Insulin-like growth factor 2 expressed in a novel fetal liver cell population is a growth factor for hematopoietic stem cells

In vivo stromal cells form a complex microenvironment for hematopoietic stem cells (HSCs) that controls their multiple fates, including apoptosis and migration as well as the cell divisions that lead to formation either of daughter HSCs or of lineage-committed progenitors that are capable of limited proliferation.^1,2  Culture of HSCs in vitro, away from their natural microenvironment, leads to apoptosis or differentiation but not HSC self-renewal.

Genetic studies in mice, together with studies in which purified proteins were added to cultures of partially purified HSC populations, have led to the identification of secreted proteins that, in combination with others, maintain or expand the number of bone marrow HSCs. In mice, mutations either in stem cell factor (SCF) or its receptor, the protein tyrosine kinase Kit, lead to a profound reduction of HSC activities and numbers.^3,4  In vitro, SCF can prevent apoptosis of HSCs and is a growth factor for primitive lymphoid and myeloid hematopoietic progenitor or stem cells.^5,6  Both TPO^-/- mice and mice with mutations in its receptor, mpl, have a marked reduction in HSC activity,⁷  and thrombopoietin (TPO) can promote survival⁸  and expansion^9,10  of bone marrow HSCs in culture. Addition of interleukin-11 (IL-11) together with SCF or Flt-3 ligand (FL) maintains and expands bone marrow HSCs.¹¹  Flt-3 ligand (FL) receptor-deficient mice have many deficiencies in hematopoiesis,⁶  but elimination of the IL-11 receptor does not affect HSC numbers.¹²  FL has been included in numerous culture systems for HSCs,⁶  and several papers suggested that the FL receptor is expressed in some fetal liver long-term (LT)–HSCs but not in bone marrow LT-HSCs.^13-15 

Addition of fibroblast growth factor-1 (FGF-1) resulted in robust expansion of repopulating HSCs when total bone marrow cells were cultured in serum-free medium.¹⁶  Because purified HSCs cannot be expanded using FGF-1, the exact nature of the FGF-1 target cell population remains to be determined.¹⁶  Recently, purified and lipid-modified Wnt3a protein, when added in combination with fetal bovine serum and SCF, enhanced ex vivo expansion of purified bone marrow c-Kit⁺Thy-1.1^loSca-1⁺Lin^- hematopoietic progenitor cells.^17,18  However, mouse repopulation was studied only at 4 or 6 weeks after transplantation. Thus, it is not known whether Wnt3a supports expansion of long-term repopulating stem cells.

Although fetal liver HSCs have a greater proliferative potential than adult bone marrow HSCs and undergo significant expansion in vivo,^19-21  attempts to expand fetal liver HSCs ex vivo have failed.¹¹  We hypothesized that unknown growth proteins are produced by as yet unidentified populations of fetal liver cells that stimulate the expansion of fetal liver HSCs. Here we studied the effects of differentiated fetal liver hematopoietic cells on HSCs cultured ex vivo and identified a novel cell population, day-15 fetal liver CD3⁺ cells that support expansion of HSC numbers in culture. We show that insulin-like growth factor 2 (IGF-2) is specifically produced by fetal liver CD3⁺ cells and is the key molecule produced by these cells that stimulates HSC expansion. Furthermore, when combined with other growth factors, IGF-2 is capable of markedly enhancing ex vivo expansion of long-term repopulating fetal liver and adult bone marrow HSCs. Thus, IGF-2 is a novel growth factor for HSCs.

Lin^-Sca-1⁺Kit⁺ or side population (SP) cells were cultured alone or cocultured with the same number of fetal liver CD3⁺ cells as indicated, at a density of 50 Lin^-Sca-1⁺Kit⁺ or SP cells per 30 μL in Iscove modified Dulbecco medium (IMDM) with 10% fetal bovine serum (FBS) (StemCell Technologies, Vancouver, BC), 1% bovine serum albumin (BSA) (Sigma, St Louis, MO), and 50 μM β-mercaptoethanol (Invitrogen, Carlsbad, CA) in a well of a U-bottom 96-well plate. Murine IGF-2 (R&D Systems, Minneapolis, MN), 20 ng/mL SCF (R&D Systems), 30 ng/mL Flt-3 ligand (R&D Systems), and 10 ng/mL IL-6 (R&D Systems), or 50 ng/mL SCF and 100 ng/mL TPO (R&D Systems), as well as anti–IGF-2 neutralizing antibody (R&D Systems) were used as indicated.

Donor fetal liver or bone marrow cells were isolated from C57BL/6 CD45.2 mice. To sort Lin^-Sca-1⁺Kit⁺ cells, nucleated cells were incubated with rat serum for 15 minutes at 4°C at a density of 5 × 10⁷/mL. Cells were then stained with a biotinylated Lin⁺ antibody cocktail (for fetal liver cells, anti-CD3, anti-CD5, anti-B220, anti–Gr-1, anti-Ter119, and anti–7-4; for bone marrow cells, anti–Mac-1 was also included; StemCell Technologies and BD Pharmingen, San Diego, CA) for 15 minutes at 4°C. Cells were washed and incubated for 20 minutes with streptavidin-allophycocyanin (streptavidin-APC), anti-c-Kit–phycoerythrin (anti-c-Kit–PE), and anti-Sca-1–fluorescein isothiocyanate (anti-Sca-1–FITC). After 2 washes the cells were isolated by fluorescence-activated cell sorting (FACS) in a MoFlo cell sorter. Fetal liver CD3⁺ cells were sorted after staining with the anti-CD3–PE antibody. Adult mouse bone marrow SP cells were isolated essentially as described.²² 

To sort IGF-2–hFc⁺ and IGF-2–hFc^- cells, 10⁷ cells were resuspended in 1 mL conditioned medium (with about 1 μg/mL IGF-2–hFc determined by Western blotting) at 4°C for 30 minutes, followed by staining with anti-human immunoglobulin G1 (IgG1) Fc-PE (Jackson ImmunoResearch, West Grove, PA). The conditioned medium from mock-transfected cells was used as control. When necessary, the costaining of biotinylated Lin⁺ antibody cocktail followed by streptavidin-APC and anti-Sca-1–FITC was performed before sorting Lin^-Sca-1⁺IGF-2–hFc⁺ and Lin^-Sca-1⁺IGF-2–hFc^- cells.

For reconstitution analysis, peripheral blood cells were collected by retro-orbital bleeding, followed by lysis of red blood cells and staining with anti-CD45.2–FITC, anti-CD45.1–PE, anti-Thy1.2–PE, anti-B220–PE, anti-Mac-1–PE, anti-Gr-1–PE, or anti-Ter119–PE monoclonal antibodies (BD Pharmingen). When FACS analysis was used for characterizing CD3⁺ cells, day 15 fetal liver cells or adult splenic cells were stained with isotype control, anti-CD3–FITC, anti-CD4–PE, anti-CD8–PE, anti-Thy1.2–PE, and anti-Ter119–PE monoclonal antibodies (BD Pharmingen). FACS analyses were performed on a FACSCalibur instrument.

Both activity assays and limiting dilution assays were performed using the competitive reconstitution protocol. Indicated numbers of CD45.2 donor cells were mixed with 2 × 10⁵ (1 × 10⁵ used in Figures 4A and 6B) freshly isolated CD45.1 competitor bone marrow cells, which were then injected intravenously via the retro-orbital route into a group of CD45.1 mice previously irradiated with a total dose of 10 Gy. To measure reconstitution of mice undergoing transplantation, peripheral blood was collected by retro-orbital bleeding at the indicated times after transplantation, and the presence of CD45.1⁺ and CD45.2⁺ cells in lymphoid and myeloid compartments was measured.

In limiting dilution assays, the frequency of repopulating cells (competitive repopulating units [CRUs]) was estimated on the basis of Poisson statistics as described¹¹ ; the ranges of CRUs are given as 95% confidence intervals. For the experiment in Figure 6D, because the numbers of injected cells were insufficient for limiting dilution assays, the frequencies of HSCs were calculated as follows. If n = unknown frequency of HSCs in the population and m = injected cell number, the probability of a mouse not repopulated by the donor cells is P = (1 -n)^m. For a group of mice in Figure 6D, the P value, which is the ratio of the number of nonchimeric mice to the total number of mice, and m, which is the number of injected cells, are both known and n can be calculated. When all the recipient mice were repopulated, we could not calculate the actual n value; instead we calculated a mock n value by assuming that 1 of the recipient mice was not repopulated. The actual 1/n value was represented as less than 1 per mock n value.

Fifteen micrograms of total RNA and cRNA were prepared for hybridization to Affymetrix U74Av2 mouse chips according to manufacturer's instructions. The analysis was performed as detailed in our recent paper.²³ 

Reverse transcriptase–polymerase chain reaction (RT-PCR) was performed on cDNAs isolated from FACS-sorted murine day 15 fetal liver Lin^-Sca-1⁺Kit⁺ cells, CD3⁺ cells, CD3⁺Ter119⁺ cells, CD3⁺Ter119^- cells, IGF-2–hFc⁺ cells, and IGF-2–hFc^- cells to check expression of IGF-2, IGF type 1 receptor (IGF1R), IGF type 2 receptor (IGF2R), insulin receptor (IR), and glyceraldehydes-3-phosphate dehydrogenase (GAPDH). Forward primer for IGF-2: GTCGATGTTGGTGCTTCTCA; reverse primer for IGF-2: AAGCAGCACTCTTCCACGAT. Forward primer for IGF1R: CAAGCTGTGTGTCTCCGAAA; reverse primer for IGF1R: TGATGAGATCCCGGTAGTCC. 5′ primer for IGF2R: GCACCAAGATGAAGCAGTCA; reverse primer for IGF2R: ACATCCGGTAGCTGTTGGTC. Forward primer for IR: AAAGTTTGCCCAACCATCTG; reverse primer for IR: GTGAAGGTCTTGGCAGAAGC. Forward primer for GAPDH: AACTTTGGCATTGTGGAAGG; reverse primer for GAPDH: ACACATTGGGGGTAGGAACA. The level of murine IGF-2 in the medium of cultured cells was measured by enzyme-linked immunosorbent assay (ELISA) as described in the DuoSet ELISA Development Kit for mouse IGF-2 (R&D Systems).

The DNAs encoding Met1-Glu91 of human prepro–IGF-2 and Pro100-Lys330 of human IgG1 Fc sequences were linked by a DNA sequence encoding IEGRMD linker peptide. The whole 1.1-kilobase (kb) fragment was inserted into between KpnI and XhoI sites of pcDNA3.1, downstream of cytomegalovirus (CMV) promoter. The plasmid was transfected into 293T cells using lipofectamine 2000 (Invitrogen), and conditioned medium at 48 hours after transfection was collected. The conditioned medium with about 1 μg/mL IGF-2–hFc determined by Western blotting was used in the subsequent staining of fetal liver or bone marrow cells.

Because fetal liver HSCs undergo significant expansion during fetal development, we hypothesized that certain lineage-positive (Lin⁺) cells might produce protein(s) that support HSCs. A medium containing fetal calf serum, SCF, FL, and IL-6 cannot in itself support the maintenance of purified fetal liver HSCs,²⁴  presumably because it is missing some essential growth factor(s). In preliminary studies we confirmed this by showing that culture of day-15 fetal liver Lin^- cells in this medium for 3 or 6 days resulted in loss of HSC activity. However, adding day-15 fetal liver Lin⁺ cells in this culture system for 3 days supported a net 3-fold expansion in the number of in vivo repopulating HSCs (data not shown), suggesting that some Lin⁺ cells are expressing HSC supportive growth factor(s).

To identify the Lin⁺ cell population(s) that support HSC ex vivo expansion, we fractionated day-15 fetal liver cells by flow cytometry into 4 populations: CD3⁺, B220⁺, Gr-1⁺, and Ter119⁺, which were then individually cocultured with Lin^- cells. Only fetal liver CD3⁺ cells supported increases in HSC numbers and activity during a 3- or 6-day culture; the activity of HSCs was maintained from day 3 to day 6 and dropped after 6 days of culture (data not shown and Figure 1). Controls showed, as expected, no hematopoietic reconstitution ability in either freshly isolated or 3-day–cultured day 15 fetal liver CD3⁺ cells (10⁴ cells tested) 1 to 8 months after transplantation.

The limiting dilution experiment in Figure 1A shows that day-15 fetal liver CD3⁺ cells directly stimulate ex vivo expansion of purified LT-HSCs during a 3-day coculture. The CRU frequency for freshly isolated fetal liver Lin^-Sca-1⁺Kit⁺ (LSK) cells is 1 per 50 (95% confidence interval for mean, 1/36 to 1/57; n = 41). That is, as calculated from Poisson statistics, injection on average of 50 freshly isolated LSK cells is sufficient to repopulate 63% (1 -1/e) of mice undergoing transplantation. After culture in medium containing serum, SCF, FL, and IL-6, with or without CD3⁺ cells, the number of cells was too few to be counted reliably. However, after a 3-day culture without CD3⁺ cells, all HSC activity was lost, as judged by the fact (“▴” data point) that all 10 mice undergoing transplantation with the progeny of 100 input LSK cells failed to reconstitute (less than 1% donor contribution to nucleated blood cells; Figure 1).

After a 3-day culture of LSK cells with an equivalent number of day-15 fetal liver CD3⁺ cells, the number of functional LT-HSCs had increased about 2.5-fold (P < .05, Student t test), as evidenced by the fact that 63% of the mice undergoing transplantation with the progeny of, on average, only 22 initial LSK cells became reconstituted. In other words, the CRU frequency was 1 per 22 input equivalent LSK cells (95% confidence interval for mean, 1/13 to 1/26; n = 35). Figure 1B shows that fetal liver LSK cells cocultured with CD3⁺ cells repopulated, 4 months after transplantation, all blood lineages we tested (Figure 1B), attesting to net expansion of LT-HSCs.

At day 15, CD3⁺ cells represent 2% to 3% of total fetal liver cells. Although they are nonadherent to plastic and remain in suspension during culture, they are very different from CD3⁺ cells found in adult spleen or bone marrow (Figure 2A and data not shown). CD3⁺ cells isolated from adult spleen are mature T cells; they are Thy1.2⁺ (Figure 2Axi) and either CD4⁺ or CD8⁺ (Figure 2Aix-x). Most of them are Ter119^- (Figure 2Axii). Fetal liver CD3⁺ cells, in contrast, are Thy1.2^-CD4^-CD8^- (Figure 2Aiii-v), and half are Ter119⁺ (Figure 2Avi). In addition, the level of CD3 expression on fetal liver CD3⁺ cells is less than that on adult splenic CD3⁺ cells (Figure 2A, compare plots ii and viii). Fetal liver HSCs cocultured with CD3⁺ cells isolated from adult spleen or bone marrow lose most if not all repopulation ability (data not shown).

In an attempt to determine what factor(s) are made by day-15 fetal liver CD3⁺ cells that support HSC expansion, we compared the transcriptional profile of day-15 fetal liver CD3⁺ cells with those of 2 negative cell populations: adult splenic CD3⁺ cells and day-15 fetal liver Gr-1⁺ cells. Table 1 shows that several potentially important secreted and membrane proteins are differentially expressed, including IGF-2, Dlk1-like, CD59, and PRNP. Dlk1-like was previously shown to support maintenance of fetal liver HSCs in long-term culture.²⁵  Because IGF-2 was highly expressed in fetal liver CD3⁺ cells and not by adult CD3⁺ cells or fetal liver Gr-1⁺ cells, we hypothesized that this protein, not previously implicated in HSC biology, might mediate expansion of HSCs.

Figure 2B shows that IGF-2 as well as the 3 known IGF-2 receptors²⁶ —IGF type 1 receptor (IGF1R), IGF type 2 receptor (IGF2R), and insulin receptor (IR)—are expressed in fetal liver CD3⁺ cells. In contrast, fetal liver LSK cells express all 3 IGF-2 receptors but do not contain detectable amounts of IGF-2 mRNA.

About half of the day-15 fetal liver CD3⁺ cells are Ter119⁺ (Figure 2A, plot 6), which is a marker for terminally differentiating erythroblasts and mature erythrocytes. Figure 2D shows that the CD3⁺Ter119^- cells showed a much better ability than their CD3⁺Ter119⁺ counterparts to support the in vivo repopulating activity of LSK cells cocultured for 3 days (Figure 2D, compares lanes 3 and 2, P < .05). Thus, the HSC supportive activity of day-15 fetal liver CD3⁺ cells lies within the Ter119^- subfraction.

Addition of neutralizing anti–IGF-2 antibody to the coculture of LSK and CD3⁺Ter119^- cells completely blocked expansion of HSC activity by the cocultured CD3⁺Ter119^- cells (Figure 2D, compare lanes 4 and 3, P < .005), suggesting that IGF-2 is a key factor produced by fetal liver CD3⁺Ter119^- cells to support HSCs in culture.

Interestingly, within the fetal liver CD3⁺ cell population, both CD3⁺Ter119⁺ and CD3⁺Ter119^- cells express IGF-2 mRNA (Figure 2B, right) whereas only the latter population supports HSC expansion. Thus, we measured the level of IGF-2 secreted by cultured fetal liver CD3⁺Ter119⁺ and CD3⁺Ter119^- cells using ELISA (Figure 2C). When cultured alone, both populations secreted moderate levels (25 to 30 ng/mL) of IGF-2 (Figure 2C, lanes 1 and 2). When cocultured with fetal liver LSK cells, which in themselves produce little IGF-2 (Figure 2C, lane 3), CD3⁺Ter119^- cells express significantly more IGF-2 (250 ng/mL) than their Ter119⁺ counterparts (40 ng/mL IGF-2) (Figure 2C, compare lanes 5 and 4). Taken together, these results strongly support the notion that IGF-2, produced by fetal liver CD3⁺Ter119^- cells, is a key factor that supports the growth of HSCs ex vivo and that exposure to HSCs increases IGF-2 production by these supportive cells.

While fetal liver LSK cells express receptors for IGF-2 (Figure 2B), only a minority of these cells actually are HSCs. Existing antibodies against IGF1R and IGF2R are suitable for applications such as Western blotting but not for flow cytometry. Therefore, we generated a fusion protein of prepro–IGF-2 and a human IgG1 Fc fragment (IGF-2–hFc) that could be used in flow cytometry to detect and isolate cells expressing receptors for IGF-2. To this end, a plasmid expressing the IGF-2–hFc was constructed (Figure 3A, top) and transfected into 293T cells. The IGF-2–hFc fusion protein was secreted and could be detected by antibodies against either human IGF-2 or human IgG1 Fc (Figure 3A, bottom panels). The 2 bands observed on the Western blots suggest that some of the IGF-2–hFc has undergone partial protease degradation.

As judged by Western blotting, we estimate that medium from transfected 293T cells contains about 1 μg/mL IGF-2–hFc. Figure 3B shows that, after incubation in this conditioned medium, about 36% of total fetal liver cells stain positive for IGF-2–hFc (Figure 3Biv), as do 76% (67.0% + 9.3%) of Lin^- fetal liver cells and 83% (9.3 of 11.2) of Lin^-Sca-1⁺ fetal liver cells (Figure 3Bv). As expected, there is no IGF-2 staining of total fetal liver cells incubated in control-conditioned medium (Figure 3Biii). Both Western blot and RT-PCR analyses showed that fetal liver cells unable to bind the IGF-2–hFc fusion protein express little or undetectable IGF1R or IGF2R while those that bind the IGF-2–hFc fusion protein express both (Figure 3C-D). The insulin receptor, which binds IGF-2 weakly, is expressed by both cell populations (Figure 3D). Therefore, IGF-2–hFc specifically binds to IGF1R and IGF2R expressed on the surface of fetal liver cells.

Figure 4 demonstrates that all fetal liver HSCs express receptors for IGF-2. In Figure 4A, total fetal liver cells were sorted according to their ability to bind the IGF-2–hFc fusion protein. Those that did not bind the fusion protein were unable to reconstitute these mice, either at 3 weeks (measuring primarily short-term [ST]–HSCs) or 6 months after transplantation, primarily detecting LT-HSCs. In contrast, total fetal liver cells arbitrarily sorted according to their weak (+) or strong (++) ability to bind the IGF-2–hFc fusion protein supported both strong short-term and long-term engraftment (Figure 4A). Similarly, 100 sorted fetal liver Lin^-Sca-1⁺ cells unable to bind the IGF-2–hFc fusion protein were unable to reconstitute irradiated mice at 6 months after transplantation while 50 Lin^-Sca-1⁺IGF-2–hFc⁺ cells supported significant long-term reconstitution (Figure 4B). Further experiments (data not shown) showed that even 1000 fetal liver Lin^-Sca-1⁺IGF-2–hFc^- cells contain no HSC activity. Thus, all fetal liver HSC activity resides in the Lin^-Sca-1⁺IGF-2–hFc⁺ but not in the Lin^-Sca-1⁺IGF-2–hFc^- population, confirming the results of Figure 4A that all long-term repopulating HSCs bind IGF-2.

When 50 fetal liver LSK cells were cultured in the presence of serum supplemented with SCF, FL, and IL-6 for 3 days, all HSC activity was lost, similar to the results in Figure 1 (Figure 5Ai). After culture in serum supplemented with SCF and TPO, some HSC activity was retained (Figure 5Aii), but only half of the animals were repopulated by the progeny of 50 LSK cells. However, when 500 ng/mL IGF-2 was added to either medium, significant increases of HSC activities were observed, and all mice undergoing transplantation exhibited chimerism (Figure 5A, compare lanes iii versus i, P < .05; lanes iv versus ii, P < .005). Impressively, the combination of SCF, TPO, and IGF-2 substantially increased HSC activity, and each of 4 recipient mice was strongly reconstituted by the cultured progeny of only 50 original LSK cells (Figure 5Aiv). This suggests that IGF-2 is capable of strongly stimulating expansion HSCs in culture. Surprisingly, IGF-1, which only binds to IGF1R but not the IGF2R, does not support HSC maintenance or expansion in our culture system (data not shown).

Figure 5B definitely shows that IGF-2 increases the number of fetal liver stem cells. Freshly isolated fetal liver LSK cells were directly transplanted or were cultured for 3 days in serum-containing medium supplemented with SCF, TPO, and IGF-2. The CRU frequency for freshly isolated LSK cells was 1 per 50 (95% confidence interval for mean, 1/40 to 1/57; n = 31). After culture in medium containing serum, 500 ng/mL IGF-2, SCF, and TPO, the number of cells was again too few to be counted reliably. However, the number of functional LT-HSCs had increased about 2-fold (50 vs 25; P < .05, Student t test), as evidenced by the fact that 63% of the mice undergoing transplantation with the progeny of, on average, only 25 initial LSK cells became reconstituted. In other words, the CRU frequency was 1 per 25 input equivalent LSK cells (95% confidence interval for mean, 1/9 to 1/50; n = 26). To our knowledge, this is the first report showing the expansion of purified fetal liver HSCs ex vivo.

IGF-2 is thought not to be ubiquitously expressed in adult mice, but Figure 6A shows that about 42% of total adult bone marrow cells do bind significant amounts of the IGF-2–hFc fusion protein (Figure 6A, plot 4), as do 41% (39.1% + 2.1%) of Lin^- bone marrow cells and 47% (2.1 of 4.5) of Lin^-Sca-1⁺ bone marrow cells (Figure 6Av). Total bone marrow cells that did not bind the IGF-2–hFc fusion protein reconstituted mice at 3 weeks but not at 6 months after transplantation. In contrast, total bone marrow cells arbitrarily sorted according to their weak (+) or strong (++) ability to bind the IGF-2–hFc fusion protein supported both strong short-term and long-term engraftment (Figure 6B). Similarly, 100 sorted bone marrow Lin^-Sca-1⁺ cells unable to bind the IGF-2–hFc fusion protein were unable to reconstitute irradiated mice 6 months after transplantation. In contrast, 50 Lin^-Sca-1⁺IGF-2–hFc⁺ cells supported significant long-term reconstitution (Figure 6C). Thus, like their fetal counterparts, all long-term repopulating bone marrow HSCs bind IGF-2.

Compared with freshly isolated bone marrow SP cells (Figure 6Di), SP cells cultured for 3 days in serum-containing medium with SCF, FL, and IL-6 but without IGF-2 maintained some stem cell activity (Figure 6Dii). Importantly, addition of 1 μg/mL IGF-2 increased the repopulating activity and the stem cell frequency (Figure 6D, compare lanes iii versus i and ii, P < .05). Based on the finding that 5 of the 9 mice injected with 50 freshly isolated SP cells exhibited significant chimerism (Figure 6Di), we estimate that the HSC frequency in fresh bone marrow is 1 per 62 SP cells. After culture with IGF-2 the frequency increases to more than 1 per 26 input equivalent cultured SP cells, representing at least a 2-fold increase in HSC numbers.

Here we describe a new population of nonadherent fetal liver Ter119^-CD3⁺ cells that are capable of supporting a 2.5-fold expansion of LT-HSCs in culture and that are different from the other known population of CD3⁺ cells, adult T cells. We further showed that fetal liver CD3⁺ cells specifically express high levels of IGF-2 and that IGF-2 is the key factor in these cells that supports HSCs in culture. Importantly, all fetal liver and adult bone marrow HSCs express IGF-2 receptors, and culture of day 15 fetal liver LSK cells with IGF-2 increased HSC numbers and activities. IGF-2 also supported ex vivo expansion of adult bone marrow HSCs. Thus, fetal liver CD3⁺ cells are a novel population that is capable of supporting HSC expansion, and IGF-2, produced by these cells, is a growth factor for fetal liver and adult bone marrow HSCs.

In vivo, stromal cells form a complex microenvironment for HSCs that controls their multiple fates, including apoptosis, migration, and the cell divisions that lead to formation either of daughter HSCs or of lineage-committed progenitors that are capable of limited proliferation.^1,2  Different types of stromal and other cell types can have either positive or negative effects on fates of HSCs.²⁴  Certain differentiated hematopoietic cells negatively regulate the ability of transplanted HSCs to repopulate the hematopoietic system.^27,28  Consistent with these findings, in a preliminary study we observed that fetal liver B220⁺, Gr-1⁺, and Ter119⁺ cells did not support maintenance of fetal liver HSCs in culture while CD3⁺Ter119^- cells did. Thus, differentiated hematopoietic cells are similar to stromal cells in that they provide both positive and negative signals to cocultured HSCs.

Fetal liver CD3⁺Ter119⁺ cells are mainly erythroid precursors (C.C.Z., M. Socolovsky, H.F.L., unpublished observations, April 2003), while the nature of the CD3⁺Ter119^- population is unknown. Although both CD3⁺ fractions express IGF-2, only CD3⁺Ter119^- cells are capable of supporting HSC activity in culture. Because both CD3⁺Ter119^- and CD3⁺Ter119⁺ populations express IGF-2, it is likely that CD3⁺Ter119^- cells also produce other factor(s) that facilitate IGF-2's function in supporting HSC activity. Alternatively, CD3⁺Ter119^- but not CD3⁺Ter119⁺ cells might physically interact with HSCs; this hypothesis is supported by our observation that the level of IGF-2 secreted by CD3⁺Ter119^- cells increases when cultured in the presence of HSCs. In addition, fetal liver CD3⁺Ter119⁺ cells might express negatively acting factors that counteract the positive effects of IGF-2.

Fetal liver CD3⁺Ter119^- cells are different from a CD3⁺ “facilitating” cell population in adult bone marrow that enhances allografts.²⁹  Fetal liver CD3⁺Ter119^- cells do not enhance engraftment when freshly cotransplanted with donor cells (data not shown). They lack expression of the classical CD8 and Thy-1 T-cell markers while facilitating cells are CD8⁺Thy-1⁺.²⁹  Furthermore, the function of the facilitating cells was mediated by the CD3 T-cell receptor (TCR) β-Fcp33 complex³⁰  while HSC supportive activity of fetal liver CD3⁺ cells is mediated by IGF-2.

Normally maternally imprinted, IGF-2 has been suggested to play important roles in embryonic but not adult growth and development.^26,31  Targeted disruption of IGF-2 leads to significant fetal growth retardation,^32,33  partially as a result of decreased placenta growth.³⁴  Although no role in HSC biology was previously identified, IGF-2 has been used to establish pluripotential cell lines³⁵  and maintain undifferentiated stem cells from the inner cell mass in culture.^35,36  IGF-2 stimulates growth and/or differentiation and inhibits apoptosis of a number of cell types including hematopoietic progenitors. While stimulating the growth of human bone marrow CD34⁺ cells,³⁷  IGF-2 also increases the formation of myeloid and erythroid progenitors such as granulocyte macrophage colony-forming units (CFU-GMs), erythroid burst-forming units (BFU-Es), and erythroid colony-forming units (CFU-Es)^37,38  and regulates T-cell development and stimulates T-cell proliferation.^39,40  However, this is the first report showing that IGF-2 is a growth factor for HSCs.

Different IGF-2 receptors have either a positive or a negative effect on cell proliferation. Two receptor protein tyrosine kinases, IGF1R and IR, mediate the mitogenic activities of IGF-2, while IGF2R mediates degradation of IGF-2. Lack of IGF1R results in fetal growth retardation, and IGF1R-deficient mice invariably die within minutes of birth.⁴¹  In contrast, mice genetically deficient in the negatively acting IGF2R show increased serum and tissue levels of IGF-2 that result in cell overgrowth, developmental abnormalities, and prenatal death.^42,43 

Multiple IGF-2 receptors could explain the fact that we need to add 500 ng/mL (67 nM) IGF-2 to support HSC expansion during culture of LSK cells; preliminary experiments (not shown) indicated that neither 25 ng/mL nor 50 ng/mL IGF-2 support cultured HSCs when combined with SCF, FL, and IL-6. Given the low nanomolar values of the dissociation constants for binding of IGF-2 to both the IGF1R and IGF2R,⁴⁴  it is possible that, at low concentrations, the balance of binding of IGF-2 to its 3 types of receptors triggers the HSC differentiation; hypothetically, only higher concentrations of hormone could effectively prevent apoptosis and promote self-renewal. This is consistent with the “threshold” model⁴⁵  positing that a low concentration of a growth factor stimulates stem cell differentiation but that high levels of the factor lead to a net expansion of stem cells.

The situation may be different when fetal liver CD3⁺Ter119^- cells are cocultured with HSCs. Because cells tend to closely interact in the U-bottomed wells we used, it is possible that CD3⁺Ter119^- cells physically contact HSCs and thus that the local concentration of IGF-2 around an HSC is much higher than in the rest of the medium. On the other hand, IGF2R⁺ cells might degrade IGF-2, and this would mandate a high initial concentration of hormone to support expansion of HSCs over a 3-day culture period.

Both SCF and TPO are required to maintain normal HSC levels, mainly by preventing their apoptosis.^5,8  IGFs prevent apoptosis and trigger proliferation of different types of cells, principally through both the mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3 kinase (PI-3K) pathways.⁴⁶  Here we showed that the combination of IGF-2 with SCF and TPO supports a more than 2-fold increase in HSC numbers. Most likely, when combined with SCF and TPO, IGF-2 is both antiapoptotic and mitogenic for HSCs. We are trying to determine whether IGF-2 affects apoptosis of HSCs or whether it acts by accelerating the rate at which HSCs enter the cell cycle. IGF-2 may also increase the proportion of the daughter cells that become HSCs.

In adult rodents, IGF-2 is reportedly not expressed and its functions are suggested to be replaced by IGF-1. However, we found that some non-HSC and stromal bone marrow cells express high levels of IGF-2 (C.C.Z., B. Luo, H.F.L., unpublished observations, July 2003). Therefore, IGF-2 likely is functionally important in the adult bone marrow to maintain homeostasis of adult stem cells, consistent with our observations that all adult bone marrow HSCs express receptors for IGF-2 and that IGF-2 supports expansion of bone marrow HSCs in culture.

In summary, we identified IGF-2 as a novel stem cell growth factor and for the first time achieved a 2- to 3-fold net expansion of fetal liver HSCs in the presence of IGF-2. Although rodent IGF-2 is thought to play important roles only in fetal development, we showed that IGF-2 also stimulated the ex vivo expansion of adult bone marrow HSCs. Because IGF-2 has pleiotropic effects, the expression of receptors for IGF-2 may define pan–stem cell properties, and IGF-2 may have the potential to act on a broader spectrum of stem cells, both ex vivo and in vivo.

We thank Dr Nir Hacohen for advice and insightful discussions on the project and Drs Biao Luo, Chang-Zheng Chen, Xiao-Wu Zhang, Wei Tong, Alec Gross, and Christopher Hug for reading of the manuscript and help in experiments. We are grateful for help with FACS provided by Glenn Paradis and Michael Doire at the Massachusetts Institute of Technology (MIT) flow cytometry core facility and the technical assistance by Sonali Rudra. We appreciate the kind gift of anti-IGF2R by Dr Stuart Kornfeld at Washington University.