Lin− PU.1dimGATA‐1− defines haematopoietic stem cells with long‐term multilineage reconstitution activity

Abstract Despite extensive characterization of the state and function of haematopoietic stem cells (HSCs), the use of transcription factors to define the HSC population is still limited. We show here that the HSC population in mouse bone marrow can be defined by the distinct expression levels of Spi1 and Gata1. By using a double fluorescence knock‐in mouse model, PGdKI, in which the expression levels of PU.1 and GATA‐1 are indicated by the expression of GFP and mCherry, respectively, we uncover that the HSCs with lymphoid and myeloid repopulating activity are specifically enriched in a Lin−PU.1dimGATA‐1− (LPG) cell subset. In vivo competitive repopulation assays demonstrate that bone marrow cells gated by LPG exhibit haematopoietic reconstitution activity which is comparable to that of classical Lin−Sca1+c‐kit+ (LSK). The integrated analysis of single‐cell RNA sequence data from LPG and LSK‐gated cells reveals that a transcriptional network governed by core TFs contributes to regulation of HSC multipotency. These discoveries provide new clues for HSC characterization and functional study.


| INTRODUCTION
Haematopoiesis is a hierarchical and dynamic process involving haematopoietic cells at distinct differentiation stages like multipotent stem cells, lineage-restricted progenitors and specialized blood cells.
The isolation and identification of functional haematopoietic stem cells (HSCs) is a major challenge due to their rarity and heterogeneity.
An early study showed that HSCs existing in a subpopulation called Lin À Sca1 + c-kit + (LSK), which is negative for markers of mature blood cells and positive for Sca1 and c-kit (LSK), have the ability to reconstitute all blood cell types in mouse bone marrow. 1 Subsequently, a series of surface markers have been combined with LSK for purification and characterization of HSCs.[4] The surface marker EPCR, encoded by the Procr gene, has been suggested to track the maturation process from pre-HSCs to HSCs in mouse embryos. 5Cells purified solely based on EPCR expression show potent haematopoietic reconstitution activity, which is comparable to the cells gated by the Hoechst 33342 staining method in mouse bone marrow. 6,7Similarly, the expression of platelet integrin CD41 (αIIb), a surface marker regulated by GATA-1, successfully marks the transient initiation of definitive haematopoiesis during murine ontogeny. 8CD41 is expressed in a subset of long-term myeloid-biassed CD150 + CD48 À LSK cells in adult mouse bone marrow, and the expression of CD41 in adult long-term HSCs becomes prevalent with age. 91][12][13] Accumulating evidence suggests that haematopoietic transcription factors (TFs) play pivotal roles in the maintenance and differentiation of HSCs.5][16][17][18] Lineage-specific TFs play essential roles in controlling the progression of haematopoietic differentiation, and their coordinated expression primes the developmental potential of cells with mixed-lineage states.0][21] PU.1 and GATA-1 proteins directly bind to each other as a negative regulator of a network of downstream genes. 22,23They are cross-antagonistic transcription factors involved in erythroid and myeloid cell lineage priming in HSCs. 244][35][36][37] However, without evaluation of the long-term haematopoietic capacity, many bone marrow cells expressing those TFs have been mislabelled as HSCs. 25In addition, platelet-related TFs such as GATA-1, TPO-R and vWF are highly expressed in long-term HSCs and down-regulated in short-term HSCs. 38While vWF + HSCs can give rise to myeloidbiassed HSCs and lymphoid-biassed HSCs, a specific population of these cells, called platelet-biassed HSCs, also effectively and stably replenish megakaryocyte/platelet-lineage cells but not other blood cell lineages. 39,40While vWF + HSCs are highly enriched in the megakaryocyte niche, NG2 + arteriolar niche cells selectively maintain vWF À HSCs in the bone marrow. 30,41These observations suggest that and the donor templates, the knock-in mice were generated by CRISPR-Cas9-mediated gene-editing technology. 42Offspring were then genotyped using PCR analysis with the primer set for Spi1 knock-in mice (upstream arm forward primer: TCTCTGCCATCCC TCACTGACCTTC and IRES reverse primer: GCACACCGGCCTTATTC CAAGC; EGFP forward primer: ACATGGTCCTGCTGGAGTTCGTG and downstream arm reverse primer: TGCTATGCTTATCTCCGA GTCGTCCAG; forward primer on the upstream arm: GGAGGGTCCC CATAAAATC and reverse primer on the downstream arm: TGATC CCTGAGCCCTGATA) and the primer set for the Gata1 knock-in mice (upstream arm forward primer: CAAACGGGCAGGCACCCAATG and IRES reverse primer: GCACACCGGCCTTATTCCAAGC; mCherry forward primer: CGAGGACTACACCATCGTGGAACAG and downstream arm reverse primer: TCTGCCTTGCCTCTGCCACCG; forward primer on the upstream arm: TCCCTCTTTGCTCCTCTTTCT and reverse primer on the downstream arm: CTCCATGCTCCACTTGACACT).
These primers can detect both wild-type (WT) and knock-in alleles at the same time.The Spi1-IRES-puroR-p2A-eGFP and Gata1-IRES-puroR-p2A-mCherry double knock-in mice (C57BL/6 background) were obtained by mating the two types of knock-in mice together.
The homozygous mice were utilized in this study.

| CFUs assay
A total of 12,500 total bone marrow cells were isolated from the legs and hips of PGdKI mice and cultured in methylcellulose-based medium (StemCell technology, 03434) with recombinant cytokines (including EPO) for mouse cells.Number and morphology of the colonies were analysed with phase contrast microscope after 7 days of culture.

| Competitive repopulation assay
Recipient mice were treated with lethal irradiation (9.0 Gy) for bone marrow transplantation experiments.10 5 Lin À bone marrow cells (CD45.2) were sorted using a flow cytometer (BD FACS Aria II) depending on the expression of PU-1 and GATA-1, and mixed with 5 Â 10 5 whole bone marrow competitor cells (CD45.1) in 200 μL PBS, and injected into tail veins of recipient mice.For secondary transplants, 2 Â 10 6 whole bone marrow cells from primary recipient mice were transplanted into lethally irradiated recipient mice.

| Peripheral blood analysis
Sixteen weeks after transplantation, peripheral blood was collected from the tail vein of recipient mice.Collection was repeated at 4-week intervals.Then, the erythrocytes were lysed using erythrocyte lysis buffer (Shanghai Yeasen Biotech Co. Ltd), and the remaining cells were stained with antibodies against the following markers: anti-CD3 -PerCP/Cya-nine5.5 (Biolegend, 100,218), anti-CD45R/B220 -APC (Biolegend,

| Single-cell RNA sequencing
Single-cell RNA sequencing (scRNA-seq) was performed in cooperation with CapitalBio Technology Inc. (Beijing, China).Briefly, bone marrow tissues were flushed from adult mouse femurs with PBS buffer and passed through 40 μm strainers, and the LSK cells were collected and counted using a flow cytometer (BD FACS Aria II).Then single-cell RNA-seq libraries were constructed according to the instructions accompanying the single cell 3 0 Library and Gel Bead Kit V3 (10Â Genomics, 1,000,075).The cDNA libraries were subsequently generated, amplified, and assessed for quality control using the Agilent 4200, and single-cell RNA sequencing was further performed on the Illumina Novaseq6000 sequencer.Read pre-processing was performed using the 10 Â Genomics workflow.For quality control, we set the threshold and removed cells with more than 10% of reads mapping to mitochondrial genes (regarded as low-quality cells that exhibit extensive mitochondrial contamination).A final dataset of 12,814 LSK cells was preserved to map the landscape of HSC development.
LPG cells were enriched and processed as described above.
ScRNA-seq data of 12,814 LSK cells merged with 19,608 LPG cells were computed and visualized using R packages.

| Cell clustering and analysis
[46] For the Seurat analysis, we used the Read10Â function to read the output from the 10Â Genomics Cell Ranger (version = 4.0) mapped to the mm10 reference genome.This returned a unique molecular recognition (UMI) count matrix, which was used to create a Seurat object.For quality control, we visualized quality control metrics, and applied them to filter cells.We removed cells with more than 30,000 reads and fewer than 3000 reads, or over 10% mitochondrial In addition, we used pySCENIC to infer the gene regulatory networks (GRNs) based on DNA motif analysis and the co-expression of TFs and target genes.TFs were identified using GENIE3/GRNBoost and compiled into modules (regulons) that were further subjected to cis-regulatory motif analysis by RcisTarget.To identify cell states and their regulators, we assessed the network activity in each individual cell (AUCell), and scoring regulons in the cells.We identified stable cell states based on their gene regulatory network activity (cell clustering).

| Data and code availability
The raw data files for the scRNA sequencing data have been deposited in the Science Data Bank.All relevant data supporting the findings of this study are also available from the lead contact (tbzhao@ioz.ac.cn) upon request.

| ScRNA-seq analysis of LSK cells reveals that multipotent HSCs and erythroid-biassed HSCs can be distinguished by expression of Spi1 and Gata1
To explore the potential TFs that can be used to isolate and characterize haematopoietic stem cells (HSCs) in adult mouse bone marrow, we first performed single-cell RNA sequencing (scRNA-seq) on LSK cells.
We then analysed the differentially expressed genes (DEGs) among these clusters (Figure 1C).The top 5 DEGs in the HSC1 cluster are Prkg1, Fgfr1, Mecom, Hes1 and Rora, among which Fgfr1, Mecom and Hes1 have been shown to be essential for HSC mobilization and functioning. 29,47,48[51][52][53] Two TF genes, Hlf and Ctla2a, which promote haematopoietic transplantation engraftment, are highly expressed in the MPP1 cluster. 54,55The gene Diaph3, which is expressed in erythroid progenitor cells, 56 is highly expressed in both the MPP2 and MPP3 clusters.In the MPP3 cluster, there is strong expression of the gene Hmgb2, which has been shown as a regulator for HSC regeneration, 57 and Mki67, which encodes a marker of proliferation. 580][61] The expression levels of the lymphoid-related genes Dntt and Il18rap are increased in the LMPP cluster, which is consistent with the lymphoid-primed characteristics of these cells 62,63 (Figure 1C).
There are 3 broad trajectories of HSC differentiation landscape: lymphoid, myeloid (granulocyte and monocyte), and erythroid (erythrocyte and megakaryocyte). 64We next assessed the expression of haematopoietic TFs in the seven clusters to predict their differentiation trajectories (Figure 1D).6][67] Unexpectedly, we found enhanced expression of the platelet-related gene Vwf in HSC1 and HSC2. 41In addition, the HSC1 cluster strongly expressed the myeloid-and lymphoid-related genes Egr3 and Gata3. 68,69In the HSC2 cluster, the expression levels of the erythroid trajectory-specific TF genes Bmi1, Gfi1b, Gata1 and Klf1 70-73 were enhanced, while the expression levels of the myeloid-and lymphoid-related genes Gfi1, Spi1 (encoding PU.1) and Nfe2l2 [74][75][76] were reduced (Figure 1D).These results indicate that the HSC1 cluster possesses myeloid and lymphoid differentiation potential, whereas the HSC2 cluster possesses erythroid-megakaryocyte differentiation potential, which is consistent with the results from DEG analysis (Figure 1C).
Given the distinct expression patterns of Spi1 and Gata1 in haematopoietic lineages and their reciprocal activation effects on HSC specification, we aimed to evaluate whether they can be used together to distinguish the multipotent HSC populations.We found that the two putative HSC clusters in LSK cells can be distinguished by the expression differences between Spi1 and Gata1 (Figure 1E).
Interestingly, the PU.1 dim GATA-1 À subset, but not the GATA-1 + , PU.1 À GATA-1 À and PU.1 high GATA-1 À subsets, of Lin À bone marrow cells were able to undergo long-term lymphoid and myeloid lineage differentiation in irradiated recipient mice after the first and second bone marrow transplantation (Figure 3C,D).These data provide evidence that the multipotent HSCs are enriched in the LPG population of mouse bone marrow cells.
We then compared the haematopoietic reconstitution capacity between LSK and LPG cells.LPG cells account for about 50.53% ± 3.86% of LSK cells, while LSK cells occupy around 12.43% ± 1.76% of LPG cells (Figure S2E,F).To measure the frequency of functional HSCs, we performed limiting dilution assays on LSK and LPG cells, and the chimerism rates of peripheral blood cells in the recipients were examined 16 weeks after transplantation to evaluate the competitive repopulating units (CRUs).The results showed that the frequency of functional HSCs in LSK and LPG cells is 1/334 versus 1/1671.2(Figure 3F,G).Notably, Sca1 + c-kit + cells account for around 2.23% of Lin À cells, while PU.1 dim GATA-1 À gated approximately 10.5% of Lin À cells (Figure 3E,G).Thus, LPG gates the multipotent HSC population at the same level as LSK in mouse bone marrow.

| Core TFs control the multipotency of HSCs via distinct gene regulatory networks
We next performed the scRNA-seq analysis on LPG cells, and integrated these data with the scRNA-seq data from LSK cells by the "Harmony" method.Combined with canonical haematopoietic surface markers, the integrated LSK and LPG cells were designated into two HSC clusters and 10 progenitor clusters, which were visualized by UMAP 93 (Figures 4A-C and S3).Similar to the HSC1 and HSC2 clusters identified in LSK cells, the HSC cluster expressing Spi1 but not Gata1 was denoted to be the multipotent HSC cluster (multi-HSC), while the HSC cluster expressing Gata1 was classified to be the platelet-biassed HSC cluster (plt-b HSC) (Figure 4B).The Spi1 and Gata1 expression patterns in the HSC clusters of LPG cells exhibited high similarity to their counterparts in LSK cells.The distinct expression pattern of Procr and Itga2b in the two HSC clusters further supported this designation (Figure 4C).
To gain insight into the mechanism underlying HSC multipotency regulation, we searched for the transcriptional signatures within each cluster by pySCENIC, in which the positive correlation coefficient of expression between a TF and its targets can be detected.The resultant top 5 active TFs were listed for each cluster (Figure S4A). 45,94I G U R E 2 Spi1 is moderately expressed in haematopoietic stem cell (HSC) populations gated by traditional surface markers.(A) Diagram of the knock-in alleles in the mouse genes Spi1 and Gata1.IRES-puroR-p2A-eGFP and IRES-puroR-p2A-mCherry expression cassettes were inserted after the stop codons of Spi1 and Gata1, respectively.The locations of primers for genotyping analysis are indicated.(B-C) Genotyping analysis of the knock-in mice.The primers used are indicated in A. (D) FACS analysis shows that the Lin À bone marrow cells of PGdKI mice are classified into four cell subsets according to different expression levels of eGFP and mCherry.(E) The percentages of the four indicated subpopulations in Lin À bone marrow cells.Data are shown as mean ± SEM, n = 4. (F-I) Expression of PU.1 and GATA-1 in HSCs identified by traditional HSC surface markers: (F) Lin À Sca-1 + c-kit + (LSK) CD34 -/low CD48 À , (G) Lin À Sca-1 + EPCR + (LSE) CD34 -/low CD48 À , and (H) LSKCD34 -/low CD41 + cells.(I) The contribution of the indicated cell subsets in the CD34 À/dim CD48 À LSK population, the CD34 À/dim CD48 À LSE population, and the CD34 À/ dim CD41 + LSK population (mean ± SEM, n = 3).GATA-1 + cells are gated in the red box, PU.1 À GATA-1 À cells are gated in the orange box, PU.1 dim GATA-1 À cells are gated in the green box, and PU.1 high GATA-1 À cells are gated in the black box.The TFs Zfp612, Foxq1, Hoxb5, Hoxa10 and Smad6 were identified at the top of the transcriptional activity list in the multi-HSC population (Figure S4A).Interestingly, analysis of the clustered cells in a regulon activity matrix, which plotted the top five active regulons in each distinct cluster, demonstrated that the majority of cells enriched for these five TFs are multi-HSCs and MPP1s (Figure 4D).This finding indicates that these TFs play pivotal roles in multipotency regulation.
By SCENIC analysis on integrated data from LPG and LSK cells, we uncovered that a regulatory network, containing five core TFs (Zfp612, Foxq1, Hoxb5, Hoxa10 and Smad6) and five downstream co-regulated genes (Prdm16, Ski, Mecom, Nrarp and Efna3), exists in multipotent HSC subsets.This network may play important roles in HSC fate determination.While Hoxa10, Prdm16, Ski, Smad6 and Efna3 have been demonstrated to be critical for HSC multipotency maintenance, [95][96][97][98][99][100] the transcription factors MECOM and Hox-B5 have been shown to be highly expressed in HSCs of mouse bone marrow and are essential for HSC selfrenewal. 26,29,101Accordingly, we found that Mecom and Hoxb5 are highly expressed in the multipotent HSC1 population refined from the integrated LPG and LSK populations (Figure S4B).Furthermore, the majority of Lin À Hox-B5 + cells are PU.1 dim GATA-1 À , and only the PU.1 dim GATA-1 À subpopulation of Lin À Hox-B5 + cells possess longterm haematopoietic reconstitution capacity (Figure S5A,B).
Existing studies have demonstrated that the bone morphogenetic protein (BMP) signalling pathway is activated in myeloid-lymphoid balanced rather than myeloid-biassed HSC populations, 102 while the transforming growth factor beta (TGF-β) signalling pathway is generally activated in myeloid-biassed HSC populations. 103Accordingly, the core TF SMAD6 that we identified here by regulon analysis has been reported to be correlated with the BMP signalling pathway and inhibits erythropoiesis. 104The TF SKI, which is co-regulated by SMAD6 and HOXA10, inhibits the TGF-β signalling pathway. 98 addition, histone demethylation has been revealed to be required for haematopoietic stem cell maintenance. 105The TF MECOM, which is predicted to be co-regulated by SMAD6, Hox-B5, FOXQ1 and ZFP612, directly interacts with the histone methyltransferase SUV39H1 to form an active complex with methyltransferase activity, and the TF PRDM16, which is co-regulated by Hox-B5 and Hox-A10, is an H3K9me1 methyltransferase. 106,107These data may suggest undefined novel epigenetic and transcriptional mechanisms that regulate HSC multipotency.
In conclusion, by combining single-cell transcriptomics with lineage tracing, we uncover that LPG defines a population of mouse bone marrow HSCs with the capacity to reconstitute multiple haematopoietic lineages.In addition, by using pySCENIC analysis of integrated data from LPG and LSK cells, we identified a novel gene regulatory network consisting of the core TFs Zfp612, Foxq1, Hoxb5, Hoxa10 and Smad6.The detailed mechanism of how these core TFs and their co-regulated genes coordinate to regulate HSC fate requires further investigation.

2 . 2 |
both multipotent HSCs and platelet-biassed HSCs are regulated by distinct niches.Understanding the transcriptional network maintaining HSC multipotency is important for precise identification and characterization of the functional HSCs.In this study, we have demonstrated that HSCs express TFs PU.1 and GATA-1 at distinct levels, and multipotent HSCs are specifically enriched in the Lin À PU.1 dim GATA-1 À (LPG) subset of mouse bone marrow cells.By scRNA-seq analysis on the LPG and LSK populations, we reveal that a gene regulatory network, controlled by core TFs, plays pivotal roles in HSC multipotency maintenance.Our discovery provides new clues for understanding transcriptional regulation of mouse HSCs. 2 | MATERIALS AND METHODS 2.1 | Mice All animal experiments were approved by the Ethics Committee in the Institute of Zoology, Chinese Academy of Sciences in accordance with the Guidelines for Care and Use of Laboratory Animals established by the Beijing Association for Laboratory Animal Science.C57BL/6 mice, including CD45.1, CD45.2 and CD45.1/CD45.2,congenic mice, at the age of 8-12 weeks were used for transplantation assays.The Spi1-IRES-puroR-p2A-eGFP (C57BL/6 background) and Gata1-IRES-puroR-p2A-mCherry (C57BL/6 background) knock-in mice were bred at the Transgenic Research Center, National Institute of Biological Sciences (Beijing).The C57BL/6-Hoxb5 em2(2A-tdTomato)Smoc mice were purchased from Shanghai Model Organisms Center, Inc. (Shanghai), and C57BL/6 mice were purchased from Vital River Laboratory Animal Technology Co. Ltd. (Beijing) and SPF Biotechnology Co., Ltd.(Beijing) at 7-10 weeks of age.Generation of Spi1 eGFP/eGFP and Gata1 mCherry/mCherry double knock-in (PGdKI) mice IRES-puroR-p2A-eGFP and IRES-puroR-p2A-mCherry expression cassettes were inserted after the stop codons of the Spi1 and Gata1 genes, respectively, by homologous recombination.Briefly, the guide RNA sequence for the Spi1 gene (CACCTACCAGTTCAGCGGCG) or the Gata1 gene (GTTGTAGGCGATCCCAGCAG) was designed according to the CRISPR design tool (http://crispr.mit.edu/) and inserted into the BbsI-digested pX330 plasmid (Addgene, Cambridge, MA; plasmid 42230).The plasmid sequence was confirmed.A donor template for homologous recombination was constructed containing the sequences on each side of the stop codon of Spi1 (Gene ID: 20375) or Gata1 (Gene ID: 14460).Both upstream and downstream arm sequences were amplified by the polymerase chain reaction (PCR) from C57BL/6 mouse genomic DNA and cloned into the pBluescript II SK(+) vector adjacent to the insertion fragment IRES-puroR-p2A-eGFP or IRES-puroR-p2A-mCherry.Using the pX330 plasmids
reads.Next, we normalized the data using the NormalizeData function in Seurat.We selected the top 2000 variably expressed genes (VEGs) utilizing the FindVariableFeatures function in Seurat with the default setting.The VEGs were further scaled by the ScaleData function in Seurat.For principal component analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP), the first 20 principal components were performed and then used.To identify clusters of cells by a shared nearest neighbour (SNN) modularity optimization, we utilized the FindNeighbors and FindClusters functions.To visualize and explore these datasets., we used the RunUMAP function to run nonlinear dimensional reduction.To identify the DEGs among clusters, we used the FindMarkers function in Seurat.
Statistical details and number of replicates are shown in the corresponding figure legends.Analyses were performed using Prism 8 (GraphPad Software).Statistical significance was calculated using unpaired t tests between the indicated groups, p-values <0.05 were considered as statistically significant.p-Values are indicated by asterisks as follows: *p < 0.05, **p < 0.01, ***p < 0.001.
1 antibodies and observed them by immunofluorescence microscopy.The results F I G U R E 1 Single-cell RNA sequencing analysis of the Lin À Sca1 + c-kit + (LSK) cellular population in mouse bone marrow.(A) Uniform Manifold Approximation and Projection (UMAP) showing 7 clusters identified from 10 Â scRNA-seq analysis of all mouse bone marrow LSK cells.Each dot represents one cell.(B) The expression pattern of eight HSPC-related cell marker genes within the cells from A. (C) Dot plot showing the average expression levels and expression proportions of the top 5 different expression genes (DEGs) within each cluster.The size of the dot represents the proportion of cells expressing the indicated gene within a cluster, and the colour of the dot indicates the average expression level of cells within a cluster.The top 5 DEGs are labelled in the same font colour as the corresponding cluster in A. (D) Heatmap showing the mean expression of haematopoietic lineage-specific transcription factors in each cluster.(E) Violin plot showing the expression of transcription factor genes, including Spi1 and Gata1, in the HSC1 and HSC2 clusters.HSC, haematopoietic stem cell; HSPC, haematopoietic stem and progenitor cell; LMPP: lymphoid-primed multipotent progenitor cell; MPP, multipotent progenitor cell.

F
I G U R E 2 Legend on next page.

F I G U R E 4
Integrated scRNA-seq analysis of Lin À PU.1 dim GATA-1 À (LPG )and Lin À Sca1 + c-kit + (LSK) cells reveal novel transcriptional regulation networks in haematopoietic stem cells (HSCs).(A) Uniform Manifold Approximation and Projection (UMAP) visualization of integrated single-cell sequencing data from LSK cells and LPG cells.Each dot represents one cell.(B) The expression of Spi1 and Gata1 in distinct clusters of LPG and LSK cells.(C) The expression pattern of Procr and Itga2b in distinct clusters of LPG and LSK cells.(D) SCENIC results for the clustered cells.The clustered regulon activity matrix plots the top five active regulons within each distinct cluster and displays them as a heatmap of z-scored enrichment values.The coloured bars on the right of the heatmap indicate the group and the cluster of every single cell.The putative regulons of HSCs are marked by a red dotted box and are magnified.(E) HSC-related gene regulatory network.Blue dots indicate core TF genes and red dots indicate downstream genes.Smad6, Hoxa10, Hoxb5, Foxq1 and Zfp612 serve as master regulators for active regulons in multipotent HSCs.The co-regulated downstream genes of master regulators are labelled in orange text, and other target genes which are regulated by only one master transcription factor are marked in black text.CLP, common lymphoid progenitor cells, CMP, common myeloid progenitor cells, EILP, early innate lymphoid progenitor cells; MPP, multipotent progenitor cells; multi-HSCs: multipotent HSCs; plt-b HSCs: platelet-biassed HSCs.