Patient-derived neomorphic SEPTIN6 mutations induce hematopoietic stem cell senescence underlying congenital neutropenia and myelodysplasia

Septins are phylogenetically conserved proteins that belong to the RAS-like GTPase superclass and are considered the fourth component of the cytoskeleton. They are primarily known for their function in cell division and cytokinesis (1–3). Overall, there are 13 septins, divided by sequence similarity and in some cases functional redundancy into 4 subfamilies comprising SEPT2, SEPT3, SEPT6, and SEPT7, which together form hetero-oligomeric complexes that are capable of participating in more complex cytoskeletal structures such as rings, filaments, and bundles. These cell structures are dependent on the interactions between septins of different subfamilies and other proteins, including actin, tubulin, and Rho GTPases such as Cdc42 (4–6). Septins contain a proline-rich N-terminal domain and a coiled-coiled C-terminal domain (7, 8). The C-terminus is involved in septin filament stabilization as well as establishing protein-protein interactions (9, 10).

Although detailed information is available about the role of septins in mammalian cell division and cytokinesis, less is known about their role in hematopoiesis. Sept1, -4, and -5 which belong to the SEPT2 subfamily, are involved in the regulation of hematopoietic stem and progenitor cells (HSPCs) (11–13), and Sept6 has a regulatory role in lymphopoiesis (14). Since septins function in cell division in many cell types, their role in cancer is better defined. Genetic alterations in septins are seen in colon, skin, and lung epithelial cancers, with many mutations seen in SEPT6 subfamily genes (15, 16). Septins have also been implicated in leukemias and lymphomas, but their role in leukemogenesis remains incompletely understood. In acute myeloid leukemia (AML), septins undergo reciprocal translocation with the mixed lineage leukemia (MLL) gene to form MLL-SEPT fusion proteins (17). These fusion proteins enhance HSPC self-renewal while blocking lineage differentiation, although the specific contribution of septins in the transformation phenotype is not clearly defined. Several septins, including SEPT2, -5, -6, -9, -11, and -14 are known fusion partners for MLL in AML and septins as a group are the most common fusion partner in MLL-rearranged leukemias (17–21). SEPT2 and SEPT6 have also been identified as targets of mutated splicing factor genes involved in MDS and AML (22).

We have previously identified 2 male patients with genetic alterations in the X-linked SEPT6 gene with congenital neutropenia that evolved into pediatric myelodysplastic syndrome (MDS) (23, 24). A third related patient was recently identified (25). Pediatric MDS is characterized by varying degrees of cytopenia, which occurs in the setting of abnormal and clonal hematopoiesis and is pathologically and genetically distinct from adult MDS (26–28). Both the reported patients had a germline mutation at the same codon, GM1: SEPT6 c.1282T>C (p.*428Gln) and GM2: SEPT6 c.1282T>A (p.*428Lys), leading to the loss of a stop codon at the C-terminus and a predicted increase in the protein size by 9 amino acids. This expressed immunoreactive protein was mislocalized in myeloid precursors and megakaryocytes (23). The third identified infant shares the mutation with his sibling and presented with the same early disease phenotype (25). GM1 was a de novo mutation (23), whereas GM2 was maternally inherited (25). The mother was an asymptomatic heterozygous carrier of the mutation and only expressed the WT variant of the SEPT6 gene, suggesting strong negative selection of this X-linked mutant. Protein modeling of the GM1 variant predicts that the mutated structure may hinder oligomerization of SEPT6. Both patients had similar bone marrow (BM) abnormalities, including myeloid precursor multinucleation, markedly enlarged myeloid precursors, and an increase in early myeloid forms consistent with abnormal functioning of the expressed SEPT6 mutant protein. A second somatic mutation (SM: SEPT6 c.43C>T, p.Arg15*) was identified in one of these patients, in cis with GM1 (SEPT6 c.1282T>C), creating an early stop codon, which we hypothesized may rescue hematopoietic function by expression of a truncated protein (23). Although these mutations suggest an important new molecular predisposition for pediatric MDS, no mechanism has previously been shown experimentally.

Here, we use base editing and lentiviral vector (LVV) transduction of human CD34⁺ HSPCs to introduce both germline mutations and SMs in the SEPT6 gene. We demonstrate that the expressed germline mutants lead to defects in hematopoiesis, including the formation of multinucleated myeloid cells, reduced growth of hematopoietic progenitor colonies, defective engraftment, and failure of long-term reconstitution in a xenograft model. Mechanistically, we also demonstrate that expression of the SEPT6 germline mutations leads to mitochondrial defects and premature senescence of HSPCs. The senescence defect was linked to an upregulation of mTOR phosphorylation, and inhibition of the mTOR pathway resulted in rescue of the in vitro hematopoietic defect and senescence. Finally, we demonstrate rescue of the hematopoietic phenotype when the SM that introduces an early stop codon is expressed in cis with the germline mutation, further emphasizing the detrimental nature of the germline mutation as found in the first reported patient. By generating a truncated protein, the SM effectively causes the knockout of the toxic gain-of-function mutant protein made by GM1. Overall, these data establish the association of SEPT6 mutations with abnormal hematopoiesis in pediatric MDS and contribute to understanding of the mechanism through which these mutations dysregulate hematopoiesis.

Transgenic expression of GM1 in SEPT6 WT cells disrupts hematopoiesis. SEPT6 is an X-linked gene, and the reported disease phenotype manifests in males. To determine whether the phenotypes associated with these mutations resulted from reduced WT expression or, alternatively, from the acquisition of de novo protein functions, we expressed the GM1:SEPT6 c.1282T>C (p.*428Gln) mutant via LVV (Figure 1A) in WT human CD34⁺ cells. Human CD34⁺ cells were transduced and mCherry⁺ cells were purified 2 days after transduction by fluorescence-activated cell sorting (FACS). Following myeloid differentiation, CD34⁺ cells expressing GM1, but not control transduced cells, demonstrated morphological abnormalities, including cell enlargement and multinucleation (Figure 1B). Functionally, GM1 expression significantly impaired hematopoietic colony formation. Compared with controls, GM1-expressing cells produced markedly fewer granulocyte/macrophage (GM) progenitor colonies, burst-forming unit-erythroid (BFU-E) colonies, and total colonies (Figure 1C). Together, these findings indicate that GM1 expression is not only functionally defective but also exerts a negative effect on WT SEPT6, thereby disrupting hematopoiesis.

Figure 1

Transgenic expression of GM1 in CD34⁺ cells leads to defective hematopoiesis. (A–C) The SEPT6 c.1282T>C germline mutant (GM1) was introduced into WT human CD34⁺ cells. (A) Schema of the lentiviral vector (LVV) used to introduce GM1 into WT human CD34⁺ cells. A stuffer sequence was used as a negative open reading frame control. (B and C) Two days after transduction, mCherry⁺CD34⁺ cells were isolated. (B) Cytospins of sorted CD34⁺ cells cultured for 5 days showing myeloid cell morphology. Original magnification, ×100. n = 3 biological replicates with similar results. (C) Hematopoietic progenitor assays of cells plated 2 days after transduction. Colonies were counted as GM, BFU-E, or GEMM 14 days after plating. n = 3 independent experiments performed in triplicate. (D–I) Single base pair gene editing was performed to introduce GM1 into human CD34⁺ cells. (D) The arrow indicates the base pair edited. (E) SEPT6 GM1–edited CD34⁺ cells were cultured for 5 days, fixed, and stained. Original magnification, ×100. The red arrow indicates an enlarged, multinucleated cell and the black arrow indicates cytoplasmic protrusions similar to filopodia. n = 3 biological replicates with similar results. (F and G) Two hundred and fifty edited cells were plated in hematopoietic progenitor assays 2 days after electroporation. Colonies were counted after 14 days and scored by morphology (F) and colony size (G). n = 3 independent experiments performed in triplicate. (H and I) Edited CD34⁺ cells (1 × 10⁶) were injected retro-orbitally into NBSGW mice. (H) Chimerism was analyzed monthly in the peripheral blood (PB) and at the 4-month time point in bone marrow (BM) using flow cytometric analysis of human CD45⁺ (donor cells). (I) Four months after xenotransplantation, BM was sorted into myeloid, lymphoid, and CD34⁺ HSPC compartments. Sorted cells were sequenced to determine editing frequency. Graphs represent 2 independent xenotransplants combined. n = 7–11 mice. Data presented as mean ± SD. All experiments were analyzed using 2-tailed t tests. *P < 0.05; **P < 0.005; ****P < 0.0001.

Germline SEPT6 mutations dysregulate hematopoiesis. To confirm that hematopoietic defects seen after LVV transduction were not exclusively due to overexpression of the mutant protein, we introduced the GM1 mutation into human male CD34⁺ cells by adenine base editing (ABE). By electroporating an ABE mRNA and a selected gRNA, we generated a T>C substitution at the SEPT6 c.1282 codon. ABE generated an additional synonymous mutation next to the target germline mutation (Figure 1D) that does not change the protein sequence. Editing efficiency (~80%) of the target mutation was confirmed by sequencing and deconvolution analysis (TIDE) 5 days later (Figure 1D). SEPT6 GM1–edited cells were enlarged (Figure 1E) (red arrow) and displayed frequent multinucleated forms, replicating aspects of the patient’s BM cellular phenotype. Multiple cells also possessed abnormal protrusions consistent with excessive filopodia (Figure 1E) (black arrow). Functionally, SEPT6 GM1–edited CD34⁺ cells showed a significantly reduced total number of GM CFUs (Figure 1F), with no significant change in the number of BFU-E and granulocyte, erythrocyte, monocyte, and megakaryocyte (GEMM) CFUs. However, both GM and BFU-E colonies derived from edited cells were notably smaller (Figure 1G).

To assess hematopoietic stem cell (HSC) function, 1 × 10⁶ SEPT6 GM1– or AAVS1-edited control CD34⁺ cells were transplanted into NBSGW (NOD.Cg-KitW-41J Tyr + Prkdc^scid Il2rg^tm1Wjl/ThomJ) mice, together with a fraction of unedited (~50%) cells serving as internal competitors. SEPT6 GM1–edited cells demonstrated markedly impaired peripheral blood (PB) chimerism compared with controls at 4 months (Figure 1H). A milder but significant engraftment defect was detected in BM at the same time point (Figure 1H), suggesting impaired terminal differentiation and release into the circulation. Lineage distribution of BM at 4 months after transplantation revealed reduced contribution of edited hematopoiesis across all compartments, including myeloid (~69%), lymphoid (~80%), and HSPCs (~63%) (Figure 1I), indicating preferential reconstitution of non-edited HSPC competitors affecting all cell lineages. Together with the in vitro data, these data suggest that the SEPT6 GM1 dysregulates engraftment and differentiation at the HSPC level.

To assess the functional relevance of SEPT6 germline mutations in human hematopoiesis, we constructed an LVV enabling simultaneous knockdown of endogenous SEPT6 and re-expression of WT or mutant SEPT6 cDNAs. Because patients carry only a single X-linked SEPT6 allele, endogenous SEPT6 was silenced using a microRNA-adapted short hairpin RNA (shmiR) driven by the PGK promoter, while siRNA-resistant WT, GM1 (SEPT6 c.1282T>C, p.*428Gln), or GM2 (SEPT6 c.1282T>A, p.*428Lys) cDNAs were expressed via a minimal CMV promoter (Figure 2A). An empty vector (EV) served as control. Human CD34⁺ cells were transduced and GFP⁺ cells were purified by FACS after 7 days of culture, and efficient SEPT6 knockdown and transgene expression were confirmed by qPCR (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.209595DS1) and Western blotting (Supplemental Figure 1B). Morphologically, enlarged multinucleated cells were observed specifically in the SEPT6 shmiR+GM1 group (black arrow, Figure 2B). Overall, a 2.3 ± 0.6-fold (mean ± SD) increase in multinucleation was observed in SEPT6 shmiR+GM1 compared with EV control (Figure 2C). Functionally, GM1 expression significantly reduced the total number of progenitor colonies formed, with the greatest loss in GM CFUs (Figure 2D). By contrast, re-expression of WT SEPT6 in the knockdown background fully rescued colony formation, yielding results comparable to the EV control (Figure 2D). Introduction of the second germline mutation, GM2, produced a phenotype similar to that of GM1, including multinucleated morphology (Supplemental Figure 1C) and a significant reduction in GM CFUs (Supplemental Figure 1D).

Figure 2

GM1 causes a hematopoietic engraftment defect accompanied with a decrease in the number of LT-HSCs and ST-HSCs. (A) Schema of bidirectional LVV to simultaneously knock down WT SEPT6 and express mutated SEPT6 in CD34⁺ cells. SEPT6 cDNAs used as inserts are shown. The location of the shmiR containing shRNA targeting WT SEPT6 embedded in microRNA30 is shown. The same schema was used to make the lentivirus used in Supplemental Figures 1, 2, and 4. (B) Photomicrographs of cytospins of CD34⁺ cells 7 days after transduction. Multinucleated myeloid precursor is shown at the black arrow. Original magnification, ×100. n = 3 biological replicates with similar results. (C) Cells were stained with 7-AAD to calculate percentage of multinucleation, n = 3. One-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. (D) Transduced GFP⁺ cells were sorted and plated 2 days after transduction in hematopoietic progenitor assays. Colonies were counted 14 days after plating. n = 3 independent experiments performed in triplicate. Two-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. (E–G) Transduced CD34⁺ cells (1 × 10⁶) 2 days after lentiviral transduction were injected retro-orbitally into NBSGW mice. (E) Percentage GFP⁺ donor chimerism was analyzed monthly in PB, and at the 4-month time point in BM. (F) Differences in donor engraftment in the myeloid and lymphoid compartment, and (G) immunophenotypically defined HSPC compartments analyzed in BM 4 months after xenotransplantation. (E–G) Graphs represent 1 of 2 xenotransplants, each containing an n of 5–10 mice and showing similar results. Two-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. (H and I) Human CD34⁺ cells were transduced and flow sorted 2 days after lentiviral transduction into GFP⁺ cells. Transduced CD34⁺ cells were cultured for 10 days in serum-free culture and analyzed for immunophenotypically defined LT-HSCs (H) and ST-HSCs (I) at each time point. Both percentages (first panel) and absolute number of cells (second panel) were analyzed and calculated. n = 3. Two-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. Data presented as mean ± SD (C–G) or mean ± SEM (H and I). *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

In xenotransplantation assays, the re-expression of WT SEPT6 rescued hematopoiesis, yielding PB chimerism and BM donor engraftment comparable to EV controls at 4 months after transplantation (Figure 2E). In contrast, GM1 expression significantly impaired engraftment. This defect was prominent in the myeloid lineage (Figure 2F), while the defect in the lymphoid compartment was less significant and variable, especially in B cells (Figure 2F), although this xenograft model does not provide robust reconstitution of lymphoid lineages. Analysis of progenitor compartments revealed a significant reduction in engrafted GFP⁺ HSCs, multipotent progenitors (MPPs), common myeloid progenitors (CMPs), and megakaryocyte-erythroid progenitors (MEPs) (Figure 2G). Reductions in the number of multi-lymphoid progenitors (MLPs), and lympho-myeloid primed progenitors (LMPPs) were also observed, but they were mild and not statistically significant (Figure 2G), suggesting partially lineage-restricted effects of GM1 expression.

To better define the hematopoietic defects caused by the expression of GM1, we assessed in vitro cell growth kinetics of LVV-transduced CD34⁺ cells. GM1 expression led to significantly reduced numbers of long-term HSCs (LT-HSCs) (Figure 2H) and short-term HSCs (ST-HSCs) (Figure 2I) over 10 days in culture. No significant differences were observed in LT-HSC and ST-HSC cell numbers between EV control and SEPT6 shmiR+WT SEPT6. Transduction of CD34⁺ cells with GM2 also resulted in a significant reduction in the percentage and number of LT-HSCs (Supplemental Figure 1E) and ST-HSCs (Supplemental Figure 1F) compared with controls. The reduced HSC numbers were not associated with increased apoptosis (Supplemental Figure 1G). Instead, cell cycle analysis revealed a modest but consistent increase in the percentage of G₂/M cells across SEPT6 shmiR, SEPT6 shmiR+GM1, and SEPT6 shmiR+GM2 groups (Supplemental Figure 2A), suggesting that the disruption of WT SEPT6 function alters cell cycle progression rather than survival. Together, these in vivo and in vitro data demonstrate that germline SEPT6 mutations impair HSC maintenance, engraftment, and differentiation.

Dysregulation of HSPC senescence by GM1. To further investigate the mechanisms through which SEPT6 mutations disrupt hematopoiesis, we performed bulk mRNA sequencing (RNA-seq) of SEPT6 GM1–edited CD34⁺ cells. Principal component analysis (PCA) plots for the samples sequenced are shown in Supplemental Figure 2B. Edited cells (45%–50% efficiency) cultured for 5 days showed 220 upregulated and 110 downregulated genes relative to controls (Supplemental Figure 2C and Supplemental Table 1). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed broad downregulation of protein synthesis, DNA replication, and DNA damage pathways (Supplemental Figure 2D), although interpretation was limited by cellular heterogeneity within the edited populations.

We next performed single-cell RNA-seq (scRNA-seq) of purified CD34⁺CD45RA^– progenitors after LVV transduction. Three groups were sequenced: (a) EV control, (b) SEPT6 shmiR+WT SEPT6, and (c) SEPT6 shmiR+GM1. Data from EV control and SEPT6 shmiR+WT SEPT6 groups were similar and pooled as a control set. Cell types were annotated into HSC and progenitor subsets (Supplemental Figure 3A). Among the top 10 targets, a number of genes involved in mitochondrial function were significantly downregulated in HSCs upon introduction of GM1, including mitochondrially encoded NADH:ubiquinone oxidoreductase (MTND3; –10.1 log₂[fold change]), translocase of outer mitochondrial membrane 7 (TOMM7; –4.4 log₂[fold change]), ATP synthase F1 subunit epsilon (ATP5F1E; –4.5 log₂[fold change]), and cytochrome c oxidase subunit 7C (COX7C; –5.5 log₂[fold change]) (Figure 3A). Ribosomal proteins were prominently represented, with 39 among the top 50 differentially expressed genes (DEGs) (Figure 3A and Supplemental Table 2; GM1 vs. Controls). Macrophage migration inhibitory factor (MIF), a gene that when knocked down in experimental models accelerates senescence (29), showed a log₂(fold change) of –18.4 upon introduction of GM1 (Figure 3B and Supplemental Figure 3B). Although not a top target, cyclin-dependent kinase inhibitor A1 (CDKN1A/p21), another gene associated with senescence and cell cycle arrest, showed a log₂(fold change) of 2 upon introduction of GM1 (Supplemental Table 2; GM1 vs. Controls). Additional downregulated transcripts included chromatin regulators such as H4 clustered histone 3 (HIST1H4C), which showed a log₂(fold change) of –2.8.

Figure 3

Transcriptional profiling of human CD34⁺ cells transduced with SEPT6 GM1. (A–D) Human CD34⁺ cells were transduced and isolated by FACS 5 days after lentiviral transduction into GFP⁺CD34⁺CD45RA^– cells. Sorted cells were analyzed by scRNA-seq. (A) Volcano plots were generated to determine the differential gene expression of SEPT6 shmiR+GM1 versus EV control and SEPT6 shmiR+WT SEPT6. Red boxes denote the top target genes associated with senescence, mitochondrial, and histone regulation. The threshold was set at 0.3/–0.3 log₂(fold change), n = 2 biological donors, P < 0.05. (B) Violin plots determining the differential mRNA expression of MIF through scRNA-seq. (C) Upregulated and (D) downregulated biological processes were identified and plotted through GO analysis. False discovery rate (FDR) was set at 0.05. Gene count in each pathway is represented by circle size.

Gene ontology (GO) analysis of the scRNA-seq data indicated upregulation of biological pathways involved in cellular metabolic processes, macromolecule metabolic processes, biosynthetic processes, and nitrogen compound metabolic processes (Figure 3C), and a downregulation of biological pathways mainly involved in mitochondria synthesis and mitochondrial respiration, including oxidative phosphorylation, mitochondrial ATP synthesis–coupled electron transport, proton motive force–driven mitochondrial ATP synthesis, and mitochondrial electron transport, NADH to ubiquinone (Figure 3D). In addition, pathways involved in cellular respiration were also downregulated.

Since enlarged, multinucleated morphology, dysregulated mitochondrial/translation pathways, reduced MIF expression, and increased p21 expression are features of senescence, we directly assessed the senescence of GM1-transduced progenitors. β-Galactosidase assays revealed significantly increased senescent cells on day 7 after transduction (Figure 4A), which further increased by day 15 (Figure 4B). Similar increases were observed in GM2 (Supplemental Figure 4A). GM1 also induced an approximately 2-fold upregulation of cyclin D1 at both mRNA (Figure 4C) and protein levels (Supplemental Figure 4B and Figure 4D), consistent with senescence-associated cell cycle arrest. Optic atrophy 1 (OPA-1), a mitochondrial fusion protein, demonstrated a log₂(fold change) of 0.35 in mRNA expression (Supplemental Table 2; GM1 vs. Controls) and a 1.8-fold change in protein expression (Supplemental Figure 4C and Figure 4E). Mitochondrial fusion is another feature of senescence and imaging with MitoBright revealed changes in mitochondrial morphology in GM1-expressing cells compared with controls (Figure 4F). EV control– and SEPT6 shmiR+WT SEPT6–expressing cells exhibited predominantly punctate and fragmented mitochondria, while the presence of GM1 resulted in elongated tubular mitochondria with significantly reduced fragmentation (Figure 4G) consistent with enhanced mitochondrial fusion and mitochondrial network remodeling. Additionally, Seahorse analysis revealed an elevated oxygen consumption rate (OCR) in GM1-transduced HSCs (Figure 4H), confirming mitochondrial perturbations predicted by scRNA-seq.

Figure 4

GM1 causes premature senescence in human CD34⁺ cells. (A and B) A β-galactosidase (β-gal) assay was performed on human CD34⁺ cells to determine the percentage of senescent cells on day 7 (D7) (A) and D15 (B) after LVV transduction. β-Gal⁺ cells were counted to determine percentage senescence. n = 3. (C) On D7 after lentiviral transduction, GFP⁺ cells were collected to quantify CyclinD1 mRNA expression. n = 3. (D) Densitometric analysis of cyclin D1 based on the Western blot shown in Supplemental Figure 4B. n = 3. (E) Densitometric analysis of OPA-1 based on the Western blot shown in Supplemental Figure 4C. n = 3. (F and G) Transduced GFP⁺ cells were stained with MitoBright LT Deep Red and DAPI to image for the mitochondrial network. (F) Images were acquired at ×63 oil immersion, and quantified to determine (G) the mitochondrial fragments per cell. n = 3 donors. (H) Transduced cells were sorted 5 days after LVV transduction for GFP⁺CD34⁺CD45RA^–EPCR⁺. One day after sorting, cells were subjected to a mitochondrial stress test to determine the oxygen consumption rate (OCR) at various time points. Arrows indicate injection of various stressors: oligomycin (Oligo), carbonyl cyanite-4 (trifluoromethoxy) phenylhydrazone (FCCP), rotenone/antimycin A (R/A), and 2-deoxy-D-arabino-hexose (2DG). n = 3. Data presented as mean ± SD (A–E and G) or mean ± SEM (H). All experiments were analyzed using 1-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

Collectively, these findings indicate that SEPT6 germline mutations accelerate HSC senescence through dysregulation of mitochondrial function, protein synthesis, and chromatin regulation, providing a mechanistic basis for defective hematopoiesis and engraftment.

SM mitigates the hematopoietic defect caused by GM1. The SEPT6 c.43C>T (p.Arg15*) SM present in a low variant allele frequency (VAF) in the BM of the first patient creates a stop codon and would likely cause early truncation of the mutant protein (23) and may mitigate the deleterious effects of the germline mutant by effectively reducing expression of the C-terminal mutant. To determine the potential functional relevance of the SM present in the BM, an LVV similar to that shown in Figure 2A but with a SEPT6 cDNA containing both the GM1 and SM was constructed to efficiently introduce both mutations in cis into human CD34⁺ cells. Introduction of the SM in cis with GM1 led to reduced SEPT6 protein expression, similar to the SEPT6 shmiR group, as demonstrated by Western blot analysis (Figure 5A). Notably, the enlarged, multinucleated morphology observed in GM1-expressing cells was absent in the double mutant (SEPT6 shmiR+GM1+SM) (Figure 5, B and C). To directly test whether the SM rescues the proliferative defect caused by GM1, we performed competition assays in which GM1 (BFP⁺) and double mutant (GFP⁺) were cocultured at a 1:1 ratio. Over 10 days of culture, GM1-expressing cells were progressively outcompeted by cells transduced with the double mutant (Figure 5D). By day 10, 41% of cells in the GM1 group remained, whereas 59% of cells were still present in the double mutant group, supporting rescue of cell proliferation by the double mutant presumably by early truncation and reduced expression of the GM1 SEPT6 mutant protein.

Figure 5

Rescue of the hematopoietic defect caused by GM1. (A) Western blot of sorted GFP⁺CD34⁺ cells 7 days after transduction to probe for SEPT6. GAPDH was used as a loading control. (B) Photomicrographs of cytospins of CD34⁺ cells 7 days after transduction. Original magnification, ×100. n = 3 biological replicates with similar results. (C) Transduced CD34⁺ cells were stained with 7-AAD to calculate percentage of multinucleation. n = 3. One-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. (D) Transduced GFP⁺ SEPT6 shmiR+GM1+SM and BFP⁺ SEPT6 shmiR+GM1 cells were sorted and mixed in a 1:1 ratio 2 days after transduction. Mixed cells were cultured up to 10 days after transduction. Flow cytometry was used to determine percentage of GFP⁺ and BFP⁺ cells at regular intervals. n = 3. Two-tailed t test. (E) Transduced GFP⁺ cells were sorted and plated 2 days after transduction in hematopoietic progenitor assays. Colonies were counted 14 days after plating. n = 3 independent experiments performed in triplicate. Two-way ANOVA. (F) Transduced CD34⁺ cells (1 × 10⁶) 2 days after lentiviral transduction were injected retro-orbitally into NBSGW mice. Percentage GFP⁺ donor chimerism was analyzed monthly in PB, and at the 4-month time point in BM. Graphs represent 1 of 2 xenotransplants, each containing an n of 5–10 mice and showing similar results. Two-way ANOVA. (G and H) Human CD34⁺ cells were transduced and flow sorted 2 days after lentiviral transduction into GFP⁺ cells. Transduced CD34⁺cells were cultured for 10 days in serum-free culture and analyzed for immunophenotypically defined LT-HSCs (G) and ST-HSCs (H) at each time point. Controls are similar between Figure 2, H and I, and Figure 5, G and H. n = 3. Data presented as mean ± SD (C, E and F) or mean ± SEM (D, G, and H). Two-way ANOVA. *P < 0.05; **P < 0.005; ****P < 0.0001.

Colony-forming assays further supported the rescue of hematopoietic activity. GM1 expression decreased both total CFU and GM CFU numbers, while the introduction of the double mutant restored the colony output (Figure 5E). In xenotransplantation assays, as previously seen, the expression of GM1 reduced PB chimerism and BM engraftment, whereas the double mutant completely rescued this defect (Figure 5F). Donor chimerism and engraftment with the double mutant group were not significantly different from EV control. Similarly, the double mutant restored immunophenotypically defined LT-HSC (Figure 5G) and ST-HSC (Figure 5H) number in culture compared with GM1, reaching levels indistinguishable from EV control. Our data demonstrate that expression of the SM rescues the hematopoietic defects of GM1 both in vitro and in vivo.

To define the mechanisms underlying this phenotypic rescue, double mutant–transduced cells were analyzed by scRNA-seq, similarly to the studies outlined above. We analyzed GM1 HSCs versus the double mutant HSCs and both controls. Several histone and cytoskeletal accessory proteins dysregulated by GM1 were normalized by the double mutant, including HIST1H4C (–2.8 log₂[fold change]) and HIST2H2AC (–1.8 log₂[fold change]) (Figure 6A and Supplemental Table 2; GM1 vs. GM1+SM vs. Controls). GO analysis indicated reversal of GM1-induced defects in protein modification and mitochondrial respiration/stress pathways (Figure 6, B and C).

Figure 6

Transcriptional profiling of the genes rescued by the double mutant. (A–C) Human CD34⁺ cells were transduced and subjected to FACS 5 days after lentiviral transduction to isolate cells into GFP⁺CD34⁺CD45RA^– cells. Sorted cells were analyzed by scRNA-seq. (A) Volcano plots were plotted to determine the genes uniquely expressed in SEPT6 shmiR+GM1 that were significantly different from EV control, SEPT6 shmiR+WT SEPT6, and SEPT6 shmiR+GM1+SM. Red boxes denote the top target genes associated with senescence, mitochondrial, and histone regulation. Threshold was set at 0.3/–0.3 log₂(fold change), n = 2 biological donors, P < 0.05. (B) Upregulated and (C) downregulated biological processes were identified and plotted through GO analysis. FDR was set at 0.05. Gene count in each pathway is represented by circle size.

Mitochondrial genes were also prominently rescued. Among the top 10 gene targets dysregulated by GM1 but rescued by the double mutant, 6 genes were associated with mitochondria. NADH:ubiquinone oxidoreductase subunit A1 (NDUFA1) demonstrated a –2 log₂(fold change) in mRNA expression; NDUFA13 a –1.9 log₂(fold change), ubiquinol-cytochrome c reductase, complex III subunit XI (UQCR11) a –2.8 log₂(fold change), and UQCR10 a –1.9 log₂(fold change) (Figure 6A and Supplemental Table 2; GM1 vs. GM1+SM vs. Controls). Expression of MIF, a key antisenescence factor suppressed by GM1, was fully restored by the double mutant (Supplemental Figure 3B).

Functional β-galactosidase assays confirmed these transcriptomic findings. EV control (dotted line in Figure 7, A–E) and GM1 data are shown in Figure 4, A–E, and H. The double mutant significantly reduced β-galactosidase⁺ senescent cells compared with GM1 on day 7 after transduction (Figure 7A), with the rescue still evident on day 15 (Figure 7B). The percentage of β-galactosidase⁺ cells in the double mutant group was not significantly different compared to the EV control. Along with the rescue of premature senescence, double mutant cells also showed reduced mRNA expression (Figure 7C) and protein expression (Supplemental Figure 4B and Figure 7D) of cyclin D1. OPA-1 protein expression was also rescued by the double mutant (Supplemental Figure 4C and Figure 7E). To determine whether this rescue was extended to the mitochondrial dysfunction associated with GM1, we next assessed the response of the double mutant to a mitochondrial stress test. Seahorse assays showed that the double mutant restored the OCR to control-like levels, contrasting with the elevated respiration observed in GM1 cells (Figure 7F vs. Figure 4H).

Figure 7

Rescue of premature senescence in human CD34⁺ cells caused by GM1. (A and B) A β-gal assay was performed on human CD34⁺ cells to determine the percentage of senescent cells on day 7 (D7) (A) and D15 (B) after LVV transduction. β-Gal⁺ cells were counted to determine percentage senescence. n = 3. Dotted line represents EV control. Controls are the same as Figure 4, A and B. (C) On D7 after lentiviral transduction, GFP⁺ cells were collected to quantify CyclinD1 mRNA expression. n = 3. (D and E) Western blot densitometric analysis of cyclin D1 (D) and OPA-1 (E). Blots are shown in Supplemental Figure 4, B and C, respectively. n = 3. (F) Transduced cells were sorted 5 days after LVV transduction for GFP⁺CD34⁺CD45RA^–EPCR⁺. One day after sorting, cells were subjected to a mitochondrial stress test to determine the OCR at various time points. Arrows indicate injection of various stressors: oligomycin (Oligo), carbonyl cyanite-4 (trifluoromethoxy) phenylhydrazone (FCCP), rotenone/antimycin A (R/A), and 2-deoxy-D-arabino-hexose (2DG). Figure 4H and Figure 7F were performed during the same experimental set up. n = 3. (G) Western blot (left panel) and densitometric analysis (right panel) of phosphorylated mTOR at the Ser2481 position. GAPDH was used as an internal loading control. (H) Transduced GFP⁺ cells were sorted and plated 2 days after transduction in methylcellulose containing 25 nM rapamycin. Colonies were counted 14 days after plating. n = 3 independent experiments performed in triplicate. (I and J) A β-gal assay was performed on human CD34⁺ cells to determine the percentage of senescent cells on D7 (I) and D15 (J) after LVV transduction. Rapamycin (25 nM) was added to the culture every 7 days. β-Gal⁺ cells were counted to determine percentage senescence. n = 3. Data presented as mean ± SD (A–E and G–J) or mean ± SEM (F). All experiments were analyzed using 1-way ANOVA followed by Tukey’s multiple comparisons test as a post hoc analysis. *P < 0.05; **P < 0.005.

Due to association of senescence with the mTOR pathway, we analyzed changes in expression of mTOR phosphorylation. GM1 expression in human CD34⁺ cells resulted in the upregulation of phosphorylated mTOR at the Ser2481 position (Figure 7G). To further define mTOR pathways involved downstream of GM1, we performed rapamycin-induced mTOR inhibition. Rapamycin-induced inhibition of the mTOR pathway restored the in vitro hematopoietic defect caused by GM1 (Figure 7H). Both numbers of total and GM colonies were rescued and comparable to controls. Functional β-galactosidase assays confirmed that addition of rapamycin to transduced cells also rescued the GM1-induced senescence defect, evident on days 7 and 15 (Figure 7, I and J). These studies confirm the GM1-associated senescence defect is linked to the mTOR pathway.

Interestingly, cells transduced with SEPT6 shmiR demonstrated a modest increase in the G₂/M phase of the cell cycle (Supplemental Figure 2A), although there was no significant effect on short-term hematopoietic progenitor cell production in vitro (Supplemental Figure 1D). Knockdown studies support the hypothesis that the phenotype observed due to the germline mutations are caused by the acquisition of new functions and not because of the hypomorphic expression of WT SEPT6. Despite all these broad functional and transcriptional rescues, some abnormalities persisted in cells transduced by the double mutant. Double mutant cells exhibited a modest G₂/M cell cycle accumulation (Supplemental Figure 2A) and retained dysregulation of ribosomal proteins and certain metabolic pathways (Supplemental Figure 5), similarly to GM1, which suggests decreased WT SEPT6 expression may lead to subtle functional abnormalities. Overall, these data demonstrate that expression of SEPT6 SM, when present in cis with GM1, mitigates the hematopoietic, mitochondrial, and senescence defects caused by GM1. This highlights mitochondrial dysfunction and premature senescence as central mechanisms of GM1-associated hematopoietic failure.

We have recently identified and reported 2 unrelated, non-syndromic newborns presenting with severe neutropenia with both novel X-linked germline mutations in the C-terminus of the SEPT6 gene (23, 24). Both identified patients had a high degree of myeloid tetraploidy in the BM, were refractory to G-CSF, and progressed to BM failure with multilineage MDS, suggesting a dysfunction in the HSC compartment. Septins are highly conserved GTP-binding proteins that regulate a variety of cellular functions, including cytokinesis, polarity, cell cycle, vesicle trafficking, and exocytosis (1–3, 10). Of the SEPT6 subfamily, Sept6, -8, and -11 are expressed in primitive hematopoietic cells and septins are the most common fusion partners in MLL-rearranged leukemias. Here, using orthogonal approaches of gene editing and LVV-mediated over- and under-expression of SEPT6, we demonstrate that expression of these germline mutations in SEPT6 lead to hematopoietic failure in experimental systems. Engraftment studies suggest a defect in terminal differentiation of HSCs, which was more severe in the myeloid compartment, and a premature loss of engraftment. Mechanistically, these hematopoietic defects were associated with increased senescence of HSPCs linked to dysregulation of mitochondrial function. Defective senescence was accompanied by an increase in mTOR phosphorylation and pharmacologic inhibition of mTOR activation reversed the senescence phenotype. The hematopoietic and senescence defects were also rescued by a somatic mutation identified at a low allelic frequency in one of the patients, which was predicted to cause the early truncation of the deleterious germline mutation.

Most studies of septins in hematopoiesis have been related to somatic mutations in human leukemia. Septins are known fusion partners with the MLL gene, where the N-terminus of MLL is fused in frame with the entire open reading sequence of the septin partner (17, 21, 30, 31). Septins as a group make up the largest of the many MLL partners. In MLL:SEPT fusions studied, including specifically MLL:SEPT6, the endogenous septin is under-expressed but the significance of this under-expression has not been fully elucidated. The link of septins with leukemia phenotypes suggest that altered expression is relevant to myeloid biology. MLL is fused with approximately 90+ other genes but the largest group of genes are the septin family genes, including SEPT2, -5, -6, -9, and -11, among which SEPT6 is the most common fusion partner. Fusions with non-septin genes are associated with an acute lymphoblastic leukemia (ALL) phenotype in more than 70% of cases; in contrast, nearly 90% of MLL fusions with septins are associated with AML. The role of septins in normal hematopoiesis has not been well studied but there appears to be alterations in septins in HSCs from old mice that are associated with altered polarity and symmetry of cell divisions (5, 32).

Senescence is the irreversible arrest of replication in cells. It is induced by various stressors such as replicative stress, dysfunctional telomerase, DNA damage, and oncogenes and is typically characterized by enlarged, multinucleated cells with increased lysosomal activity and abnormal mitochondria (33, 34). It is commonly observed in MDS and is accelerated by the upregulation of CDKN1/p21 (35, 36). Here, we have demonstrated that 2 different germline mutations in SEPT6 cause enlargement and multinucleation of HSPCs and a measurable increase in senescence, along with dysregulation of several senescence-associated genes, including p21 and MIF. Increased p21 levels are known to inhibit the cell cycle, whereas MIF is an antisenescence gene known to inhibit miR34a and protect cells via targeting sirtuin 1 (SIRT1) (37).

Mitochondrial dysfunction is known to be an important driver and maintainer of senescence, whereas senescence directly contributes to mitochondrial dysfunction, making these processes tightly interlinked. Mitochondrial dysfunction in senescence is coupled with an increased OCR (38), which we observed in transduced HSCs expressing the germline SEPT6 mutations. We also observed an upregulation of OPA-1, a mitochondrial fusion protein (39) and mitochondrial elongation and fusion, another feature associated with induction of senescence (40). Depletion of OPA-1 rescues cells from senescence-associated phenotype (41) but also dysregulates pathways important for regular cell function (42). Dysregulated mitochondrial, metabolic, and translational pathways are also associated with aberrant RNA splicing, which overall predicts survival in patients with MDS (22). In particular, mutations in the spliceosomal component gene SF3B1 affects the expression of SEPT2 and SEPT6 in patients with MDS.

Mitochondrial dysfunction also facilitates the exchange of mitochondrial DNA and mitochondria between damaged and intact cells. This occurs through the formation of tunneling nanotubes (TNTs), which are made of actin and microtubules and dependent on actin regulatory factors, specifically the Rho GTPases Cdc42 and Rac1 (43, 44). Cdc42 also mediates mTOR signaling, which is required for the transport of mitochondria through TNTs (43). Although our work does not show a direct association between SEPT6 and Cdc42 activity (data not shown), we observed increased mTOR phosphorylation in SEPT6 mutants. Rapamycin-induced inhibition of mTOR phosphorylation resulted in recovery of both in vitro hematopoietic defects and senescence caused due to the GM1 SEPT6 mutation. The mTOR pathways are known contributors of metabolic dysfunction and SASP in senescence (45). Inhibition of mTORC ablates mitochondrial dysfunction, which in turn rectifies p21 expression and certain features of senescence (46, 47).

The premature acceleration of senescence in transduced HSPCs expressing SEPT6 germline mutations poses important questions regarding their association with mTOR and Rho GTPases and may provide broader insights into senescence. Moreover, the connection between these pathways and actin and tubulin warrants studies regarding the pathways through which these mutations cause morphological abnormalities and affect polarity. Given the noted changes in transcription of epigenome genes, the link between SEPT6 mutants, nuclear actin organization (48), and regulation of gene expression, need further exploration. Future studies involving ablation of dysfunctional mitochondria to rescue senescent HSPCs containing SEPT6 mutations may clarify the causative role(s) of mitochondrial abnormalities in SEPT6-induced HSC senescence.

Germline mutations leading to MDS are present in approximately 30% of pediatric cases (27). Acquisition of SMs leading to a genetic rescue is a feature of several pediatric germline MDS predisposition syndromes, most notably those linked to SAMD9/9L variants. In MDS associated with SAMD9/9L, expression of SMs can restore multilineage hematopoiesis during evolution of the disease process (27, 28, 49, 50). Gain-of-function mutations seen in SAMD9/9L and now SEPT6 are not only a feature of MDS but also other BM failure syndromes such as severe congenital neutropenia, wherein heterozygous mutations in ELANE cause disease and predispose patients to MDS and AML (51). The presence of an SM in the SEPT6 gene at a low VAF in some BM cells in one of the affected patients led us to speculate that the SM may lead to a premature truncation of the disease-causing C-terminal mutant protein. Early truncation of the SEPT6 protein was confirmed in our results by the presence of a smaller band size on Western blots and visible protein staining through immunohistochemistry (data not shown). Here we provide direct experimental evidence supporting this hypothesis. While we demonstrate that SEPT6 mutant proteins acquire what we believe are novel functions that appear to interfere with normal SEPT6 functioning, an important observation is that some abnormalities in cell cycle progression were observed with knockdown of SEPT6 expression alone and some pathway abnormalities induced by the germline mutations were not fully corrected by expression of the SM. Both observations suggest the possibility that SEPT6 plays an important role in normal HSC functions. Finally, it is interesting to note that the phenotype described in both patients was restricted to blood cells and both patients had complete correction with allogeneic HSC transplantation (data not shown for GM1) (25). Whether this is due to lineage-specific expression levels or to redundancy of other septins in critical pathways in non-hematopoietic tissues, as can be seen in yeast (52), is not known and requires additional studies. Overall, we have described a mechanism by which SEPT6 germline mutations cause congenital neutropenia, BM failure, and propensity to MDS.

Sex as a biological variable

This study focuses on an X-linked condition that manifests symptoms in males.

LVV cloning and transduction of HSPCs

An shRNA against SEPT6 was designed (5′-AGCGACCTGTTCAACACCAAATTCGATAGTGAAGCCACAGATGTATCGAATTTGGTGTTGAACAGGG-3′) and embedded in a miR30A background to create a SEPT6 shmiR, similar to previous studies (53). The shmiR and mutant SEPT6 cDNA were cloned into a pCCL bidirectional vector that contained the PGK and minimal CMV promotors. All cloning was performed using the NEBuilder HiFi DNA Assembly kit (New England Biolabs). Cloned plasmids were used to prepare lentivirus for transducing male human CD34⁺ cells at an MOI of 50. Male human CD34⁺ cells from mobilized PB were purchased from the Fred Hutchinson Cancer Center. All CD34⁺ cells in this publication were cultured in Serum Free Expansion Media II (STEMCELL Technologies) supplemented with 1% penicillin/streptomycin, 100 ng/μL SCF, 100 ng/μL Flt3, 50 ng/μL TPO, 0.75 μM SR1 (STEMCELL Technologies), and 35 nM UM171 (ExcelThera) (54). All cytokines were purchased from Peprotech. Three days after lentivirus transduction, GFP⁺ cells were isolated through FACS. All sorting was performed on a FACSAria Fusion (BD Biosciences).

Base editing of CD34+ cells

HSPCs were cultured and electroporated as previously described (55). Briefly, SpRY-ABE8e mRNA was prepared by T7 run-off in vitro transcription with cotranscriptional capping with ARCA (3′-O-Me-m7G[5′]ppp[5′]G cap analog) using a linearized plasmid template. Human male CD34⁺ cells were thawed and cultured at 0.75 × 10⁶ cells/mL in StemSpan SFEMII media supplemented with FLT3L, SCF, TPO, UM171 and SR1 as noted above. After 48 hours of expansion culture, cells were electroporated with 5–7 μg of base editor mRNA and 300–360 pmol of SEPT6 sgRNA in 20 μL volume using the Lonza 4D-Nucleofection system. Editing outcomes were evaluated 5 days after electroporation. Overall editing efficiency was between 50% and 80%.

Morphological studies

One week after electroporation or lentiviral transduction, 10,000 cells resuspended in 100 μL of phosphate-buffered saline (PBS) were cytospun onto frosted slides at 200g for 5 minutes. Cells were then fixed and stained using a Dip Quick Stain Kit (Jorgensen Laboratories) as per the manufacturer’s instructions. Slides were observed at ×100 magnification using an Olympus BX41 microscope attached to an Excelis HD Camera and monitoring system.

Clonogenic assays

To assess the hematopoietic progenitor colony–forming capacity of human CD34⁺ cells, 250 cells were added to 1 mL of H4435 Methocult (STEMCELL Technologies). For inhibition of the mTOR pathway, 25 nM rapamycin (MilliporeSigma) was supplemented in the Methocult. Cell mixtures were gently vortexed and plated onto 35-mm dishes (STEMCELL Technologies) and incubated at 37°C in 5% CO₂ for 14 days. Colonies were counted and distinguished into lineages using STEMvison (STEMCELL Technologies), an automated counter for clonogenic assays.

Xenotransplantations

Each xenotransplant was a combination of 4 different human CD34⁺ male donors mixed and distributed between the various groups. Two days after electroporation or lentiviral transduction, 1 × 10⁶ human CD34⁺ cells were retro-orbitally injected into 6- to 8-week-old non-irradiated female NOD.Cg-KitW-41J Tyr + Prkdc^scid Il2rg^tm1Wjl/ThomJ (NBSGW) mice (Jackson Laboratory). Chimerism was assessed monthly for 4 months using human CD45 as a marker for donor cells, after which BM was isolated and used for flow cytometric analysis. Isolated cells were stained for 30 minutes at 4°C with a cocktail of antibodies listed in Supplemental Table 3. All antibodies were purchased from BioLegend. Cells were washed and data acquired on a Cytek Aurora (Cytek Biosciences). Flow cytometry data were analyzed using FlowJo (BD Biosciences).

Bulk RNA-seq

Male human CD34⁺ cells were electroporated to introduce a single base pair (T>C) change, as described above. Editing efficiency was approximately 50%. Edited cells were counted 5 days after electroporation. One thousand cells were resuspended in buffer TCL with 1% 2-mercaptoethanol and loaded in a 96-well plate. Two human CD34⁺ donors were used for these experiments and 6 replicates of each donor were plated into a 96-well plate for each group. The SmartSeq2 RNA-seq platform at the BROAD Institute was utilized to perform bulk RNA-seq. Briefly, RNA was first purified and converted to cDNA. cDNA libraries were constructed using a Nextera XT DNA Library Preparation kit (Illumina). Pooled libraries underwent flowcell cluster amplification and sequencing according to the manufacturer’s instructions using a NextSeq (Illumina). TAG analysis was performed to determine the quality of the sequencing data.

The quality control of raw fastq files was performed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and if necessary, the adapters were trimmed by fastp (56). The reads were mapped and quantified using default parameters with Salmon (57). The DEGs were identified by DESeq2 (58). The threshold of DEGs was adjusted to P less than 0.05 by the Benjamini-Hochberg method and a fold change of greater than 2. The GO analysis was conducted using clusterProfiler (59).

RNA extraction and qRT-PCR

Total RNA was extracted using a RNeasy Plus Micro Kit (QIAGEN). cDNA was prepared using the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen) with random hexamers. qRT-PCR was performed using the SYBR Green PCR Master mix (Applied Biosystems). Primers are provided in Supplemental Table 4.

Western blotting

Western blotting was performed on sorted GFP⁺CD34⁺ cells cultured for 7 days after transduction. Cells were lysed using RIPA buffer with Protease and Phosphatase Inhibitors (Thermo Fisher Scientific). Total protein was quantitated using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Approximately 20–30 μg of protein was loaded onto a NuPAGE 4%–12% Bis-Tris gel (Invitrogen). The proteins were then transferred to a polyvinylidene difluoride (PVDF) membrane using the Trans-Blot Turbo Transfer System (Bio-Rad). The membranes were treated with a blocking buffer to prevent non-specific binding followed by overnight primary antibody incubation at 4°C. The following primary antibodies were used: anti-Sept6, clone 9E7 (MilliporeSigma) or clone B-8 (Santa Cruz Biotechnology), anti–cyclin D1, clone SP4 (Abcam), anti–OPA-1, clone D6U6N (Cell Signaling Technology), anti–p-mTOR (Ser2481) polyclonal (Thermo Fisher Scientific), anti-mTOR polyclonal (Thermo Fisher Scientific), and GAPDH loading control monoclonal antibody, clone GA1R, HRP (Invitrogen). The membranes were washed and then incubated with a secondary HRP-conjugated antibody for 1 hour at room temperature. The secondary antibodies used were HRP-linked anti-rabbit (cyclin D1 and E1) or anti-mouse (SEPT6) immunoglobulin G (IgG) secondary antibody (Cell Signaling Technology). Blots were imaged and densitometric analysis was performed using ImageJ (NIH).

Cell growth kinetics studies

To analyze the percentage and number of HSCs, CD34⁺ cells were transduced and cultured in SFEMII and cytokines as described above. Cells were collected on days 3, 7, and 10 for analysis. A cocktail of antibodies described in Supplemental Table 3 was used to phenotypically define human CD34⁺ cells into LT-HSCs (CD34⁺CD45RA^–ITGA3⁺EPCR⁺CD90⁺) and ST-HSCs (CD34⁺CD45RA^–ITGA3^–EPCR^–CD90⁺). Cells were stained for 30 minutes at 4°C. All cell growth kinetic studies were performed on a BD Biosciences LSRFortessa. Flow cytometry data were analyzed in FlowJo.

Apoptosis studies were performed on sorted GFP⁺CD34⁺ cells cultured for 7 days after transduction. Briefly, transduced cells were counted on day 7 and 100,000 cells per group were stained with Annexin V and 7-AAD. The kit used was the APC Annexin V Apoptosis Detection Kit with 7-AAD (BioLegend) and the staining protocol was as per the manufacturer’s instructions.

Cell cycle studies were performed on sorted GFP⁺CD34⁺ cells cultured for 4 days after transduction. A total of 100,000 cells were stained using the APC BrdU Flow Kit (BD Biosciences), as per the manufacturer’s instructions.

Senescence studies were performed on sorted GFP⁺CD34⁺ cells on days 7 and 15 after transduction. For mTOR inhibition studies, 25 nM rapamycin (MilliporeSigma) was added every 7 days, and 100,000 cells were stained for β-galactosidase using the Cell Senescence β-Galactosidase Staining Kit (MedChemExpress). Stained cells were imaged and captured at ×20 magnification using an Olympus BX41 microscope attached to an Excelis HD camera and monitoring system. Cells that turned blue were counted as senescent cells.

Mitochondrial studies

Mitochondrial imaging was performed on sorted GFP⁺CD34⁺ cells on day 7 after LVV transduction. Approximately 5,000 cells were seeded onto chamber slides (Polysciences) and allowed to adhere for 1 hour at 37°C in a 5% CO₂ incubator. For mitochondrial labeling, cells were stained with 0.1 μmol/L MitoBright LT Deep Red (Dojindo Laboratories) for 30 minutes at 37°C, washed twice with PBS, and fixed with 4.2% paraformaldehyde (BD Biosciences) for 15 minutes at room temperature. Fixed cells were subsequently washed twice with PBS and counterstained with DAPI (Thermo Fisher Scientific) for 10 minutes, mounted using ProLong Glass Antifade Mountant (Thermo Fisher Scientific), and sealed. Images were acquired using a Zeiss LSM 980 confocal microscope equipped with a ×63 oil-immersion objective at 0.3 μm z-stack intervals. Automated quantification was performed using CellProfiler software, following described analytical pipelines (60). Quantitative measurements of mitochondrial morphology were subsequently calculated from segmented mitochondrial structures.

OCR was measured using the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies). Transduced CD34⁺ cells were cultured for 5 days followed by FACS to isolate GFP⁺CD34⁺CD45RA^–EPCR⁺ cells. Antibodies used are listed in Supplemental Table 3. Sorted cells were resuspended in Seahorse XF RPMI medium (Agilent Technologies) supplemented with 10 mM glucose, 2 mM glutamine, and 1 mM Na-pyruvate, and 25,000 cells were seeded onto a Seahorse XFe96 plate (Agilent Technologies) previously coated with fibronectin. The plate was centrifuged at 800g for 3 minutes followed by incubation for 50 minutes at 37°C in a no-CO₂ incubator. Mitochondrial stress test parameters were measured under basal conditions followed by sequential injection of 15 μM oligomycin A (MilliporeSigma), 20 μM carbonilcyanide p-triflouromethoxyphenylhydrazone (FCCP; MilliporeSigma), 5 μM rotenone plus 5 μM antimycin A (MilliporeSigma), and 50 mM 2-deoxy-D-arabino-hexose (Cayman Chemical). The cell number was normalized using the CyQUANT Cell Proliferation Assay kit (Thermo Fisher Scientific). Results were analyzed using the Wave program (Agilent Technologies) after being normalized by cell number/well.

scRNA-seq

Cell collection. Human CD34⁺ cells were transduced using a lentivirus into the following 4 groups: (a) EV control, (b) SEPT6 shmiR+WT SEPT6, (c) SEPT6 shmiR+GM1, and (d) SEPT6 shmiR+GM1+SM. The 4 groups were cultured for 5 days as described above. On day 5, cells were sorted into a GFP⁺CD34⁺CD45RA^– fraction and submitted to the Translational Immunogenomics Lab (TIGL) at the Dana-Farber Cancer Institute for further scRNA-seq analysis. Extracted human cells were resuspended in PBS with 0.4% bovine serum albumin (BSA) at the concentration of 1,000 cells/μL. Approximately 20,000 cells were loaded onto a Chromium X instrument (10X Genomics) according to the manufacturer’s instructions. The scRNA-seq libraries were processed using a Chromium GEM-X Single Cell 5′ Kit v3 (10X Genomics). Quality controls for amplified cDNA libraries and final sequencing libraries were performed using a Bioanalyzer High Sensitivity DNA Kit (Agilent). The sequencing libraries for scRNA-seq were normalized to 4 nM concentration and pooled. The pooled sequencing libraries were sequenced on the Illumina NovaSeq S4 300 cycle platform. The sequencing parameters used were: Read 1 of 28 nt, Read 2 of 90 nt, Index 1 of 10 nt, and Index 2 of 10 nt.

Data preprocessing and quality control. Raw scRNA-seq data were aligned to the human reference genome (GRCh38) using the Cell Ranger pipeline (v8.0.0, 10X Genomics). Quality control, integration, clustering, annotation, and differential expression analyses were performed with the Seurat R package (v5.0.2) (61). Cells were retained if they contained more than 500 detected genes, more than 500 transcripts, and less than 20% of reads mapped to mitochondrial genes. Potential doublets were identified and removed using the scDblFinder package (62). Genes expressed in fewer than 10 cells were excluded.

Analysis of scRNA-seq data. The count matrix was normalized using the sctransform function in Seurat (63), filtering out mitochondrial variation. Principal component analysis (PCA) was applied to variable features. The top 30 PCs were used for batch correction and sample integration through an anchor-based reciprocal PCA integration process, nearest-neighbor graph construction, and Uniform Manifold Approximation and Projection (UMAP) dimensional reduction. Clustering was tested across multiple resolutions, and a resolution of 0.5 was selected.

Cell type annotation was performed by combining the top 20 cluster-specific markers identified with FindAllMarkers function in Seurat with predictions from the Azimuth annotation package (64). A non-parametric Wilcoxon rank-sum test was applied to perform differential expression tests and volcano plots were generated using Enhanced Volcano package (65).

Pathway analysis. In the differential expression tests, genes with an absolute log₂(fold change) of greater than 0.25 and P value of less than 0.01 were defined as DEGs. Up- or downregulated DEGs were separately submitted to STRING (https://string-db.org/) for the protein-protein interaction and GO Biological Process (GOBP) enrichment analyses. To consider the weights of all the genes, genes were ordered in a ranked list based on the fold change. Gene set enrichment analysis (GSEA) (v4.3.3) (66, 67) was used to screen the enrichment of a priori–defined sets of genes in the preranked gene list. The gene set database was c5.go.bp.v2025.1.Hs.symbols.gmt from the Molecular Signatures Database (MSigDB) (66, 68) and the number of permutations was 1,000. Top 10 up- or downregulated gene sets were visualized as bubble plots.

Statistics

Data are presented as mean ± SD unless otherwise indicated. Experiments were performed a minimum of 3 times using different biological donors for human CD34⁺ cells. Xenotransplantation and scRNA-seq and bulk RNA-seq were repeated twice. Statistical differences were determined using 1-way ANOVA or 2-way ANOVA followed by a Tukey-Kramer post hoc analysis. Significance was set at a P value of less than 0.05. Statistical analyses were performed using Prism (GraphPad Software).

Study approval

All mice experiments were approved by the Boston Children’s Hospital Institutional Animal Care and Use Committee.

Data availability

Values for all data points in graphs are reported in the Supporting Data Values file. The scRNA-seq data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (GEO) (69) and are accessible through GEO Series accession number GSE338788 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE338788).