Introduction

During development, hemogenic ECs generate hematopoietic cells through a process of endothelial-to-hematopoietic transition (EHT) at geographically defined anatomical sites^1–9. At these locations, the hemogenic ECs represent a small fraction of all ECs^10–12, and their competency to produce hematopoietic stem and progenitor cells (HSPC) is temporally restricted to short developmental windows, and the hemogenic potential differs^13–18. In the dorsal aorta, the hemogenic endothelium produces hematopoietic stem cells (HSC) and other multi-potent progenitors between E10.5 and E11.5^17–19 that persist in the adult and are generally believed to sustain adult hematopoiesis throughout life.

Efforts to reliably generate ex vivo HSC from Cdh5-expressing ECs identified difficulties in reproducing the microenvironmental clues coming from the inductive niche cells^20–22 and have generally relied on transcription factor-induced reprogramming of the ECs to drive hematopoietic specification^23,24. The transcription factor RUNX1, which marks the hemogenic endothelium, can confer a hemogenic potential to embryonic ECs lacking such potential^25–28. When transcription factors RUNX1, FOSB, GFI1, and SPI1 were co-expressed in adult murine ECs co-cultured with “vascular niche ECs”, EHT was induced in vitro producing HSC with long-term self-renewal capacity²⁹. A similar approach was used with human ECs enabling hematopoietic specification³⁰.

An unresolved question is whether hemogenic ECs, thought to be mostly restricted to the early stages of mouse development^12,23, may persist in the adult mouse^2,12,31,32. Exploiting advances in cell lineage tracking and single cell analyses^33,34, we report the identification of hemogenic ECs in the adult mouse bone marrow (BM) that produce functional hematopoietic progenitors and mature blood cells.

Results

Evidence that adult bone marrow ECs generate hematopoietic cells

We assessed the hemogenic potential of ECs in adult mice using Cre-reporter-based lineage tracing. Since Cdh5, encoding vascular endothelial cadherin (VE-Cadherin), is selectively expressed by ECs, Cdh5-Cre^ERT2 recombinase activity allows tracking the hematopoietic cell output from adult hemogenic ECs^35,36. Therefore, we generated three Cdh5-based lineage-tracing models using inducible Cdh5-Cre^ERT2(PAC)^37,38 and Cdh5-Cre^ERT2(BAC)³⁹ mouse lines, in combination with the Cre-reporter lines ZsGreen and mTmG (Figure S1A). We then treated eight- to twelve-week-old mice with three doses of tamoxifen (10 mg·kg⁻¹, gavage) on consecutive days, and four weeks later we evaluated peripheral blood and BM (Figure 1A). Expectedly³⁶, most BM Endomucin⁺ ECs were ZsGreen⁺ in tamoxifen-treated Cdh5-Cre^ERT2(PAC)/ZsGreen and Cdh5-Cre^ERT2(BAC)/ZsGreen mice, and virtually no ZsGreen⁺ ECs were present in Cre negative or peanut oil-treated controls (Figures S1B and S1C). By flow cytometry, >90% CD31⁺VE-Cadherin⁺ BM ECs were tracked by tamoxifen-induced fluorescence in the three mouse lines (Figure 1B; Figure S1D). A low-level tamoxifen-independent reporter fluorescence was also detected in BM ECs, which was low in the mTmG reporter line (Figure 1B), and higher in the ZsGreen reporter lines, previously attributed to “basal” Cre^ERT2 activity⁴⁰ (Figure S1D).

Figure 1.

To evaluate the hemogenic potential of adult BM ECs, we analyzed the expression of the hematopoietic marker CD45 in Cdh5-tracked cells. Notably, tracked CD45⁺ hematopoietic cells, presumed progeny of VE-Cadherin⁺ ECs, were detected by flow cytometry in BM and blood of mice from all three Cdh5-Cre^ERT2 mouse lines (Figure 1C and D; Figures S1E and F; representative flow cytometry gating in Figure S1G and H). Moreover, confocal microscopy identified isolated ZsGreen⁺CD45⁺ cells in BM (Figure S1I) and blood of tamoxifen-induced mice (Figure 1E). Importantly, while a small fraction of fluorescent CD45⁺ cells were detected in EGFP and ZsGreen (Figures 1C and D; and Figure S1D-F) reporter mice without tamoxifen, as reported^40–43, the significant increase of tamoxifen-induced fluorescent CD45⁺ cells (Figure 1C and 1D, Figure S1D-F) indicates the occurrence of Cdh5-Cre recombination in the adult mouse, presumably tracking EHT. This tamoxifen-induced increase in CD45⁺ZsGreen⁺ peripheral blood mononuclear cells (PBMC) was also observed in individual mice (Figure S1J).

We tested the kinetics of tamoxifen-induced ZsGreen expression in BM ECs and PBMC of Cdh5-Cre^ERT2(PAC)/ZsGreen mice (Figure 1F). By flow cytometry, virtually all BM ECs were ZsGreen⁺ by day 4 after tamoxifen administration and this level persisted over 50 days. Instead, the percentage of ZsGreen⁺ PBMCs increased more gradually, plateauing by day 14 and this level persisted over 50 days of observation. This gradual increase is likely attributable to tamoxifen-induced ZsGreen expression in hemogenic ECs.

Further analysis showed that most BM EGFP⁺CD45⁺ cells in Cdh5-Cre^ERT2(PAC)/mTmG mice were CD11b⁺Ly6G⁺ granulocytes (48.85%) and CD11b⁺Ly6G⁻ monocytes (32.06%), while CD19⁺ B (2.46%) and CD3⁺ T (0.84%) lymphocytes were detected at lower frequencies (Figure 1G). All EGFP⁺ BM cell populations were significantly induced by tamoxifen administration, including rare LSK (Lin⁻Sca1⁺cKit⁺) progenitors (Figure 1H). Also, the peripheral blood of tamoxifen-induced Cdh5-Cre^ERT2(PAC)/mTmG mice contained EGFP⁺ granulocytes and monocytes, and fewer B and T lymphocytes (Figure S1K and L). Similarly, in ZsGreen-reporter mice, tracked LSK progenitors, B lymphocytes, granulocytes, and monocytes were all present in the BM and blood (Figure S2A-D).

We further characterized the tracked BM hematopoietic LSK stem/progenitors by flow cytometry (Figure S2E). The results, displayed by Uniform Manifold Approximation and Projection (UMAP) dimensional reduction, showed that the ZsGreen-tracked progenitor population includes cell subsets defined phenotypically as hematopoietic stem and progenitor cells (HSPC) (Figure 1I). Each ZsGreen⁺ progenitor population represented a similarly small proportion of the corresponding non-tracked progenitor cell population (Figure 1J). These results support the existence of EHT in the adult mouse contributing to generation of hematopoietic progenitors and mature cells.

Adult BM ECs cultured ex vivo generate transplantable HSPCs

To investigate whether adult BM ECs can generate hematopoietic cells ex vivo, we cultured BM cells isolated from tamoxifen-treated Cdh5-Cre/ZsGreen mice. Initially, BM single-cell (sc) suspensions were cultured under two conditions (Figure 2A): (1) on “Primaria” pro-adhesive flasks, and (2) on OP9 stromal cell⁴⁴ monolayers grown on gelatin-coated conventional tissue culture flasks. To attempt recreating BM niches, fresh wild-type (WT) BM cells were added to the ZsGreen-tracked BM cell cultures twice/week throughout the culture period (see Methods for details). By week 8, BM cells cultured on “Primaria” dishes formed a confluent monolayer of ZsGreen-tracked cells with a typical EC morphology, whereas BM cells cultured on OP9 monolayers exhibited a fibroblast-like morphology with the ZsGreen⁺ cells clustering in foci, without forming a monolayer (Figure 2A). We visualized some ZsGreen-tracked (ZsGreen⁺) cells in cultures of BM cells grown onto OP9 monolayers supplemented with fresh BM cells, but not in cultures of BM cells grown onto “Primaria” surfaces supplemented with fresh BM cells (Figure S2F). This hinted that the OP9 and BM cell culture system may allow hematopoietic cells emergence from Cdh5-Cre/ZsGreen-tracked EC ex vivo.

Figure 2.

To test this possibility, we generated a pool of BM cells (from 20 adult mice; age 10-18 weeks) and used the pool to derive 3 populations: sorted BM CD45⁻VE-Cadherin⁺ZsGreen⁺ ECs; sorted CD45⁺ZsGreen⁺ hematopoietic cells; and unsorted BM cell populations (Figure 2B). We then cultured these 3 cell populations under the same conditions (OP9 monolater and supplementation with fresh BM cells twice/week). After 8-week culture, flow cytometry analysis detected CD45⁺ZsGreen⁺ cells from the unsorted BM and the sorted CD45⁻VE-Cadherin⁺ZsGreen⁺ cell cultures, but the CD45⁺ZsGreen⁺ were virtually absent from the CD45⁺ZsGreen⁺ cell cultures (Figure 2C and 2D). These results show that CD45⁻VE-Cadherin⁺ZsGreen⁺ ECs can generate CD45⁺ZsGreen⁺ ex vivo. Notably, most (>80%) CD45⁻ZsGreen⁺ cells retained expression of VE-Cadherin and Endomucin, thereby confirming their endothelial identity (Figure 2C, most right panel). Importantly, the virtual absence of CD45⁺ZsGreen⁺ cells in 8-week cultures of sorted CD45⁺ZsGreen⁺ cells shows that pre-existing CD45⁺ZsGreen⁺ hematopoietic cells, derived from tamoxifen-dependent and independent processes, are effectively removed during the extended culture. This further suggests that the CD45⁺ZsGreen⁺ hematopoietic cells detected in the CD45⁻VE-Cadherin⁺ZsGreen⁺ EC cultures likely originated from ECs during the final week of culture.

Next, we tested whether the ex vivo-derived CD45⁺ZsGreen⁺ cells are functional in lethally irradiated recipients. To this end, we obtained unfractionated BM cells from 50 tamoxifen-treated Cdh5-Cre/ZsGreen mice, cultured the cells onto OP9 monolayers with fresh BM supplementation, sorted the ZsGreen⁺ cells (96% of which were CD45⁺ by flow cytometry) at the end of 8-week culture, and transplanted these cells into 8 lethally irradiated WT recipients; 5×10³ cells; 2.5×10³ cells; 1.25×10³ cells; 6.25×10² cells, n=2/group (Figure 2E-G). Two mice (recipients of 6.25×10² and 1.25×10³ cells) died on days 4 and 6, but the remaining six mice survived and remain well at the time of manuscript submission (12 weeks post-transplant). Ten weeks post-transplant, peripheral blood WBC counts were within normal range in all surviving transplant recipients, indicative of hematopoietic reconstitution (Figure 2H). Flow cytometry revealed that 74.85% of PBMC were ZsGreen⁺ (Figure 2I). Most myeloid cells (granulocytes and monocytes, 91.97%) and B cells (93.78%) were ZsGreen⁺, whereas only 30.63% of T cells were ZsGreen⁺, likely attributable to long-lived ZsGreen⁻ T cells.

These results confirm that BM ECs propagated ex vivo onto OP9 monolayers with BM cell supplementation produce hematopoietic cells and show that this output includes HSPC capable of engrafting and generating hematopoietic cell progeny of myeloid and lymphoid lineages in lethally irradiated recipients.

Adult BM ECs can give rise to hematopoietic cells after transfer to conditioned recipients

Next, we examined whether adult BM ECs are hemogenic in transplant recipients. To this end, we FACS-sorted CD45⁻VE-Cadherin⁺ZsGreen⁺ ECs from the BM of tamoxifen-pretreated (4 weeks prior to BM harvest) Cdh5-Cre^ERT2(PAC)/ZsGreen mice (Figure 3A; Figure S2G and S2H) and transplanted these cells into adult WT C57Bl/6 mice untreated (PBS) or conditioned by fluorouracil (5-FU treated) prior to transplant⁴⁵. The choice of 5-FU conditioning rather than lethal irradiation of adult mice was driven by previous experiments showing the difficulties at reconstituting lethally myelo-ablated adult recipients with hemogenic yolk sac cells, which reconstituted conditioned newborns⁴⁶. Four weeks after transplant, we observed a significant increase in the proportion of ZsGreen⁺ ECs in the BM and ZsGreen⁺CD45⁺hematopoietic cells in the BM and peripheral blood of the transplant recipients conditioned by 5-FU but not controls (PBS) (Figure 3B and 3C; Figure S2I). These CD45⁺ZsGreen-tracked cells in BM and blood included granulocytes, monocytes, and lymphocytes (Figure 3D). Thus, BM-derived CD45⁻VE-Cadherin⁺ZsGreen⁺ cells, when transferred into 5-FU-conditioned recipients, gave rise to detectable CD45⁺ZsGreen⁺ hematopoietic cells.

Figure 3.

We further examined the effect of mouse age on the endothelial hemogenic potential by treating the mice with tamoxifen between week 6 and 32 of age. Four weeks later (week 10 to 36 of age), we measured the percentage of ZsGreen-tracked CD45⁺ hematopoietic cells in the BM. We observed that tamoxifen inductions beyond week 10 of age resulted in a progressive decrease of the CD45⁺ hematopoietic cell output, and detected an inverse correlation (Pearson’s r -0.63, P<0.0001) between age and EC hemogenic potential (Figure 3E). This progressive decline of CD45⁺ cell output was not coupled with a loss of BM EC fluorescence, since virtually all BM ECs were ZsGreen⁺ throughout the duration of the experiment (Figure 3F), consistent with the stability of Cre-mediated labeling of Cdh5-expressing cells ³⁶. These observations suggest an age-related loss of hemogenic capability of BM ECs.

Additionally, we examined whether tracked hematopoietic cells from adult BM EC are functional. Since lymphocyte trafficking from the peripheral blood to the peritoneal cavity is critical for their function at this site^47,48, we first evaluated the spontaneous migration of tracked CD45⁺ hematopoietic cells to the peritoneal cavity. Compared to no-tamoxifen controls, tamoxifen-treated Cdh5-Cre^ERT2(PAC)/ZsGreen mice displayed a significant increase of CD45⁺ZsGreen⁺ monocytes, macrophages, and B1, B2, and T lymphocytes in the peritoneal cavity (Figure S2J and S2K).

After inducing peritonitis with thioglycolate (TGL, 4 hours) in tamoxifen-treated Cdh5-Cre^ERT2(PAC)/ZsGreen mice, the overall number of peritoneal leukocytes increased substantially compared to untreated (PBS) controls (Figure 3G), mostly attributable to neutrophils (Figure 3H). In addition, the peritoneal ZsGreen⁺ and ZsGreen⁻ cell populations exhibited a similar cell type distribution in TGL-treated mice (Figure 3I). We further examined Escherichia coli (K-12 strain) phagocytosis and reactive oxygen species (ROS) production in peritoneal cell exudates in response to TGL (Figure 3J-3O). Both ZsGreen⁺ and ZsGreen⁻ peritoneal neutrophils and macrophages comparably phagocytosed Escherichia coli and generated ROS (Figure 3J, 3K, 3M, and 3N), except that the phagocytic and ROS production of ZsGreen⁺ macrophages was somewhat higher than that of ZsGreen⁻ macrophages (Figure 3L and 3O).

Thus, Cdh5-tracked mature neutrophils and macrophages are functional at trafficking and homing, and presumably capable of contributing to the host response to tissue inflammation.

Adult EHT is independent of preexisting hematopoietic cell progenitors

Faithful Cre-reporter lineage tracing requires that Cre recombinase activity be restricted to the intended cell type⁴⁰. In our system, Cdh5-Cre^ERT2 is expected to drive recombination specifically in ECs, as Cdh5 is an established EC-specific marker. However, flow cytometry has occasionally revealed rare VE-Cadherin⁺CD45⁺ in mouse BM, exemplified in Figure S2G, potentially reflecting double-positive cells. To address the possibility that cells co-expressing VE-Cadherin/Cdh5 and CD45/Ptprc exist in the mouse BM, we analyzed publicly available sc-RNAseq data from adult mouse BM⁴⁹. This analysis showed that only a small subset of plasmacytoid dendritic cells (pDCs) co-express VE-Cadherin and CD45, but not HSPC or other mature blood cells (Figure S3A-S3D). pDCs are terminally differentiated cells and unlikely progenitors of tracked CD45⁺ multilineage progeny in our Cdh5-cre mice. Nonetheless, it is plausible that other, currently unidentified, hematopoietic cells may also co-express Cdh5/VE-Cadherin and Ptprc/CD45. We additionally considered the possibility that CD45⁺ progenitors with functional Cre^ERT2 activity may contaminate the sorted populations of Cdh5⁺CD45⁻ BM cells and could become fluorescent upon tamoxifen administration.

To address both possibilities, we performed transplant experiments with two populations of Lin⁻Sca1⁺cKit⁺ (LSK) progenitors from tamoxifen untreated Cdh5-Cre^ERT2(PAC)/ZsGreen mice: (1) a ZsGreen⁻ LSK population (>99% purity; Figure S3E-S3G) to assess whether ZsGreen⁻ hematopoietic progenitors with a functional Cdh5-Cre^ERT2 activity could become fluorescent after tamoxifen, and (2) an LSK population enriched for ZsGreen⁺ cells (45.9% purity, with the remaining 54.1% comprising ZsGreen⁻ LSK cells (Figure S3F)), to assess the contribution of pre-tracked, tamoxifen-independent, hematopoietic progenitors. We then transplanted these two LSK populations into lethally irradiated (11 Gy) WT C57Bl/6 recipients (Figure 4A). Four weeks post-transplant, we administered tamoxifen and monitored the peripheral blood for the presence of ZsGreen⁺ hematopoietic cells over six months.

Figure 4.

Expectedly, all LSK recipients (ZsGreen⁻ LSKs or LSK enriched with ZsGreen⁺ cells) showed successful hematopoietic reconstitution as evidenced by normal blood WBC counts by 10 weeks post-transplant (Figure 4B). Importantly, the mice transplanted with ZsGreen⁻ LSKs did not produce ZsGreen⁺ PBMCs post-tamoxifen administration, indicating that the transplanted LSKs and their progeny did not express tamoxifen inducible Cdh5-Cre^ERT2 recombinase activity (Figure 4C and 4D). Instead, the mice transplanted with ZsGreen⁺-enriched LSKs displayed a similar proportion of ZsGreen⁺ PBMCs prior to and after tamoxifen administration (Figure 4D), indicating that the ZsGreen⁺ LSK has not expanded after tamoxifen administration. Collectively, these results demonstrate that LSK progenitors in adult BM lack of tamoxifen-inducible Cdh5 expression and do not contribute to tamoxifen-induced adult EHT in our Cdh5-reporter mice. Rather, these results strongly support the conclusion that Cdh5⁺CD45⁻ ECs are a source of hematopoietic cells in the adult mouse.

In additional experiments, we took advantage of the relative insensitivity of BM ECs subsets to irradiation relative to BM hematopoietic cells⁵⁰ to examine the possibility that EC surviving after lethal irradiation may be hemogenic. As a lethal dose of irradiation effectively eliminates HSPCs and requires hematopoietic reconstitution for survival, we transplanted WT (untracked) BM cells into lethally irradiated Cdh5-Cre/mTmG mice and treated the mice with tamoxifen 4 weeks after transplantation (Figure 4E). Prior to tamoxifen administration, >99% of PBMCs were not fluorescent, indicating that these cells derived from the transplanted WT BM rather than host-derived (Figure 4F). After tamoxifen treatment, a progressive increase in EGFP⁺ PBMCs was observed, reaching ∼0.55% by six weeks, which included myeloid cells and B and T lymphocytes (Figure 4F and 4G). Although we cannot exclude the possibility that rare EGFP⁺ hematopoietic progenitor (tracked tamoxifen-dependently or independently) may have survived the irradiation, the presence of EGFP⁺ hematopoietic cells in the circulation of lethally irradiated Cdh5-Cre/mTmG mice suggests their derivation from radioresistant Cdh5⁺ ECs rather than from radiosensitive hematopoietic progenitors. These results further support the view that adult ECs possess hemogenic potential and can produce hematopoietic cells in vivo.

Single cell tracking confirms the presence hemogenic ECs in adult BM

To directly trace hematopoietic cell progeny arising from individual adult ECs, we exploited the PolyloxExpress sc genetic barcoding system. We generated Cdh5-Cre^ERT2/ZsGreen/PolyloxExpress mice, in which both the ZsGreen and Polylox transgenes are inserted in the Rosa26 locus, such that individual ZsGreen⁺ cells contain a single Polylox barcode^33,51,52. To evaluate EC-derived hematopoietic cell output, we harvested BM from tamoxifen-treated Cdh5-Cre^ERT2/ZsGreen/PolyloxExpress mice (n=3), sorted the ZsGreen⁺VE-Cadherin⁺Endomucin⁺ (purity >99 %) and the ZsGreen⁺VE-Cadherin⁻Endomucin⁻CD45⁺ hematopoietic cells (purity >99%), mixed these populations (1:1 ratio; total 147,446 cells) and processed for 10x Illumina Sequencing and PacBio sequencing (detailed in Methods). We recovered sc transcriptome from 93,553 cells; of these, 4,072 cells had a barcode (Figure 5A, Figure S5A, detailed in Methods).

Figure 5.

Unsupervised clustering of sc transcriptome data revealed 34 clusters, 31 of which remained after doublet removal (Figure 5B; Figure S5A-S5D). Cell clusters 0, 1, 22, and 13 were annotated as “Endothelial cells” based on expression of the classical EC markers Cdh5, Pecam1, Kdr and Flt1, and absence of Ptprc/CD45, Runx1, and the mesenchymal cell markers Cxcl12, Lepr, Pdgfrb, and Col1a2 expression (Figure 5B; Figure S4E). Cell cluster 14 was annotated as “Mesenchymal type” based on co-expression of Cxcl12, Lepr, Pdgfrb, Col1a2, but expressed Runx1 and the EC markers Cdh5 and Pecam1 (Figure 5B; Figure S5E). The remaining cell clusters included the hematopoietic progenitors and mature blood cells of various lineages (Figure 5B; Figure S4E).

Barcode analysis revealed robust Polylox barcode diversity among cell populations, including 296 “true” barcodes, defined as barcodes with low generation probability (P _gen <1×10^{− 4})⁵¹ consistent with rare, unique recombination events (Figure S5A and S5B). Expectedly, “true” barcodes linked HSPC to downstream progenitors and mature hematopoietic cells, validating the system (Figure 5C).

Notably, 83 ECs (from “endothelial” cell clusters 0, 1, 22, and 13) were marked by “true” barcodes, 31 of which linked EC to hematopoietic cells based on their shared barcode (Figure 5D). These hematopoietic cells linked to ECs by shared barcodes included HSPC, EPC, GMP, and mature blood cells, encompassing granulocytes, monocytes, dendritic cells, B and T lymphocytes, plasma cells, and pDCs (Figure 5D). These results provide direct evidence, at sc resolution, that adult mouse BM ECs can generate hematopoietic progenitors and mature blood cells.

Additionally, 30 “Mesenchymal type” cells (from cluster 14; co-expressing mesenchymal cell markers, Runx1, and the EC markers Cdh5 and Pecam1) were also marked by 13 “true” barcodes, 9 of which were shared with hematopoietic cell progenitors (HSPC and Erythroid) and mature blood cells (Figure 5E and F). Also, 6 “true” barcodes linked “Mesenchymal type” cells to ECs (from clusters 0, 1, 22 and 13), suggesting either a shared precursor or derivation from each other. Although it cannot be excluded that barcoding missed identification of mesenchymal cell links to other ECs or cells, these results raise the possibility that certain BM “Mesenchymal type” cells may produce hematopoietic cell progeny. Despite similarities of sc tracing results linking ECs and “Mesenchymal type” cells to hematopoietic cells, these two cell populations display a distinctive transcriptome profile (Figure 5G; Figure S4E).

Additional analysis of all tracked cells showed that several ECs and to a lower extent “Mesenchymal type cells” shared “true” barcodes (Figure 5D-5F), indicating clonal expansion. Transcriptome-based sc cell cycle analysis, confirmed the presence of ECs and “Mesenchymal type cells” in the S and G2/M phases, albeit to a much lower degree than HSPC (Figure S5C). Together, these results demonstrate at a sc level that adult BM ECs can generate hematopoietic cell progeny of HSPC and mature blood cells. The results further raise the possibility that “Mesenchymal type” cells, marked by a hybrid endothelial and stromal phenotype, may represent an additional source of hemogenic activity in adult BM.

Single cell transcriptome identifies a Cdh5⁺Col1a2⁺ Runx1⁺cell population in the adult BM

To further characterize adult hemogenic cell populations, we analyzed publicly available scRNAseq datasets comprising BM cells from adult mice (1-16 months of age)^53–59 and embryonic caudal artery ECs (9.5–11.5 days post coitum)²⁴. After quality control and dimensional reduction, the remaining 434,810 cells clustered into 71 distinct populations (Figure 6A-D). Among the Cdh5-expressing endothelial clusters, two clusters, cluster 8 composed predominantly of embryonic cells (98%) and cluster 50 composed largely of adult BM-derived cells (95.5%) (Figure 6E), were notable in comprising cells co-expressing Cdh5 and Runx1, a transcription factor that marks the hemogenic EC identity during development⁶⁰ (Figure 6B).

Figure 6.

We jointly analyzed the transcriptome from the publicly available embryonic cluster 8 and adult cluster 50 (Figure 6 A-C), and from our sc Polylox dataset, including adult clusters 14 (Mesenchymal-type) and adult EC clusters 0, 1, 13 and 22 (Figure 5B). All populations expressed canonical EC markers (Cdh5, Pecam1, Kdr, Flt1), though at different levels, but expression of Runx1 was mostly confined to clusters 8, 50 and Polylox 14 (Figure 6F). Interestingly, adult clusters 50 and Polylox 14 co-expressed the mesenchymal cell-associated genes Col1a2, Lepr, Cxcl12, and Pdgfrb, distinguishing these two clusters from the embryonic cluster 8 and adult Polylox clusters 0, 1, 13 and 22 (Figure 6F). Additionally, embryonic cluster 8 exhibited higher expression of key EHT-related transcription factors (FLI1, LMO2, TAL1, ERG)^61–64, some of which were variably expressed by the Polylox clusters 0,1,13 and 22 (Figure 6F).

To explore whether cells expressing Cdh5, Col1a2 and Runx1 are unique to BM (Figure 6G), we looked at other adult mouse tissues. Analysis of public scRNAseq datasets from 11 adult mouse tissues⁶⁵ showed that Cdh5, Runx1, and Col1a2 expressing ECs are largely restricted to the BM, with only 3 such cells detected out of ∼32,000 EC from other tissues (Figure 6H and I)⁶⁵. These observations, together with the Polylox sc tracing experiments, indicate that adult mouse BM harbors a cell population with mixed endothelial and mesenchymal phenotype marked by Runx1, Col1a2 and Cdh5 expression and raises the possibility that this population has hemogenic potential.

Col1a2 and Runx1 expression in BM ECs

To evaluate a potential contribution of Col1a2 expression to EHT in adult BM, we generated the mouse lines Col1a2-Cre^ERT2/mTmG and Col1a2-Cre^ERT2/ZsGreen to track cells expressing the Col1a2 gene (Figure S6A). Four weeks after tamoxifen administration, the BM of Col1a2-Cre^ERT2/ZsGreen mice contained abundant ZsGreen⁺ cells (Figure S6B). By flow cytometry, a subset of these BM Col1a2-tracked ZsGreen⁺ cells were VE-Cadherin⁺CD45⁻ and RUNX1⁺ consistent with an endothelial identity and perhaps hemogenic potential (Figure S6C). Noteworthy, a similar subset of VE-Cadherin⁺CD45⁻RUNX1⁺ cells was detected in BM of adult WT C57Bl/6 mice (Figure S6D) and in BM of Cdh5-Cre^ERT2(PAC)/ZsGreen mice after tamoxifen treatment (Figure S6E).

To evaluate hemogenic potential, we looked for Col1a2-tracked CD45⁺ hematopoietic cells in Col1a2-Cre^ERT2/mTmG and Col1a2-Cre^ERT2/ZsGreen mouse lines after tamoxifen treatment. In both these mouse lines, we detected CD45⁺ hematopoietic cells tracked by EGFP or ZsGreen fluorescence in BM and blood, which were rare but significantly more abundant than in control mice not treated with tamoxifen (Figure 7A and B; Figure S6F-H). These results indicated that a proportion of the Col1a2-tracked cells in the adult mouse are hemogenic.

Figure 7.

To further evaluate this possibility, we first sorted VE-Cadherin⁺CD45⁻ ZsGreen (Col1a2)⁺ cells from the BM of tamoxifen-induced adult Col1a2-Cre^ERT2/ZsGreen mice (Figure S7A), examined expression of selected genes, and used these cells in transplant experiments. The sorted VE-Cadherin⁺CD45⁻ ZsGreen (Col1a2)⁺ cells expressed Cdh5, Col1a2, Cxcl12, and Runx1 mRNAs distinctively from other BM cell population (Figure S7B), but resembled the BM hemogenic population annotated as “mesenchymal-type” (cluster 14) identified by Polylox s.c sequencing (Figure 5G; Figure S4E) and subsets of adult BM Cdh5⁺ cells identified in public adult datasets (Figure 6E and F).

We transplanted the BM VE-Cadherin⁺CD45⁻ ZsGreen (Col1a2)⁺ cells (1×10⁴ cells/mouse) into 5-FU-conditioned adult WT C57Bl/6 recipients and looked for tracked CD45⁺ cells in the BM and blood (Figure 7C). Four weeks after transplantation, BM and peripheral blood of transplant recipients contained ZsGreen⁺CD45⁺ cells (Figure 7D; Figure S7C and D), indicating that the transplanted ZsGreen⁺ (Col1a2)-tracked CD45⁻ ECs had produced CD45⁺ hematopoietic cell progeny. These ZsGreen⁺CD45⁺ cells in the recipient mice comprised mainly granulocytes and monocytes, and few B and T lymphocytes (Figure 7E). These results indicate that Col1a2-tracked ECs, like Cdh5(VE-Cadherin)-tracked ECs, can give rise to hematopoietic progeny in vivo, but display a more restricted multilineage potential compared to ECs from Cdh5-Cre^ERT2(PAC)/ZsGreen mice.

In additional experiments, we evaluated the role of Runx1 expression in adult BM hemogenic EC since our analyses identified Runx1 as a putative marker of adult EHT (Figure 6E and F; Figure S4E). Therefore, we generated a Runx1^EC-KI mouse line (Cdh5-Cre^ERT2/ZsGreen/Runx1-Knock-In), in which Cre-mediated excision of a floxed STOP codon enables co-expression of ZsGreen and Runx1 in ECs upon tamoxifen induction (Figure S7E)⁶⁶. In these Runx1^EC-KI mice, tamoxifen treatment increased significantly the frequency of ZsGreen⁺ cells in PBMC over 60 weeks compared to control tamoxifen-treated (Cdh5-Cre^ERT2/ZsGreen) mice (Figure 7F).

Since these circulating ZsGreen-tracked cells presumably represent hematopoietic cell progeny of ECs co-expressing ZsGreen and Runx1, we tested the hemogenic potential of these tamoxifen-induced BM EC ex vivo. Using the culture system that successfully supported ex vivo hematopoiesis in Cdh5-tracked BM (Figure 2A-H), we now compared the hemogenic potential of BM cells from tamoxifen pretreated Runx1^EC-KI mice to that of BM cells from tamoxifen pretreated Cdh5-Cre^ERT2/ZsGreen) mice. We detected a significantly greater number of ZsGreen⁺ clusters in cultures from Runx1^EC-KI BM cells compared to control BM cells (Figure 7G and H), and flow cytometry showed that more CD45⁺ZsGreen⁺ hematopoietic cells were produced in cultures of Runx1^EC-KI BM cells compared to BM from controls (Figure 7I and J). Although we cannot exclude the possibility that RUNX1 promotes EC proliferation in culture, these results support a role of RUNX1 in promoting adult EHT.

Discussion

Our results provide evidence for the presence of ECs with hemogenic potential in the adult mouse BM. Previously, hemogenic ECs were detected during embryonic development or perinatally but not thereafter^1–9,67. The current findings argue that EHT is not limited to the prenatal or perinatal development but is present up to 28 weeks after birth, decreasing thereafter. This conclusion is supported by Cdh5-based bulk and Polylox sc lineage tracking, culture of hemogenic ECs, transplant analyses and characterization of EC-derived hematopoietic cell progeny, which link key features of adult EHT to embryonic EHT^25–28,68.

Previous experiments found that EHT, present in the late fetus/neonatal BM, disappears shortly after birth⁶⁷. This contrasts with the current experiments showing persistence of adult EHT well beyond 20 weeks of age. The divergent results likely stem from functional differences of the Cdh5-based tracking systems. In the previous experiments, attempts to activate Cdh5-fluorescence in BM ECs by injecting tamoxifen at different time-points after birth were unsuccessful starting 20 days after birth. Expectedly, the absence of tamoxifen-induced fluorescence in BM ECs was associated with the absence of traced hematopoietic cell output from these cells. In the current experiments and those of others³⁶, tamoxifen administration after 10, 20, and 30 weeks of age consistently induces fluorescence in most BM ECs. We conclude that the absence of adult EHT had not been firmly established.

Adult hemogenic ECs identified here express Cdh5, Pecam1, Kdr, and Flt1, resembling other BM EC populations and embryonic hemogenic ECs⁶⁹. However, the current studies identify yet another, small hemogenic cell population in the adult BM, distinctive in co-expressing typical EC markers, Cdh5 and Pecam1, and the mesenchymal-type markers Lepr, Col1a2 and Cxcl12. These two hemogenic populations are clonally linked, but it is currently unclear whether they have a common progenitor, or one population derives from the other.

ECs derive from two sources during development: the splanchnic mesoderm that gives rise to the primitive aorta where ECs located on the aortic floor are hemogenic⁶⁸, and the somites from the paraxial mesoderm, that produce ECs contributing the endothelial vascular network of the trunk and limb^70,71. Somite-derived ECs are not hemogenic in situ, but they can transiently acquire a hemogenic potential when variously induced⁷² and may also include a cell subset with hemogenic potential⁷³. They can also express Runx1 and trigger aortic hematopoiesis from hemogenic ECs in zebrafish⁷⁴ and in the chick⁶⁸. Besides generating hemogenic ECs, blood and other tissues, the mesoderm in conjunction with the neural crest, gives rise to mesenchymal stem cells (MSCs) and perhaps a precursor of both MSCs and ECs^75–77. However, MSCs do not produce blood cells, although they can differentiate into many other tissues⁷⁸, and their potential for endothelial differentiation remains controversial^79–82. By contrast, endothelial to mesenchymal transition (EndMT) is a well-established process in development and disease states⁸³.

It was suggested that the transient wave of EHT occurring in the BM of the late fetus and perinatally may serve to mitigate the slow HSC colonization of BM from the fetal liver or perhaps prepare the BM niches to accommodate the incoming HSC from the liver, but a function has not yet been firmly established⁶⁷. Similarly, the function of adult BM EHT identified here is currently unclear. We hypothesize that adult EHT plays a role under conditions of hematopoietic stress or disease rather than under steady-state conditions, but this will need future investigation. Interestingly, previous studies found that subsets of EC can regenerate after irradiation that has eliminated HSC⁴⁵, and we traced the emergence of some hematopoietic cells from ECs that persisted after lethal mouse irradiation. We also found that 5-FU treatment was required for the successful engraftment and function of adult hemogenic ECs suggesting that this population may require inductive signals from a BM that is recovering from an insult^45,84. The identification of such inductive factors may enable effective propagation of BM hemogenic ECs ex vivo and motivate a search for hemogenic ECs in human BM.

Altogether, our results provide evidence for a previously unrecognized capacity of ECs in the adult mouse BM to generate blood cells. These results suggest that hematopoiesis in the adult mouse may arise through the contribution of cells and processes beyond the HSCs generated through aortic EHT during development and BM EHT perinatally.

Limitations of the study

A limitation of our study is that the function of adult EHT is currently unknown. Under steady-state conditions, the hematopoietic cell output from adult EHT is low in comparison to that of embryonic EHT, which effectively sustains adult hematopoiesis, and our experiments detected no functional differences between hematopoietic cells from embryonic and adult EHT. To address this limitation, we will examine whether adult EHT is a greater contributor to adult hematopoiesis when stress is imposed on the adult hematopoietic system, and more fully characterize the cell populations arising from adult EHT, focusing on minor cell population known to serve critical functions in specific settings. Another limitation of our study is that we could not extend the in vivo single-cell PolyloxExpress genetic lineage tracing to the ex vivo single-cell genetic lineage tracing due to technical limitations of the PolyloxExpress system. Finally, rigorous serial transplantation experiments, including CD45.1/CD45.2 competitive repopulation assays, will be necessary to conclusively determine whether bona fide long-term hematopoietic stem cells are generated by adult hemogenic endothelial cells both in vivo and ex vivo.

Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Giovanna Tosato (tosatog@mail.nih.gov).

Materials availability

This study did not generate new unique reagents.

Experimental model and study participant details

Mouse strains

All animal studies were approved by the Institutional Animal Care and Use Committee of the CCR (Bethesda, MD), National Cancer Institute (NCI), NIH and conducted in adherence to the NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and approved protocols. Cdh5-Cre^ERT2(PAC) mice ³⁷ (MGI:3848982) were a gift from Dr. R. Adams and Cdh5-Cre^ERT2(BAC) mice ³⁹ (MGI:5705396) were a gift from Dr. Kubota. Col1a2-CreER mice ⁸⁵ were purchased from the Jackson Laboratory (JAX#029567). Cre-dependent Ai6-ZsGreen ⁸⁶ (JAX#007906) and mTmG ⁸⁷ (JAX#007676) fluorescent reporter mice were purchased from the Jackson Laboratory. In ZsGreen reporter mice, Cre activity leads to constitutive expression of ZsGreen1 in cell bodies. In mTmG reporter mice, Cre activity leads to an irreversible switch from cell membrane-localized tdTomato (mT) to membrane-localized EGFP (mG). PolyloxExpress mice ³³ were a kind gift of Drs. Hans-Reimer Rodewald and Avinash Bhandoola. The Cdh5-Cre^ERT2/ZsGreen/PolyloxExpress mouse line was generated in house. Runx1 conditional knock-in mice⁶⁶ (Gt(ROSA)26Sor^{tm1(CAG–Runx1)Lzjg}, MGI:7490340) were generously provided by Drs. Qiufu Ma and Nancy Speck. Runx1 endothelial-specific knock-in (Runx1^EC-KI) mice were generated by crossing Runx1^Ki/+ mice with Cdh5-Cre^ERT2(PAC)/ZsGreen^Tg/Tg mice. Upon tamoxifen treatment, Cre-mediated excision of the two floxed STOP codons enables co-expression of Runx1 and ZsGreen in ECs.

All animals were bred in the animal facilities of CCR/NCI (Bethesda, MD). The mice were maintained in a C57BL/6J background. Mice were identified with ear tags and routinely genotyped by PCR. No mouse was excluded from the experiments, unless assessed as sick by the veterinarians or fight wounds were observed at harvest. Tamoxifen (Sigma-Aldrich, #T5648) dissolved in peanut oil (Sigma-Aldrich, #P2144) (10 mg·mL⁻¹) was administered orally (via gavage using 22g feeding needles) at 100 mg·kg⁻¹. Unless otherwise specified, three doses on consecutive days were administered. Unless indicated otherwise, 8- to 12-week-old male and female mice were used when tamoxifen was administrated. No randomization or blinding was used to allocate experimental groups. Mice were typically sacrificed between 9 AM and 11 AM local time.

Method details

OP9 cell culture

OP9 cells, a gift from Dr. T. Nakano⁴⁴ (also available from ATCC, #CRL-2749), were maintained in α-MEM (Gibco #12561056; without ribonucleosides and deoxyribonucleosides, with 2.2 g/L sodium bicarbonate) supplemented with 20% fetal bovine serum (FBS; Sigma-Aldrich #2442). Culture dishes and flasks (Corning #353003, #430641U) were precoated with gelatin (Sigma-Aldrich #G9391; ∼100 µL/cm², 60 minutes at 37°C). Cells were incubated at 37°C in a humidified atmosphere of 95% air and 5% CO₂.

Bone marrow cell harvesting, culture and terminal harvest for analysis

To harvest BM cells to be cultured, mice were euthanized by cervical dislocation and soaked in 70% ethanol for 5 minutes. Long bones (femurs and tibiae) were dissected, and surrounding skin, muscle, and connective tissue were carefully removed. Cleaned bones were immediately transferred into ice-cold sterile PBS and kept on ice. Once all bones were harvested, they were soaked in 70% ethanol for 1 minute and rinsed three times with ice-cold PBS.

Each bone was cut into two pieces using sterile scissors and placed--with the open end facing downward--into a sterile 500 µL microcentrifuge tube pre-perforated with a 16G needle (sterilized in advance). Each 500 µL tube was loaded with two femurs and two tibiae. To each tube, 200 µL of DMEM supplemented with 1 mM EDTA was added. The 500 µL tubes were then placed inside sterile 1.5 mL microcentrifuge tubes, sealed with Parafilm, and centrifuged at 12,000 rpm for 20 seconds. The inner tubes were discarded, and the BM pellet collected in the 1.5 mL tubes was resuspended in 1 mL DMEM + 1 mM EDTA.

Cells were filtered through 70 µm cell strainers, centrifuged at 350 × g for 5 minutes at 4°C, and resuspended in complete DMEM (DMEM containing 15% FBS (Sigma-Aldrich #2442-500ML, LOT:24G002), Penicillin-Streptomycin [Gibco #15140-122], and Anti-Anti [Gibco #15240-062]). A single-cell suspension was prepared by pipetting 30 times with a 1 mL pipette.

Bone marrow cells were cultured either on Corning Primaria™ dishes/flasks (Corning #353808, #353810, #353846) or on OP9 stromal cell monolayers in standard tissue culture dishes/flasks (Corning #353003, #430641U) precoated with gelatin (Sigma-Aldrich #G9391; ∼100 µL/cm², incubated 60 min at 37°C).

Cells were seeded at a density equivalent to BM cells from two femurs and two tibiae per ∼75 cm² culture surface (e.g., T75 flask) in 15 mL complete DMEM. After ∼32 hours, non-adherent cells and medium were removed, and adherent cells were gently washed three times with PBS. For Primaria cultures, 15 mL fresh complete DMEM was replenished twice weekly. For OP9 co-cultures, non-adherent cells and medium were removed twice weekly and replaced with fresh complete DMEM supplemented with freshly isolated unfractionated WT BM cells (from two femurs and two tibiae per ∼75 cm² surface, using the same BM isolation method mentioned above).

For terminal collection, at the start of the final culture week, all non-adherent cells were removed, and adherent cells were washed three times with PBS. Fresh complete DMEM (15 mL) was added. After three days, a second 15 mL complete DMEM addition was made. No further medium changes occurred before harvest. For final collection, non-adherent cells and supernatant were collected first. Adherent cells were washed with PBS, and the wash was pooled with the supernatant. Remaining adherent cells (including OP9 and BM-derived cells) were detached using 0.25% Trypsin-EDTA (Gibco #25200-056) for 5 minutes at 37°C and added to the pooled suspension. After another PBS wash, a second trypsinization (10 minutes at 37°C) was performed. All collected material from each step was combined for downstream flow cytometric analysis. For transplantation experiments, only non-adherent cells (with supernatant) and loosely adherent cells recovered after the initial 5-minute Trypsin-EDTA incubation were pooled and subjected to FACS sorting to isolate ZsGreen⁺ cells for injection.

Blood collection

For terminal collection, blood was obtained from the mouse abdominal aorta with BD Vacutainer™ EDTA tubes (BD #367856) and Vaculet™ blood collection needles (23G, EXELINT #26766). Blood smears were prepared with 10 µL of collected blood. For flow cytometry analysis, ACK buffer (Lonza, #BP10-548E) was added to the blood to lyse red blood cells before Fc receptor blocking and antibody staining. For non-terminal blood collection, ∼20-50µL blood was collected by submandibular blood sampling, using a 3mm animal lancet (BRAINTREE SCI., GR-3MM) and a 250µL BD Microtainer^® K2EDTA tube (BD #365974).

Kwik Stop^® Styptic Powder was applied to stop the bleeding (Miracle Corp. #423615). For long-term, repeated blood collection, 2 drops of blood were collected from the tail vein with Microhematocrit Capillary Tubes (Fisherbrand # 22-362574). White blood cell (WBC) counts were determined using acridine orange/propidium iodide (AO/PI, Logos Biosystems, #F23001) staining and quantified with a LUNA-FL™ fluorescence cell counter (Logos Biosystems). For flow cytometry detection of ZsGreen/EGFP positive PBMCs, blood was collected with one of the methods above. For flow cytometric detection of ZsGreen⁺/EGFP⁺ PBMCs, blood was collected using one of the three methods described above. Red blood cells were lysed with ACK buffer, and DAPI and DRAQ5 were used to exclude dead cells and to identify nucleated PBMCs.

Bone marrow harvest for flow cytometry analysis

For flow cytometry analysis, bone marrow was harvested using one of the two methods described below. For hematopoietic cell isolation, bone marrow was harvested by flushing femurs and tibiae with ice-cold Sort Buffer (1× PBS [Gibco, #10010-031] supplemented with 5 mM EDTA, 25 mM HEPES, and 2% FBS [Sigma-Aldrich, #F2442]). Red blood cells were then lysed using ACK lysing buffer (Lonza, #10-548E) according to the manufacturer’s instructions. Cells were then washed with Sort Buffer and passed through a 40μm cell strainer (GREINER BIO-ONE, #542040, #542140). For greater preservation of endothelial cells, bone marrows were harvested by gently crushing mouse femurs and tibiae in Sort Buffer (5mM EDTA, 25mM HEPES, 2% FBS in 1× PBS). Red cell lysis was performed using ACK buffer. Bone marrow cells were then incubated with 0.1U·mL⁻¹ Collagenase (Worthington Biomedical Corp., #LS004176), 0.8U·mL⁻¹ Dispase (Worthington Biomedical Corp., #LS02109), and 0.5mg·mL⁻¹ DNase (Worthington Biomedical Corp., #LS006344) in 1x Hanks’ Balanced Salt Solution (HBSS) with Ca²⁺ and Mg²⁺ (Gibco, # 14065056) at 37°C for 30 min on a rotating mixer. Cells were then washed with Sort Buffer and passed through a 40μm cell strainer.

Endothelial cell transplantation experiments

Four days prior to transplant, recipient mice received one dose of 5-FU (150 mg·kg⁻¹ in PBS, Sigma-Aldrich, #F6627) intraperitoneally under isoflurane anesthesia. Sorted bone marrow cells (Col1a2-Cre/ZsGreen: 5,000 cells; Cdh5-Cre/ZsGreen: 20,000 cells in 100μL PBS) were inoculated retro-orbitally under isoflurane anesthesia. Bone marrows and blood were harvested from transplant recipients four weeks after the transplant unless otherwise specified.

Bone marrow ablation and transplantation

Recipient mice were lethally irradiated (11 Gy) using a Cesium-137 (¹³⁷Cs) gamma irradiator three days prior to transplantation.

For whole BM cell transplantation into mTmG recipients, BM cells were harvested from WT donor mice by flushing (as described above); 5 × 10⁶ cells were transplanted per recipient via tail vein.

For LSK transplantation into WT recipients, BM from Cdh5-Cre^ERT2(PAC)/ZsGreen mice (not treated with tamoxifen) was harvested by flushing, followed by red blood cell lysis with ACK buffer. Fc receptor blocking was performed using Azide-Free Fc Receptor Blocker (Innovex, #NB335-60) per manufacturer’s instructions. Cells were stained with antibodies to Lineage cocktail, Sca-1, and c-Kit. DAPI was used to exclude dead cells. LSK populations were sorted using Sony SH800S and its Ultra Purity mode. Transplanted cell numbers were as follows: ZsGreen⁻ LSKs (5 × 10⁴, n = 3; 2.5 × 10⁴, n = 3) and ZsGreen⁺-enriched LSKs (2,800 cells, n = 2).

For transplantation of ex vivo–cultured BM cells, non-adherent and loosely adherent ZsGreen⁺ cells were collected from 8-week OP9 co-cultures (as described above), sorted by FACS for live (DAPI⁻) ZsGreen⁺ cells, and transplanted into irradiated recipients at 5 × 10⁴, 2.5 × 10⁴, or 1.25 × 10⁴ cells per mouse (n = 2 per group).

Flow cytometry and cell sorting

For intracellular antigen detection, single cell suspensions of bone marrow and blood were first incubated with Azide-Free Fc Receptor Blocker (Innovex, #NB335-60), following the manufacturer’s instructions. After washing, cells were first stained with surface marker antibodies at the concentration of 2μg per 1×10⁷ cells in Sort Buffer for 30 minutes at 4°C and then stained with live/dead cell discriminating BioLegend Zombie Dyes (UV, NIR, Violet, or Yellow, BioLegend #423108, #423106, #423114, and # 423104) following the manufacturer’s instructions. After washing, cells were fixed in 4% paraformaldehyde for 10 minutes at 37°C, and then permeabilized with 1% saponin (Sigma-Aldrich, # 47036) /Sort Buffer for 30min on ice. Subsequently, the cells were stained with rat monoclonal primary RUNX1-PE Ab (Invitrogen, # 12-9816-80) in 0.1% saponin/Sort Buffer overnight. After washing and resuspension, propidium iodide (PI, 0.5μM, Millipore Sigma, # P4170), 7-AAD (1μg·mL⁻¹, Millipore Sigma, #A9400) or DAPI (0.5μg·mL⁻¹, BioLegend, #422801) was added as a nuclear counterstain. For live cell staining without cell permeabilization, after cell surface antibody staining, cells were washed and suspended in Sort buffer containing propidium iodide (PI, 0.5μM, Millipore Sigma, # P4170), 7-AAD (1μg·mL⁻¹, Millipore Sigma, #A9400) or DAPI (0.5μg·mL⁻¹, BioLegend, # 422801), to distinguish dead cells from the live cells. For FACS sorting of live endothelial cells, after Fc receptor blocking, bone marrow cells were first depleted of CD45⁺ cells with MojoSort mouse CD45 nanobeads (BioLegend #480028) following the manufacturer’s protocol and then stained with specific antibodies. For flow cytometry analysis, compensation beads (BD Biosciences, #552844) were used for flow cytometer compensation.

Flow cytometric data were acquired with BD FACSCanto-II, BD LSRFortessa, BD FACSymphony A5 (BD Biosciences), Sony SA3800 or Sony ID7000 cell analyzers. Cell sorting was performed with BD FACS Aria III, BD FACS Aria Fusion or Sony SH800S cell sorters.

FSC and SSC profiles were used for excluding dead cells and debris. 7-AAD, PI, DAPI or BioLegend Zombie Dye was used for excluding dead cells. FSC-W versus FSC-H and SSC-W versus SSC-H were used to gate on single cells. Unless otherwise specified, fluorescence minus one (FMO) controls are used for negative gating reference. For BM HSPC analysis, BM cells were harvested followed by lineage positive cell depletion (Biolegend #480004). Data were analyzed with FlowJo (BD, v10.8.1), SONY ID7000 Software (Version 1.2.0.28212) or FACS Diva (BD, v6.1 and v9.0). Flowjo Plugins UMAP_R (v4.0.4) and FlowSOM (v4.1.0) were used for UMAP dimensional reduction and unsupervised clustering of flow cytometry data.

Bone marrow cryosections

Deeply anesthetized mice were transcardially perfused with 20mL ice-cold 1x PBS, followed by perfusion with 15mL ice-cold hydrogel solution (5% acrylamide/bis-acrylamide 19:1 (Sigma-Aldrich, #A2917), 2.5mg/mL polymerization initiator VA-044 (FUJIFILM, Wako, VA-044, Water soluble Azo initiators), 4% PFA in 1× PBS, 5mL/min flow speed). Femurs and tibiae were collected in tubes containing 5mL hydrogel solution and incubated at 4°C for 4 hours. The bones were then washed with PBS and incubated at 37°C for 2 hours. Bone decalcification was performed by incubating the bones in 40mL 0.5M EDTA pH 8.0 (KD Medical, #RGF-3130) for 3 days on a rotate mixer, with daily refreshed 0.5M EDTA solution. The bones were then dehydrated in 20% sucrose and 2% polyvinylpyrrolidone in PBS overnight. Dehydrated bones were then embedded in OCT (SAKURA, #4583) blocks using Precision Cryoembedding System (IHC WORLD, #IW-P101). Cryosections (10μm) for immunofluorescence staining were prepared from OCT frozen bone blocks using Leica CM3050S microtome, low-profile microtome blades (Leica 819, #14035838925), and TruBond™ 380 adhesion slides (Electron Microscopy Sciences, # 63701-W10).

Immunofluorescence staining and imaging

Tissue sections were rehydrated with 1× PBS (15 minutes), permeabilized in 0.3% Triton X-100 (Sigma-Aldrich, #T9284)/PBS (15 minutes), washed in 1× PBS, and incubated (2 hours) with blocking solution (2% BSA, 5% donkey serum (SIGMA, #D9663), and 0.3% Triton X-100/PBS). Samples were then washed 3 times with PBS and incubated with primary antibodies (5ng/mL; 4°C overnight). When secondary antibodies were used, 3 PBS washes were performed before incubating with fluorescent-labeled secondary antibodies (2ng/mL, room temperature, 2 hours). After washing (3x, 10 minutes each with 1× PBS), DAPI was added (300nM in PBS, 10 min). After 3 washes (5 min each with 1× PBS), coverslips were mounted (EPREDIA, #9990402), dried and sealed with nail polish. For blood smear staining, slides were first soaked in acetone/methanol/PFA (19:19:2 for 90 seconds) before rehydration ⁸⁸. Confocal imaging was performed with Zeiss LSM 780, Zeiss LSM 880 NLO Two Photon, or Nikon ECLIPSE Ti2-E SoRa systems, according to the experimental specific needs (resolution, speed, wavelength capabilities). Images were processed with Zen (Zen Black v2.3, release Version 14.0.12.201, Zen Blue Lite v2.5, Carl Zeiss), Bitplane Imaris (v9.7.0, Oxford instruments), and Photoshop (v23.3.0, Adobe, for whole image contrast and brightness adjustments).

Isolation of peritoneal cavity cells

To isolate peritoneal cavity cells, mice were euthanized by cervical dislocation, injected intraperitoneally (i.p.) with cold FACS sort buffer (5 mL), massaged and flicked gently on the abdomen. Peritoneal fluid was then withdrawn slowly and transferred into a polypropylene centrifuge tube on ice prior to centrifugation (350×g, 10min, at 4°C) and analysis.

Thioglycolate induced sterile peritonitis

Mice were injected i.p. with 1 mL PBS or 4% thioglycolate (Sigma-Aldrich, #70157). Peritoneal cavity cells were harvested 4 hours after injection for further analysis.

Phagocytosis assay

Phagocytosis assay was performed using pHrodo™ Red Escherichia coli BioParticles (Invitrogen, # P35361) following the manufacturer’s instructions. Briefly, mice were first treated with thioglycolate (as described). Peritoneal cavity cells were incubated with the fluorescent Escherichia coli particles for 60 min at 37°C. Cells were then stained with surface marker antibodies and analyzed by flow cytometry.

ROS assay

ROS assay was performed using CellROX™ Orange Flow Cytometry Assay Kit (Invitrogen, # C10493) following the manufacturer’s instructions. Briefly, mice were first treated with thioglycolate (as described). CellROX detection reagent was added to peritoneal cavity cells (final concentration of 500nM) prior to incubation for 60 min at 37°C. Cells were then stained with the surface marker antibodies and analyzed by flow cytometry.

PolyloxExpress single cell lineage tracing

These experiments essentially followed published procedures ³³. Briefly, BM cells from tamoxifen treated Cdh5-Cre/ZsGreen/Polylox mice (treated at 10 weeks old and harvested at 16 weeks old) were enriched for ZsGreen positive cells by FACS. Single cell capturing was performed with 10x Genomics Chromium Single Cell 3′ Reagent Kits, following manufacturer’s protocols. After reverse transcription (RT) in droplets, pooled cDNA was amplified and split into two aliquots for parallel transcriptome library preparation and barcode enrichment. Initial library quality control was performed with Agilent TapeStation D5000. For transcriptome analysis, 10 μL (25%) of a 10x cDNA library was fragmented and a gene expression library, generated following protocols in Single Cell 3’ Reagent Kits v3 and v4, was sequenced with Illumina NextSeq 2000 P3/P4 Reagents (28+74 bp read length). For Polylox barcode amplification, targeted amplification of barcodes from a 5-10 ng aliquot of a 10x cDNA library was performed by nested PCR. In the first round, primers #2,652 (5′-GCATGGACGAGCTGTACAAG-3′, annealing at the 5′ end of Polylox) and #2,674 (5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTC-3′, annealing at the adaptor site (read 1) were used for amplification for 5 min at 95°C; (30 s at 95°C, 30 s at 57°C, 3 min at 72°C) 12 times; 10 min at 72°C. PCR products purified with 0.7x AMPure beads were used for the second round of PCR using primers #2,426 (5′-CGACGACACTGCCAAAGATTTC-3′, annealing at the 5′ end of Polylox) and #2,676 (5′-AATGATACGGCGACCACCGA-3′, annealing at the 5′ end of primer #2,674), for 5 min at 95°C; (30 s at 95°C, 30 s at 60°C, 3 min at 72°C) 18 times; 10 min at 72°C. The PCR products were purified with AMPure PB beads according to the manufacturer’s protocol. Long read amplicon-seq libraries were sequenced by PacBio Sequel II with PacBio Amplicons Library Preparation using SMRTbell prep kit 3.0. A custom transcriptome reference was built from mouse reference MM10 to include ZsGreen1.

A snakemake workflow with custom python script was used to retrieve cell indexes and Polylox barcodes from the PolyloxExpress amplicons (https://github.com/CCRSF-IFX/SF_Polylox-BC).

Single cell transcriptome and single cell barcodes were linked using the 10x 3’ kit cell index and group information. Further analysis and illustrations were generated using Scanpy (https://github.com/scverse/scanpy). Doublet detection was performed using Scrublet, with the predicted doublet rate calculated based on 10x Genomics guidance about the cell numbers loaded to the microfluidic chips. Batch correction was done using BBKNN method (https://github.com/Teichlab/bbknn). Cell types were manually annotated based on canonical marker gene expression, guided by results from three automated annotation tools: (1) decoupleR ((https://saezlab.github.io/decoupleR/), using PangLaoDB (https://panglaodb.se/) as reference; (2) scANVI (https://github.com/scverse/scvi-tools) using ImmGne (https://www.immgen.org/) as reference; (3) CellTypist, using the embedded Immune_All_Low model. The rare barcodes and their pGen were identified using the MatLab script from the Höefer’s Lab (https://github.com/hoefer-lab/polylox). True barcodes are defined as pGen < 1×10⁻⁴, such that the expectation of a True barcode in detected 4,072 cells is 0.407. This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov) and Frederick Research Computing Environment (FRCE). Python version: 3.10.; R version 4.5.0.

Public single-cell RNA-seq data analysis

Raw FASTQ and BAM files were downloaded from publicly available datasets (GSE108885 ⁵⁴, GSE108891 ⁵⁴, GSE118436 ⁵⁴, GSE123078 ⁵⁴, GSE122465 ⁵⁵, GSE128423 ⁵³, GSE145477 ⁶⁵, GSE156635 ⁵⁶, GSE137116 ²⁴, E-MTAB-8077 ⁶⁵, GSE230260⁵⁸, GSE259382⁵⁹). The BAM files were first converted to FASTQ files with bamtofastq (10x Genomics, v2.0.1). All FASTQ files were then processed by the count function of Cell Ranger (10x Genomics, v8.0.1) and aligned to the mouse genome (mm10, version 2020-A), to generate read matrixes. Further analysis and illustrations were generated using Scanpy as described above.

Definition of cultured fluorescent BM cell clusters and quantification

A cluster was defined as (1) a spatially distinct group of ZsGreen⁺ cells not contiguous with other fluorescent cells, (2) containing a central core with at least five extending branches, and (3) exhibiting a roughly radial organization, with projections spreading outward in multiple directions, and (4) extending at least ∼200 µm in one direction. Isolated or irregularly scattered fluorescent cells were not counted. Due to their large size, clusters were counted manually across the culture dish using a tally counter under fluorescence microscopy.

RNA isolation and quantitative RT PCR

RNA was extracted using RNeasy^® Micro Kit (QIAGEN, #74004), following the manufacturer’s protocol. Cells were sorted directly into RNA lysis buffer (Buffer RTL of RNeasy^® Micro Kit). cDNA samples were prepared with SuperScript IV Reverse Transcriptase (Invitrogen, #18091050), following manufacturer’s instructions. Real Time PCR was performed using Applied Biosystems™ TaqMan™ Fast Advanced Master Mix (4444557) and Applied Biosystem™ QuantStudio™ 5 Real-Time PCR System. TaqMan™ probes used were purchased from Applied Biosystems™: Spp1 (Mm00436767_m1); Cxcl12 (Mm00445553_m1); Col1a2 (Mm00483888_m1); Cdh5 (Mm00486938_m1); Runx1 (Mm01213404_m1); Ptprc (Mm01293577_m1); Gapdh (Mm99999915_g1); Actb (Mm02619580_g1). The reaction condition was set as follows: 50°C 2 minutes, 95°C 20 seconds, 45 cycles of 95°C 1 second, 60°C 20 seconds. Ct values were determined using the ABI QuantStudio™ Design & Analysis Software (v1.5.2). Relative gene expression was assessed using the 2^−ΔΔCt method, normalized to Gapdh expression level for each sample. The data was further normalized to gene expression levels in the unsorted bone marrow sample to calculate relative gene expression levels in each sample. Data reflect triplicates real-time PRC experiments.

Quantification and statistical analysis

No statistical method was used to predetermine sample size. No data were excluded from the analyses. Mice with the correct genotypes were randomly assigned to control or treated groups. Unless otherwise specified, data are represented as mean ± S.D, and individual dots in the graphs indicate individual mice. Comparisons between two groups were performed using two-tailed unpaired Student’s t-tests (except for Figure S1-H, which used a paired t-test). Spearman rank correlation test was used for Figure 3E. Statistical analyses were performed with GraphPad Prism (v9.0.1). A statistical difference of P<0.05 was considered significant: ns, not significant, * P < 0.05, ** P < 0.01, *** P< 0.001.

Data availability

The results of scRNAseq and PacBio SmrtSeq of Polylox barcodes are deposited in NCBI SRA (PRJNA1079369, public at time of publication). A custom script was used to retrieve cell indexes from the PolyloxExpress amplicons (https://github.com/CCRSF-IFX/SF_Polylox-BC). Bash, R, and Python codes are available from the corresponding author upon reasonable request.

Acknowledgements

This project is supported by the Intramural Program of CCR, NCI, NIH. The findings are those of the authors and do not necessarily represent the official views of the NIH or the Department of Health and Human Services. This work used the computational resources of the NIH High Performance Computing (HPC) Biowulf cluster (http://hpc.nih.gov) and Frederick Research Computing Environment (FRCE). Flow cytometry and cell sorting were performed at the CCR Flow Cytometry Core Facility; microscopy analyses at the CCR Microscopy Core Facility in Building 37 of the NCI, NIH. We thank Drs. Ralf Adams and Manfred Boehm for mouse line Cdh5-Cre^ERT2(PAC); Drs. Yoshiaki Kubota and Yosuke Mukoyama for mouse line Cdh5-Cre^ERT2(BAC); Dr. Hans-Reimer Rodewald for the PolyloxExpress mouse line. We thank Dr. S. Banerjee, Dr. S. Siddiqui and Ms. K. M. Wolcott for flow cytometry support; Dr. M. Kruhlak and Mr. L. Lim for confocal microscopy support and Mr. A. Abdelmaksoud for bioinformatics assistance. We thank Drs. D. Lowy, R. Yarchoan, M. DiPrima and H. Ohnuki for thoughtfully commenting on aspects of this work.

Additional information

Author contributions

Concept and design: J-X.F. and G.T. Acquisition, analysis, or interpretation of data: J-X.F., G.T., C.L., F.L., N.T., A.B., L.L, M-T.Y., D.W., J.C., and Y.Z. G. Drafting of the manuscript: J-X.F., G.T. Critical revision of the manuscript for important intellectual content: J-X.F., G.T., A.B. and N.T.

Funding

Intramural Program of CCR/NCI/NIH

• Giovanna Tosato

Additional files

Supplemental Figures