Notchless-dependent ribosome synthesis is required for the maintenance of adult hematopoietic stem cells

To address Nle function in the adult, we generated a conditional allele of Nle (Fig. S1, A–C) and performed acute conditional inactivation using the Rosa26^CreERT2 line (Hameyer et al., 2007). Adult Rosa26^CreERT2/+; Nle^Flox/null mice (named Nle^cKO hereafter) were injected three or five consecutive days with tamoxifen. Because activation of the CreERT2 recombinase could potentially affect cell physiology (Naiche and Papaioannou, 2007; Higashi et al., 2009), we used Rosa26^CreERT2/+; Nle^Flox/+ (named control hereafter) littermates exposed to the same regimen as control. In both control and Nle^cKO mice, efficient conversion of Nle^Flox into Nle^Del allele was observed in all organs except the brain (not depicted). Noticeably, all Nle^cKO mice died within 10–13 d after the first injection (Fig. 1 A). Examination of Nle^cKO mice euthanized on days 7–9 revealed alterations of intestinal gut epithelium histology (not depicted) and major defects in BM and secondary lymphoid organs (see below). This led us to analyze in more detail the role of Nle in hematopoiesis.

First, we monitored Nle expression in hematopoietic cells by Western blot and real-time RT-QPCR. NLE protein was detected in BM and spleen in control mice (Fig. 1 B). In the BM, lineage-negative (Lin⁻) immature cells expressed higher levels of protein than lineage-positive (Lin⁺) cells (Fig. 1 B). Lin⁻ cells from noninjected control were sorted into HSCs (Lin⁻ Sca-1⁺ c-Kit⁺ CD34⁻ [LSK CD34⁻]), MPPs (Lin⁻ Sca-1⁺ c-Kit⁺ CD34⁺ [LSK CD34⁺]), common myeloid progenitors (CMPs; Lin⁻ Sca-1⁻ c-Kit⁺ CD34⁺), and common lymphoid progenitors (CLPs; Lin⁻ Sca-1^low c-Kit^low Il7rα⁺; see the gating strategy in Fig. S2). Higher Nle mRNA levels were found in LSK CD34⁻, LSK CD34⁺, CMPs, and CLPs compared with CD19⁺ B cells or total BM containing a majority of Lin⁺ cells (Fig. 1 C). We then verified that in control and Nle^cKO mice, Nle deletion was effective in the hematopoietic tissue. Control and Nle^cKO mice were injected with tamoxifen, and PCR analysis was performed on BM and peripheral blood (PB) genomic DNA at various time points after Cre induction. Efficiency of recombination was estimated through comparison with samples of DNA from hematopoietic cells containing different ratios of Nle^Flox and Nle^Del alleles (Fig. S1, D and E). After Cre induction, high deletion efficiency was detected in BM, thymic, and splenic cells from control and Nle^cKO mice (Fig. 1 D and not depicted). This was further confirmed by the dramatic decrease in NLE protein levels in Nle^cKO BM and spleen (Fig. 1 B). At later time points (up to 9 mo), the vast majority of hematopoietic cells from control mice were carrying an Nle^Del allele (Fig. 1 E), indicating that Nle deletion mediated by the Rosa26^CreERT2 allele was effective in HSCs.

During the autopsy, we noticed a marked reduction in the size of hematopoietic and lymphoid organs from Nle^cKO mice. At day 9, the numbers of cells in thymus, spleen, and BM were strongly reduced (Fig. 1 F). In addition, hematoxylin and eosin staining of bone sections revealed severe hypocellularity in Nle^cKO BM, showing capillaries saturated with erythrocytes (Fig. 1 G). Flow cytometry of Nle^cKO BM revealed that most hematopoietic lineages were severely affected (Fig. 1 H). Together, these results show that Nle is critically required for hematopoiesis.

We next monitored the consequences of Nle inactivation on the BM immature cell populations and observed a strong diminution in the number of CMPs and CLPs 1 wk after tamoxifen injection (Fig. 2 A). Flow cytometry analysis also showed a rapid and dramatic decrease in the number of immature Lin⁻ Sca-1⁺ c-Kit⁺ (LSK) cells after Nle inactivation. Both HSC-enriched LSK CD34⁻ and MPP LSK CD34⁺ cell populations were virtually undetectable as early as 5 d after the first tamoxifen treatment (Fig. 2 B and not depicted). We noticed on the Nle^cKO FACS profiles a transitory Lin⁻ Sca-1⁺ c-Kit^low cell population (see the cells within the black circle in Fig. 2 D), suggesting that in the absence of Nle, HSCs and MPPs down-regulate c-Kit expression at their surface. A similar situation was previously reported in 5-FU–treated mice, in which HSC activation after abrupt loss of hematopoietic mature cells and precursors is associated with low c-Kit expression at their membranes (Randall and Weissman, 1997). We therefore decided to reanalyze HSCs in Nle^cKO BM irrespectively of c-Kit expression, using the SLAM family receptors CD48 and CD150 (Kiel et al., 2005). Although less pronounced than for LSK CD34⁻ cells, we observed a rapid decrease in the number of HSC-enriched Lin⁻ Sca-1⁺ CD150⁺ CD48⁻ cells (SLAM) in Nle^cKO BM compared with control (Fig. 2 C). Strikingly, although a vast majority of control SLAM cells expressed high levels of c-Kit receptor at their surface, the proportion of c-Kit^high cells among SLAM cells was drastically reduced in Nle^cKO BM 5 and 7 d after Cre induction (17% and 2%, respectively). Down-regulation of c-Kit expression was probably effective on other populations, including CMPs (see for example Fig. 2 D). Nonetheless, by itself, it did not account for the reduction of phenotypic CMPs and CLPs in Nle^cKO mice as shown by the decreased number of Lin⁻ CD34⁺ and Lin⁻ Il7rα⁺ cell populations that include CMPs and CLPs, respectively (Fig. 2 A). We also monitored Nle^cKO mice 5 d after Cre induction for the presence of Lin⁻ PB cells and did not find any signs of increased HSC mobilization (not depicted). Altogether, these data show that phenotypic HSCs and immature progenitors are rapidly lost after Nle deficiency.

To address the possibility that functional HSCs with a modified phenotype may nevertheless persist in Nle^cKO BM, we performed total BM transplantation experiments. BM cells were recovered from control and Nle^cKO mice 5 d after Cre induction and transferred into sublethally irradiated mice (Fig. 2 E). By the time of transplantation, efficient deletion of the Nle^Flox allele was observed in both control and Nle^cKO donor cells (Fig. 2 F, left). In contrast to control, the Nle^Del allele could neither be detected in the blood of mice transplanted with Nle^cKO BM 3.5 and 7 wk or 9 mo after engraftment nor in the BM after 9 mo (Fig. 2 F). This indicates that the reconstituting activity of Nle^cKO BM was limited to Nle-proficient nonrecombined cells. These data show that Nle function is critically required for HSC and immature progenitor activity and function.

To determine the mechanisms by which Nle deletion induces stem and progenitor cell loss, we analyzed Nle^cKO BM cells shortly after Cre induction. Because Nle has previously been shown to play a role in cell survival (Cormier et al., 2006; Orkin and Zon, 2008), we first examined whether increased cell death could be responsible for the rapid exhaustion of HSCs and MPPs in Nle^cKO mice. Using Annexin V/7AAD labeling, we analyzed freshly isolated Nle^cKO BM cells at days 3 and 5, when no or moderate defects in BM homeostasis were observed (Figs. 1 F and 2, A–D). At either time point, no increase in the frequency of Annexin V⁺7AAD⁻ (apoptotic) or Annexin V⁺7AAD⁺ (dead) cells among total BM cells, SLAM cells, and immature progenitors (LSK CD34⁺, CMPs, and CLPs at day 3 or Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ at day 5) was detected (Fig. 3, A and B; and not depicted). At day 5, a significant decrease in the frequency of apoptosis among SLAM cells was observed in Nle^cKO mice (Fig. 3 A), although no HSC activity was found in the BM of such animals (Fig. 2 F). Similarly, a TUNEL assay performed on control and Nle^cKO mice at day 4 did not show an increased cell death of HSCs and immature progenitors after Nle inactivation (not depicted). We then examined the cell cycle status of control and Nle^cKO BM cells using Ki67 and Hoechst staining. Importantly, as early as 3 d after Cre induction, there was a significant decrease in the frequency of G₀ cells between control and Nle^cKO SLAM or LSK CD34⁻ and MPPs (Fig. 3, C [left] and D; and not depicted). In all populations, Ki67⁺ cells were identically distributed among G1 and S-G2/M between control and Nle^cKO mice (Fig. 3 C, right; and not depicted), indicating that cycling Nle-deficient BM cells have cycling parameters similar to control cells. We next examined SLAM cells for various cell cycle regulators by RT-QPCR (Fig. 3 E). As expected, a massive reduction of Nle mRNA levels was observed in Nle^cKO SLAM cells. For most genes tested, no differences were observed between control and Nle^cKO SLAM cells. Consistent with the higher proportion of cycling cells, cyclin A2, cyclin B1, and cyclin D1 mRNA levels were slightly increased in Nle-deficient SLAM cells. Strikingly, a strong up-regulation of Cdkn1a (p21) mRNA level was observed in Nle^cKO SLAM cells, suggesting that depletion of NLE protein in HSCs causes the activation of p53 (see below). However, this up-regulation was not accompanied by a potent cell cycle arrest, as visualized by FACS analysis (Fig. 3, C and D). A similar situation has been previously reported during HSC mobilization (Passegué et al., 2005). Up-regulation of Cdkn1a (p21) together with increased cell cycle entry of HSCs has also been described after x-ray exposure (Insinga et al., 2013). Altogether, our data indicate that Nle is required for the maintenance of HSC quiescence.

In Nle^cKO mice, systemic Nle inactivation might contribute to BM failure. To examine this possibility, we performed noncompetitive transplantation between wild-type and control or Nle^cKO mice. In a first set of experiments, we transplanted BM cells from wild-type mice into sublethally irradiated untreated control or Nle^cKO mice (Fig. 4 A). In the reciprocal experiment, untreated control or Nle^cKO BM cells were transplanted into sublethally irradiated wild-type mice (Fig. 4 B). 2 mo after grafting, efficiently reconstituted (>80%) mice were injected with tamoxifen and their BM were analyzed after 7 d, at a time point where a drastic BM phenotype was observed in Nle^cKO mice (Fig. 1, F and H; and Fig. 2, A–C). The number of wild-type LSK cells was not modified in an Nle-deficient environment compared with control (Fig. 4 C). In contrast, wild-type mice reconstituted with Nle^cKO BM exhibited a reduced number of LSK cells after Nle inactivation (Fig. 4 D). At longer time points, we observed the progressive disappearance of the Nle^Del allele in BM and PB (not depicted). This shows that Nle-deficient HSCs are not able to maintain hematopoiesis in a wild-type environment and supports the notion that loss of HSCs and progenitors in Nle^cKO mice was caused by Nle loss of function in BM cells. Death of Nle^cKO mice reconstituted with wild-type BM 2 wk after tamoxifen injection indicates that Nle function is also required in nonhematopoietic cells for mouse survival (not depicted).

To address whether Nle acts in a cell-autonomous manner in hematopoietic cells under homeostatic conditions, we adopted a competitive chimera approach. Either untreated control or Nle^cKO CD45.2 donor BM cells were transplanted with an excess (1:3 to 1:4 ratio) of competitor wild-type CD45.1 BM cells into irradiated (CD45.1×CD45.2)F1 recipient mice (Fig. 4 E). In such chimeras, hematopoiesis can be sustained by competitor cells, therefore preventing secondary effects caused by the cytopenia that occurs both in Nle^cKO mice and noncompetitive chimeras. After stable reconstitution, mice were injected with tamoxifen, and PB and BM cells were analyzed at various time points. No differences in BM cellularity were observed between mice engrafted with Nle^cKO and control cells, confirming that BM homeostasis was maintained in these chimeras (not depicted). We also verified that efficient deletion of Nle^Flox allele was obtained in these mice (Fig. 4 F). In PB, Nle^cKO CD45.2 Mac1⁺ myeloid cells were no longer detected from 20 d after injection onwards (Fig. 4 G). Long-lived CD3⁺ T and CD19⁺ B lymphoid cells were also affected, although with a delayed kinetics (not depicted). In agreement with the data obtained in the PB, Nle^cKO CD45.2 cells rapidly disappeared in the BM (Fig. 4 H, left). The evolution of CD45.2 cells was addressed at day 7 and at later time points, as soon as 3 wk after Nle inactivation, and contrary to control CD45.2 cells, virtually no Nle^cKO CD45.2 cells were detected in BM and within the SLAM compartment (Fig. 4 H), indicating that Nle-deficient HSCs were not maintained even in a wild-type microenvironment and during homeostasis. Under these conditions, the proportion of c-Kit^high cells among Nle^cKO CD45.2 SLAM cells was still reduced (Fig. 4 I), indicating that Nle deficiency in HSCs can intrinsically down-regulate c-Kit expression. Altogether, our data show that Nle is cell-autonomously required for the maintenance of functional HSCs under both homeostasis and cytopenic conditions.

Because hematopoietic cells are continuously and actively fueled by the HSC/MPP compartment, the rapid exhaustion of HSCs and immature progenitors in Nle^cKO mice could preclude the detection of a more direct role of Nle in restricted progenitors and differentiated cells. To examine this possibility, we first used an ex vivo differentiation assay. Untreated control and Nle^cKO HSC-enriched (Lin⁻ Sca-1⁺ c-Kit⁺ Flt3⁻) cells were FACS-sorted and plated on OP9 stromal cells. After 5 d of culture, immature LSK cells were no longer detected (not depicted), and a pulse of 4-hydroxy-tamoxifen (4-OHT) was added to the culture medium to induce loss of Nle function in progenitor cells. During the following days, both control and Nle-deficient cells underwent massive expansion (Fig. 5 A). Cells were harvested after a total of 12 d in culture and analyzed for differentiation markers. Under the conditions used, differentiation was biased toward myeloid lineage, but control and Nle^cKO cells were indistinguishable in their ability to give rise to Mac1⁺Gr1⁺ granulocytes (Fig. 5 B, right). Importantly, we checked that Nle was efficiently inactivated at days 6 and 12 by PCR analysis (Fig. 5 B, left). This suggests that myeloid progenitors were able to actively proliferate and differentiate in the absence of Nle.

We also evaluated the role of Nle in vivo during B cell development using the CD19-Cre mouse line (Rickert et al., 1997). We generated CD19^Cre/+; Nle^Flox/+ and CD19^Cre/+; Nle^Flox/null mice, named, respectively, CD19-control and CD19-Nle^KO mice. As already observed for other Flox alleles, CD19^Cre-mediated Nle inactivation was an on-going process along B cell maturation (Fig. 5 C). Comparison of CD19-control and CD19-Nle^KO mice revealed no significant differences in the deletion efficiency of Nle^Flox allele and in the numbers of BM pre-B cells, immature B cells, and splenic mature B cells (Fig. 5, C–E). Moreover, CD19-Nle^KO B cells were able to secrete steric IgM (not depicted), indicating that Nle-deficient mature B cells were not functionally impaired. Altogether, these data strongly suggest that contrary to HSCs and MPPs, Nle is dispensable for survival, proliferation, and maturation of more restricted progenitors.

In yeast, NLE orthologue Rsa4 acts as a pre-60S maturation factor, being essential for progression of ribosomal large subunit biogenesis and nuclear export (de la Cruz et al., 2005; Ulbrich et al., 2009; Bassler et al., 2010). Based on its high degree of conservation among eukaryotes (Cormier et al., 2006; Gazave et al., 2009), we hypothesized that NLE might also participate in pre-60S maturation in mouse cells. To address this possibility, we took advantage of Rosa26^CreERT2/+; Nle^Flox/null embryonic stem (ES) cells, in which Nle loss of function can be induced by addition of 4-OHT in the culture medium (Fig. 6 A). In ES cells, NLE localized to the nucleolus as indicated by the specific nucleolar immunostaining that partially colocalized with B23/nucleophosmin and was undetectable after 4-OHT treatment (Fig. 6 B). To investigate the involvement of Nle in ribosome biogenesis, total cellular extracts from treated and untreated ES cells were fractionated on a sucrose gradient. Immunoblot of untreated ES cells fractions indicated that NLE accumulated mainly in fractions 8–11, corresponding to 60S to 80S sedimentation coefficients (Fig. 6 C). In treated cells, the sedimentation profile was dramatically altered with a decrease in the amount of 60S subunit and 80S ribosomes and an increase of 40S subunit (Fig. 6 D). These observations are consistent with a role of NLE in pre-60S maturation.

In yeast, Rsa4 deficiency has been shown to interfere with pre-60S nuclear export. We thus expected to observe a similar phenotype leading to accumulation of 5.8S and 28S rRNA precursors and decrease of the corresponding mature rRNA species in Nle-deficient ES cells (Fig. 7 A). We monitored the levels of different rRNA species in treated and untreated ES cells by Northern blot analysis (Fig. 7, A–D; and not depicted). Total RNA recovered from the same number of treated and untreated cells was probed with sequences specific for 5.8S, 18S, 28S, and its2. We observed a progressive reduction in the steady-state levels of 5.8S and 28S mature rRNA concomitantly with the accumulation of 12S and 32S rRNA precursors in mutant cells (Fig. 7, B–D). In contrast, processing of 18S rRNAs was not affected (Fig. 7 B). We confirmed these observations by RNA fluorescence in situ hybridization (FISH) using 5′ets, its1, and its2 probes recognizing various combinations of nucleolar rRNA species (Fig. 7, A and E). By flow cytometry, we measured an increase of the its2/its1 ratio, indicating a specific accumulation of rRNA precursors of the large subunit (Fig. 7 F). 45/47S levels measured by 5′ets signal intensity were unchanged (not depicted). Importantly, its2 levels were not affected after 4-OHT treatment of control ES cells (Fig. 7 G), demonstrating that the accumulation of immature rRNA precursors is indeed caused by Nle loss of function. Altogether, these data demonstrate that Nle is required in mouse ES cells for the maturation and the export of the large ribosomal subunit.

Having established the role of Nle in ES cells, we next wondered whether defects in ribosome biogenesis could be responsible for the BM phenotype observed in Nle^cKO mice. We combined cell surface antibody staining with intracellular RNA FISH and used flow cytometry to measured steady-state levels of rRNA precursors in several BM cell populations. We first checked that as in ES cells, RNA FISH with its2 probes labeled subnuclear structures, poor in chromatin, likely to correspond to nucleoli in immature cells and B cells (Fig. 8 A). 3 d after Cre induction, we measured a shift in its2 intensity in Nle-deficient SLAM and Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ MPPs but not in more committed Lin⁺ cells (Fig. 8, B and C, left). Importantly, neither its1 nor 5′ets levels were modified in stem cells and MPPs, pointing to defective large ribosomal subunit maturation in HSCs/MPPs (Fig. 8, B and C; and not depicted). Accumulation of immature rRNAs of the large subunit in SLAM and Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ cells was confirmed by RT-QPCR (Fig. 8 C, right). 5 d after Cre induction, a similar situation was observed (Fig. 8 B, right). We performed similar analysis on B cell lineage populations from CD19-control and CD19-Nle^KO mice. Consistent with the lack of a phenotype in CD19-Nle^KO mice, no changes in the level of its2, its1, or 5′ets were observed in BM and spleen B cells (Fig. 8, D and E; and not depicted). We thus conclude from these experiments that Nle is differentially required for 60S ribosomal subunit biogenesis between HSCs/MPPs and more committed or differentiated hematopoietic cells.

Defective ribosome biogenesis was shown to trigger the activation of p53 pathway in vertebrate cells (Zhang and Lu, 2009). We therefore monitored p53 expression levels in BM cells shortly after Nle inactivation. 4 d after Cre induction, we measured a clear increased in p53 levels in both Nle-deficient SLAM and Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ cells (Fig. 9, A [left] and B). In contrast, in Nle^cKO Lin⁺ cells and CD19-Nle^KO B cells, in which no increases in pre-60S rRNA intermediates were observed, p53 levels were not augmented (Fig. 9, A [right] and C). We next assessed the transcriptional activation of various p53 target genes in these compartments. In addition to Cdkn1a (p21; Fig. 3 E), mRNA levels of Bax, Noxa, Puma, and Sestrin2 were significantly increased in SLAM or Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ cells (Fig. 9 D), further confirming the activation of p53 after Nle inactivation. To determine the functional implication of p53 in the severe hematopoietic phenotype of Nle^cKO mice, we crossed p53^−/− mice with Nle^cKO mice and analyzed BM cell populations after tamoxifen injection. Strikingly, loss of p53 totally rescued the number of total BM and LSK cells in Nle^cKO mice 5 d after Cre induction (Fig. 9 E). Accordingly, the number of SLAM cells in Nle^cKO; p53^−/− mice was unchanged compared with control mice (not depicted). Importantly, Nle^cKO; p53^−/− SLAM and Lin⁻ Sca-1⁺ CD150⁺ CD48⁺ cells exhibited increased its2 levels compared with controls (Fig. 9 F). The fact that its2 levels were significantly higher in these mice compared with Nle^cKO mice suggests that the p53-dependent checkpoint operates at a relatively early step of pre-60S rRNA intermediate accumulation in these cell populations. Collectively, our data suggest that Nle deficiency leads to defective ribosome biogenesis and consecutive p53 activation in HSCs and MPPs that in turn causes their disappearance.

In this study, we show that Nle is critically required for the maintenance of adult HSCs. Collectively, our data demonstrate that Nle is an intrinsic regulator, acting in a cell-autonomous manner, and necessary for HSC self-renewal. By demonstrating that Nle function in the pre-60S ribosomal subunit maturation has been conserved during evolution, we underscore the importance of ribosome biogenesis as a critical pathway for HSC maintenance. In contrast, Nle function is dispensable in restricted progenitors and differentiated hematopoietic cells, indicating that ribosome biogenesis is differentially regulated during differentiation.

Strikingly, Nle-deficient cells were rapidly eliminated from the HSC population. Although several factors have been shown to participate in HSC regulation, rapid loss of the HSC pool has been reported after the disruption of only a few genes: N-Myc/c-Myc, NKAP, Cited2, Mll, Survivin, Mcl-1, or APC (Opferman et al., 2005; Jude et al., 2007; Leung et al., 2007; Laurenti et al., 2008; Qian et al., 2008; Kranc et al., 2009; Pajerowski et al., 2010). Of note, one of the primordial functions of Myc is the regulation of ribosome biogenesis (van Riggelen et al., 2010; Ji et al., 2011). Combined deficiency of c-Myc and N-Myc has been shown to severely impair HSC function, and interestingly, down-regulation of 31 ribosomal proteins coding genes has been reported in mutant HSCs (Laurenti et al., 2008). However, contribution of defective ribosome biogenesis to the phenotype of c-Myc/N-Myc double-deficient HSCs has not been investigated.

Concomitantly with HSC loss, we observed the down-regulation of c-Kit expression at the cell surface of Nle^cKO immature hematopoietic cells. c-Kit is a major regulator of HSC survival and proliferation, and its down-regulation may therefore participate in defective maintenance of Nle^cKO HSCs. In Gfi1^−/− mice, a slight decrease of c-Kit in LSK cells correlates with progressive loss of HSC self-renewal activity (Hock et al., 2004). Reduced c-Kit expression might also reflect HSC activation consecutive to BM depletion (Randall and Weissman, 1997; Laurenti et al., 2008). Nle^cKO HSCs displayed reduced expression of c-Kit receptor under homeostatic conditions, indicating that such deregulation is not solely the consequence of BM cytopenia. How c-Kit expression is regulated in HSCs is not fully understood. Recently, the Shp2 tyrosine phosphatase was shown to promote Kit gene expression and HSC pool maintenance through a positive c-Kit/Shp2/c-Kit feedback loop (Zhu et al., 2011). 4 d after Cre induction, a ≈30% decrease in c-Kit mRNA levels was observed in Nle^cKO HSCs/MPPs (unpublished data), suggesting that c-Kit down-regulation at their membranes was controlled at least in part at the transcriptional level. Future work will be necessary to understand the mechanisms underlying c-Kit down-regulation in Nle-deficient HSCs.

Based on genetic study in Drosophila, Nle has been proposed to code for a modulator of Notch signaling. Absence of canonical Notch signaling does not impact HSC maintenance at homeostasis (Maillard et al., 2008). Conversely, sustained Notch activity promotes HSC expansion ex vivo and enhanced self-renewal of hematopoietic progenitors (Bigas and Espinosa, 2012). Thus, modulation of Notch signaling cannot simply account for the failure of Nle-deficient HSCs/MPPs. Contrary to genes coding for key components of the Notch pathway that are specific to metazoans, Nle is an ancient eukaryotic gene (Gazave et al., 2009), and its ancestral function as a ribosome biogenesis factor has been deciphered in yeast (de la Cruz et al., 2005; Ulbrich et al., 2009). Here, we show that Nle function has been conserved during evolution. Although we cannot formally exclude the possibility that NLE has acquired additional nonribosomal functions during evolution, it seems more likely that Notch activity modulation by NLE is related to its role in ribosome biogenesis. Genetic interactions between Notch and Minute mutations, which affect a class of ribosomal genes, have been reported a very long time ago (Schultz, 1929). More recently, genetic screens for modifiers of Notch activity have identified several ribosomal proteins (Hall et al., 2004; Mourikis et al., 2010). In particular genetic and cellular contexts, Notch signaling activity must therefore be sensitive to variations in ribosome supply as previously discussed for the Hedgehog signaling pathway (Nybakken et al., 2005).

Using a cellular system, we showed that Nle loss of function induced defective pre-60S subunit maturation and export, leading to the rapid diminution of free ribosomes. In vivo, by combining rRNA FISH and flow cytometry, we were able to detect alterations in ribosome biogenesis in HSCs and immature progenitors at early time points after Nle inactivation. This suggests that these defects cause HSC/MPP exhaustion in Nle^cKO mice. Defective ribosome biogenesis might cause HSC elimination through several mechanisms. A shortage of ribosomes would affect translation at a global level, although IRES-mediated translation may be more sensitive to ribosome insufficiency than a cap-dependent one (Horos et al., 2012). Preferential translation of mRNAs with a 5′ terminal oligopyrimidine tract has also been reported upon defective ribosome biogenesis (Fumagalli et al., 2009). In addition, suppressed translation of a specific subset of mRNAs has been observed upon ribosomal protein haploinsufficiency/knockdown (Kondrashov et al., 2011; Horos et al., 2012). Therefore, altered translation of transcripts essential for HSC function might contribute to the phenotype of Nle^cKO mice. Impaired ribosome biogenesis is known to efficiently trigger the up-regulation of p53 function through binding of MDM2 by free ribosomal proteins RPL11, RPL23, RPL5, or RPS7 (Zhang and Lu, 2009). Consistently, we demonstrated the rapid activation of the p53 pathway in Nle-deficient HSCs/MPPs. Up-regulation of the proapoptotic genes Bax, Puma, and Noxa in these compartments suggests that apoptosis may actively contribute to their elimination. However Annexin V and TUNEL stainings failed to unravel increased cell death in Nle-deficient HSCs/MPPs. The reasons for this discrepancy are yet unclear but may be caused by the rapid clearance of dying mutant cells in the BM or the fact that these dying cells lose some of their surface markers and therefore were not included in our analyses. As a complementary approach, we performed deletion of Nle in vitro in sorted HSC/MPP populations. Unfortunately, in vitro Cre activation in these populations compromises their maintenance and disabled us from using this assay to assess apoptosis in these cells. Other cellular responses to p53 activation may also participate in the exhaustion of Nle-deficient HSCs/MPPs. Hence, it has been suggested that p53 negatively regulates self-renewal of HSCs and MPPs, although how p53 regulates this process is not precisely understood (Akala et al., 2008; Chen et al., 2008). Induction of differentiation was also shown to represent an effective response to stress in various stem cell populations (Inomata et al., 2009; Li et al., 2012; Wang et al., 2012). Thus, it will be interesting to determine whether premature differentiation could be induced by defective ribosome biogenesis in Nle-deficient HSCs, thereby contributing to their exhaustion. Regardless the underlying molecular mechanisms, disappearance of Nle-deficient HSCs/MPPs is genetically dependent on p53. In the absence of p53, Nle^cKO HSCs and immature progenitor populations were unaffected, although accumulation of pre-60S rRNA intermediates was still observed. Further analyses at later time points will be required to determine whether and how impaired ribosome biogenesis consecutive to NLE depletion affects HSC functionality independently of the p53-dependent ribosome biogenesis checkpoint.

Considering its conserved and critical role in large ribosomal subunit biogenesis, our observation that NLE is dispensable in restricted progenitors was unexpected. Importantly, we did not detect accumulation of rRNA intermediates in Nle-deficient B and Lin⁺ cells, suggesting that pre-60S maturation occurs normally in the absence of NLE in these cells. Notably, normal proliferation, differentiation, and maturation of B cell lineage after constitutive inactivation of Nle with the CD19-Cre driver strengthens this conclusion and argues against the hypothesis that long-lasting inherited ribosomes could compensate for a defective ribosome biogenesis. Collectively, our findings support the hypothesis that alternative Nle-independent pathways can assure 60S subunit production in restricted progenitors but not in HSCs/MPPs (Fig. 10). An NLE-related WD-repeat protein, WDR12/Ytm1, containing an N-terminal domain homologous to the NLE N-terminal domain is present in all eukaryotes and also participates in pre-60S maturation (Nal et al., 2002; Bassler et al., 2010). However, WDR12/Ytm1 and NLE/Rsa4 have been shown to act sequentially and in a nonredundant manner during pre-60S subunit maturation in yeast (Ulbrich et al., 2009; Bassler et al., 2010). Ribosome biogenesis has been extensively studied in yeast, but is yet poorly described in higher eukaryotes. Several differences have been reported between yeast and higher eukaryotic cells in terms of rRNA maturation steps or composition of preribosomal particles (see for example, Turner et al., 2009; Srivastava et al., 2010). Whether ribosome biogenesis is qualitatively modulated during development or between cell types in a pluricellular organism is still an open question. This study is to our knowledge the first in vivo demonstration that differences in ribosome biogenesis exist between cell types in mammals. Recently, the ribosomal filter hypothesis postulating that differences in composition of ribosomal subunit or posttranslational modifications of ribosomal proteins may differentially affect the translation of particular transcripts has received support from studies performed in budding yeast and mouse embryo (Komili et al., 2007; Kondrashov et al., 2011). If this is true, ribosome heterogeneity implies some degrees of plasticity during preribosomal subunit assembly. Differential activities of alternative pathways for 60S subunit maturation between HSCs/MPPs and restricted progenitors could thus lead to the generation of qualitatively distinct ribosomes and therefore would represent an additional mode of gene expression regulation.

The expression of several genes involved in ribosome biogenesis is highly enriched in Drosophila ovarian stem cells (Kai et al., 2005) and fish neural stem cells (Joly, J.-S., personal communication). A genome-wide gain of function screen in mouse ES cells identified multifunctional nucleolar proteins involved notably in ribosome biogenesis as novel regulators of self-renewal (Abujarour et al., 2010). Similarly, genetic screens for mutants that affect neuroblast and ovarian germline stem cell maintenance in Drosophila identified genes involved in ribosome biogenesis (Fichelson et al., 2009; Neumüller et al., 2011). Together with our findings, these observations link ribosome biogenesis with stem cell biology. Beyond overall protein synthesis and cell growth, specificities in preribosomal factors or ribosomal protein expression may contribute to stem cell/immature progenitor identity. In the future, it will be important to address whether such specificities and in particular dependency on the NLE pre-60S maturation pathway contribute to self-renewal capacity of malignant cells.

Experiments on live mice were conducted according to the French and European regulations on care and protection of laboratory animals (EC Directive 86/609, French Law 2001-486 issued on June 6, 2001) and the National Institutes of Health Animal Welfare (Insurance #A5476-01 issued on 02/07/2007). Control and Nle^cKO animals were obtained by crossing mice harboring Nle^Flox (Fig. S1), Nle^null (Cormier et al., 2006), Rosa26^CreERT2 (Hameyer et al., 2007), CD19^Cre (Rickert et al., 1997), or p53⁻ (Jacks et al., 1994) alleles. Mice were injected intraperitoneally with 75 mg/kg tamoxifen for three or five consecutive days. Nle^cKO mice were systematically compared with control littermates treated under the same conditions.

ES cells were cultured in DMEM + GlutaMAX-I (Gibco) supplemented with 15% FCS, 100 µM β-mercaptoethanol, 100 U/ml penicillin, 0.1 mg/ml streptomycin, and 10³ U/ml LIF. Nle^Flox/+ ES cells were obtained by homologous recombination in CK35 129/Sv ES cells (Kress et al., 1998; for details see Fig. S1). Control and Nle^cKO ES cells were obtained by iterative gene targeting. Nle^Flox/+ ES cells were electroporated with the Nle^nLacZ targeting vector (Cormier et al., 2006) to generate Nle^Flox/null ES clones. Then, SfiI-linearized pRosaCreERT2 targeting vector (Hameyer et al., 2007) was electroporated into Nle^Flox/+ and Nle^Flox/null ES cells to generate control (Rosa26^CreERT2/+; Nle^Flox/+) and Nle^cKO (Rosa26^CreERT2/+; Nle^Flox/null) ES cells, respectively. The deletion of the Nle^Flox allele in culture was induced by treatment with 1 µM 4-OHT for two consecutive days.

We raised an anti-NLE polyclonal antibody in rabbit. Recombinant fusion peptide corresponding to amino acids 1–174 of the NLE protein was produced in Escherichia coli as a His-tagged protein from pET15b expression vector (EMD Millipore) and purified using immobilized metal ion affinity chromatography. Purified peptide was then used as an immunogen for the production of rabbit antiserum (Agro-Bio). Antibody was used at 1:50 and 1:500 dilutions for immunofluorescence and immunoblotting, respectively. Antibodies used for flow cytometry are listed in Table S1.

Single cell suspensions of BM, spleen, and PB were stained with different combinations of antibodies. Marker combinations and FACS gating strategy used to define the various BM cell populations are indicated in Fig. S2. Apoptosis was monitored using 7AAD and Annexin V (Roche) according to the manufacturer’s instructions. For proliferation analysis and p53 labeling, cells were incubated with fixation/permeabilization solution (eBioscience) and stained either with anti-Ki67 (B56; BD) antibody and 10 µg/ml Hoechst 33342 or with p53 antibody (CM5; Novocastra) and Alexa Fluor 488–conjugated anti–rabbit secondary antibody, respectively. CyAn ADP, FACSCanto, and LSR Fortessa (BD) flow cytometers were used for analysis. Sorting was performed using the MoFlo cytometer (Beckman Coulter) or the AutoMACS technology (Miltenyi Biotec). Data were analyzed with FlowJo software (Tree Star).

HSCs were cultured on an OP9 monolayer in Opti-MEM, 10% FCS, 100 µM β-mercaptoethanol, 100 U/ml penicillin, 0.1 mg/ml streptomycin, Il-7, Flt3 ligand, and GM-CSF (supernatants of cell lines transfected with expression vectors for IL-7, Flt3 ligand, and GM-CSF, respectively; a gift from F. Melchers, Max Planck Institute for Infection Biology, Berlin, Germany).

For noncompetitive transplantations, ∼8 × 10⁵ Lin⁻ cells from control or Nle^cKO mice (129Sv/C57BL/6 mixed background) were injected into the retroorbital sinus of sublethally irradiated (11 Grays) (129Sv×C57Bl6)F1 mice carrying the Act-GFP transgene (Okabe et al., 1997). BM was depleted from lineage-positive cells to avoid graft versus host reaction. The converse transplantation was performed similarly using donor and recipient mice selected for histocompatibility matching through genotyping of the chromosome 17 region encompassing major MHC class I loci. For competitive transplantations, Lin⁻ control or Nle^cKO cells (CD45.2, 129Sv/C57Bl6 mix) were mixed with an excess (3:1 or 4:1 ratio) of competitor cells (CD45.1, C57BL/6) and grafted into a sublethally irradiated recipient (CD45.1×CD45.2, (129Sv×C57Bl6)F1). 2 mo after grafting, the reconstitution rate was assayed by flow cytometry analysis of PB cells using either GFP expression or CD45.1/45.2 immunostaining (noncompetitive and competitive transplantations, respectively).

Total RNA from ES cells was prepared with TRIzol reagent (Sigma-Aldrich) according to the supplier’s instructions. Migration and hybridization were performed as previously described (Fichelson et al., 2009). The probes used in this study were 5.8S (5′-GCGTTCGAAGTGTCGATGATCAATGTGTCCTGCAATTCAC-3′), 18S (5′-ACGGTATCTGATCGTCTTCGAACC-3′), 28S (5′-GTTCCCTTGGCTGTGGTTTCGCTAGATA-3′), and its2 (5′-ACTGGTGAGGCAGCGGTCCGGGAGGCGCCGACG-3′).

Cells were treated for 10 min with 10 mg/ml cycloheximide (Sigma-Aldrich) in cell culture medium, collected, and then resuspended in buffer lysis (50 mM Tris-Hcl, pH 7.4, 50 mM Kcl, 10 mM MgAc, 1 mM DTT [Invitrogen], 200 U/ml RNasin [Promega], 1 tablet/10 ml proteases inhibitors [Roche], and 10 mg/ml cycloheximide [Sigma-Aldrich]). 200-µl glass beads were added to cell suspension, and lysis was performed by vortexing cells at 4°C, for 1 min, five times. After centrifugation, supernatants were layered on a 15–50% sucrose gradient and centrifuged for 15 h at 33,000 rpm in an SW41 rotor (Beckman Coulter). The gradient was fractionated with a fractions collector system coupled to a UV detector to monitor absorbance at 260 nm. Half of each fraction was treated with trichloroacetic acid to recover proteins, whereas the remaining half was treated with phenol-chloroform to extract RNA.

ES cells or stained hematopoietic cells were fixed for 30 min with 4% PFA, rinsed in PBS, and placed in 70% ethanol at 4°C. The hybridization was performed as previously described (O’Donohue et al., 2010). Cy3 conjugated probes were purchased at Eurogentec (Its1, 5′-TAGACACGGAAGAGCCGGACGGGAAAGA-3′; its2, 5′-GCGATTGATCGTCAACCGACGCTC-3′; and 5′ets, 5′-AGAGAAAAGAGCGGAGGTTCGGGACTCCAA-3′).

Genotyping was performed by PCR after lysis of tissue and ES cell pellet in the following buffer: 50 mM Tris, pH 8.5, 100 mM NaCl, 0.5% Tween 20, and 100 mg/ml proteinase K at 56°C overnight, followed by a 10-min incubation at 96°C. For Nle, PCR amplification with primers a (5′-CAGACTTTCTTATTGGGGTTTGGT-3′), b (5′-TAGCCCTGGCTGACCTAGAA-3′), b′ (5′-TCCCCTCGAGGGACCTAATA-3′), and c (5′-ATGCCCGGCGTGTTAAGATT-3′) allowed us to discriminate Nle^wt (538 bp, primers a and b), Nle^Flox (439 bp, primers a and b′), and Nle^Del (773 bp, primers a and c) alleles. D17mit24 and D17Mit135 polymorphic markers were used to discriminate between 129Sv and C57BL/6 alleles around MCH class I loci.

For all bar graphs with pooled data, means (SD) are shown. Statistical analyses were performed using the Mann–Whitney Wilcoxon test. Significance is indicated on the figures with the following convention: *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

mRNA from sorted cells was purified using either the Dynabeads mRNA direct kit (Invitrogen) or RNeasy Plus Micro kit (QIAGEN) and reverse-transcribed using the SuperScript Vilo cDNA synthesis kit (Invitrogen), according to the manufacturers’ instructions. Real-time PCR was then performed using custom-designed primers (compare in Table S2) and SYBR green PCR master mix on a StepOne instrument (StepOne software version 2.2; Applied Biosystems) or, alternatively, with the Fluidigm Biomark EvaGreen qPCR system (Fluidigm Real-Time PCR analysis software version 3.1.3; Fluidigm). Expression levels were normalized using several housekeeping genes, including TBP, Tubulin β 5, Aldolase, TfIID, and Rrm.

Fig. S1 describes the generation of the Nle^Flox allele by homologous recombination in ES cells. Fig. S2 shows hierarchical relations of various hematopoietic cell populations defined by marker combinations and illustrates the FACS-gating strategy used for analysis. Table S1 lists the antibodies used for flow cytometry. Table S2 lists the sequence primers used for real-time RT-PCR.

Imaging and cell sorting analyses were performed at the Institut Pasteur Imagopole. We are grateful to P.-H. Commere for his assistance with FACS analysis and sorting. We thank J. Jonker and A. Berns for the Rosa26CreERT2 targeting vector and mouse line, K. Rajewsky for the CD19-Cre mouse line, and L. Guillemot, M. Fromont-Racine, S. Duncker, M.-F. O’Donohue, P.-E. Gleizes, C. Moch, J.-R. Huyn, A. Mougin, F. Morlé, S. Tajbakhsh, and all members of the laboratory for technical advice and helpful discussions.

This work was supported by the Centre National de la Recherche Scientifique, the Institut Pasteur, the Agence Nationale de la Recherche (contract no. JC05_41835 and ANR-10-LABX-73-01 REVIVE), and the Institut National du Cancer (INCa 2007-1-COL-6-IC-1 and PLBIO09-070). M. Le Bouteiller was supported by the Université Pierre et Marie Curie and received fellowship from the Ministère de l’Education Nationale, de la Recherche et de la Technologie, and the Ligue Nationale Contre le Cancer.

The authors have no conflicting financial interests.