Decoding functional hematopoietic progenitor cells in the adult human lung

Hematopoietic stem cells (HSCs) are self-renewing cells residing in the bone marrow (BM) that are ultimately responsible for production of all mature circulating blood cell lineages.¹ Despite the burden of maintaining hematopoiesis, HSCs are a rare cell type in the BM accounting for <0.01% of nucleated cells.^2,3 They occupy and are maintained by a specific BM niche yet are able to exit the BM environment and enter the circulation, although the explanation of this behavior is largely unknown.⁴ HSCs are central to the pathogenesis of serious disorders including myelodysplasia, acute and chronic leukemia, aplastic anemia,⁵ and clonal hematopoiesis,⁶ and HSC transplantation can be a lifesaving therapy. Previously, we identified a population of hematopoietic stem and progenitor cells (HSPCs) in the mouse lung, but their function was unclear.⁷ Motivated by this background, in this study we sought to determine whether HSPCs occupied the adult human lung, an organ with a vast vasculature that contains a wide-ranging repertoire of stromal and immune cells, and to determine the niche and function of these cells.

Fresh human tissues were obtained from deceased organ donors after either brain death or circulatory death (supplemental Table 1, available on the Blood website). All patients were on mechanical ventilation and treated in the intensive care unit until organ retrieval. Lungs and vertebral bodies were surgically recovered and immediately placed on ice. Peripheral blood (PB) was collected in sodium heparin tubes. Healthy human stem cell donors were treated with a granulocyte colony-stimulating factor mobilization regimen, and PB stem cells were collected by apheresis. We received 1 mL of the mononuclear fraction that would have otherwise been discarded. Isolation of cells, magnetic enrichment, antibody staining, cell sorting, and immune phenotyping by flow cytometry are described in the supplemental Methods.

Lineage depleted BM or lung cells were plated in MethoCult or MegaCult-C medium according to the manufacturer’s instructions. After incubation for 10 to 14 days at 37°C in 5% CO₂, hematopoietic colonies were scored based on morphologic and phenotypic criteria and quantified by manual counting using brightfield microscopy. Further details are described in the supplemental Methods.

NSG-SGM3 mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl Tg[CMV-IL3, CSF2, KITLG]1Eav/MloySzJ, Stock No: 013062) were purchased at the Jackson Laboratory and housed in a specific pathogen–free animal facility. One day before reconstitution, mice were preconditioned by sublethal irradiation (2.4 Gy).⁸ On the day of transplantation, lineage-negative (Lin⁻) lung and BM cells were thawed and 1.5 × 10⁶ viable cells were IV injected into the tail vein of recipient mice.⁸ Matched lung and BM cells were transplanted into recipient mice where possible. Ten weeks after transplantation, the BM, lung, and blood were investigated for engraftment of human cells. Further details are described in the supplemental Methods.

Live, Lin⁻ CD34⁺ cells from the lung and BM were sorted for subsequent 10× Genomics single-cell RNA sequencing (scRNA-seq). Live, Lin⁺ cells were collected for demultiplexing the samples from different individuals based on single nucleotide polymorphisms.⁹ Further details of scRNA-seq workflow and the bioinformatic analysis are provided in the supplemental Methods.

Human lung tissue from deceased organ donor samples was frozen in optimal cutting temperature on dry ice, and tissue blocks were processed at the facility of Resolve Biosciences, San Jose, for spatial transcriptomics using a custom 100 gene panel. Further details on spatial transcriptomics workflow and the bioinformatic analysis are provided in the supplemental Methods.

Statistical analysis was done with GraphPad Prism version 10.0.2, R/RStudio version 4.0.3, or Python version 3.12.1. Statistics for each analysis are described in the relevant section.

Our study does not involve human subjects because the tissue involved was obtained from deceased research donors or from samples obtained during routine clinical collections from apheresis stem cell donors that would have been otherwise discarded. The mouse xenotransplantation experiments were performed under the University of California San Francisco Institutional Animal Care and Use Committee approval (AN201629).

We had the unique opportunity to receive matched adult human lungs, vertebral bodies (BM), and PB freshly recovered from deceased research donors (supplemental Table 1). The lungs were extensively perfused at the time of collection, and we selected healthy-appearing lung tissue for experiments (supplemental Figure 1A). After tissue dissociation and the rendering of single-cell suspensions, we characterized live, Lin⁻ cells (Figure 1A) using standard surface markers for HSPCs (supplemental Table 2).^10-12 Notably, our lineage panel contained markers for mature immune, endothelial, and epithelial cells, allowing for the phenotyping of rare Lin⁻ CD34⁺ cells. We discovered a distinct population of multipotent (MP) cells (CD34⁺/CD38⁻) in the lung and BM and very few of these cells in the PB (Figure 1B). The lung and BM contained cells with surface staining consistent with HSCs (CD34⁺/CD38⁻/CD90⁺/CD45A⁻) but these cells were absent in the PB. Multipotent progenitor (MPP) cells (CD34⁺/CD38⁻/CD90⁻/CD45A⁻) were observed in all 3 tissues. More committed hematopoeitic progenitor cells (HPCs; CD34⁺, CD38⁺), such as common myeloid progenitor (CMP; CD34⁺/CD38⁺/CD45A⁻/Flt-3⁺), granulocyte-macrophage progenitor (GMP; CD34⁺/CD38⁺/CD45A⁺/Flt-3⁺), and megakaryocyte-erythroid progenitor cells (MEP; CD34⁺/CD38⁺/CD45A⁻/Flt-3⁻) were observed in all 3 tissues but were less common in the lung (Figure 1B-C). Overall, the BM and PB had nearly identical proportions of hematopoietic progenitors, whereas Lin⁻ lung cells were enriched for immunophenotypic HSCs and MPPs (Figure 1C-D).

To rule out residual blood as a source of HSPCs in perfused lungs, we estimated the numbers of progenitors in equivalent volumes of blood and lung tissue, yielding lung HSPC numbers that could not be explained by retention of intravascular blood in lungs (supplemental Figure 1B). Furthermore, the distinct proportions of cell subsets in the lung and blood indicate a tissue-specific progenitor composition. Remarkably, the frequency of the immunophenotypic MP pool in the lung is similar to the BM (Figure 1E), whereas the pool of more committed hematopoietic progenitor cells (HPCs) is much smaller in the lung (Figure 1E). We also tested for the effects of donor age and sex on our results and found that increasing age was associated with fewer numbers of HSPCs in the BM but not in the lung (supplemental Figure 1C). Sex was not associated with changes in HSPC frequency in the BM or lung (supplemental Figure 1D).

Given that fibroblasts are generally Lin⁻ cells, but some can be CD34 positive,¹³ we included a marker for fibroblasts (platelet-derived growth factor receptor α [PDGFRA]) in our lineage panel to determine whether fibroblast contamination could affect our results, but we did not detect any changes in HSPC frequencies between our 2 lineage panels (supplemental Figure 2A-C).

We next tested the functional capacity of lung HSPCs using in vitro colony-forming assays. We plated Lin⁻ cells from the lung or BM in MethoCult and observed a variety of colonies from both tissues. The Lin⁻ cells from the lung produced overall fewer colonies, but with a significant increase in the relative proportion of erythroid colonies (burst-forming unit erythroid [BFU-E]; Figure 1F; supplemental Figure 3A). After morphologic colony assessment, we used flow cytometry (supplemental Figure 3B-D) to confirm cellular colony composition and found that lung colonies were enriched with cells expressing the erythrocyte marker GlyA (supplemental Figure 3D). There were not enough Lin⁻ cells from the PB to perform a comparative study, so we determined the colony-forming unit (CFU) potential of nucleated blood cells. This yielded few colonies (supplemental Figure 3E-G), including only a few BFU-E colonies, which is the predominant colony type produced by the lung. These results strengthen our conclusion that the presence of lung HSPCs is not an artifact of blood contamination. We also plated Lin⁻ cells in MegaCult to test the potential to produce megakaryocytes, a cell population that we previously described as resident cells in the mouse lung.⁷ Both BM and lung cells were capable of producing megakaryocyte colonies although the lung produced fewer and small colonies (Figure 1F). In addition, we sorted lung and BM HSCs and HPCs into MegaCult and observed that lung HSCs produced comparatively more megakaryocytes (supplemental Figure 4A), whereas in BM most cells with CFU megakaryocyte capacity are derived from precursors in the HPC fraction (supplemental Figure 4B).

Overall, the fewer colonies observed from lung cells may also relate to reduced cell cycling compared with the BM (Figure 1H). The molecular cues that regulate the transition to activity in pulmonary HSPCs are not known and could be different from growth factors added in classical in vitro CFU assays. Together, we conclude that the human lung contains functional HSPCs that exhibit a bias toward erythroid and megakaryocyte lineages within the hematopoietic progenitor tree.

We next tested whether HSPCs isolated from the lung are capable of engraftment when xenotransplanted into immunodeficient mice.⁸ We chose NSG-SGM3 mice (human stem cell factor, granulocyte-macrophage colony-stimulating factor, interleukin-3) to support human myeloid cell engraftment¹⁴ for these experiments in which we adoptively transferred magnetic-enriched Lin⁻ cells from the lung or BM into mice after sublethal irradiation.¹⁵ Given that the production of erythroid cells in xenograft models is not reliable owing to the lack of cross-reactivity between mouse erythropoietin (EPO) and the human EPO receptor, mice received recombinant human EPO injections in the final 3 weeks of the experiment (Figure 2A) as described previously.⁸ Human cells in the BM, lung, and PB of recipient mice were assessed after 10 weeks, a time point that is commonly used to measure human HSC activity in vivo.^15-17 Engraftment was rigorously defined as the presence of human CD45⁺ cells, using 2 different antibody clones, with the threshold for engraftment set to ≥0.01% CD45⁺⁺ cells of all CD45⁺ cells (mouse and human) with at least 30 cells recorded in the CD45⁺⁺ gate for BM and lung, and ≥15 cells for PB⁸ given that low levels of engrafted human cells are expected from previous studies.¹⁷ Examples of positive and negative engraftment (sublethal irradiation without adoptive cell transfer) are shown in Figure 2B. Overall, 6 of 7 mice with BM-derived HSPCs and 5 of 7 mice with lung-derived HSPCs engrafted in the BM and in the lung tissue (Figure 2C; supplemental Figure 5A). In the PB, engraftment was observed in 4 of 7 mice with BM-derived HSPCs and 2 of 7 mice with lung-derived HSPCs (Figure 2C; supplemental Figure 5A). Examples of BM- or lung-derived cells in the mouse BM or lung are shown in Figure 2D.

We also determined erythroid engraftment using the gating strategy in Figure 2E to detect CD45⁻, human GlyA⁺, and human CD71⁺ cells and observed similar erythroid engraftment of BM- or lung-derived HSPCs in the mouse BM, lung, or PB (Figure 2F-G; supplemental Figure 5B). Similar to BM-derived HSPCs, lung-derived HSPCs were capable of multilineage blood cell production (Figure 2H; supplemental Figure 5C-E).

The availability of matched lung and medullary HSPCs allowed us to directly compare their transcriptional profiles using scRNA-seq (supplemental Figure 6A). Our gating strategy is shown in supplemental Figure 6B-C. We integrated all samples using Harmony¹⁸ and generated a batch-corrected uniform manifold approximation and projection (UMAP) to identify clusters of transcriptionally similar cells. We filtered out stromal and mesothelial cell populations and subsetted progenitor cells (supplemental Figure 6C-E). Using the function “findConservedMarkers,” we identified genes that were consistently expressed across BM- and lung-derived cells and annotated clusters based on a reference data set.^19,20 Dimensionality reduction yielded a visual representation consistent with HSC and MPP production of progenies with progressive commitment to more differentiated fates (Figure 3A; supplemental Figure 6F-G). To validate our annotation, we generated coregulated modules of differentially expressed genes using Monocle3.²¹ The module highly specific for HSCs contains genes such as AVP, SPINK2, SELL, and HOPX and was strongly expressed in cells from both the BM and lung (Figure 3B; supplemental Figure 7A-B). We ordered the cells along a pseudotime trajectory to reconstruct their developmental path for each tissue individually, suggesting a relationship between HSCs and stromal cells in the lung that was absent in cells derived from the BM (Figure 3C). As previously discussed, some pulmonary fibroblasts are CD34⁺, and these results could point to the ontogeny of a subset of lung stromal cells.

Next, we compared the differential gene expression between medullary and lung cells within the HSC/MPP cluster and plotted the median expression of both cells on a scatterplot (Figure 3D). Using the Wilcoxon rank-sum test in Seurat’s “FindMarkers,” we identified 50 genes that were upregulated in lung HSCs and 10 genes that were higher in BM cells (Figure 3D). Among the top upregulated genes in lung HSCs, several genes (CEBPB, SOD2, PLCG2, HSPA1A) were associated with maintaining HSC quiescence and fitness (Figure 3E).^22-25 We attributed the highest biological relevance to genes that were upregulated in most of the cells, as validated by analyzing the distribution of gene expression values (Figure 3F-H). HSCs from the lung have unique features (Figure 3F) and share characteristics of the hematopoietic lineage (Figure 3G), whereas, as expected, cells from the BM have a high expression of classical stem cell genes (Figure 3H).¹⁹ 

Next, we performed single-sample gene set enrichment analysis to identify pathways that are differentially regulated between cells from the lung and the BM. The enrichment scores were calculated across all individual cells and tissues for gene set collections from the Molecular Signature Database (H, hallmark; CP, canonical pathways; C5, ontology). In the gene sets analyzed, repeatedly pathways associated with EPO signaling, platelet function, and immune responses were found to be upregulated in pulmonary HSCs (Figure 3I). Side-by-side comparison of selected pathways indicates that HSCs from the lung are enriched for megakaryocyte (R-HSA-8936459) and EPO-induced erythroblast (R-HSA-9027277, R-HSA-9006335) differentiation, as indicated by higher normalized enrichment scores (Figure 3J). Our finding of increased EPO signaling pathways in lung HSCs is consistent with our data in Figure 1E on the erythroid-biasing of lung colonies. Our finding of platelet and megakaryocyte-skewing of lung HSCs is provocative in light of previous work on platelet production in the lung and tissue-resident immune-like megakaryocytes.^7,26,27 In addition, we found inflammatory signaling to be upregulated in pulmonary HSCs (Figure 3K), suggesting that these cells could impart unique immunological functions of their progeny.

A small cluster of cells has recently been suggested as an HSC population in an analysis of the healthy and diseased human lung tissue based on CD34, SPINK2, STMN, and PRSS57 expression,²⁸ but the self-renewal and differentiation potential and the location of these cells are not known. In addition, CD34⁺ cells can be found in endothelial, lymphatic, and fibroblast clusters (supplemental Figure 8A-B). To identify potential HSCs based on their transcriptomic profile, we used UCell scoring to find Lin⁻ CD34⁺ cells with an HSC signature in a data set combining 9 human lung scRNA-seq studies (Human Lung Cell Atlas V2²⁹; supplemental Figure 8C-D). We projected these cells on our UMAP structure of HSPCs from the lung and BM (Figure 3A) in supplemental Figure 8E and mapped 43 cells to the HSC/MPP cluster (supplemental Figure 8F) across all integrated data sets in the Human Lung Cell Atlas (supplemental Figure 8G). We conclude that, given their rarity and the coexpression of CD34 in multiple lung cell types, lung HSCs are mostly masked and overlooked when using standard unsupervised clustering techniques.

The lung is composed of diverse cell entities, such as epithelial, endothelial, stromal, and immune cell subpopulations that could provide a unique niche for HSPC maintenance and differentiation.²⁰ We used immunostaining to localize putative HSCs (Lin⁻/CD34⁺/CD90⁺) in human lung and BM and found HSCs that localized to the alveolar interstitium (Figure 4A; supplemental Figure 9A-B). To characterize the pulmonary HSC niche, we used a spatial transcriptomics approach based on combinatorial single-molecule fluorescence in situ hybridization (Figure 4B). We designed a marker panel that characterizes HSPCs and the common cell entities of the lung (supplemental Table 3). After QuPath-based cell segmentation, we annotated the lung cells by performing unsupervised clustering on transcript expression values (Figure 4C; workflow shown in supplemental Figure 10A-D). We filtered for putative HSPCs defined by CD34⁺ positivity, expression of HSPC-associated transcripts, and negativity for marker genes of other lung cell entities (Figure 4C; workflow shown in supplemental Figure 10E-F; supplemental Figure 11A-B). We visually validated all candidate cells and assigned them to their anatomic location. More than 90% of cells matching the criteria for HSPCs localized to the extravascular lung, with most cells in the alveolar interstitium or in proximity of bronchi (peribronchial) and vasculature (perivascular) (Figure 4D; supplemental Figure 11A). To define the neighborhood of HSPCs on the cellular level, we used Squidpy for co-occurrence analysis across all samples³⁰ suggesting that the immediate HSPC niche is mainly formed by endothelial cells, although epithelium and fibroblasts have a steady presence (Figure 4E-F; supplemental Figure 11C).

Finally, we evaluated the potential function of pulmonary HSCs by evaluating their contribution to the mobilized blood progenitor pool. In 8 healthy volunteers undergoing mobilized blood donations for stem cell transplantation (apheresis donations), we analyzed the transcriptomic profile of Live/Lin⁻/CD34⁺ cells by scRNA-seq (Figure 5A-B). After cell annotation (Figure 5C), we subsetted the HSCs per donor (Figure 5D) and calculated tissue-specific signature scores using UCell to infer their source (supplemental Table 4). We determined that ∼25% of mobilized HSCs were of extramedullary origin including ∼15% of total cells being of pulmonary origin (Figure 5E-G). These results indicate that extramedullary HSCs are mobilized into the HSC pool used for transplantation and points to their potential biological function.

For many years, HSCs were viewed as unbiased cells that initiated hematopoietic development and specialization. Enabled by new technologies, our understanding of hematopoiesis has been refined to include the possibility that developmental biases may be present even in these most undifferentiated cells, such as with megakaryocyte-biased HSCs.¹¹ However, the mechanisms responsible for this early biasing or specialization are not clear.^31,32 Here, we propose that the traditional view of HSC residency in the BM should be reconsidered to include extramedullary tissues, such as the lung. Indeed, and remarkably, we found an equal frequency of multipotent cells residing in the BM and lung. Owing to the low frequency, we used approaches to enrich for the presence of HSPCs in our studies, which were not done in previous studies and likely enabled the profiling of this rare subset of cells among the >30 different lung cell types.

It is clear from our studies that the lung HSPCs have unique features compared with their medullary counterparts, but also to HSPCs found at other extramedullary sites (PB and spleen³³). Chief among these, and perhaps obvious from our understanding of lung biology, is that lung HSPCs, like spleen HSPCs, are less active in terms of cell division and the production of more differentiated and specialized hematopoietic cells than their BM counterparts.³³ These results from in vitro experiments would suggest that lung HSPCs function as a reserve pool that could be mobilized in the setting of hematopoietic stress or, as we have shown here, during stem cell collections for transplantation. Further studies are needed to define physiological and pathological stimuli that trigger hematopoiesis in the lung.

Our xenotransplantation experiments indicated that lung HSPCs performed similarly to medullary HSPCs, when given the challenge of engraftment after sublethal irradiation. A wide range of engraftment levels and analysis time points have been used for measuring human HSC activity in vivo, and there is no standard in the field.¹⁷ Thus, despite low graft expansion in our model (NSG-SGM3 mice, sublethal irradiation, IV HSPCs), our data are consistent with previous outcomes and demonstrate self-renewal capacity of pulmonary HSPCs in vivo. In addition, we included donors of any age and sex (supplemental Table 1), which might have contributed to larger variations in our results.

Our transcriptomics analysis revealed a distinct molecular program in lung HSPCs that was further distinguishing compared with medullary HSPCs. We found clear megakaryocyte biasing of lung vs medullary HSPCs indicating that perhaps the lung is a source of these biased progenitors. Our colony analysis of lung HSCs also supports megakaryocyte biasing. This is an intriguing finding given (1) the role of the lung in platelet biogenesis⁷ and (2) the presence of tissue-resident immune-like megakaryocytes in the lung that may be derived from a lung progenitor.^26,27 We further found an erythroid bias of lung HSPCs, akin to that reported in PB and spleen HSPCs.³³ Therefore, erythroid bias is a shared feature of steady-state extramedullary HSPCs, perhaps determined by distinct access to environmental oxygen vs the relatively hypoxic environment of the BM.³⁴ 

Our immunofluorescence imaging and spatial transcriptomic studies were essential in confirming the presence of HSPCs in the lung and defining their precise locations and niche. Given the prevailing dogma that hematopoietic precursors widely circulate,^35-37 it was important to rule out that blood contamination was producing our results. We found a few intravascular HSPCs, confirming previous studies, but most were extravascular and predominately in vascular-rich zones of the lung alveoli. This anatomic location in the lung is similar to the location of HSPCs in the BM, which are closely positioned next to the vascular sinusoids.^38,39 In the lung, this positioning could be important for seeding of the lung with circulating HSCs and potentially for exiting the lung during hematopoietic stress or during mobilization for stem cell collections. In this niche, it was notable that lung fibroblasts were in close proximity. The lung mesenchyme is well known to be a critical niche supporting epithelial cell development and repair and similar mechanisms could be operable influencing lung HSPCs.^40,41 In addition, we identified a developmental trajectory between lung HSPCs and lung stromal cells. Subpopulations of lung fibroblasts are known to be CD34⁺, and previous lineage-tracing experiments in pulmonary fibrosis have shown a hematopoietic contribution to fibroblasts.^42,43 Future studies are needed in this area to understand how lung HSPCs could be involved in fibrotic lung diseases.

Our study does have limitations. We used human deceased donors for tissue collection, and there may be biases introduced in the mechanisms of their deaths and how that influences HSC biology. There is also donor heterogeneity, variability in ischemia times during organ recovery, and differences in cell preparation methods for lung and BM affecting cell viability that could influence our results. Our studies rely heavily on transcriptomics and future studies will need to identify a lung HSC-specific marker to facilitate their identification. We have not addressed the ontogeny of lung HSPCs, something that is not possible given the restraints of our human tissue study. There is irrefutable evidence that HSCs commonly enter the circulation and microcirculatory beds, and given the functional and molecular similarities of lung HSPCs to other extramedullary HSPCs, this is perhaps the source of the tissue-resident HSPCs in the lung.^36,44 However, there is precedence for tissue residency to be endowed during development, such as with yolk sac–derived macrophages.⁴⁵ Future studies will be needed to answer this question, including the possibility that hemogenic endothelium in the fetal lung could be the source.^46,47 Our scRNA-seq analysis in blood stem cell donors provides important insights into the potential contribution of pulmonary HSCs to the mobilized blood progenitor pool, but these results need further validation. It is important to recognize that, owing to the limitations of direct methodologies in humans, our approach relies on organ-specific HSC signature scores to infer the tissue of origin.

Our findings reframe our understanding of the HSPC pool and its molecular diversity and should enable future studies that could potentially lead to therapeutic advances, such as for lifesaving HSC transplantation for BM malignancies and failure. In the modern era, transplantation is mainly accomplished using mobilized HSCs obtained from the PB. We have now shown that this pool is heterogeneous with a sizable extramedullary component that implies functional heterogeneity given our findings on lung HSC biases. These results have important implications for treatment responses and complications. Our findings may also help to understand the mechanisms of leukemogenesis with the possibility that lung HSPCs are direct targets of environmental carcinogens.⁴⁸ Furthermore, our findings add to our expanding understanding of rare cell types in the lung and their potential functions.⁴⁹ 

The authors thank Donor Network West, Michael Matthay (University of California San Francisco [UCSF]), and the Nina Ireland Program for Lung Health (UCSF) for providing human lung tissue from deceased organ donors.

This work was supported by an International Anesthesia Researach Society (IARS) Mentored Research Award to C.C., the Deutsche Forschungsgemeinschaft Research Fellowships to C.C. (CO 2096/1-1) and N.M. (ME 5209/1-1), Bakar UCSF ImmunoX support to A.J.C, a Wellcome-Royal Society Sir Henry Dale Fellowship (107630/Z/15/Z) and funding in part by the Wellcome Trust [203151/Z/16/Z, 203151/A/16/Z] and the United Kingdom Research and Innovation (UKRI) Medical Research Council [MC_PC_17230] to E.L., and National Institutes of Health (NIH), National Heart, Lung, and Blood Institute grant R35HL161241 to M.R.L. Sequencing was performed at the UCSF Center for Advanced Technology, supported by the UCSF Program for Breakthrough Biomedical Research, Research Ressource Program Institutional Matching Instrumentation Award, and NIH, Office of the Director 1S10OD028511-01 grants.

Contribution: C.C. designed and conducted experiments, analyzed the data, and wrote the manuscript; M.M. designed and conducted experiments; J.T., B.S., U.V., and A.J.C. assisted in designing and conducting experiments and helped analyzing the data; H.W. and M.N. helped analyzing the data; S.J.C., L.Q., J.J.T., and M.D.G. assisted in conducting experiments; A.D.L. provided human apheresis blood samples; E.P., E.L., and N.M. assisted in designing experiments, provided technical expertise with hematopoietic progenitor analyses, and provided editorial support on the manuscript; and M.R.L. designed the experiments, supervised the study, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Mark R. Looney, Department of Medicine, University of California San Francisco, 513 Parnassus Ave, HSE 1355A, San Francisco, CA 94143-0130; email: mark.looney@ucsf.edu.