Positive and negative regulation of granulopoiesis by endogenous RARα

The retinoic acid receptor α (RARα) gene is altered by chromosomal translocations in all cases of acute promyelocytic leukemia (APL). These translocations lead to the synthesis of chimeric proteins, resulting from the fusion of novel N-terminal sequences to RARα amino acids 60-462.1-3 To date, 5 distinct fusion partners for RARα have been identified, all of which belong to distinct protein families. PML, the most frequent partner, is a nuclear protein containing a RING finger and a coiled coil. PML has been implicated in regulating the formation and function of nuclear bodies, nuclear substructures whose function is not completely understood.4 Less frequent partners include PLZF (a zinc finger transcription factor),5 NuMA (a mitotic apparatus protein),6 nucleophosmin,7 and STAT5b.8 Forced expression of either PML/RARα or PLZF/RARα induces an APL-like disease in transgenic mice and establishes the causal role of these fusion proteins in APL leukemogenesis.9-11 At the molecular level, both PML/RARα and PLZF/RARα are strong repressors, efficiently recruiting a complex containing the nuclear receptor corepressors N-CoR/SMRT and histone deacetylase to target promoters.9,12-15 Importantly, these repressive properties appear to be closely linked with their leukemogenic potential, as histone deacetylase inhibitors can reverse the leukemic phenotype and lead to remission in transgenic mouse models of APL.9 

Despite accumulating data implicating the fusion proteins in leukemia, relatively little is known about how their action relates to a normal function of RARα during granulopoiesis, the process by which immature myeloid cells, such as promyelocytes, differentiate to mature neutrophils. In the hematopoietic system, RARα is expressed at significant levels in granulocytes, indicating a possible function within this lineage.16-18 RA has been shown to act as a differentiating agent for both normal and transformed immature myeloid cells,19,20 suggesting that the fusion proteins might antagonize normal RA-dependent differentiation. Conversely, overexpression of wild type (WT) RARα (as well as dominant negative forms of this receptor) has also been shown to block promyelocyte differentiation in culture.21-23 Thus, at limiting RA concentrations, an increase in the proportion of unliganded RAR molecules may lead to a dominant-negative inhibition of RA-controlled differentiation. Lastly, RA appears to orient the differentiation of pluripotent hematopoietic progenitors toward the granulocyte lineage, indicating that RA may be involved in granulocyte fate specification.24 These various studies, however, which employ either overexpressed receptors or exogenous ligands, fall short of demonstrating a physiological role for endogenous RARα in granulopoiesis. By studying cells derived from RARα-deficient mice, we demonstrate here that RARα is a bidirectional modulator of granulocyte differentiation: In the absence of RA, RARα acts to limit granulocyte differentiation; in the presence of RA, RARα stimulates differentiation.

RARα- and RARγ-deficient mice were described previously.25,26 Mice were genotyped by polymerase chain reaction (PCR) (primers available upon request). All mice used in this study were between 2 and 6 months of age. Vitamin A deficiency (VAD) in cellular retinol-binding protein type I (CRBPI)^−/− mice was produced by maintaining CRBPI^−/− mice (all males) on a VAD diet for at least 16 weeks after birth.27 VAD was confirmed by the observation of several VAD-related phenotypes: keratinization of the seminal vesicle and prostate epithelia, testis degeneration, and retinal defects.27 These symptoms were specifically observed in CRBPI^−/− mice fed a VAD diet, not in control mice (ie, WT mice fed a VAD diet or CRBPI^−/− mice fed a normal diet). All-trans-retinoic acid (ATRA) (Sigma Chemical Co, St Louis, MO) and the pan-RAR antagonist BMS493 were dissolved in ethanol and then placed in an emulsion with corn oil by vigorous vortexing. The emulsion was administered to test animals by gavage every 12 hours (BMS493) or every 36 hours (ATRA) for 3 days. Bone marrow (BM) cells were harvested by flushing femurs with medium. Fetal liver (FL) cell suspensions were prepared by sequential passage of dissected FLs through 18-, 22-, and 25-gauge needles.

Liquid myeloid cultures were performed by seeding 2 × 10⁶ BM cells in 10 mL Iscoves modified Dulbecco medium supplemented with 20% fetal bovine serum (Gibco-BRL Life Technologies, Grand Island, NJ), 5 ng/mL granulocyte colony-stimulating factor (G-CSF), and 5 ng/mL stem cell factor (SCF) (R&D Systems, Minneapolis, MN), penicillin/streptomycin, and L-glutamine (2 mM final concentration). Ligands were added in 10 μL volumes from a 10⁻³ M solution in ethanol. Labeling with 5- and 6-carboxy fluorescein diacetate, succinimidyl ester (CFSE) (Molecular Probes, Eugene, OR) was performed by incubating BM cells for 10 minutes at 37°C at a concentration of 10⁷ cells per mL in medium supplemented with 2 μL/mL of a 10 mg/mL CFSE solution in dimethyl sulfoxide followed by 2 washes with 10 mL medium. Clonogenic cultures were performed with commercial methylcellulose media (Methocult 3434; Stem Cell Technologies, Vancouver, BC, Canada). The media contained erythropoietin (EPO), interleukin-3 (IL-3), IL-6, and SCF for the experiments described in Figure1C and Figure2, and the media contained Methocult 3234 supplemented with granulocyte-macrophage (GM)–CSF (R&D Systems) at a final concentration of 50 ng/mL for the experiments described in Figure4.

The following antibodies were used: fluorescein isothiocyanate (FITC)–labeled GR-1 (Caltag, South San Francisco, CA) or phycoerythrin-labeled GR-1 (Pharmingen, San Diego, CA) and biotin-labeled Mac-1 (CD11b, Pharmingen). Streptavidin-Cy5 was purchased from Jackson Immunoresearch Laboratories, West Grove, PA. Antibody stainings were performed on 10⁶ BM or FL cells and analyzed (Coulter Elite cytometer; Coulter Electronics, Hialeah, FL). Listmode data were collected on 3 × 10⁴ cells.

Reverse transcriptase (RT)–PCR reactions were performed on 50 ng complementary DNA (cDNA) using the following oligonucleotides: cathepsin G: 5′-GATTGTAATCAGGATGGCGG-3′, 5′-CTGACTAAGCAACGGTTCTGG-3′; myeloperoxidase: 5′-ATGCAGTGGGGACAGTTTCTG-3′, 5′-GTCGTTGTAGGATCGGTACTG-3′; elastase: 5′-CACCATCAGTCAGGTCTTCC-3′, 5′-AGTCTCCGAAGCATATGCC-3′; β-actin: 5′-GTGACGAGGCCCAGAGCAAGAG-3′, 5′-AGGGGCCGGACTCATCGTACTC-3′; RARβ: 5′-CTCGTCCCGAGCCCACCATCTCCACTT-3′, 5′-GAGGTCGTCTAGCTCCGCTGTCATCTC-3′; and hypoxanthine phosphoribosyltransferase (HPRT): 5′-GTAATGATCAGTCAACG-3′, 5′-CCAGCAAGCTTGCAACCTTAACCA-3′. RARβ and HPRT amplification products (Figure 1D) were detected after Southern blot analysis and hybridization with phosphorous 32 (³²P)-labeled oligonucleotide probes (5′-CCATAGTGGTAGCC-3′ and 5′-CACTGGTAAAACAATGCAAA-3′, respectively).

Granulocyte populations appeared normal in RARα^−/− mice at the level of blood neutrophil counts (data not shown) and in the BM (GR-1⁺Mac-1⁺population; Figure 5C, upper right panel, and data not shown). To investigate whether functional compensation by other RARs occurs in RARα^−/− mutants, we examined RARα^−/−RARγ^−/− double mutants, as RARγ is also expressed in granulocytes.17 Because RARα^−/−RARγ^−/− mutants die in utero,28 granulopoiesis was studied in the FL. Granulocytes developed normally in RARα^−/−RARγ^−/− E14.5 FLs as tested by several criteria: double-mutant myeloid cells exhibited a normal GR-1⁺Mac-1⁺ phenotype (Figure 1A); differentiated cells (containing ring-shaped or segmented nuclei) were detected at similar frequencies on cytospin preparations from mutant and WT FLs (data not shown); granulocyte-specific markers were expressed at similar levels (Figure 1B); and morphologically normal neutrophils were generated in myeloid colonies in vitro (Figure 1C). Because RARβ expression was not detected in freshly isolated or cultured FL cells (Figure 1D), granulopoiesis can thus proceed efficiently in the absence of any RAR.

To determine whether granulocyte progenitors from RARα^−/− mice could still respond to retinoids, we compared the response of WT and RARα^−/− BM cells to 9-cis retinoic acid (9C-RA, a pan-RAR and pan-RXR [retinoid X receptor] agonist) in a myeloid colony-forming assay. As previously reported,29 the total number of colonies developing from WT BM cells, as well as the size of pure granulocyte colonies, were markedly reduced in the presence of 9C-RA (Figure 2A,B). Neither of these effects were observed in cultures of RARα^−/− BM cells (Figure 2) (P < .01), indicating that RARα^−/− granulocyte progenitors are impaired in their response to retinoids. These results, together with the apparent normal granulopoiesis in RARα^−/−RARγ^−/− mutants, demonstrate that RARs and retinoids are dispensable for granulopoiesis.

Although the RA/RAR pathway is not required, it can clearly modulate granulocyte differentiation, as shown by the response of WT cells to retinoids. Consequently, we analyzed the effects of RAR agonists and antagonists on granulocyte differentiation in liquid cultures of WT BM cells stimulated with G-CSF and SCF. To monitor cells that had undergone division during the culture period, freshly isolated BM cells were cytoplasmically labeled with the vital dye CFSE (a fluorescent compound that is gradually diluted upon cell division) prior to culture, and analysis was confined to CFSE^−/lowcells (Figure 3A). Increased levels of GR-1 and Mac-1 expression were used to measure differentiation.30 9C-RA induced an increase in the proportion of de novo GR-1^high cells in cultures of WT cells (Figure 3B, upper right panel, gate C, and Figure 3C) as well as a decrease in the proportion of CFSE^−/low cells (data not shown), which is consistent with the premise that RA limits the proliferation of granulocyte precursors and induces their differentiation.

In contrast, the pan-RAR antagonist BMS493 (H. Gronemeyer and P.C., unpublished observation, February 1996) induced an increase in the proportion of immature granulocytes (Figure 3B, upper middle panel, gate B, and Figure 3C), a finding corroborated by the higher frequency of promyelocytes/myelocytes on cytospin preparations from these cultures (data not shown). Interestingly, the magnitude of the response to BMS493 greatly exceeded that to 9C-RA, suggesting that retinoids in the serum might play an important part in controlling the differentiation that occurs in the absence of exogenous ligands. When RARα^−/− BM cells were grown under identical culture conditions, the ratio of immature versus mature cells generated was unaffected by treatment with either 9C-RA or BMS493 (Figure 3B,C), confirming the resistance of RARα^−/− granulocytes to the effects of RAR ligands. In addition, the proportions of mature cells were consistently increased when compared with WT samples in the control and BMS493-treated cultures to levels usually observed in 9C-RA–treated WT cultures (Figure 3B,C). Thus granulocyte differentiation appears to be accelerated in the absence of RARα.

This apparent acceleration of maturation was also seen in clonogenic methylcellulose cultures stimulated with GM-CSF (Figure 4). After 5 days of culture, granulocytes from pooled RARα^−/−colonies were more mature than those from WT colonies (compare Figure4A with B and inset). By day 8, there was a marked underrepresentation of granulocytes in the mutant cultures, which now consisted mainly of macrophages (Figure 4C,D), suggesting that the prematurely differentiated neutrophils seen on day 5 of culture had died off over time. At the colony level, dense granulocyte clusters were less abundant and smaller in RARα^−/− cultures after 8 days (Figure 4E,F), a phenotype that reliably allowed for the blind separation of dishes containing WT or RARα^−/− cells. It is noteworthy that both types of experimental cultures used different cytokines (G-CSF and SCF for the liquid cultures and GM-CSF for the methylcellulose cultures), thus excluding a cytokine-specific effect. These observations provide evidence that lack of RARα relieves a physiological restraint on granulocyte differentiation normally exerted by this receptor in the absence of ligand.

To understand how retinoids may modulate granulopoiesis in vivo, we evaluated granulocyte differentiation in WT mice treated with BMS493. Of 11 treated WT mice, 7 mice exhibited a clear increase in the proportion of immature granulocytes in the BM (Mac-1^low/+GR-1^−/low, Figure 5A). As this increase was not seen in vehicle-treated or untreated animals, its occurrence in BMS493-treated mice most likely reflects a role for endogenous retinoids in regulating granulocyte differentiation. To investigate further, we analyzed VAD mice using CRBPI-deficient mice that, owing to low vitamin A stores, become readily VAD when starved for vitamin A.27 In all VAD CRBPI^−/− mice tested (n = 6), there was a marked expansion in the size of the entire granulocyte population (Figure 5B, upper right panel) when compared with control animals (ie, CRBPI^−/− mice on a normal diet or CRBPI^+/+ mice on a VAD diet) as well as an increase in the proportion of immature Mac-1^low/+GR-1^−/low cells (Figure 5B, lower right panel). This effect was accompanied by an increase of neutrophils in the spleen and blood (data not shown), which is consistent with a previous report showing enhanced neutrophil numbers in VAD rats.31 Importantly, the VAD-induced increase in the immature granulocyte compartment of the BM mirrors that seen following BMS493 administration (Figure 5A), underscoring a role for endogenous retinoids in the positive control of granulopoiesis in vivo.

In contrast, ATRA treatment of WT mice (n = 7) elicited an increase in the levels of GR-1 and Mac-1 expressed by the Mac-1^low/+GR-1^−/low immature granulocyte population (Figure 5C, gated region, compare upper and lower left panels), an effect that did not occur in RARα^−/− mice (n = 4) (Figure 5C, right panels). This indicates that RARα is the direct mediator of exogenous (and, by inference, endogenous) retinoid-induced differentiation. Significantly, the fact that all 7 ATRA-treated WT mice responded in a similar manner also suggests that under normal dietary conditions, RARα proteins are not fully liganded in vivo, thus placing them in an appropriate state to negatively regulate differentiation.

We have shown here that (1) endogenous RARα is a key mediator of the effects of retinoids on granulopoiesis, and (2) the RARα receptor can bidirectionally modulate granulopoiesis as a differentiation factor when liganded to RA or as an inhibitor in the absence of ligand (or when liganded to an antagonist), probably by acting as a transcriptional repressor.32-34 The fact that RA is a differentiating agent for granulocytes in culture has been well established for both normal and leukemic cells.19,20,35However, the role of retinoids during normal granulocyte differentiation in vivo has so far not been documented. The increase in the immature granulocyte populations, seen both in the VAD and BMS493-treated mice here, show for the first time that retinoid-controlled differentiation actively contributes to the regulation of granulocyte numbers in vivo. Our VAD results are in agreement with those of Kuwata et al,36 who recently reported a collective increase in granulocyte numbers during murine VAD as well as with an earlier report documenting elevated neutrophils in the peripheral blood of VAD rats.31 Although Kuwata et al36 did not study the relative contributions of immature versus mature cells in their granulocyte population, their VAD mice also seemed to accumulate more cells with a GR-1^lowphenotype, as seen in the Kuwata et al Figure 3A,36 and may therefore exhibit a phenotype similar to that of our VAD mice. We have noted that optimal resolution of the GR-1^low and GR-1^high populations appears to depend on the fluorochrome to which the anti–GR-1 antibody is conjugated; FITC-conjugated anti–GR-1 (as used in our study) provided the best resolution, while other conjugated forms of this antibody appeared to give more compact stainings similar to those shown by Kuwata et al.36 These authors also found less apoptosis within the granulocyte population of their VAD mice, an observation that they used to explain their increase in granulocyte numbers. Although we cannot exclude that retinoids directly regulate cell survival, it is possible that the decrease in apoptosis seen by Kuwata and colleagues is an indirect consequence of a shift toward the more immature stages because apoptosis is the ultimate endpoint of differentiation.

The finding that endogenous unliganded RARα negatively regulates differentiation is consistent with results from previous studies showing that overexpression of WT RARα blocks promyelocyte differentiation in vitro.22 23 However, the non-physiological levels of receptor expressed in these systems have prevented any clear conclusion about the role of the endogenous molecule. In fact, these latter studies often claimed that the overexpressed receptors functioned in a dominant-negative manner by titrating out retinoids present in the medium, thereby countering the differentiating role of these compounds. Our present results suggest otherwise; the differentiation block exerted by overexpressed RARα corresponds to an amplification of an intrinsic inhibitory activity of the unliganded receptor. It is important to stress that in contrast to the overexpression studies, blocking the activation of the endogenous receptor in vivo or in vitro (with VAD or BMS493) never resulted in a complete arrest in differentiation such as the arrest achieved by overexpressed receptors, thereby reflecting the possibility that RARα is expressed in limiting amounts.

Intriguingly, our results appear to contradict those of Labrecque et al,17 who reported that granulocytes from myeloid colonies derived from RARα1^−/−RARγ^−/−neonatal BM cells tend to be more immature than those derived from WT cells (RARα1 being one of the 2 major isoforms produced from the RARα gene).37 As we never observed such a block in colonies derived from RARα^−/−RARγ^−/− FL cells, the apparent difference in the behavior of the RARα1^−/−RARγ^−/− cells might be due to the persistent expression of RARα2, an isoform not present in our double mutants. It should be noted that we used FL cells in our study, while Labrecque and colleagues17 used BM cells. However, we observed a similar premature differentiation in GM-CSF–dependent colonies derived from either FL or adult BM RARα^−/−cells (unpublished observations, February 1998). Thus, it appears unlikely that RARs control granulopoiesis differently in FL and BM cells. With respect to RARγ, its role appears to be minor at best, as cells lacking only RARα are already resistant to the effects of retinoids.

The present results demonstrate that granulocyte differentiation can proceed in the absence of RARs. The RA/RARα pathway is therefore not an obligatory regulator of granulopoiesis, but rather a modulator of this differentiation. Both retinoid excess and deficiency affect granulopoiesis, suggesting that a balance between liganded and unliganded receptor molecules exists in vivo under normal conditions. The question therefore arises as to which physiological or pathological conditions modulate retinoid levels. Dietary VAD, a condition that is associated with higher susceptibility to bacterial infections, is an obvious candidate. The negative regulation of granulopoiesis by unliganded RARα (which ultimately translates into an elevated granulocyte pool) might thus represent an innate response used to off-set VAD-associated infections. In this view, an important role of RARα during granulopoiesis could paradoxically be to mediate the effect of ligand deficiency. Analysis of the VAD response in RARα^−/− mice should help to clarify this issue.

At the molecular level, several downstream candidates might mediate the actions of RARα. A prime candidate is C-EBPε, an RA-target gene induced by RA in myeloid cell lines38,39 and shown to be essential for terminal granulocyte differentiation.40However, because granulocytes can fully differentiate in the absence of RARs or during retinoid deficiency, it is unlikely that retinoids are the only regulators of C-EBPε expression in vivo. In this respect we have found that overall, C-EBPε transcript levels were not altered in the BM of RARα^−/− mice (unpublished data, March 1998). Other interesting candidates are the cdk inhibitors p21 and p27, whose increases have been shown to be associated with the transition between immature, proliferating cells and post-mitotic differentiated cells for several cell types including myeloid cells.41-45Interestingly, p21 has been proposed to be a target gene for nuclear receptors, RARs included.41,46 How these proteins are regulated in cells lacking RARα thus merits further investigation. Hox genes, which are regulated by retinoic acid in the embryo, might also contribute to mediating RAR function during hematopoiesis47; their precise role during granulopoiesis, however, and their possible regulation by RARs remain unknown. Finally, Mad1 has been shown to positively regulate cell cycle exit during granulopoiesis, possibly by antagonizing the effects of myc proteins.48 Whether the Myc/Mad and the RAR pathways are parallel or convergent modulators of granulopoiesis remains to be seen.

The role of retinoid signaling during hematopoiesis is likely to extend beyond its role of regulating granulocyte differentiation described here. RA has also been suggested to act as a commitment factor for the granulocyte lineage, as exogenous RA is able to redirect erythroid-, monocyte-, or eosinophil-committed progenitors toward a granulocyte fate.24,49 Although our study does not directly address this issue, it is clear that other factors besides RA can influence granulocyte lineage commitment, as commitment was not impaired in RARα^−/−RARγ^−/− double mutants or upon VAD. However, the proportion of granulocytes was often reduced in myeloid colonies derived from RARα^−/− BM cells, and the frequency of granulocyte-only colonies was diminished (unpublished data, October 1998), phenotypes that could indicate defective granulocyte commitment in the mutants. The present effects of RAR agonists and antagonists during terminal granulocyte differentiation also contrast with those observed when similar compounds were used on hematopoietic stem cells, where RAR agonists and antagonists antagonized and promoted their differentiation, respectively.50 These differences suggest a complex pattern of retinoid receptor function during hematopoiesis.

RARα chromosomal rearrangements represent the genetic hallmark of APL. The tight association between RARα alterations and APL strongly suggests that a normal RARα-dependent function must be altered to transform promyelocytes. Generally, it has been assumed that the APL RARα fusion proteins act as dominant-negative inhibitors of retinoid-dependent differentiation.51,52 However, because granulocyte differentiation is only partially blocked when retinoid signaling through RAR is prevented, dominant-negative inhibition of retinoid-induced differentiation cannot account for the complete differentiation block seen in APL cells. Our data suggest an alternative mechanism of action for the RAR fusion proteins: enhancement of the normal inhibitory function of unliganded RARα. This enhancement may be due to an overexpression of the fusion proteins vis-à-vis WT RARα, as is often observed in APL cells,53,54 and/or to the more potent repressive properties of the fusion proteins themselves. In this respect, PML-RARα and PLZF-RARα are powerful transcriptional repressors that efficiently recruit histone deacetylases to target promoters.9,12-15 Interestingly, overexpression of thyroid hormone receptor α (TRα) in erythroid precursors blocks their differentiation,55 suggesting that the differentiation arrest mediated by v-erbA, the oncogenic form of TRα, during erythropoiesis might also correspond to an enhancement of a normal property of TRα. Different nuclear receptors may therefore control differentiation and leukemogenesis in distinct hematopoietic lineages via similar mechanisms.

If interference with RARα-regulated differentiation appears important for APL pathogenesis, it probably represents only one facet of the action of RARα fusion proteins in APL. Fusion of RARα with PML or PLZF has been shown to be required to provoke APL.11Dominant interference with other functions, such as those ofPML (a tumor suppressor gene that controls apoptosis and p53 acetylation), is likely to be important.1,4,56-58 In this respect the coiled coil moiety of PML appears to mediate the formation of high molecular weight oligomeric complexes containing both PML-RARα and PML, complexes which could disrupt normal PML function.59 Simultaneous targeting of several cellular functions by the RARα fusion proteins might therefore cooperate to transform promyelocytes.

We thank P. Reczek (Bristol-Myers Squibb, Princeton, NJ) for providing BMS493 and M. Sporn for 9cis-RA; C. Bronn, B. Bondeau, and B. Feret for excellent technical assistance; C. Ebel for help with flow cytometry; and A. Pailleux, F. Fischer, and P. Michel for maintaining the mice.