In vivo hematopoietic stem cell modification by mRNA delivery

Abstract

Hematopoietic stem cells (HSC) are the source of all blood cells over a lifetime. Diseased HSC can be replaced with gene-engineered or healthy HSC through HSC transplantation (HSCT). However, current protocols carry significant side effects and have limited access. We developed CD117/LNP-mRNA, a novel lipid nanoparticle (LNP) encapsulating mRNA that is targeted to stem cell factor receptor (CD117) on HSC. Delivery of the anti-human CD117/LNP-based editing system yielded near-complete correction of hematopoietic sickle cells. Furthermore, in vivo delivery of pro-apoptotic PUMA mRNA with CD117/LNP affected HSC function and permitted non-genotoxic conditioning for HSCT. The ability to target HSC in vivo offers a non-genotoxic conditioning regimen for HSCT and this platform could be the basis of in vivo genome editing to cure genetic disorders, abrogating the need for HSCT.

One Sentence Summary:

mRNA-encapsulating lipid nanoparticles targeted to hematopoietic stem cells allow efficient in vivo genome editing and control of cellular function.

Hematopoietic stem cells (HSCs) reside in the bone marrow, where they divide throughout life to produce all cells of the blood and immune system due to their self-renewal ability. Their multipotency enables the formation of myeloid (i.e., erythroid, megakaryocytic, and myeloid-immune) and lymphoid cell progenitors. Bone marrow HSC transplantation (HSCT), which replaces diseased HSCs with healthy ones, can be a curative treatment for non-malignant hematopoietic disorders, such as hemoglobinopathies and immunodeficiencies. Non-malignant hematopoietic disorders can be cured by allogeneic HSCT (where the HSC source is obtained from a sibling, parent, or unrelated donor), but only a fraction of patients have a suitable immunologic match to minimize the potentially fatal complication of graft versus host disease (GVHD). Gene therapy can eliminate the risk of GVHD and correct non-malignant hematopoietic disorders by using autologous HSCs (where the HSC are obtained from the actual patient) and replace the genetic defect either by gene addition or editing. Current hematopoietic gene therapy requires isolation of HSC from the patient, and ex vivo lentiviral transduction for gene addition or electroporation with purified reagents for genome editing. A “conditioning” regimen, such as chemotherapy or radiation, is used to eliminate the patient’s own HSCs. This makes space in the bone marrow niche to allow engraftment of infused allogeneic donor or genetically modified autologous HSC. The conditioning procedure carries significant acute and chronic systemic toxicities, including infertility and secondary malignancies due to accumulated DNA damage. Additionally, some non-malignant hematopoietic disorders are due to DNA repair pathway mutations, such as radiosensitive severe combined immunodeficiency (SCID) or Fanconi anemia. These patients do not tolerate existing conditioning due to excessive toxicity with alkylating chemotherapy or radiation, as well as increased rates of malignancy long-term. Therefore, we sought to address two major challenges by developing a novel flexible methodology that can modify HSC in vivo and separately establish a non-genotoxic conditioning method. Here, we describe an HSC-targeted lipid nanoparticle (LNP) encapsulating mRNA that utilizes antibodies against CD117 conjugated to LNP (CD117/LNP-mRNA). HSCs are dependent on stromal-derived factors, including stem cell factor (SCF), which binds to the receptor c-Kit (CD117). CD117 is expressed on both short- and long-term HSCs and some hematopoietic progenitors (1). CD117 is internalized after binding of SCF, which we hypothesize may facilitate or augment LNP internalization (2). Nucleoside-modified and purified mRNA is non-immunogenic, stable, extensible, and can be used to express virtually any protein of interest (3–5). Lipid nanoparticles (LNP) are thus far the most promising delivery systems to fulfill the therapeutic potential of mRNA (6, 7). These LNP contain ionizable lipids (positively charged at pH < 6.4), which aid in packaging the mRNA and endosomal escape. Such LNP were first approved in 2018 for siRNA (8), but became widely utilized in 2020, due to the LNP-mRNA platform for Moderna and Pfizer COVID-19 vaccines. The LNP-mRNA in these FDA-approved vaccines drive antigen expression, but do not actively target specific cells or organs. By decorating the surface of LNP with targeting moieties, we have demonstrated effective targeting to specific cell types, such as endothelial cells and T cells, with therapeutic efficacy upon single i.v. injection in mice, as described in our previous reports (9–11). Here, we utilized nucleoside-modified mRNA encoding Cre recombinase, a cas9 base editor fusion gene, or the pro-apoptotic BH3-only gene PUMA in CD117/LNP-mRNA to genetically alter HSC, correct a disease mutation, or deplete HSC by non-genotoxic conditioning, respectively. Prior studies have shown that HSC depletion via immunotoxins or radioimmunotherapy can be performed as conditioning for HSCT (12, 13), but they can only serve as platforms for HSC depletion, rather than delivering other cargoes. Our proof of principle data reveals an innovative and flexible approach to target HSC in vivo, which may pave the way to modify HSC behavior and correct genetic mutations by delivering targeted mRNA-based therapeutics capable of genome engineering.

Results

Anti-CD117 LNP efficiently target bone marrow cells in vitro

We first incubated C57BL/6 lineage depleted (Lin⁻) bone marrow (BM) cells or whole bone marrow (WBM) in vitro with either unconjugated LNP encapsulating 0.1, 1, or 3 μg of nucleoside-modified luciferase mRNA (unmodified LNP-Luc), anti-CD45-conjugated LNP (CD45/LNP-Luc), anti-CD117-conjugated LNP (CD117/LNP-Luc), or isotype control IgG-conjugated LNP (control IgG/LNP-Luc). CD45/LNP and CD117/LNP were hypothesized to bind all hematopoietic-derived cells or stem and progenitor cells, respectively. Control IgG/LNP and unconjugated LNP were utilized as controls. The highest levels of luciferase activity in WBM were detected with CD117/LNP-Luc (Fig. 1A). Luciferase activity was further increased when Lin⁻ cells were treated with CD117/LNP-Luc (Fig. 1B). Increased activity of CD117/LNP-Luc in Lin⁻ cells was consistent with a 23-fold increase in the proportion of CD117⁺ in Lin-selected cells (2.8% CD117⁺ in WBM cells vs. 65% CD117⁺ in Lin⁻ cells). CD117/LNP luciferase activity was 500 and 700-fold higher than CD45/LNP luciferase activity in WBM and Lin⁻, respectively, when normalized to the frequency of CD45 and CD117 positive cells in WBM and Lin⁻ cells (fig. S1A). Normalized luciferase activity suggests that CD117 mediated targeting and delivery is superior to CD45 mediated targeting in vitro. This demonstrates efficient targeting and functional delivery of mRNA with CD117/LNP.

Fig. 1. In vitro targeting of whole bone marrow or hematopoietic progenitors (lin^-) cells incubated with LNPs encapsulating luciferase (CD117/LNP-Luc) or Cre recombinase (CD117/LNP-Cre) mRNAs.

(A) Luciferase activity normalized by total protein in whole bone marrow cells incubated with varying doses (indicated on x-axis) of targeted or control LNP-Luc for 18 hours in vitro. Data indicate mean +/− SD of n=3 replicate experiments. P values are from Dunnett’s multiple comparison after two-way ANOVA. **** p<0.0001. (B) LNP-Luc treatment of Lineage negative (Lin-) bone marrow cells (N=3). Data indicate mean +/− SD of n=3 replicate experiments. P values are from Dunnett’s multiple comparison after two-way ANOVA. **** p<0.0001. (C-G) Assessment of ZsGreen+ reporter induction after CD117/LNP-Cre treatment in Ai6 bone marrow (BM) cells triggered by removal of loxP flanked stop cassette by Cre. Treatment of (C, E, G) bone marrow cells or (D, F) Lin-BM cells at doses and culture intervals stated in figure. (D, F) Lin-Sca1⁺cKit⁺ (LSK) subset shown when treating Lin-cells. No difference between CD117/LNP-Cre editing in Lin^- cells treated with 0.1 and 0.5 or 0.5 and 1μg;. In C-G, data represent mean +/− SD of n=3 replicate experiments. P values are from Dunnett’s multiple comparison after two-way ANOVA. Specifically, in C-G, ** p<0.01, *** p<0.001 **** p<0.0001.

CD117/LNP encapsulating Cre recombinase mRNA (CD117/LNP-Cre) was used to test LNP-mediated genetic recombination in HSCs and persistence of the editing in conjunction with three reporter murine models. These murine models (Ai6, Ai9 and Ai14) are engineered with a Creresponsive reporter allele comprised of a loxP-flanked STOP cassette preventing transcription of a CAG promoter-driven green or red fluorescent reporter gene (ZsGreen for Ai6 and tdTomato for Ai9 and Ai14, respectively) inserted into the Gt(ROSA)26Sor locus (14). The fraction of edited WBM cells (Fig. 1C), and the subset of edited Lin⁻Sca1⁺cKit⁺ cells (LSK) within the BM (Fig 1D), exhibited a dose dependency (0.1 to 1 μg mRNA) when incubated with CD45/LNP-Cre and control IgG/LNP-Cre. The majority of LNP-mediated transfection occurred within 6 hours (Fig. 1C–F). Targeting rates in the LSK subset were consistently and significantly higher with CD117/LNP-Cre than with CD45/LNP-Cre or control IgG/LNP-Cre, suggesting saturation of cKit⁺ cells by CD117/LNP-Cre at the lowest dose tested (Fig. 1C and D). CD117/LNP-Cre showed higher efficacy in LSK cells at lower concentrations: treatment with 0.1 μg CD117/LNP-Cre was 2.5-fold more effective at targeting LSK cells compared to treatment with 0.1 μg CD45/LNP-Cre (Fig. 1D). There was no significant difference between targeted cell frequency in the LSK subset with the 0.1 μg and 0.5 μg dose or 0.5 μg and 1 μg dose. We also replaced media of cells treated for 18h with LNP-Cre and kept them for 3 additional days in culture to assess the maximum targeting achieved after exposing WBM to LNP. The rate of targeted cells increased over 3 days without additional LNP exposure (Fig.1G): at a dose of 0.1 μg CD117/LNP-Cre, 88.5% WBM cells were ZsGreen⁺ at 90h vs. 43.5% at 18h (Fig. 1G and 1E), indicating that additional mRNA translation, cre-mediated recombination, and ZsGreen⁺ transcription/translation occurred beyond the 18-hour LNP exposure. Importantly, LNP-Cre treatment had no consistent effect on cell viability across formulations, regardless of the targeting antibody (fig. S1B–D). Hence, we determined that the use of CD117/LNP-Cre was superior to that of CD45/LNP-Cre to modify HSCs and selected CD117-LNP-Cre for subsequent experiments.

Anti-CD117 LNPs edit multi-potent and self-renewing long-term HSC ex vivo

To evaluate multipotency in cells edited by CD117/LNP-Cre, we transplanted lethally irradiated congenic C57BL/6 CD45.1 recipient mice with Ai14 BM cells treated ex vivo with increasing doses of CD117/LNP-Cre and control IgG/LNP-Cre. Since HSCs give rise to all blood cell lineages, we followed reporter gene expression in peripheral blood cells over time and analyzed the BM at the 4-month endpoint (Fig. 2). The percent of CD117/LNP-Cre-mediated tdTomato positive Ai14 erythroid cells in recipient mice increased with time post-HSCT, consistent with engraftment of donor HSC (Fig. 2–C). Mice had durable editing in all lineages, specifically myeloid cells (Gr1+, Fig. 2A), lymphoid cells (CD3+ and B220+, Fig. 2B), and erythroid cells (Fig. 2C) at 4 months post-HSCT, consistent with genome editing of multipotent HSC.

Fig. 2. CD117/LNP-Cre treatment ex vivo leads to near complete tdTomato gene editing upon transplantation.

(A) Percent tdTomato marking in myeloid (Gr1+), (B) lymphoid (CD3+, left and B220+, right) cells measured at 16 weeks post HSCT in lethally irradiated congenic CD45.1 recipients receiving Ai14 bone marrow treated ex vivo with 0.1 μg (and 1 μg of control IgG/LNP-Cre or CD117/LNP-Cre. In A and B data represent mean +/− SEM of n=4 (for IgG/LNP-cre at 1μg only) or n=5 experimental animals per cohort. P-values are from Tukey’s multiple comparison test after one-way ANOVA. In A**** p<0.0001. In B *p<0.05, **** p<0.0001. (C) Kinetic analysis of erythroid editing measured up to 16 weeks post HSCT. Data represent mean +/− SD of n=4 or 5 experimental animals per cohort. (D) tdTomato marking in in the bone marrow (BM) and BM subsets: c-Kit⁺ (Lin^-c-Kit⁺), LSK (Lin^-c-Kit⁺Sca1⁺), and LT-HSC (LSK CD150⁺CD48^-). Data represent mean +/− SEM of n=4 or 5 experimental animals per cohort (same animals as in A and B). P-values are Tukey’s multiple comparison test after one-way ANOVA. *** p<0.001, and **** p<0.0001. (E) Colony forming unit assay from Ai14 bone marrow treated ex vivo with 0.1 μg or 1 μg of control IgG/LNP-Cre or CD117/LNP-Cre formulations or untreated. Semi-quantitative PCR of (F) bone marrow and (G) spleen genomic DNA isolates from the groups in A-C at 4 months post BMT. **271bp Cre-recombinase edited gDNA region, *1142bp unedited region are indicated.

Editing rates in long-term HSC (LT-HSC, LSK CD150+ CD48-aka SLAM) was 95% at the 0.1 and 1 μg mRNA dose with CD117/LNP-Cre compared to 13.5 and 20% with control IgG/LNP-Cre, respectively (Fig. 2D), which was similar to that seen in the whole BM (WBM), the c-Kit⁺, and the LSK subsets, respectively. Donor chimerism was consistently high among all groups (over 94% at 4 months, fig. S2A). Gene editing rates of ex vivo treated bone marrow cells was dose dependent (fig. S2B, C). RBC and leukocyte editing rates with CD117/LNP-Cre were ≥99% at 0.05, 0.1, and 1 μg mRNA dose, and 91.8% at the 0.01 μg dose (Fig. 2A–C and fig. S2D). By comparison, targeting mediated by control IgG/LNP-Cre was near 0% at 0.01 μg (fig. S2C, D). tdTomato⁺ Gr1⁺ cells had the fastest rise (fig S2B–C), which is expected given their rapid turnover of 2–3 days. BM cells harvested from these animals showed similar editing rates in colony forming assays, a functional assay for clonogenic potential, thus corroborating the flow cytometry results of LT-HSC (Fig. 2E and fig. S2E, F). At four months post-HSCT, splenocytes had genome editing levels comparable to those in the WBM (Fig. 2F, G), consistent with migration of edited BMderived cells to the spleen. To assess stem cell potential of ex vivo edited bone marrow cells, we performed secondary transplants using the BM from two primary chimeras that were recipients of Ai14 BM cells treated ex vivo with either CD117/LNP-Cre or control IgG/LNP-Cre (0.1 μg dose of mRNA). Editing levels in secondary chimeras phenocopied those observed in the primary transplantation, including sustained editing in the LT-HSC subset and editing in multiple hematopoietic lineages (fig. S3).

In vivo editing of multi-potent and self-renewing long-term HSC

Given the near complete targeting of LT-HSC ex vivo with CD117/LNP-Cre and our prior ability to target lung endothelial and T cells in vivo (9–11, 15), we hypothesized that LT-HSC could be targeted in vivo, as well. Intravenous (i.v.) administration of CD117/LNP-Luc generated luciferase activity in the femur at 24 h, whereas IgG/LNP-Luc did not (Fig. 3A). Of note, both control IgG/LNP-Luc and CD117/LNP-Luc showed comparable luciferase activity in the liver, as LNP bind ApoE and are non-specifically targeted to the LDL receptor, which is expressed on hepatocytes (8). We tested in vivo multilineage editing by quantifying tdTomato expression in peripheral blood cells of i.v. CD117/LNP-Cre treated animals over time (up to 4 months) and tdTomato expression in bone marrow, and specifically the LT-HSC, at 4 months. At the same dose (5 μg) CD117/LNP-Cre treated mice had significantly higher editing in all peripheral blood lineages (Fig. 3B–C), and importantly 3-fold higher editing in LT-HSC (55 vs 19%, respectively) compared to that observed in control IgG/LNP-Cre treated mice (Fig. 3D). HSC editing after in vivo treatment with CD117/LNP-Cre was dose dependent in peripheral blood and bone marrow at 16 weeks with a 5.5-fold increase in the percentage of gene-edited LT-HSC with 5 μg versus 1 μg (Fig. 3E–G). LNP-Cre in vivo editing led to the appearance of edited RBC with similar kinetics to transplantation of ex vivo treated bone marrow (Fig. 3H–I). At 4-months post-treatment with CD117/LNP-Cre, marking of HSC was confirmed by visual inspection of tdTomato⁺ CFU (Fig. 3J and fig. S7A), while Cre-mediated genomic deletion in the BM and splenic DNAs was confirmed by PCR (Fig. 3K and L). To further confirm in vivo LT-HSC targeting, we investigated editing of the endothelial protein C receptor (EPCR)+ LT-HSC SLAM subpopulation (16), whose self-renewal properties are enriched compared to that of the LT-HSC SLAM population (17), using the Ai6 model (fig. S4). Editing rates in the SLAM LT-HSC population and the EPCR+ LT-HSC subpopulation were comparable within each cohort (CD117/LNP-Cre and Control IgG/LNP-Cre) (fig. S4E). Mice injected with CD117/LNP-Cre had 55%±10 edited SLAM LT-HSC versus 46%±14 edited EPCR+ LT-HSC, while mice in the control group had 9%±2.3 edited SLAM LT-HSC versus 8%±1.9 edited EPCR+ SLAM LT-HSC (fig. S4E). CFU from the bone marrow of primary chimeras generated from in vivo treated donors confirmed editing differences between the two cohorts and yielded no difference in the number colonies (fig. S4F–H). To demonstrate that LNP mediated editing targeted bona fide HSC, chimeras from the initial in vivo experiment (Ai9 strain) were generated by transplanting irradiated congenic (C57BL/6 CD45.1) recipients with BM from mice four months post-in vivo treatment with a 5 μg dose of CD117 or control IgG/LNP-Cre. Assessment of the hematopoietic-derived lineages, including LT-HSC in the BM, in these chimeras recapitulated editing found in the donor cells (fig. S5). LT-HSC editing in secondary chimeras was 52% for those derived from the CD117/LNP-Cre treated primary and 19% for those derived from the control IgG/LNP-Cre treated primary. Absolute count of viable LT-HSC was comparable among cohorts in both primary ex-vivo transplants and in mice injected in vivo (fig S6).

Fig. 3. CD117/LNP-Cre formulations lead to over 50% tdTomato marking in LT-HSC after in vivo injection.

(A) Biodistribution of i.v. injection of 1 μg of targeted LNP-mRNA expression in vivo by luminescence imaging at 24 hours. A representative sample set of dissected mouse organs were analyzed 5 min after the administration of D-luciferin. tdTomato⁺ cell frequency in peripheral blood (B) myeloid (Gr1+) and (C) lymphoid cells (CD3+ [T-cells], B220+ [B-cells]) and in (D) bone marrow (BM) subsets (c-Kit, Lin^-c-Kit⁺ subset, LSK, Lin^-c-Kit⁺Sca1⁺, SLAM/LT-HSC, LSK CD150⁺ CD48-) at 4 months after 5 μg of CD117/LNP-Cre or control IgG/LNP-Cre. In B, C, and D data represent mean +/− SEM of n=5 experimental animals per cohort. P-values are reported from paired t-test. ** p<0.01, *** p<0.001, and **** p<0.0001. tdTomato⁺ cell frequency in peripheral blood (E) myeloid and (F) lymphoid cells, and in (G) bone marrow subsets at 4 months after 5 or 1 μg of CD117/LNP-Cre. In E, F, and G data represent mean +/− SEM of n=7 (1μg) and 5 (5μg) experimental animals per cohort. P-values are reported from t-test. *** p<0.001, and **** p<0.0001. Edited RBC frequency over time in Ai9 mice treated in vivo with (H) 5 μg of CD117/LNP-Cre or control IgG/LNP-Cre or with (I) 1 or 5 μg of CD117/LNP-Cre. In H data represent mean +/− SD of n=5 experimental animal per cohort. P-values are reported from paired t-test. **** p<0.0001. In I data represent mean +/− SD of n=7 (1μg) and 5 (5μg) experimental animals per cohort. P-values are reported from t-test. **** p<0.0001. J) Colony forming unit assay from bone marrow at 4 months after in vivo treatment with 5 μg control-IgG/LNP-Cre (top), no treatment (middle), or 5 μg CD117/LNP-Cre (bottom). Semi-quantitative PCR of (K) bone marrow and (L) spleen genomic DNA isolates from the groups in A-C at 4 months post BMT. **271bp Cre-recombinase edited gDNA region, *1142bp unedited region are indicated.

Non-hematopoeitic targeting after targeted LNP treatment

To quantify non-specific cellular uptake, we compared tdTomato expression levels in lung and liver cells 4-months after in vivo treatment with a single dose of CD117/LNP-Cre (1 and 5 μg dose) or control IgG/LNP-Cre (5 μg dose). At 5 μg, liver editing was high (76–79% of cells), and editing was comparable between the two treatments (fig. S7B), consistent with known non-specific ApoE and LDL receptor axis mediated LNP uptake (8). In the lung, tdTomato expression mediated by CD117/LNP-Cre delivery was significantly higher (7-fold) than that of mice injected with control IgG/LNP-Cre (fig. S7C). Editing observed in the perfused lung was 3-fold higher with 5 μg of CD117/LNP-Cre compared to 1 μg. This effect was partly “on-target” editing: ~8% of lung cells were cKit⁺ and ~90% of lung cKit⁺ cells were edited (fig. S7D). Cells collected from the testis were also analyzed and did not show significant variations from baseline levels in control mice (fig. S7E). Additionally, none of 50 offspring sired by male mice treated with CD117/LNP-Cre in vivo (N=4) or 39 offspring sired by male mice treated with control IgG/LNP-Cre (N=3) in vivo expressed tdTomato. A complete list of animals tested is provided in supplemental table 3.

Efficient in vitro editing of primary Sickle Cell Disease hematopoietic stem and progenitor cells with anti-human CD117

To assess the feasibility of using this platform for therapeutic human genome editing, we adapted our targeting to human CD117 and utilized LNP containing mRNA encoding a cas9 adenine base editor (ABE) fusion and LNP carrying a single guide RNA (sgRNA) targeted to the beta-globin sickle cell mutation. Adenine base editing of the A to G leads to conversion of the pathogenic E6V (HBB^S) mutation to a non-pathogenic E6A variant (HBB^G-Makassar)(18). We applied this therapeutic strategy to convert pathogenic sickle hemoglobin (HBB^S) to non-pathogenic G-Makassar hemoglobin (HBB^G) on 4 sickle cell specimens from separate donors (fig. S8A–B). We found that a molecular excess of sgRNA to ABE mRNA containing LNPs lead to efficient editing with highest rates (88%) at 10 pg/cell dose (fig. 4A). This led to a corresponding increase in HBB^G protein (up to 91.7% of beta-like globin) and HBB^S decrease after in vitro erythroid differentiation, and a nearly complete absence of sickled cells upon exposure of the erythroblasts to hypoxic conditions (Fig 4B–C). Editing levels and increase of HBB^G were directly correlated (Fig. 4D). We observed that LNP doses from 3 pg/cell up to 10 pg/cell did not alter viability and proliferation rate of erythroid progenitor cells in vitro (fig. S8C–D).

Fig. 4. Base editing of the E6V sickle cell mutation with Human CD117 targeted LNP.

(A) Representative reverse-phase (RP) HPLC chromatograms of in vitro differentiated sickle cell disease erythroid progenitor lysates after treatment with anti-human CD117 (hCD117)/LNPNRCH cas9 ABE-8e mRNA and hCD117/LNP gRNA. Base editing yields non-pathogenic Hbb^G-Makassar (β^G), which elutes before Hbb^S (pathogenic, β^S) and the α-globin protein (α). % shown is β^G/(β^G+β^s) *100. (B) Representative images of sickling of in vitro differentiated erythroid progenitors under hypoxic conditions at the treatments in (A). Arrowheads indicate sickled morphology. Scale bar 20 microns. (C) % sickled cells from unedited and edited (varying mRNA doses) sickling assays. Data represent mean +/− SD of n=10 high powered fields (hpf) (unedited specimens) and n=30 hpf (edited specimens). P-values are reported from unpaired t-test. **** p<0.0001. (D) Correlation of %β^G by RP-HPLC (protein) to base edited allele frequency (DNA).

PUMA mRNA depletes HSC from mouse bone marrow in vitro

Survival of human and mouse HSC depend on the anti-apoptotic gene Mcl-1 (19, 20), thus we sought to test the ability of CD117/LNP to deplete BM cells using pro-apoptotic mRNA. We tested a variety of pro-apoptotic mRNAs that act within this pathway. Among those genes tested on murine C57BL/6 BM cells, treatment with PUMA mRNA reduced BM and LSK viability after 48h and 6 days in culture, respectively (fig. S9A). To confirm LNP-PUMA mRNA treatment depleted multi-lineage HSPC, we performed competitive HSCT in which C57BL/6 CD45.2 BM was treated with CD117/LNP-PUMA ex vivo (5 μg) and transplanted at equal or increasing ratios against untreated GFP⁺ C57BL/6 CD45.2 BM cells into lethally irradiated congenic C57BL/6 CD45.1 recipients (schema fig. S9C). If CD117/LNP-PUMA efficiently depletes HSC, mice receiving only CD117/LNP-PUMA treated BM (C57BL/6 CD45.2) would experience BM failure due to depletion of HSC and those receiving competitive BM would have an over-representation of untreated GFP+ BM. The results were consistent with our expectations: mice injected with only CD117/LNP-PUMA treated GFP⁻ BM cells died within 2 weeks from the HSCT, indicating that HSC were not viable and did not engraft. Mice receiving 50 or 75% CD117/LNP-PUMA-treated C57BL/6 CD45.2 GFP-BM had less than 0.5% donor GFP⁻ Gr1⁺ cells or RBCs (Fig.5A–B) at 4-months (endpoint), versus the expected 50–75%. The remainder of donor cells (CD45.2) were GFP+ (untreated) cells. This is consistent with essentially complete depletion of engrafting, multilineage HSC with ex vivo treatment of CD117/LNP-PUMA. By comparison, mice injected with control untreated GFP^+/− C57BL/6 CD45.2 BM at a 1:1 ratio had 25% GFP+ cells (Fig. 5AD). At endpoint, all groups had similar donor chimerism (≥94% C57BL/6 CD45.2; Fig. 5E).

Fig. 5. HSC depletion and transplantation conditioning with CD117/LNP-PUMA.

(A) GFP⁺ granulocytes and (B) RBC in peripheral blood, as well as %GFP⁺ (C) CD45⁺ splenocytes and (D) BM cells of C57BL/6 CD45.1 chimeras competitively transplanted with indicated proportion of GFP+ C57BL/6 BM untreated and C57/BL6 (GFP-) BM treated with CD117/LNP-PUMA. Data represent mean +/− SD of n=4 (recipients of a 25:75 ratio of GFP:C57/BL6+CD117/LNP-PUMA BM), n=8 (recipients of a 50:50 ratio of GFP:C57/BL6+CD117/LNP-PUMA BM), and n=4 (recipients of a 50:50 ratio of GFP:C57/BL6 untreated BM) experimental animals per cohort. P-values calculated by Dunnett’s multiple comparison test after one-way ANOVA.**** p<0.0001. (E) Donor chimerism 4 months post HSCT. Chimerism calculated as CD45.2%/(CD45.1%+CD45.2%). Data represent mean +/− SD of the same cohorts indicated in A-C. One-way ANOVA not significant (p>0.05). (F) Granulocyte, (G) RBC, and (H) hematopoietic cells of the BM, BM subsets, and spleen in recipients conditioned with 0.05 mg/kg CD117/LNP-PUMA and receiving 10×106 GFP+ C57BL/6 BM cells at 6.5 days post treatment. Data in F-H represent mean +/− SD of n=3 recipient animals. Levels of GFP+ granulocytes and RBC in unconditioned controls (N=2) were nearly undetected (0.06±0.03 and 0.05±0.02, respectively) 2 months after BMT. (I) Persistence upon secondary transplantation of CD117/LNP-PUMA conditioned GFP+ donor BM in lethally irradiated congenic mice. Data represent mean +/− SD of n=8 recipient animals generated from 3 primary chimeras.

HSC depletion with CD117/LNP PUMA allows for bone marrow engraftment

HSC depletion in vivo was confirmed with i.v. injection of CD117/LNP-PUMA at 0.05 mg/kg in C57BL/6 mice, which showed a 71% and 58% decrease in the frequency of LSK and LT-HSC cells in BM isolates 6 days post-treatment, respectively (fig. S9B). 0.05 mg/kg mRNA was found to be the maximum tolerated dose. Animals treated with 0.15mg/kg, or more, CD117/LNP-PUMA displayed decreased activity, elevations in the AST/ALT, venous congestion of the lungs and liver, and mortality.

We tested in vivo CD117/LNP-PUMA HSC depletion as conditioning for HSCT. After confirming a liver-specific miRNA binding site (mir-122) could decrease expression in the liver (21) (fig. S10), we incorporated liver-specific miRNA binding sites for mir-122 into the 3’UTR of our PUMA mRNA cargo. mir-122 is expressed in vertebrate hepatocytes and can decrease expression of transgenes in hepatocytes. C57BL/6 recipients received 0.05 mg/kg mRNA CD117/LNP-PUMA-miR i.v. 7 days prior to infusion of 10⁶ GFP⁺ C57BL/6 BM cells. The level of engraftment was evaluated after 2 weeks and up to 16 weeks (endpoint) and confirmed progressive increase and stabilization of GFP⁺ Gr1⁺ cells and RBCs, as well as hematopoietic cells in the spleen (CD45⁺) and BM (Fig. 5F–H): 3.8% of BM LSK cells were donor. By comparison, C57BL/6 recipient mice not treated with CD117/LNP-PUMA conditioning failed to engraft donor cells. Secondary transplantation of the cells that engrafted with PUMA conditioning phenocopied the donors (Fig. 5I). This shows that in vivo targeting with CD117/LNP-PUMA effectively depleted HSC, allowing GFP⁺ BM cells to successfully engraft without need of chemotherapy or irradiation. These engraftment rates are consistent with those reported to be sufficient for the cure of SCID with healthy donor BM (22–24) and may overcome BMF syndromes.

Concluding Remarks

Our results suggest that LNP loaded with diverse mRNA cargos can access HSC in the murine BM niche in situ, with a single systemic injection. Delivery efficacy to long-term HSC in the BM niche is greatly increased by conjugation of a targeting moiety (anti-CD117 antibody). Here, we showed that LNP loaded with a Cre mRNA cargo can induce durable genome editing in long-term HSC ex vivo and in vivo at, or above, the levels reportedly required for cure of non-malignant hematopoietic disorders affecting the erythroid lineage with allogeneic or autologous gene modified cells (25, 26). This approach was translated to primary human cells, where we were able to achieve high rates of therapeutic base editing in hematopoietic cells from individuals with sickle cell disease. Additionally, we demonstrated for the first time that a genetic medicine, targeted LNP-mRNA, can leverage our understanding of HSC biology (Mcl-1 pathway dependence) to effect cellular state change in vivo with physiologic effects. We utilized this system to deplete HSC in vivo without genotoxic conditioning regimens that often result in pulmonary, liver, and reproductive toxicity (20, 27, 28). Although this conditioning approach requires additional refinement to reduce toxicity, such as modifications to restrict LNP tropism and/or further limit gene expression in unintended cells, this has the capacity to replace current myeloablation approaches. These findings may potentially transform gene therapy in two ways. First, the cure of monogenic disorders, including non-malignant hematopoietic disorders (hemoglobinopathies, congenital anemias or thrombocytopenias, and immunodeficiencies) and non-hematopoietic diseases (cystic fibrosis, metabolic disorders, myopathies) with a simple i.v. infusion of targeted genetic medicines. Second, effecting cell-type specific state changes in vivo with minimal risk could allow previously impossible manipulations of physiology. Such novel delivery systems may help translate the promise of decades of concerted genetic and biomedical research to treat a wide array of human diseases.

Data and materials availability:

All data are available in the main text or the supplementary materials. Requests for materials should be addressed to S.R. or H.P.