Mutant Samd9l expression impairs hematopoiesis and induces bone marrow failure in mice

SAMD9 and SAMD9L germline mutations have recently emerged as a new class of predispositions to pediatric myeloid neoplasms. Patients commonly have impaired hematopoiesis, hypocellular marrows, and a greater risk of developing clonal chromosome 7 deletions leading to MDS and AML. We recently demonstrated that expressing SAMD9 or SAMD9L mutations in hematopoietic cells suppresses their proliferation and induces cell death. Here, we generated a mouse model that conditionally expresses mutant Samd9l to assess the in vivo impact on hematopoiesis. Using a range of in vivo and ex vivo assays, we showed that cells with heterozygous Samd9l mutations have impaired stemness relative to wild-type counterparts, which was exacerbated by inflammatory stimuli, and ultimately led to bone marrow hypocellularity. Genomic and phenotypic analyses recapitulated many of the hematopoietic cellular phenotypes observed in patients with SAMD9 or SAMD9L mutations, including lymphopenia, and pinpointed TGF-β as a potential targetable pathway. Further, we observed nonrandom genetic deletion of the mutant Samd9l locus on mouse chromosome 6, mimicking chromosome 7 deletions observed in patients. Collectively, our study has enhanced our understanding of mutant Samd9l hematopoietic phenotypes, emphasized the synergistic role of inflammation in exaggerating the associated hematopoietic defects, and provided insights into potential therapeutic options for patients.

SAMD9 and SAMD9L are located on human 7q21 and share 60% amino acid homology, while the mouse genome only encodes Samd9l, suggesting functional redundancies (9). They are IFNresponsive genes, and their expression dramatically increases following exposure to IFNs and TNF-α (10)(11)(12). In addition to the germline mutations observed in patients with BMF and myeloid neoplasms, SAMD9L mutations have been associated with inflammatory diseases (13). Conversely, dysregulation of inflammatory pathways, such as TNF-α and IFNs, negatively influence hematopoietic cells and are associated with hematological disorders, including BMF (14)(15)(16)(17). We recently highlighted the detrimental effects of overexpressing SAMD9 and SAMD9L mutations ex vivo in primary hematopoietic cells, including the suppression of proliferation and ultimately the induction of cell death (18). However, the impact of Samd9l mutations on in vivo hematopoietic cell functional output, fitness, and differentiation from the endogenous IFN-responsive locus has yet to be evaluated.
Here, we describe a conditional knockin mouse model carrying a pathogenic mutation in Samd9l that is homologous to a mutation identified in our previously published patient cohort (4). Phenotypic and single-cell analyses of the hematopoietic cells from the Samd9l-mutant mouse showed dramatic impairment of cell fitness and lineage composition, leading to a variably penetrant BMF and nonrandom chromosomal loss involving the mutant allele. Inflammatory stimuli upregulated Samd9l expression and exacerbated the mutant Samd9l-associated hematopoietic phenotypes. We also showed that the TGF-β pathway is activated in Samd9l-mutant cells and its inhibition restored their clonogenicity via reverting cell death. These collective data enhance our understanding of pediatric MDS/BMF arising from germline SAMD9L mutations and provide insights into potential therapeutic options for this recently described new class of genetic predispositions.
SAMD9 and SAMD9L germline mutations have recently emerged as a new class of predispositions to pediatric myeloid neoplasms. Patients commonly have impaired hematopoiesis, hypocellular marrows, and a greater risk of developing clonal chromosome 7 deletions leading to MDS and AML. We recently demonstrated that expressing SAMD9 or SAMD9L mutations in hematopoietic cells suppresses their proliferation and induces cell death. Here, we generated a mouse model that conditionally expresses mutant Samd9l to assess the in vivo impact on hematopoiesis. Using a range of in vivo and ex vivo assays, we showed that cells with heterozygous Samd9l mutations have impaired stemness relative to wild-type counterparts, which was exacerbated by inflammatory stimuli, and ultimately led to bone marrow hypocellularity. Genomic and phenotypic analyses recapitulated many of the hematopoietic cellular phenotypes observed in patients with SAMD9 or SAMD9L mutations, including lymphopenia, and pinpointed TGF-β as a potential targetable pathway. Further, we observed nonrandom genetic deletion of the mutant Samd9l locus on mouse chromosome 6, mimicking chromosome 7 deletions observed in patients. Collectively, our study has enhanced our understanding of mutant Samd9l hematopoietic phenotypes, emphasized the synergistic role of inflammation in exaggerating the associated hematopoietic defects, and provided insights into potential therapeutic options for patients. Mutant Samd9l expression impairs hematopoiesis and induces bone marrow failure in mice (Supplemental Figure 2D). Notably, we utilized a well-established flow cytometric assessment of B cell maturation stages (20) and demonstrated a clear suppression of the mature B cell populations in Samd9l-Mut mice relative to Samd9l-WT mice, whereas the less mature cells were not affected (Supplemental Figure 2, E and F). Collectively, these findings indicate that Samd9l-Mut expression affects all levels of the hematopoietic hierarchy, with the most significant impact on mature B cells, likely due to their elevated levels of Samd9l expression.
Single-cell transcriptomics was then used to further study the BM compartments of these mice. Using established lineage markers (21, 22) (Supplemental Figure 3A), 11 cell clusters were identified, and further consolidated into 5 main populations (myeloid, B cells, T cells, erythroid, and HSPCs) ( Figure 2A). As noted by flow cytometry, the Samd9l-Mut have a clear decrease in B cell populations, an increase in myeloid cells, and additional alterations in the HSPC population ( Figure 2B and Supplemental Figure 3, B-D). The HSPC subcompartments were further defined by the expression of Cd34, Car1, Mpo, Fn1, and Il7 (Supplemental Figure  3E) using single-cell RNA sequencing (scRNA-seq). Differentially expressed gene (DEG) analysis of the HSPC population showed enrichment of genes involved in translation and cell division in the Samd9l-Mut relative to Samd9l-WT mice (Supplemental Figure 3F). Of the myeloid cell populations, Samd9l-Mut mice had a relative increase in MPs and a decrease in monocytes and macrophages, along with expression changes associated with cell cycle regulation and immune response ( Figure 2C and Supplemental Figure 3G). Moreover, Samd9l-Mut mice exhibited a relative increase in the less differentiated B cell population (Pro-B) and a relative decrease in the more mature populations (Pre-B and immature B) ( Figure 2D). Notably, we observed an enrichment of genes related to translation, apoptosis, and DNA replication in B cells in Samd9l-Mut (Supplemental Figure 3H). Similarly, complete blood counts (CBCs) from these mice demonstrated global lymphopenia with a relative increase in the percentage of myeloid cells ( Figure 2E). These changes in the mature cells were supported by immunohistochemistry (IHC) on BM and spleen sections, which showed a relative decrease in B cells (stained by anti-PAX-5) and a slight increase in myeloid precursors (stained by anti-MPO) in Samd9l-Mut versus Samd9l-WT (Figure 2, F and G). Together, these scRNA-seq data confirm the global hematopoietic lineage alterations observed by flow cytometry and further highlight the clear deterioration of B cells.
Samd9l mutation impairs hematopoietic cell fitness and repopulation potential. Next, we assessed the functional output of Samd9lmutant cells and their relative hematopoietic fitness. We first examined the serial colony formation of BM cells from Samd9l-Mut mice as a correlate of self-renewal potential ( Figure 3A). Relative to wild-type Samd9l controls, Samd9l-Mut cells exhibited a gradual decrease in clonogenicity that almost completely diminished at the third replating ( Figure 3B). Samd9l-KO cells showed a mildly stronger replating potential and a greater number of cells per colony than wild-type controls ( Figure 3B and Supplemental Figure 4A), reminiscent of the reported repopulation advantage resulting from Samd9l haploinsufficiency (19). We observed no apparent differences in colony subtypes (Supplemental Figure 4B). Next, we transplanted Lin -BM cells from CD45.2 mice (C57BL/6,

Results
Generation of a conditional mutant Samd9l-knockin allele. We generated a conditional knockin mouse model carrying the Samd9l-W1171R point mutation, corresponding to the human W1180R mutation previously reported by our group (4) to study the in vivo impact of Samd9l mutations. Gene targeting was used to introduce LoxP sites flanking a GFP-fused Samd9l-WT exon 2 (the first coding exon) followed by an FRT-flanked neomycin selection cassette upstream of a stop codon and an mCherry-fused mutant Samd9l exon 2 containing the W1171R mutation (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI158869DS1). Following removal of the neomycin cassette, the generated Samd9l fl/fl (Samd9l cKI+/+ ) mice were bred with transgenic mice containing the hematopoietic cell-specific Vav1-Cre to generate a heterozygous mutation (which is universally observed in patients) in the hematopoietic compartment (Supplemental Figure 1, B and C). These Vav1-Cre +/Tg Samd9l cKI+/mice (Samd9l-Mut) are viable and fertile with offspring born at the expected Mendelian ratios ( Figure 1, A and B). Of note, attempts to induce the mutation throughout the animal early in development by crossing with CMV-Cre mice led to a lack of viable offspring. Three-month-old Samd9l-Mut mice demonstrated a distinct reduction in white blood cell (WBC) counts, with no apparent difference in red blood cell (RBC) and platelet counts ( Figure 1C) when compared to the Samd9l cKI+/mice (Samd9l-WT) and the previously reported Samd9l -/mice (Samd9l-KO) (19).
Samd9l-mutant hematopoiesis favors myeloid commitment. We next compared the BM compartments of 3-month-old Samd9l-Mut mice with Samd9l-KO, Samd9l-WT, and native C57BL/6 mice and observed a significant relative increase in the percentage of Lin -cKit + myeloid progenitors (MPs), with a decrease in the lymphoid lines, including the common lymphoid progenitor (CLP; Lin -cKit lo Sca-1 lo CD127 + ) ( Figure 1D and Supplemental Figure 1, D and E). We also observed an increase in granulocytic/monocytic progenitors (GMPs). Within the hematopoietic stem and progenitor cell (HSPC) compartment, both the MPP2 population (Lin -cKit + Sca-1 + CD48 + CD150 + , the precursors for myeloid cells) and the long-term HSCs (LT-HSCs; Lin -cKit + Sca-1 + CD48 -CD150 + ) were increased in the Samd9l-Mut mice ( Figure 1E). Notably, lymphoid progenitors had lower percentages of EdU incorporation in Samd9l-Mut BM cells relative to other groups, whereas MP and KSL (Lin -cKit + Sca-1 + ) populations had higher relative percentages (Supplemental Figure 1F). This skewing was further noted in the mature cells, as we observed a significant decrease in B cells accompanied by a relative increase in the percentage of the mature myeloid cells in the PB, BM, and spleen (Figure 1, F and G, and Supplemental Figure 1, G and H). This effect was pronounced in B cells in Samd9l-Mut mice, which also showed a lower EdU incorporation percentage of the total proliferating cells relative to myeloid and T cells (Supplemental Figure 1I). We then wanted to examine whether the basal expression level of the Samd9l gene in different cell types could correlate to their diverse sensitivity to the mutations. To test this, we analyzed publicly available data sets and observed a higher Samd9l level in B cells relative to other cell types in humans or mice, which we also confirmed by qPCR analysis of sorted murine cells (Supplemental Figure 2, A-C). Moreover, Samd9l expression was directly proportional to B cell maturation  observed similar patterns in the BM and spleen at sacrifice (Supplemental Figure 4D). After the adoptive transfer into irradiated recipients, the Samd9l-Mut donor compartment showed a relative increase in the percentage of myeloid cells, with a decrease in B cells in comparison with Samd9l-WT donors (Supplemental Figure  4, E-G), which mirrors the impacts shown in Figure 1. Altogether, Samd9l-mutant cells demonstrated a profound lack of fitness and repopulation capacities ex vivo and in vivo.
Inflammation upregulates mutant Samd9l and further potentiates its pathogenicity. Considering that Samd9l is a known IFNresponsive gene (10), we next addressed the impact of inflammation-induced upregulation of Samd9l-Mut expression on hematopoietic cell function. We confirmed an increase in SAMD9 and SAMD9L RNA and protein levels after IFN-α treatment in HEK293T and human cord-blood CD34 + cells (Supplemental  Figure 3C). Among the compared groups, only Samd9l-Mut cells were significantly outcompeted by wild-type CD45.1 cells ( Figure 3D). This phenotype was sustained even at 16 weeks after transplantation (data not shown). Upon gating on the CD45.2 populations, we observed a proportional increase in myeloid cells at the expense of lymphoid cells in the Samd9l-Mut relative to other groups (Supplemental Figure 4C). To determine whether the competitive disadvantage of Samd9l-Mut cells resulted from impaired stem cell fitness and not a failure of homing to the BM, we performed a competitive transplantation experiment via intrafemoral injections ( Figure 3E). Consistent with the i.v. data, weekly PB analysis showed a gradual competitive disadvantage of Samd9l-Mut cells, but not the Samd9l-WT cells ( Figure 3F). We also  For panels G and J, a longitudinal mixed-effects regression model was used for statistical analysis followed by pairwise Tukey-adjusted tests. For panel K, the data were not normally distributed; therefore, the Kruskal-Wallis test was initially performed and followed by Wilcoxon's rank-sum test for pairwise comparisons. For all other panels, 2-way ANOVA was used to evaluate the main effects and genotype/treatment interactions and followed by Tukey's pairwise comparisons. Error bars indicate the SEM for biological replicates. Blue, Samd9l-KO, black, Samd9l-WT; red, Samd9l-Mut. *P < 0.05; **P < 0.01; ***P < 0.001 compared with vehicle-treated groups. # P < 0.05 compared with pI:pC-treated groups. Color indicates the comparison group.
our previous in vitro studies (18). Next, we utilized polyinosinic/ polycytidylic acid (pI:pC), a Toll-like receptor 3 agonist, to induce type-I IFNs in vivo. We treated Samd9l-WT, Samd9l-Mut, and Samd9l-KO mice with 5 mg/kg pI:pC twice a week for 4 weeks (23,24). Similar to IFN-α, pI:pC increased SAMD9L protein expression in both Samd9l-WT and Samd9l-Mut mice ( Figure  4D), with a resulting significant reduction in colony formation and induction of apoptosis in BM cells from Samd9l-Mut ( Figure  4, E and F). Likewise, BM cells isolated from a patient harboring the SAMD9L-S626L mutation treated with IFN-α exhibited a complete loss in colony formation relative to vehicle-treated cells or control cord-blood healthy donor cells ( Figure 4G). Our data collectively demonstrate that inflammatory stress mediates an increase in Samd9l expression, ultimately enhancing the cellular phenotypes associated with the expression of a mutant allele.
Inflammatory stimulus further impairs hematopoietic cell fitness. Next, we assessed the in vivo impact of inflammation on hematopoietic cell fitness using a 5:1 competitive BM transplantation model of Samd9l-WT or Samd9l-Mut mice (CD45.2) pretreated with pI:pC or vehicle ( Figure 4H). Despite the initial higher ratio of input CD45.2 cells at injection, Samd9l-Mut groups were outcompeted by CD45.1 cells. Notably, we observed a further competitive disadvantage of pI:pC-treated Samd9l-Mut cells in comparison with vehicle-treated Samd9l-Mut (Figure 4, I and J, and Supplemental Figure 5G). Similarly, pI:pC decreased Samd9l-WT fitness relative to vehicle-treated cells (Figure 4, I and J, and Supplemental Figure 5G). We observed no significant difference in mature cell composition in both Samd9l-Mut and Samd9l-WT treated with pI:pC relative to vehicle (Supplemental Figure 5H). Importantly, while vehicle-treated Samd9l-Mut donor cells had significant increases in B cell apoptosis, pI:pC-treated mice had significant increases in apoptosis for B and myeloid cells as a result of the combined transplant and inflammatory stresses ( Figure 4K). Similarly, 1:1 competitive transplantation also showed lowered Samd9l-Mut fitness with inflammation (Supplemental Figure 5, I-K).
The lack of fitness in Samd9l-mutant cells is partly mediated by TGF-β activation. We next performed RNA-seq analysis of Lin -cKit + HSPCs from Samd9l-WT or Samd9l-Mut mice with or without pI:pC treatment and found a variety of upregulated pathways (including inflammatory signaling such as TNF-α/NF-κB and JAK/STAT signaling and hematopoietic lineage) and downregulated pathways (MYC, ribosome, and DNA repair) ( Figure 5, A-C, and Supplemental Figure 6, A and B) (18). Importantly, the data revealed an upregulation of genes involved in TGF-β pathways in Samd9l-Mut mice treated with pI:pC relative to all other groups ( Figure 5, B and C, and Supplemental Figure 6, B and C). Gene set enrichment analysis (GSEA) confirmed the activation of genes involved in the TGF-β pathway in Samd9l-Mut cells in response to pI:pC ( Figure  5D). We also found a consistent TGF-β enrichment in hCD34 + cells with SAMD9L-W1180R overexpression (Supplemental Figure 7A). Mechanistically, TGF-β transmits intracellular signals through the SMAD family and its activity is measured by SMAD2/3 phosphorylation (25). Intracellular phospho-flow revealed hyperphosphorylation of SMAD2/3 in total BM cells from Samd9l-Mut after IFN-α or pI:pC treatment (Supplemental Figure 7, B and C). Among BM cells, SMAD2/3 activation was more obvious in the B cell population of Samd9l-Mut in response to inflammatory stimuli ( Figure 5, E and F, and Supplemental Figure 7D). Consistent with previous reports on other BM disorders (25,26), treatment with SD-208, a small-molecule TGF-β inhibitor, rescued the mutation-dependent reduction in clonogenicity and reverted the apoptotic phenotype of Samd9l-Mut cells, with no effect on Samd9l-WT cells ( Figure  5G and Supplemental Figure 7E). Similarly, SD-208 improved the impaired clonogenicity in BM cells isolated from a patient with the SAMD9L-S626L mutation ( Figure 5H).
Inflammation augments B cell lymphopenia and impairs erythroid maturation. We next treated Samd9l-Mut and Samd9l-WT mice with pI:pC to assess the impact of inflammatory stimuli on the hematopoietic compartments in vivo ( Figure 6A). PB counts, particularly lymphocytes, were decreased after pI:pC treatment in Samd9l-Mut and Samd9l-WT mice; however, there were no changes in myeloid cells ( Figure 6B). Morphologic evaluation of BM revealed predominantly mature neutrophils, with scattered lymphocytes, myeloblasts, and erythroblasts in Samd9l-WT mice, as expected, regardless of whether they were treated with vehicle or pI:pC (Supplemental Figure 8A). In contrast, Samd9l-Mut mice had a higher abundance of myeloblasts and fewer lymphocytes relative to Samd9l-WT groups, and these effects were remarkably exaggerated with pI:pC, with no significant dysplasia (Supplemental Table  1). Histological examination of BM demonstrated no architectural changes; however, a clear distortion of splenic architecture in the pI:pC-treated Samd9l-Mut mice was observed (Supplemental Figure 8, B and C). Additionally, thymuses from Samd9l-Mut treated with pI:pC exhibited either cortical apoptosis with tingible body macrophages or atypical hyperplasia (Supplemental Figure 8D). Importantly, IHC studies confirmed the reduction in lymphocytes and moderate myeloid hyperplasia in pI:pC-treated Samd9l-Mut mice compared with all other groups ( Figure 6, C and D).
We next performed scRNA-seq on both whole BM (WBM) and sorted Lin -cKit + HSPCs from Samd9l-WT and Samd9l-Mut mice with or without pI:pC (Supplemental Figure 9, A-C). Among the WBM populations, consistent with Figure 2, A and B, we observed a significant decrease in B cells and a slight increase in the percentage of the myeloid populations in the Samd9l-Mut mice, which were further exacerbated with pI:pC (Figure 7, A and B, and Supplemental Figure 9, C-E). DEGs of the HSPC population included enriched genes involved in DNA replication, translation, and response to viral genes in the pI:pC-treated Samd9l-Mut mouse relative to the vehicle-treated mouse (Supplemental Figure  9E). Myeloid differentiation trajectory supported our previous observation of the increased percentage of MPs in Samd9l-Mut, which appear more pronounced with pI:pC (Supplemental Figure  9F). B cell maturation trajectory analysis showed a decrease in the relatively mature cells in the BM (immature-B) in Samd9l-Mut versus Samd9l-WT. Treatment with pI:pC almost abolished B cells at all stages, likely due to cellular damage indicated by the increased proapoptotic and inflammatory markers and decreased proliferation markers (Supplemental Figure 9, G and H). Our single-cell data from Lin -cKit + cells showed a remarkable decrease in the more mature erythroid progenitors in Samd9l-Mut after pI:pC (Figure 7, C and D). Correspondingly, staining of BM sections using anti-GATA1 or anti-CD235a showed a distinct decrease in erythroid cells and an increased myeloid/erythroid ratio in the pI:pC-treated Samd9l-Mut ( Figure 7E and Supplemental Figure  Figure 10G). The mice also demonstrated marked splenomegaly (Supplemental Figure  10H). Of note, the survivors did not have a change in BM cellularity, nor any acquired somatic revertant mutations in Samd9l that would provide a rescue mechanism indicated by targeted Sam-d9l sequencing, as observed in patients (10). Strikingly, however, we observed a significant downregulation of 42 genes (including Samd9l) in a specific locus on chr6 in 1 out of 3 pI:pC-treated Samd9l-Mut mice analyzed by RNA-seq (Supplemental Figure  11A). The affected region extended for approximately 20 megabase pairs along chr6 qA1-qA3.1 (chr6:3,322,257-23,605,136) and is syntenic to human chr7 ( Figure 8E). Interestingly, this partial deletion in chr6 only occurs on the mutant allele, leaving an intact Samd9l wild-type allele ( Figure 8F). This phenomenon was further investigated with fluorescence in situ hybridization (FISH) on splenic sections using custom probes to detect chr6:3,496,083-3,687,193 (proximal to Samd9l within the affected region) and chr6:28,129,437-28,303,622 (distal end of chr6 outside the affected region) loci ( Figure 8G). We observed deletions at the proximal locus but not the distal locus in pI:pC-treated Samd9l-Mut mice ( Figure 8H). As predicted, we also observed a deletion in an inter-10A). The spleens of pI:pC-treated Samd9l-Mut mice had a markedly expanded red pulp with distortion of white pulp, possibly due to extramedullary hematopoiesis, which is commonly seen in mice in response to the diminished ability of BM to generate erythroid cells in stress conditions (Supplemental Figure 10, B and C) (27). We complemented these findings with flow cytometric assessment of erythroid maturation in both the spleen and PB, as previously reported (28). Treatment with pI:pC resulted in a significant increase in immature erythroid precursors in both spleen and PB and a decrease in the circulating mature erythroid cells in Samd9l-Mut mice (Supplemental Figure 10, D and E). In support of these data, RBCs and platelets were significantly reduced exclusively in the pI:pC-treated Samd9l-Mut mice, consistent with pancytopenia (Supplemental Figure 10F). Together, our data demonstrated that inflammation potentiates Samd9l-Mut phenotypes, in particular, pancytopenia with lymphopenia, corresponding to clinical observations in patients with SAMD9L mutations (13,(29)(30)(31).
Samd9l-Mut mice recapitulate cellular phenotypes seen in patients. Finally, prolonged evaluation of these Samd9l-Mut mice demonstrated a phenotype mimicking BMF with incomplete penetrance, with approximately one-third of mice becoming symptomatic at the age of 6 months ( Figure 8A). The moribund mice  For panel B, the Kruskal-Wallis test was used to perform an initial comparison across all groups, and followed by pairwise comparisons with Wilcoxon's rank-sum test. *P < 0.05, **P < 0.01, ***P < 0.001 compared with vehicle-treated groups. # P < 0.05, ### P < 0.001 compared with pI:pC-treated groups. Color indicates the comparison group. Error bars indicate the SEM for biological replicates. Black, Samd9l-WT, red, Samd9l-Mut; stripes, pI:pC; solid, vehicle.
stress, can impair HSC fitness and ultimately lead to BM hypocellularity, a hallmark of pediatric MDS and BMF syndromes.
Additionally, we demonstrated a distinct lymphopenia that was exacerbated with inflammation. We correlated this observation to the higher expression levels of Samd9l in mature B cells, potentially rendering them more sensitive to the effects of expression of the mutant allele. In fact, this is not completely surprising, as clinical reports have shown that lymphopenia in general, and particularly B cell ablations, are observed in BMF syndromes, including patients with mutant SAMD9L (29,31). A recent report also demonstrated B lymphopenia in mice with the Samd9l-D764N mutation leading to phenotypes that mimic MIRAGE syndrome (32). Inflammation is known to compromise leukocyte production and affect B cell maturation and likely synergizes with mutant Samd9l expression in our model (33). We also observed a concomitant increase in relative percentages of myeloid precursors in Samd9l-Mut mice, especially after inflammation, a potential response to the associated lymphopenia, a process known as emergency granulopoiesis (34).
Inflammation resulting from viral infections is a potential mediator of BM disorders via suppression of hematopoiesis (35)(36)(37). In fact, SAMD9 and SAMD9L are host restriction factors activated as a defensive mechanism in response to infection with viruses such as vaccinia, myxoma, and rhinovirus (38)(39)(40). Our data emphasize the profound impact of inflammatory stimuli on inducing mutant Samd9l expression from the endogenous locus, triggering exaggerated hematopoietic suppression and apoptosis. Our hypothesis, supported by our data, is that inflammatory pathways can be particularly harmful to patients with SAMD9L muta-  Figure 8H). In total, we identified 4 mice with chr6 deletions within a cohort of Samd9l-Mut mice treated with pI:pC (n = 26) using either FISH and/or a validated qPCR screening analysis. The deletions were quantified by FISH in at least 3 mice and were observed in 70%, 53%, and 24% of the examined spleen cells, including those in the red pulp where myeloid cells are predominant.

Discussion
Here we generated a mouse model to evaluate the in vivo hematopoietic impact of mutant Samd9l expression from the native locus. Our data demonstrated that heterozygous Samd9l mutation expression within hematopoietic cells impaired cell proliferation and self-renewal, which was potentiated by inflammation, leading to the induction of apoptosis, consistent with our previous ex vivo observations (18). Using a range of phenotypic and single-cell transcriptomic analyses, we found a decrease in WBC counts and significant alterations in the frequencies of hematopoietic compartments in the BM of Samd9l-mutant mice relative to wild-type or Samd9l-KO mice, including MPs and LT-HSCs. However, the mutant cells paradoxically lack stem cell self-renewal, as demonstrated by serial colony-forming unit cell (CFU-C) replating and competitive transplantations. These data suggest that Samd9l-mutant mice have elevated numbers of poorly functioning stem cells, and this increase may occur to compensate for the impaired steady-state hematopoiesis caused by the mutations. Collectively, these data show that in vivo expression of mutant Samd9l, in part induced by inflammatory with mutant SAMD9L may contribute to the disease progression, which requires further evaluation. Our in vivo model recapitulates many of the clinical manifestations of patients with SAMD9L mutations, including pan-tions and could accelerate BM suppression, potentially leading to mortality or adaptive responses, such as the outgrowth of clones with chromosomal deletions. This also raises an important question of whether microenvironmental inflammation in patients EDTA-coated tubes. BM was collected by flushing the long bones with 2% FBS in PBS. Spleens were harvested, crushed, and strained through a 0.4 μm filter. Lineage depletion was done using the EasySep Mouse Hematopoietic Cell Isolation kit (STEMCELL Technologies).
Flow cytometry. Cell surface staining was performed using fluorescently labeled antibodies (Supplemental Table 2). For apoptosis assessment, cells were washed with annexin V binding buffer (Thermo Fisher Scientific) and stained with annexin V for 15 minutes followed by DAPI. Flow cytometry was done using an LSR FORTESSA II (BD Biosciences) and analyzed using FlowJo software (TreeStar).
Colony-forming unit. Lineage-depleted BM (3,000 cells) was cultured in MethoCult GF M3434 medium (STEMCELL Technologies) and incubated at 37°C with 5% CO 2 for 7 days. For serial replating, colonies were counted and harvested by washing in PBS and 10,000 cells were recultured in MethoCult medium for an additional 7 days for 2 rounds.
RNA-seq and scRNA-seq. For RNA-seq, RNA was isolated from Lin -cKit + sorted BM cells and sequenced on a NovaSeq 6000 (Illumina) as previously reported (18). scRNA-seq was performed on WBM or Lin -cKit + sorted BM cells (n = 1 per group). Cells were prepared and quantified following the 10× Genomics protocol. Cells (4,000) were isolated for single-cell barcoding and 5′ GEX library preparation using the 10× Genomics Chromium Next-GEM Single Cell V(D) J Reagent Kit v1.1 following the manufacturer's protocol. Libraries were pooled and sequenced on a NovaSeq 6000 using SP Reagent kit v1.5 (Illumina, 20028401).
RNA-seq read, mapping, and data analysis. Gene expression data were analyzed using the following guidelines. Reads from aligned bam files were assigned to genes and counted using HTSeq v0.11.2 (47) with the GENCODE mouse release 67 gene annotation. We generated the gene count matrix, and the log 2 CPM (counts per million) values were used for downstream analysis. We required that at least 2 samples (equal to the smallest group in the RNA cohort) should have 10 or more read counts per million reads sequenced in order to consider a gene as expressed. Differential gene expression analysis was done using R package Limma v3.32.10 (48). P values were adjusted by the Benjamini-Hochberg method to calculate false discovery rate (FDR). Genes with FDR less than 0.05 were considered significantly differentially expressed. Clustering and visualization were performed by R package heatmap v1.0.12 with Euclidian distance and Ward.D2 linkage. Pathway enrichment analysis was performed by R package clusterProfiler (v3.6.0). Rich Factor and other plots were created using R package ggplot 2 (v3.0.0). GSEA (49) was performed by GSEA v.1.0 using MSigDB genesets c2.all v7.4 for each mouse group (http://www.gsea-msigdb.org/). scRNA-seq barcoding, mapping, and data analysis. The featurebarcode count data matrix for each sample was generated from raw sequence FASTQ files using Cell Ranger v3.0 (50) and the GRCm38 mouse reference genome. The output data from Cell Ranger was further analyzed in R environment (v4.0.2, RRID: SCR_001905) using R package Seurat (v3.2.1, RRID: SCR_007322) with default settings. Count data matrices in each experiment were combined using the merge function, followed by excluding cells with less than 250 or greater than 6,000 features, or greater than 5% mitochondrial transcripts. Feature counts for each cell were normalized to 1 × 10 4 counts/cell and natural log transformed using the Normalize Data function. The dimensional reduction was performed by principal component analysis (PCA) using the top 500 variable genes identified by vst analysis using the Find Variable Features function, followed by Jack Straw analysis and UMAP (uni-cytopenia, BM hypocellularity, and incomplete clinical penetrance. Ribosome regulatory pathways were significantly reduced after stimulating with inflammation, which upregulates mutant Samd9l expression, leading to translation suppression reminiscent of our ex vivo overexpression model (18) and consistent with recent clinical observations (13,29). Further, our model highlights the variable clinicopathological features in patients with germline SAMD9 and SAMD9L mutations. In some patients, the mutant hematopoietic cells overcome the associated impairment by a potentially beneficial somatic-revertant mosaicism that restores normal hematopoiesis (via copy-number neutral loss of heterozygosity or loss-of-function mutations in cis with the germline allele). In contrast, other patients develop a clone with a potentially deleterious removal of the pathogenic allele on chr7, which can ultimately lead to MDS/AML (5,8,41,42). In our mouse model, we did not observe cis revertant mutation acquisition. However, importantly, we observed a deletion of a specific region in mouse chr6 syntenic to human chr7 that includes common genes like Samd9l, Hepacam2, and others. Part of this genetic locus, including Samd9l, was also reported to modulate proinflammatory cytokines, including TNF-α and IFN-α, after viral infections (43). We concluded that the observed chr6 focal deletions in our mouse model mimics the monosomy 7 deletions seen in SAMD9L patients. To the best of our knowledge, our mouse model is the first to demonstrate this clinically relevant and rare phenomenon of chromosomal deletion affecting the Samd9l locus secondary to a deleterious Samd9l variant. Given that the chromosomal deletions affect the mutant allele, we postulate that this genetic haploinsufficiency is an adaptive response to release the stress caused by inflammation-induced Samd9l mutation expression. This concept is referred to as adaptation by aneuploidy (44).
Finally, our data exposed the TGF-β pathway as a potential therapeutic target in patients with SAMD9L mutations. Congruently, its upregulation is observed in similar BMF syndromes, such as Fanconi anemia and Shwachman-Diamond syndrome and is associated with impaired hematopoietic self-renewal (25,26). Importantly, treatment with SD-208 rescued the defective human and mouse cells with SAMD9L mutations, consistent with previous observations in other BM disorder syndromes (25,26). We propose TGF-β inhibition as a potential adjuvant therapy to further study as a protective measure against SAMD9L-mutationassociated hematopoietic impairment. In fact, a TGF-β inhibitor exhibited promising results in patients with low-and intermediaterisk MDS (45). Altogether, the demonstrated clinical relevance of our model establishes it as an outstanding reagent to study disease pathogenesis and identify potential therapeutic targets.

Methods
Animals. Samd9l-KO mice were provided by Hiroaki Honda and Toshiya Inaba at Hiroshima University (19). We enlisted Ingenious Targeting Laboratory to generate the conditional mouse model with a Samd9l-W1171R mutation corresponding to the human W1180R mutation using embryonic stem cell-based gene targeting, as previously reported (46). The targeting vector was constructed by subcloning from a C57BL/6 BAC clone using a homologous recombination-based technique. CD45.1 and C57BL/6 mice were obtained from The Jackson Laboratory. Blood was collected via retroorbital bleeding and kept in deconvolution. The camera and objective lens that were used were a Hamamatsu Orca Flash 4.0 camera and a 60× PlanApo N.A. 1.4 lens; 300 cells from each sample were analyzed.
Data availability. RNA-seq and scRNA-seq data were deposited into the NCBI Gene Expression Omnibus (GEO GSE190566 and GSE191147, respectively).
Statistics. Details about statistical comparisons are provided in each figure legend. For 2-group comparisons, Wilcoxon's test was performed to test the distribution difference. Comparisons across more than 2 groups were performed using ANOVA models followed by Tukey-adjusted pairwise comparisons or by performing the Kruskal-Wallis test followed by pairwise Wilcoxon comparisons. Longitudinal trends were evaluated with mixed-effects longitudinal regression models with the R packages lmer4 and emmeans R package (lmer4) (53,54). In one case, rank-based methods were used due to violation of the normality assumption. For survival data, Kaplan-Meier estimates were computed and plotted by genotype. The difference in survival distribution between genotypes was examined using an exact log-rank test. All the computations were done using R (55) and all P values are 2-sided.
Study approval. All animal studies including husbandry, breeding, and experimental procedures were performed in accordance with protocols approved by St. Jude Children's Research Hospital Institutional Animal Care and Use Committee.