AUTHORS
Christi Salisbury-Ruf, PhD, Mohd Azhar, PhD2, Isabel Vallecillo-Viejo, MD, PhD3, Richard Smith, PhD4, Anirudh Rao, MD5, So Gun Hong, PhD6, Kevin Quiroz Caceda, BS7, Olga Vasalatiy, PhD8, Noriko Sato, PhD9, Sofia Roitman, PhD10, Neelam Redekar, PhD11, Andre Larochelle, MD, PhD12
AFFILIATIONS
1 National Institutes of Health, Bethesda, Maryland, United States
2 National Institutes of Health, Bethesda, Maryland, United States
3 National Institutes of Health, Bethesda, Maryland, United States
4 National Institutes of Health, Bethesda, Maryland, United States
5 National Institutes of Health, Bethesda, Maryland, United States
6 National Institutes of Health, Bethesda, Maryland, United States
7 National Institutes of Health, Bethesda, Maryland, United States
8 National Institutes of Health, Bethesda, Maryland, United States
9 National Institutes of Health, Bethesda, Maryland, United States
10 National Institutes of Health, Bethesda, Maryland, United States
11 National Institutes of Health, Bethesda, Maryland, United States
12 National Institutes of Health, Bethesda, Maryland, United States
ABSTRACT
Introduction. In vivo gene editing of hematopoietic stem and progenitor cells (HSPCs) could eliminate apheresis and ex vivo manufacturing and may reduce myeloablative conditioning. Recent nonhuman primate (NHP) studies have demonstrated lipid nanoparticle (LNP)-mediated reporter mRNA delivery to marrow CD34? cells and marrow populations (Kim, Nat Biotechnol 2025; Zenhausern, Nat Biotechnol 2026), yet in vivo NHP HSPC editing has not been achieved. Whether this reflects inadequate marrow access, cell-type mistargeting, a block downstream of uptake, or failure of short reporters to predict editor-sized cargo remains unclear.
Methods. Seven ionizable lipids (ILs) encapsulating EGFP mRNA were screened in human CD34? cells from 3 donors. The lead IL, 4A3-SC8 (Farbiak, Adv Mater 2021), was formulated with sodium-iodide symporter (NIS) mRNA and infused intravenously into dexamethasone-premedicated NHPs (1 mg/kg mRNA; n = 2). NIS enables PET-CT imaging after 18F-tetrafluoroborate administration. PET-CT was performed at baseline and 24 hours. Bone marrow mononuclear cells (BM-MNCs) collected at 4 hours were profiled by scRNA-seq. A separate NHP received 4A3-SC8 LNPs co-encapsulating Cas9 mRNA and CCR5 single-guide RNA (sgRNA; 1 mg/kg). Indels were quantified through day 14 in FACS-purified marrow populations on 2 independent sequencing platforms. Cargo-size dependence was tested in vitro using molarity-matched ~1-kb EGFP mRNA, 2-, 3-, and 4-kb EGFP-tagged mRNAs, and full-length ABE8e-EGFP mRNA (5.7 kb) with or without 10% Cy5-labeled mRNA (n=1). For in vivo analyses, no inferential statistics were applied across animals.
Results. Among 7 formulations, 4A3-SC8 produced the highest EGFP mean fluorescence intensity in CD34? cells (1.91 × 10^4 ± 8.77 × 10^3) compared to the next-ranked formulation, SM-102 (1.55× 10^4 ± 8.75 × 10^3), and 1.5 to 4.7-fold above the remaining 5 formulations, with >93% viability. Physiologic NIS uptake in the thyroid, nasal area and stomach, kidneys, and bladder was unchanged from baseline. In contrast, PET signal at 24 hours (total uptake) increased 1.6- to 1.7-fold in marrow-containing humeral, femoral, and vertebral regions and 2.4- to 8.5-fold in the liver in both NHPs, demonstrating marrow localization and functional protein translation. Infusions were tolerated. Complete blood counts remained normal for >500 days, with a transient 2- to 3-fold increase in alanine aminotransferase.
scRNA-seq detected NIS transcripts in 1,392 of 7,911 post-infusion BM-MNCs (17.6%), compared with 3 of 8,670 pre-infusion cells (0.03%). NIS transcripts were detected in all 13 annotated marrow populations, ranging from 6.0% of T cells to 76.1% of stromal cells. NIS positivity was 27.0% in phagocytic myeloid cells, 19.6% in granulocyte-monocyte progenitors, and 8.0% in a CD34-enriched putative HSPC cluster (11 of 137 cells). Two cluster-independent measures yielded concordant estimates: 10% of CD34? cells and 11.9% of MLLT3? cells were NIS?.
Despite this delivery, CCR5 indel frequencies in the separate editing NHP remained at background in all sorted marrow and blood populations on both sequencing platforms. Monocytes and granulocytes, which showed the highest documented mRNA uptake, were also unedited. The LNPs generated 33.0% indels in NHP CD34? cells ex vivo, confirming reagent competence. In the in vitro cargo-size study, EGFP positivity was 94.4%, 66.1%, 29.8%, and 1.3% with the 1kb-EGFP mRNA and 2-, 3-, and 4-kb EGFP-tagged reporters, respectively, and 1.7% with 5.7-kb ABE8e-EGFP at matched molarity. In contrast, Cy5 uptake was equal or greater for the larger cargos, demonstrating a sharp loss of productive expression between 3 and 4 kb despite preserved mRNA uptake.
Conclusions. 4A3-SC8 LNPs produced PET-detectable functional expression of a small NIS reporter cargo (~2 kb) in marrow and distributed NIS transcripts across all 13 annotated marrow populations, including putative HSPCs. High stromal positivity may reflect vascular-niche interception, warranting investigation. Nevertheless, editing with Cas9 mRNA (~4.5 kb) was undetectable in every sorted population. Cargo-length testing identifies a payload-dependent bottleneck downstream of cell association with this untargeted formulation. Marrow access and short-reporter delivery are therefore insufficient surrogates for HSPC editing. Future LNP selection should evaluate therapeutic-length editor cargo and direct editing readouts in rigorously defined HSPC subsets.
DISCLOSURES
Christi Salisbury-Ruf, PhD: No relevant conflicts of interest to declare.