Evidence that Growth factor independence 1b regulates dormancy and peripheral blood mobilization of hematopoietic stem cells

Murine hematopoietic stem cells (HSCs) are highly enriched in a bone marrow fraction defined by a combination of markers (Lin⁻, Sca-1⁺, c-kit⁺ [LSK], CD150⁺, CD48⁻)¹  and are either in a quiescent (dormant) state or undergo cell cycling.^2,3  During cell division, one daughter cell retains its stem cell properties, whereas the other daughter cell remains a stem cell or differentiates into multipotential progenitors (MPPs; LSK, CD150⁺, CD48⁺ or CD150⁻, CD48⁺), which in turn develop into myeloid, lymphoid, and erythroid cells. Expression level of CD34 allows differentiating between dormant and activated stem cells.^2,4,5  It has been shown previously that CD34⁻ LSK cells represent long-term stem cells whereas CD34⁺ LSK cells have only short-term repopulating potential.⁴  More specifically, CD34⁺ LSK, CD150⁺, CD48⁻ HSC are activated and gradually lose their long-term repopulating capacity.^2,5,6  In contrast, CD34⁻ LSK, CD150⁺, CD48⁻ HSCs are quiescent and comprise a pool of stem cells with long-term repopulating capacity.^2,5,6 

A critical regulatory element for hematopoietic differentiation is provided by the orchestrated activity of transcription factors.^6-12  The zinc finger proteins Growth factor independence 1 (Gfi1) or Growth factor independence 1b (Gfi1b) are paradigmatic examples that illustrate the importance of transcriptional regulation during hematopoiesis and in particular for HSCs.¹³  Previous work has shown that Gfi1b is required for erythroid and megakaryocytic development.^14-21  In contrast, Gfi1, the closely related paralogue of Gfi1b, is a key regulator for lymphoid and myeloid development but seems not to be required for erythro- or megakarypoiesis.^13,22,23  Expression of Gfi1 and Gfi1b is complementary in many hematopoietic cell types, but in HSCs and some lymphoid subsets both transcription factors are expressed.^24-27  Experimental evidence obtained from lymphoid cells demonstrated that the transcription at the Gfi1 and Gfi1b loci is controlled by auto- as well as cross-regulation.^24-26 

It has been reported by 2 different groups that Gfi1 is required to maintain the ability of HSCs to self-renew and to restrict their proliferation,^13,27,28  but the exact mechanisms how Gfi1 exerts this function is unknown. Whether Gfi1b has the same or a different function in adult HSCs could not be analyzed since Gfi1b knockout mice arrest development at midgestation.^29,30  Here, we describe the generation of Gfi1b conditional knockout mice, which permits the analysis of the role of Gfi1b in regulating the function and numbers of adult HSCs. Our data point to a pivotal role of Gfi1b in the restriction of HSC proliferation, which is different and more profound than the role of Gfi1, in particular concerning the regulation of HSC dormancy and proliferation.

Gfi1b^fl/fl mice were generated by homologous recombination in R1 embryonic stem cells. All mice were backcrossed with C57Bl/6 mice, and the C57Bl/6 background was verified by specific satellite polymerase chain reaction. Gfi1^fl/fl, Gfi1^GFP/WT, and Gfi1b^GFP/WT mice were described previously.^24,30,31  All mice were housed in specific pathogen–free conditions, and all animal experiments were approved by the institutional animal ethics committee at the Institut de recherches cliniques de Montréal.

MxCre tg Gfi1^fl/fl or Gfi1b^fl/fl mice were injected with polyinosinic-polycytidylic acid (pIpC; Sigma-Aldrich) at a dose of 500 μg per injection every other day for a total of 5 injections. As control, wild-type (wt) or Gfi1b^fl/fl mice not carrying the MxCre tg were injected with pIpC. With regard to N-acetyl-cystein (NAC; Sigma-Aldrich) treatment, mice were fed every day with 500 μL of NAC (50 mg/mL).

HSCs and progenitors were analyzed with an LSR (BD Biosciences) or CyAn (Beckman Coulter) flow cytometer, and HSCs were sorted with a MoFlo (Cytomation) from adult mouse bone marrow as described previously.^1,32  The 5-bromo-2-deoxyuridine (BrdU) experiments and the determination of cell cycle phases by Hoechst staining was done according to described procedures.²  Reactive oxygen species (ROS) were analyzed by staining HSCs with 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (carboxy-H2DCFDA; Invitrogen) for 30 minutes at 37°C. After staining, cells were analyzed by fluorescence-activated cell sorting for level of ROS in HSCs.

An amount of 20 000 bone marrow cells were seeded on methylcellulose (StemCell Technologies) supplemented with erythropoetin, interleukin-3, interleukin-6, and stem cell factor. After 10 days, the number of colonies was determined. Subsequently cells were resuspended, and 10 000 cells of the suspension were replated on fresh methylcellulose medium.

The number of functional stem cells was determined in vivo using a limiting dilution assay.³³  Different amounts of bone marrow cells from pIpC-treated Gfi1b^fl/fl and MxCre tg Gfi1b^fl/fl mice (both CD45.2⁺) were transplanted together with 200 000 CD45.1⁺ bone marrow cells into lethally irradiated CD45.1⁺ mice. Eighteen weeks after transplantation, the peripheral blood of the recipient mice was analyzed for the contribution of CD45.2⁺ cells, and a percentage of higher than 1% was considered a positive call. Using the L-Calc software (Version 1.1) from Invitrogen, the frequency of functional stem cells was determined.

Genotyping of Gfi1b^fl/fl mice was performed using the following primers:

LP5-3s GGTTTCTACCAGTCTGGCCCTGAACTC; LP3-3r CTCACCTC TCTGTGGCAGTTTCCTATC; LP5-4r TACATTCATGCTTAGAAACTTGAGTC.

The product length of the wt allele is 256 bp, 295 for the floxed allele, and 540 for the deleted allele.

Microarray data have been deposited in the public database Gene Expression Omnibus (National Center for Biotechnology Information; accession no. GSE20655). Samples were hybridized with Affymetrix Mouse Gene 1.0 ST Arrays. Data were processed using the Affymetrix Expression Console software; algorithm-name: rma-gene-default. Only genes up- or down-regulated more than 2-fold were taken in consideration.

The unpaired Student t test was chosen for analyzing the differences in the number of HSCs, common-monocyte progenitors, granulocyte-monocyte progenitors, and platelets. Analysis of variance was used to compare plating efficiency between wt and Gfi1b-deficient bone marrow cells. All P values were calculated 2-sided, and values of P < .05 were considered statistically significant. Statistical analysis was done with GraphPad Prism software (GraphPad Software).

Using previously described Gfi1b:green fluorescent protein (GFP) knock-in mice (Gfi1b^GFP/+), in which the level of GFP follows Gfi1b promoter activity and indicates Gfi1b expression levels,³⁰  we observed that Gfi1b is highly expressed in virtually all HSCs (defined as: LSK, CD150⁺, CD48⁻) but is significantly down-regulated in the more differentiated MPP subsets (defined as MPP1: LSK, CD150⁺, CD48⁺, and MPP2: LSK, CD150⁻, CD48⁺; Figure 1A-B). However, the dormant CD34⁻ HSC fraction^2,4,6  from Gfi1b^GFP/+ mice showed similar mean fluorescence intensities (MFI) as the activated CD34⁺ HSCs (Figure 1B). In addition, using similar reporter mice for Gfi1 (Gfi1:GFP knock-in mice, Gfi1^GFP/+), in which the Gfi1 promoter activity and mRNA levels can be measured by following green flourescence,^24,30  we observed that expression levels of Gfi1 and Gfi1b were different in HSC and MPP subsets. In particular, the Gfi1b promoter is highly active in HSCs and down-regulated upon differentiation to the MPPs (Figure 1B-C), whereas Gfi1 shows lowest levels in HSCs and is up-regulated in the MPP fractions. This suggests that both transcription factors are differentially regulated and thus may have different roles in these cells.

These observations prompted us to further investigate whether Gfi1b plays a particular role in HSCs. Since constitutively Gfi1b-deficient mice die at midgestation²⁹  and thus cannot be used for an analysis of adult hematopoiesis, we generated a Gfi1b-conditional mouse carrying floxed Gfi1b alleles and a MxCre transgene (Figure 1D). In these MxCre tg Gfi1b^fl/fl mice, Gfi1b exons 2-4 can be deleted by injection of pIpC, leading to the abrogation of Gfi1b expression (Figure 1E-G). To exclude possible effects of pIpC and Interferon-α on HSCs, MxCre tg Gfi1b^fl/fl and Gfi1b^fl/fl mice were examined 21 days after the last pIpC injection. It has been shown that this time period is sufficient to wean off effects of pIpC or interferon-α on HSCs.³⁴  When pIpC-injected MxCre tg Gfi1b^fl/fl-deficient mice were analyzed, we observed drastically increased frequencies of HSCs in bone marrow, spleen, and in the peripheral blood (between 30- and 100-fold, respectively; Figure 2A-E and Table 1). This is new feature that is not observed in Gfi1-deficient mice (data not shown and as previously reported^27,29 ). The expansion affected both short-term (defined as CD34⁺ LSK, CD150⁺, CD48⁻) and long-term (CD34⁻ LSK, CD150⁺, CD48⁻) HSCs (Figure 2C and Table 1).

The deletion of Gfi1b increased the number of Lin⁻ cells in the bone marrow but did not significantly alter the overall cellularity of the bone marrow (Table 1). In contrast, there was an increase in the number of splenocytes (Table 1) in Gfi1b-deleted mice, which was mainly the result of an expansion of erythroid progenitors in the spleen (data not shown). Since the total number of bone marrow cells was not altered, the increased frequencies of HSCs correlated well with the increased absolute numbers of HSCs in bone marrow, spleen, and blood, indicating an expansion between 39- and more than 100-fold, respectively (Table 1). We also found that the number of platelets and erythrocytes in the peripheral blood was reduced compared with wt mice, albeit to a different extent, whereas the total number of leukocytes was not changed (Figure 2F-H). This is consistent with the established role of Gfi1b in the erythroid-megakaryocytic lineage.^14-21,29  Finally, we also verified whether the excision of the floxed Gfi1b regions was efficient in HSCs after pIpC induction and observed that cells with nonexcised Gfi1b alleles were below detection level (Figure 2I).

The increased numbers of HSCs in Gfi1b-deficient mice could be the result of a lower rate of spontaneous cell death or more proliferation. Gfi1b-deficient (Gfi1b^ko/ko) HSCs underwent a slightly higher rate of spontaneous apoptosis than wt HSCs but remained still less than 2.5% (Figure 3A). Using a BrdU pulse chase approach, we found that loss of Gfi1b correlated with significantly increased frequencies of cycling HSCs but had no or little effect on cell cycle progression of cells from the MPP subsets (Figure 3B). Staining with Hoechst showed similarly that a higher percentage of Gfi1b^ko/ko HSCs was in S and G₂/M phases than wt HSCs (Figure 3C) but that cell cycle progression of the MPPs was not altered. These two results suggested that Gfi1b restricts specifically the proliferation of HSCs and hence might control HSCs dormancy but does not affect the rate of cell cycle progression in the different MPP fractions (Figure 3C). In support of this, a label retention assay showed that only 10% of Gfi1b^ko/ko HSCs were quiescent (ie, did not divide during our observation period; Figure 3D). In contrast, 45% of the pIpC-treated wt HSCs did not undergo a cell division at the end of the same time period (Figure 3D). These findings indicate that a significant proportion of Gfi1b^ko/ko HSCs is no longer dormant and has entered the cell cycle. It is known that HSCs are kept in a dormant state at the endosteal niche, which provides a hypoxic environment and protects them against oxidative damage by ROS, whereas high ROS are characteristic for activated HSCs and MPPs.^35,36  Interestingly, we observed that Gfi1b^ko/ko HSCs had a significantly increased level of ROS, compared with the wt HSC population (Figure 3E).

One possible explanation for the increased number of HSC could be that loss of Gfi1b activates HSCs and that this activation leads to increased level of ROS, which in turn could lead to an expansion of HSCs as previously described.³⁷  To verify this hypothesis, we fed mice NAC, which counteracts the effects of ROS.³⁷  Indeed, we found that administration of NAC significantly limited the expansion of Gfi1b^ko/ko HSCs in the bone marrow, spleen, and peripheral blood both with regard to frequencies and absolute numbers (Figure 3F-H and Table 2) but did not affect the pIpC-mediated excision of the floxed Gfi1b exons in HSCs (Figure 3I). This suggested that elevated levels of ROS are at least partially responsible for the expansion of Gfi1b-deficient HSCs.

Next, we examined whether loss of Gfi1b might change the self-renewal capacity of HSCs, as was reported in the case of Gfi1 deficiency.^27,28 Gfi1b^ko/ko bone marrow cells generated the same type of colonies (including colony forming unit [CFU[-E, burst-forming unit-E, CFU-G, CFU-M, CFU-GM, CFU-GEMM) as wt cells when seeded in methylcellulose (Figure 4A), and showed initially a higher replating efficiency and generated a higher number of colonies than wt bone marrow (Figure 4A). However, after the fourth replating cycle, Gfi1b^ko/ko cells exhausted their replating ability similar to wt cells (Figure 4A). We also performed a limiting dilution assay³³  to verify the number of functional HSCs in vivo and found an HSC frequency of 1/7000 cells in Gfi1b^ko/ko mice compared with 1 HSC in 46 000 cells in wt mice (Tables 3–4; P ≤ .03). These findings suggested that Gfi1b deficiency enhances the number of functional HSCs approximately 6-7 times (Table 4).

To further examine whether loss of Gfi1b alters self-renewal and multipotency of HSCs, we first transplanted 200 000 bone marrow cells from either wt or Gfi1b-deficient CD45.2 mice in competition with wt CD45.1 bone marrow cells (Figure 4B). Transplanted Gfi1b-deficient bone marrow cells were able to compete with wt CD45.1 cells with regard to blood, bone marrow, spleen, and thymus repopulation, and recipient mice transplanted with Gfi1b-deficient bone marrow cells even showed a significantly higher level of chimerism (measured as the percentage of CD45.2⁺ cells) in blood than recipients that received wt CD45.2 cells (Figure 4C-D). However, when frequencies of CD45.2⁺ myeloid or lymphoid cells were measured in bone marrow, spleen, and thymus, there was no difference between mice that had received wt or Gfi1b-deficient bone marrow (Figure 4E). In addition, we observed a strong and highly significant expansion of transplanted CD45.2⁺Gfi1b-deficient HSCs in blood and bone marrow (Figure 4F-M). Gf1b-deficient (CD45.2⁺) HSCs represented almost 90% of all HSCs in the recipient animals (Figure 4F-J). A similar expansion of Gfi1b-deficient HSCs was also detectable in the peripheral blood of recipients that received Gfi1b-deficient bone marrow, indicating that the phenotype of HSC expansion observed in mice lacking Gfi1b is cell autonomous (Figure 4K-L).

The bone marrow of Gfi1b-deficient mice contains approximately 39 times more phenotypically defined stem cells (HSCs; Figure 2B and Table 1). Yet, limiting dilution experiments suggested only 6 time more functional stem cells in Gfi1b-deficient bone marrow (Tables 3–4). One possible explanation for this discrepancy would be that, as a result of activation, Gfi1b^ko/ko HSCs are at least partially compromised in their stemness and their ability to compete with wt HSCs. To test this, we transplanted a mixture of 50 sorted wt CD45.1⁺ HSCs (defined as above as LSK, CD48⁻, CD150⁺) with either 50 sorted CD45.2⁺wt HSCs or 50 sorted CD45.2⁺ HSCs from Gfi1b^ko/ko mice into syngenic recipient animals (CD45.1⁺; Figure 5A). We observed that Gfi1b^ko/ko HSCs could contribute to the same extent to myeloid and lymphoid lineage differentiation in blood and peripheral organs as wt CD45.2⁺ HSCs (Figure 5B-D). We measured again a significant expansion of Gfi1b-deficient CD45.2⁺ HSCs and LSK cells in the bone marrow and peripheral blood of recipient animals (Figure 5E-I). This expansion of HSCs is comparable with the result obtained after transplantation of the same number of wt and Gfi1b-deficient bone marrow cells (Figures 4I-L and 5E-I).

We next examined whether loss of Gfi1b might exhaust the self-renewal capacity of Gfi1b-deficient HSCs in a serial transplantation assay. We transplanted syngeneic mice (CD45.1) with wt (CD45.1) and CD45.2⁺Gfi1b-deficient bone marrow and tested the degree of chimerism in the primary and secondary recipient by measuring the percentage of CD45.2⁺ cells in the blood (Figure 5J). The experiment showed that the degree of chimerism in a secondary transplantation is maintained. The results of these experiments suggest that Gfi1b^ko/ko HSCs maintain their stemness and multipotency but also their ability to expand in blood and bone marrow beyond wt HSC numbers. It is thus unlikely that the difference between more than 30-fold elevated numbers of phenotypically defined HSCs on one hand and a 6-fold elevated number of functional HSCs (limiting dilution assay) on the other hand is due to a loss of multipotency and self-renewal capacity.

It has been previously shown that HSCs residing in peripheral blood of mice have long-term potential capacity.³⁸  Since we observed a significant expansion of phenotypically defined HSCs in the blood of Gfi1b-deficient mice, we wished to verify whether these blood HSCs represent true functional stem cells. To test this, we transplanted 50 μL of blood originating either from wt or Gfi1b^ko/ko (both CD45.2⁺) mice alongside with 200 000 bone marrow cells from wt CD45.1 mice. We observed that Gfi1b^ko/ko HSCs from peripheral blood were able to give rise to CD45.2⁺ cells (Figure 5K), indicating that Gfi1b^ko/ko HSCs found in blood are functionally intact stem cells. Taken together, these data suggest that Gfi1b^ko/ko HSCs are not compromised in their ability to compete with wt HSCs and maintain their stemness, self-renewal capacity, and multipotency.

A direct comparison of both Gfi1- and Gfi1b-deficient mice showed that loss of Gfi1 led to an increase of HSCs, very likely due to higher cell proliferation as previously reported,^27,28  but that this increase was by far not as pronounced as in Gfi1b-deficient mice (Figure 6A-B). However, when both Gfi1 and Gfi1b were deleted and mice were examined 15 days after the first pIpC injection, we observed a drastic (> 5-fold) reduction of HSC frequencies compared to wt controls (Figure 6A,C). Genotyping of the few HSCs remaining in these double deficient mice showed repeatedly that one Gfi1b allele was not excised, but both Gfi1 alleles were deleted, indicating a functional Cre recombinase (Figure 6D). We also found that, if double Gfi1/Gfi1b deficient mice were observed for a longer time (40 days after the first pIpC injection), HSCs numbers were restored to wt levels (Figure 6C), but these HSCs showed again only a partial excision of the Gfi1b locus. In addition, we found that loss of Gfi1b leads to an enhanced activity of the Gfi1 promoter in HSCs (Figure 7A) and that HSCs, in which Gfi1b was deleted, up-regulated the expression of Gfi1 mRNA (Figure 7B), confirming previously shown ability of Gfi1b and Gfi1 for cross-regulation.^25,26  These data demonstrate that down-regulation of Gfi1b leads to up-regulation of Gfi1 in HSCs and suggest that the complete deletion of both Gfi1 and Gfi1b is incompatible with the generation or maintenance of HSCs.

To further explore how Gfi1b might function in HSCs and how its function differs from Gfi1, we analyzed relative gene expression levels of wt and Gfi1b^ko/ko HSCs using Affymetrix gene arrays. We found that genes encoding cell adhesion molecules and integrins were profoundly deregulated in Gfi1b^ko/ko HSCs (Figure 7C-D). Notably, such proteins as vascular cell adhesion protein 1 (VCAM-1), CXCR4, and integrin α4 that are important to retain HSCs in their endosteal niche^1,6,39-42  were expressed at significantly lower levels on Gfi1b^ko/ko HSCs compared with wt HSCs (Figure 7D-E and Table 5). On the other hand, adhesion molecules such as integrin β1 and β3 that mediate endothelial adhesion^43,44  were significantly up-regulated at mRNA and protein level (Figure 7D-E and Table 5), suggesting that loss of Gfi1b directly or indirectly affects expression of cell surface molecules that have a role in niche organization.

Gfi1 and Gfi1b are 2 distinct and unique proteins encoded by 2 different genes. Both are nuclear zinc finger proteins sharing their N-terminal SNAG repressor domain and 6 C₂H₂ C-terminal zinc-coordinating motifs.¹³  During hematopoiesis, the expression of Gfi1 and Gfi1b is often mutually exclusive,³⁰  but in HSCs both are expressed at the same time, albeit, as we report here, Gfi1 at a much lower levels than Gfi1b. This observation and our observation that Gfi1b expression levels are highest in the quiescent (dormant) CD34⁻ HSC fraction and decrease in subsequent MPP subsets was intriguing and pointed to the possibility that Gfi1b is implicated in regulating the function of HSCs that may be different from the role that its paralogue Gfi1 exerts in stem cells.

The role of Gfi1 in HSCs has been inferred from previously reported experiment, which characterized the LSK bone marrow subset that contains HSCs. Two independent reports suggested that Gfi1 restricts HSCs proliferation and is required to maintain their self-renewal capacity.^27,28  Our analysis provides evidence that Gfi1b-deficient HSCs cycle at a higher rate than wt HSCs, which has also been observed, albeit more indirectly for Gfi1 deficient HSCs.^27,28  However, Gfi1b-deficient mice show a very strong increase of HSC numbers in the bone marrow, spleen, and peripheral blood, which was never observed in Gfi1-deficient mice. AnnexinV staining, BrdU label experiments, and Hoechst staining showed that loss of Gfi1b does not noticeably affect cell survival but increases specifically cell cycle progression in the HSC fraction but not in the MPP subsets. The absolute difference between these BrdU and Hoechst measurements (4% wt HSC undergoing cell cycle in the Hoechst staining as opposed to 11% for the BrdU staining) is probably due to the fact that application of BrdU itself activates stem cells and promotes their entry into cell cycling.²  Our observations with regard to cell cycle progression and expansion of HSCs suggest that Gfi1b very likely exerts functions in stem cells. These roles and functions of Gfi1b are on one hand similar to the function of Gfi1, as loss of Gfi1 and Gfi1b increases cell cycling. On the other hand, the function of Gfi1b in HSCs is different from Gfi1, as Gfi1b-deficient HSCs do not lose their transplantation capacity, as is the case for Gfi1-deficient HSCs.

According to our label retention assay, the fraction of quiescent HSCs is significantly smaller in Gfi1b-deficient mice than in wt animals. In addition, Gfi1b-deficient HSCs show an increased level of ROS that is not seen in wt HSCs. It is likely that ROS play an important role in the expansion of Gfi1b-deficient HSCs since we can demonstrate that NAC, which counteracts ROS, can significantly limit the expansion of HSCs. The high level of ROS also suggests that Gfi1b-deficient HSCs are metabolically more active than wt HSCs and have a higher mitochondrial activity. However, it is also possible that they are exposed to an environment with higher oxygen levels or that Gfi1b regulates genes involved in the regulation of ROS levels.

We do not have evidence for a genetic regulation of ROS levels, but our observations suggest that Gfi1b-deficient HSCs, in addition to being metabolically more active than wt HSCs, may also be localized in another bone marrow microenvironment that is different from the hypoxic endosteal niche, where they are normally kept in a dormant state. HSCs only leave this niche and enter the cell cycle every 60-100 days or after external activation.²  It is possible that loss of Gfi1b causes a delocalization of HSCs away from the hypoxic endosteal niche to a more oxygen rich environment, which drives them into the cell cycle and causes a significant expansion in bone marrow and blood. Our finding that VCAM-1, integrin α4, and CXCR4, which have been described to play an important role in keeping HSCs in a dormant state in their endosteal niche,^1,6,39-42  are all down-regulated in Gfi1b^ko/ko HSCs, points to this as a possible explanation for the activation of Gfi1b-deficient HSCs. Dysregulation of surface proteins regulating retention of HSCs in the endosteal niche could lead to the release of Gfi1b^ko/ko HSCs from the hypoxic endosteal niche.^45,46  Subsequently, these HSCs would become more exposed to oxygen,^35,36  which would result in the appearance of higher ROS triggering entry of these HSCs into the cell cycle.³⁷ 

Several results reported here would be consistent with this notion as the increased numbers of Gfi1b-deficient HSCs in bone marrow, spleen, and blood, their higher cell cycle rate, their higher levels of ROS, and their lower expression levels of CXCR4, VCAM-1, and integrin α4. However, to definitely show whether Gfi1b-deficient HSCs are indeed localized in different anatomically defined locations in the bone marrow would require future studies with in-depth in vivo imaging as demonstrated elegantly in different recent publications.^45,46  In addition, it remains to be shown how precisely Gfi1b regulates the promoters of genes encoding surface and adhesion molecules necessary for niche organization, since we found integrins and adhesion molecules both up- and down-regulated in Gfi1b-deficient HSCs. One exception is CXCR4, which is a Gfi1 target gene⁴⁷  and might be repressed by Gfi1, since we found Gfi1 mRNA expression to be significantly increased in Gfi1b-deficient HSCs because of cross-regulation between these 2 factors.^25,26  A repression of CXCR4 by this mechanism is conceivable and may be implicated in the Gfi1b null phenotype since it has been described that down-regulation of CXCR4 activates HSCs.⁴⁸ 

The most dramatic phenotype in Gfi1b^ko/ko mice remains the strong expansion of HSCs in bone marrow and blood. This feature is clearly cell autonomous since it reappears upon transplantation in recipient mice. It is of interest that such a significant expansion, as seen in Gfi1b adult knockout mice, was not reported for Gfi1b-deficient fetal HSCs²⁹  nor for Gfi1-deficient HSCs.^27,28  This underlines that Gfi1b and Gfi1 indeed exert different functions in stem cells and that both also have different roles in adult and fetal stem cells. In contrast to Gfi1, Gfi1b seems to not only restrict HSCs proliferation but also their expansion, and Gfi1b^ko/ko HSCs do not loose their potential to self-renew nor to initiate multilineage differentiation as was described for Gfi1^−/− HSCs. This is supported by our competitive and serial transplantation experiments. Loss of Gfi1b leads to a massive expansion of HSCs in blood and spleen. One attractive hypothesis for this observation would be that Gfi1b regulates mechanisms that govern egress of HSCs from bone marrow to blood. The fact that the expansion of HSCs in spleen is more than 100-fold in the spleen and 95-fold in the blood of Gfi1b-deficient mice compared with wt mice would support such a hypothesis. In addition, it has been shown that antagonizing the binding of stromal cell–derived factor-1 to its receptor CXC motif receptor-4 (CXCR4) can cause the egress of HSCs to the peripheral blood.⁴⁸  The fact that we see a reduction of CXCR4 on Gfi1b^ko/ko HSCs and a strong presence of HSCs in the blood and spleen of Gfi1b-deficient mice is certainly consistent with our hypothesis. However, further experiments in particular microscopic analyses need to be done to prove whether Gfi1b indeed regulates HSC egress or blood mobilization. Moreover, it is also possible that the few HSCs that are always present in the blood or spleen expand independently after deletion of Gfi1b following a similar mechanism as their bone marrow counterparts.^25,26 

Although Gfi1b^ko/ko mice have a larger pool of both phenotypically and functionally defined HSCs than wt mice, there is a discrepancy between numeric expansion (over 30 times) determined by flow cytometry and functional expansion of HSCs (approximately 6 times) determined by limiting dilution and transplantation. This cannot be explained by the hypothesis that Gfi1b^ko/ko HSCs may be compromised in their self-renewal capacity, because the transplantation of equal numbers of phenotypically defined wt and Gfi1b^ko/ko HSCs resulted in an almost identical contribution to the different hematopoietic lineages. In addition, all transplantation and repopulation experiments confirm that Gfi1b-deficient HSCs are functionally intact. However, all these experiments also showed that the population of HSCs itself was always significantly expanded in both bone marrow and blood. One possibility to reconcile these findings would be that only a fraction of the expanded Gfi1b^ko/ko HSC population is able to give rise to differentiated progeny possibly because the loss of Gfi1b increases the rate of symmetrical over asymmetrical HSC divisions.⁴⁹  A higher rate of symmetrical divisions could lead to an expansion of the HSCs without increasing the contribution to the different lineages. This could explain why an over 30-fold increased number of HSCs does not translate into an equal higher number of functional HSCs as determined by the limiting dilution assay or into equally increased numbers of differentiated cells after transplantations. Albeit compelling, this can only be proven in future studies by monitoring symmetrical and asymmetrical divisions of HSCs using tools such as the described Notch target reporter mice.⁴⁹ 

Since we found that Gfi1 is up regulated in Gfi1b-null HSCs, it is conceivable that the loss of either Gfi1 protein can be compensated by a cross-regulatory mechanism in HSCs as was described for lymphoid cells^25,26  and that the knockout of either gene only reveals those parts of the function of both transcription factors that cannot be compensated by the other. It was thus of interest to investigate the consequences of a simultaneous deletion of both Gfi1 and Gfi1b. The experiment showed that a counter-selection occurs for HSCs that lack both Gfi1 and Gfi1b, since the few HSCs that are still found in pIpC-treated MxCre tg Gfi1^fl/flGfi1b^fl/fl mice always retain one intact Gfi1b allele. This indicates that the deletion of both Gfi1 and Gfi1b is incompatible with the maintenance of an HSC population, and only those HSCs emerge in double KO mice that have escaped Cre-mediated deletion of at least one Gfi1b allele. This also demonstrates that Gfi1b is indispensible in the absence of Gfi1. However, one important conclusion from our data is that in the presence of Gfi1, the modulation of Gfi1b function affects HSC numbers but does not affect their function. This particular feature may render Gfi1b a suitable target for therapeutic intervention in situations where HSC expansion is required,⁵⁰  either to facilitate HSCs harvesting from donors or to enable faster recovery of patients after bone marrow transplantation. This would protect patients against the lethal effects of the procedure and would improve outcome of leukemia and lymphoma.

We are indebted to Mathieu Lapointe and Rachel Bastien for technical assistance; Nancy Laverriere, Marlène Bernier, and Marie-Claude Lavallée for excellent animal care; Eric Massicotte and Julie Lord for fluorescence-activated cell sorting; and Qinzhang Zhu for embryonic stem cell techniques and help generating the conditional mouse strain. We thank the Genome Quebec for performing the arrays and Mark Schlissel and Danae Shultz for providing the Abelson B-cell lines from Gfi1b-deficient bone marrow cells.

This work was supported by grants from the Roche Foundation for Anemia Research and Canadian Institutes of Health Research (MOP-84 238, MOP-94 846; T.M.) and the Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health (J.Z., W.E.P.). T.M. holds a Canada Research Chair (Tier 1), and C.K. is a Cole Foundation fellow.

Contribution: C. Khandanpour performed research, analyzed data, and wrote the manuscript; E.S.-A. performed research and analyzed data; L.V. and M.-C.G. performed research, analyzed data, and edited manuscript; J.Z. and W.E.P. provided vital reagents; T.O. assisted in generating the described mouse strain; C. Kosan assisted in generating the described mouse strain and edited the manuscript; and T.M. designed the research, analyzed data, wrote the manuscript, and provided funding.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Tarik Möröy, Institut de recherches cliniques de Montreal, IRCM, 110 avenue des Pins Ouest, H2W 1R7, Montreal, QC, Canada; e-mail: Tarik.Moroy@ircm.qc.ca.