GATA-3 regulates the self-renewal of long-term hematopoietic stem cells.

The transcription factor GATA-3 is expressed and required for differentiation and function throughout the T lymphocyte lineage. Despite evidence it may also be expressed in multipotent hematopoietic stem cells (HSCs), any role for GATA-3 in these cells has remained unclear. Here we found GATA-3 was in the cytoplasm in quiescent long-term stem cells from steady-state bone marrow but relocated to the nucleus when HSCs cycled. Relocation depended on signaling via the mitogen-activated protein kinase p38 and was associated with a diminished capacity for long-term reconstitution after transfer into irradiated mice. Deletion of Gata3 enhanced the repopulating capacity and augmented the self-renewal of long-term HSCs in cell-autonomous fashion without affecting the cell cycle. Our observations position GATA-3 as a regulator of the balance between self-renewal and differentiation in HSCs that acts downstream of the p38 signaling pathway.

GATA3 is a zinc finger transcription factor expressed and essential to differentiation and function throughout the T lymphocyte hierarchy [1]. In common with many transcription factors, its localization between cytoplasm and nucleus is governed by modification of its classical nuclear localization signal motif. In human T lymphocytes, the GATA3 nuclear localization motif is serine phosphorylated by activated p38/MAPK downstream of T cell receptor activation, leading to binding of importin α and carriage into the nucleus [8]. Besides expression in developing T cells, mRNA-level expression of Gata3 has also been reported further upstream in sorted populations enriched for murine long-term multipotent hemopoietic stem cells (LT-HSC) [2-4]. While the RNA evidence might indicate a functional role in HSC, it is puzzling that deletion of Gata3 targeted to HSC has not yielded a hemopoietic deficiency phenotype outside the T lymphocyte lineage [5-7]. The absence of phenotype raises questions that remain outstanding, including whether Gata3 transcripts were correctly attributed to LT-HSC as opposed to contaminating cell types in the impure HSC fractions analyzed, or whether RNA is translated in HSC to functional GATA3 protein. Further, as GATA3 depends on upstream signaling for its activation [8], it is not clear whether effects of deletion were fully tested in circumstances where GATA3 was in an active state.

HSC are capable of reconstituting all blood lineages from a single purified cell after transplant into irradiated mice [2,9]. Two distinct classes of HSC have recently been resolved based on differing biology and phenotypes. LT-HSC express SLAMF1 (CD150 antigen) [2,3,14] but not α2 integrin (CD49b antigen) [2] and generate grafts that indefinitely sustain myeloid, erythroid and lymphoid cell output. They are deeply quiescent in the normal bone marrow steady state, with intermitotic intervals estimated at 50 – 100 d from histone H2B-GFP dilution rates [10,11]. Intermediate-term (IT-) HSC do express α2 integrin [2] allowing their prospective separation from LT-HSC, are 3-fold more numerous [2,11] and generate systemic grafts from single injected cells which sustain myeloid and erythroid cell production for 12 wk before declining [2,3,12]. They are similarly quiescent in the normal steady-state, but reenter cycle more frequently (every 10 – 20 d) as evidenced by more rapid H2B-GFP dilution rates [10,11]. Their greater number, their extensive proliferative capacity and their more frequent exit from quiescence imply a dominant role in maintaining blood cell production in the steady-state [13]. Gata3 transcripts are detected preferentially in α2 integrin −ve fractions suggesting preferential expression in LT- rather than IT-HSC [2]. Given the known involvement of GATA3 in differentiation of progenitor hierarchies in T lymphocyte and mammary luminal epithelial lineages, the observation raised the intriguing possibility that GATA3 might be involved in specifying the sustained self-renewal property of LT-HSC or the transition from LT- to IT-HSC.

Here we report results that resolve the ambiguities in the earlier evidence and identify a novel role for GATA3 as an HSC-autonomous regulator of cell fate. Direct proof is provided for Gata3 mRNA and protein expression in LT-HSC, but not more advanced HSC. Further, in LT-HSC purified to near functional homogeneity, GATA3 protein is shown to relocate from cytoplasm to nucleus in response to external signals, a change associated with both exit from quiescence and reduction in capacity for long-term reconstitution in irradiated hosts. Finally, we use conditional deletion of Gata3 in LT-HSC to establish a causal link between GATA3 activation and reduction in long-term reconstitution activity, and thus to demonstrate gain in HSC self-renewal, without change in cycling parameters, as the essential deletion phenotype.

RESULTS

Gata3 RNA and protein are expressed in LT- but not IT-HSC

Quantitative RT-PCR confirmed that expression of Gata3 transcripts in the LineageloSca-1+c-Kit+Rholoα2Integrinlo (LSKRα2lo) fraction, which contains LT-HSC at 30% functional purity[2], was 47-fold more abundent than in the LSKRα2hi fraction consisting entirely of IT-HSC (Supplementary Fig. 1). To link Gata3 expression more definitively to LT-HSC, we took advantage of a mouse strain (Gata3) in which exon 4 of the Gata3 gene is replaced with an eGFP cassette expressed from the endogenous Gata3 promoter [15]. Gata3 bone marrow cells were sorted into eGFP negative, low and high fractions (Fig. 1a). Competitive 32 wk erythroid reconstitution assays (Fig. 1b) showed that all long-term reconstituting activity (as well as phenotypic LT-HSC; Supplementary Fig. 1) was present in the eGFPlo and eGFPhi fractions. In contrast, nearly all intermediate 8 – 16 wk activity remained in the eGFP− fraction. These results indicate preferential transcriptional activity of Gata3 genes in functionally defined LT-HSC.

Figure 1

Gata3 is actively transcribed in LT- but not IT-HSC

(a) GFP fluorescence intensity distributions in bone marrow cells from wild-type or Gata3 mice expressing a GFP cassette under control of the endogenous Gata3 promoter. Total bone marrow cells were gated on forward and side scattering channels to exclude outliers. (b) Erythroid reconstitution from C57BL/6J-GPI1b-Gata3 (left) or C57BL/6J-GPI1b-Gata3 (right) bone marrow cells fractionated according to eGFP fluorescence and transferred in competition with wild-type cells of host genotype into lethally irradiated C57BL/6J-GPI1a mice. Blood samples were analysed at 8 and 32 wk post-transplant (3 mice per fraction in each of 2 independent experiments). Results from both experiments were pooled. The number of 8 wk (intermediate-term, top) and 32 wk (long-term, bottom) erythroid repopulating cells per fraction was calculated. The results for each fraction were normalized to percentage of total repopulating cells recovered from all 3 fractions. The number of intermediate term cells was the number calculated in a fraction at 8 wk minus the number determined at 32 wk.

Absence of a deletion phenotype might be expected if Gata3 mRNA were not translated to functional protein in HSC. We therefore looked for GATA3 protein expression by immunostaining. GATA3 protein was detected in 30% of LSKRα2lo cells, which corresponds to the proportion of cells in this fraction able to generate long-term hemopoietic grafts [2], but was not detected in LSKRα2hi cells consisting entirely of IT-HSC (Fig. 2a,b). When SLAMF1 was added to the isolation strategy (Fig. 2c), 85% of LSKRα2loSLAMF1hi cells immunostained for GATA3 (Fig. 2e). In 32 wk erythroid reconstitution assays 50 LSKRα2loSLAMF1hi cells reproducibly competed 1:1 with 106 bone marrow cells (Figure 2D), which contain about 50 LT-HSC [2], suggesting essential functional homogeneity of this fraction. These observations showed direct correspondence between the proportion of cells immunostained for GATA3 protein and the extent of HSC purification, and suggested that the LSKRα2loSLAMF1hi marker combination achieves the closest approach to functional long-term HSC purification yet reported.

Figure 2

GATA3 protein is expressed in LT- but not IT-HSC

(a) Immonufluorescence analysis of GATA3 expression in LSKRα2lo and α2hi cells. LSKRα2lo and α2hi cells were sorted, fixed, permeabilized and stained for GATA3. Representative images are shown. GATA3 staining was detected in approximately 1/3 of LSKRα2lo but not in α2hi cells. (b, e) Quantification of GATA3 flurorescence intensity in purified HSC. Total fluorescence signal was measured in individual cell images stained with mouse anti-GATA3 (left panel) or isotype control (IgG, right panel). Frequencies were normalized to a mode of 100 for α2hi cell images and equal total areas on each plot. The proportion of cells expressing GATA3 above background was 30% and 85% for LSKRα2lo and LSKRα2loSLAMF1hi cells respectively. Results were aggregated from 5 and 3 independent separations respectively. (c) Isolation of LSKR α2loSLAMF1hi and α2hiSLAMF1lo cells. Events shown were gated on LSKR parameters. (d) High functional purity of the LSKRα2loSLAMF1hi cell fraction. 50 sorted cells were coinjected with 106 bone marrow cell competitors of host genotype into lethally irradiated recipients in 2 independent experiments. Historically this competitor dose would contain about 50 LT-HSC [2]. Donor red blood cell reconstitution was measured at 8 wk intervals after transplantation with SEMs indicated. Achievement of 50% donor reconstitution at 24 – 32 wk is indicative of functional homogeneity of the LSKRα2loSLAMF1hi fraction.

Localization of GATA3 protein responds to cellular signaling

Having established the presence of GATA3 protein in purified LT-HSC, we wanted to determine whether its behavior or activation state would be responsive to upstream signaling. GATA3 was cytoplasmic in quiescent, freshly isolated LSKRα2lo cells (Fig. 2a). To assess the responsiveness of GATA3 protein to cytokine signaling, LSKRα2lo cells were cultured with serum, Kit ligand, Flt3 ligand, Interleukin-11 (IL-11) and Interleukin-7 (IL-7). Cultured cells began dividing by 40 h 2. GATA3 was readily detectable in cell nuclei by 24 h and was mainly nuclear by 48 h (Fig. 3a–c).

Figure 3

GATA3 relocalizes to the nucleus in cycling cells and the effect is inhibited by p38α inhibitors

(a) Subcellular localisation of GATA3 in quiescent and cultured LT-HSC. LSKRα2lo cells were stained for GATA3 and DNA counterstained with DAPI, directly after sorting or after culture. (b,c) Whole cell and nuclear fluorescence pixel intensities were summed for individual cells in microscopy images. The nuclear perimeter was traced in the DAPI-stained images and duplicated in the GATA3-stained images. The measurements are plotted as the mean ratio of summed nuclear to cytoplasmic fluorescence pixel intensities (n=6 or 7 cells for each value) with SEMs and 1-tail T test probabilities indicated (* = .031; ** = .00026). (d) Phospho-p38 MAPK staining in cultured LT-HSC. LSKRα2loSLAMF1hi cells were cultured for 1 h in serum- and cytokine-free medium after which serum and cytokines were added. Controls received medium without serum and cytokines. After a further 15 min incubation cells were stained with anti-phospho-p38α antibody. Pixel intensities were summed for individual cells in fluorescence microscopy images. Frequencies were normalized as in figure 2b. Results are aggregated from 3 independent experiments. (e) Effect of p38α inhibitors. LSKRα2loSLAMF1hi cells were cultured for 2 d with cytokines and SB239063 prior to staining and assessment of GATA3 localization by confocal microscopy. (f) Quantitation of GATA3 fluorescence localization in the presence of p38α inhibitors. Means + SEM from 2 independent experiments are shown with 1-tail T test probabilities (* = .026; ** = .00090, *** = 4.4E-04).

To determine whether nuclear entry of GATA3 might be controlled by p38, as in human T lymphocytes, we first asked whether p38 was activated in cultured LSKRα2loSLAMF1hi cells. Activation of p38α in cultured LSK cells was previously described [17]. However, LSKRα2loSLAMF1hi cells represent less than 3% of LSK cells and therefore activation of p38 specifically in LT-HSC expressing GATA3 remained to be established. As assessed by immunofluorescence microscopy, phospho-p38α was measurably elevated after 15 min exposure to medium containing serum and cytokines (Fig. 3d). To test directly for dependence of nuclear relocalization on activation by p38, we cultured HSC in the presence of specific inhibitors of p38α catalytic activity. Two agents were tested, SB203580, a “first generation” pyrimidinyl imidazole inhibitor [18], and SB239063, a “second generation” inhibitor of the same class with greater specificity for p38α and higher potency [19]. Both significantly inhibited nuclear localization of GATA3 at 10 μM (Fig. 3 e,f), within the concentration range originally established for high p38α specificity [19,20]. Inhibition at 3 μM was more marked with SB239063. Neither inhibitor affected subsequent growth in culture (not shown). The observations show GATA3 protein translocates to the nucleus in LT-HSC following cytokine exposure and activation from quiescence, implicate activated p38 as a required signal for activation and nuclear entry of GATA3, and suggest a functional role for GATA3 in cycling as opposed to quiescent LT-HSC.

Poly(I:C) affects LT-HSC cycling, GATA3 localization and HSC performance

We next examined the functional consequences of Gata3 deletion with a particular focus on regenerative activity following bone marrow transplantation, when HSC would be actively cycling. We deleted Gata3 using an Mx1-Cre transgenic background, in which Cre expression is triggered in hemopoietic cells by type I interferons, typically induced by administration of poly(I:C), which signals thorugh Toll-like receptor 3 to induce multiple downstream effects in addition to type I interferon expression [21]. Because of the likelihood that poly(I:C) administration might itself affect HSC independent of Gata3 deletion, we first characterized its effects on HSC in wild-type mice. Poly(I:C) is known to induce HSC cycling in vivo, with a single injection having maximal effect at 48h followed by return to quiescence by 4 d [22]. We analysed the in vivo cell cycle response of LSKα2loSLAMF1hi cells one and ten days after 3 successive poly(I:C) injections, omitting the Rhodamine123lo marker because it is selective for quiescent cells. Freshly isolated cells were stained intracellularly with Ki67 and Hoechst 33342 (Fig. 4a). Most (81%) LSKα2loSLAMF1hi cells were in G1 or G2/M one day after poly(I:C) and returned to quiescence by ten days. We also examined GATA3 protein localization in these cells on day 1, 10 and 5 months after poly(I:C) injection (Fig. 4b). GATA3 was still mostly nuclear at 10 d, but 5 months after treatment a substantial proportion of GATA3 was again cytoplasmic. Thus, as observed in vitro, relocation of GATA3 to the nucleus of HSC also occurred in vivo after treatment that induced HSC cycling, and was followed by delayed return to the cytoplasm after reversion of HSC to quiescence.

Figure 4

Poly(I:C) treatment induces HSC cycling and GATA3 relocalization in vivo and reduces the long-term reconstituting capacity of LSKRα2lo cells

(a) Cell cycle analysis in LSKα2loSLAMF1hi cells at 1 and 10 d after in vivo treatment with poly(I:C). Cells were stained with anti-human Ki67 and Hoechst 33342 and analyzed by flow cytometry (left). Cycle phase distributions are plotted (right) showing means + SEM of 2 independent experiments. (b) Subcellular localization of GATA3 in LSKRα2lo cells after poly(I:C) treatment analysed by confocal microscopy. Whole cell and nuclear fluorescence intensities were summed in images of individual GATA3-positive cells (n=6 for each condition). The ratios of nuclear to cytoplasmic intensities are plotted (right) showing means + SEM. UT = Untreated. (c) 100 LSKRα2lo cells from control or poly(I:C) treated mice were injected in competition with 106 bone marrow cells of host genotype into lethally irradiated recipients and erythroid reconstitution was tracked. Means and SEMs from 5 independent experiments are plotted.

We next asked whether nuclear relocated GATA3 might affect the regenerative activity of LT-HSC. Following treatment of wild-type mice with poly(I:C), bone marrow LSKRα2lo cells were isolated 10 days or 5 months later and transplanted into irradiated hosts in competition with unfractionated wild-type recipient-genotype bone marrow cells. At 8 wk after transplantation, cells obtained at either timepoint after poly(I:C) treatment achieved similar levels of erythroid reconstitution to cells from untreated mice (Fig. 4c), suggesting that IT-HSC (normally the majority HSC type in the LSKRα2lo fraction [2]) were not affected. However, 24 – 32 weeks after transplantation, LSKRα2lo cells that had been treated with poly(I:C) 10 days before transplantation - when the cells had returned to quiescence but in which GATA3 was still intranuclear - yielded markedly lower levels of erythroid reconstitution (Fig. 4c) which represented a 6-fold reduction in number of long-term reconstituting cells by poly(I:C) treatment. This result was concordant with earlier reports of a depleting effect of poly(I:C) treatment on the long-term reconstituting potential of bone marrow cells [22,23]. In contrast, robust reconstitutions were achieved by either non-poly(I:C) treated control cells or LSKRα2lo cells treated with poly(I:C) 5 months previous to transplantation (Fig. 4c), in which GATA3 was again cytoplasmic (Fig. 4b). These observations indicate that HSC containing nuclear GATA3 had diminished long-term reconstituting capacity, suggesting a correlation between LT-HSC regenerative performance and localization of GATA3, while regenerative activity of IT-HSC, which lack GATA3, was close to control levels after poly(I:C) treatment.

Gata3 deletion has little effect on steady-state hemopoiesis

For deletion of Gata3 in hemopoietic cells we used a mouse strain (Gata3fl/fl) in which the 4th exon of Gata3 encoding the first zinc finger is flanked by LoxP sequences [24]. Gata3fl/fl mice were crossed with Mx1-Cre mice to yield Gata3fl/fl-Mx1-Cre mice. Induction of Cre recombinase in these mice yields a truncated Gata3 transcript predicted to splice out-of-frame to exon 5 containing the second zinc finger, translating to an inactive protein lacking both zinc fingers [24]. Gata3fl/fl mice that received the same poly(I:C) treatment as Gata3fl/fl-Mx1-Cre mice were routinely used as undeleted controls.

Gata3fl/fl-Mx1-Cre mice were treated with poly(I:C) to induce deletion of Gata3, and blood cells and bone marrow precursor populations were examined 3 – 9 mo later. Little effect of Gata3 excision was evident either on the proportions (Fig. 5a) or absolute numbers (not shown) of circulating myeloid or lymphoid cells or on the number of LSKRα2loSLAMF1hi and LSKRα2hiSLAMF1lo HSC in the bone marrow. Because a previous report had suggested that Gata3-deficient HSC were more quiescent in steady state conditions than intact HSC [7], we also examined the cell cycle status of LSKα2loSLAMF1hi cells by Ki67/Hoechst 33342 intracellular staining in the bone marrow of Gata3fl/fl-Mx1-Cre and Gata3fl/fl mice 3 mo after treatment with poly(I:C). 85% of LSKα2loSLAMF1hi cells were quiescent, with no significant difference in Ki67/Hoechst staining distributions between undeleted Gata3fl/fl control or Gata3fl/fl-Mx1-Cre mice (Fig. 5b). We also tested the cell cycle response of Gata3-deficient LSKα2loSLAMF1hi cells to a second round of poly(I:C) treatment and found they entered cell cycle to a similar degree as Gata3-intact controls (Fig. 5b). The lack of apparent effect of Gata3 deletion on steady state bone marrow maintenance confirms an earlier report [6] and is consistent with the cytoplasmic localization of GATA3 protein in quiescent LT-HSC. Diagnostic PCRs on total bone marrow cells and circulating myeloid cells at various times after 3 successive poly(I:C) injections regularly indicated a minimum 97% efficiency of Gata3 excision (Fig. 5c). Additionally, in clones grown in culture from individual LSKR cells obtained 10 d following poly(I:C) treatment, Gata3 excision was complete in 99 of 101 the clones examined. These results showed that Gata3 deficiency had little effect on blood cell production or LT-HSC maintenance in steady-state conditions where LT-HSC were quiescent and where GATA3 was mainly cytoplasmic.

Figure 5

Gata3 excision has little effect on steady state bone marrow and blood populations

Gata3fl/f (“Control”) or Gata3fl/fl-Mx1-Cre (“Excised”) mice were treated with poly(I:C) and analyzed 3 – 9 months later (“steady state”). (a) Myeloid and lymphoid cells (proportion of total cells, n=7 independent experiments) in blood and absolute numbers of phenotypic LT- and IT-HSC (n=4 and 3) in steady state bone marrow. (b) Cell cycle analysis in LSKα2loSLAMF1hi bone marrow cells in steady state bone marrow or 1 d after a new poly(I:C) treatment. The bar graphs show means + SEM of 2 independent experiments. (c) Representative PCR reactions on DNA for assessment of Gata3 excision at the indicated times after treatment with poly(I:C). Image contrast and gamma were adjusted to allow visualization of the faint Gata3fl band. 1, bone marrow 2 wk; 2, bone marrow 8 wk; 3, bone marrow 24 wk; 4, blood myeloid 7 mo; 5, blood myeloid 10 mo; 6–8, LSKR clones 10 d after excision; 9, Gata3fl/fl unexcised control.

Gata3 excision protects LT-HSC from depletion by poly(I:C)

We next looked for possible consequences of Gata3 deficiency in conditions where GATA3 would be intranuclear. Unfractionated bone marrow cells, 106, from Gata3fl/fl-Mx1-Cre or Gata3fl/fl mice 10 days after poly(I:C) treatment - when GATA3 was mainly nuclear (Fig. 4b) -were each injected in competition with 106 untreated wild-type bone marrow cells into irradiated recipients. Hematopoietic reconstitution in primary transplant recipients was tracked at 8, 16, 24 and 32 weeks as a measure of the long-term reconstitution potential of transferred cells. While long-term reconstituting activity was depleted in poly(I:C)-treated wild-type bone marrow, no defect was seen with poly(I:C)-treated Gata3fl/fl-Mx1-Cre bone marrow which robustly sustained erythroid, myeloid and B lymphocytic lineages in primary recipients up to 32 weeks (Fig. 6a), matching or exceeding the regenerative activity of the co-injected untreated wild-type competitor cells. As expected, T lymphocyte regeneration from poly(I:C)-treated Gata3fl/fl-Mx1-Cre bone marrow was weak relative to myeloid and B lymphocyte reconstitution (Fig. 6a). These results showed that poly(I:C)-induced nuclear translocation of GATA3 impaired the long-term regenerative capacity of bone marrow LT-HSC and that deletion of the Gata3 gene reversed the effect.

Figure 6

Regenerative activity and self-renewal are enhanced in vivo and in vitro after Gata3 deletion

(a) Gata3fl/fl-Mx1-Cre (“excised”) and control Gata3fl/fl (“intact”) mice were treated with poly(I:C). 106 marrow cells taken 10 d later were injected with 106 host-genotype bone marrow cells into irradiated recipients. Donor red cells were measured at intervals post-transplant (left). Proportions of donor myeloid and lymphoid cells in blood were measured at 32 wk (right). Means and SEMs of 3 independent experiments are indicated. (b) Gata3-deleted bone marrow cells were transplanted into irradiated 1° recipients with normal competitors. After 24 wk, bone marrow was transfered to irradiated 2° recipients. The bar graphs show the cumulative fold-expansion in LT-HSC numbers in 1° and 2° recipients with SEMs and 1-tail T test probabilities indicated. (c) Gata3fl/fl or Gata3fl/fl-Mx1-Crel bone marrow cells were injected with normal competitors into irradiated wild type recipients. Hosts were treated with poly(I:C) 8 wk later. At 24 wk, bone marrow cells were assayed competitively in vivo for long-term erythroid reconstituting activity. The bar graph shows the fold-expansion in LT-HSC numbers calculated to have occurred in the 1° recipients with SEMs indicated (single experiment, 3 – 5 mice per point, * = .04, 1-tail T test). (d) Bone marrow from Gata3fl/fl-Mx1-Crel and control Gata3fl/fl mice was analyzed 2–3 months after treatment with poly(I:C). Single LSKRα2lo (n=3 independent experiments) or LSKRα2loSLAMF1hi (n=1) cells were cultured per well with serum and cytokines, and cell numbers were recorded every 4–8 hr. Each point represents the sum of cells in 30 wells. (e) LSKRα2loSLAMF1hi cells, 25, from Gata3fl/fl-Mx1-Cre or control Gata3fl/fl mice treated 3 – 5 months earlier with poly(I:C) were cultured at 1 per well, and also assayed competitively in vivo for long-term erythroid reconstitution. After 7 d all harvested cells were again assayed in vivo. The bar graph shows SEM and mean relative number of LT-HSC recovered from 7 d cultures (“Out”) relative to the number initiating the cultures (“In”) in 2 independent experiments.

Gata-3 deletion enhances LT-HSC expansion in vivo

Because GATA3 relocated to the nucleus when LT-HSCs were induced to cycle, we next asked whether functional consequences of Gata3 deletion might be apparent in LT-HSC after transfer into irradiated mice, where they enter active cell cycle and have been shown to expand in number by a factor of 10 after transplantation of 106 bone marrow cells [25,26]. Gata3fl/fl-Mx1-Cre mice were treated with poly(I:C) to delete Gata3. Bone marrow was taken 10 d later and erythroid regenerative activity was measured competitively in primary recipient mice by coinjection of 106 unfractionated cells with the same number of bone marrow cells from untreated wild-type mice of recipient genotype. The proportions of donor and competitor erythrocytes were measured in blood at 24 wk by GPI1 assay #1, from which numbers of Gata3-deleted LT-HSC initially injected were estimated as in [2].

Unfractionated bone marrow obtained at 24 wk from 1° recipients was transplanted into irradiated 2° recipients. After a further 24 wk, proportions of donor and competitor erythrocytes were again determined in GPI1 assay #2, and unfractionated bone marrow from 2° recipients was injected in varying numbers into 3° recipients for enumeration of LT-HSC originating from Gata3-deleted and wild-type competitors by limiting dilution of 20 wk erythroid reconstituting activity. Results of GPI1 assay #2, the limiting dilution assay and the historically determined content of 106 normal competitors were combined as in [2] to yield the numbers of LT-HSC initially injected into secondary recipients and the number recovered from them. These numbers were used to compute the net numerical expansions of LT-HSC occuring in primary and secondary recipients (Fig 6b). Untreated wild-type competitor HSCs expanded in primary and secondary recipients by 10- and 7.5-fold in number during the primary and secondary reconstitution respectively to achieve a cumulative expansion of 75-fold. In contrast, Gata3-deficient LT-HSC expanded by 20-fold during regeneration in primary recipient mice and a further 13-fold in secondary recipients to achieve a cumulative 260-fold increase over the number of initially transferred cells (Fig. 6b). These results showed that Gata3-deficient LT-HSC expanded to a greater extent during sequential regeneration in vivo than wild-type control HSC, suggesting that GATA3 must restrain the extent to which normal LT-HSC expand during bone marrow regeneration.

GATA3 effects on HSC self-renewal are HSC-autonomous

To ensure that the regenerative advantage of Gata3-deficient HSCs was HSC-autonomous, and not due to systemic effects of Gata3 deletion in the poly(I:C)-treated Gata3fl/fl-Mx1-Cre mice, we transplanted 0.5 x 106 unfractionated Gata3fl/fl-Mx1-Cre or Gata3fl/fl bone marrow cells together with 106 wild-type control bone marrow cells into irradiated wild-type mice and induced Gata3 deletion with poly(I:C) treatment 8 wk post-transfer. In this experimental strategy Gata3 would be deleted only in the regenerated Gata3fl/fl-Mx1-Cre cells. 24 wk after reconstitution, unfractionated bone marrow cells from the mixed chimeras were injected into secondary irradiated hosts and the extent of erythroid regeneration from Gata3-deleted versus control competitor cells was quantified. Gata3-deficient bone marrow cells again had a marked erythroid regenerative advantage compared to wild-type cells (Fig. 6c). These results suggest a cell-autonomous effect of Gata3 on the regenerative advantage of HSCs, independent of the environment in which deletion was induced.

Gata3 excision protects LT-HSC from depletion in culture

We also tested the functional consequences of Gata3 deletion in cultured LT-HSC, in which GATA3 also relocates to the nucleus (Fig. 3). Purified, quiescent LSKRα2lo or LSKRα2loSLAMF1hi cells isolated from Gata3fl/fl-Mx1-Cre mice treated with poly(I:C) 3 – 5 months before isolation or from wild-type untreated mice were cultured for 7 days. During culture, neither the cloning efficiency (90 – 95%), the latency to onset of growth[2] (not shown), cell numbers nor subsequent division rate (Fig. 6d) were significantly affected by Gata3 deletion in comparison with wild-type controls. Purified wild-type LT-HSC normally maintain their in vivo competitive reconstituting activity for the first 4 d in culture. Activity falls off steeply thereafter and little or no long-term activity is recovered by 7 d (Fig. 6e). In contrast, Gata3-deficient HSCs maintained their long-term competititve reconstituting activity throughout the 7 days of culture. These results add to the evidence that activated GATA3 exerts a negative regulatory action on the long-term regenerative capacity of LT-HSC, in a way unlikely to involve changes in cell cycling.

DISCUSSION

Earlier work has suggested that GATA3 has little impact on steady-state hemopoiesis beyond its role in T lymphocyte production and function [5-7]. Here we report novel observations that explain the earlier results and define a strong functional role for GATA3 in LT-HSC. The quiescence of LT-HSC and sequestration of GATA3 to the cytoplasm in steady state bone marrow provide a plausible basis for the observations, here and earlier, demonstrating little effect of Gata3 deletion on steady-state hemopoiesis: most GATA3 protein is inactive in quiescent LT-HSC and its deletion should not influence their behavior, while IT-HSC which perform the bulk of bone marrow maintenance do not express Gata3. Deficiency of Gata3 in regenerative contexts did not lead to loss of function, which earlier studies were configured to detect, but rather gain in extent of self-renewal, whose detection requires suitably configured quantitative assays. In every context tested - growth in culture, response to poly(I:C) in vivo and regeneration following transfer into irradiated mice - deletion of Gata3 led to enhancement in numbers of LT-HSC detected in functional assays. Moreover, the experiments measuring expansion in vivo showed that the effects of Gata3 deletion were LT-HSC-autonomous. Thus, deletion of Gata3 consistently led to gain in self-renewal during LT-HSC proliferation, suggesting a role for GATA3 in restraining self-renewal in proliferating LT-HSC.

Although our findings confirm a lack of effect of Gata3 deletion on steady state bone marrow function as documented earlier [6], the marked enhancement of LT-HSC self-renewal documented here in irradiated hosts was not apparent in the earlier study. Differing Gata3 deletion strategies could account for the discrepancy. The previous study [6] used a Vav-Cre background to achieve Gata3 excision at first emergence of HSC from hemogenic endothelium in the embryo. It is possible that homeostatic adjustments in expression of compensatory genes were more complete in that model than in our Mx1-Cre model that deleted Gata3 in adult life in already formed LT-HSC.

The means by which GATA3 restrains self-renewal remain to be determined. One possible mechanism would be restraint of proliferation. Such an effect would be contrary to studies suggesting that GATA3 actually promotes cycling of LT-HSC [7]. However, we did not observe any effects of Gata3 deletion on cell cycle either in vitro or in vivo. Promotion of apoptosis in cycling cells is another formal possibility but was similarly not supported by our kinetic observations in culture. An attractive alternative mechanism would be promotion of differentiation of LT-HSC toward IT-HSC, a step that we show here also involves sharp down-regulation of Gata3. Such a role would place GATA3 as a regulator of the balance between self-renewal and differentiation in LT-HSC, mediating their reprogramming from a multipotent cell with unlimited regenerative lifetime to a multipotent IT-HSC with a 3 month regenerative potential.

An analogous “pro-differentiation” role for GATA3 has been shown in luminal breast epithelial progenitors, where GATA3 is required for differentiation to a mature ductal epithelial phenotype and where its deletion results in numerical expansion of undifferentiated precursors [27]. While in luminal stem cells the GATA3 requirement for differentiation appears absolute, LT-HSC are still able to differentiate in its absence, suggesting redundancy in elements controlling the differentiation process [6]. In the mammary system, GATA3 collaborates with FOXA1 as a “pioneer” factor [28-31] involved in nucleating a remodelling complex at heterochromatic estrogen receptor target regions that leads to opening and epigenetic marking of the sites for active transcription. A similar role in LT-HSC could provide a mechanism for stable transformation of cellular identity to the IT-HSC stage.

If GATA3 suppresses self-renewal in proliferating LT-HSC, an apparent paradox arises: when bone marrow is transplanted into irradiated recipient mice, the number of LT-HSC normally increases 10-fold over the number injected in the weeks following the transplant [9,25,26]. LT-HSC are induced to proliferate in bone marrow-ablated milieus [32,33], and GATA3 might be expected to be activated and relocate to the nucleus. If so, how would a net increase in LT-HSC occur? We were unable to assess the extent of nuclear entry of GATA3 by immunofluorescence microscopy in this setting because marker sets that allow purification to homogeneity of quiescent LT-HSC do not yield sufficient functional purity when applied to regenerating bone marrow [34]. Conceivably, the degree of GATA3 activation and nuclear relocation in physiologically regenerating bone marrow could be less than what we observed in culture or in vivo consequent to treatment with poly(I:C). Positive signals may also be available in regenerating bone marrow which partially balance out the negative action of GATA3. Nevertheless, deletion of Gata3 augmented expansion of LT-HSCs injected, suggesting that in physiologically regenerating bone marrow GATA3 is activated at a level sufficient to constrain, but not to abolish, increases in LT-HSC numbers.

Translocation of GATA3 to the nucleus was shown in human T lymphocytes to depend on serine phosphorylation at the nuclear localization signal sequence via direct catalysis by p38 and consequent interaction with importin-α [8]. To test for corresponding p38α dependence in LT-HSC, we used two selective p38α inhibitors whose few known off-target kinases are affected only at concentrations 1 – 2 orders of magnitude higher than those used here [19,20,35,36]. Our results in murine LT-HSC confirm in a new cellular context that activation of GATA3 depends on activated p38. Crucially, we also confirmed that p38α is indeed activated in cultured LT-HSC. Activation of p38α has also been documented downstream of TLR3 receptors responding to poly(I:C) in dendritic and NK cells [37,38], as well as c-kit [39], TNF and LTα [40] and reactive oxygen signaling [41] in a variety of hemopoietic cells. Our identification of GATA3 as a negative agonist in LT-HSC responding to the activation state of the p38 signaling pathway implicates it as a potentially critical mediator of the LT-HSC-depleting effects of inflammatory cytokines and reactive oxygen in the earliest hemopoietic stem cells. These effects should be mitigated by inhibition of p38 signaling, and enhanced preservation of LT-HSC cultured in the presence of p38α inhibitors has indeed been observed [17]. Our findings collectively identify a physiological mechanism that constrains the capacity of bone marrow stem cells to engraft long-term, and show that interference with that mechanism can lead to enhanced self-renewal of LT-HSC. The observations have important potential implications for enhancing stem cell expansion both in culture and in vivo.

METHODS

Mice

Gata3GFP/+ and Gata3fl/fl mice were maintained by M.B. [15,24] and backcrossed to C57BL/6J mice for 5 or more generations before use. The genotyping primers for Gata3GFP/+mice were: 5′-CGCCGCCGGGATCACTCTCG-3′ and 5′-GATCCAGACATGATAAGATACA-3′. Gata3fl/fl mice were bred with C57BL/6J-Mx1-cre deleter mice obtained from The Jackson laborarory, Bar Harbor, ME. The genotyping primers for Cre were 5′-GCGGTCTGGCAGTAAAAACTATC-3′ and 5′-GTGAAACAGCATTGC TGTCACTT-3′, and for Gata3 fl and WT alleles 5′-GTCAGGGCACTAAGGGTTGTT-3′ and 5′-TGGTAGAGTCCGCAGGCATTG-3′. Excision of the Gata3fl allele was detected using 5′-GTCAGGGCACTAAGGGTTGTT-3′ and 5′-TATCAGCGGTTCATCTACAGC-3′.

Bone marrow donor genotypes were C57BL/6J-Ly5.2-Gpi1b/b, C57BL/6J-Gata3GFP/+-Ly5.2-Gpi1b/b, C57BL/6J-Gata3fl/fl-Ly5.2-Gpi1b/b, and C57BL/6J-Mx1-cre-Gata3fl/flLy5.2-Gpi1b/b. Recipient mice were C57BL/6J-Ly5.1-Gpi1a/a. HSC purifications were performed on bone marrow from mice that were at least 16 wk old. Animal experiments were conducted under the ethical oversight of the Animal Care Committee of the Ontario Cancer Institute.

For excision of Gata3 in Gata3fl/fl-Mx1-Crel mice, 300 μg high MW poly(I:C) (InvivoGen, San Diego, CA) in isotonic saline was administered intraperitoneally on alternate days for a total of 3 injections. Mice did not exhibit adverse effects of treatment.

Cell Culture

Cells were incubated at 37°C in 5% CO2 in U-bottom microtiter wells (Nunclon) in 100 μl IMDM containing 7.5 x 10−5 M α-thioglycerol, 4% FBS, 0.1% BSA, 5 μg/ml transferrin, 5 μg/ml insulin, 50 ng/ml c-kit ligand (KL), 50 ng/ml Flt3 ligand (FL), 10 ng/ml interleukin-11 (IL-11), and IL-7 conditioned medium all as detailed in [2,9]. For growth kinetics of single cells, sorted cells were plated at limiting dilution and wells containing exactly 1 cell were identified visually after 18 hr culture.

p38 MAPK inhibitors

The inhibitors SB239063 and SB203580 (Sigma-Aldrich) were dissolved in DMSO at 10 mM concentration. The final concentration of DMSO added to the cultures was 0.03 – 0.1% v/v. Control cultures contained the same amounts of DMSO only. No effect on growth was seen at these concentrations of inhibitor or solvent.

Reconstitution Assays

Quantitative competitive assays were performed by coinjecting C57BL/6J-Ly5.2-Gpi1b/b bone marrow cells or 30 to 100 purified HSC with 0.5 – 1 x 106 C57BL/6J-Ly5.1-Gpi1a/a bone marrow competitors into irradiated (9 Gy, Cs137) C57BL/6J-Ly5.1-Gpi1a/a mice. Limiting dilution assays were performed in sublethally irradiated (4 Gy, Cs137) C57BL/6J-Ly5.1-KitW-41J/W-41J-Gpi1a/a hosts. The ratio [(proportion donor) / (1 – proportion donor)] is linearly related to the number of HSC injected in competitive assays [42]. This value was combined with the known frequency of LT-HSC in normal competitor bone marrow (1:18200, [2]) to derive from competitive assay outcomes, typically erythroid reconstitution values, an estimate of the absolute numbers of HSC in the assayed samples as detailed in [2,26].

For the experiment shown in Fig 6b, Gata3-deleted bone marrow cells, 106, were transplanted into irradiated 1° recipients with 106 normal competitors. The proportions of donor and competitor erythrocytes were measured in blood at 24 wk, followed by transplant of 1.5 x 107 1° bone marrow cells without fresh competitors into irradiated 2° recipients. Proportions of donor and competitor erythrocytes were measured at 24 wk. Erythroid reconstituting cells in bone marrow from 2° recipients were enumerated by limiting dilution analysis. 4 Gy irradiated B6-KitW–41J/W–41J mice were transplanted with 2, 6 or 8 x 105 cells per mouse followed by enumeration of mice positive for erythroid reconstitution at 20 wk. Mice had either undetectable reconstitution or at least 15% donor erythrocytes.

For the experiment shown in Fig. 6c, 5 x 102 Gata3fl/fl or Gata3fl/fl-Mx1-Crel bone marrow cells were injected with 106 normal competitors into irradiated wild type recipients. Hosts were treated with poly(I:C) 8 wk later. The proportion of donor erythrocytes in blood was measured at 24 wk and was used to estimate the number of LT-HSC that were originally injected. Host bone marrow cells, 1.5 x 107, were subsequently passaged to irradiated recipients and proportions of donor erythrocytes were determined after 20 wk. These values were used to estimate the number of LT-HSC regenerated in the 1° recipients. The bar graph shows the fold-expansion in LT-HSC numbers calculated to have occurred in the 1° recipients with SEMs indicated (single experiment, 3 – 5 mice per point, * = .04, 1-tail T test).

For the experiment shown in Fig. 6d, bone marrow from Gata3fl/fl-Mx1-Crel and control Gata3fl/fl mice was analyzed 2–3 months after treatment with poly(I:C). Single LSKRα2lo (n=3 independent experiments) or LSKRα2loSLAMF1hi (n=1) cells were cultured in microwells with serum and cytokines, and the number of cells per well was recorded every 4–8 hr. 30 clones from each HSC fraction were tracked. Each point represents the sum of cells in 30 wells.

For the experiment shown in Fig. 6e, Gata3fl/fl-Mx1-Cre and control Gata3fl/fl mice were sacrificed 3 – 5 months after treatment with poly(I:C). LSKRα2loSLAMF1hi cells, 25, were cultured at 1 per 100 μl well in medium containing serum and cytokines, and also coinjected with 0.5 x 106 host-genotype competitor cells into each irradiated recipient in order to measure the amount of long-term reconstituting activity initially placed into culture. After culture for 7 d, all harvested cells were injected with 0.5 x 106 competitors into each irradiated recipient. Erythroid reconstitution was measured at 32 wk.

Detection of Donor Erythrocytes, Myeloid and Lymphoid Cells

Erythrocyte Gpi1 isoforms were resolved by flat bed electrophoresis on Super Sepraphore membranes (cellulose acetate on Mylar, VWR/Pall) and Gpi1 bands quantitated as described [2]. For leukocytes, erythrocytes were lysed in NH4Cl, Fc receptors were blocked with anti-CD16/32 antibody (eBioscience, clone 93) followed by reaction with fluorochrome-conjugated anti-Ly5.1 (eBioscience clone A20), anti-Ly5.2 (eBioscience clone 104), anti-B220 (eBiosciences clone RA36B2), anti-CD11b (eBioscience clone M1/70), anti-Ly6Gr1 (eBioscience clone RB6-8C5) and anti-TCRβ (eBioscience clone H57-597) antibodies. Labelled cells were analysed by flow cytometry (LSRII, Beckton-Dickinson, Mountain view, CA, USA). The proportion of donor cells was expressed as proportion donor within the stated lineage.

Purification of HSC

Femoral and tibial bone marrow cells were treated with NH4Cl for lysis of red cells, stained/destained with Rhodamine123 (Eastman Kodak), incubated with Fc receptor blocking antibody for 5 min (anti-CD16/32, eBioscience, clone 93) and then with PE-Cy5-conjugated anti-B220 (eBioscience, clone RA3-6B2) and anti-CD3 (eBioscience, clone 145-2C11) for 30 min at 4°C as detailed in [2]. Washed cells were separated on a ARIA (BD) sorter and the RholoB220−CD3− fraction collected. Recovered cells were labeled with fluorochrome-conjugated PE-anti-CD49b (BD Pharmingen, clone HMα2), PE-Cy7-anti-c-kit (eBioscience, clone 2B8), and APC-anti-Sca-1 (eBioscience, clone D7) antibodies for 30 min, washed and sorted for LSKRα2 integrin low and high cells. For isolations using both α2 integrin and CD150 markers, cells after Rhodamine123 staining/destaining were lineage depleted (CD5, B220, CD11b, Gr1, 7/4 and Ter-119) using magnetic beads and an Automacs separator (Miltenyi Biotec, BergischGladbach, Germany)). Lineage depleted cells were incubated with Fc receptor blocking antibody and then stained with PE-Cy5-anti Streptavidin (eBioscience), PE-Cy5 anti-CD3 (eBioscience, clone 145-2C11), PE-Cy7-anti-c-kit, APC-anti-Sca-1, Pacific-Blue-anti-CD150 (Biolegend clone TC15-12F12.2) and PE-anti-CD49b antibodies for 30 min, washed and sorted.

Transcript Expression in Purified Cells

Globally amplified cDNAs from LSKR-α2lo and -α2hi cells were described in [2]. cDNAs from the erythromyeloid hierarchy were described in [43,44]. RNA from purified stages in the B [45] and T lymphocyte [46] hierarchies was globally amplified as described in [47]. PCR primers targeting sequence within a 300 bp range upstream of polyadenylation sites [48] were used for detection of specific transcripts in amplified cDNA. Primers for Gata3 were: 5′-GTCACTTTTCTTGCAGCCTA-3′ and 5′-CAGACTGTTTAAAGGCAGTG-3′. Gata3 transcript expression was analysed by Q-RT-PCR using the QuantiTect SYBR Green PCR Kit (Qiagen, Hilden, Germany) on an ABI 7900HT Fast Real-Time PCR system (Applied Biosystems, California USA).Ct (threshold cycle) values were normalized to the endogenous control gene GAPDH (ΔCT = Cttarget−Ctendogenous) and compared to the mean of the LSKRα2hiΔCT calibrator using the ΔΔCT method (ΔΔCT = ΔCTsample−ΔCTcalibrator).

Immunofluorescence and Confocal Analysis

Freshly sorted cells, typically 100–200, were dropped on polylysine coated slides in a humid chamber at room temperature for 30 min followed by fixation for 10 min in PBS 2% fomaldehyde, permeabilization for 10 min in PBS/0.2% TritonX and blocking for 20 min in PBS containing 10% goat serum. Fixed and permeabilized cells were stained for 1h at room temperature with a mouse anti-human GATA3 antibody (HG3-31, IgG, Santa Cruz Biotechnology, or mouse IgG control (Santa Cruz Biotechnology) followed by labeling with Alexa Fluor 555-conjugated goat anti-mouse IgG antibodies for 45 minutes at room temperature in the dark. Slides were mounted with mowiol 4–88 medium (Calbiochem-Merck Chemicals, Darmstadt, Germany) containing DAPI. Cells were imaged by conventional (Zeiss AxioImager) or confocal (Zeiss LSM700) fluorescence microscopy. Fluorescence signal was quantified in individual cell images using ImageJ (http://rsb.info.nih.gov/ij/index.html).

For phospho-p38 MAPK staining, LT-HSC were sorted and cultured for 1 h at 37°C in IMDM without cytokines and serum followed by a 15 min pulse in IMDM containing 4% FBS, 0.1% BSA, 5 μg/ml transferrin, 5 μg/ml insulin, 50 ng/ml c-kit ligand (KL), 50 ng/ml Flt3 ligand (FL), 10 ng/ml interleukin-11 (IL-11), and IL-7. Cells were then fixed for 10 min in PBS/2% formaldehyde, permeabilized for 10 min in PBS/0.2% TritonX, blocked in PBS containing 10% goat serum for 20 min and stained with anti-phospho-p38 MAPK AlexaFluor 555 conjugated antibody (Cell signaling technology, clone D3F9) for 1 h.

DNA purification

2 to 10 x 106 bone marrow or peripheral blood cells were lysed in 0.2 to 1 ml DNAzol lysis buffer (Invitrogen), gDNA was precipitated using 100 to 500 μl EtOH 100%, washed twice with EtOH 70% and diluted into H2O. Gata3 excision was assessed by semi-quantitative PCR.

Cell cycle analysis

Bone marrow cells were lineage depleted using magnetic beads and an Automacs separator (Miltenyi Biotec, Bergisch Gladbach, Germany) followed by labeling with APC-efluor780-anti-Streptavidin (eBiosciences), PE-Cy7-anti-c-kit, APC-anti-Sca-1, PEcy5-anti-CD150 (Biolegend clone TC15-12F12.2) and PE-anti-CD49b antibodies for 30 min followed by fixation and permeabilization in cytofix/cytoperm buffer (BD Pharmingen, Franklin Lakes, NJ, USA). After fixation/permeabilization, cells were stained with anti-human Ki67 antibody (BD Pharmingen, clone B56) for 20 minutes and Hoechst 33342 at 20 μg/ml (Molecular probes) for 5 minutes. Labelled cells were analysed by flow cytometry (LSRII, Beckton-Dickinson, Mountain view, CA, USA).