Suppression of SRCAP chromatin remodelling complex and restriction of lymphoid lineage commitment by Pcid2

Lymphoid lineage commitment is an important process in haematopoiesis, which forms the immune system to protect the host from pathogen invasion. However, how multipotent progenitors (MPP) switch into common lymphoid progenitors (CLP) or common myeloid progenitors (CMP) during this process remains elusive. Here we show that PCI domain-containing protein 2 (Pcid2) is highly expressed in MPPs. Pcid2 deletion in the haematopoietic system causes skewed lymphoid lineage specification. In MPPs, Pcid2 interacts with the Zinc finger HIT-type containing 1 (ZNHIT1) to block Snf2-related CREBBP activator protein (SRCAP) activity and prevents the deposition of histone variant H2A.Z and transcription factor PU.1 to key lymphoid fate regulator genes. Furthermore, Znhit1 deletion also abrogates H2A/H2A.Z exchange in MPPs. Thus Pcid2 controls lymphoid lineage commitment through the regulation of SRCAP remodelling activity. Haematopoiesis and the generation of lymphoid cell subsets are controlled by delicate genetic programs enforced via epigenetic regulation. Here the authors show that Pcid2 interacts with ZNHIT1, a component of the SRCAP chromatin remodelling complex, to critically modulate the differentiation of multipotent progenitors.

A dult haematopoiesis depends on a rare population of haematopoietic stem cells (HSC) in the bone marrow (BM) that possess the capacity for self-renewal and differentiation 1 . HSCs comprise long-term HSCs (LT-HSC) and short-term HSCs (ST-HSC). LT-HSCs, at the very top of the cellular hierarchy, are endowed with the ability to continuous supply of blood cells owing to their self-renewal and differentiation 2,3 . ST-HSCs, losing self-renewal ability, are doomed to differentiate and give rise to multiple blood cell lineages. Multipotent progenitors (MPPs), a downstream progenitor of ST-HSCs, can generate either common lymphoid progenitors (CLPs) or common myeloid progenitors (CMPs) [4][5][6] . CLPs produce all lymphoid cells but lose myeloid potential 7 , whereas CMPs give rise to myeloid cells and lose lymphoid capacity 8 . The differentiation into lymphoid-or myeloid-restricted progenitors are tightly controlled by intrinsic and extrinsic signals 9,10 . However, the underlying mechanism regulating MPP fate decisions into CLPs or CMPs remains elusive. Pcid2 (PCI-domain containing protein 2) is a homologue of yeast protein Thp1 that participates in the export of mRNAs from the nucleus to cytoplasm 11 . A report showed that Pcid2 is in the human TREX2 complex and prevents RNA-mediated genome instability 12 . Through genome-scale RNA interference (RNAi) screening, Pcid2 was identified to be an important factor that is involved in the self-renewal of mouse embryonic stem cells (ESCs) 13 . We demonstrated that Pcid2 modulates the pluripotency of mouse and human ESCs via regulation of EID1 protein stability 14 . Moreover, Pcid2 is selectively involved in the transport of MAD2 mRNA that modulates the mitotic checkpoint during B-cell development 15 . However, how Pcid2 modulates the HSC fate decision in mammalian haematopoiesis is still unclear.
During differentiation, the haematopoietic lineage development follows a strict hierarchical pattern programming emanating from a few HSCs. Both genetic and epigenetic modulations are involved in the regulation of haematopoietic lineage specification 16,17 . DNA organized in loose chromatin (euchromatin) is readily available for gene expression, while DNA tightly packed into dense chromatin (heterochromatin) becomes inaccessible to genetic reading and transcription. Chromatin remodelling is a prerequisite for eukaryotic gene transcription 18 , which relies on ATP-dependent remodelling complexes. These remodelling complexes are divided into four major subfamilies, including SWI/SNF, ISWI, CHD and INO80 subfamilies, based on a common SWI2/SNF2-related catalytic ATPase subunit 19,20 . The SNF2-related CBP activator protein (SRCAP)-contained remodelling complex, termed SRCAP complex, belongs to the INO80 subfamily. Eleven protein subunits, including SRCAP, ZNHIT1, Arp6, and YL-1, have been identified in the SRCAP complex 21 . The SRCAP complex can exchange histone H2A for the variant H2A.Z in the nucleosomes, rending accessible DNA for gene transcription 22 . H2A.Z is proposed to activate target gene transcription enhancing the promoters' accessibility of the target genes 23 . Moreover, in the haematopoietic system, increased H2A.Z serves as a chromatin signature during the differentiation of haematopoietic stem or progenitor cells 24 .
Here we show that Pcid2 is highly expressed in the BM and restricts lymphoid lineage specification. PCID2 binds to ZNHIT1 to block the SRCAP complex remodelling activity and prevents H2A.Z/H2A exchange of key lymphoid fate regulator genes in MPPs, leading to skewed lymphoid lineage commitment.
These observations were further validated in BM by immunofluorescence staining (Fig. 1k). Altogether, Pcid2 deficiency increases lymphoid cells but decreases myeloid cells.
Enhanced lymphoid differentiation of Pcid2-deficient MPPs. We showed above that Pcid2 deficiency caused increased CLPs and decreased CMPs. Their upstream progenitor MPPs were also markedly declined in Pcid2-deficient mice. Ki67 staining can discriminate cells in the G0 phase of the cell cycle from those in the cycling state 25 . We noticed that MPPs exhibited no obvious alteration of resting state (G0) vs. cycling state (S/G2/M) ratio between Pcid2 KO and WT littermate control mice (Fig. 2a). In addition, MPPs, CMPs and CLPs did not undergo apparent apoptosis in Pcid2 KO vs. WT control mice (Fig. 2b). These results suggest that the skewed ratio of the CLPs and CMPs in Pcid2 KO mice may be caused by lineage specification of the MPP progenitor.
To further confirm that Pcid2 deficiency induced lymphoid lineage commitment in vivo, we sorted LT-HSCs from the BM of Pcid2 −/− mice and transplanted them into lethally irradiated recipient mice (CD45.1). Donor-derived BM haematopoietic progenitor cells in recipient mice were analysed 8 weeks after transplantation. We rescued Pcid2 in Pcid2 −/− LT-HSCs via infection with Pcid2-overexpressing retrovirus. Pcid2 was restored in Pcid2 −/− LT-HSCs, which was comparable to that of WT LT-HSCs (Fig. 2g). We found that Pcid2-deficient HSC transplantation caused declined BM cellularity (Fig. 2h), as well as decreased numbers of MPPs and CMPs, but increased numbers of CLPs in the BM ( Fig. 2i and Supplementary Fig. 3A). Additionally, Pcid2-deficient HSC transplantation resulted in increased numbers of CD3 + T cells and NK1.1 + NK cells but decreased numbers of granulocytes and monocytes in peripheral blood ( Fig. 2j and Supplementary Fig. 3B). We observed that transplantation of Pcid2-rescued LT-HSCs could restore the biased lymphoid differentiation caused by Pcid2 deletion (Fig. 2h-j). Collectively, Pcid2 is involved in the regulation of branchpoint lineage commitment of MPPs.
PCID2 interacts with ZNHIT1 to block SRCAP complex activity. To further explore the molecular mechanism that Pcid2 regulated lineage specification, we screened a mouse BM cDNA library using Pcid2 as bait by a yeast two-hybrid approach. Of the 57 positive clones we screened out, 13 clones were identified to be ZNHIT1. ZNHIT1 was further validated to be a interacting  protein of PCID2 (Fig. 4a). ZNHIT1, also called as p18 hamlet , is a main regulatory component of the SRCAP chromatin remodelling complex. Their direct interaction was further confirmed by a glutathione S-transferase (GST) pulldown assay (Fig. 4b).
Pcid2 KO causes H2A.Z deposition to lymphoid fate genes. We next sought to explore how Pcid2 deficiency-mediated H2A.Z/ H2A exchange drove MPPs toward lymphoid lineage commitment. We sorted MPPs from Pcid2 −/− and Pcid2 +/+ mice and lysed them for chromatin immunoprecipitation (ChIP) assays with anti-H2A.Z and anti-H2A antibodies. We noticed that Pcid2 deletion enriched more H2A.Z deposition to the promoters of lymphoid fate regulator genes (Fig. 5a). In contrast, Pcid2 KO almost lost H2A deposition to these promoters of lymphoid fate regulator genes (Fig. 5b). Consistently, the promoters of lymphoid fate regulator genes in Pcid2 −/− MPPs bound substantial amounts of SRCAP (Fig. 5c). Of these lymphoid fate regulator genes, the Ets transcription factor PU.1 (encoded by Spi1 gene) is indispensable for the primary expression of lymphoid fate regulator transcripts for lymphocyte fate determination 27 . By contrast, GATA1 is a critical transcription factor for myelopoiesis 28 . Of note, PU.1 could associate with SRCAP by co-IP assay, while GATA1 did not interact with SRCAP in MPPs (Fig. 5d). Consistently, in Pcid2 −/− MPPs, PU.1 was enriched at the promoters of lymphoid fate regulator genes (Fig. 5e). By contrast, Pcid2 deletion did not cause H2A.Z, SRCAP and PU.1 deposition onto the promoters of myeloid lineage effector genes ( Supplementary  Fig. 5A). These data suggest that PU.1 behaves differently in the loci of lymphoid and myeloid lineage effector genes. Consequently, lymphoid fate regulator genes were highly expressed in Pcid2 −/− MPPs, whereas myeloid fate regulator genes were not significantly (P > 0.05, t-test) changed in Pcid2 −/− MPPs compared to Pcid2 +/+ MPPs (Fig. 5f).
In addition, Pcid2 deficiency augmented chromatin accessibility to DNase I digestion at the promoters of lymphoid fate regulator genes, such as Il7r and Ikzf1 (encoding Ikaros) (Fig. 5g). We next performed luciferase assays to confirm our observations. We found that Pcid2 depletion remarkably promoted Il7r transcription (Fig. 5h). However, depletion of Znhit1 or Srcap could dramatically suppress Il7r expression (Fig. 5h). Parallelly, an Ikzf1 promoter reporter luciferase assay obtained similar results (Fig. 5h). We next performed fragment mapping on the Il7r promoter-enhancer loci. We observed that PU.1, SRCAP and ZNHIT1 were co-occupied at the same site of Il7r promoterenhancer locus (−1400 to −1200 nt), whereas PCID2 did not bind to the same locus ( Supplementary Fig. 5B). The binding of ZNHIT1 to SRCAP prevented ZNHIT1 from binding to PCID2. Meanwhile, we used the myeloid differentiation gene Csf1r as an assay control. Csf1r harbours a PU.1-binding site on the promoter region 29 . We noticed that PU.1 was accumulated at the position of −200 to 0 nt of the Csf1r promoter ( Supplementary Fig. 5B). However, SRCAP and ZNHIT1 were not co-occupied at the same locus on Csf1r promoter-enhancer region ( Supplementary  Fig. 5B). Furthermore, depletion of Pcid2, Znhit1 or Srcap could not suppress Csf1r transcription (Supplementary Fig. 5C). These data suggest that Pcid2 deficiency causes H2A.Z and SRCAP deposition to lymphoid fate regulator genes in MPPs to drive lymphoid lineage specification.
Znhit1 deletion abrogates H2A/H2A.Z exchange. To explore the role of ZNHIT1 in the regulation of lymphoid lineage commitment, we deleted Znhit1 gene in HSCs. We generated Rosa26-LSL-Cas9 + ;Vav-Cre + mice by crossing B6;129-Gt(ROSA) 26Sor tm1(CAG-cas9*,−EGFP)Fezh /J knockin mice with Vav-Cre transgenic mice and sorted their HSCs for sgRNA against Znhit1 lentivirus-mediated genome editing 31 , followed by in vivo transplantation. After successfully reconstituting the recipient BM, we sorted MPPs from these mice for further studies. As expected, Znhit1 was completely deleted in MPPs (Fig. 6a). Of note, Znhit1 −/− MPPs abrogated H2A.Z deposition to lymphoid fate regulator genes (Fig. 6b). Consequently, Znhit1 −/− MPPs inhibited the lymphoid lineage differentiation but induced myeloid lineage differentiation via in vitro differentiation assays (Fig. 6c-e and Supplementary Fig. 6A-C). Finally, we next analysed donor-derived BM haematopoietic progenitor cells and peripheral blood cells in Znhit1 −/− HSCs reconstituted recipient mice. We observed that Znhit1 deficiency led to reduced BM cellularity and a decreased number of CLPs but an increased number of CMPs compared with WT mice (Fig. 6f, g). In parallel, Znhit1 deficiency caused a shifted myeloid lineage specification (Fig. 6h). Therefore, Znhit1 deficiency results in an opposite phenotype vs. Pcid2 KO mice. Collectively, Znhit1 deletion abrogates H2A/H2A.Z exchange in MPPs to disrupt lymphoid lineage commitment capacity.

Discussion
Lymphoid lineage commitment is an important process in haematopoiesis, which forms an immune system to protect a host from pathogen invasion. Within the BM, haematopoietic progenitors undergo lineage commitment towards a given lineage via the loss of ability to generate other lineages 32 . With lineage commitment signals, MPPs, differentiating from short-term HSCs (ST-HSC), give rise to two different progenitors CLPs and CMPs 1,33 , causing generation of mature lymphoid and myeloid cells in periphery, respectively. However, how MPPs switch into CLPs or CMPs has not been defined yet. In this study, we show that Pcid2 is highly expressed in the BM and MPPs. Pcid2 deletion in the haematopoietic system causes skewed lymphoid lineage specification. Pcid2-deficient HSCs prefer to differentiate into lymphoid cells but not into myeloid cells. In MPPs, PCID2 interacts with ZNHIT1 to block the SRCAP complex activity and prevents the H2A/H2A.Z exchange at nucleosomes of key lymphoid fate regulator genes. Pcid2 acts as one of the upstream inhibitors of PU.1 to impede the expression of key lymphoid fate regulator genes. Finally, Znhit1 deletion abrogates  Znhit1 deletion abrogates H2A/H2A.Z exchange to disrupt skewed lymphoid lineage commitment. a B6;129-Gt(ROSA)26Sor tm1(CAG-cas9*,−EGFP)Fezh /J knockin mice were crossed with Vav-Cre mice to obtain tissue-specific Cas9 expression. HSCs were sorted and infected with sgZnhit1-containing lentivirus and then transplanted into lethally irradiated recipient mice (CD45.1) for 8 weeks. ZNHIT1 was detected by immunoblotting. β-Actin were used as a loading control. b Indicated MPPs were sorted and lysed for ChIP assays with anti-H2A.Z antibody. Indicated promoters were examined by real-time qPCR. Signals were normalized to input DNA. c Indicated LMPPs were sorted and co-cultured with OP9-DL1 stromal cells as in Fig. 2c. Data are representative of six independent transfection and transplantation mice for each group. **P < 0.01. d Cells were sorted, infected and cultured as in c and used Flt3 ligand alone thereafter. After 4 days, cells were harvested and analysed by flow cytometry. n = 6 for each group. **P < 0.01. e Indicated MPPs were sorted and cocultured with OP9 stromal cells as in Fig. 2c. Data are representative of six independent transfection and transplantation mice for each group. **P < 0.01. f Paraffin sections from the femurs of Znhit1-deficient mice were stained with H&E. Scale bars, 20 μm. g Flow cytometric analysis of LSKs, MPPs, CLPs and CMPs from the BM of the indicated mice. (n = 5); **P < 0.01. h Peripheral blood cell numbers were analysed from the indicated transplanted recipient mice. Lymphocytes (Lym), monocytes (Mon), and granulocytes (Gran) were analysed. Results are shown as means ± S.D. Data are representative of six independent transfection and transplantation mice for each group. **P < 0.01. Student's t-test was used as statistical analysis H2A/H2A.Z exchange in MPPs to disrupt lymphoid lineage commitment capacity (Supplementary Fig. 6D). Upon sensing differentiation signals, CLPs generate all lymphoid lineages but not any myeloid cells 7,26 , whereas its counterpart CMPs produce all myeloid cell types 33,34 , supporting that the lymphoid and myeloid developmental programmes independently operate downstream of HSCs. Many reports showed that extrinsic cytokine signals can direct lineage-restricted capacities on HSCs. Erythropoietin (Epo) initiates erythroid lineage bias at all lineage bifurcations between HSCs and erythroid progenitors. We demonstrated that extrinsic insulin signalling drives lymphoid lineage commitment in early lymphopoiesis 35 .
Here we show that HSCs from Pcid2-deficient mice cause decreased MPPs and CMPs but increased CLPs in BM, causing skewed lymphoid lineage specification. The decreased numbers of MPPs and CMPs might be caused by dysregulated proliferation or cell death. Actually, we noticed that Pcid2-deficient MPPs and CMPs exhibit a similar cycling ratio to WT mice and have no apparent apoptosis. Through in vivo engraftment, we validate that Pcid2 acts as a cell intrinsic factor to restrict lymphoid lineage specification.
Pcid2 was originally identified as a partner to export mRNAs from the nucleus to cytoplasm 11 . PCID2 can interact with the repair factor BRCA2 of the human TREX-2 complex to prevent RNA-mediated genome instability 12 . Pcid2 also participates in the regulation of self-renewal of mouse ECSs 13 . We showed that PCID2 is in the CBP/p300-EID1 complex to sustain the pluripotency of human and mouse ESCs via regulation of EID1 stability 14 . In this study, we show that Pcid2 deficiency fails to impact the number of HSCs (Fig. 1g and Supplementary  Fig. 2B). These data suggest Pcid2 is not implicated in the regulation of self-renewal of HSCs. We demonstrate that PCID2 interacts with ZNHIT1, a master regulatory subunit of SRCAP complex, to replace ZNHIT1 out from the SRCAP complex of the nucleosomes of lymphoid fate regulator genes in MPPs. Replacement of ZNHIT1 by PCID2 causes inactivation of ATPase activity of the SRCAP complex to block H2A/H2A.Z exchange, causing suppression of lymphoid fate regulator gene expression.
Epigenetic modulations are involved in a variety of biological processes, including gene transcription, DNA replication, DNA repair and DNA recombination 36 . These processes include posttranslational modulations of histones, DNA methylation, incorporation of histone variants and nucleosome remodelling activity. Of note, the nucleosome remodelling and incorporation of histone variants largely rely on assistance of ATP-dependent chromatin remodelling complexes 37 . The chromatin remodelling complexes have been reported to participate in the maintenance of HSC pluripotency 38 . BAF53a, a subunit of the SWI/SNF-like BAF complex, is implicated in the survival and maintenance of HSCs 18 . SNF2-like ATPase mi-2β subunit of the nucleosome remodelling deactylase (NuRD) complex is required for the maintenance of HSCs 39 . We previously showed that WASH protein associates with the nucleosome remodelling factor (NURF) complex to trigger c-Myc expression, which directs the differentiation commitment of HSCs 40 . However, how the chromatin remodelling complexes regulate lymphoid and myeloid branchpoint lineage specification still remains elusive. Here we show that Pcid2 deficiency causes the SRCAP complex enrichment at the lymphoid fate regulator genes to drive H2A.Z exchange for these gene expression, resulting in shifted lymphoid lineage differentiation. ZNHIT1, also called as p18 Hamlet , is a master regulatory subunit of the SRCAP complex 41 . The SRCAP complex harbors ATPase activity to supply ATP for exchange of H2A to the variant H2A.Z at nucleosomes [42][43][44] , ensuring the loose chromatin structure necessary for transcriptional activation. Mutations of SRCAP are related to the pathogenesis of Floating-Harbor syndrome 45 , suggesting the SRCAP complex has a critical function in physiological and pathological processes. The SRCAP complex causes H2A.Z exchange at the muscle-specific gene promoters to direct muscle differentiation 21 . Here we show that Znhit1 deletion in the haematopoietic system abrogates H2A.Z exchange at lymphoid fate regulator gene promoters to impede the shifted lymphoid lineage commitment.
Accumulating evidence supports that combinations of transcription factors coordinately and sequentially modulate lymphopoiesis. For example, PU.1 (encoded by Spi1 gene) and Ikaros (encoded by Ikzf1 gene) are expressed in earlier haematopoietic progenitors that take part in the regulation of lymphoid lineage commitment 46,47 . Neonatal Ikzf1-null mice display complete defect in fetal thymocyte development 33 . Adult Ikzf1-deficient mice impair lymphoid differentiation and exhibit thymocyte development shifted to CD4 T cells 48 . PU.1 is one of the most important regulators of lymphoid and granulocyte/monocyte (GM) lineage differentiation 49 . PU.1 is also involved in the generation of early GM and lymphoid progenitors such as CMPs, GMPs and CLPs 27 . Interestingly, PU.1 + MPPs exhibit granulocyte/monocyte/lymphoidrestricted progenitor activity without megakaryocyte/erythroid differentiation capacity 28 , while GATA-1 + MPPs display potent myeloerythroid potential without producing lymphocytes. Furthermore, PU.1 and GATA-1 mutually suppress each other's expression and transactivation functions 50,51 , suggesting that the reciprocal activation of PU.1 and GATA-1 primarily organizes the haematopoietic lineage fate decision to generate the earliest lymphoid or myeloid haematopoietic progenitors.
Besides the antagonistic relationship with GATA-1, PU.1 acts also by its expression levels. High levels of PU.1 promote myeloid development, while low levels drive B-cell development 52 . In this study, we show that Pcid2 acts as an upstream regulator of PU.1 to suppress its activity in MPPs, resulting in restrict lymphoid lineage specification. In the presence of PCID2, it binds to ZNHIT1 to block the SRCAP complex assembly causing prevention of H2A/H2A.Z exchange on Spi1 promoter, which deposits H2A to suppress its expression. In the absence of PCID2, freed ZNHIT1 assembles the SRCAP complex to enrich H2A.Z onto the Spi1 promoter, which initiates its expression to drive lymphoid lineage commitment. Our results demonstrate that PU.1 expression in MPPs promote lymphoid lineage specification via Pcid2-mediated H2A/H2A.Z exchange on Spi1 promoter. Whether differing concentrations of PU.1 affects MPPs to lymphoid lineage specification remains to be further studied.
It has been reported that other factors such as Gfi-1 and Egr proteins are also involved in the modulation of myeloid and lymphoid lineage decisions. Gfi-1 is implicated in the regulation of lymphocyte development and activation 53 . Gfi-1 deficiency causes defects in B-and T-cell differentiation 53 . Moreover, Gfi-1 also takes part in regulating homeostasis of HSCs by coordinating proliferation and migration 54 . Gfi-1 also directly associates with and suppresses transactivation by PU.1 in favouring granulopoiesis 55 . In addition, PU.1 also induces the expression of Egr proteins. Egr-1 and Egr-2 are zinc-finger transcription factors that are able to activate or repress transcription. Gene activations induced by Egr-1 or Egr-2 apparently promotes monocytic maturation 55 . However, how these factors tightly coordinate in the regulation of myeloid and lymphoid lineage decisions still need to be further investigated.
Here we show that DKO Pcid2 and Spi1 genes display a similar ratio of lymphoid vs. myeloid progenitors and mature cells in comparison with Spi1 −/− mice. Pcid2 acts as an upstream regulator of PU.1 to suppress its activity in MPPs, causing restrict lymphoid lineage specification. In addition, PCID2 binds to ZNHIT1 to impair the SRCAP complex activity that suppresses the expression of lymphoid fate regulator genes. Therefore, we conclude that Pcid2 has a critical function in the earliest haematopoietic branchpoint specification that directs the myeloid and lymphoid progenitor populations.

Methods
Cell culture. Human 293T cells (ATCC, CRL-3216) were cultured with Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS) and 100 U/ml penicillin and 100 mg/ml streptomycin. Retrovirus and lentivirus infecting primary BM cells were produced in 293T cells by using the standard protocols. Transfection was performed using lipofectin reagent (Invitrogen). For in vitro differentiation, MPPs were plated on monolayer of OP9-DL1 in α-Minimum Essential Medium containing 20% FBS, 100 μg/ml streptomycin and 100 U/ml penicillin and supplemented with 1 ng/ml rmIL-7 and 5 ng/ml rmFlt3L. MPPs were also sorted and cultured on OP9 feeder cells (ATCC, CRL-2749) supplemented with recombinant mouse SCF (10 ng/ml), recombinant mouse Flt3-ligand (20 ng/ml) and recombinant mouse IL-7 (1 ng/ml). Cell lines were obtained from ATCC and authenticated by PCR. Mycoplasma contamination had been tested by PCR and excluded.
Littermates with the same age and gender for each group were used. We excluded the mice 5 g thinner than other littermates before any treatment or analysis. Mice were bred in specific pathogen-free animal facility. The euthanasia method was cervical dislocation when needed.
Histology. Mouse thymus and cervical lymph nodes were fixed in 4% paraformaldehyde (PFA) for 12 h. Fixed tissues were washed for twice using 75% ethanol and embedded in paraffin, followed by sectioning and staining with haematoxylin and eosin according to standard laboratory procedures. For BM histology analysis, femurs were fixed in 4% formaldehyde followed by decalcifying in 10% EDTA-PBS (phosphate-buffered saline) buffer. Longitudinal paraffin sections were prepared for haematoxylin and eosin staining. Mouse femurs were fixed in PFA-lysine-periodate buffer for 12 h, then decalcified in decalcifying buffer (10% EDTA in PBS (w/v), pH 7.4) for 48 h and then rehydrated in 30% sucrose solution for 48 h before snap frozen in OCT (TissueTek). Cytosections were obtained according to standard procedures 57 .
Analysis of peripheral blood cells. Mice were anesthetized and peripheral blood was collected from the inferior vena cava with EDTA as anticoagulant. Blood was analysed by XFA6030 automated hemacytometer (Slpoo). Cell numbers and percentages of each population were counted. For morphological analysis, peripheral blood smears were stained by Wright's. Images were taken in oil-immersion 100× lens.
In vitro differentiation assay. In vitro lymphoid differentiation assay was described previously 26 . LMPPs were sorted and cultured for 14 days on OP9-DL1 feeder cells supplemented with recombinant mouse Flt3-ligand (5 ng/ml), and recombinant mouse IL-7 (1 ng/ml). Then cells were collected for staining with anti-CD44 and anti-CD25 antibodies. For mature T-cell differentiation, LMPPs were sorted and cultured for 14 days on OP9-DL1 feeder cells supplemented with recombinant mouse IL-7 (1 ng/ml) and recombinant mouse Flt3-ligand (5 ng/ml), followed by co-culturing with OP9-DL1 using Flt3 ligand (5 ng/ml) alone for 4 days. Cells were then harvested and stained with anti-CD4 and anti-CD8 antibodies. MPPs were also sorted and cultured on OP9 feeder cells supplemented with recombinant mouse SCF (10 ng/ml), recombinant mouse Flt3-ligand (20 ng/ml) and recombinant mouse IL-7 (1 ng/ml) for 10 days. Then cells were collected for staining with anti-CD19 and anti-Mac1 antibodies. For in vitro myeloid differentiation, MPPs were sorted from BM and cultured in methocult medium (Stemcells) for CFC formation assay. CFU-GEMM, CFU-M, CFU-G, CFU-GM and BFU-E colonies were counted, respectively, according to the manufacturer's instruction.
Immunofluorescence assay. Cells were placed on 0.01% poly-L-lysine-treated coverslips and fixed with 4% PFA for 20 min at room temperature, followed by 0.5% Triton X-100 permeabilization and 5% donkey serum blocking. Cells were then incubated with appropriate primary antibodies at 4°C overnight followed by incubation with corresponding fluorescence-conjugated secondary antibodies. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). Images were obtained with Olympus FV1200 laser scanning confocal microscopy (Olympus, Japan). Confocal sections were taken every 1 μm through the whole cell for z-stack projection. The software ImageJ was used for co-localization quantitation. For each experiment, at least 50 typical cells were observed. For BM histology analysis, mouse femur cytosections were blocked with 10% donkey serum, followed by incubation with appropriate primary antibodies and corresponding fluorescenceconjugated secondary antibodies. Nuclei were stained with DAPI. Cytosections were washed and dehydrated in gradient EtOH followed by rinsing in xylene and coversliping. For thymus histology analysis, thymus was fixed in 4% formaldehyde followed by paraffin embedding, sectioning and antibody staining. Images were obtained with Zeiss Axio scan.Z1 scanning system (Zeiss, Germany).