Hematopoietic stem cells (HSCs) are the best-characterized tissue-specific stem cells, yet experimental study of HSCs remains challenging, as they are exceedingly rare and methods to purify them are cumbersome. Moreover, genetic tools for specifically investigating HSC biology are lacking. To address this we sought to identify genes uniquely expressed in HSCs within the hematopoietic system and to develop a reporter strain that specifically labels them. Using microarray profiling we identified several genes with HSC-restricted expression. Generation of mice with targeted reporter knock-in/knock-out alleles of one such gene, Fgd5, revealed that though Fgd5 was required for embryonic development, it was not required for definitive hematopoiesis or HSC function. Fgd5 reporter expression near exclusively labeled cells that expressed markers consistent with HSCs. Bone marrow cells isolated based solely on Fgd5 reporter signal showed potent HSC activity that was comparable to stringently purified HSCs. The labeled fraction of the Fgd5 reporter mice contained all HSC activity, and HSC-specific labeling was retained after transplantation. Derivation of next generation mice bearing an Fgd5-CreERT2 allele allowed tamoxifen-inducible deletion of a conditional allele specifically in HSCs. In summary, reporter expression from the Fgd5 locus permits identification and purification of HSCs based on single-color fluorescence.
Hematopoietic stem cells (HSCs) function to maintain blood homeostasis throughout life via their unique ability to differentiate into all blood cell types and to self-renew. These properties, along with the robust ability of HSCs to engraft myeloablated recipients in the setting of BM transplantation, have established the clinical paradigm for therapeutic stem cell use (Weissman, 2000).
Originally described by Till and McCulloch (1961), HSCs were first experimentally defined by their ability to form macroscopic colonies in the spleens (CFU-S) of irradiated recipients after BM transplantation that histological examination revealed contained multiple blood lineages, and cytological examination revealed were clonally derived (Becker et al., 1963). Together with the demonstration that a subset of CFU-S colonies had the potential to reform colonies when transplanted into secondary recipients (Siminovitch et al., 1963), the defining properties of hematopoietic stem cells—multipotency and self-renewal—were established. In the 50 yr since these seminal studies were conducted, the experimental study of HSCs has flourished, leading to a profound level of understanding of their biology. These efforts were enabled through the development of several in vivo and in vitro assays that permitted evaluation of HSC self-renewal and multilineage potential, and by methods that allowed purification of HSCs by FACS. HSCs were initially reported to be enriched within the Thy1lowLineage− fraction of the murine BM (Muller-Sieburg et al., 1986), and subsequently cells with a Thy1lowLineage−Sca1+ immunophenotype were shown to possess long-term multilineage repopulating activity (Spangrude et al., 1988). The immunophenotype of HSCs was further refined, culminating with the demonstration that single cells purified from the Lineage−Sca1+c-kit+ (LSK)CD34−/low fraction of the BM of adult mice could function to long-term multilineage reconstitute irradiated recipients at the clonal level (Osawa et al., 1996). Additional cell surface markers that have also been used to enrich for HSC activity include: CD105 (Chen et al., 2002), Flk2/Flt3 (Christensen and Weissman, 2001), CD201/PROCR (Balazs et al., 2006), ESAM (Ooi et al., 2009; Yokota et al., 2009), and CD150, CD48, and CD244 (Kiel et al., 2005a) among others. In addition to immunophenotype, intravital dye efflux activity has also proven to be an effective strategy for enriching for HSC activity (Bertoncello et al., 1985; Wolf et al., 1993; Goodell et al., 1996).
Although immunophenotype combined with flow cytometry has become the principle technique used for identifying and studying diverse cells types, genetically engineered reporter strains have also enabled the identification and study of other cell types, including tissue-specific stem cells from other organs. For example, rapidly cycling intestinal stem cells were identified with the use of an Lgr5 reporter (Barker et al., 2007), whereas a population of more slowly cycling stem cells in the intestinal crypt were marked with a reporter for telomerase (Montgomery et al., 2011). In the developing embryo, reporter strains for Isl1 (Laugwitz et al., 2005) and WT1 (Zhou et al., 2008) have been combined with lineage-tracing experiments to identify cardiac progenitors in the developing heart. In the skin, a Tet-inducible H2B-GFP reporter stain was used in conjunction with a keratinocyte-specific driver to isolate label-retaining stem cells in the epidermis (Tumbar et al., 2004). A similar H2B-GFP label retention strategy was later used by two independent groups to explore the turnover of HSCs, showing that a label-retaining population of cells with potent HSCs activity resides in a state of prolonged dormancy during steady-state homeostasis (Wilson et al., 2008; Foudi et al., 2009). Importantly, depending on vector design, introducing reporter cassettes into specific genomic loci (knock-in) can also lead to the disruption of the targeted gene, permitting analysis of the null (knockout) genotype when targeted alleles are crossed to homozygosity.
With the goals of identifying novel genes that could be used to specifically report on HSC activity within the murine BM, we performed a system-wide microarray screen of hematopoietic stem, progenitor, and effector cells, and identified a set of genes whose expression was highly restricted to the HSC compartment. Generation of mice with targeted reporter knock-in/knock-out alleles at three of the identified genes, Sult1a1, Clec1a, and Fgd5 revealed that whereas knockout of Sult1a1 and Clec1a were viable and had normal HSC function, nullizygosity of Fgd5 was embryonic lethal at midgestation, though the generation and function of definitive HSCs was not affected by loss of Fgd5. Of the three reporter alleles, only Fgd5 explicitly marked immunophenotypic HSCs in the adult marrow at steady state and after transplantation. BM cells isolated based solely on reporter signal of the Fgd5 reporter mice showed robust HSC activity, with all stem cell activity residing within the labeled fraction. These results demonstrate that HSCs can be identified and purified from the BM of Fgd5 reporter mice by single color fluorescence. Finally, the development of Fgd5-CreERT2 mice permitting inducible deletion of Floxed alleles specifically in HSCs represents an invaluable genetic resource for exploring gene function in HSC biology.
Systems-wide microarray screen identifies genes with HSC-restricted expression in the adult hematopoietic system
To identify genes specifically expressed in HSCs within the hematopoietic system, we compiled the expression profiles of 37 different hematopoietic cell types comprising the vast majority of hematopoietic progenitor and effector cells (Fig. 1 A). These datasets include published microarray data from our group and others that were carefully curated from publicly available databases. As many of these datasets were generated in different laboratories, we subjected them to a number of quality control (QC) measures in accordance with current standards using the ArrayQualityMetrics package of R/Bioconductor. In total, 122 expression profiles passed QC and were normalized together in a single database. Using this database, we were readily able to identify genes that showed highly restricted expression in diverse hematopoietic cell types (Fig. 1 B). Analysis of such cell type–specific gene lists indicated that previously established and validated cell type–specific genes could be identified (Fig. 1 B). These included genes known to mediate critical functions in specific cell types, as well as genes whose products are routinely used to phenotypically define different cell types with no known function in the specific cell type (Fig. 1 B). For example, Ncr1, which is critical in NK cells (Gazit et al., 2006), Bcl11b, which is involved in specifying T cell identity (Wakabayashi et al., 2003), and the adult α-globin in erythroblasts (Pászty et al., 1995) were highly restricted to these cells in our database (Fig. 1 B). Similar results were obtained for other hematopoietic cell types and, in all cases the cell type–specific genes were associated with a very high degree of statistical confidence due to the fact that only FACS-purified cells were used in the generation of the data, and also due to the large number of samples and biological replicates analyzed (Tables S1 and S2).
We next sought to identify genes predominantly expressed in HSCs (immunopurified as LSKFlk2−CD34low/−) in comparison to their downstream progenitor and effector progeny. This analysis identified 323 probe sets corresponding to 235 unique annotated genes with highly restricted expression in HSCs (Fig. 1 C, Fig. S1, and Table S2). This list overlapped significantly with the results of previous studies in which expression profiling of different hematopoietic cell types has been used to define HSC transcriptional signatures and explore their biology (Chambers et al., 2007a; Gazit et al., 2013; Fig. 1 D). Interestingly, many of the genes we identified were also expressed at low levels in multipotent progenitors (MPP1s), which represent the most proximal progenitor to HSCs and have previously been referred to in the literature as short-term HSCs (ST-HSCs; Fig. 1 C). Among the HSC-enriched genes were several that encode proteins with well-established functional roles in HSC biology, such as Cdkn1c (Matsumoto et al., 2011; Zou et al., 2011), Meis1 (Pineault et al., 2002), and Ndn (Kubota et al., 2009), in addition to many genes that have not yet been reported to have a functional role in HSC biology. We chose to focus on three genes of this latter group, Clec1a, Fgd5, and Sult1a1, which showed highly restricted expression in our database (Fig. 1 C). In addition to expression specificity, these genes were selected, in part, based on consideration of their genomic structure (intron/exon, repetitive elements), which suggested that they would be amenable to targeting by homologous recombination. Of these, Clec1a encodes a C-type lectin type II transmembrane receptor that has been shown to be expressed in human and rat dendritic and endothelial cells (Colonna et al., 2000; Sobanov et al., 2001; Thebault et al., 2009), and has been reported to play an immunomodulatory role in allograft tolerance in rats (Thebault et al., 2009). Sult1a1 encodes a cytosolic transferase studied in human cells for its ability to conjugate sulfate to various phenolic substrates (Wilborn et al., 1993; Raftogianis et al., 1997; Hildebrandt et al., 2009). Fgd5 encodes a protein predicted to have GTP-exchange (guanine nucleotide exchange factor [GEF]) activity that has been studied exclusively in the context of endothelial cell biology (Cheng et al., 2012; Kurogane et al., 2012). To our knowledge, Fgd5, Clec1a, and Sult1a1 have not previously been studied in the context of HSC biology. Collectively, these results demonstrate that expression of Fgd5, Clec1a, and Sult1a1 is predominantly confined to HSCs in the adult hematopoietic system.
Immunophenotypic HSCs are labeled by mCherry in Fgd5mCherry/+ mice
To generate reporter mice and begin to explore the possible roles for Clec1a, Sult1a1, and Fgd5 in HSC biology, constructs were made to target each locus by homologous recombination in embryonic stem (ES) cells using a knock-in/knock-out strategy. In all cases, constructs were designed to place a fluorescence reporter cassette (either mCherry into Fgd5, or eGFP•CreERT2 into Clec1a and Sult1a1) in-frame with the protein coding sequence that would be expected to be expressed under endogenous regulatory control. At the same time, correct targeting of the loci is expected to generate null alleles for each of the genes. Sequence verified targeting constructs were introduced into ES cells derived from C57BL/6 mice, correctly targeted ES cell clones were identified by Southern blotting, and germline transmission of the targeted alleles was established (not depicted).
To characterize the utility of the targeted alleles to label HSCs, we isolated BM cells from young adult Fgd5mCherry/+, Sult1a1GFP/+ and Clec1aGFP/+ mice and control littermates and immunostained for HSCs using a panel of well-defined markers (LSKCD48−CD150+). Disappointingly, the BM of young adult Clec1aGFP/+ mice showed almost negligible expression in HSCs, whereas HSCs identified from the BM of Sult1a1GFP/+ mice were completely negative for reporter fluorescence (Fig. 2 A). In contrast, immunophenotypic HSCs from the Fgd5mCherry/+ mice were almost uniformly positive for reporter signal (mCherry; Fig. 2 A). Although there could be several explanations for the absence of signal in HSCs of Clec1aGFP/+ and Sult1a1GFP/+ mice, we focused on the Fgd5mCherry/+ mice for further characterization.
Whereas Fgd5mCherry/+ mice clearly showed labeling of immunophenotypic HSCs, we sought to further determine the spectrum of BM cells expressing mCherry. To address this, we gated BM cells from the Fgd5mCherry/+ mice into mCherry+ and mCherry− fractions, and then determined the immunophenotype of these cells by co-staining with a panel of markers. This analysis revealed that in contrast to mCherry− cells, a significant fraction of the mCherry+ cells were negative for the lineage markers associated with mature blood cells (B220, Mac1, GR-1, Ter119, CD3, CD4, and CD8); positive for c-Kit and Sca1; negative for CD48; and positive for CD150 (Fig. 2 B). By gating sequentially through these markers, the vast majority of the mCherry+ cells of Fgd5mCherry/+ BM co-stained with markers of immunophenotypic HSCs (LSKCD48−CD150+; Fig. 2 B, bottom).
Our expression profiling data showed that whereas Fgd5 expression in the murine BM was almost exclusively restricted to HSCs, low-level expression was also detected in multipotent progenitor cells. To examine this, we stained the BM of Fgd5mCherry/+ mice with several different marker combinations that are commonly used to identify HSCs and discriminate them from multipotent progenitor cell subsets within the primitive LSK fraction of the murine BM. Using markers associated with the “Slam code” (Kiel et al., 2005b), the vast majority of HSCs (LSKCD150+CD48−) were fluorescently labeled (Fig. 2 C). In contrast, the most proximal multipotent progenitors (LSKCD150−CD48−) showed low-level expression of the reporter, whereas the more distal LSKCD150−CD48+ progenitors were predominantly negative. Similarly, when CD34 and Flk2 were used to immunophenotypically define HSCs and multipotent progenitors (Osawa et al., 1996; Christensen and Weissman, 2001; Rossi et al., 2005), HSCs (LSKFlk2−CD34−) were largely positive, whereas the MPP1/ST-HSC (LSKFlk2−CD34+) population expressed lower levels of signal, and the MPP2 subset (LSKFlk2+CD34+) showed very little signal (Fig. 2 D), consistent with our microarray data (Fig. 1). Similar results were found using other marker immunophenotypic strategies to identify HSCs and multipotent progenitors including PROCR/CD201 (Balazs et al., 2006; Fig. 2 E) and ESAM (Ooi et al., 2009; Yokota et al., 2009; Fig. 2 F). Collectively, these results indicate that immunophenotypic HSCs are almost exclusively labeled in the BM of Fgd5mCherry/+ mice.
Fgd5 deficiency does not impair HSC function and is not required for definitive hematopoiesis
Because the targeting of the Fgd5 locus places a mCherry cassette into the first exon of the Fgd5 coding region and is predicted to generate a null allele, we wanted to determine if inactivation of one or both Fgd5 alleles would affect HSC function. To address this in the setting of Fgd5 heterozygosity, we competitively transplanted 106 BM cells from Fgd5mCherry/+ or wild-type (Fgd5+/+) control littermates (CD45.2) against 1 × 106 wild-type BM cells (CD45.1) into lethally irradiated congenic recipients (CD45.1). Transplant recipients were bled at monthly intervals and reconstitution of CD45.2 test cells and their contribution to myeloid lineage granulocytes, macrophages/monocytes, lymphoid lineage B cells, and T cells, was determined (Fig. 3 A). This showed that Fgd5 heterozygosity had no adverse effect on HSC function in respect to total repopulation or lineage potential in primary (1°) transplant recipients. To further challenge the Fgd5mCherry/+ and wild-type HSCs, serial transplantation into secondary (2°) recipients was performed. Peripheral blood analysis up to 20 wk after transplant revealed that HSCs derived from both Fgd5mCherry/+ and Fgd5+/+ mice robustly reconstituted 2° hosts showing comparable repopulating activity, and no differences in lineage output (Fig. 3 B). Thus, inactivation of one Fgd5 allele has no adverse consequence on the long-term repopulating or self-renewal potential of HSCs.
We next sought to determine if Fgd5 nullizygosity would have an impact on HSC function. To address this, we set Fgd5mCherry/+ X Fgd5mCherry/+ crosses but were unable to obtain any viable Fgd5mCherry/mCherry offspring, indicating that ablation of Fgd5 is lethal for embryos. We examined the requirement for Fdg5 during embryonic development and found that whereas no Fgd5mCherry/mCherry embryos could be identified at embryonic day 13.5 (E13.5) or later, null embryos could be obtained at Mendelian numbers at E11.5 (Fig. 3 C). Gross examination and genotyping of embryos from timed matings showed that although many Fgd5-null embryos appeared morphologically normal at E11.5, the presence of resorbed embryos and Fgd5mCherry/mCherry embryos with clear morphological abnormalities at E12.0 indicate that most Fgd5-null embryos die around E11.5–E12.0 (Fig. 3, C and D). Histological examination of the aorta-gonad mesonephros (AGM) region of E11.5 control and morphologically normal Fgd5mCherry/+ embryos revealed that all endothelial cells, including the aorta, were mCherry+ (Fig. 3 E), which is consistent with the previously reported pan-endothelial cell expression of Fgd5 in developing embryos (Cheng et al., 2012). Because the timing of death (E11.5–E12.0) is shortly after the developmental time point at which definitive HSCs first emerge in the embryo at the AGM region (Dzierzak and Speck, 2008), the possibility that definitive hematopoiesis may be defective or impaired in absence of FGD5 was raised. To test this possibility directly, we dissected the AGM region of E11.5 embryos derived from Fgd5mCherry/+ X Fgd5mCherry/+ crosses and cultured them for 4 d at an air/liquid interface using an adapted protocol (Medvinsky and Dzierzak, 1996; Taoudi et al., 2008), and then competitively transplanted all the cells from the AGM explants into irradiated recipients (Fig. 3 F). These experiments showed that although there was some variability in lineage output, as might be expected with this protocol, AGM explants arising from Fgd5+/+, Fgd5mCherry/+, or Fgd5mCherry/mCherry embryos all gave rise to HSCs capable of long-term multilineage reconstitution in 1° recipients (Fig. 3 G). To further test the functional capacity of the Fgd5-deficient HSCs, 2 × 106 BM cells from the 1° hosts were serially transplanted into 2° recipients. These experiments showed that AGM-derived HSCs of all Fgd5 genotypes were able to give rise to long-term, multilineage reconstitution in 2° hosts (Fig. 3 H).
Collectively, these results indicate that Fgd5 is required for embryonic development, but is not required for the generation of definitive HSCs, and further that loss of one or both Fgd5 alleles does not impair the long-term self-renewal or multilineage differentiation potential of HSCs in serial transplantation assays.
It is also important to note that although the targeting of Clec1a or Sult1a1 did not faithfully label HSCs, we were nonetheless able to determine that neither gene was required for embryonic development as homozygous null mice were born at Mendelian numbers and appeared phenotypically normal. Moreover, BM from mice harboring Clec1a or Sult1a1 alleles functioned normally in primary transplantation assays, indicating that neither gene was required for HSC long-term repopulating activity (unpublished data).
Fgd5mCherry identifies BM cells with potent hematopoietic stem cell activity
Having determined that the mCherry+ BM fraction of Fgd5mCherry/+ mice labels cells that express markers consistent with immunophenotypic HSCs (Fig. 2), and also that the targeted allele had no adverse effect on HSC function (Fig. 3), we next sought to test the functional activity of the labeled cells directly. We sorted defined numbers of cells based solely on mCherry positivity from the BM of Fgd5mCherry/+ mice (CD45.2), and competitively transplanted them into lethally irradiated congenic recipients (CD45.1). We also sorted immunophenotypic HSCs (LSKCD150+CD48−, hereafter referred to as HSCSlam) and competitively transplanted these in parallel. In a series of independent experiments, mCherry+ cells, and HSCSlam were transplanted at 200, 120, 40, 20, or 5 cells per recipient, and peripheral blood reconstitution was monitored for 16 wk (Fig. 4, A–E). At all transplant doses, the mCherry+ cells and HSCSlam gave rise to long-term donor chimerism that was statistically comparable (Fig. 4 F). In each of the 200, 120, 40, and 20 cell transplants, all recipient mice transplanted with either mCherry+ cells or HSCSlam showed donor-derived multilineage reconstitution 16 wk after transplant, whereas at the 5-cell dose, 11/13, and 8/13 recipients were multilineage reconstituted when mCherry+ cells or HSCSlam were transplanted, respectively (Fig. 4, A–E). These results demonstrate that the mCherry+ fraction of Fgd5mCherry/+ BM is highly enriched with potent repopulating activity that is functionally comparable, on a per cell basis, to HSCs sorted by rigorous immunophenotypic markers.
In addition to the ability to give rise to long-term multilineage reconstitution in 1° transplant recipients, HSCs are most rigorously experimentally defined by their ability to sustain activity during serial transplantation. To address this, we again isolated mCherry+ cells or HSCSlam from Fgd5mCherry/+ or Fgd5+/+ mice (CD45.2), respectively, and competitively transplanted 250 cells of each into irradiated congenic (CD45.1) recipients. Analysis of 1° recipients revealed, as before, that the mCherry+ cells and HSCSlam performed comparably (Fig. 5 A). 21 wk after transplant, 2 × 106 BM cells derived from the 1° recipients were transplanted into 2° hosts (CD45.1). Throughout the 28-wk-long experiment, all 2° hosts showed donor-derived multilineage reconstitution (Fig. 5 B). 31 wk after transplant, BM cells were harvested from the 2° hosts and 5 × 106 cells were transplanted into tertiary (3°) recipients (CD45.1). As we had observed in the 2° hosts, both the mCherry+ donor cells and HSCSlam controls continued to show potent long-term multilineage repopulating activity in all of the 3° recipients (Fig. 5, C and D). These experiments demonstrate that the mCherry+ fraction of Fgd5mCherry/+ BM contains potent multilineage repopulating potential, and extensive self-renewal potential in vivo.
We next analyzed the BM reconstitution of recipients that had been transplanted with either mCherry+ cells or HSCSlam. Similar to the chimerism observed in the peripheral blood (Fig. 4), the BM was robustly reconstituted with CD45.2 donor-derived cells regardless of whether mCherry+ cells or HSCSlam had been transplanted (Fig. 6). Co-staining the BM with a panel of markers showed, as we had observed in the steady state (Fig. 2), that the mCherry+ signal was restricted to the immunophenotypic HSC (LSKflk2−CD34−) compartment, with only a minor fraction of the LSK multipotent progenitors expressing lower levels of labeling (Fig. 6 C). These results show that the mCherry+ cells are able to self-renew to give rise to immunophenotypic HSCs in vivo and, further, that the near exclusive labeling of HSCs observed in the BM of Fgd5mCherry/+ mice is faithfully maintained even after the extensive challenge of primary transplantation.
All HSC activity resides within the mCherry+ fraction of Fgd5mCherry/+ BM
We next sought to determine if all HSC activity was confined to the mCherry+ fraction of the Fgd5mCherry/+ BM. To address this, we sorted the BM of Fgd5mCherry/+ mice into mCherry+ and mCherry− fractions, and competitively transplanted either 100 or 100,000 cells, respectively, into irradiated recipients in two independent experiments (Fig. 7, A–C). As we had previously observed (Figs. 4–5), the 100-mCherry+ cell transplants all yielded robust long-term, multilineage reconstitution in both experiments (Fig. 7, B–D). In contrast, the 100,000-mCherry− cell transplants yielded only short-term reconstitution that gradually diminished to very low levels at later time-points (Fig. 7, B and C). Peripheral blood analysis of the few remaining mCherry− donor-derived cells at 16-wk (experiment #1) or 24-wk (experiment #2) after transplant revealed that virtually all were B and T cells, which can be long-lived (Fig. 7 D). Importantly, although the mCherry+ cell transplants all gave rise to sustained, high levels of granulocyte chimerism indicating robust, ongoing HSC activity (Bhattacharya et al., 2006; Bryder et al., 2006), granulocyte reconstitution was progressively extinguished in the recipients transplanted with the mCherry− fraction indicating an absence of HSC activity (Fig. 7, E and F). The sole exception to this was one of the recipients of 100,000 mCherry− cells in experiment #1 showed very low donor-derived granulocyte reconstitution 16 wk after transplant (Fig. 7 E). Although we cannot formally exclude the possibility that HSC activity resides in the mCherry− fraction of the BM at very low frequency, the fact that only 1 out of 10 independent recipients transplanted with 100,000 mCherry− cells showed minor granulocyte reconstitution 16 wk after transplant, indicates that most, if not all, of HSC activity resides in the mCherry+ fraction of the Fgd5mCherry/+ mice.
Pan-endothelial Fgd5 expression in adult BM, and inducible HSC-specific Cre-mediated deletion revealed by next generation Fgd5 mice
Our functional data to this point indicates that Fgd5 would be an ideal locus for the construction of additional genetic tools designed to study the functional, molecular, and therapeutic properties of HSCs. To expand the repertoire of HSC-specific genetic resources we generated two additional strains by targeting the Fgd5 locus in ES cells. With the intent of generating a brighter Fgd5 reporter that could be detected using 488-nm lasers common to most flow cytometers and FACS machines, we knocked-in tandem ZsGreen cassettes separated by a 2A peptide into the Fgd5 locus. Similar to the Fgd5mCherry/+ mice (Fig. 2), the vast majority of ZsGreen+ cells were lineage (lin−) and CD48−, and c-Kit+, Sca1+, and CD150+ (Fig. 8 A). Consistent with the brighter expression of the Fgd5ZsGr•ZsGr/+ reporter, and also with the expression of Fgd5 transcript observed in multipotent progenitor subsets (Fig. 1 C), LSKCD48−CD150− (MFI 6,031) and to a lesser extent, LSKCD48+CD150− (MFI 1,566) progenitor subsets, expressed Fgd5-ZsGreen signal, albeit at lower levels than observed in LSKCD48−CD150+ HSCs (MFI 11,343; unpublished data). Fgd5 reporter signal was detected in all HSC subtypes fractionated by differential CD150 expression in young (4 mo) and mid-aged (14 mo) mice (Fig. 8, B and C; Beerman et al., 2010; Morita et al., 2010). Interestingly, CD150hi HSCs, which are believed to be the most primitive HSC subset (Beerman et al., 2010; Morita et al., 2010), showed the highest mean fluorescence intensity of Fgd5 reporter signal, followed by CD150lo HSCs and CD150neg HSCs (Fig. 8 C), suggesting that the intensity of Fgd5 reporter expression tracks with HSC potency, and further that the Fgd5 reporter faithfully tracks the changes in the HSC population during aging.
Using this strain we next sought to examine Fgd5 expression in the adult BM using laser-scanning cytometry (Nombela-Arrieta et al., 2013). Consistent with previous studies suggesting pan-endothelial expression of Fgd5 (Cheng et al., 2012; Kurogane et al., 2012), all blood vessels in the adult BM expressed high levels of the Fgd5 reporter (Fig. 8 D and Video 1).
The most widely used genetic resources for investigating gene function in HSCs via loss-of-function approaches are the Mx1-Cre (Kühn et al., 1995) and Vav1-Cre (de Boer et al., 2003) strains. Although both strains are robust, they both have limitations for specifically studying gene function in HSCs as both are pan-hematopoietic. Moreover, Vav1-Cre is noninducible, thus precluding temporal control of gene deletion, and although Mx1-Cre is inducible, its induction by the synthetic double-stranded RNA polyinosinic-polycytidylic acid (polyI:polyC) elicits a systemic interferon response that is known to activate and drive HSCs into cycle (Baldridge et al., 2010; Essers et al., 2009) potentially confounding interpretations using this strain, particularly in regard to HSC cell cycle status and regulation of quiescence. To circumvent these issues, we targeted an inducible CreERT2 cassette along with a ZsGreen reporter cassette into the Fgd5 locus in ES cells to derive Fgd5ZsGr•CreERT2/+ mice. These mice were then crossed to mice bearing a floxed-stop dTomato reporter allele at the Rosa26 locus. After tamoxifen induction, we examined the specificity and efficacy of Cre-mediated deletion. These experiments showed that dTomato expression after Cre-mediated excision of floxed-stop cassette was highly restricted to ZsGreen+ LSKCD150+ HSCs (Fig. 9 A), with essentially no dTomato expression detected in other hematopoietic compartments (Fig. 9 B).
In this study, we set out to identify genes with restricted expression in the HSC compartment of the murine BM, and then target the endogenous loci of several identified genes in mouse ES cells to generate reporter knock-in/knock-out alleles. Mice bearing such alleles could potentially be used to identify HSCs by single-color fluorescence without the need for immunostaining, which could have great utility for addressing outstanding questions related to HSC biology. At the same time, our knock-in/knock-out approach would allow us to examine the requirement of the targeted genes for HSC development and function. To achieve these goals, we used a microarray approach in which we compared the expression profiles of highly purified HSCs to that of 36 downstream progenitor and effector cell types. Previous studies using related approaches have been successful in identifying genes that function in HSCs or in downstream populations (Ivanova et al., 2002; Park et al., 2002; Forsberg et al., 2005; Kiel et al., 2005a; Shojaei et al., 2005; Balazs et al., 2006; Luckey et al., 2006; Chambers et al., 2007a,b; Seita et al., 2012; Gazit et al., 2013), as well as genes whose products serve as antigens that have been used to facilitate identification of HSCs such as Esam (Forsberg et al., 2005), the Slam code (CD150, CD48, and CD244; Kiel et al., 2005a), and Procr/CD201 (Balazs et al., 2006). Mindful of the fact that HSCs share several functional attributes with their proximal multipotent progenitor progeny, and also, to a lesser degree, with downstream oligo-potent and lineage-restricted progenitors, we included such populations in our microarray screen reasoning that this would allow us to more precisely identify genes with HSC-restricted expression. With this said, it must be recognized that identifying a gene as “specific” to any cell type is ultimately limited by the spectrum and comprehensiveness of the samples studied. Bearing this potential caveat in mind, using the hematopoietic database assembled for this study, we were able to identify 235 annotated genes with highly restricted expression in HSCs, many of which have not been previously studied in the context of HSC biology.
To assess the potential functional role that such novel identified genes might play in HSCs, and to increase the likelihood of identifying a genomic locus that upon targeting would lead to faithful HSC labeling, we focused on three genes with highly HSC-restricted expression. The knock-in/knock-out targeting strategy we used allowed us to determine that neither Clec1a nor Sult1a1 are required for normal mammalian development, and both appeared to be dispensable for HSC function in transplantation experiments (unpublished data). In contrast, whereas Fgd5 heterozygotes developed normally and showed no deficit in HSC function, Fgd5 nullizygosity was embryonic lethal at mid-gestation, indicating a critical, nonredundant function for FGD5 during development. Recent studies have shown that Fgd5 expression outside of the hematopoietic system is predominantly restricted to endothelial cells in developing and adult mice and in zebrafish embryos (Cheng et al., 2012). Fgd5 is also expressed in several human endothelial cell lines, where it has been suggested to play a role regulating CDC42 activity during capillary formation (Kurogane et al., 2012). The importance of Fgd5 in endothelial cell biology was recently demonstrated in a study in which murine Fgd5 was knocked-down by siRNA, and overexpressed, showing that FGD5 regulates vascular pruning during endothelial cell remodeling (Cheng et al., 2012). These studies raise the possibility that the embryonic lethality associated with loss of FGD5 may result from defective vasculogenesis. Consistent with this we observed robust endothelial expression of FGD5 in the developing embryo and also in adult BM. With regards to hematopoiesis, we showed that despite the midgestation lethality, Fgd5-nullizygosity did not impair the formation or function of definitive HSCs (Fig. 3), which retained self-renewal and multilineage differentiation potential comparable to wild-type controls.
Several previous studies have targeted a variety of loci with the intent of establishing a reporter that specifically identifies HSCs. Mice bearing a reporter for the Gata2 transcription factor that is not specifically expressed in HSCs (Fig. S1) were found to be useful for enriching for HSC activity when sorted in combination with immunostaining for Sca1 (Suzuki et al., 2006). Similarly, although mice bearing a reporter allele at the Abcg2 locus predominantly labeled Ter119+ erythroid cells in the murine BM as expected from expression analysis (Fig. S1), HSCs could be identified when used in combination with side-population activity, and antibody staining to exclude lineage+ cells (Tadjali et al., 2006). Interestingly, although Bmi-1 is broadly expressed throughout hematopoiesis (Fig. S1), mice bearing a GFP knock-in allele at the Bmi-1 locus, which is critical for HSC function (Park et al., 2003), were used to demonstrate that BM cells expressing the highest levels of GFP contained HSC activity when sorted in combination with additional HSC markers (Hosen et al., 2007). In an elegant study, Goyama et al. (2008) targeted the Evi-1 locus, which is required for HSC function, by cleverly knocking-in an Evi-1 cDNA-IRES-GFP rescue/reporter cassette, which largely rescued HSC activity (Kataoka et al., 2011). This study showed that although the majority of cells of the primitive LSK compartment were GFP-labeled in these mice, only the GFP+ fraction when sorted in combination with additional HSC markers including Lin−Sca1+c-kit+ exhibited HSC activity (Kataoka et al., 2011). Thus, to the best of our knowledge, although several reporter mice have been developed that label hematopoietic stem and progenitor cells, the goal of establishing a reporter strain that faithfully and more specifically labels HSCs based on single-color fluorescence has not been previously achieved.
Although Clec1a and Sult1a1 were identified as being among the most HSC-restricted genes from our expression screen, targeting expression reporter cassettes into the coding region of these genes did not lead to expression in the HSC compartment. Several reasons could underlie this disappointing outcome, including the possibility that targeting the loci in the manner that we did sufficiently disrupted regulatory elements necessary for expression in HSCs. In contrast to Clec1a and Sult1a1, targeting of the Fgd5 locus yielded a reporter that almost exclusively labels HSCs in the murine BM. In addition to being essentially synonymous with a myriad of markers that are used to immunophenotypically identify HSCs (Osawa et al., 1996; Kiel et al., 2005b; Balazs et al., 2006; Bryder et al., 2006; Ooi et al., 2009), cells sorted solely on reporter signal from the Fgd5mCherry/+ mice had potent HSC activity that matched the functional potential of rigorously immunopurified HSCs (Kiel et al., 2005b). We demonstrated that the Fgd5 reporter effectively marked all HSC activity in the steady-state BM, which is remarkable in light of evidence showing that even widely used protocols that have proven very effective in identifying HSCs do not strictly identify all cells with HSC activity (Morita et al., 2006; Weksberg et al., 2008). Moreover, the demonstration that Fgd5 reporter activity exclusively identified immunophenotypic HSCs in the BM of transplanted mice suggests that even under experimental settings requiring extensive HSC activity and self-renewal Fgd5 reporter expression remained a powerful tool for identifying HSCs. The Fgd5 reporter and Fgd5-CreERT2 mice reported here represent invaluable resources for studying the pathways and mechanisms that govern the central properties of HSCs under diverse settings.
MATERIALS AND METHODS
Hematopoietic expression database.
Microarray data were generated on the Affymetrix 430 2.0 platform and included previously unpublished data generated in our laboratory in addition to datasets that were curated from Gene Expression Omnibus (complete list of accession nos. available in Table S1). All datasets were subjected to quality control (QC) measures provided in the ArrayQualityMetrics package of R/Bioconductor. Datasets were normalized (gcRMA) using R/Bioconductor. To identify cell type–specific genes we applied three filters: (1) only probe sets that had expression values ≥ log 7.0 were included; (2) Ratio of expression difference for the cell type of interest to all others had to be greater than fourfold; (3) Statistical significance for the cell type of interest to all others had to be P < 10−6 by Student’s t test.
Gene targeting and mice.
The targeting constructs were built using the W vector (provided by J. Segal and K. Rajewsky, Immune Disease Institute, Boston). Homologous arms were cloned by PCR from genomic DNA of C57BL/6 mice. Constructs were sequenced validated. ES cells (B6/3) derived from C57BL/6 mice were electroporated with linearized constructs and clones were positively selected with G418 and negatively selected by the diphtheria toxin cassette of the constructs. Clones were manually picked, expanded, and screened by Southern blot. Positive clones were expanded, validated, and injected into blastocysts, and germline transmission was confirmed. All strains used were on a C57BL/6 background including CD45.1 and CD45.2 congenic, and the Rosa26-LoxP-STOP-LoxP-tdTomato (stock# 007914) was obtained from The Jackson Laboratory. All experiments involving mice were done per institutional guidelines of The Harvard Medical School Standing Committee on Animals, with IACUC approval.
FACS analysis and cell sorting.
BM mononuclear cells prepared over Histopaque 1083 (Sigma-Aldrich) were stained for 1.5 h in PBS 2 mM EDTA, 2% FBS at 4°C with combinations of the following antibodies: the lineage markers Ter119, Mac-1 (m1/70), Gr-1 (8C5), CD3 (17A2), CD4 (RM4-5), CD8 (53–6.7), B220 (RA3-6B2), and IL7Ra (A7R34); CD34 (RAM34), Flk2 (A2F10), c-kit (2B8), Sca1 (D7), CD150 (TC15-12F12.2), CD48 (HM48-1), CD201 (eBio1560), and ESAM (1G8; all from BioLegend or eBioscience). After staining, cells were washed and suspended with DAPI (1 µg/ml), and kept on ice. FACSAria II (BD) or MoFlo Astrios (Beckman Coulter) equipped with 590-nm lasers for optimal detection of the mCherry signal were used for cell sorting and analysis. Transplanted cells were double sorted for purity.
Transplantations and peripheral blood analysis.
Congenic recipient mice were lethally irradiated (900 rad), and donor cells, mixed with 2 × 105 whole BM competitor cells in 200 µl PBS 2 mM EDTA, and 2% FBS, were injected into the tail vein. At the indicated time points, 2–3 drops of blood were collected from the tail and added to 150 µl Alsever’s solution (Sigma-Aldrich). Blood samples were treated with 10 ml ACK solution (0.15 M NH4Cl, 1 mM KHCO3, and 0.1 mM EDTA) for 5 min at room temperature and washed two times with PBS. Leukocytes were stained with PerCP/Cy5.5-Ter119 (Ter119), PE/Cy7-Mac1 (m1/70), Fitc-Gr1 (8C5), PE-CD3 (17A2), APC/Cy7-B220 (RA3-6B2), A647-CD45.1 (A20), and PacBlue-CD45.2 (104); all from BioLegend. Cells were washed and suspended with propidium iodide (1 µg/ml) and analyzed on FACSCanto II (BD). Analysis was done using FlowJo software (Tree Star). For serial-transplants (secondary and tertiary), 2 × 106 cells whole BM cells were transferred into lethally irradiated congenic recipients.
AGM explants, culturing, and imaging.
The procedure we used was based on a previously published protocol (Medvinsky and Dzierzak, 1996; Taoudi et al., 2008). In brief, the aorta-gonad-mesonephros (AGM) of E11.5 embryos (CD45.2) were individually dissected (excluding somites) and cultured at the air-liquid interface on Durapore 0.65-µm filters (Millipore) suspended on IMDM media containing 20% serum and 100 ng/ml SCF, IL-3, and Flt3L (PeproTech). After 4 d, cells were dissociated with Collagenase I (Worthington), filtered, and mixed with whole BM competitor cells (CD45.1) and transplanted into lethally irradiated congenic recipients (CD45.1). For microscopic visualization of the AGM region, E11.5 embryos from Fgd5mCherry/+ x Fgd5mCherry/+ timed matings were fixed in 2% PFA for 30 min, equilibrated in 30% sucrose, embedded in OCT, and cryosectioned, followed by staining with FITC-conjugated Isolectin and DAPI. Images were taken using a Ti-E microscope (Nikon).
3D immunohistology of BM.
Thick femoral bone slices from Fgd5ZsGreen•ZsGreen/+ mice were generated as previously described (Nombela-Arrieta et al., 2013). BM slices were blocked overnight in 0.2% PBS, 1% Triton, and 10% BSA and stained with rabbit anti-laminin (Sigma-Aldrich) for 2 d in blocking solution, washed overnight in PBS, and stained for 2 d with DyLight549 donkey anti–rabbit IgG (Jackson ImmunoResearch Laboratories). Whole-mount stained slices were washed in PBS and incubated overnight in FocusClear (CelExplorer Laboratory). For observation under the confocal microscope, BM slices were mounted in glass slides embedded in FocusClear. Confocal microscopy was performed with a LSM700 system (Carl Zeiss). Image stacks were rendered into 3D volumes using Volocity Software (Improvision) and exported to Photoshop (Adobe Inc.) for processing.
Tamoxifen treatment of mice.
Tamoxifen was purchased from Sigma-Aldrich and suspended at 200 mg/ml in ethanol and mixed with sunflower oil to a final concentration of 10 mg/ml (10% ethanol). Tamoxifen was administered by intraperitoneal injection to 8–12-wk-old mice at 100 mg/kg body weight for 5 consecutive days. 2 d after tamoxifen treatment, the mice were sacrificed and analyzed for expression of dTomato in conjunction with immunostaining to reveal diverse hematopoietic populations.
All gene expression datasets used in this study are available at the Gene Expression Omnibus accession no. GSE56952.
Online supplemental material.
Fig. S1 shows relative expression of all 323 probe sets and corresponding genes identified as HSC-specific in our dataset. Video 1 shows 3D reconstruction of the microvascular network of femoral BM. Table S1 lists cell types and microarrays used in this study. Table S2 is a list of genes identified as cell type specific in our study.
We wish to thank Trista North, Christophe Bock, Isabel Beerman, Jane Segal, and Dvora Ghitza for expertise, input and assistance and Natasha Barteneva, Ken Ketman, John C. Tigges, and Vasilis Toxavidis for flow cytometry expertise.
R. Gazit was supported by a JCF Machiah post-doctoral fellowship. This work was supported by the National Institutes of Health grants RO1HL107630 (D.J. Rossi) and UO1DK072473-01 (D.J. Rossi), grants from GlaxoSmithKline (D.J. Rossi), The Leona M. and Harry B. Helmsley Charitable Trust (D.J. Rossi), the New York Stem Cell Foundation (D.J. Rossi), and the Harvard Stem Cell Institute (D.J. Rossi). D.J. Rossi is a New York Stem Cell Foundation Robertson Investigator.
The authors declare no competing financial interests.
R. Gazit and P.K. Mandal contributed equally to this paper.
R. Gazit’s present address is the Shraga Segal Dept. of Microbiology, Immunology, and Genetics, BGU Center for Regenerative Medicine & Stem Cells, National Institute for Biotechnology in the Negev, Ben-Gurion University in the Negev, Be’er-Sheva, Israel.