs and HSPCs in vitro. MkNPs were added to HSPCs that had been cultured for 3 days as described in Section 5.5. For these studies, DiD-loaded PLGA NPs and the PKH26-labeled Mk membranes were prepared prior to extrusion. MkNPs are visible by the signal colocalization of the DiD cargo in the NPs and the stained Mk membranes (Figure 3a).
Using confocal microscopy, we observed that more than 90% of the HSPCs contained overlapping NP and membrane signals across multiple levels of Z-stacks after 24 h of incubation, suggesting that MkNPs were internalized intact by the HSPCs. To further confirm that the MkNPs were inside the cells, HSPCs were stained with AF488conjugated phalloidin, labeling actin in the outer HSPC membrane and analyzed using super-resolution structured illumination microscopy (SR-SIM). As before, minimal MkNP signals were observed along the HSPC membrane at the peripheral Z-stacks; the colocalization signals were well inside the cells in the medial Z-stacks, thus confirming that the MkNPs were indeed internalized by HSPCs (Figure 3b). Our analysis further revealed several regions in the HSPC cytoplasm where the NP cargo and membrane signal were colocalized, suggesting the MkNPs were intact following uptake by HSPCs (Figure 3b). Quantitative uptake of MkNPs by HSPCs is presented in Section 2.3.
We also investigated the endocytic pathway by which bare PLGA NPs and MkNPs are taken up by the HSPCs. We followed the experimental design and associated experience from our study using MkMPs. 14,15 HSPCs were preincubated with endocytic inhibitors including dimethylamiloride (DMA), methyl-β-cyclodextrin (MβCD), Dynasore, or LY29400, which block macropinocytosis, lipid raftmediated uptake, dynamin-dependent endocytosis, and macropinocytosis through PI3K, respectively. HSPCs were treated with each of the F I G U R E 1 Overview of the synthesis and application of membrane-wrapped nanoparticles (MWNPs). Megakaryocytic (Mk) membrane vesicles, derived from Mk-like CHRF-288 cells, can be wrapped around polymeric nanoparticles (NPs) loaded with desired cargo to produce membrane-wrapped NPs (MWNPs) that can selectively target, bind, and enter hematopoietic stem and progenitor cells (HSPCs) to deliver their cargo.

| MkNPs exhibit interaction specificity with CD34 + HSPCs in comparison to MSCs and HUVECs
To test their target specificity, MkNPs containing DiD were incubated with either CD34 + HSPCs or two additional cell types that are physiologically related to HSPCs. MSCs co-reside in the BM microenvironment with HSPCs, and provide stroma functions for HSPCs. 19 Mature and immature blood cells, including HSPCs and Mk, enter the systemic circulation through gaps of the endothelium of BM sinusoids 13,20,21 where they interact with endothelial cells. Here, we used HUVECs as (c) Before exposure to either bare DiD-loaded NPs or MkNPs, HSPCs were preincubated with inhibitors against specific endocytic pathways, including dimethylamiloride (DMA), methyl-β-cyclodextrin (MβCD), Dynasore, and LY29400, which block macropinocytosis, lipid raft-mediated uptake, dynamin-dependent endocytosis, and macropinocytosis through PI3K, respectively. Uptake of bare NPs and MkNPs was analyzed by assessing the fraction of DiD + cells via flow cytometry following 30 min of incubation, and NP uptake for each inhibitor-treated culture is shown relative to either bare NP or MkNP uptake in untreated (none) HSPC cultures. Data represent the average of four (bare NPs) and four (MkNPs) biological replicates ± standard error of the mean. *p < 0.05 versus untreated control (Student's t test) a model endothelial cell. We hypothesized that MkNPs would be preferentially taken up by HSPCs over MSCs and HUVECs. Both flow cytometry and confocal microscopy were used to examine and characterize NP-cell interactions.
First, we examined the impact of various concentrations of bare PLGA NPs on the viability of each cell type to determine if there was an upper limit in the ratio of NPs to cells that could safely be administered. PLGA NPs are biodegradable and biocompatible, but excess NPs could induce cytotoxicity due to adsorption of cytosolic proteins onto the surface of the NP. 22 Based on the results of a cytotoxicity assay (Supplemental Figure S1a), we applied a ratio of 40,000:1 NPs per cell as this yielded high NP uptake with minimal tradeoff in viability. As expected, bare DiD NPs were indiscriminately taken up by all examined cells, as >90% of the HSPCs, MSCs, and HUVECs were DiD using natively produced MkMPs. 14 To summarize, we found that coating PLGA NPs with Mk membranes enables their preferential interaction with and uptake by HSPCs versus HUVECs and MSCs.
F I G U R E 5 Legend on next page.

| Mk-like CHNPs facilitate delivery of siCD34 to HSPCs in vitro
To determine if MWNPs could be viable cargo delivery vehicles, we needed to test if functional cargo could be successfully delivered and deployed to HSPCs. We selected siRNA as a model cargo for our proof-of-concept studies. In previous studies, siRNA has proven stable when encapsulated in PLGA NPs. [23][24][25] We chose to load the NPs with siRNA designed to disrupt the expression of CD34 (siCD34) since CD34 is a characteristic surface marker of immature, undifferentiated HSPCs. 26 Thus, the effectiveness of siCD34 delivery could be quickly assessed by flow cytometry. Importantly, reducing CD34 expression with siCD34 should have no impact on HSPC viability, as HSPCs display gradually reduced expression of CD34 as they differentiate into different blood cell lineages. 27,28 For these experiments, nontargeting siRNA (siNeg) was used as a control.
For these experiments, CHRF-288-11 cells, a megakaryoblastic cell line, were used to generate membranes for wrapping siRNA- CHNPs. While the CHNPs yielded slower rates of CD34 disruption in HSPCs than bare NPs, they offer the advantage of specific targeting afforded by the Mk-like CHRF membranes. This specificity is expected to enable improved delivery of cargo to target cells if implemented in a heterogeneous environment (more than one cell type present) and is explored for in vivo work in Section 2.5.

| DISCUSSION
PLGA has emerged as an attractive choice for harboring therapeutics for drug delivery, largely due to its stability and broad biocompatibility. 29-32 PLGA NPs may be loaded with a broad range of therapeutic molecules, from small molecules to miRNAs and DNA with minimal loss of functionality. 24,25,33,34 However, bare or PEGcoated PLGA NPs administered in vivo lack tissue tropism, thus limiting the potency of the encapsulated therapeutic. 32,35 PLGA and other NPs can be modified to target specific tissues by coating them with different types of cell-derived membranes. 36  Beyond PLGA carriers, many other materials have been used to create MWNPs. 36,40 This includes synthetic and naturally occurring polymers that offer stability, biocompatibility, and easy manipulation, as well as metallic NP cores that offer unique imaging and photo-responsive properties. 41 In this work, PLGA was chosen due to its wide application in nanomedicine, ability to encapsulate both hydrophilic and hydrophobic cargo, ease of synthesis, and proven use in membranewrapped carrier systems. Here, we demonstrate that wrapping PLGA NPs in Mk membranes can enable targeted cargo delivery to HSPCs.
We have previously demonstrated that MkMPs can recognize and fuse with HSPCs for HSPC-specific cargo delivery. 14 While we demonstrated robust and preferential uptake of MkNPs into HSPCs both in vitro and in vivo, additional work will be needed to further validate MkNPs as a full-fledged targeted drug delivery system. Our data indicate that siRNA-loaded CHPPNPs successfully disrupted expression of CD34 upon delivery to HSPCs. However, as we were unable to quantify the specific amount of siRNA encapsulated within the CHNPs, our dosing strategy was based on CHPPNP particle counts and particle to HSPC ratios rather than the total amount of siRNA delivered to each cell. Future studies will need to quantitate the amount of siRNA contained within each NP, the release kinetics of the siRNA from the NPs, and the total amount of siRNA delivered to each target cell. Additionally, using siRNA-loaded CHPPNPs in vivo with a murine disease model could further showcase their potential for facilitating HSPC-specific RNA-based therapies for a variety of genetic hematological diseases. [42][43][44] As mentioned earlier, future studies could also explore the compatibility of MkNPs or CHNPs with various classes of therapeutic cargo rather than the model (non-therapeutic) cargo we used as a proof of concept here.

| CONCLUSION
In conclusion, we have shown that MkNPs and CHNPs specifically and robustly interact with hard-to-transfect HSPCs both in vitro and in vivo and can be used to deliver functional nucleic acid cargo to HSPCs. With further development and optimization, these tools may be used to enable the treatment of a broad spectrum of blood disorders.

| Mk and CHRF-288-11 cultures
Human CD34 + D0 (Day 0) cells were cultured in IMDM supplemented with 20% BIT 9500, rhSCF, rhTPO, rhIL-3, rhIL-6, rhIL-11, and human LDL to produce Mks and incubated in a hypoxic (5% O 2 , 5% CO 2 ) humidified (95% rH) environment at 37 C as described by Panuganti et al. 26 After a Day-5 media exchange, Day 7 CD61 + cells (Mks) were enriched using anti-CD61 magnetic microbeads (Miltenyi Biotec) and were cultured in IMDM containing 20% BIT 9500. rhSCF, rhTPO, human LDL, and nicotinamide as described by Panuganti et al., 26 and incubated in a 20% O 2 , 5% CO 2 , and 85% rH environment at 37 C. On Day 12, Mks were collected, and membranes were isolated as explained in Section 5.3. Dounce homogenizers with tight-fitting pestles (Kimble) as described. 11 The entire solution was subjected to 30 passes with each pestle before spinning down at 3200 Â g for 5 min at 4 C. The supernatants were collected and spun down at 20,000 Â g for 20 min at 4 C, after which the pellets were discarded or used for protein analysis, and subsequently centrifuged at 100,000g for 90 min at 4 C. 11 The supernatant was discarded, and pelleted cell membranes were  MkNPs were selected from a population containing both PKH26 and DiD; presence of CD41a on the MkNPs was determined from this subpopulation. Analysis presented is the average of four replicates.
Live cells were examined using confocal microscopy with a 40Â oil objective (Carl Zeiss LSM880 multiphoton confocal microscope).    45 After several washes with filtered PBS, the fixed slides were mounted using mounting media (Slow Fade with DAPI) and sealed with coverslips. Slides were imaged using confocal microscopy with a 40Â oil objective (Carl Zeiss LSM880) and measured using their respective fluorescent channels.
To assess the optimal dose of bare NPs to cell, unwrapped DiDlabeled PLGA NPs were incubated with CD34 + cells, MSCs, and HUVECs at ratios of 1 Â 10 4 , 1Â10 5 , and 1 Â 10 6 NPs per cell in a 12-well plate (2 biological replicates) and stored at 37 C, 20% O 2 , 85% humidity. After 24 h, cells were removed from their respective wells, washed, and incubated in filtered PBS containing 0.1% ethidium homodimer for 15 min at 37 C to assess viability. Cells were then analyzed and gated for live and dead cells using size and fluorescence under the PE channel. DiD NP uptake was tabulated by measuring the number of DiD + live cells under the APC channel.

| Generation of siRNA-loaded CHNPs
To encapsulate hydrophilic siRNA cargo in PLGA NPs, a previously established double emulsion water-in-oil-in-water procedure was adapted. 46 Briefly, the first water phase was created by adding 0.4 nmol siCD34 or siNeg to 100 μl of 1% PVA in deionized (DI) water. The siRNA/PVA solution was then added dropwise, while stirring, to 1 ml of a 4 mg/ml PLGA-in-acetone solution that was prepared by dissolving 50:50 PLGA (LACTEL Absorbable Polymers) (inherent viscosity: 0.67 dl/g) in acetone. The mixture was stirred for 5 min at 800 rpm to create a water-in-oil emulsion that was then added dropwise while stirring to 0.1% aqueous PVA at a 1:3 ratio, to produce the water-in-oil-in-water solution. The solvent was evaporated overnight under continuous stirring at 800 rpm and the resultant NPs were purified by 50 kDa MWCO centrifugal filtration (Sigma) at 3200g for 30 min at 4 C. siRNA-NPs were wrapped as described in Section 5.4 using CHRF membranes and the final product characterized by TEM, zeta potential, and NTA.  To extract the BM cells for flow cytometry and microscopy, each femur was placed in a clean petri dish and scraped clean of any other tissue (muscle, connective tissue, etc.) and rinsed with 1Â PBS containing antibiotic-antimycotic. Next, the epiphyses were trimmed, and a 3-ml syringe equipped with a 20G needle and filled with the RPMI-FBS tissue storage buffer was used to flush the BM out of the bone.
The BM cells were subsequently added atop a 30-μm pre-separation filter (Miltenyi Biotec). Finally, additional RPMI tissue storage buffer was added to the petri dish, and the evacuated femur was crushed.

ACKNOWLEDGMENTS
Tissue sectioning and H&E staining was performed by J. Cooper at the Delaware Biotechnology Institute. Microscopy equipment (Carl Zeiss LSM880 confocal microscope) was acquired with a shared instrumentation grant (S10 OD016361) and access was supported by the NIH-NIGMS (P20 GM103446), the NSF (IIA-1301765) and the State of Delaware. The authors would also like to thank E. Muñoz and the Hudson Lab at the University of Delaware for initial NTA equipment access.

CONFLICT OF INTEREST
The authors have a patent pending related to the MkNP technology under international PCT application number PCT/US2019/063685.

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