Combinatorial Gata2 and Sca1 expression defines hematopoietic stem cells in the bone marrow niche

Abstract

The interaction between stem cells and their supportive microenvironment is critical for their maintenance, function, and survival. Whereas hematopoietic stem cells (HSCs) are among the best characterized of tissue stem cells, their precise site of residence (referred to as the niche) in the adult bone marrow has not been precisely defined. In this study, we found that a Gata2 promoter directs activity in all HSCs. We show that HSCs can be isolated efficiently from bone marrow cells by following Gata2-directed GFP fluorescence, and that they can also be monitored in vivo. Each individual GFP-positive cell lay in a G₀/G₁ cell cycle state, in intimate contact with osteoblasts beside the endosteum, at the edge of the bone marrow. We conclude that the HSC niche is composed of solitary cells and that adult bone marrow HSC are not clustered.

Stem cells possess two intrinsic capacities, self-renewal and multilineage differentiation (1–3). Potential clinical applications for stem cells have been widely touted, but the development of human therapeutic applications are intimately dependent on elucidating the basic mechanisms by which stem cells maintain this unusual dual capability. Although hematopoietic stem cells (HSCs) are among the best characterized of all pluripotent lineages, we do not know specifically where HSCs reside in the bone marrow.

Various transcription factors are key regulators of hematopoiesis, and their mutant expression often results in hematopoietic disease, including leukemia (4). Because transcription factor GATA-2 is expressed in HSCs (5, 6), we hypothesized that lineage marking of Gata2 expression might enable the purification from, and perhaps the in vivo visualization of, HSCs in living bone marrow.

Transcription factor GATA-2 is widely expressed and is required for multiple aspects of vertebrate embryogenesis (7–10). Two alternative promoters are used to transcribe the mouse Gata2 gene, a feature that is conserved in the human and chicken orthologs (8, 11). Transcription from the IS (I-specific) promoter is limited exclusively to hematopoietic and neural cells, whereas transcription from the IG (I-general) promoter is more widespread (8), and we previously found that IS-mediated transcription is preferred in human CD34⁺ (immature) hematopoietic progenitor cells (11). Within hematopoietic lineages, GATA-2 is most abundant in immature progenitors (5, 6, 12), and germ-line loss-of-function studies demonstrated that GATA-2 is essential for multilineage hematopoietic progenitor proliferation (5, 13). Furthermore, the Gata2 IS promoter alone can direct expression in a very early hematopoietic progenitor compartment (14, 15).

In this study, to visualize HSCs in the living bone marrow, we established GFP “knock-in” mutant lines in which the GFP cDNA was inserted into the Gata2 IS first exon by germ-line targeting. As expected, HSCs could be efficiently isolated by monitoring GFP expression from the knock-in mice. Real-time monitoring of Gata2-expressing HSCs in living bone marrow showed that individual GFP-positive HSCs are immobile, in intimate contact with osteoblasts beside the endosteum. Taken together, we conclude that the adult HSC niche is composed of solitary cells and that adult bone marrow HSCs are not clustered.

Results and Discussion

A GFP reporter gene was inserted into Gata2 exon IS by targeted mutagenesis (to create EIS-KI mice) to determine where GATA-2 activity was most abundant during early hematopoiesis (Fig. 1 A and B). At the same time, we established a second GFP-targeted mutant line that disrupted the first Gata2 coding exon (in exon II, referred to as EII-KI mutant mice, Fig. 1 A and B).

Fig. 1.

Generation of two murine lines with GFP inserted into the Gata2 locus. (A) Strategy of GFP knock-ins into exon IS or exon II of the Gata2 gene. The WT Gata2 gene (w) has two alternative untranslated first exons (IS and IG). The GFP gene with a polyadenylation signal sequence (pA) and Neo cassette were integrated into exon IS (ISĜFP:neo) or 5′ to the translational start site in exon II (GFP:neo). Neo cassettes were removed by Cre/loxP-mediated site-specific recombination (ISĜFP, GFP). (B) BglII plus NotI, BglII digested genomic DNA from each mouse line was hybridized with radiolabeled exon III probe as indicated in A. (C) Fluorescent images of EIS-KI and EII-KI whole-mount embryos at E11.5. Green fluorescence was observed only in the midbrain (mb) of EIS-KI embryos whereas, in contrast, EII-KI embryos expressed GFP in the placenta (pl), mesonephros (me), midbrain (mb), and the sensory organs (n, nose; e, ear). (D) Mononuclear cells recovered from EII-KI and EIS-KI bone marrow were stained with PE-conjugated antibodies for hematopoietic lineage markers and analyzed by flow cytometry. The percentages of each quadrangle are shown.

The EII-KI mutant was anticipated to result in embryonic lethality, because it should disrupt normal GATA-2 translation (10, 14). As anticipated, homozygous EII-KI mutant mice died at embryonic day 10 (E10) with severely reduced numbers of hematopoietic cells, in good agreement with the phenotype originally reported for Gata2-null mutant mice (Table 1 and ref. 13). In contrast, homozygous EIS-KI mutant mice were born in a normal Mendelian ratio and were fertile (Table 1). The level of GATA-2 transcript in EIS-KI mice was similar to that of WT littermates (data not shown), initially implying (which was subsequently confirmed, below) that EIS-KI mice did not suffer from any hematopoietic deficiency. These results demonstrate that, whereas GATA-2 is indispensable for proper development of the hematopoietic system, the contribution of the IS promoter to GATA-2 function is fully compensated by IG promoter activity.

Table 1.

Genotype of pups and embryos

Parents	Stage	Litter		Pups or embryos
(IS∧GFP/w) (IS∧GFP/w)	3–5 wks	6	(w/w) 12	(IS∧GFP/w) 23	(IS∧GFP/IS∧GFP) 14
(IS∧GFP/w) (GFP/w)	3–5 wks	5	(w/w) 19	(IS∧GFP/w) 18 (GFP/w) 20	(IS∧GFP/GFP) 18
(IS∧GFP/GFP) (IS∧GFP/w)	3–5 wks	3	(IS∧GFP/w) 6	(IS∧GFP/IS∧GFP) 5 (GFP/w) 4	(IS∧GFP/GFP) 6
(GFP/w) (GFP/w)	3 wks	3	(w/w) 6	(GFP/w) 14	(GFP/GFP) 0
	E11.5	4	(w/w) 10	(GFP/w) 24	(GFP/GFP) 0 (16)
	E10.5	5	(w/w) 8	(GFP/w) 16	(GFP/GFP) 9 (2)
	E9.5	4	(w/w) 16	(GFP/w) 22	(GFP/GFP) 18

Genotypes of pups and embryos were determined by PCR. The numbers of dead embryos are indicated in parentheses.

We then examined the expression of GFP in these mutant mice. Tissues expressing GFP in EII-KI heterozygous mutants (summarized in Table 2) are the same organs that are affected by Gata2 null mutation. For example, GFP was expressed at the urethra: bladder junction, in the placenta, and in adipose tissue, all organs where GATA-2 function has been shown to be indispensable (9, 16–18). Although the expression of GFP in EII-KI mice was visualized in virtually all of the anticipated Gata2-expressing tissues, GFP fluorescence in the EIS-KI mice was highly restricted to neural and hematopoietic cells (Fig. 1C and Table 2), demonstrating the fidelity of GFP expression to documented Gata2-IS promoter activity (14, 15).

Table 2.

Comparison of GATA-2 expression with GFP expression

Tissues	GATA-2 (RT-PCR)*			GFP (FM/IHC)†
	eIV/eV	eIS/eII	eIG/eII	EII-KI	EIS-KI
Embryonic
E7.5	+	−	+	Trophoectoderm, extraembryonic mesoderm	—
E9.5	+	±	+	mb, ec, ys, aorta, ur, pl, ht	mb
E11.5	+	+	+	liver, brain, sc, ec, ys, aorta, ur, pl, ht, ear, nose	Liver, brain, sc
E13.5	+	+	+	liver, brain, sc, ec, kidney, urogenital, pl, ht, ear, nose	Liver, brain, sc
Adult
Brain	+	+	+	n, ec	n
Thymus	−	−	−	ec	—
Heart	+	−	+	ec	—
Lung	+	+	+	ec, nd	nd
Liver	−	−	−	ec	—
Spleen	+	+	+	h, ec	h
Kidney	+	+	+	ec, tubular cell, nd	nd
Uterus	+	+	+	Endometrium, ec	Endometrium
Ovary	−	−	−	ec	—
Testis	+	−	+	ec, nd	—

*RT-PCR was performed using exon IS, IG, II, IV, and V Gata2 gene-specific primers (eIS, eIG, eII, eIV, and eV, respectively).

^†GFP expression in EIS-KI and EII-KI mice was detected by fluorescent microscope (FM) or immunohistochemistry (IHC) using anti-GFP antibody. mb, midbrain; ec, vascular endothelial cell; ys, yolk sac; ur, urogenital ridge; pl, placenta; ht, heart; sc, spinal cord; n, neural cell; h, hematopoietic cell; nd, not determinant cell.

Most studies distinguish between rare early hematopoietic progenitors by using a host of empirically defined cell surface markers, so we next analyzed the hematopoietic cells from these mutant mice by flow cytometry. EII-KI and EIS-KI mice expressed GFP in either 17.0% or 0.7% of total bone marrow cells, respectively (Fig. 1D). GFP was expressed in the majority of Lin (specific hematopoietic lineage)-negative cells in the bone marrow of EII-KI mice whereas, in contrast, GFP-positive cells represented only a very small fraction of Lin⁻ cells among those recovered from EIS-KI mouse bone marrow (Fig. 1D). These results indicated that GFP activity in EIS-KI mice is restricted to very immature hematopoietic progenitors that do not display any of the six cell surface markers that define the more mature hematopoietic lineages (see Materials and Methods). A similar enrichment was not observed in EII-KI GFP⁺ bone marrow cells, demonstrating the breadth of transcriptional potential of the IG promoter in the hematopoietic system. EII-KI-directed GFP expression was detected in both Ter119⁺ erythrocytes and B220⁺ B lymphocytes (Lin⁺ cells in Fig. 1D, and data not shown). Because GATA-2 mRNA was undetectable in these EII-KI-derived GFP⁺Lin⁺ cells by RT-PCR (data not shown), the data suggest that extended GFP perdurance (19) was responsible for the labeling of cells that no longer express GATA-2. We therefore did not further examine the EII-KI mice in these studies.

Many previous reports have shown that Lin-negative (L⁻) bone marrow cells that express the Sca1 and c-Kit markers (L⁻S⁺K⁺) are enriched in HSCs (20, 21). Intriguingly, 8% of the L⁻S⁺K⁺ cells recovered from the EIS-KI mouse bone marrow also express GFP (Fig. 2A, Lower Center). Another tantalizing observation was that the mean fluorescence intensity of GFP⁺L⁻S⁺K⁺ cells (G⁺L⁻S⁺K⁺) was ≈6-fold higher than in comparable cells lacking Sca1 (G⁺L⁻S⁻K⁺; Fig. 2A, Lower Center: mean green fluorescent intensity of G⁺L⁻S⁺K⁺ and G⁺L⁻S⁻K⁺ fractions were 980 and 170, respectively). Because all GFP⁺Sca1⁺ bone marrow cells also express high levels of c-Kit (Fig. 2A, Lower Left), the data demonstrate that G⁺L⁻S⁺K⁺ bone marrow cells can be effectively discriminated by following only the Sca1 and GFP markers.

Fig. 2.

High-level expression of GFP and GATA-2 mRNA in HSC of E1S-KI mice. (A) Lineage negative (Lin⁻) cells recovered from EIS-KI bone marrow were analyzed by flow cytometry after staining with c-Kit and Sca1 antibodies. Cells were also stained by Hoechst 33342, and blue and red fluorescence were measured (H. Blue and H. Red, respectively) to detect the side population (SP). The percentage of cells in each quadrangle is shown. (B) Cells in each quadrangle of the Lin⁻c-Kit⁺ fraction (Lower Left in A) were compared for levels of GATA-2 mRNA by quantitative RT-PCR (normalized to GAPDH). (C) The morphology of Lin⁻GFP⁺Sca1⁺c-Kit⁺ (G⁺L⁻S⁺K⁺) cells from EIS-KI bone marrow are shown after Wright-Giemsa staining. (Scale bar: 10 μm.) (D) Cell-cycle status of G⁺L⁻S⁺K⁺ and G⁻L⁻S⁺K⁺ cells. GFP⁺ and GFP⁻ cells in Sca1⁺c-Kit⁺ fraction of fixed EIS-KI bone marrow were analyzed for their DNA content by Hoechst 33342 flow cytometry. The percentages of cells in S/G₂/M-phases are shown.

When we quantified GATA-2 mRNA abundance in c-Kit⁺ subfractions that express or do not express Sca1 by quantitative RT-PCR, we found that GATA-2 mRNA abundance in the G⁺L⁻S⁺K⁺ cells was 4-fold higher than in the comparable Sca1⁻ fraction (Fig. 2B), but was undetectable in GFP-negative cells. These results demonstrate that the expression of GFP in the EIS-KI hematopoietic cells correlates well with endogenous GATA-2 mRNA accumulation, and thus that GFP is a good quantitative marker for GATA-2 expression in the bone marrow cells of EIS-KI mice.

Further detailed analysis revealed that HSCs are even more highly enriched in the G⁺L⁻S⁺K⁺ fraction than in the comparable GFP⁻ fraction of EIS-KI mouse bone marrow. HSCs are known to be abundant not only in L⁻S⁺K⁺ cells, but also in the “side population (SP)” flow cytometry fraction characterized by reduced Hoechst dye uptake (22). Although the SP fraction constitutes only a small proportion of total L⁻K⁺ cells (1.1%; Fig. 2A, Upper Center), 87.5% of G⁺L⁻S⁺K⁺ cells were found in the SP fraction (7% of the 8% of total GFP⁺ L⁻S⁺K⁺ cells; Fig. 2A, Upper Right). G⁺L⁻S⁺K⁺ cells were especially abundant in the low fluorescence tip of the SP fraction, which is also most enriched for HSC (22). CD34 is another well characterized cell surface antigen that is differentially expressed in immature hematopoietic cells, and mouse HSCs are highly enriched in CD34⁻L⁻S⁺K⁺ cells (20). To further refine the identity of HSC within the G⁺L⁻S⁺K⁺ cells, we found that 25% of the G⁺L⁻S⁺K⁺ cells were also CD34-negative (data not shown).

Morphologically, the G⁺L⁻S⁺K⁺ cells were quite uniform in appearance, with amorphous nuclei and a large nuclear to cytoplasmic ratio (Fig. 2C). More than 95% of G⁺S⁺ cells were found to be in the G₀/G₁ phase of the cell cycle, whereas <80% of the comparable GFP⁻ cells are in the same cell cycle phase (Fig. 2D). In summary, these data taken together indicate that a significant fraction of the GFP⁺Sca1⁺ cells (25%) within the EIS-KI mouse bone marrow are brightly fluorescent, uniform in appearance, Hoechst dye-excluding, noncycling, CD34⁻Lin⁻Sca1⁺c-Kit⁺ HSC (20, 22, 23).

To independently assess whether bona fide HSC were enriched in the GFP⁺ fraction of the bone marrow from EIS-KI mice, competitive long-term reconstitution assays were performed (20). Each lethally irradiated mouse was injected with either 4 × 10³ L⁻G⁺ or 2 × 10⁵ L⁻G⁻ cells based on their fractional representation in EIS-KI Lin⁻ bone marrow (Fig. 2A). The results clearly demonstrate the presence of long-term (8 months) reconstituting HSCs in both Lin⁻GFP⁺ and Lin⁻GFP⁻ fractions. Even though 50 times as many GFP⁻ cells were injected, the fractional chimerism of GFP⁺ recipients 8 months after transplant was much higher than that of the corresponding GFP⁻ recipients (Fig. 3A).

Fig. 3.

Long-term hematopoietic reconstitution activities of G⁺L⁻S⁺K⁺ and equivalent GFP⁻ cells. (A) The contribution of donor (Ly5.1⁺ EIS-KI mice)-derived myeloid (Mac⁺ and Gr1⁺, Left) and lymphoid (Thy1.2⁺ and B220⁺, Right) cells in lethally irradiated Ly5.2⁺ recipient mice was measured for 8 months after transplantation. Total Lin⁻ cells (2 × 10⁵, dotted lines/triangles), 4 × 10³ Lin⁻GFP⁺ (black circles), or 2 × 10⁵ Lin⁻GFP⁻ (gray boxes) cells were examined. (B) The frequency of long-term (5 months) reconstituting HSC in G⁺L⁻S⁺K⁺ (black circles) and comparable GFP⁻ (gray boxes) fractions was investigated by limiting dilution assay. The gray lines show the 37%-negative points based on a Poisson distribution. (C) Bone marrow cells recovered from primary recipients were divided into GFP⁺ and GFP⁻, and 2 × 10⁴ cells from each population were secondary transplanted. The chimerism of transplanted cells in the peripheral myeloid (Left) and lymphoid (Right) compartments after transplantation is shown.

The frequency of HSC in the G⁺L⁻S⁺K⁺ and comparable GFP⁻ fractions of EIS-KI bone marrow was determined by limiting dilution analysis of cells containing reconstitution activity (20, 21). The frequency of HSC was >10-fold higher in the GSK fraction (12.5%) than in an otherwise identical fraction lacking green fluorescence (1.1%; Fig. 3B). Five months after initial transplantation of the GSK cells, 2 × 10⁴ Lin⁻GFP⁺ or Lin⁻GFP⁻ cells derived from primary recipient mice were serially transplanted: only the GFP⁺ fraction was capable of sustaining hematopoiesis for >3 months in the secondary recipients (Fig. 3C). Based on this most stringent criterion for in vivo activity of HSC, we conclude that all adult bone marrow-derived HSC in the EIS-KI mice express GFP.

Because there are possible technical limitations to the long-term reconstitution assay (e.g., the ability of HSCs to home to the marrow niche), the frequency of HSC in the G⁺L⁻S⁺K⁺ fraction could be even higher than the percentage implied from the transplant data (Fig. 3B). Because the flow results indicated that 25% of the G⁺L⁻S⁺K⁺ cells are also CD34⁻ (data not shown), and because all serially transplantable HSCs in the EIS-KI bone marrow express GFP (and GATA-2), when taken together the data indicate that the G⁺L⁻S⁺K⁺ fraction accounts for between 0.05% and 0.10% of total bone marrow mononuclear cells, and that ≈20% of G⁺L⁻S⁺K⁺ cells are HSCs.

Because the G⁺L⁻S⁺K⁺ fraction represents a highly enriched HSC compartment in vivo (Fig. 3), and because virtually all Sca1⁺GFP⁺ cells in the EIS-KI bone marrow also express c-Kit (Fig. 2A), we hypothesized that murine HSCs might be readily visualized in EIS-KI mouse bone marrow simply by monitoring the coordinate expression of GFP and Sca1. Furthermore, because the GFP fluorescence intensity in G⁺L⁻S⁺K⁺ cells was 6-fold higher than in comparable Sca1⁻ cells (Fig. 2A), we suspected that GFP expression in those HSC should be distinguishable by fluorescence microscopy. To test this hypothesis, frozen sections from the trabecular bones of EIS-KI mouse femurs were stained with phycoerythrin (PE)-labeled anti-Sca1 antibody, and then the red (Sca1) and green (GFP/GATA-2) images were recorded. Through the examination of hundreds of sections, we were able to visualize a few GFP-positive cells (arrows in Fig. 4A and B). The anti-Sca1 antibody stains both HSCs and some endothelial cells (Fig. 4 A and B). The coincidence of cells that were both GFP⁺ and Sca1⁺ was extremely low, but a small number of GFP⁺Sca1⁺ cells were detected in close proximity to the trabecular bone (short arrows in Fig. 4 A and B). Immunohistochemical analysis of fixed sections of the bone marrow employing an anti-GFP antibody further revealed that all cells that abundantly express GFP (the brown cells in Fig. 4 C and D) were in direct contact with osteoblasts (purple, alkaline phosphatase-positive cells in Fig. 4C), in accord with earlier observations, thus supporting the notion that HSC and osteoblast association may be functionally significant (3, 24). GFP⁺ cells were never found adjoining the cartilage (Alcian blue in Fig. 4D). These results show that the endogenous HSC niche is near the endosteum and includes directly contacted osteoblasts.

Fig. 4.

HSCs are dormant and solitary in the bone marrow niche. (A and B) Frozen bone sections from EIS-KI mice were stained with PE-conjugated anti-Sca1 antibody. The green fluorescent images (Left) were merged with Nomarski and PE images (Right). A and B are photographs from independent mice. GFP or Sca1 single-positive cells are indicated with green (long arrows) or red, respectively. Yellow cells (short arrows) indicate GFP/Sca1 double-positive cells that reside by the edge of the bone marrow. The bone marrow sections were also stained with anti-GFP antibody (arrows, brown) and alkaline phosphatase (purple in C) or Alcian blue (D). GFP⁺ cells are in direct contact with purple-stained osteoblasts (C), but not with the Alcian blue⁺ cartilage (D). (E) Green fluorescent (Left) and bright field (Right) images of living bone marrow from the scapula of an EIS-KI mouse 1 day after 5-FU injection. GFP⁺ cells are detected near the edge of the bone marrow. (F and G). Two days after 5-FU administration, GFP⁺ cells are apparently quiescent, deep in the bone marrow. Time-lapse images of E, F, and G are exhibited in Movies 1–3. The movies demonstrate that GFP⁺ cells are almost completely immobilized in the bone marrow, whereas the surrounding GFP⁻ cells move vigorously.

We next examined the movement of HSCs in living bone marrow. To enrich for HSCs, EIS-KI mice were treated with 5-fluorouracil (5-FU) to eliminate more differentiated, actively dividing cells (25). As anticipated, the number of GFP⁺ cells diminished ≈5-fold after 5-FU treatment (data not shown). Of those few remaining slowly cycling or noncycling GFP⁺ cells, they invariably expressed Sca1 (data not shown), indicating that the GFP⁺ cells in their native residence in the marrow are exclusively noncycling, consistent with well established properties of HSC (25).

When we monitored the few remaining GFP⁺ cells within the 5-FU-treated marrow using a fluorescent microscope equipped with a time-lapse recording device, we found that those GFP⁺ cells residing at the bone marrow edge did not move, whereas in contrast the surrounding round (presumptive hematopoietic progenitor) GFP⁻ cells were very actively motile (Fig. 4 E–G and Movies 1–3, which are published as supporting information on the PNAS web site). Whereas almost all of the GFP⁺ cells are relatively inert, a few GFP⁺ cells can be observed to be more motile, although far less than the surrounding (presumptively more differentiated progenitor) control cells (Movies 1–3). We speculate that these more active GFP⁺ cells may be entering a differentiation or proliferation stage. These data further suggest that HSCs might be maintained in a quiescent state by fixation in the niche through specific cell surface interactions (e.g., by selectin, cadherin, chemokine, or δ/notch adhesion; ref. 26)

In summary, the data presented here demonstrate that detailed examination of Gata2 promoter activity enabled the purification of HSCs from the bone marrow of EIS-KI mice to high purity. The combination of GFP expression from the Gata2-IS promoter with Sca1 antibody staining allowed visualization of HSC staining in situ, where they were found to reside quietly near the endosteum in direct contact with osteoblasts in the bone marrow (3, 24).

Among GFP⁺ bone marrow cells, there is clearly a population of 75–85% that are not HSCs; however, we would emphasize the following: (i) all EIS-KI-derived bone marrow HSC express brightly fluorescent GFP; and (ii) EIS-KI GFP⁺ progenitor cells express low-level GATA-2 and are 6-fold lower in fluorescence intensity (Fig. 2). Thus, we conclude that the estimate of 10–25% purity of HSC among the G⁺L⁻S⁺K⁺ bone marrow cells (evidence from flow cytometry and long-term reconstitution assays) is in perfect harmony with the data demonstrating that only 20% of (brightly fluorescent) GFP⁺ cells remain after 5-FU elimination of cycling cells (data not shown). One might further argue that any GFP⁺ (dull) progenitor cell population would not be distinguishable above the background fluorescence visualized in living bone marrow, but we have no evidence to directly support this contention.

Regardless of whether or not the microscopic images (Fig. 4) most often depict progenitors or, at a lower (10–25%) frequency, bona fide HSC, these data show that there is always only one brightly fluorescent GFP⁺ cell in a given niche, in contrast to the idea that stem cells cluster in the niche [ref. 1; e.g., in the mammalian epidermis (27) or the Drosophila germ line (28)]. Given that all E1S-KI-derived bone marrow cell HSC express brightly fluorescent GFP and Sca1, these data demonstrate a property that is (thus far) unique to stem cells of the hematopoietic lineage.

In the adult bone marrow of EIS-KI mice, all HSCs express GFP. GFP⁺ cells were not detected in the embryonic hematopoietic progenitor compartment, where endogenous GATA-2, as well as EII-KI GFP, are expressed (Table 2 and data not shown). These data indicate that promoter usage in the Gata2 gene differs between fetal and adult HSCs; that is, the IS promoter is transcribed only in adult HSC, but not at the fetal stage. In contrast, however, our previous transgenic reporter assays showed that a 3.1-kbp exon IS promoter fragment was able to direct GFP expression in early embryonic hematopoietic progenitor cells derived from the aorta-gonads-mesonephros (AGM) region or the fetal liver (14, 15). This discrepancy may be due to intrinsic stage-specific differences in the regulatory mechanisms controlling Gata2 transcription. We would speculate that an enhancer sequence within the 3.1-kbp IS promoter may be used to activate the IG promoter in AGM hematopoietic cells because we can detect IG transcript there (Table 2), whereas this same enhancer can activate the isolated IS promoter (in a construct lacking the IG promoter) such as the 3.1-kbp transgene used in our previous study (15). However, we have not detected GFP expression in the bone marrow cells of these same 3.1-kbp IS promoter-directed transgenic mouse lines (data not shown). Additionally, we found that the regulatory sequences controlling fetal and embryonic Gata2 expression differ even within the 3.1-kbp promoter fragment (M. Kobayashi-Osaki and M.Y., unpublished results). In summary, these results indicate that the regulatory controls over Gata2 transcription in embryonic, fetal, and adult hematopoietic cells are different and complex.

Because inactivation of the IS promoter by the GFP knock-in mutation affected neither the level of Gata2 transcript nor hematopoiesis (data not shown), probably due to compensation by IG promoter activity, we conclude that IS inactivation seems to generate animals that are free from GATA-2-deficient phenotypes (14, 29) but nonetheless faithfully report endogenous Gata2 gene expression in the HSC compartment (6, 8). Therefore, the localization and behavior of HSCs examined here is likely to reflect intrinsic properties of HSCs in vivo. Further analyses of the EIS-KI mice should provide even deeper insights into the inherent properties of HSCs, including the properties that distinguish between their capacity for self-renewal vs. their ability to repopulate (differentiate into) all hematopoietic lineages.

Materials and Methods

Establishment of Gata2 GFP Knock-in Lines.

Two GFP knock-in vectors (Vector-EIS and Vector-EII, Fig. 1A) were constructed from mouse genomic fragments, the GFP gene, and ploxNATA-DT3 plasmids (8, 30). These vectors were electroporated into E14 ES cells. Through PCR and Southern blotting analyses, seven and three homologously recombined clones were identified from the Vector-EIS- and Vector-EII-transfected ES colonies (240 each), respectively. Chimeric mice were obtained by injection of the ES clones into the C57BL/6 blastocysts, and the mice were mated with C57BL/6 mice to obtain heterozygous knock-in mice. To excise the neomycin resistance gene cassette (Neo), heterozygous knock-in mice were crossed with Ayu1-Cre transgenic mice that express ubiquitous Cre recombinase (31). Mice in generation 8 to 10 were used in this study. To detect the GFP gene, tail DNA was extracted, and PCR was performed by using GFPs4 (5′-CTGAAGTTCATCTGCACCACC-3′) and GFPas4 (5′-GAAGTTGTACTCCAGCTTGTGC-3′) primers (32).

Flow Cytometry and Cell Sorting.

Sorting and marker analysis of the bone marrow cells were performed by using a FACSVantage SE and cellquest software (Becton Dickinson). Mononuclear cell suspensions from mouse bone marrow were prepared and incubated with biotinylated monoclonal antibodies recognizing Mac-1, Gr.1, Ter119, B220, CD4, and CD8 (Lin markers). Hematopoietic lineage marker-negative (Lin⁻) cells were enriched by magnetic separation by using streptavidin-conjugated magnetic beads (Polysciences, Warrington, PA), and Lin⁻ cells were stained with PE-conjugated anti-Sca1 and allophycocyanin (APC)-conjugated anti-c-Kit antibodies (32). Live Lin⁻ cells were sorted into fractions after staining with streptavidin-conjugated energy-coupled dye (ECD) (Beckman Coulter) and propidium iodide (PI, Sigma). All antibodies were obtained from BD Pharmingen. For SP analysis, raw cells were stained in phosphate-buffer saline containing 5 μg/ml Hoechst 33342 fluorescent dye (Calbiochem) and analyzed by FACSVantage SE (22). For cell-cycle analysis, cells were fixed with 2% paraformaldehyde for 10 min, and stained for DNA by using 5 μg/ml Hoechst 33342 (23).

Quantitative RT-PCR Analysis.

RNA samples were purified from 5,000 cells by using RNeasy (Qiagen, Basel, Switzerland) and reverse-transcribed by using SenscriptRT (Qiagen) and random hexamers. Quantitative RT-PCR analyses were performed by using an ABI PRISM 7700 (PerkinElmer), and GATA-2 mRNA levels were quantitatively measured as described (32).

Long-Term Reconstitution Assay.

EIS-KI mice were backcrossed to C57BL/6-Ly5.1 mice, and the bone marrow cells were sorted from these mice. Sorted cells were mixed with 20,000 Lin⁻ competitor cells from Ly5.2 mice, and injected into the tail vein of lethally irradiated (9.5 Gy) male C57BL/6-Ly5.2 mice (20). The chimerism of cells from the donor was measured by Ly5.1⁺ cell ratio in B220⁺ and Thy1.2⁺ lymphoid fractions and by Gr.1⁺ and Mac1⁺ myeloid fractions in the peripheral blood (20). For the limiting dilution assays, 20 recipient mice were used. Five months after transplantation, the percentage of negative mice whose peripheral blood chimerism was <10% was plotted, and the frequency of HSC was calculated based on a Poisson distribution (20, 21).

Immunostaining.

Femurs were fixed in 4% paraformaldehyde for 30 min and embedded in OCT compound (Sakura Finetek, Tokyo) in liquid nitrogen. For immunofluorescence staining, frozen sections (10 μm) were incubated with PE-conjugated anti-Sca1 antibody (BD Pharmingen). After washing, sections were examined by fluorescence microscopy (Leica Microsystems). For immunohistochemistry, sections were incubated with rabbit anti-GFP antibody (diluted 1:1,000, Molecular Probes). Specific antibody binding was visualized as brown staining by reaction with chromogen (250-mg/ml diaminobenzidine with 0.01% H₂O₂) after incubation with horseradish peroxidase-conjugated anti-rabbit IgG (BioSource, Camarillo, CA). Samples were then incubated with Fast Blue RR (Muto, Tokyo) at 37°C for 2 h for alkaline phosphatase staining or counterstained with Alcian blue.

Movies.

Mice were injected with 5-FU (150 mg/kg; Kyowa Hakko Kogyo, Tokyo). Twenty-four to 72 h later, the bones were split and immersed in culture medium (DMEM with 10% FBS) saturated with 5% CO₂. The bones were cultured on a heated microscope stage, and images were recorded every minute by using a Leica AS MDW microscope. For more information, see Movies 1–3.

Glossary

Abbreviations:

HSC

hematopoietic stem cell

IS

I-specific

IG

I-general

En

embryonic day n

SP

side population

PE

phycoerythrin

5-FU

5-fluorouracil.