Systemic delivery of phagocytosis-shielded retroviral vectors enables in vivo HSC gene therapy for sickle cell disease

SUMMARY

Ex vivo hematopoietic stem cell (HSC) gene therapy is effective for non-malignant blood disorders including sickle cell disease (SCD), but requires hospitalization, ex vivo cell manipulation, and conditioning. Direct in vivo gene delivery could remove these barriers and widen access. Here, we explore phagocytosis-shielded lentiviral and alpha-retroviral vectors pseudotyped with the Baboon endogenous retrovirus glycoprotein variant BaEVRLess for in vivo HSC gene transfer. Vector injection into mobilized humanized mice yielded up to 8.8% gene marking in hCD45+ cells, which chemoselection enriched to 70% of hCD45+ cells and 54% of HSCs. Barcode analysis showed polyclonal reconstitution in over 90% of mice, which remained stable in secondary recipients. For SCD, we targeted BCL11A and ZNF410 selectively in erythroid cells using miRNA-embedded shRNAs to derepress gamma-globin, reaching therapeutically relevant levels of 61.5% of beta-like globins. In summary, BaEVRLess-pseudotyped lentiviral vectors enable clinically relevant in vivo gene transfer and fetal globin induction for SCD.

Graphical Abstract

In brief

Klatt et al. show that phagocytosis-shielded, BaEVRLess-pseudotyped retroviral vectors transduce human hematopoietic stem cells directly in vivo in mobilized humanized mice. Chemoselection enriches gene-marked cells, and erythroid-specific knockdown of BCL11A and ZNF410 derepresses fetal globin to therapeutically relevant levels for sickle cell disease.

INTRODUCTION

Hematopoietic stem cell (HSC) gene therapy has been successfully used to treat several devastating inherited blood disorders, including cerebral adrenoleukodystrophy, severe immunodeficiencies and sickle cell disease (SCD)^1–4. Despite these successes, the technologies and procedures used in these trials require complex ex vivo manufacturing of autologous cell products, are expensive and difficult to scale, and are associated with significant risks related to myeloablative conditioning. Direct in vivo genetic modification of HSCs through a simple injectable approach could address these shortcomings and potentially render gene therapy more accessible, cost-effective, and safer. This is particularly relevant for SCD, for which the global number of affected individuals is estimated to exceed 7 million, the majority residing in Africa and India^5–7.

Curative treatment options for SCD include allogeneic hematopoietic stem cell transplantation (HSCT) from a matched donor, available to only approximately 15% of patients and carrying significant risks of graft-versus-host disease (GvHD)^8–11, or ex vivo HSC gene therapy using autologous cells. Two FDA-approved ex vivo gene therapies for SCD are available in the United States: Lyfgenia^®, based on lentiviral transfer of a T87Q-globin transgene, and Casgevy^®, which disrupts the BCL11A erythroid enhancer via genome editing^12,13. Additional gene therapy strategies are in clinical development, including a Phase II multicenter study employing a lentiviral vector encoding microRNA-adapted shRNA (shmiR) targeting BCL11A (NCT05353647)^4,14,15. Downregulation of BCL11A reactivates the fetal γ-globin gene and downregulates the mutated β^S-globin, largely attenuating the hematologic manifestations of SCD. Erythroid-specific regulatory elements derived from the β-globin locus control region (LCR) confer tight transcriptional control, restricting shmiR expression to erythroid cells. This strategy induces protective fetal hemoglobin (HbF) in erythrocytes, while minimizing the risk of off-target effects in other cell types. Such transcriptional specificity could also be advantageous in the context of in vivo gene delivery, where it could complement or potentially substitute for delivery specificity conferred by the delivery system.

To successfully transduce HSCs in vivo, viral vectors need to (1) efficiently transduce non-dividing cells, (2) be resistant to the innate immune system, (3) have safe integration profiles and (4) not be mitigated by any pre-existing immunity that would be a barrier for wide adoption. Several promising delivery platforms have been evaluated across diverse settings and model systems, including helper-dependent adenoviral vectors (HDAd), lipid nanoparticles (LNPs), lentiviral vectors, and engineered virus-like particles (eVLPs)^16–20. HDAd vectors have entered clinical testing for in vivo HSC gene therapy in a Phase 1/2 study to treat X-linked chronic granulomatous disease (X-CGD; NCT06876363, Ensoma, Inc.). This trial includes chemoselection to enrich gene-modified cells and has demonstrated a promising early safety profile²¹. LNPs encapsulating mRNA have been evaluated in murine and non-human primate (NHP) models. While efficient for delivery of small payloads such as Cre recombinase, durable editing of therapeutic targets in long-term HSCs using complex gene-editing mRNA cargoes remains to be demonstrated^17,18 and another key limitation of current LNP systems is substantial off-target biodistribution, particularly to the liver, lung, and spleen^17,18,22. One notable exception is data from Tessera Therapeutics, reporting up to 74% in vivo gene editing of HSCs in NHP using LNPs, though detailed methodology and data remain limited in the public domain²³.

Recent reports position retroviral vectors and retroviral-derived VLP systems as promising candidates for in vivo HSC gene therapy, offering flexibility to deliver integrating nucleic acids, mRNA or genome editors as ribonucleoprotein (RNP) complexes^{19,20,24–28}. VSVg-pseudotyped lentiviral vectors decorated with the CD47 antigen encompassing a “Don’t eat me” signal achieved single-digit levels of long-term gene transfer in newborn mice, and to a lesser extent in mobilized juvenile animals¹⁹. Baboon endogenous retrovirus (BaEV) glycoprotein-pseudotyped lentiviral vectors achieved comparable efficiencies in humanized mouse models, with a marked reduction in vector sequestration in the liver and other off-target organs²⁰.

Here, we extend these findings by investigating the potential of phagocytosis-shielded lentiviral and alpha-retroviral vectors pseudotyped with the BaEV glycoprotein for in vivo HSC gene therapy with a future goal of treating SCD. We employ the clinically validated strategy of shmiR-mediated downregulation of BCL11A and ZNF410 to redirect the fetal-to-adult hemoglobin switch. Lentiviral vectors may be particularly suited for gene transfer into quiescent cells²⁹, whereas alpha-retroviral vectors offer a potentially favorable safety profile attributable to their near-random genomic integration pattern and absence of vector-encoded splice donor/acceptor signals^30–32. Alpha-retroviral vectors are also compatible with stable producer cell lines, an important feature with respect to scalability and cost-effectiveness^33,34. Both vector types can be efficiently pseudotyped with BaEV, which engages the neutral amino acid transporters ASCT1 and ASCT2, receptors that are broadly expressed across many cell types and confer efficient transduction of unstimulated, quiescent HSCs in vitro^35,36 and in vivo²⁰. Because BaEV exhibits tropism restricted to human and non-human primate cells, we employed humanized mouse models to evaluate in vivo transduction of human HSCs. To achieve therapeutically relevant HbF levels, we incorporated a chemoselectable MGMT-P140K transgene cassette, enabling enrichment of gene-modified HSCs through low-dose pharmacological selection. In vivo HSC transduction was validated by serial transplantation into secondary recipients, and clonal diversity was confirmed by genetic barcoding. This strategy yielded greater than 60% therapeutic γ-globin expression, establishing the translational potential of in vivo BaEV-pseudotyped vector delivery combined with chemoselection as a treatment paradigm for SCD.

RESULTS

BaEVRLess-pseudotyped alpha-retroviral vectors demonstrate serum-resistance and efficiently transduce CD34+ HSPCs

With the aim of identifying a suitable retroviral vector system for in vivo gene delivery into HSCs, we compared envelope glycoproteins derived from BaEV, RD114 and the commonly used VSVg for their ability to transduce quiescent HSCs and their stability in human serum. We tested two different variants of the BaEV envelope, BaEVTR and BaEVRLess, which both utilize ASCT1 and ASCT2 as entry receptors³⁵. In contrast to the BaEV envelopes, the feline endogenous retrovirus derived glycoprotein RD114TR binds exclusively to ASCT2, while VSVg engages the broadly expressed LDL receptor (Figure 1A)^37,38. When used to pseudotype alpha-retroviral vectors^31–33, the BaEVRLess envelope achieved ~1-log higher titers compared to BaEVTR and RD114TR, and slightly lower titers than VSVg (Figure 1B). We subsequently used an MOI of 5 to transduce pre-stimulated CD34+ HSPCs in the presence or absence of transduction enhancers and observed 70% to 80% gene transfer with BaEVRLess in the presence of the soluble transduction enhancer Vectofusin^®−1 or precoated Retronectin^®, respectively, outperforming VSVg (Figure 1C and Supplement Figure S1)^39–41. Vectofusin^®−1 was used as transduction enhancer in all subsequent experiments, as it can be premixed with the viral vectors prior to in vivo delivery. Next, we tested VSVg and BaEVRLess for the transduction of freshly thawed quiescent mobilized peripheral blood-derived (mPB) CD34+ HSPCs⁴². CD34+ HSPCs were transduced for two hours in cytokine- and serum-free medium, then washed to remove remaining viral vectors and cultured for three days in cytokine-containing medium prior to analysis. Quiescent HSCs are known to be refractory to VSVg-mediated transduction due to the low expression of its cognate receptor, LDL-R⁴³. Consistent with this, VSVg achieved only 2.4% transduction. In contrast, BaEVRLess-pseudotyped virus achieved 27.6% gene marked cells (Figure 1D), underscoring its suitability for quiescent HSC targeting. To assess serum stability, vector preparations were diluted 1:5 in fresh human serum or heat-inactivated serum for one hour prior to titration on PLB985 cells. VSVg-pseudotyped vectors were partially neutralized in the fresh serum. In contrast, BaEVRLess was resistant to complement-mediated inactivation and exhibited a 2.4-fold increase in transduction after incubation in fresh serum compared to heat-inactivated serum (Figure 1E). These findings confirm complement sensitivity of VSVg-pseudotyped vectors, which is absent with BaEVRLess. In summary, BaEVRLess combines robust titers, efficient transduction of quiescent CD34+ HSPCs, and resistance to human serum, supporting its selection for in vivo applications.

Figure 1: Comparison of different envelope glycoproteins for the transduction of quiescent HSPCs.

(A) Schematic representation of different viral glycoproteins, VSVg, BaEV derivatives BaEVTR and BaEVRLess and RD114TR and their respective binding receptors, LDL-R, ASCT1+ASCT2 and ASCT2.

(B) Titration of alpha-retroviral vectors pseudotyped with different viral envelopes on HEK293T cells. I.P.= infectious particles.

(C) Frequency of gene marking of transduced CD34+ HSPCs. Pre-stimulated CD34+ HSPCs were transduced in the absence or presence of Vectofusin^®−1 as transduction enhancer with alpha-retroviral vectors pseudotyped with different viral glycoproteins at an MOI 5.

(D) Frequency of gene marking of quiescent CD34+ HSPCs. Freshly thawed, unstimulated CD34+ HSPCs were transduced in the absence of cytokines for 2h at an MOI 100 using VSVg- or BaEVRLess-pseudotyped alpha-retroviral particles and Vectofusin^®−1 as transduction enhancer.

(E) Frequency of gene marked PLB985 cells in a neutralization assay. VSVg- and BaEVRLess-pseudotyped alpha-retroviral particles were incubated for 1 hour in fresh or heat-inactivated human serum prior to transduction of PLB985 cells.

In all experiments, the transduction rate was measured 3–4 days after transduction.

Statistics: n = 3–6 using 3 independent vector batches, bars represent mean±SD, *p≤0.05, **p<0.01, ****p<0.0001 by one-way ANOVA (B), two-way ANOVA (C and E) or Student’s t-test (D).

Both lentiviral and alpha-retroviral vectors transduce CD34+ HSPCs in vivo

Lentiviral and alpha-retroviral vectors are both promising candidates for in vivo HSC gene transfer and are compatible with the BaEVRLess envelope, but differ in their ability to transduce quiescent cells, their safety features, and manufacturing scalability. Due to the species-specificity of the BaEVRLess envelope, which targets only human and primate cells, we employed the NOD.B6.SCID Il2rγ−/− kit (W41/W41) (NBSGW) humanized mouse model⁴⁴ to study in vivo transduction of HSPCs using both vector platforms. In pilot studies, mice were injected with 5 × 10⁵ CD34+ HSPCs via one tail vein and five to ten minutes later received the alpha-retroviral eGFP-reporter vector aRV.SEW at three different MOIs (4, 20, 100) via the contra-lateral tail vein (Figure 2A). Sixteen weeks after co-injection, we observed that an MOI 100 was required to achieve clearly detectable gene marking levels in peripheral blood and bone marrow (Figure 2B–C).

Figure 2: Evaluation of MOI, CD47 expression and type of retroviral vector on the transduction of CD34+ HSPCs in vivo using co-injection experiments.

(A) Schematic representation of experimental design showing co-injection of 500k CD34+ HSPCs and retroviral vectors into contra-lateral tail veins in NBSGW mice. Peripheral blood was monitored for 16 weeks before the bone marrow was assessed for gene marking.

(B and C) Alpha-retroviral particles were co-injected at MOI 4, 20 and 100. The frequency of gene marked hCD45+ cells was assessed in peripheral blood (PB) (B) and bone marrow (BM) (C) 16 weeks after injection.

(D) Alpha-retroviral particles produced from CD47 wild type (WT) or CD47 overexpressing (OE) packaging cells were co-injected at MOI 100. Gene marking of hCD45+ cells is shown over time in PB and BM.

(E) Pooled analysis of in vivo gene transfer rates using alpha-retroviral particles derived from CD47 WT and CD47 OE producer cell lines across different time points.

(F-I) Competitive co-injection experiment of alpha-retroviral and lentiviral vectors expressing fluorescent reporters at MOI 50 each.

(F) 100k CD34+ HSPCs were transduced in vitro using the same mixture of alpha-retroviral and lentiviral vectors that was applied in vivo to measure the exact mixing ratio for normalization. Gene marking was measured three days after transduction.

(G) Gene marking in PB at weeks 6, 12 and 16 after co-injection.

(H) Sixteen weeks after co-injection, BM was harvested, and gene marking was measured in different lineages.

(I) Gene marking frequencies were normalized to the in vitro gene marking rates and are represented as the ratio of lentiviral gene marking over alpha-retroviral gene marking in different samples.

Statistics: n=3–9 from 2–3 independent experiments, bars represent mean±SD, *p≤0.05, **p≤0.01 by one-way ANOVA (B and C), two-way ANOVA showing comparison of row means only (D, G and H) or Student’s t-test (E and F).

We next investigated whether CD47 overexpression on viral particles could enhance in vivo HSC gene transfer by limiting non-specific uptake and sequestration of viral particles⁴⁵. CD47 engages SIRPα, which is expressed on myeloid cell populations and functions as “Don’t eat me” signal suppressing phagocytosis⁴⁶. Since NBSGW mice harbor a Sirpa polymorphism that increases binding affinity for human CD47⁴⁷, elevated CD47 expression on viral particles should reduce clearance by both murine and human phagocytes. We compared BaEVRLess-pseudotyped viral vectors produced in HEK293T cells with wild type (WT) and overexpressed (OE) CD47 levels³³. Overexpression of CD47 on viral vectors increased gene transfer into human CD45+ cells, most prominently in the bone marrow at 16 weeks post injection, where gene marking was 5-fold higher (Figure 2D–E). Consequently, overexpression of CD47 on the viral particles was used in subsequent experiments.

To directly compare the in vivo performance of lentiviral or alpha-retroviral vectors, we performed competitive co-injection experiments in NBSGW mice. A 1:1 mix of color-coded lentiviral and alpha-retroviral vectors was injected at an MOI of 50 each. A small aliquot of CD34+ HSPCs were simultaneously transduced with the same vector mix ex vivo to define the exact ratio of lentiviral and alpha-retroviral vectors that were injected into each animal. Ex vivo, mean gene marking rates were 16.7% for lentiviral vectors and 13.1% for alpha-retroviral vectors (Figure 2F). At week 16, the gene marking in peripheral blood was 0.4% for the alpha-retroviral vectors and 1.2% for the lentiviral vectors (Figure 2G). Within each vector group, the gene marking was similar across different hematopoietic lineages in the bone marrow (Figure 2H, Figure S2A–B; gating strategy is shown in Figure S3). After normalizing gene marking based on the ex vivo transduction rates, no significant differences were observed between alpha-retroviral and lentiviral vectors (Figure 2I). In summary, both alpha-retroviral and lentiviral vectors have the capacity to transduce CD34+ HSPCs in vivo at comparable frequencies.

AMD3100 and Gro-beta efficiently mobilize humanized NBSGW mice

After confirming that phagocytosis-shielded BaEVRLess-pseudotyped viral vectors can transduce CD34+ HSPCs in vivo, we extended these studies to NBSGW mice humanized eight weeks prior to vector administration, a time point at which HSPCs predominantly reside in the bone marrow medullary cavity, mirroring physiological conditions in humans. Transduction of HSCs has been reported to be enhanced after mobilization from the bone marrow into the blood stream prior to infusion of a viral vector⁴⁸. To identify which mobilization regimen most efficiently mobilizes humanized NBSGW mice, we concurrently tested three protocols: AMD3100 alone, AMD3100+G-CSF and AMD3100+Gro-beta (Figure 3A)^49,50. Human peripheral blood (PB) chimerism increased in all test groups from 8.5% baseline, reaching 60.2% hCD45+ cells for the AMD3100+Gro-beta group, 30.5% using AMD3100 alone or 18.7% hCD45+ with AMD3100+G-CSF (Figure 3B). This increase of human cells into the peripheral blood mainly consisted of granulocytes, especially in the AMD3100+Gro-beta group (Figure 3C). HSC mobilization, defined as Lin⁻ (hCD19⁻, hCD33⁻, hCD3⁻) hCD34+ HSPCs, increased 8.6-fold in the AMD3100+Gro-beta group, reaching ~1.3% of total CD45+ cells (Figure 3D). The absolute count of Lin- hCD34+ cells in this group significantly increased up to 6347 cells per mL of mobilized peripheral blood, representing a 11.3-fold increase of HSPCs compared to unmobilized peripheral blood (Figure 3E). Efficient HSPC mobilization using AMD3100+Gro-beta was confirmed in colony-forming unit (CFU) assays, which on average showed 168 CFUs (CFU-G + BFU-E) compared to 12 colonies per mL using unmobilized peripheral blood (Figure 3F). Approximately 71% of hCD45+ hCD34+ cells remained in G0, consistent with the predominantly quiescent state of mobilized HSPCs (Figure 3G). Injection of BaEVRLess-pseudotyped lentiviral particles at peak of mobilization resulted in 1.7-fold higher gene marking rates one week after injection compared to mice that did not undergo HSC mobilization (Figure 3H). The higher gene transfer rates were maintained over time, demonstrating significantly higher gene marking 6 weeks after viral vector injection. Additional mobilization regimens, including simultaneous versus staggered AMD3100+Gro-beta administration (Figure S4A–C) or the addition of the VLA-4 inhibitor BIO5192^48,51 (Figure S4E–G), did not further increase Lin-hCD34+ cell frequency or absolute numbers. Notably, staggered AMD3100+Gro-beta treatment slightly increased circulating CFUs compared to simultaneous administration (Figure S4D). Mobilization of human cells from the bone marrow resulted in a reduction of the percentage of mouse cells in the peripheral blood and did not increase the absolute number of circulating mouse cKit+ mCD45+ cells (Figure S4H–K). In conclusion, staggered AMD3100+Gro-beta administration achieved the most effective HSPC mobilization in humanized NBSGW mice.

Figure 3: Optimization of a mobilization regimen for humanized NBSGW mice.

(A) Schematic representation of the different mobilizing agents and dosing schedule used in the experiment. NBSGW mice were mobilized eight weeks after humanization.

(B) Human chimerism in the peripheral blood (PB) of treated mice before and after mobilization. Pre-Mob = Pre-Mobilization

(C) Frequency of granulocytes in PB of treated mice.

(D) Frequency of Lin- hCD34+ HSPCs of total (mouse+human) hCD45+ in PB of treated mice.

(E) Cell counts of Lin- hCD34+ HSPCs per mL of PB after mobilization quantified using counting beads.

(F) Progenitor content in PB was measured by colony-forming unit (CFU) assay before and after mobilization. Colonies in methylcellulose were quantified 14 days after seeding.

(G) Cell cycle state of mobilized human HSPCs was analyzed in mobilized peripheral blood (mPB) and bone marrow (BM). Cells of mobilized mice were stained with pyronin Y and Hoechst to assess the frequency of different cell cycle phases.

(H) Gene marking of hCD45+ cells in PB of mice treated with or without mobilization prior to BaEVRLess-pseudotyped vector injection over time.

Statistics: n=3–9 from three independent experiments, data of all groups was pooled for the “pre-mobilization” time point, bars represent mean±SD, *p≤0.05, **p<0.01, ***p<0.001, ****p<0.0001 by one-way ANOVA (B-E) or two-way ANOVA (F-H).

Serial transplantation confirms in vivo transduction of primitive HSPCs

After optimization of the mobilization regimen, we tested BaEVRLess-pseudotyped lentiviral vectors for in vivo gene transfer in pre-engrafted and mobilized NBSGW mice. We chose lentiviral vectors due to their trend toward higher gene marking rates in the HSPC compartment in co-injection experiments (Figure 2H–I). At the peak of mobilization ~4×10⁷ viral particles containing the pCCL.BC-PMEW vector were injected intravenously (Figure 4A). This vector encodes a MGMT-P140K cassette conferring chemoresistance to the alkylating drug regimen of BCNU and O6-Benzylguanine (O6-BG), thereby enabling enrichment of gene modified cells^48,52–54. BCNU selection was initiated four weeks after in vivo transduction and consisted of three cycles of escalating doses (2.5, 5 and 10 mg/kg BCNU) combined with 20 mg/kg O6-BG. Gene marking in peripheral blood hCD45+ increased from 2.2% (range 0.8–3.7%) one week post vector injection to ~16% post chemoselection at week 10 (Figure 4B) and was slightly higher in myeloid lineages during chemoselection, correlating with altered lineage distribution (Figure S5A–B)⁵⁴. Chemoselection significantly depleted untransduced hCD45+ cells, whereas absolute numbers of gene marked hCD45+ (Week 4: 0.6k vs Week 10: 1.3k), myeloid and lymphoid cells slightly increased during selection (Figure 4C, Figure S5C). Absolute cell numbers of murine CD45+ cells remained unchanged in the peripheral blood over time (Figure S5D). Bone marrow gene marking in primary recipients reached 8.7% two weeks after the final round of selection (Figure 4D), corresponding to 90k gene marked cells in both hind limbs (Figure 4E). At 10 weeks post gene transfer, the bone marrow from primary mice was transplanted into secondary recipients for an additional 16 weeks without further selection. The gene marking in the bone marrow of the secondary recipients averaged 58.5% (Figure 4F). The increase in gene marking compared to primary recipients suggests that selection of primitive HSPCs in the primary recipients was not fully reflected in week 10 peripheral blood and bone marrow analyses, likely reflecting reduced BCNU sensitivity and slow turnover of mature postmitotic cells⁵⁵. Lineage representation normalized in secondary recipients in the absence of selection (Figure S5E). Although the percentage of gene marked human cells was high, absolute engraftment in secondary recipients was low (<1.0k eGFP+ cells) likely due to the previously described limited serial transplantation efficiency in the NBSGW mouse model (Figure 4G, Figure S5F)⁴⁴. Monitoring of complete blood counts (CBCs) showed minimal difference between chemoselected and untreated control mice in white blood cell (WBC) and platelet counts (PLT) (Figure S5G+I). Red blood cell (RBC) counts were reduced, a typical feature after successful humanization of NBSGW mice caused by the displacement of murine progenitor cells by human cells (Figure S5H)⁵⁶. This reduction was more pronounced in the chemoselected group but recovered following cessation of chemoselection.

Figure 4: In vivo gene therapy and serial transplantation of in vivo selected HSPCs.

(A) Schematic representation of the in vivo gene therapy protocol including transplantation of bone marrow into secondary recipients with steps numbered 1–4. The lentiviral vector used for transduction expresses the mutant methyltransferase MGMT-P140K for selection, eGFP to track transduced cells and contains a genetic barcode to study the clonality of transduced cells (lower part of diagram).

(B) Frequency of gene marked and untransduced cells in the peripheral blood (PB) of treated mice over time. Numbers indicate the average.

(C) Absolute cell counts of gene marked and untransduced hCD45+ cells in total PB.

(D) Frequency of gene marked cells in the bone marrow (BM) of the primary recipients 10 weeks after in vivo gene transfer.

(E) Absolute cell number of gene marked cells in the BM of both hind limbs.

(F) Frequency of gene marked cells in the BM of secondary recipients.

(G) Absolute cell counts of gene marked hCD45+ cells in the BM (both hind limbs only) of secondary recipients.

(H) Number of unique barcodes identified in the BM of primary and secondary recipients.

(I) Shannon diversity index represents the barcode clonality in primary and secondary recipients.

(J) Graphical representation of the barcode clonality for selected primary and secondary recipients showing the overlap of identical barcodes in donors and recipients.

All barcode data in primary mice is post BCNU selection.

Statistics: n=3–12 from three independent experiments, depicted are mean±SEM (B-G) or mean±SD (H-I), ****p<0.0001 by Student’s t-test (H and I).

Genetic barcodes embedded in the lentiviral vector were used to assess the clonality of transduced cells, with unique barcodes identified in the bone marrow serving as a surrogate for individually transduced cells (Figure 4A). Primary recipients displayed on average 1280 unique barcodes, while secondary recipients maintained 523, confirming successful in vivo transduction of definitive HSCs (Figure 4H). Clonal diversity was calculated using the Shannon index, which considers unique barcode numbers and their relative abundance. A Shannon index value of 2.2 in primary mice indicates moderate polyclonality and balanced output⁵⁷. Secondary recipients retained clonal diversity with a Shannon index of 1.7 after serial transplantation (Figure 4I). More than 25% of barcodes detected in primary mice were also recovered in secondary recipients, demonstrating self-renewal of transduced cells. Interestingly, dominant clones differed between primary and secondary recipients (Figure 4J, Supplemental Table S1). In summary, BaEVRLess-pseudotyped lentiviral vectors can transduce long-term repopulating cells in vivo in humanized mice after mobilization. Low-dose BCNU selection increased gene marking in hCD45+ cells, reaching 58.5% in secondary recipients, while preserving diverse clonality^48,54,58.

Higher donor chimerism and efficient HSPC mobilization increase gene marking rates and clonal diversity

We next sought to identify key parameters affecting gene marking and clonality after in vivo gene transfer, focusing on human engraftment levels, HSC mobilization efficiency, and the relative multiplicity of infection (MOI). NBSGW mice were humanized with increasing CD34+ cell doses (62.5K, 125K, 250K, 500K). As expected, peripheral blood chimerism increased dose-dependently from 1.3% (62.5K group) to 15.0% (500K group) eight weeks post-transplantation and rose to 40% and 82%, respectively, after mobilization with AMD3100+Gro-beta (Figure 5A and S6A). Mobilized Lin- hCD34+ HSPC counts increased from 7793 to 22580 cells/mL in the lowest and the highest cell dose groups, respectively (Figure 5B), establishing a positive correlation between cell dose, human cell engraftment, WBC counts and HSPC mobilization (Figure S6B–C). All animals received 4×10⁷ BaEVRLess-pseudotyped pCCL.BC-PMEW lentiviral vector particles during peak mobilization (illustrated in Figure 4A), resulting in different relative MOIs due to varying humanization levels. Gene marking one week post-transduction positively correlated with the number of mobilized Lin- hCD34+ HSPCs (Figure 5C) and inversely correlated with the MOI, demonstrating that higher circulating HSPC frequencies at the time of vector administration enhance transduction efficiency. Chemoselection initiated four weeks post vector injection increased gene marking from <1% at week 1 in peripheral blood to ~13–32% at week 13 in bone marrow hCD45+ myeloid and lymphoid cells (Figure 5D and Figure S6D–E). As expected, the absolute number of gene marked cells in the bone marrow was slightly higher in mice receiving higher cell doses (Figure 5E and Figure S6F). Sustained gene marking in short-lived myeloid cell populations supports the conclusion of successful gene transfer into repopulating HSPCs. Barcode sequencing identified between 107 to 4460 unique barcodes in individual mice and a high degree of polyclonality in most animals (Figure 5F). Initial gene marking strongly correlated with the number of unique barcodes long term (R²=0.9, Figure 5G). One mouse (125K group) showed oligoclonal hematopoiesis, and dominant clones (>20% contribution) were observed in three additional animals. Except for the single mouse with oligoclonality, Shannon index values of >4 indicated balanced polyclonal output (Figure 5H). Chemoselection had limited hematologic impact, with no WBC changes, transient RBC reduction, and delayed moderate PLT reduction compared to healthy controls (Figure S6G–I). A complete correlation analysis of all parameters is found in Supplemental Figure S7. Overall, the number of mobilized HSPCs at vector administration was a key determinant of initial and long-term gene marking efficiency and clonal diversity.

Figure 5: Effect of the level of human chimerism on mobilization of HSCs, gene transfer rates and clonality in vivo.

(A) Mice were humanized with increasing doses of CD34+ HSPCs (62.5k, 125k, 250k and 500k) prior to in vivo gene therapy and in vivo selection using BCNU/O6-BG. Human chimerism in the peripheral blood (PB) before (Pre-Mob) and after (Post-Mob) mobilization.

(B) Quantification of Lin- hCD34+ HSCPs per mL of mobilized PB (mPB) in treated mice.

(C) Pearson’s correlation of HSPCs mobilization efficiency and gene marking; Lin- hCD34+ per mL mPB versus the gene marking frequency in the peripheral blood one week after mobilization and vector injection.

(D) Frequency of gene marked hCD45+ cells in PB and bone marrow (BM) of treated mice over time. Numbers indicate the mean gene marking of the highest group.

(E) Absolute cell counts of gene marked hCD45+, myeloid, lymphoid and HSPCs in the BM (both hind limbs) across the different cell dose groups.

(F) Graphical representation of the abundance of individual barcodes at week 13 in the BM of treated mice. (G) Pearson’s correlation showing the association of barcode abundance in the BM long-term and early gene transfer rates in PB. Mice with more than 220 unique barcodes (UMI) are included.

(H) Pearson’s correlation of the number of unique barcodes compared to the Shannon diversity index.

Statistics: n=3–4 (or pooled n=9–16), data from all groups is pooled for the Pearson’s correlations (C, G and H), depicted are mean±SD, *p≤0.05, ****p<0.0001 by two-way ANOVA comparing row means only (A, D and E) or one-way ANOVA (B).

In vivo gene therapy using a therapeutic SCD vector leads to robust fetal hemoglobin induction

Finally, we tested a therapeutic vector for in vivo gene therapy of SCD (Figure 6A). The lentiviral vector contains constitutively expressed MGMT-P140K and eGFP cassettes linked via a 2A element and a reverse erythroid-specific expression cassette encoding two shmiRs targeting BCL11A and ZNF410, key repressors of γ-globin^59,60. NBSGW mice humanized with 500k CD34+ HSPCs were mobilized 8 weeks later with AMD3100+Gro-beta and injected with 3 × 10⁷ lentiviral vectors at peak mobilization. One week after vector injection, gene marking in peripheral blood hCD45+ averaged ~4% (up to 8.8%) and increased to 45.7% after chemoselection at week 10 and 69.5% at week 16 (Figure 6B). As observed previously, untransduced cells were depleted during chemoselection, whereas absolute cell counts of gene marked cells remained stable after selection that began at week 4 (Figure 6C). Lineage analysis confirmed comparable gene transfer rates in myeloid and lymphoid lineages (Figure S8A–C). Bone marrow was analyzed at weeks 10 and 16 post vector injection. Gene marking increased from 5.7% at week 10 to 69.1% at week 16 (Figure 6D), correlating with a substantial increase in the absolute number of gene marked cells at week 16 (Figure 6E). This is consistent with selection effects requiring extended follow up after chemotherapy treatment as noted above during secondary transplantation. At week 16, gene marking reached 53.7% on average in immunophenotypically defined HSPCs and was comparable across bone marrow lineages (Figure 6F–G, Figure S8D). Because CD235a+ erythroid cells generated in vivo in this model were insufficient for reliable globin analysis (Figure S8E), harvested bone marrow cells from fully engrafted mice were differentiated in vitro. γ-globin mRNA levels in erythroid cells ranged from 34.9% (week 10) to 61.5% (week 16) (Figure 6H). CBCs showed no difference between treated and untreated animals (Figure S8F–H). Murine contribution to gene marking was excluded by showing the absence of eGFP expression in mCD45+ cells, confirming human cell specificity of the BaEVRLess envelope (Figure S9A–B). Taken together, these data demonstrate the feasibility of in vivo gene therapy to induce therapeutically relevant levels of fetal hemoglobin for SCD.

Figure 6: In vivo gene therapy using a therapeutic vector for the treatment of sickle cell disease.

(A) Schematic representation of the in vivo gene therapy protocol and in vivo selection using O6-BG/BCNU with steps 1–4 and the lentiviral vector expressing the mutant methyltransferase MGMT-P140K for selection, eGFP to track transduced cells and two miRNA embedded shRNA cassettes to knockdown BCL11A and ZNF410 to induce fetal globin for the treatment of sickle cell disease (shown below).

(B) Gene marking frequency in the peripheral blood (PB) of treated mice over time.

(C) Absolute cell counts of gene marked and untransduced hCD45+ cells in PB over time.

(D) Frequency of gene marked and untransduced hCD45+ cells in the bone marrow (BM) of treated mice 10 and 16 weeks after in vivo gene therapy.

(E) Absolute cell counts of gene marked and untransduced hCD45+ cells in BM (both hind limbs only) at week 10 and 16.

(F) Gene marking frequency among different lineages in BM at week 10 and 16. Numbers indicate mean gene marking rate.

(G) Absolute cell counts of gene marked cells among different lineages at week 10 or 16 after in vivo gene therapy in the BM (both hind limbs only) of treated mice.

(H) Expression of gamma-globin of in vitro differentiated BM samples collected on week 10 or 16 after in vivo gene therapy.

Statistics: n=3–11 from three independent experiments, depicted are mean±SD, ***p<0.001, ****p<0.0001 by two-way ANOVA comparing row means only (B-E) or one-way ANOVA (H).

DISCUSSION

We demonstrate that BaEVRLess-pseudotyped lentiviral and alpha-retroviral vectors enable in vivo transduction of long-term repopulating human HSPCs in a humanized mouse model. MGMT-P140K-mediated chemoselection can enrich the small initial pool of transduced cells to achieve therapeutically relevant levels of fetal hemoglobin induction for SCD^61,62. A single intravenous injection of 1–2 × 10⁹ infectious lentiviral particles per kg bodyweight in humanized mice yielded initial gene marking of up to 8.8% one week after viral vector injection using the therapeutic vector, with most animals in the 1–4% range. MGMT-P140K mediated chemoselection with BCNU led to on average 33-fold enrichment per animal, with post-selection gene marking of ~70% in hCD45+ peripheral blood and bone marrow cells and reaching over 50% in HSPCs while maintaining robust polyclonality. Sustained gene marking in short-lived myeloid cell populations and serial transplantation experiments confirmed transduction of bona fide long-term repopulating HSCs.

Our findings corroborate and extend recent findings using retroviral based delivery systems^19,20. Milani et al. exploited naturally elevated circulating HSC numbers in newborn mice to achieve single-digit long-term gene transfer with phagocytosis-shielded VSVg-pseudotyped lentiviral vectors and demonstrated that mobilization extends this window into the juvenile period¹⁹. Botchkarev et al. used BaEVTR-pseudotyped VLPs in NBSGW mice early after humanization and demonstrated in vivo editing of phenotypically defined HSPCs²⁰; however, their functional hemoglobin readout was assessed at a single early timepoint using F-cell frequency, a low-threshold metric since erythrocytes are classified as F-cells at HbF levels as low as ~2–3 pg/cell, well below the ~10 pg/cell (~30% of β-like globins) estimated to be required to prevent sickling^62–65. By contrast, we assessed γ-globin mRNA in erythroid cells at weeks 10 and 16 post-transfer, observing levels of 34.9% and 61.5%, respectively, with the progressive increase consistent with ongoing enrichment of transduced primitive HSCs by chemoselection, a pattern mirrored by substantially higher bone marrow gene marking in secondary transplant recipients. Notably, shmiR-mediated γ-globin induction is mechanistically advantageous over delivery of a functional β-globin chain for SCD, since reactivation of the fetal hemoglobin switch simultaneously suppresses the sickling βS-globin chain and maintains α:β globin chain balance, such that therapeutic benefit is achieved at low vector copy numbers per cell^4,14,62,66. Gene transfer rates and clonal diversity in our system are also similar to results achieved with engineered HDAd vectors in mobilized mouse models⁵⁰, despite HDAd requiring infectious doses more than three orders of magnitude higher, likely because HDAd integration requires co-delivery of two distinct viral particles whereas the retroviral system packages all components within a single particle.

CD47 overexpression on viral particles enhanced in vivo HSC transduction, consistent with phagocytosis shielding described by Milani et al. and earlier non-human primate studies⁴⁵. The magnitude of this benefit was comparable to reports of VSVg-pseudotyped vectors. Both lentiviral and alpha-retroviral vectors performed comparably after normalization based on ex vivo transduction rates. While lentiviral vectors showed a modest trend toward higher HSPC gene marking, alpha-retroviral vectors offer favorable integration profiles with reduced insertional mutagenesis risk and compatibility with stable producer cell lines potentially relevant to scalable manufacturing^30,33. Ultimately, vector choice will depend on the specific therapeutic context.

A relevant finding of this study was the positive correlation between mobilized Lin⁻CD34⁺ HSPC numbers at the time of vector injection and both initial gene marking and downstream clonal diversity. This directly addresses an open question in the field: Botchkarev et al. propose transduction primarily within the bone marrow niche²⁰, while Milani et al. establish a connection between HSC gene transfer and elevated circulating HSC numbers¹⁹. Our data indicate that circulating HSPC frequency is a primary rate-limiting variable, positioning mobilization protocol optimization as impactful as vector engineering for improving outcomes. This has direct implications for clinical translation, given that AMD3100-based mobilization is already in routine clinical use⁴⁹.

Initial in vivo gene marking rates remain below the 20–30% correction threshold required for meaningful clinical benefit in SCD^65,67,68, necessitating an enrichment step. MGMT-P140K chemoselection with BCNU achieved this threshold while preserving polyclonal hematopoiesis, with Shannon diversity indices above 4 in the majority of animals. Although clinically validated in other settings^21,69,70, BCNU-based selection carries risks including secondary malignancies and infertility^71,72. Future strategies should therefore prioritize non-genotoxic alternatives, such as antibody-drug conjugate-based selection targeting epitope-edited surface antigens^73,74, with the ultimate goal of improving primary transduction rates sufficiently to eliminate selection entirely, since any selection strategy may increase the long-term risk of clonal outgrowth^75–78.

An alternative enrichment-independent strategy is to improve on-target delivery through receptor-directed vector targeting. Engineered envelopes targeting HSC surface markers such as cKit or CD90 have demonstrated promising HSPC specificity in vitro and in vivo^20,79–81, and the CD133-targeted VLP system reported by Botchkarev et al. represents a recent proof of concept for receptor-directed in vivo HSPC editing²⁰. However, exclusive targeting of the most primitive long-term HSC compartment may paradoxically delay clinical benefit, since long-term HSCs contribute to steady-state hematopoiesis only slowly^82–84. Broader targeting of the HSPC hierarchy, including short-term HSCs and multipotent progenitors, may therefore be preferable for diseases requiring rapid production of corrected effector cells, as modification of more committed progenitors would more quickly translate into circulating corrected fully differentiated cells^85,86. The BaEVRLess-based approach described here is inherently non-selective, transducing HSPCs broadly via the widely expressed ASCT1 and ASCT2 receptors. In the absence of delivery-level cell type specificity, an additional layer of control is required to restrict therapeutic payload activity to the intended target population, achieved here through erythroid-specific regulatory elements derived from the β-globin LCR that confine shmiR expression to erythroid cells, limiting off-target effects in non-erythroid lineages where BCL11A plays essential roles^14,87.

Limitations of the study

This study has several limitations. As discussed above, therapeutically relevant gene marking required BCNU-based chemoselection, which acts on a limited founder population. Although Shannon diversity indices indicated balanced polyclonal output in most animals, one mouse showed oligoclonal hematopoiesis and dominant clones exceeding 20% contribution were observed in three others, indicating that in vivo selection can narrow clonal output.

The humanized NBSGW model imposes further constraints. Human HSPCs progressively lose repopulation potential in the xenograft niche and murine cells are less sensitive to O6-BG/BCNU than human cells^88–90, so the absolute expansion of selected human HSCs cannot be reliably assessed. The human cell tropism of BaEVRLess precludes adequate assessment of vector biodistribution and toxicity in non-hematopoietic tissues²⁰, GvHD limits follow-up duration⁹¹, and murine stroma does not fully recapitulate the human bone marrow hematopoietic microenvironment. Inefficient erythropoiesis in NBSGW mice necessitated in vitro erythroid differentiation to assess γ-globin expression, a step that can itself induce stress-erythropoiesis-associated HbF reactivation⁹². Furthermore, immunodeficient mice do not permit evaluation of immune responses to systemic vector delivery. Comprehensive non-human primate studies will be required to assess safety, biodistribution, immunogenicity, and therapeutic efficacy prior to clinical translation.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Christian Brendel (christian.brendel@childrens.harvard.edu).

Materials availability

All plasmids generated in this study are available from the lead contact upon request.

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• All original code has been deposited at Zenodo and is publicly available as of the date of publication. The DOI is https://doi.org/10.5281/zenodo.22116424.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Primary CD34+ hematopoietic stem and progenitor cells

Human CD34+ HSPCs isolated from mobilized peripheral blood of healthy adult male and female donors (unknown ethnicity) were obtained from the Fred Hutchinson Cancer Center’s Cooperative Centers of Excellence in Hematology Core. HSPCs were thawed and pre-stimulated for 24 hours using SCGM medium supplemented with 1% L-Glutamine (Gibco, Miami, FL), 1% Penicillin/Streptomycin, 100 ng/mL TPO (Peprotech), 100 ng/mL SCF (Peprotech, Cranbury, NJ) and 100 ng/mL FLT3-Ligand (Peprotech, Cranbury, NJ) prior to transduction.

Animals

All animal experiments were approved by the Boston Children’s Hospital Institutional Animal Care and Use Committee. Four-week-old, female NOD.Cg-Kit^W−41J Tyr⁺ Prkdc^scid Il2rg^tm1Wjl/ThomJ (NBSGW) were purchased from the Jackson Laboratory (Bar Harbor, ME) and housed in individually ventilated cage with littermates at a 12h/12h day/night cycle. Food and water were autoclaved and given ad libitum.

METHOD DETAILS

Viral vectors

Viral vectors expressing a green fluorescent protein (eGFP, (E)) or blue fluorescent protein (eBFP2, (B)) under control of the spleen focus-forming virus (SFFV, (S)) promoter and woodchuck post-transcriptional regulatory element (wPRE, (W)) were used in the alpha-retroviral (aRV.SEW and aRV.SBW) and lentiviral background (pCCL.SEW and pCCL.SBW). For in vivo application and chemoselection, we used a lentiviral vector containing a physiological PGK promoter (P) driving expression of the mutant methylguanine methyltransferase (MGMT-P140K, (M)) followed by a T2A site and an eGFP (pCCL.PMEW)⁹⁴. For barcode analysis, the genetic barcode (BC) 5’-ACGNNCACNNAGANNCTTNNCGANNCTANNGGANNCTTNNCGT-3’ was cloned in front of the PGK promoter (pCCL.BC-PMEW)⁹⁵. For therapeutic approaches to treat SCD, we cloned the MGMT-P140K selection cassette into the erythroid specific double shmiR (DS) vector that knocks down BCL11A and ZNF410 to induce fetal hemoglobin (pCCL.MiniG.DS.PMEW)⁶⁰.

Viral vector production and titration

Alpha-retroviral and lentiviral vectors were produced by transfection of HEK293T cells on 15 cm plates (~75% confluent) using polyethyleneimine (PEI linear, MW 25K, Stock: 1 μg/μL, Polysciences Inc., Warrington, PA). All HEK293T cells are cultured in Iscove modified Dulbecco’s medium (IMDM, Cytiva, Marlborough, MA) supplemented with 10% FBS (Biowest, Bradenton, FL) and 1% Penicillin/Streptomycin (Gibco, Miami, FL). For alpha-retroviral vectors, 17.5 μg transfer vector, 15 μg alpha.Gag/Pol and 14 μg envelope (VSVg, RD114TR, BaEVTR or BaEVRLess) plasmids, and for lentiviral vectors, 25 μg vectors, 30 μg Lenti.Gag/Pol, 15 μg Rev and 15 μg envelope plasmids were mixed at a 1:5 (w/w) ratio with PEI and incubated for 15–20 minutes at room temperature before adding the DNA:PEI mix to the HEK293T cells. After incubation of the cells for 8 hours at 37°C, the medium was exchanged to fresh, supplemented IMDM medium. Viral supernatants were harvested after 36 hours and concentrated using ultracentrifugation (24,000 rpm, 4°C, 2h, Beckman Coulter, Danvers, MA) including a 5 mL layer of 20% Sucrose in PBS on the bottom of the centrifugation tube for purification of viral particles. The viral particles were resuspended in SCGM medium (CellGenix, Freiburg, Germany) containing 0.5% bovine serum albumin (Sigma Aldrich, Burlington, MA), aliquoted and stored at −80°C. For in vivo applications, viral supernatants were harvested a second time 50 hours after the initial medium change, pooled and purified using tangential flow filtration (500kDa mPES hollow fiber, Repligen, Waltham, MA). All vector batches were titrated on 10⁵ HEK293T cells using a serial dilution of concentrated viral supernatants. Three days after transduction, gene marking based on eGFP fluorescence was measured via flow cytometry (Fortessa, BD, Franklin Lakes, NJ) and titers were calculated based on transducing units.

Transduction of human CD34+ HSPCs

Transduction was performed using an MOI of 5 with the alpha-retroviral vector aRV.SEW pseudotyped with VSVg, RD114TR, BaEVTR or BaEVRLess. Vectofusin^®−1 (10 μg/mL, pre-mixed with virus, Miltenyi, Charlestown, MA) was added to the viral vectors to enhance transduction. Additional transduction enhancers were tested individually or in combination with Vectofusin^®−1. These enhancers include RetroNectin^® (50 μg/mL, pre-coated on the plate, Takara Bio) and Lentiboost A+B (1 mg/mL, mixed with the cell suspension, Revvity). For transduction of quiescent cells, CD34+ cells were thawed, resuspended in cytokine-free medium and immediately transduced at an MOI 100 plus Vectofusin^®−1 as transduction enhancer using only VSVg- and BaEVRLess-pseudotyped alpha-retroviral vectors. Two hours after transduction, the cells were washed three times with PBS and resuspended in complete SCGM medium supplemented with cytokines. The transduction rates were measured three days after transduction via flow cytometry.

Neutralization assay

VSVg- and BaEVRLess-pseudotyped viral vectors were incubated at a 1:5 (v/v) ratio with fresh or heat-inactivated serum from three different donors for 1 hour at 37°C prior to transduction of 10⁵ PLB985 cells using 10 μL of the mix. Transduction rates were measured three days after transduction.

Co-injection experiment

At 6–8 weeks of age, each mouse was first injected with 5×10⁵ freshly thawed CD34+ HSPCs via one tail vein and 5–10 minutes later received an infusion of BaEVRLess-pseudotyped vectors studying (1) different MOIs (4, 20 and 100), (2) CD47-low versus CD47-high vectors at MOI 100 and (3) alpha-retroviral (aRV.SEW or aRV.SBW) versus lentiviral vectors (pCCL.SBW or pCCL.SEW) each at an MOI 50. All vectors were pre-mixed with Vectofusin^®−1 as transduction enhancer and injected via the contra-lateral tail vein. For competitive repopulation experiments, 10⁵ CD34+ HSPCs were transduced in vitro with the same vector mix and cultured for three days to assess the actual transduction rates, which was used to normalize the in vivo data and to adjust for any differences in baseline transduction. The peripheral blood of all animals was monitored and analyzed in weeks 6, 12 and 16 after co-injection. On week 16, the bone marrow was harvested and gene marking as well as lineage distribution was assessed by flow cytometry. All antibodies used for analyses are listed in the supplemental materials. The gating strategy is shown in Figure S3.

Mobilization of NBSGW mice

NBSGW mice were humanized with 5×10⁵ CD34+ HSPCs and mobilized eight weeks later using AMD3100 (Sigma Aldrich, Burlington, MA), AMD3100+G-CSF (Peprotech, Cranbury, NJ) or AMD3100+Gro-beta (GenScript, Piscataway, NJ). AMD3100 was injected subcutaneously for 1 hour at a dose of 5 mg/kg body weight. G-CSF was administered subcutaneously twice daily for four consecutive days at a dose of 250 μg/kg body weight. Gro-beta was given for 15 minutes subcutaneously at a dose of 2.5 mg/kg body weight⁴⁹. Peripheral blood samples were taken from each mouse pre-mobilization and at peak of mobilization and analyzed for human chimerism, lineage distribution and frequency/number of Lin- hCD34+ HSPCs via flow cytometry. Counting beads (CountBright, Thermo Fisher, Cambridge, MA) were added to each sample prior acquisition to assess the cell number of HSPCs in mobilized peripheral blood. Complete blood counts were performed using the Hemavet 950FS blood counter (Drew Scientific, Plantation, FL). The combination of AMD3100+Gro-beta was not only tested in the staggered approach (60 minutes for AMD3100 and 15 minutes for Gro-beta) but also in a simultaneous approach for 15 minutes both together. Additionally, the staggered AMD3100+Gro-beta approach was tested with the addition of BIO5192 (Tocris) given for 60 minutes subcutaneously at a concentration of 1 mg/kg body weight.

Colony-forming unit (CFU) assay

To assess colony-forming potential, 10 μL mobilized peripheral blood was seeded in 1 mL MethoCult H4434 (Stem Cell Technologies, Vancouver, Canada) according to manufacturer’s instructions. Cell suspensions were seeded in duplicates into 6-well plates and cultured for 14 days at 37°C in a humidified chamber and 5% CO₂. Colonies were counted and classified based on cellular morphology.

Cell cycle analysis

Bone marrow and peripheral blood samples of mobilized mice were stained for 30 minutes at room temperature for surface antigens using mouse and human FcR block, hCD45-PE-Vio770 and hCD34-APC. After washing the cells, cells were stained for 45 minutes with 2 μg/mL Pyronin Y (Sigma Aldrich, Burlington, MA) and 2 μg/mL Hoechst 33342 (Thermo Fisher Scientific, Cambridge, MA) at room temperature. The cells were directly acquired by flow cytometry without any additional washing step using an LSRII flow cytometer (BD, Franklin Lakes, NJ) with a UV laser.

In vivo gene therapy and chemoselection

Six-week-old, female NBSGW mice were humanized with either 62.5k, 125k, 250k or 500k CD34+ HSPCs isolated from mobilized peripheral blood of healthy donors. Human engraftment was validated six weeks later in the peripheral blood (“pre-mobilization”). Eight weeks after humanization, mice were mobilized using a staggered approach of 5 mg/kg body weight AMD3100 for 60 minutes and 2.5 mg/kg body weight Gro-beta for 15 minutes. At peak of mobilization, a blood sample was obtained from each animal (“post-mobilization”) and ~3–4 × 10⁷ BaEVRLess-pseudotyped viral particles were intravenously injected per mouse. Human chimerism and gene marking were monitored in the peripheral blood on week 1, 4, 6, 8, 10, 13 and 16 after vector injection using surface antigen staining and flow cytometry. To enrich transduced cells in vivo, three cycles of chemoselection were given on weeks 4, 6 and 8 using an escalating dose of 2.5, 5 and 10 mg/kg bodyweight 1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU) (Carmustine, Sigma Aldrich, Burlington, MA) intraperitoneally. One hour before BCNU administration, the mice received 20 mg/kg bodyweight O⁶-benzylguanine (O6-BG, Sigma Aldrich, Burlington, MA), an inhibitor of endogenous MGMT activity^48,52. At the final timepoint, bone marrow was harvested and analyzed for human chimerism, lineage distribution and gene marking by flow cytometry. For secondary transplantation, 3–4 × 10⁷ total bone marrow cells were transplanted into 6-week-old, female, secondary NBSGW recipients. Sixteen weeks after transplantation, peripheral blood, bone marrow and spleen were analyzed for human chimerism, lineage distribution and gene marking.

UMI-based clonal barcode analysis

To assess the clonal composition of transduced cells, we generated an amplicon sequencing library of the genetic barcode and analyzed the resulting FASTQ files using a custom bioinformatics pipeline. Therefore, genomic DNA (gDNA) of 10⁷ total bone marrow cells (viability >90%) of primary and secondary recipients was isolated using DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) and a total of 1500 ng gDNA was PCR amplified using the Q5 polymerase according to manufacturer’s instructions and the primers UMI_fw 5’- ACACTCTTTCCCTACACGACGCTCTTCCGATCTatcgatcacgagactagcctcg-3’ and UMI_rev 5’-GACTGGAGTTCAGACGTGTGCTCTTCCGATCTccaaccccgtggaattcgatatc-3’. The PCR products were purified using the Monarch^® PCR & DNA Cleanup Kit (New England Biolabs, Ipswich, MA) prior to amplicon sequencing (Azenta, Chelmsford, MA). Initial quality control of the reads was performed using Flexbar (version 3.5.0), with a minimum quality threshold set to qt=30 and additional parameters, such as q=TAIL and qf=i1.8, to ensure data quality. Reads passing the quality filter were processed to identify and extract the 16 bp unique molecular identifier (UMI) sequences using Flexbar. To increase detection accuracy and stringency, a 78 bp sequence was utilized, incorporating 20 bp anchor sequences both upstream (ttccgatctcaccggagacg) and downstream (cgtctcttcgaagatcggaa) of the consensus barcode sequence (NNCACNNAGANNCTTNNCGANNCTANNGGANNCTTNN). The maximum allowed error rate for detection was set at 0.1 (option -ae). The 16 bp UMI sequences detected from each sample were saved to a text file and further processed using a custom R script. UMI sequences were clustered using the dbscan algorithm, with a Levenshtein distance threshold of 1 to account for minor sequencing errors. The total number of unique UMI clusters (species richness, S) and their abundance, quantified by cumulative read counts, were calculated. To evaluate clonal diversity, various diversity indices were computed, including Shannon diversity index (H), Simpson’s diversity index (Simp), and Pielou’s evenness (J). Barcode composition and UMI frequency data for each sample were visualized as stacked bar plots generated using ggplot2.

In vitro erythroid differentiation

Whole bone marrow cells (10⁷ cells) were seeded for erythroid differentiation using a three-phase protocol adapted from Giarratana et al.⁹². The cells were cultured in erythroid differentiation medium (EDM) consisting of IMDM supplemented with 1% L-glutamine (Gibco, Miami, FL), 1% penicillin-streptomycin (Gibco), 330 μg/mL human holo-transferrin (Sigma Aldrich, Burlington, MA), 10 μg/mL insulin (Sigma Aldrich), 2 IU/mL heparin (Sigma Aldrich), 5% human AB serum (Gemini, West Sacramento, CA), and 3 IU/mL erythropoietin (Amgen, Cambridge, MA). In the first phase (days 0–7), the bone marrow cells were cultured in EDM plus 10⁻⁶ mol/L hydrocortisone (Sigma Aldrich), 100 ng/mL SCF (Peprotech) and 5 ng/mL IL-3 (Peprotech, Cranbury, NJ) (EDM-1). For the second phase (days 7–11), the cells were resuspended in EDM supplemented with SCF only (EDM-2). Finally in the third phase (days 11–18), cells were cultured in EDM without additional supplements (EDM-3).

Quantification of γ-globin using RT-qPCR

At the end of erythroid differentiation, about 10⁶ erythroid cells were harvested for RNA isolation using the RNeasy^® Mini Kit (Qiagen, Hilden, Germany) according to the manual’s instructions including a step for DNA digestion. After elution of the RNA into 20 μL RNase-free water, 500 ng RNA were reverse transcribed into cDNA using the SuperScript^™ IV VILO^™ Master Mix (Invitrogen, Carlsbad, CA). To quantify γ-globin, 20 ng cDNA were applied per qPCR reaction using the PowerSYBR^® Green PCR Master Mix (Applied Biosystems, Waltham, MA) including 2.5 nM forward and reverse primers in a total volume of 20 μL. For γ-globin the primers HBG_fw 5’-TGGATGATCTCAAGGGCAC-3’ and HBG_rev 5’-TCAGTGGTATCTGGAGGAC-3’ were used and as reference β-globin was amplified using the primers HBB_fw 5’-CTGAGGAGAAGTCTGCCGT-3’ and HBB_rev 5’-AGCATCAGGAGTGGACAGA-3’. The following cycle conditions were used at the QuantStudio 3 qPCR cycler (Applied Biosystems): 1 cycle at 50°C for 2 min, 1 cycle at 95°C for 10 min, 50 cycles at 95°C for 15s followed by 60°C for 1 min. The results were analyzed with the QuantStudio^™ Design & Analysis Software v1.5.2 (Applied Biosystems). The proportion of γ-globin was calculated as frequency of total globin (γ-globin+β-globin).

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistics were calculated with GraphPad Prism 9. The bar graphs represent the mean±SD or mean±SEM. After confirming parametric data distribution, Student’s t-test for comparing two groups or one-way ANOVA for comparing multiple groups was used. Two-way ANOVA was used to compare multiple parameters among more than two groups. The Pearson correlation coefficient was calculated for all correlation analyses. Statistical significance is indicated by asterisks. Details for each statistical analysis is found in the respective figure legend.

Supplementary Material

Supplemental information containing Supplemental Figures S1–S9 and Supplemental Table S1 can be found online at

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rat monoclonal anti-mouse CD45 PE-eFluor610 (clone 30-F11)	Thermo Fisher Scientific	Cat# 61-0451-82; RRID:AB_2574555
Human monoclonal anti-human CD45 PE-Vio770 (clone REA747)	Miltenyi	Cat# 130-110-772; RRID:AB_2658248
Mouse monoclonal anti-human CD45 PE (clone HI30)	BD Biosciences	Cat# 555483; RRID:AB_395875
Mouse monoclonal anti-human CD19 SuperBright 600 (clone HIB19)	Thermo Fisher Scientific	Cat# 63-0199-42; RRID:AB_2744834
Mouse monoclonal anti-human CD33 PerCP-Cy5.5 (clone WM53)	BioLegend	Cat# 303414; RRID:AB_2074241
Mouse monoclonal anti-human CD3 APC-Cy7 (clone HIT3a)	BioLegend	Cat# 300318; RRID:AB_314054
Mouse monoclonal anti-human CD34 BV785 (clone 561)	BioLegend	Cat# 343626; RRID:AB_2632727
Mouse monoclonal anti-human CD34 APC (clone 581)	BD Biosciences	Cat# 555824; RRID:AB_398614
Mouse monoclonal anti-human CD45RA PE-Cy7 (clone HI100)	BioLegend	Cat# 304108; RRID:AB_314412
Mouse monoclonal anti-human CD90 APC (clone 5E10)	BioLegend	Cat# 328114; RRID:AB_893431
Mouse monoclonal anti-human CD235a AlexaFluor700 (clone HI264)	BioLegend	Cat# 349122; RRID:AB_2894482
Bacterial and virus strains
Chemically competent Top10 E.coli	Brendel’s lab	N/A
Biological samples
Human mPB CD34 HSPCs	Fred Hutch Institute	N/A
Human AB serum	Boston Children’s Blood donor center	N/A
Chemicals, peptides, and recombinant proteins
Polyethyleneimine (PEI)	Polysciences	Cat# 23966–1
Iscove’s modified Dulbecco’s medium (IMDM)	Cytiva	Cat# SH30228
Fetal bovine serum (FBS)	Biowest	Cat# S1620
Penicillin/Streptomycin	Thermo Fisher Scientific	Cat# 15140122
Sucrose	Sigma Aldrich	Cat# S9378
Phosphate buffered saline (PBS)	Thermo Fisher Scientific	Cat# 14190250
Stem cell growth medium (SCGM)	Sartorius CellGenix®	Cat# 20802–0500
Bovine serum albumin (BSA)	Sigma Aldrich	Cat# A9418
L-Glutamine	Gibco	Cat# 25030081
Thrombopoietin (TPO)	Peprotech	Cat# 300–18
Stem cell factor (SCF)	Peprotech	Cat# 300–07
Fms-related tyrosine kinase (FLT) 3 ligand	Peprotech	Cat# 300–19
Vectofusin®−1	Miltenyi	Cat# 130-111-163
Retronectin	Takara Bio	Cat# T100A
Lentiboost A+B	Revvity	N/A
AMD3100	Sigma Aldrich	Cat# A5602
G-CSF	Peprotech	Cat# 300–23
Gro-beta	GenScript	Cat# Z03304
BIO5192	Tocris	Cat# 5051
MethoCult H4434	Stem Cell Technologies	Cat# 04434
Carmustine (BCNU)	Sigma Aldrich	Cat# C0400
O6-benzylguanine (O6-BG)	Sigma Aldrich	Cat# B2292
Human holo-transferrin	Sigma Aldrich	Cat# T0665
Human insulin	Sigma Aldrich	Cat# I9278
Heparin	Sigma Aldrich	Cat# H3149
Human AB serum	GeminiBio	Cat# 100-318-100
Erythropoietin (EPO)	Amgen	Cat# NDC 5513-144-10
Hydrocortisone	Sigma Aldrich	Cat# H0135
Human interleukin-3 (IL-3)	Peprotech	Cat# 200–03
Hoechst 33342	Thermo Fisher Scientific	Cat# H3570
Pyronin Y	Sigma Aldrich	Cat# 213519–1G
Counting Beads	Thermo Fisher Scientific	Cat# C36950
Human TruStain FcX	BioLegend	Cat# 422302
Mouse TruStain FcX	BioLegend	Cat# 101320
Zombie Aqua Viability Dye	BioLegend	Cat# 423102
Critical commercial assays
RNeasy® Mini Kit	Qiagen	Cat# 74104
SuperScript™ IV VILO™ Master Mix	Invitrogen	Cat# 11756050
PowerSYBR® Green PCR Master Mix	Applied Biosystems	Cat# 4367659
Experimental models: Cell lines
HEK293T (female)	ATCC	Cat# CRL-3216; RRID:CVCL_0063
PLB985 (HL-60) (female)	ATCC	Cat# CCL-240; RRID:CVCL_0002
Experimental models: Organisms/strains
NOD.Cg-KitW-41J Tyr+ Prkdcscid Il2rgtm1Wjl/ThomJ (NBSGW)	Jackson Laboratory	Cat# 026622
Oligonucleotides
UMI_fw 5’-ACACTCTTTCCCTACACGACGCTCTTCCGATCTatcgatcacgagactagcctcg-3’	This paper	N/A
UMI_rev 5’-GACTGGAGTTCAGACGTGTGCTCTTCCGATCTccaaccccgtggaattcgatatc-3’	This paper	N/A
HBG_fw 5’-TGGATGATCTCAAGGGCAC-3’	Brendel et al.14	N/A
HBG_rev 5’-TCAGTGGTATCTGGAGGAC-3’	Brendel et al.14	N/A
HBB_fw 5’-CTGAGGAGAAGTCTGCCGT-3’	Brendel et al.14	N/A
HBB_rev 5’-AGCATCAGGAGTGGACAGA-3’	Brendel et al.14	N/A
Recombinant DNA
Alpha-retroviral vector aRV.SEW	Schambach’s lab, Suerth et al.32	N/A
Alpha-retroviral vector aRV.SBW	This paper	N/A
Lentiviral vector pCCL.SEW	Demaison et al.93, Brendel et al.14	N/A
Lentiviral vector pCCL.SBW	Brendel et al.14	N/A
Lentiviral vector pCCL.PMEW	This paper	N/A
Lentiviral vector pCCL.BC-PMEW	This paper	N/A
Lentiviral vector pCCL.MiniG.DS.PMEW	This paper	N/A
Packaging plasmid alpha.Gag/Pol	Schambach’s lab, Suerth et al.32	N/A
Packaging plasmid pRSV.REV	Addgene	Cat# 12253
Packaging plasmid Lenti.Gag/Pol (pMDLg/pRRE)	Addgene	Cat# 12251
VSVg	Addgene	Cat# 12259
RD114TR	Verhoeyen’s lab	N/A
BaEVTR	Verhoeyen’s lab	N/A
BaEVRLess	Verhoeyen’s lab, Addgene	Cat# 225962
Software and algorithms
QuantStudio™ Design & Analysis Software v1.5.2	Thermo Fisher Scientific	https://www.thermofisher.com/us/en/home/technical-resources/software-downloads/quantstudio-3-5-real-time-pcr-systems.html
GraphPad Prism 9	GraphPad Software	https://www.graphpad.com/
FlowJo v10	FlowJo LLC	https://www.flowjo.com/
Custom R scripts for barcode extraction, merging and clonal diversity analysis	This paper	https://doi.org/10.5281/zenodo.22116424

Highlights.

• Phagocytosis-shielded BaEVRLess vectors transduce human CD34+ HSPCs in vivo in mice

• In vivo transduced HSPCs repopulate secondary recipients and maintain clonal diversity

• Chemoselection enriches gene-marked cells to ~70% of hCD45+ cells and 54% of HSPCs

• Erythroid-specific BCL11A/ZNF410 shmiRs induce therapeutic gamma-globin in vivo