Studies of hematopoiesis based on mouse embryonic stem cells (ESCs) have intricately defined specific genes and proteins that regulate development of hematopoietic stem cells (HSCs) and specific blood cell lineages [1, 2]. However, key differences between the earliest stages of mouse and human embryogenesis have prompted interest in using ESCs derived from human preimplantation blastocysts to characterize mechanisms that regulate human developmental cell-fate decisions starting from a well-characterized, homogeneous cell population . We and others have demonstrated that myeloid, erythroid, lymphoid, megakaryocytic, and antigen-presenting cells can all be routinely derived from human ESCs (hESCs) [4 –9]. Genes known to regulate hematopoietic development are expressed in an orderly fashion during differentiation of hESCs, much as during embryogenesis [4, 7, 8, 10, 11]. Gene expression analysis of hESC-derived CD34+ CD38− cells compared with the same phenotypic cord blood (CB)-derived cell populations finds the ESC-derived cells to have a genetic profile similar to the embryonic stage of blood development . Moreover, these analyses of hematopoietic development from hESCs demonstrate that this model system can be used to define key cytokines and growth factors required for early human hematopoiesis [5, 12, 13].
hESCs have been proposed as a novel source of cells and tissues to replace those damaged by disease, degeneration, or trauma. For example, allogeneic hematopoietic cell transplantation (HCT) is used to successfully treat thousands of patients a year with a wide range of malignant and nonmalignant conditions. However, many patients who do not have an appropriately immune-compatible related or unrelated donor for HCT must receive alternative, often less effective, therapies . Complications from currently used alternative cell sources remain, such as immune compatibility and delayed time to effective hematopoietic engraftment [15, 16]. Therefore, the evaluation of alternative sources of cells for HCT, such as those that may be derived from hESCs, remains an important goal. However, to advance hESC-based regenerative medical therapies, the ability of hESC-derived cells to effectively function in vivo must be better demonstrated.
Because transplantation into human recipients is not feasible for experimental cell populations such as those derived from hESCs, severe combined immunodeficient (SCID)-repopulating cells (SRCs) can be evaluated as a surrogate assay to define human hematopoietic cell populations with stable long-term engraftment potential [17, 18]. Here, we examined hESCs allowed to differentiate on S17 stromal cells and injected into nonobese diabetic (NOD)/SCID mice. SRCs capable of long-term primary and secondary engraftment can be demonstrated. Additionally, we evaluated the immune response against the hESC-derived cells. We find that treatment of NOD/SCID recipients with anti-ASGM1 antibody improves engraftment, suggesting that natural killer (NK) cells pose an important immune barrier in this xenogeneic transplantation model.
Materials and Methods
The hESC lines H1 and H9 (WiCell Research Institute, Inc., Madison, WI, http://www.wicell.org) were maintained as undifferentiated cells as previously described . Both lines were used for in vitro experiments because previous studies by our group and another group demonstrated equivalent results [10, 13]. All transplantation and teratoma experiments were done exclusively with H1 cells. Briefly, undifferentiated ESCs were maintained on mouse embryonic fibroblasts in Dulbecco's modified Eagle's medium (DMEM)/F-12 supplemented with 20% knockout serum replacer (Invitrogen Corporation, Carlsbad, CA, http://www.invitrogen.com), 1% minimal essential medium-nonessential amino acids (Invitrogen), 1 mM l-glutamine, 0.1 mM β-mercaptoethanol (Sigma, St. Louis, http://www.sigmaaldrich.com), and 4 ng/ml basic fibroblast growth factor (Invitrogen). All cell lines were routinely evaluated by cytogenetics and found to maintain a normal karyotype. The mouse bone marrow (BM) stromal cell line S17 (kindly provided by Kenneth Dorshkind, University of California at Los Angeles) was grown as previously described .
NOD/SCID (NOD/LtSz-Prkdcscid) mice (The Jackson Laboratory, Bar Harbor, ME, http://www.jax.org) were bred and maintained in the animal facility of the University of Minnesota in microisolator cages and provided with autoclaved food and water. SCID/Beige (CB17/Icr.Cg-PrkdcscidLystbg/Cr) mice were purchased from Charles River Laboratories (Wilmington, MA, http://www.criver.com). Mice were housed, treated, and handled in accordance with the guidelines set forth by the University of Minnesota Institutional Animal Care and Use Committee and by the National Institutes of Health's Guide for the Care and Use of Laboratory Animals.
For the teratoma assays, undifferentiated hESCs at 70% confluence were treated with 1 mg/ml collagenase type IV (Invitrogen) to make small colonies. Cells were resuspended in DMEM/F-12 media containing 5% fetal bovine serum (FBS). Equivalent numbers of cells were injected intramuscularly into the left flank. Indicated mice were injected intraperitoneally with 400 μl of phosphate-buffered saline (PBS) containing 20 μl of anti-asialo GM1 (ASGM1) antiserum (Wako Chemicals USA, Inc., Richmond, VA, http://www.wakousa.com) the day before injection of hESCs (day 0). Additional antibody treatment was given at days 11, 22, and 33. For the anti-CD3 control group, each mouse was injected intraperitoneally with 300 μl of PBS containing 100 μg of anti-mouse CD3ε (eBioscience, San Diego, http://www.ebioscience.com) the day before injection of hESCs. Another 30 μg was injected on days 11 and 22. For statistical analysis, mice were considered positive for tumor development when the diameter of tumor reached 1.5 cm. The statistical significance among different groups was evaluated using the proportional hazard model. The proportional hazard assumption was checked and the forward selection was used to determine the final model. All the statistical analyses were performed with SAS 9.1 software (SAS Institute Inc., Cary, NC, http://www.sas.com).
Hematopoietic Differentiation of hESCs
Both H1 and H9 cell lines were used for the hematopoietic differentiation by coculture with mouse BM stromal cell line S17 as previously described [4, 19]. S17 cells were irradiated (3,000 cGy) prior to coculture with hESCs. After indicated days of differentiation, flow cytometric analysis and colony-forming assays were used to define the hematopoietic progenitor cells. For flow cytometric analysis, a single-cell suspension was incubated with either appropriate isotype control antibody or antigen-specific antibodies: anti-human CD34-allophycocyanin (APC), anti-human CD45-phycoerythrin (PE; BD Pharmingen, San Diego, http://www.bdbiosciences.com/pharmingen), and anti-human CXCR4-PE (eBioscience). Propidium iodide was used to exclude dead cells, and results were analyzed with CellQuest Pro Software (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) and FlowJo 6.0 (Tree Star, Inc., Ashland, OR, http://www.treestar.com). Colony-forming unit (CFU) assays were carried out as described previously using Methocult GF+ media (StemCell Technologies, Vancouver, BC, Canada, http://www.stemcell.com) with CFU scored after 14 days [4, 19].
H1 cells allowed to differentiate on S17 cells for 7–24 days were used for intravenous and intra-BM injection. NOD/SCID male and female mice at 6–10 weeks of age were sublethally irradiated with a dose of 300 cGy via cesium irradiator 3–6 hours before injection. Cells were passed through a 70-μm nitex mesh to remove clumps, and 2 × 106 to 4 × 106 cells were injected intravenously by tail vein into each NOD/SCID mouse. Intra-BM injection was carried out as described previously . Cells (0.5 × 106 − 3.8 × 106) in 50-μl total volume was transplanted into the left femurs of all recipients mice. After 3–6 months, BM cells were flushed from the tibiae and femurs of the transplanted mice. Engraftment of human cells in the mice was determined by flow cytometry and polymerase chain reaction (PCR) analysis (as described below); 1 × 105 human umbilical CB CD34+ cells were transplanted into similar mice as a positive control. Secondary transplantation studies were performed using BM cells from one primary mouse injected into two or three secondary sublethally irradiated (300 cGy) mice. Approximately 5 × 106 unfractionated BM cells were injected into each mouse. Human engraftment was assessed after 3 months post-transplantation. For indicated animals, anti-ASGM1 treatment was done as above.
Flow Cytometry Analysis of Transplanted Cells
Single-cell suspension from the BM of transplanted mice was treated with ammonium chloride red cell lysis buffer. The BM cells were blocked with PBS containing 5% FBS and 5% human serum (Nabi Biopharmaceuticals, Boca Raton, FL, http://www.nabi.com) for 20 minutes on ice before incubation with antibody. Directly conjugated antibodies were used to identify human-specific blood cells: anti-CD45-APC, anti-CD34-APC, anti-CD33-PE, anti-CD33-fluorescein isothiocyanate (FITC) (BD Pharmingen), and anti-CD38-PE (BD Pharmingen). Levels of nonspecific staining were established by parallel analysis of cells incubated with isotype-matched control antibody labeled with the same fluorochromes. Twenty thousand cells were routinely analyzed per mouse. The percentage of cells in positive gated region of isotype controls was subtracted from the percentage of cells indicated as stained with CD45-APC antibody.
PCR and Quantitative PCR Analysis of BM Engraftment
Genomic DNA was isolated from the BM or lungs of mice that received transplants using DNeasy tissue kit (Qiagen, Valencia, CA, http://www1.qiagen.com). Peripheral blood DNA was isolated using QIAamp DNA Blood Mini kit (Qiagen). One hundred nanogram (ng) DNA samples were used to amplify a 1,171-bp fragment of human chromosome 17-specific α-satellite using the following primers: forward 5′ ACACTCTTTTTGCAGGATCTA-3′ and reverse 5′-AGCAATGTGAAACTCTGGGA-3′. The HotStar Taq DNA polymerase system was used under the following conditions: 95°C for 15 minutes (1 cycle), 94°C for 1 minute, 61.5°C for 1 minute, 72°C for 2 minutes (35 cycles), and 72°C for 7 minutes.
Quantitative PCR was performed using an ABI PRISM 7700 sequence detector (software version 1.7a; Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). Human microsatellite DNA was amplified under the following conditions: 40 cycles of three-step PCR (95°C for 15 seconds, 60°C for 30 seconds, and 72°C for 30 seconds) after initial denaturation (95°C for 10 minutes) with 100 ng of DNA as template, 50 μM of each primer, and 1× SYBR-green PCR Master Mix (Applied Biosystems). The sequence of human microsatellite primers was forward 5′ ATTCACGTCACAAACTGAACATTC 3′ and reverse 5′CG-TTTGAAATGTCCGTTTGTAGAT 3′. The level of human cell engraftment in mouse BM was determined by comparing the human microsatellite product with that of human/mouse DNA mixture standard curve (range, 100%–0.0001% human DNA).
Statistical significance of human CD45+ cells in different groups of transplanted mice was evaluated with Mann-Whitney U test.
Hematopoietic Differentiation of hESCs
We used the hESC lines H1 and H9 to derive hematopoietic cells by coculture with the S17 mouse BM-derived stromal cell line [4, 13]. We and others have previously demonstrated that these two hESC lines give very similar results when used for studies of hematopoiesis [4, 5, 13]. Under these conditions, distinct subsets of CD34+ cells are routinely identified (Fig. 1). Initially, a population of CD34+CD31+CD45− cells develops, with this population peaking at approximately 14 days of hESC coculture with S17 cells. Next, a population of CD34+CD45+ cells develops, with a peak at 17–21 days of coculture (Fig. 1A, 1B). This ordered differentiation in S17 stromal-based cultures is very similar to what has been reported by other groups examining embryoid body-mediated differentiation of hESCs and differentiation supported with OP9 cells [7, 8, 10]. In agreement with this developmental pattern, the CD34+CD45+ cells in these cultures also correlate with derivation of hematopoietic progenitors, as demonstrated by ability to give rise to myeloiderythroid colonies in standard CFU assays. CFU development peaks during the same 17- to 21-day period in which maximal CD34+CD45+ cell development is seen (Fig. 1C), and sorting for CD34+, CD45+, or CD34+CD45+ cells markedly enriches for CFU [4, 5, 13]. Interestingly, CD34+CXCR4+ cells that represent a population best able to home and engraft in BM  also develop under these conditions, something that has not been previously well characterized in cultures of in vitro differentiated hESCs (Fig. 1B).
Evaluation of Mouse Anti-hESC Immune Response
As an initial model for in vivo engraftment and growth of hESC-derived cells, we more closely evaluated teratoma development from undifferentiated hESCs in immunodeficient mice. Because different strains of immunodeficient mice have varying immune capacity, we compared teratoma formation in NOD/SCID and SCID/Beige (SCID/Bg) mice. Both strains lack cells of the adaptive immune system (B and T cells) due to the SCID mutation. However, effectors of innate immunity differ in these mice [22, 23]. Although NOD/SCID mice are reported to have defective NK cell activity, several analyses of peripheral blood mononuclear cells demonstrate that these mice retain some NK cell activity [24, 25]. In contrast, the Beige mutation leads to a defect in granule formation, and these mice demonstrate little NK cell-mediated cytotoxicity [18, 23]. We injected both NOD/SCID and SCID/Bg mice intramuscularly with undifferentiated hESCs. We found that teratomas consistently grew faster in the SCID/Bg mice as compared with NOD/SCID mice, although eventually, most NOD/SCID mice did develop tumors (Fig. 2A). To address whether mouse NK cells play a role inhibiting engraftment and growth of hESC-derived teratomas, we treated NOD/SCID mice with anti-asialo GM1 antibody as a means to deplete NK cell activity . Of note, macrophages also express this surface glycolipid [26, 27] and may be affected by this antibody treatment. However, NOD/SCID mice have a defect in myeloid cell development and function that may make activity against these cells less important in this model . We found that teratomas developed significantly faster in anti-ASGM1-treated NOD/SCID mice compared with untreated mice or mice treated with anti-CD3 antibody used as a control to deplete any residual T cells. Indeed, anti-ASGM1-treated NOD/SCID mice had teratomas that grew at essentially the same rate as in SCID/Bg mice (Fig. 2A).
These results suggest that mouse NK cells exhibit reactivity against hESCs. Of note, hESCs express relatively little HLA class I expression compared with adult cell types and no HLA class II expression. (Fig. 2B, 2C) [29, 30]. We also found that hESC-derived CD34+ cells demonstrate only low levels of HLA class I expression and that this expression is less than that seen on CD34+ cells isolated from CB (Fig. 2D). Because NK cells preferentially lyse cells that have low expression of HLA class I molecules , failure to engage inhibitory receptors on mouse NK cells may represent a mechanism that leads to decreased engraftment and teratoma development from hESCs.
Transplantation of hESC-Derived Cells
The SRC assay effectively evaluates cell populations capable of long-term in vivo engraftment. Here, we examined SRCs derived from hESCs that had been allowed to differentiate for a varying number of days on S17 stromal cells. For these studies, we elected not to sort for specific cell populations such as CD34+ cells, because some CD34+ cells may also represent SRCs . Also, there may be other cells within this heterogeneous cell population, such as mesenchymal cells, that can aid hematopoietic cell engraftment . Sublethally irradiated NOD/SCID mice were injected intravenously with hESC-derived cells (H1 cell line) that had been allowed to differentiate on S17 cells for 7–24 days (termed H1/S17 cells). Initial studies were done with groups defined based on receiving cells that had been allowed to differentiate for 7–12, 14–19, or 21–24 days. These groups were determined based on consistent phenotype and CFU potential of cells at these times (Fig. 1) . That is, the cells at days 7–12 of differentiation correspond to the CD34+CD31+ developmental subpopulation, cells at days 14–19 correspond to increased CD34+CD45+ cell development, and cells at days 21–24 represent primarily more mature hematopoietic cells . Another group of control animals was injected with CD34+ cells isolated from human CB. BM was evaluated for the presence of hESC-derived cells 3–6 months after transplant. Human CD45+ cells were consistently detected in the BM of transplanted mice, with better engraftment in animals that received cells 7- to 12-day H1/S17 cells (Figs. 3A, 4).
Because these initial studies represent a pooled analysis of mice transplanted with H1/S17 cells from different days of differentiation, we next focused on two time points of H1/S17 cell development: day 12 cells corresponding to peak of CD34+CD31+ cell development and day 20 cells corresponding to increased CD34+CD45+ cell development. Again, we found that both populations contained cells able to mediate long-term engraftment (Fig. 3B).
PCR of a human-specific sequence was done on BM samples to confirm presence of H1/S17 cells in the marrow (Fig. 5A). Peripheral blood and lung tissue also demonstrated the presence of human cells (Fig. 5A). Quantitative PCR results correlated with the level of engraftment demonstrated by flow cytometric analysis (Fig. 5B).
Analysis of phenotypic subsets of engrafted H1/S17-derived hematopoietic cells found that most were CD45+CD33+ myeloid cells (Fig. 6A). However, CD34+ cells were also detected in the marrow of most animals, with most CD34+ cells being CD38+. In contrast, in animals that received CB-derived cells, most human CD34+ cells in the BM also coexpressed CD38. Human hematopoietic progenitors within the mouse BM were demonstrated by CFU assay and PCR (Figs. 5, 6).
Secondary transplants were done from mice originally injected with H1/S17 cells from 7–19 days of differentiation. In each case, BM from one primary recipient was injected intravenously to two or three NOD/SCID mice. Twenty-eight secondary transplant recipients were evaluated, and 13 recipients showed evidence of engraftment 3–6 months after transplant, ranging from 0.08%–0.2% human CD45+ cells in the BM. Again, PCR analysis confirmed stable engraftment of hESC-derived cells (Fig. 5C).
Measures to Enhance Engraftment of hESC-Derived Blood Cells
We next transplanted H1/S17 cells into NOD/SCID mice that were treated with anti-ASGM1 antibody. As demonstrated above, anti-ASGM1-treated NOD/SCID mice had more rapid growth of hESC-derived teratomas, suggesting that these murine NK cells inhibit engraftment and growth of hESCs with low expression of HLA class I. Moreover, we found that hESC-derived CD34+ cells had lower HLA class I expression than did CD34+ cells isolated from CB (Fig. 2). Anti-ASGM1 treatment resulted in significantly better engraftment of human CD45+ cells in the BM of most of the anti-ASGM1-treated mice (Fig. 3A, 3B). Again, studies that used a single time point of H1/S17 differentiation confirmed the pooled analysis and demonstrated increased engraftment in the anti-ASGM1-treated recipients. Results were confirmed by PCR analysis (Fig. 5A).
As another means to potentially improve engraftment of hESC-derived blood cells, we also injected H1/S17 cells directly into the BM of NOD/SCID mice. This method has been found to improve engraftment in some models, most likely by decreasing the need for homing of SRCs to the BM as is needed after i.v. injection [20, 34]. In all cases, the left femur was injected, and BM was analyzed separately from both left and right femurs 3–6 months after injection. Human CD45+ cells were consistently demonstrated in the marrow of the mice, both on the left and right sides (Figs. 3C, 4B). Again, PCR analysis was done to confirm presence of human cells in both femurs (Fig. 5A). These results demonstrate that hESC-derived SRCs can circulate in these mice to allow for engraftment on the contra-lateral side. However, the level of engraftment was not significantly different compared with intravenously injected mice. This suggests that homing was not a limiting factor for engraftment after i.v. injection of H1/S17 cells.
Because development of hESC-derived teratomas is a safety concern for hESC-based therapies, the spleen, liver, and lungs of these transplanted mice were carefully examined for evidence of tumors. Given that mice were evaluated 3 or more months after transplant, teratomas would have ample time to develop (Fig. 2); however, none were seen in any animal. PCR analysis did show hESC-derived cells in the lungs of some transplanted mice, suggesting either that H1/S17 cells were trapped and survived in the lungs or that the H1/S17 cell population consists of some cells that engraft pulmonary tissue (Fig. 5A). Finally, all mice injected with H1/S17 cells appeared healthy and survived normally until used for experimental analysis, without evidence of early death after i.v. injection of hESC-derived cells .
Here, we examined the ability of cells derived from hESCs to provide hematopoietic engraftment and survive long-term after injection into immunodeficient mice. This SRC assay represents the best means to characterize putative human hematopoietic precursor populations with similar characteristics and in vivo potential as HSCs [17, 18]. We find that SRCs can be derived from hESCs that have been induced to differentiate by coculture with the S17 BM stromal cell line. These H1/S17 cells can engraft both primary and secondary recipients after i.v. injection. Moreover, after direct intra-BM injection, circulating SRCs are demonstrated by the ability to recover hESC-derived blood cells in the contra-lateral femur of these mice. These results suggest that hESCs can serve as an effective source of human HSCs.
Although an ordered hematopoietic development from hESCs in vitro has been described, the in vivo potential of these cells is less well characterized. It remains unclear whether putative HSCs derived from hESCs will act more like those derived from mouse ESCs that are relatively inefficient at in vivo engraftment, especially without genetic modification (reviewed in ), or whether these cells will be more like hematopoietic precursors isolated from human CB that can be routinely transplanted into immunodeficient mouse recipients. The studies presented here, as well as other recent studies using human and nonhuman primate ESCs that also demonstrate derivation of cells capable of hematopoietic engraftment [35–37], suggest that hESC-derived hematopoietic cells are capable of stable engraftment without the need for genetic modification, much like SRCs from CB. One of these studies used the fetal sheep transplantation model to further demonstrate that hESC-derived hematopoietic cells (also H1/S17 cells) can exhibit stable primary and secondary engraftment .
The xenogeneic SRC model presents several impediments for stable engraftment of hESC-derived blood cells. Hematopoietic precursor cells must appropriately home to the BM, then engraft and survive in the mouse hematopoietic microenvironment. Specific receptors, such as CXCR4, are required for successful homing , and we demonstrate that hESC-derived CD34+ cells are also CXCR4+ (Fig. 1). Somewhat surprisingly, direct intra-BM injection did not lead to enhanced engraftment in our studies. This suggests that poor homing may present less of a problem in this model.
Immune mechanisms present a second barrier to engraftment. Whereas some studies suggest that hESCs may have poor immune-stimulatory activity or be immune-privileged [30, 38], others clearly show that immune-stimulating dendritic cells and other antigen-presenting cells can be established from hESCs . Similar to our results, engrafted mouse ESC-derived hematopoietic cells are also rejected by host NK cells . The xenogeneic immune response against undifferentiated and differentiated hESCs important to this preclinical model has not been well established. We first noted that NOD/SCID mice may have anti-hESC activity, because undifferentiated hESCs developed into teratomas at a much slower rate after injection into NOD/SCID mice, compared with the SCID/Bg mice that were used for earlier studies [40, 41]. Indeed, a direct comparison using the same hESCs injected into either NOD/SCID or SCID/Bg mice confirmed that the tumors developed more rapidly in the SCID/Bg hosts (Fig. 2). We hypothesized that depletion of NK cells in the NOD/SCID mice would lead to more rapid tumor development. This hypothesis was proven correct by treatment of the NOD/SCID mice with anti-ASGM1 antiserum and by finding that teratomas then developed at essentially the same rate as in untreated SCID/Bg mice. Macrophages that also express ASMG1 [26, 27] may also play a role in inhibiting engraftment in the untreated NOD/SCID mice. However, NOD/SCID mice also demonstrate defects in myeloid development and function, and these mice lack the complement protein C5 and cannot generate complement-mediated killing . Also, a recent report that treated immune-competent mice with anti-ASGM1 demonstrated depletion of NK cells, but not Mac-1+ macrophages . Therefore, depletion of NK cells by anti-ASGM1 treatment probably provides the primary activity in this model.
We found that CD34+ cells derived from hESCs expressed lower levels of HLA class I molecules on the cell surface than did CD34+ cells isolated from CB. Similar results were obtained in a gene array analysis of CD34+ cells derived from hESCs or isolated from CB . Again, several models demonstrate that mouse NK cells can inhibit engraftment of human HSCs/SRCs [24, 25, 43–45]. Also, engraftment of mouse ESC-derived HSCs is inhibited by mouse NK cells when injected in a syngeneic recipient . Interestingly, mouse ESC-derived HSCs, as well as HSCs isolated from yolk sac and other sites of early hematopoiesis, also typically express a low level of MHC (major histocompatibility complex) class I molecules [39, 46–48]. Therefore, not only can mouse NK cells inhibit engraftment of both human CB CD34+ cells and mouse ESC-derived HSCs, but they also pose a barrier to engraftment of hESC-derived hematopoietic precursor cells.
One proposed measure to avoid immune-mediated rejection of hESC-derived cells and tissues is to engineer these cells to have deficient HLA class I expression as a means to avoid T cell-mediated immune response [29, 49]. However, our results suggest that this strategy may have the unintended consequence of increased NK cell-mediated rejection. A number of other immune-avoidance strategies remain feasible to facilitate hESC-based regenerative medicine therapies [29, 49, 50].
The H1/S17 cells tested here represent a heterogeneous cell population. The cell phenotype(s) that mediate engraftment are not specifically defined in these studies. These results also do not determine whether one, or multiple, phenotypic cell populations engraft and survive. However, CD34+ cells are routinely derived from hESCs after a few days of differentiation (Fig. 1) [4, 13], with 5%–10% of H1/S17 cells typically being CD34+ and most being CD34+CD38−. CD45+ cells can typically be detected at day 10 and increase at later time points (Fig. 1; ). Previously, we have noted that the intensity of CD45 expression on cells from S17-mediated differentiation can be lower than CD45 expression on human hematopoietic cells from other sources such as blood or BM [4, 13]. However, our ability to sort and enrich for functional populations such as CFU  unambiguously demonstrates CD45 to be an important hematopoietic cell marker for hESC-derived cell populations. We also routinely observe CD45− cells that express the surface antigens and receptors CD31 and Flk1. This CD34+CD31+CD45− cell population has been previously identified as an hESC-derived population that may serve as a common precursor to both blood and endothelial cells [8, 10]. Therefore, some component of cells engrafted in mice injected with H1/S17 cells may represent endothelial cells. Recent studies using cynomologous monkey ESCs allowed to differentiate on OP9 cells for 6 days before injection into fetal sheep also demonstrated hematopoietic chimerism . As in our studies, this cell population was also heterogeneous and not sorted for a specific phenotypic cell type prior to injection. Therefore, these surviving cells in the sheep BM may arise from mesodermal cells capable of both blood and endothelial cells development. Similarly, our studies suggest a higher level of engraftment in mice injected with H1/S17 cells allowed to differentiate for 7–12 days as compared with 14–19 days or 21–24 days (Fig. 3). Therefore, we hypothesize that, as seen in these other studies of human and nonhuman primate ESC-derived hematopoietic cells [35, 36], the CD34+CD31+CD45− H1/S17-derived cell population at these earlier time points of differentiation may include more mesodermal cells capable of blood and endothelial cell development within this earlier cell population, and these less mature cells may have better engraftment capacity. This corresponds to our finding that these day 7 to 12 cells contain fewer CFU than the cells allowed to differentiate for longer periods of time. These putative bipotential mesodermal cells would not be expected to contain mature hematopoietic progenitors, and therefore development of CFU may not correlate with SRCs.
Because mice in these studies were injected intravenously with a minimum of 2 × 106 H1/S17 cells, this would represent 1–2 × 105 CD34+ cells. A similar quantity of CD34+ cells isolated from CB would be expected to provide robust SRC activity [24, 25, 51]. However, the hESC-derived CD34+ cell population may be composed of a wider variety of cell types, such as cells of endothelial lineage, leading to a relatively low level of long-term engraftment. This heterogeneous population with relatively fewer SRCs may account for the difficulty to demonstrate lymphoid cells in vivo. Clearly, sorting for specific cell populations within the H1/S17 cells is needed as a next step to define the phenotype of cells with SRC activity. This would allow better understanding of in vitro culture conditions that best support development of SRCs (and presumably HSCs).
Safety of hESC-based therapies has been raised as a concern due to the ability of the undifferentiated cells to form teratomas. Importantly, we examined the lungs, liver, and spleen of animals injected with differentiated H1/S17 cells for evidence of tumors, and none were seen. Because the NOD/SCID mice were sublethally irradiated and many were injected with anti-ASGM1 antibody, these mice would be highly susceptible to teratoma development. Moreover, no unexplained premature deaths of the mice that received H1/ S17 cells were noted. In contrast, mice injected intravenously or by intra-BM with undifferentiated hESCs routinely develop teratomas. Therefore, it is likely that few, if any, undifferentiated ESCs remained in the H1/S17 cell population that was injected. Additionally, no deaths were noted at time of injection, as has been reported for mice that received i.v. injection of sorted cell populations derived from hESCs , suggesting that use of the mixed H1/S17 cell population may provide advantages over the use of specific populations of sorted cells. It is of interest that another study was able to demonstrate significant engraftment after intra-BM injections, but not after i.v. injection of hESCs induced toward hematopoietic differentiation by embryoid body formation in media supplemented with multiple cytokines and growth factors . These results suggest that both the method used to support lineage-specific differentiation hESCs, and the method used to transplant specific cell types play a key role in determining levels of engraftment. Further studies to directly compare engraftment and survival of mice that receive unsorted versus sorted populations of different hESC-derived blood cells are needed to evaluate the strengths and weaknesses of various phenotypic cell populations.
Multiple studies have now defined means to support hESC differentiation into a wide range of cell types [3, 50]. In addition to blood, the in vivo potential of other hESC-derived lineages is now being realized, as seen for cell types such as cardiomyocytes  and neuralglia cells . Well designed clinical trials will soon be needed to define the safety and efficacy of hESC-based therapies [29, 50].
This work was supported by the National Institutes of Health (HL-72000) (D.S.K.) and an American Society of Hematology Scholars award (D.S.K.). We thank Elizabeth Johnson and Ibi-Yinka Tolliver for expert technical assistance, Jingbo Du for management of the NOD/SCID mouse core, Drs. Jeff Miller and Catherine Verfaillie and members of their labs for helpful discussions and assistance, and Dr. Colin Martin for review of the manuscript and assistance. Statistical analyses were done with the assistance of Yan Zhang and the Biostatistics core of the University of Minnesota Cancer Center.