Prolonged Cell Cycle Transit Is a Defining and Developmentally Conserved Hemopoietic Stem Cell Property¹

Hemopoietic stem cells (HSCs)³ constitute a small population of cells defined by their unique ability to differentiate into all blood cell lineages as well as to generate identical progeny with the same unrestricted hemopoietic potential through self-renewal (1). The regulation of HSC fate decisions remains largely unknown, and no unique characteristics of the HSC cell cycle transit have been identified, which might be decisive for securing symmetrical and asymmetrical self-renewing divisions.

Formation of definitive HSCs in the developing embryo occurs in the dorsal aorta region around 10 days postcoitum (dpc) (2). Embryonic blood cell production starts as these stem cells migrate and populate the liver. The fetal liver microenvironment not only supports blood formation but importantly also an exponential expansion of HSC numbers through self-renewing cell divisions (3, 4). This is reflected by the fact that fetal liver HSCs, despite being largely similar to and the ancestors of adult HSCs, are much better than their adult counterparts in repopulating the bone marrow (BM) and reconstituting hemopoiesis following transplantation into myeloablated hosts (5, 6, 7). Thus, the in vivo repopulating ability of highly proliferative fetal HSCs is superior to that of the more quiescent and slowly cycling adult HSCs (8, 9). As much as this might seem reasonable from a developmental point of view, it is seemingly in conflict with several studies in which HSC function (repopulating ability) has been suggested to be compromised while transiting through the S-G₂-M phases of the cell cycle (10, 11, 12, 13, 14, 15).

In steady-state adult BM, HSCs transit the cell cycle on average as infrequently as once every 4–8 wk (8, 9) and are therefore highly enriched in the G₀ cell cycle phase (16). For this reason, it has not been possible to establish whether the cell divisions of adult HSCs when entering active cell cycle in steady state, occur with similar kinetics to that of progenitor cells that have lost the ability to self-renew. However, based on in vitro studies, it has been proposed that the first cell division of HSCs is prolonged due to the need to exit G₀, and although complicated by most HSC daughter cells in such cultures being progenitors rather than HSCs, it has been assumed that subsequent ex vivo HSC divisions occur with the same kinetics as downstream progenitors (10, 13, 17). This combined with ex vivo expanding HSCs being highly enriched in G₁ have led to the conclusion that HSCs are severely compromised in their engraftment potential when transiting S-G₂-M phases of the cell cycle. If so, it could have important clinical implications (15), because a reduced repopulating ability of cycling HSCs could explain why extensive efforts to ex vivo expand HSCs have in most cases failed (15, 18), and only in a few cases resulted in marginal increases in HSC numbers (19, 20).

Herein, to better establish the cell cycle transit of HSCs actively undergoing self-renewing divisions under physiological conditions, we investigated for the first time the cell cycle kinetics of HSCs in the 14.5 dpc fetal liver, a stage when all HSCs were found to divide within 48 h. We demonstrate that the average cell division of HSCs (10.6 h) is twice that of hemopoietic progenitors (5.6 h), due to a prolonged G₁ transit and passage through a state of relative G₀ quiescence, resulting in also fetal HSCs being highly enriched in G₁. Furthermore, we demonstrate for the first time that also ex vivo expanding adult HSCs sustain a prolonged G₁ and cell cycle transit when compared with hemopoietic progenitors, and consequently the relative enrichment of HSCs in G₀-G₁ does at least in part reflect the uniquely prolonged and developmentally conserved cell cycle transit of HSCs.

Congenic C57BL/6 strains differing only at the CD45 locus were used in all experiments. Experiments were approved by the ethical committee at Lund University. All mice were given sterile food and autoclaved acidified water, and housed under pathogen-free conditions in individually ventilated cages.

Pregnant mice were given an i.p. injection of BrdU (Sigma-Aldrich) in 0.9% saline (1 mg of BrdU per 6 g of body weight) and for 2–48 h allowed to freely drink water containing BrdU (1 mg/ml), before isolation of livers from fetuses at 14.5 dpc for analysis. Evaluation of BrdU incorporation as well as cell cycle analysis was done using a BrdU and a Ki67 and 7-aminoactinomycin D intracellular staining kit (BD Pharmingen) according to the manufacturer’s protocol.

Isolation of LSK and LSKCD34⁻ cells from the BM of >10-wk-old C57BL/6 mice was performed using FACS, as previously described (20). Cells were sorted on a FACSVantage or FACSDiva (BD Biosciences). Reanalysis of sorted cells reproducibly showed a high purity (>96%).

For isolation of LSKMac1^low HSCs from fetal liver, C57BL/6 breeders were put together in late afternoon and were checked for vaginal plugs the following morning (designated 0.5 dpc). At 14.5 dpc, livers were dissected from fetuses and single-cell suspensions were filtered through a nylon mesh (70 μm). Cells were lineage depleted and stained with specific Abs as previously described (20), using a mixture of lineage-specific Abs against B220, CD3, CD8, Gr-1, Ter-119, and Abs against Mac-1, Sca-1, and c-kit or isotype-matched control Abs (all from BD Pharmingen). LSKMac1^low cells were subsequently sorted on a FACSDiva with a purity of >96%.

To separate HSCs in G₀-G₁ and S-G₂-M fractions, cells were diluted to 10⁶ cells/ml in PBS supplemented with 5% FCS (BioWhittaker) and Verapamil (100 μM) (Abbott Scandinavia) and incubated at 37°C for 30 min with 2.5 μg/ml Hoechst 33342 (Molecular Probes). Cells were kept in PBS containing Verapamil at 4°C to prevent dye efflux and fractionated based on cell cycle distribution using a FACSVantage or FACSDiva. As a purity control of sorted S-G₂-M cells, we in one experiment transplanted recipients with a number of G₀-G₁ cells corresponding to the impurity of the S-G₂-M sort (S-G₂-M impurity was at 0.7%).

Ex vivo-expanded BM HSCs (300–600 cells) or fresh fetal liver HSCs (100 cells) from C57BL/6 mice (CD45.2) were transplanted into lethally irradiated (925 rad) congenic recipients (CD45.1 or CD45.1/CD45.2) together with 200,000 unfractionated BM competitor cells (CD45.1), allowing quantification of reconstitution activity and ensuring survival of lethally irradiated mice. Hemopoietic donor cell reconstitution and lineage distribution was evaluated in peripheral blood at different time points posttransplantation by FACS, as previously described (20). Total reconstitution from 1000 control LSKCD34⁻ HSCs transplanted directly from the primary adult HSC culture or following the Hoechst 33342 staining did not differ significantly (56.7 ± 1.9 and 51.7 ± 6.8%, respectively), suggesting that staining procedures did not affect their viability or repopulating ability. Calculation of reconstituting units (RUs) was performed based on a definition of 1 RU as the repopulating ability of 10⁵ competitor BM cells, as previously described (21).

Sorted HSCs cells were expanded ex vivo (20) using serum-free medium X-Vivo 15 supplemented with 1% detoxified BSA, 1% penicillin/streptomycin, 1% l-glutamine (all from BioWhittaker), 1% 2-ME (Sigma-Aldrich), and cytokines (50 ng/ml each of stem cell factor (SCF; Immunex), Flt3 ligand (Immunex), and thrombopoietin (TPO; Amgen), and 20 ng/ml IL-3 (PeproTech)). Cell densities were never allowed to exceed 0.5 × 10⁶ cells/ml.

Cells were plated in duplicate in IMDM, supplemented with 20% FCS, 1% penicillin/streptomycin, 1% l-glutamine, 1% 2-ME, methylcellulose (StemCell Technologies), and cytokines (50 ng/ml each of SCF, Flt3 ligand, and TPO, 20 ng/ml IL-3, and 10 ng/ml granulocyte CSF (Amgen)) in 35-mm Petri dishes. Cultures were incubated at 37°C in a humidified atmosphere with 5% CO₂ in air for 7 days. Only colonies with >50 cells were scored.

Sorted HSCs cells were expanded ex vivo under conditions described above and subsequently stained with the mitotic tracker dye PKH26 (Sigma-Aldrich) according to the manufacturer’s protocol and after 2 days of additional expansion culture, sorted on a FACSDiva or FACSVantage to fractionate the rapid (PKH26^low, 35% of population with lowest PKH26 intensity) and slow (PKH26^high, 20% of population with highest PKH26 intensity) proliferating cells. On average, 1,000 PKH26^high cells or 3,000–5,000 PKH26^low cells were transplanted together with 200,000 unfractionated BM cells, and hemopoietic reconstitution was evaluated as described above. Note that, to compensate for the potentially higher number of asymmetric cell divisions in PKH26^low compared with PKH26^high cells, three to five times as many PKH26^low cells were transplanted into each recipient. High-resolution cell division tracking using CFSE (Molecular Probes) was conducted as previously described (20).

All data are reported as means ± SD or SEM. Statistical comparisons were made using Student’s t test for unpaired samples and differences with p < 0.05 were regarded as significant.

We investigated the cell cycle status and transition time of HSCs in the fetal liver, at a time when HSC numbers expand extensively in vivo (3, 4). Based on their rapid expansion, it was expected, although not previously demonstrated, that 14.5 dpc fetal liver HSCs would be actively dividing. Indeed, we found that 100% of Sca-1⁺c-kit⁺ HSCs (22) as Sca-1⁻c-kit⁺ hemopoietic progenitor cells (HPCs; Fig. 1,A) had proliferated within 48 h, as demonstrated through uptake of the thymidine analog BrdU (Fig. 1,B). Although the Sca-1⁺c-kit⁺ cell surface phenotype does not exclusively identify HSCs, all HSCs are Sca-1⁺. As the goal was to measure times for complete BrdU incorporation within each population and because long-term HSCs (LT-HSCs) have the slowest division kinetics, impurities within the Sca-1⁺c-kit⁺ population would in fact tend to underestimate differences in cell cycle transit times between LT-HSCs and progenitors. In light of the BrdU incorporation data, the distinct differences in cell cycle distribution between HPCs and HSCs were striking (Fig. 1 C). Whereas the majority of fetal liver HPCs resided in S-G₂-M (78%), the majority of fetal liver HSCs were in G₁ (51%), with just 34% in S-G₂-M. Moreover, whereas virtually no HPCs were found in G₀, as much as 14% of the HSC population appeared to be in a G₀ state, as indicated by undetectable Ki67 expression (23). Thus, despite extensive expansion and continuous proliferation, fetal liver HSCs appear to re-enter a transient state of relative quiescence and prolonged G₁ transit.

To compare the cell cycle kinetics of fetal liver HSCs and HPCs, pregnant females were treated with BrdU, and enriched HSC (Sca-1⁺c-kit⁺) or HPC (Sca-1⁻c-kit⁺) populations of 14.5 dpc fetal liver cells were analyzed for BrdU incorporation after 2–48 h. Within 2 h, as much as 42% of enriched HSCs and 74% of HPCs (p < 0.05) had incorporated BrdU (Fig. 2,A). After 9 h, 64% of HSCs and virtually all (99%) HPCs had incorporated BrdU (p < 0.05), indicating that the HSC population in the fetal liver has a much slower turnover rate than the HPCs. Plotting the log values of undivided cells (that had not incorporated BrdU) over time gives a linear regression curve (8) (Fig. 2,B). Based on this, the constant fraction of HSCs and HPCs entering the cell cycle in a random fashion was calculated to be 8 and 18% per hour, respectively, and the 50% BrdU incorporation point for HSCs and HPCs was 3.8 and 1.7 h, respectively. Although the HSC and HPC populations analyzed are heterogeneous and cycling asynchronously, the average cell cycle time for each population can be calculated by dividing the BrdU incorporation rate (0.08 and 0.18 for HSCs and HPCs, respectively) with the fraction of proliferating cells (G₁ + S-G₂-M = 0.85 and 0.99 for HSCs and HPCs, respectively; Fig. 1,C). Thus the calculated mean cell cycle transit time of enriched HSCs and HPCs in fetal liver was 10.6 and 5.6 h (p < 0.05), respectively. Although we used the same methods as in previous studies of HSC kinetics in adult mice to calculate the above cell cycle transit times (8), the calculated times might be unrealistically short. As an almost complete (99%) BrdU labeling was observed within 25 h for HSCs and within 11 h for HPCs (Fig. 2 B), these times might be more reasonable estimates for the generation times.

The above studies of cell cycle distribution and BrdU incorporation in phenotypically defined fetal liver HSCs suggested that fetal liver LT-HSCs, like their adult counterparts, would preferentially reside in G₀-G₁ of the cell cycle. To test this, we isolated G₀-G₁ and S-G₂-M fractions of Lin⁻Sca-1⁺c-kit⁺Mac-1^low (LSKMac1^low) fetal liver cells (22), a population which is highly enriched for LT-HSCs but predominately contains short-term HSCs (ST-HSCs) and multipotent progenitors (11) (Fig. 3,A). We then evaluated the long-term multilineage competitive repopulating activities of the two fractions by transplantation of limited numbers (100 cells per mouse) into lethally irradiated congenic adult recipients. Although long-term reconstitution ability was detected in both fractions, G₀-G₁ cells were enriched for multilineage LT-HSC repopulating activity and provided as much as 4-fold higher reconstitution (Fig. 3 B) when compared with S-G₂-M cells. Thus, as LT-HSCs expand in the fetal liver, they are preferentially distributed in G₀-G₁, as would be predicted due to their uniquely protracted G₀-G₁ transit.

Lin⁻Sca-1⁺c-kit⁺CD34⁻ (LSKCD34⁻) cells represent only 0.01% of total BM cells, but contain virtually all LT-HSCs in adult mice (24). Defined conditions efficiently promoting in vitro self-renewal of mouse LT-HSCs have been established (19, 20, 25). However, because HSCs at best are maintained in such cultures, HSC divisions (as in vivo) seem to be primarily asymmetric or non-self-renewing, sustaining rather than expanding the number of LT-HSCs, resulting in the vast majority of cells produced in such cultures representing committed progenitors rather than HSCs. It is well established that the first in vitro cell division of LT-HSCs in mouse BM requires prolonged cytokine stimulation (26, 27) and we have previously demonstrated that by 3 days of combined stimulation with the early acting cytokines SCF, Flt3 ligand, TPO, and IL-3, all LSKCD34⁻ HSCs have undergone their first cell division (20, 25). Thus, in the present studies, we used the same ex vivo self-renewing conditions for 6 days (Fig. 4), to ensure that all HSCs have divided at least once, and found that expanding cells in these expansion cultures re-enter G₁ but not G₀ (Fig. 5,A). Subsequently, using the viable DNA dye Hoechst 33342, ex vivo-expanded cells were sorted into highly purified G₁ and S-G₂-M populations (11) (Fig. 5, B and C) and evaluated for their progenitor and HSC activities.

Whereas G₁ and S-G₂-M populations contained indistinguishable frequencies of myeloid progenitor cells (Fig. 5,D), G₁ cells were most highly enriched in LT-HSC repopulating activity (Fig. 5, E and F), as evaluated using an in vivo competitive repopulation assay (28), where we found as much as a 10-fold enrichment of RUs in the G₁ compared with the S-G₂-M fraction (21) (Table I). Notably, there was less enrichment of ST-HSCs from ex vivo-expanded LSKCD34⁻ cells; 3-wk engraftment levels were only 2-fold higher from the G₁ compared with the S-G₂-M fraction (Fig. 5 E).

In agreement with previous studies, detectable LT-HSC activity was present in the S-G₂-M population (11) (Fig. 5,E). Because BM HSCs in S-G₂-M have been suggested to be severely compromised in their repopulating ability (10, 11, 12, 13, 14), we addressed whether the ST- and LT-HSC activity of transplanted S-G₂-M cells could be due to contaminating G₁ cells by transplanting a low number of G₁ cells corresponding to the impurities of G₁ cells in FACS-sorted S-G₂-M cells (Fig. 5,C). Importantly, the marginal reconstituting activity of this low number of contaminating G₁ cells (Fig. 5,E), demonstrate that ex vivo-expanded adult LSKCD34⁻ cells residing in S-G₂-M in fact contain ST-HSC and LT-HSCs, which can reconstitute lethally irradiated recipients. In further support of this, S-G₂-M HSCs showed similar long-term contributions to B, T, and myeloid lineage reconstitution as HSCs residing in G₁ (Fig. 5 F) and were also able to self-renew, as demonstrated through their ability to reconstitute secondary recipients (J. M. Nygren, unpublished data).

Although our initial experiments clearly demonstrated that transplantable ex vivo self-renewing adult HSCs reside in G₁ as well as S-G₂-M phases of the cell cycle, LT-HSCs capable of reconstituting lethally ablated recipients remained highly enriched in G₁. This was in agreement with previous studies (12, 14, 15, 18) and has been proposed to reflect that S-G₂-M HSCs less efficiently than G₁ HSCs home and/or engraft upon transplantation. Alternatively or additionally, it could reflect a prolonged transit of ex vivo expanding LT-HSCs through G₁, similar to that of physiologically expanding fetal liver HSCs. If so, this would unavoidably result in an accumulation of HSCs in the G₁ fraction of the cell cycle, when compared with progenitors and ST-HSCs with a shorter G₁ transit, which therefore rather accumulate in S-G₂-M.

If the reduced HSC repopulating ability in S-G₂-M would reflect a truly compromised HSC engrafting ability, it should be possible to improve their reconstitution potential by promoting their re-entry into G₀-G₁. Thus, G₁ and S-G₂-M fractions were purified after 6 days of ex vivo expansion of LSKCD34⁻ cells, recultured under identical conditions for an additional 6 days to promote additional HSC self-renewing cell divisions, and then investigated for changes in HSC activity upon redistribution of cells into the different phases of cell cycle (Fig. 4). Notably, G₁ and S-G₂-M populations purified from primary (1°) expansion cultures showed comparable cellular expansion in secondary (2°) cultures (J. M. Nygren, unpublished data), displayed indistinguishable cell cycle profiles after 6 days of 2° expansion (Fig. 6,A), and as in primary cultures contained similar frequencies of myeloid progenitors (Fig. 6,B). However, despite the acquisition of indistinguishable cell cycle distribution profiles following 2° expansion, with most cells now residing in G₁ (Fig. 6,A), competitive transplantation of cells from the unfractionated 2° expansion cultures showed that the differences in ST-HSC and LT-HSC repopulating activities were retained (Fig. 6, C and D, and Table I). Thus, re-entry of S-G₂-M HSCs into G₁ did not to affect their long-term repopulating ability.

To reconcile our inability to enhance the repopulating activity of S-G₂-M cells by promoting their transit to G₁, with the predominant localization of HSCs in G₁ during active proliferation, we hypothesized that actively in vitro proliferating HSCs might have a significantly longer cell cycle (G₁) transit than committed progenitors, as demonstrated for fetal liver HSCs. To test this hypothesis, PKH26 was used to label HSCs that had already undergone at least one division ex vivo, as indicated by decreased CFSE levels (Fig. 7 A), to test the timing of subsequent cell divisions. Because PKH26 is distributed proportionally among daughter cells following mitosis (29), it allows high-resolution cell division tracking.

Following two additional days of ex vivo expansion, cells that continued to divide with rapid (PKH26^low) and slower (PKH26^high) kinetics could be distinguished (Fig. 7,B). These were FACS sorted to high purities (Fig. 7,C) and transplanted in competition with 200,000 BM cells into lethally irradiated recipients for evaluation of their total (Fig. 7,D) and myeloid (Fig. 7 E) long-term reconstitution potentials. Only the PKH26^high (slowly dividing) cells contained LT-HSC activity, whereas PKH26^low (rapidly dividing) cells provided only short-term multilineage reconstitution, despite the fact that 3- to 5-fold more PKH26^low cells were transplanted. These findings are compatible with ex vivo self-renewing adult LT-HSCs having slower cell cycle kinetics, and therefore preferentially reside in the G₁ phase of the cell cycle.

Whereas it was previously believed that only a very small fraction of mouse LT-HSCs cycle actively in steady-state adult BM (30), more recent studies have suggested that all adult LT-HSCs cycle within a period of 4–8 wk (8, 9). In the present studies, we investigated the cell cycle kinetics of HSCs in the 14.5 dpc fetal liver at a stage of normal development when HSCs expand extensively (4). Our studies demonstrate that fetal liver HSCs, although all dividing within as short a time as 48 h, have extensively prolonged cell cycle kinetics when compared with fetal HPCs. This striking difference in cell cycle dynamics reflects that HSCs pass through a stage of relative quiescence (Ki67^low/neg) and prolonged G₁, translating into a doubling in the cell cycle transit time when compared with HPCs. Such G₁ prolongation might reflect a specific G₀ state entry of HSCs, a notion supported by recent findings that long-term repopulating HSCs in adult bone marrow are highly enriched in G₀ compared with G₁ (16). Thus, although mechanisms must be in place to dramatically enhance the speed of HSC expansion at this stage of development, opposing mechanisms are obviously distinctly prolonging the HSC cell cycle transit time.

Purified adult BM HSCs have been demonstrated to need considerably longer time to complete the first cell division ex vivo when compared with more committed progenitors. However, subsequent ex vivo divisions have been assumed to have a G₁ transit time comparable to progenitors (13, 26, 27), but as current ex vivo expansion conditions, dramatically expand cell numbers but only maintain or at best slightly expand LT-HSCs over time (19, 20, 31, 32, 33, 34, 35, 36), the vast majority of cells are committed progenitors rather than LT-HSCs, complicating the interpretation of such experiments. Furthermore, our finding of considerable prolonged cell cycle kinetics of fetal HSCs expanding under physiological conditions raised the possibility that also actively proliferating adult HSC might have cell cycle kinetics distinct from those of adult progenitor cells. Thus, herein we took a number of functional approaches, to directly establish the cell cycle kinetics of adult LT-HSCs actively proliferating ex vivo (20). We were able to demonstrate that LT-HSCs, not only in steady-state adult BM (8, 9), but unexpectedly also during active proliferation reside predominately in G₁ of the cell cycle. We also found for the first time that purified S-G₂-M cells from ex vivo-expanded HSC cultures, upon re-entering G₁ in reculture, fail to regain repopulating activity and that purified G₁ cells, upon cell culture-induced S-G₂-M re-entry, did not lose significant HSC potential.

Our data would be best compatible with LT-HSCs having a prolonged cell cycle transit, such that the primary sort of S-G₂-M cells mainly consisted of short-term stem/progenitor cells. To obtain further and direct support for this, a second line of experiments was performed using advanced viable cell division tracking. Through this approach, we obtained direct evidence that the cell cycle transit time of ex vivo self-renewing LT-HSCs is prolonged compared with their committed progeny, not only during the first cell division but also subsequent cell divisions. Of interest, our data also suggest that ST-HSCs have a prolonged cell cycle transit relative to that of committed progenitors, although shorter than that of LT-HSCs. The prolonged cell cycle transit of LT-HSCs must primarily be due to a slower progression through the G₁ phase, as we observed an accumulation of proliferating LT-HSCs specifically in G₁. This is in agreement with G₁ being the cell cycle phase that is most regulatable in time (37).

Previous studies have suggested that the expression of adhesion molecules and BM homing properties of HSCs fluctuate with cell cycle passage (12, 38). Although our studies support that cells in all phases of the cell cycle phase can repopulate lethally irradiated mice, and that the main explanation for enrichment of repopulating HSCs in G₀-G₁ is the prolonged G₀-G₁ transit of HSCs, it is important to emphasize that this does not exclude the possibility that HSCs in S-G₂-M might have somewhat reduced repopulating ability, perhaps reflecting changes in adhesion molecule expression or function. However, when performing intrafemoral transplantations, we have observed the same differences in repopulating ability between S-G₂-M and G₀-G₁ cells, as when transplanted i.v. (J. M. Nygren, unpublished observations).

Our studies point to the importance in evaluating LT-HSCs by functional (long-term competitive repopulation) rather than phenotypical (cell surface marker expression) methods. Consequently, the conclusions from our studies could only be reached by identifying and evaluating LT-HSCs through their function, because there is no reliable phenotype of LT-HSCs following in vitro and in vivo expansion (39, 40). Thus, previous studies implicating cell cycle-specific effects on LT-HSC engraftment (10, 12, 13, 14) as well as gene expression (41), should be re-evaluated in light of the fact that functionally defined LT-HSCs, due to a prolonged cell cycle transit, always are enriched in G₁ relative to S-G₂-M.

In conclusion, the prolonged transit through G₀-G₁ is not limited to HSCs in steady-state adult BM but is a defining and developmentally conserved stem cell property. We propose that this distinct regulation of cell cycle transit in HSCs (and potentially other somatic stem cells) could be a requirement for their self-renewal (42) and to avoid exhaustion of the LT-HSC compartment by limiting their proliferative capacity (43, 44). Thus, it will be of considerable importance to identify the unique regulatory mechanisms of HSC cell cycle transit during development as well as in steady-state adult hemopoiesis.

We thank Lilian Wittman for expert technical assistance; the FACS Facility at Lund Strategic Research Center for Stem Cell Biology and Cell Therapy for assistance with cell sorting; and Drs. Connie Eaves, Norman Iscove, and Jennifer Antonchuk for helpful discussions and review of the manuscript.

The authors have no financial conflict of interest.