The endothelial antigen ESAM marks primitive hematopoietic progenitors throughout life in mice

Hematopoietic stem cells (HSCs) are defined as cells with the capacity for self-renewal as well as differentiation into multilineage blood cells, maintaining the immune system throughout life. A large body of information exists about molecular mechanisms involved in maintaining their integrity, and many studies have attempted to identify unique markers associated with these extremely rare cells. In bone marrow of adult mice, the Lin⁻ c-kit^Hi Sca1⁺ CD34^−/Lo Thy1.1^Lo subset is known to include HSCs with long-term repopulating capacity.¹  However, several of these parameters differ between strains of mice, change dramatically during developmental age or inflammation, and are expressed on many non-HSCs.^2-4  The recent identification of CD150/SLAM as stable markers made it possible to increase the purity of HSCs even in aged mice or cyclophosphamide/granulocyte colony-stimulating factor-treated mice with mobilized progenitors.⁵  However, even the most highly purified HSCs are heterogeneous, and it may eventually be possible to associate discrete functions or activity states with subpopulations. Additional authentic HSC markers could have utility in attempts to rescue hematopoietic disorders using hematopoietic progenitors obtained from reprogrammed adult tissues.^6,7 

HSCs are thought to arise initially from hemogenic endothelium, which can produce hematopoietic cells as well as endothelial cells. Therefore, it is not surprising that HSCs share some endothelial properties at early developmental stages.^8,9  For example, the CD34 sialomucin and Tie2, an angiopoietin receptor, are expressed on HSCs in E10–11 embryos.^10,11  Endoglin and vascular-endothelial cadherin are additional endothelial markers found on fetal HSCs.^8,12  However, the expression of many of these antigens declines on HSCs at later stages of development.^3,4,13  It is interesting that the expression of CD34 is restored when adult HSCs are driven into cycle by 5-fluorouracil or granulocyte colony-stimulating factor administration.^14,15  CD11b/Mac-1 is an adhesion molecule that is similarly dependent on developmental age and activation status.¹⁶  In contrast to these patterns, endomucin is a CD34-like sialomucin that marks HSCs from E10 and throughout subsequent development.¹⁷  Each of these advances offered the promise of learning more about how HSCs arise de novo and function throughout life.

We previously determined that the most primitive cells with lymphopoietic potential first develop in the para-aortic splanchnopleura (PSp)/aorta-gonad-mesonephros (AGM) region of embryos, and we tracked expression of the Rag1 lymphoid gene.^18,19  To extend those findings, we searched for genes that might be differentially expressed at the very earliest stages of lymphopoiesis. We here show that endothelial cell-selective adhesion molecule (ESAM) is a durable and effective marker of HSCs. Indeed, ESAM was expressed throughout life and could be used as a gating parameter for sorting long-term repopulating HSCs.

Rag1/GFP knockin mice (CD45.2 alloantigen) were described.^20,21  Mice of the corresponding wild-type (WT) C57BL/6 were obtained from Japan Clea (Shizuoka, Japan). Mating homozygous male Rag1/GFP knockin mice with WT C57BL/6 female mice generated heterozygous Rag1/GFP knockin fetuses. The day of vaginal plug observation was considered as day 0.5 postcoitum (E0.5). In some experiments, we purchased pregnant C57BL/6 mice from Japan Clea and used their fetuses. The congenic C57BL/6 strain (C57BL/6SJL; CD45.1 alloantigen) was purchased from The Jackson Laboratory (Bar Harbor, ME) and used in transplantation experiments. The experimental designs of this study were approved by the committee of Osaka University for animal studies.

Phycoerythrin (PE)–conjugated anti-Sca1 (Ly6A/E; D7), CD48 (HM48-1), CD11b/Mac-1 (M1/70), Gr-1 (RB6-8C5), CD19 (1D3), CD4 (L3T4), and CD8a (53-6.7) monoclonal antibodies (mAbs), biotinylated anti-CD45.2(104) mAb, allophycocyanin (APC)–conjugated anti-CD11b/Mac-1 (M1/70) and c-kit (2B8) mAbs, and PE-Texas red tandem-conjugated (PE-TR) streptavidin were purchased from BD Biosciences PharMingen (San Diego, CA). PE-conjugated anti-CD34 (RAM34), CD31/PECAM-1(390), CD105/Endoglin (MJ7/18), and Tie2 (TEK4) mAbs, PE-Cy7–conjugated anti-Sca1 (Ly6A/E; D7) mAb, and APC-conjugated anti CD45.1 (A20) mAbs were purchased from eBioscience (San Diego, CA). A rat anti–mouse ESAM mAb (1G8), a rabbit anti–mouse ESAM polyclonal Ab (VE19), and a rabbit preimmune IgG were prepared in our hands.²²  A fluorescein isothiocyanate (FITC)–conjugated goat anti–rat IgG (H+L) Ab purchased from Southern Biotechnology (Birmingham, AL), a PE-conjugated goat antirat Ig Ab purchased from BD Biosciences PharMingen, or AlexaFluor 488 goat anti–rabbit IgG (H+L) Ab purchased from Invitrogen (Carlsbad, CA) was used as a second Ab for the anti-ESAM Abs. A PE-conjugated hamster IgG1 was purchased from BD Biosciences PharMingen and used as a control for a PE-conjugated anti-CD48. FITC-conjugated anti-CD11b/Mac-1 (M1/70), Gr-1 (RB6–8C5), TER-119, CD45R/B220 (RA3-6B2), and CD3e (145-2C11) were purchased from BD Biosciences PharMingen and used as lineage markers in adult studies.

Fetal liver or cells obtained from adult femurs and tibias of heterozygous Rag1/GFP knockin mice were harvested and subjected to cell sorting as previously described.¹⁸  In the first step, Rag1⁻ or Rag1^Lo cells were sorted according to levels of green fluorescent protein (GFP) expression. Background auto-fluorescence was discriminated from authentic GFP by collecting data in 2 fluorescence channels without compensation. Under these conditions, even extremely low levels of fluorescence specific to GFP knockin mice were obvious on 2-color diagonal plots. The purity of the sorted cells in the first step was more than 95%. The sorted cells were incubated with anti-FcR (2.4G2) before staining with PE-anti-Sca1 and APC-anti-c-kit antibodies, suspended in 7-amino-actinomycin D (7-AAD)–containing buffer, and subjected to a second round of sorting. Dead cells stained with 7-AAD and a few contaminating cells with inappropriate GFP levels were gated out, and the cells were then fractionated according to Sca1 and c-kit to obtain Rag1⁻ c-kit^Hi Sca1⁺ HSCs and Rag1^Lo ckit^Hi Sca1⁺, early lymphoid progenitors (ELPs). The cell sorting was performed with FACSaria using the FACSDiva program (BD Biosciences, San Jose, CA). The sorting gates used to isolate HSCs and ELPs from E14.5 fetal liver for gene array experiments (“Gene arrays”) are shown in Figure S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article).

In the sorting experiments with ESAM expression, Rag1/GFP⁻ cells were first sorted from fetal tissues or adult bone marrow. Then the cells were stained with rat anti–mouse ESAM mAb (1G8) followed by FITC-goat anti–rat IgG or PE-goat anti–rat Ig, respectively. After the staining for ESAM, E14.5 fetal liver Rag1/GFP⁻ cells were stained with PE–anti-Sca1 and APC-anti–c-kit antibodies. For PSp/AGM or yolk sac (YS) cells of E9.5-10.5 embryos, a PE-anti–Tie2 Ab was used instead of a PE-anti–Sca1 Ab. The adult marrow Rag1/GFP⁻ cells were stained with FITC-anti-Lin (Mac-1, Gr-1, TER119, CD45R/B220, CD3e), PE-Cy7–anti-Sca1, and APC-anti–c-kit Abs. Then the cells were suspended in 7-AAD–containing buffer and subjected to a second round of sorting.

Total RNAs were isolated from the Rag1⁻ ckit^Hi Sca1⁺, HSC-enriched fraction and the Rag1^Lo ckit^Hi Sca1⁺, ELP-enriched fraction, both of which were isolated from mouse E14.5 fetal liver. From the aliquot of each RNA (24 ng), biotin-labeled cRNA was prepared using GeneChip 2-Cycle Target Labeling and Control Reagents (Affymetrix, Santa Clara, CA) according to the manual. The cRNA was fragmented and hybridized to Mouse Genome 430 2.0 Array (Affymetrix) using a GeneChip Hybridization, Wash, and Stain Kit (Affymetrix). Fluorescent signals were scanned by GeneChip Scanner 3000, and the data were analyzed by GeneSpring GX software (Agilent Technologies, Palo Alto, CA). Bacillus subtilis genes on the arrays were used as negative controls for background subtraction. All values of the genes in each array were divided by the value of glyceraldehyde-3-phosphate dehydrogenase (Gapdh) gene. The Cross-Gene Error Model was used to estimate measurement precision by combining variability of gene expression data. The experiments were duplicated to confirm reproducibility of the data. All microarray data have been deposited with CIBEX, National Institute of Genetics DDBJ (DNA Data Bank of Japan) under the accession number CBX73.

Rag1/GFP⁻ cells of E14.5 fetal liver were incubated with rat anti–mouse ESAM mAb (1G8) followed by FITC-goat anti–rat IgG. Then the cells were incubated with a rat anti–mouse FcRII/III (2.4G2) and subsequently stained with PE-Cy7–anti-Sca1 and APC-anti–c-kit in combination with PE-anti–CD34, CD31/PECAM1, CD105/Endoglin, or Tie2 mAbs. The cells were suspended in 7-AAD–containing buffer and subjected to flow cytometry analyses, performed with FACSaria using the FACSDiva program (BD Biosciences).

In the other experiments, cultured cells or recovered cells from transplanted mice were incubated with anti-FcR and then stained with PE-, APC-conjugated mAbs and/or a biotinylated Ab followed by PE-TR-streptavidin as indicated in each figure. The flow cytometry analyses were performed with FACScalibur using the Cellquest program (BD Biosciences). The data analyses were done with FlowJo software (TreeStar, Ashland, OR).

A total of 500 or 100 cells of each sorted fraction (Rag1/GFP⁻ ckit^Hi Sca1⁺ ESAM1^−/Lo or Rag1/GFP⁻ ckit^Hi Sca1⁺ ESAM1^Hi of fetal liver or adult marrow, respectively) were cultured in Iscove modified Dulbecco medium-based methylcellulose medium supplemented with 50 ng/mL of recombinant mouse (rm) stem cell factor (SCF), 10 ng/mL of rm interleukin-3 (IL-3), 10 ng/mL of recombinant human IL-6, and 3 U/mL of recombinant human erythropoietin (Methocult GF 3434; StemCell Technologies, Vancouver, BC). After 8 to 10 days, colonies were enumerated and classified as granulocyte colony-forming units (CFU-G); macrophage colony-forming units (CFU-M); granulocyte-macrophage colony-forming units (CFU-GM); erythroid burst-forming units (BFU-E); or mixed erythroid-myeloid colony-forming units (CFU-Mix) according to shape and color under an inverted microscope. In some experiments, colony types were morphologically certified by examination of cytospin preparations after May-Grünwald/Giemsa staining.

The murine bone marrow stromal cell line MS-5 was generously provided by Dr J. Mori (Niigata University, Niigata, Japan). Nonirradiated MS-5 stromal cells were prepared at a concentration of 5 × 10⁴ cells/well in 24-well tissue plates 1 day before the seeding of sorted cells at a concentration of 50 cells/mL. Cells were cultured in α-minimum essential medium (Invitrogen) supplemented with 10% fetal calf serum (FCS), rm SCF (10 ng/mL), rm Flt3-ligand (20 ng/mL), and rm IL-7 (1 ng/mL). The cultures were fed every 4 days by removing half of the medium and replacing it with fresh medium, and maintained for 6 or 10 days. Cytokines were freshly added with each feeding.

The frequencies of lymphohematopoietic progenitors in the fetal liver–derived Rag1/GFP⁻ ckit^Hi Sca1⁺ ESAM^−/Lo or Rag1/GFP⁻ ckit^Hi Sca1⁺ ESAM^Hi fraction were determined by plating the sorted cells in limiting dilution assays using 96-well flat bottom plates. Pre-established MS-5 stromal cell layers were plated with 1, 2, 4, 8, or 50 cells each using the Automated Cell Deposition Unit of the FACSaria (BD Biosciences). Cells were cultured in α-minimum essential medium supplemented with 10% FCS, rm SCF (10 ng/mL), rm Flt3-ligand (20 ng/mL), and rm IL-7 (1 ng/mL). At 10 days of culture, wells were inspected for the presence of hematopoietic clones. Positive wells were harvested, stained, and analyzed by flow cytometry for the presence of CD19⁺ Gr-1⁻ B lineage cells. The frequencies of progenitors were calculated by linear regression analysis on the basis of Poisson distribution as the reciprocal of the concentration of test cells that gave 37% negative cultures.

The Ly5 system was adapted to a competitive repopulation assay. A total of 1000 Rag1/GFP⁻ c-kit^Hi Sca1⁺ ESAM^−/Lo or Rag1/GFP⁻ c-kit^Hi Sca1⁺ ESAM^Hi cells sorted from fetal liver of E14.5 Rag1/GFP heterozygous embryos (CD45.2) were mixed with 2 × 10⁵ unfractionated adult bone marrow cells obtained from WT C57BL/6-Ly5.1 (CD45.1) mice and were transplanted into C57BL/6-Ly5.1 mice irradiated at a dose of 10 Gy. At 5 weeks after transplantation, peripheral blood cells of the recipients were colleted by retro-orbital bleeding and were stained with APC-conjugated anti-CD45.1 and biotinylated anti-CD45.2 Abs followed by PE-TR streptavidin. The cells were simultaneously stained with PE-conjugated anti–Mac-1 and Gr-1 Abs, or PE-conjugated anti-CD19 Ab, or PE-conjugated anti-CD4 and CD8a Abs. Twenty weeks after transplantation, all recipients were killed, and reconstitution of CD45.2⁺ myeloid, B, or T cells was confirmed by flow cytometry in the bone marrow, spleen, and thymus, respectively. For the second transplantation, bone marrow cells of primary recipients with CD45.2 engraftment were transferred into 10 Gy-irradiated C57BL/6-Ly5.1 mice (10⁶ whole bone marrow cells per recipient). After 20 weeks, contribution of CD45.2 cells to the hematopoietic reconstitution was evaluated in the second recipients.

To examine the phenotype or function of hematopoietic progenitors in the PSp/AGM and the YS of early embryos, cells were prepared as previously described.¹⁹  Briefly, tissues were dissociated by incubation with dispase II (Roche Diagnostics, Mannheim, Germany) for 20 minutes at 37°C and cell dissociation buffer (Invitrogen) for 20 minutes at 37°C followed by vigorous pipetting. The cells were then suspended in phosphate-buffered saline containing 3% FCS and stained with the Abs indicated in each experiment.

Statistical analysis was carried out by standard Student t tests. Error bars used throughout indicate SD of the mean.

Microarrays were performed as part of an effort to characterize the initial transition of fetal HSCs to primitive lymphopoietic cells. The comparisons involved mRNA from Rag1^Lo ckit^Hi Sca1⁺, ELPs, and the HSC-enriched Rag1⁻ ckit^Hi Sca1⁺ fraction isolated from E14.5 fetal liver (Figure S1).

Consistent with previous analyses,^12,23,24 Evi1, CD41, and Endoglin were highly expressed by fetal HSCs (Figure 1). Esam transcripts were conspicuous in the HSC fraction and attracted attention because of a sharp down-regulation on differentiation to ELP (Figure 1). Real-time PCR using Esam specific primers verified the results of these microarrays (data not shown).

The availability of a rat antimouse ESAM mAb (clone 1G8) facilitated characterization of mononuclear cells obtained from E14.5 fetal liver. A majority of ESAM⁺ cells were found in the c-kit^Hi fraction (Figure 2A left). ESAM expression also correlated with Sca1 (Figure 2A middle). These observations suggested that ESAM might substitute for at least one of the 2 most widely used HSC markers. Indeed, cells in the Sca1^Hi ESAM^Hi gate were almost homogeneous with respect to high c-kit expression (Figure 2A right).

The c-kit^Hi Sca1⁺ fraction, known to include all conventional HSCs, was divided into 2 categories according to ESAM staining (Figure 2B left and middle). The subpopulation with the highest density of ESAM was enriched for c-kit^Hi Sca1^Hi cells, whereas ones with negative or low levels of ESAM were found in the c-kit^Hi Sca1^Lo subset (Figure 2B right). The latter are thought to be enriched with respect to committed myelo-erythroid progenitors.²⁵  Similar results were obtained with a rabbit antimouse ESAM polyclonal Ab (VE19) (data not shown).

When E14.5 fetal liver cells of WT C57B6 mice and traditional markers of fetal liver HSCs were used instead of the Rag1/GFP knockin mice, essentially identical results were obtained (Figure 2C). This analysis was performed with CD48 expression, one of the SLAM family members recently shown to conversely relate with HSC activity in fetal liver.²⁶  Lin⁻ c-kit^Hi Sca1⁺ cells could be divided into ESAM⁻ and ESAM⁺ fractions, and a subpopulation with higher Sca1 expression was more enriched with ESAM⁺ cells (Figure 2C middle and right). CD48 expression tended to increase along with down-regulation of ESAM. Indeed, the Lin⁻ c-kit^Hi Sca1⁺ fraction was found to consist of 2 major subpopulations, CD48^Hi ESAM⁻ and CD48^−/Lo ESAM⁺. Thus, ESAM is conspicuously expressed on immature hematopoietic cells in fetal liver and seems to conversely relate to lineage progression.

Next we evaluated how ESAM corresponds to other endothelial antigens that were previously identified on fetal hematopoietic progenitors. CD34 and CD31/PECAM1 were uniformly present on Rag1⁻ c-kit^Hi Sca1⁺ cells in E14.5 fetal liver (Figure 3A,B). Neither could resolve the Rag1⁻ c-kit^Hi Sca1⁺ cells into ESAM^Hi and ESAM^−/Lo fractions, and even showed a slightly inverse relationship with ESAM expression on early progenitors. Expression profiles of Endoglin and Tie2 did correlate with ESAM (Figure 3C,D). However, although the primitive ESAM^Hi fraction uniformly expressed high levels of Endoglin and Tie2, many of the more differentiated ESAM^−/Lo cells still retained the 2 markers. These results suggest that ESAM might be a more useful marker of HSCs than other endothelial antigens and could represent an important tool for fetal stem cell studies.

To evaluate whether ESAM expression can enrich primitive progenitors in fetuses, we compared the clonogenic potential in methylcellulose cultures of the ESAM^−/Lo and ESAM^Hi cells in the Rag1⁻ ckit^Hi Sca1⁺ HSC fraction of E14.5 fetal liver. Cells in the ESAM^Hi fraction formed more colonies with larger size than those in the ESAM^−/Lo fraction (Figure 4). In particular, most CFU-Mix, primitive progenitors with both myeloid and erythroid potential, were found in the ESAM^Hi fraction (Figure 4A).

Next we analyzed the lymphopoietic potential of cells resolved on the basis of ESAM in cocultures with the MS5 bone marrow stromal cell line. The culture media contained SCF, Flt3-ligand, and IL-7, factors that effectively generate CD19⁺ B lymphoid cells as well as Mac1⁺ myeloid cells.¹⁸  When 500 cells were cultured in individual wells of 6-well plates, both ESAM^−/Lo and ESAM^Hi fractions produced CD19⁺ cells and Mac1⁺ cells (Figure 5A). However, we observed that most of the hematopoietic colonies from ESAM^Hi cells grew beneath the MS5 stromal cell layer, although this was not the case with ESAM^−/Lo cells (data not shown). Moreover, the lymphohematopoietic cells from ESAM^Hi cells continued to expand explosively after 6 days of coculture and gave rise to approximately 50 000 cells per input progenitor by day 10 (Figure 5B).

Authentic HSCs are characterized as having both lymphoid and myeloid potential.^8,27  To compare numbers of primitive lymphohematopoietic progenitors in the ESAM^Hi and ESAM^−/Lo fractions, we performed in vitro limiting dilution assays in MS5 cocultures. One in 2.1 ESAM^Hi cells and 1 in 3.5 ESAM^−/Lo cells gave rise to blood cells, indicating that both fractions are extremely potent sources of hematopoietic progenitors (Figure 5C left). However, we observed drastic differences between the 2 fractions regarding the frequencies of primitive progenitors with lymphopoietic potential. Whereas 1 in 8 ESAM^Hi cells produced CD19⁺ B lineage cells, only 1 in 125 ESAM^−/Lo cells were lymphopoietic under these conditions (Figure 5C right). These results suggest that primitive stem/progenitor cells, which are multipotent for myeloid, erythroid, and lymphoid lineages, are present in the ESAM^Hi fraction of fetal liver.

To evaluate ESAM expression on long-term reconstituting HSCs in E14.5 fetal liver, we transplanted cells of the ESAM^−/Lo or ESAM^Hi fraction into lethally irradiated mice (Figure 6A). Five weeks after transplantation, it was obvious that CD45.2⁺ ESAM^Hi cells contributed highly to the recovery of hematopoiesis in recipients, but no chimerism was detected in mice transplanted with ESAM^−/Lo cells (Figure 6B). Indeed, although 9 of 11 mice transplanted with 1000 ESAM^Hi cells had clear donor CD45.2⁺ populations (> 1.0%) among peripheral leukocytes, none of 10 mice given 1000 ESAM^−/Lo cells had evidence of chimerism. Five months after transplantation, we analyzed the contribution of CD45.2⁺ cells to long-term lymphohematopoietic reconstitution of the recipients. Although most of the 11 mice transplanted with ESAM^Hi cells had clear CD45.2⁺ populations in bone marrow (mean ± SD of percentage CD45.2⁺ CD45.1⁻ in total CD45⁺ cells; 16.3% ± 22.0%), none of the ESAM^−/Lo-transplanted mice contained detectable CD45.2⁺ cells in that organ (mean ± SD of percentage CD45.2⁺ CD45.1⁻ in total CD45⁺ cells; 0.02% ± 0.03%) (Figure 6C). Furthermore, multilineage recovery was observed in the bone marrow, spleen, and thymus of mice transplanted with ESAM^Hi cells (Figure 6D). In addition, bone marrow cells recovered from primary recipients with ESAM^Hi cell transplantation effectively reconstituted CD45.2⁺ lymphohematopoietic cells in secondary recipients at 5 months after transplantation (data not shown). These results indicate that high levels of ESAM expression correspond to fetal HSCs with long-term repopulating potential.

The definitive HSCs that account for lymphohematopoieisis in adults first arise in the AGM region.^28,29  To examine whether ESAM was present on those HSCs, flow cytometry analyses were performed with the Rag1/GFP⁻ cells isolated from E10.5 embryos. Tie2 and c-kit were used as additional parameters because we previously identified lymphohematopoietic cells in the early embryos as Rag1⁻ Tie2⁺ c-kit⁺.¹⁹  A small but conspicuous Tie2^Hi population was detected in E10.5 AGM cells (Figure 7A left). Interestingly, the Tie2^Hi AGM cells were clearly divided into 2 discrete populations according to ESAM expression (Figure 7A middle). When the 2 populations were back-plotted on the Tie2 and c-kit profile, ESAM⁺ cells were exclusively c-kit⁺ (Figure 7A right). In addition, lymphohematopoietic potential in the MS5 cocultures was highly enriched in the ESAM⁺ fraction among the Tie2^Hi cells. That is, the Tie2^Hi ESAM⁺ fraction could effectively produce both CD19⁺ lymphoid cells and Mac1⁺ myeloid cells, although the Tie2^Hi ESAM⁻ fraction generated only a small number of Mac1⁺ cells (Figure 7B). We concluded from these analyses that ESAM marks the primitive hematopoietic cells endowed with lymphopoietic activity in the E10.5 AGM as well as in the E14.5 liver.

A small number of Tie2^Hi c-kit^Lo ESAM^Hi cells were also observed among extraembryonic YS cells (Figure 7C). In addition, the E10.5 YS contained a conspicuous population whose phenotype was Tie2^Lo c-kit^Hi, and cells in that fraction expressed low levels of ESAM (Figure 7C). Cells with the same phenotype were also clearly observed in the E9.5 YS, although they were absent in the caudal half of the embryo proper (Figure S2A). These Tie2^Lo c-kit^Hi ESAM^Lo cells in the YS effectively produced myeloid and/or erythroid colonies in methylcellulose culture but showed little lymphopoietic potential (Figure S2B,C). Importantly, lymphopoietic activity was exclusive to the Tie2^Hi c-kit^Lo ESAM^Hi fraction of the PSp/AGM region, which showed no myeloid-erythroid potential in conventional methylcellulose assays.

It is well known that HSC properties change with developmental age. Indeed, Forsberg et al previously reported ESAM expression by adult marrow HSCs,³⁰  but the levels were weak compared with those on fetal HSCs. Therefore, we expected that ESAM, like the other endothelial antigens, would decline with HSC aging. However, we found that ESAM expression was detectable on the Rag1⁻ LSK fraction of bone marrow through life (Figure 7D). Indeed, the proportion of ESAM^Hi cells and the mean fluorescence intensity of ESAM expression in the Rag1⁻ LSK fraction increased with age (Figure 7D). Moreover, primitive myeloid-erythroid progenitors from adult bone marrow were also enriched in the ESAM^Hi fraction (Figure 7E). In summary, these findings demonstrate that ESAM expression is useful for exploring the biology of hematopoietic stem/progenitor cells throughout life in mice.

We conducted gene array analyses with the principal goal of learning more about the initial differentiation of fetal HSCs to cells in lymphoid lineages. In that regard, several informative genes were identified and will be presented elsewhere (manuscript in preparation). Our screen also identified genes whose expression was previously thought to correlate with HSCs in fetal or adult tissues. Among those, Esam was particularly noteworthy as being drastically down-regulated during differentiation of HSCs to ELPs. We now report that it represents a potent tool for identifying HSCs over a wide range of developmental age.

ESAM was originally identified as an endothelial cell-specific protein, although it was also shown to be expressed on megakaryocytes and platelets.^22,31  A recent study also found ESAM transcripts in adult marrow HSCs when gene arrays were used to compare Thy1.1^Lo Flk2⁻ LSK (long-term HSC-enriched), Thy1.1^Lo Flk2⁺ LSK (short-term HSCs), and Thy1.1⁻ Flk2⁺ LSK (multiple progenitors).³⁰  We now show how ESAM levels can be exploited to obtain highly enriched CFU-Mix, lymphopoietic cells, and long-term HSCs from the Rag1⁻ c-kit^Hi Sca1⁺ fraction of E14.5 fetal liver. It is important to note that long-term HSCs in mouse embryos are unique in expressing markers, such as Flk2 and CD11b/Mac1, not characteristic of adult HSCs.^18,32  This complicates cell sorting strategies and fetal/adult comparison. We previously found that Rag1/GFP⁻ c-kit^Hi Sca1⁺ cells derived from E14.5 fetal liver reconstituted lymphohematopoiesis in lethally irradiated adults, whereas Rag1/GFP^Lo c-kit^Hi Sca1⁺ cells transiently contributed to T and B lymphopoiesis. ESAM specific Abs can be used without knockin reporters and with fewer combinations of Abs to obtain highly enriched HSCs.

Although we identified Esam as one of the highly expressed genes in HSC but not in ELP, levels of this antigen strongly correlated with lymphopoietic activity. The ESAM^Hi fraction in Rag1⁻ c-kit^Hi Sca1⁺ cells of E14.5 fetal liver produced CD19⁺ B-lineage cells more effectively and contained lymphopoietic progenitors with even higher frequency than the ESAM^−/Lo fraction. The cells also gave rise to long-term reconstitution in both T and B lymphoid lineages in lethally irradiated recipients. Furthermore, in the PSp/AGM region, B lymphopoietic activity in culture was exclusive to the ESAM^Hi Tie2⁺ fraction. From these observations, we conclude that ESAM expression indicates high lymphopoietic potential in HSCs of the fetal liver and hematopoietic cells arising in the AGM region. On the other hand, we presume from the sharp down-regulation in ELPs that ESAM is not necessary for early lymphoid differentiation.

We previously reported that the first lymphopoietic cells arise in a Tie2⁺ c-kit⁺ CD34^−/Lo CD41⁻ subset in the E8.5 PSp region.¹⁹  In this report, we showed clear ESAM expression on lymphopoietic cells contained in a Tie2^Hi c-kit⁺ subset of the E9.5 to E10.5 PSp/AGM. Lymphopoietic activity in early embryos is thought to associate closely with HSC development. Because the first HSCs have few reliable surface markers, high ESAM expression must be useful to monitor how the first HSCs are developing and moving to other sites. As an additional finding, we detected high myeloid-erythroid but little lymphopoietic potential in a Tie2^Lo ckit^Hi subset of the YS, whose expression level of ESAM was apparently low. We presume that ESAM^Hi Tie2^Hi c-kit⁺ cells in the PSp/AGM and ESAM^Lo Tie2^Lo c-kit^Hi cells in the YS differ in their origins and/or as development progresses. If combined with those cell surface markers, a newly described method to trace runt-related transcription factor 1-expressing cells would address several important issues related to “primitive” and “definitive” hematopoiesis.³³ 

It is interesting to speculate that ESAM might be important for HSC functions. A previous study showed that ESAM mediates homophilic interactions between endothelial cells,²²  and endothelial cells must represent an important component of HSC-supportive niches in bone marrow.³⁴  A subsequent study of ESAM-deficient mice showed impaired migration of neutrophils through vascular walls after normal adhesion.³⁵  In addition, platelets from ESAM-deficient mice were less prone to disaggregation.³⁶  Thus, ESAM expressed by endothelial and/or stem cells could have functions associated with HSC adhesion and/or migration.

Osteoblastic stromal cells that line trabecular bone are thought to construct niches containing molecules needed to regulate HSC quiescence, proliferation, and differentiation in adult bone marrow.^37,38  Other findings indicate that HSCs also interact with sinusoidal endothelial cells at some distance from bones.^39,40  Osteoblastic and vascular niches probably function in complementary ways.^41,42  The fetal liver has no osteoblasts, and HSC niches have not been identified in that site. High level ESAM display suggests that it should be further explored within the context of endothelial-HSC interactions.

It was an unexpected but intriguing finding that ESAM levels correlate with the stem cell rich fraction even in 22-month-old mice. Although many endothelial antigens on HSCs decline with aging, endomucin and endothelial protein C receptor (CD201) have been reported to be adult HSC markers.^17,43  In humans, angiotensin-converting enzyme (CD143) was recently found to mark HSCs through embryonic and adult life.⁴⁴  However, ESAM seems unique in actually increasing with aging. Many genes involved in NO-mediated signal transduction, stress responses, and inflammation have been linked to HSC aging.⁴  P-selectin is one of the most highly up-regulated of those stress-related genes. Of particular interest, P-selectin mediates some leukocyte-vascular endothelium interactions and leukocyte extravasation, functions also attributed to ESAM on neutrophils.³⁵  Recent reports showed that stem/progenitor cells continuously circulate outside bone marrow and can actively participate in innate immune responses.^45,46  Furthermore, platelets can use P-selectin to recruit marrow hematopoietic cells into sites of injury.⁴⁷  Although direct evidence is lacking at this moment, further study might implicate elevated ESAM levels in extramedullary migration of HSCs.

Efficient HSC-based therapies and the emerging field of regenerative medicine will benefit from learning more about what defines stem cells. Although patterns of expression of transcription factors and other intracellular proteins are informative, surface markers such as ESAM that are unique to HSCs have special utility.

The authors thank Drs N. Sakaguchi (Kumamoto University) and H. Igarashi (Kawasaki Medical School) for the Rag1/GFP knockin mice.

Contribution: T.Y. conducted experiments and analyzed results; T.Y., K.O., P.W.K., and Y.K. designed the research plan and wrote the paper; S.B. and D.V. prepared anti-ESAM antibodies; K.K. and T.M. performed and analyzed microarrays.

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

Correspondence: Takafumi Yokota, Department of Hematology and Oncology, Osaka University Graduate School of Medicine, C9, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan; e-mail: yokotat@bldon.med.osaka-u.ac.jp.