Nur77 variants solely comprising the amino-terminal domain activate hypoxia-inducible factor-1α and affect bone marrow homeostasis in mice and humans

Gene targeting via homologous recombination can occasionally result in incomplete disruption of the targeted gene. Here, we show that a widely used Nur77-deficient transgenic mouse model expresses a truncated protein encoding for part of the N-terminal domain of nuclear receptor Nur77. This truncated Nur77 protein is absent in a newly developed Nur77-deficient mouse strain generated using Cre-Lox recombination. Comparison of these two mouse strains using immunohistochemistry, flow cytometry, and colony-forming assays shows that homologous recombination–derived Nur77-deficient mice, but not WT or Cre-Lox–derived Nur77-deficient mice, suffer from liver immune cell infiltrates, loss of splenic architecture, and increased numbers of bone marrow hematopoietic stem cells and splenic colony–forming cells with age. Mechanistically, we demonstrate that the truncated Nur77 N-terminal domain protein maintains the stability and activity of hypoxia-inducible factor (HIF)-1, a transcription factor known to regulate bone marrow homeostasis. Additionally, a previously discovered, but uncharacterized, human Nur77 transcript variant that encodes solely for its N-terminal domain, designated TR3β, can also stabilize and activate HIF-1α. Meta-analysis of publicly available microarray data sets shows that TR3β is highly expressed in human bone marrow cells and acute myeloid leukemia samples. In conclusion, our study provides evidence that a transgenic mouse model commonly used to study the biological function of Nur77 has several major drawbacks, while simultaneously identifying the importance of nongenomic Nur77 activity in the regulation of bone marrow homeostasis.

In response to inflammatory stimuli and other danger signals, the bone marrow can give rise to a wide range of special-ized immune cells. The production of such immune cells is tightly regulated at each step of differentiation from hematopoietic stem and progenitor cell (HSPC) 5 to functional leukocyte (1). HSPCs reside in specific areas of the bone marrow, termed the HSPC niche, where their survival is supported by low oxygen tension and concomitant induction of the transcription factor hypoxia-inducible factor (HIF)-1 (2). Tight control of HIF-1 expression and activity is crucial for both bone marrow homeostasis and the biological function of HSPCs (3), and elevated expression of this factor may enhance HSPC mobilization from the bone marrow into the circulation (4).
Nur77 (Nr4a1) is a transcription factor belonging to the NR4A nuclear receptor family. Together with its family members Nurr1 (Nr4a2) and NOR-1 (Nr4a3), it regulates a wide variety of biological processes, including inflammation, metabolism, and apoptosis (5). Like other nuclear receptors, Nur77 consists of three protein domains: the N-terminal domain, the DNA-binding domain (DBD), and the ligand-binding domain (LBD). Nur77 is known to play an important role in various aspects of hematopoiesis. For example, the differentiation of Ly6C lo patrolling monocytes from more naive myeloid progenitors crucially depends on Nur77, as Nur77-deficient mice essentially lack this subset of monocytes (6). Additionally, a dual deficiency of Nur77 and its family member NOR-1 leads to rapid development of acute myeloid leukemia (AML), with bone marrow differentiation being heavily skewed toward myeloid progeny (7). Finally, Nur77-deficient mice suffer from spontaneous, systemic inflammation with age characterized by excessive immune cell infiltration of various peripheral organs (8).
Whole-body Nur77-deficient mice generated via homologous recombination (B6;129S2-Nr4a1 tm1Jmi /J, JAX stock no. 006187) have been used in a plethora of studies to investigate the biological functions of Nur77. However, the gene-targeting strategy originally used to generate these mice disrupts neither the transcriptional nor the translational start site of the Nr4a1 gene locus (9). Instead, a neomycin resistance gene (Neo R ) cassette is inserted ϳ300 bp downstream of the Nr4a1 translational start site. Although such a targeting strategy could leave the Nr4a1 gene locus nonfunctional, previous reports have shown that disrupting gene loci downstream of the translational start site can lead to the production of truncated but biologically active proteins (10,11).
Here, we show that Nur77-deficient mice generated via homologous recombination (referred to as classical Nur77-KO mice) develop bone marrow, liver, and spleen abnormalities with age, which are absent in similarly aged WT mice or a newly generated Cre-Lox system-derived Nur77-deficient mouse line (referred to as complete Nur77-KO mice). Classical Nur77-KO mice still express part of the N-terminal domain of Nur77, which is not present in complete Nur77-KO mice. We show that this truncated Nur77 protein can stabilize and activate HIF-1 in vitro. In vivo, classical Nur77-KO mice maintain the expression of HIF-1 and its target genes in the bone marrow at WT levels, whereas complete Nur77-KO mice do not. Given that both classical Nur77-KO and complete Nur77-KO mice lack Nur77 transcriptional activity, these mice have altered expression of bone marrow mobilization genes. Finally, we show that the truncated N-terminal domain protein may be relevant in human disease, because a hitherto uncharacterized transcript variant of human Nur77, also encoding solely for the N-terminal domain (referred to as TR3␤), possesses similar HIF-1-stabilizing and activating abilities and is highly expressed in human bone marrow and AML samples, as opposed to the full-length Nur77 protein.

Classical Nur77-KO mice, but not WT or complete Nur77-KO mice, exhibit liver immune cell infiltrates and loss of splenic architecture at old age
It has previously been reported that the classical Nur77-KO mice, generated via homologous recombination, display systemic inflammation and immune cell infiltrates in liver and spleen upon aging (8). We decided to investigate whether this phenotype is a true hallmark of Nur77 deficiency by comparing WT and the classical Nur77-KO mice with the newly developed complete Nur77-KO mice, in which exons 2-4 of the Nr4a1 gene are removed using cytomegalovirus promoter-driven Cre-Lox recombinase. We confirmed the presence of liver immune cell infiltrates in our 6.5-month-old classical Nur77-KO mice (Fig. 1A). Additionally, we found that these infiltrates consist of macrophages (IBA-1-positive), T-cells (CD3-positive), and B-cells (B220-positive; Fig. 1B). Interestingly, these liver infiltrates were not present in age-matched WT or complete Nur77-KO mice (Fig. 1A). Similarly, immunohistochemical assessment of the spleen revealed a loss of splenic architecture in aged classical Nur77-KO mice, but not in aged WT or complete Nur77-KO mice (Fig. 1C). This altered spleen morphology was accompanied by a lower intensity staining for T-cells (CD3), B-cells (B220), and macrophages (IBA-1) in classical Nur77-KO mice when compared with both WT and complete Nur77-KO mice, suggesting that spleens of classical Nur77-KO mice contain more naive immune cell progenitors (Fig. 1C). These phenotypes were not overtly present in younger (3-month-old) classical Nur77-KO mice, however, which had liver and spleen morphologies comparable with that of 3-month-old WT and complete Nur77-KO mice (Fig. S1, A  and B).
In response to infections and other inflammatory stimuli, the spleen and liver may function as emergency sites of extramedullary hematopoiesis (12). We therefore speculated that the immune cell liver infiltrates and splenic disorganization observed exclusively in classical Nur77-KO mice might arise from abnormalities in their bone marrow and hematopoiesis. There were no differences in the total number of bone marrow cells between either 3-month-old or 6.5-month-old WT, classical Nur77-KO, and complete Nur77-KO mice (Fig. S1C). However, bone marrow from classical Nur77-KO mice did give rise to significantly more bone marrow-derived macrophages than either WT or complete Nur77-KO bone marrow, suggesting that the bone marrow in classical Nur77-KO mice is skewed toward myeloid progeny (Fig. S1D). Furthermore, classical Nur77-KO mice had significantly more HSPCs, defined as lineage-negative (Lin Ϫ ), Sca1-positive (Sca1 ϩ ), and cKit-positive (cKit ϩ ) cells (LSK cells), in the bone marrow than either WT or complete Nur77-KO mice at 6.5 months of age (Fig. S1E). Moreover, we observed that spleens isolated from classical Nur77-KO mice gave rise to significantly more HSPC-derived colonies than spleens from either WT or complete Nur77-KO mice did (Fig. S1F). There were no significant differences in splenic LSK cell counts or the colony-forming capacity of bone marrow between the three mouse strains (Fig. S1, E and F).
Together, these results suggest that the liver immune cell infiltrates and splenic disorganization observed exclusively in classical Nur77-KO mice are due to aging-induced irregularities in their bone marrow compartment, leading to increased seeding of peripheral tissues, such as the spleen and liver, with hematopoietic progenitor cells.

Classical Nur77-KO mice express part of the N-terminal domain of full-length Nur77
We were intrigued by the phenotypic differences between aged classical Nur77-KO and complete Nur77-KO mice and decided to investigate a possible cause for this discrepancy in more detail. Classical Nur77-KO mice were originally generated by insertion of a Neo R cassette in exon 2 of the Nr4a1 locus (9). This gene-targeting strategy leaves both the transcriptional and translational start sites of the Nr4a1 gene locus entirely intact ( Fig. 2A and Fig. S2A). It has previously been shown that such gene-targeting strategies can lead to the production of aberrant mRNA and protein products from the targeted gene locus (10,11). Therefore, we assessed whether classical Nur77-KO mice still produce mRNA and protein from the Nr4a1 gene locus. Nur77 mRNA expression was measured using primer sets targeting either exons 3 and 4, which are present in WT but absent in both Nur77-KO mouse strains, or targeting exon 2, which is present in WT and classical Nur77-KO mice but absent in complete Nur77-KO mice. In bone marrow cells and in bone marrow-derived macrophages

Nur77 variants regulate HIF-1␣ and bone marrow homeostasis
(BMDMs) stimulated with lipopolysaccharide (LPS; a known inducer of Nur77 expression (13)), cells from classical Nur77-KO mice did not produce detectable Nur77 mRNA when using a primer set targeting exons 3 and 4 ( Fig. 2B) but had high Nur77 mRNA expression when using the primer set targeting exon 2 (Fig. 2C). Importantly, neither primer set could detect Nur77 mRNA in cells from complete Nur77-KO mice, whereas both primer sets detected Nur77 mRNA in cells from WT mice (Fig. 2, B and C). Similar results were found for spleen and liver mRNA of these three mouse lines (Fig. S2B).
To assess whether the Nur77 mRNA produced in classical Nur77-KO mice is translated into protein, we used a combination of Nur77 N-terminal domain-specific antibodies for immunoprecipitation (IP) and Western blotting (WB) to detect endogenous Nur77 protein as described previously (14). Fulllength Nur77 protein was successfully detected in WT BMDMs stimulated with LPS (Fig. 2D, left). Neither classical Nur77-KO nor complete Nur77-KO BMDMs expressed full-length Nur77 protein, but we did observe a smaller protein of ϳ21 kDa exclusively in LPS-stimulated classical Nur77-KO BMDMs (Fig. 2D, right). Notably, this smaller protein (referred to as Nur77(1-117) here) was not expressed in BMDMs of WT or complete Nur77-KO mice (Fig. 2D). We hypothesized that Nur77(1-117) may be the product of the aberrant mRNA expressed in classical Nur77-KO mice, which encodes for the first 115 amino acids of full-length Nur77 plus 2 amino acids from the subsequent Neo R cassette (Fig. S2A). We cloned the putative coding domain sequence of the aberrant mRNA expressed in classical Nur77-KO mice and overexpressed it in RAW264.7 macrophages, which also resulted in a single band of ϳ21 kDa (Fig.  2E). Together, these results show that an aberrant Nur77 mRNA and protein is produced from the endogenous Nr4a1 gene locus exclusively in classical Nur77-KO mice.
The cellular localization of full-length Nur77 is tightly regulated via two nuclear localization signals in its DBD and three nuclear export signals in the LBD (15), all of which are absent in the Nur77(1-117) protein ( Fig. 2A). Therefore, we assessed the cellular localization of Nur77(1-117) protein in RAW264.7 macrophages using immunofluorescence (IF) microscopy, which showed that overexpressed full-length Nur77 is contained entirely in the nucleus, whereas Nur77(1-117) is present throughout the cell (Fig. S2C). Consistently, IF microscopy using two different Nur77 N-terminal domain-specific antibodies showed that BMDMs from classical Nur77-KO mice expressed a protein, which is localized throughout the cell, while BMDMs from complete Nur77-KO mice did not (Fig.  S2D). Taken together, these results show that the aberrant Nur77 mRNA produced in classical Nur77-KO mice is translated into a protein that encodes for part of the N-terminal domain of Nur77 and that this protein is localized in both the nucleus and cytoplasm of the cell.

Nur77(1-117) increases HIF-1␣ transcriptional activity, and HIF-1␣ protein and target gene expression are elevated in bone marrow of classical Nur77-KO mice
Having established that classical Nur77-KO mice express part of the Nur77 N-terminal domain, we next investigated possible mechanisms that may be responsible for the bone mar-row and spleen phenotypes observed exclusively in these mice. It has previously been shown that classical Nur77-KO mice have significantly reduced numbers of circulating Ly6C lo monocytes (6). We speculated that a loss of these patrolling monocytes plays a role in the immune cell infiltrates observed in these mice. However, both classical Nur77-KO and complete Nur77-KO mice showed a significant decrease in circulating Ly6C lo monocytes and an increase in circulating Ly6C hi monocytes (Fig. S3A). It is therefore highly unlikely that altered differentiation of Ly6C lo monocytes is the cause of the spleen and liver abnormalities observed exclusively in classical Nur77-KO mice.
The increased presence of hematopoietic progenitor cells and differentiated leukocytes in peripheral tissues of classical Nur77-KO mice ( Fig. 1) could also be indicative of increased HSPC mobilization from the bone marrow in these mice. HSPC mobilization can be induced by various inflammatory cytokines and chemokines (16), including the chemokine stromal cellderived factor-1 (SDF-1␣) (17). Nur77 is known to repress SDF-1␣ expression in macrophages (18,19), and we have previously reported that pro-atherosclerotic Ldlr Ϫ/Ϫ mice that receive a transplantation of classical Nur77-KO bone marrow have higher circulating SDF-1␣ levels than mice that received WT bone marrow (19). However, we now report that circulating SDF-1␣ plasma levels are not changed between unchallenged WT, classical Nur77-KO, and complete Nur77-KO mice at either 3 or 6.5 months old (Fig. S3B). Similarly, expression of the inflammatory cytokine genes Il6 and Tnf was unchanged in spleens and livers of 6.5-month-old WT, classical Nur77-KO, and complete Nur77-KO mice (Fig. S3C). As such, altered levels of the above-mentioned cytokines are probably not the cause of the phenotypes observed in classical Nur77-KO mice.
In addition to induction by external cytokines, HSPC mobilization is also tightly regulated by various signaling pathways and transcription factors. One such transcription factor is HIF-1␣, which plays an important role in survival of HSPCs (3). Pharmacological stabilization of HIF-1␣ protein has been shown to significantly enhance HSPC mobilization (4). Interestingly, it has also been shown that Nur77 can stabilize HIF-1␣ protein by inhibiting MDM2-induced degradation and that the N-terminal domain of Nur77 by itself is sufficient for this stabilization to occur (20). Therefore, we considered the possibility that the N-terminal domain-encoding Nur77(1-117) protein expressed in classical Nur77-KO mice regulates HIF-1␣ stability and thereby causes the bone marrow, spleen, and liver phenotypes observed in these mice. Consistent with this hypothesis, lentiviral overexpression of Nur77(1-117) protein in BMDMs from complete Nur77-KO mice resulted in increased HIF-1␣ protein expression in response to the hypoxia mimetic cobalt chloride (Fig. 3A). Overexpressing either fulllength Nur77 or Nur77(1-117) protein in RAW264.7 macrophages also significantly increased HIF-1 transcriptional activity in response to cobalt chloride, deferoxamine (DFO), or hypoxic growth conditions (Fig. 3B). These results demonstrate that Nur77(1-117) overexpression can increase HIF-1␣ protein stability and HIF-1 transcriptional activity in vitro.
In vivo, we found that HIF-1␣ protein expression in the bone marrow was similar between WT and classical Nur77-KO mice,

Nur77 variants regulate HIF-1␣ and bone marrow homeostasis
whereas complete Nur77-KO mice had significantly less HIF-1␣ when compared with classical Nur77-KO mice (Fig.  3C). This result is consistent with the idea that HIF-1␣ is stabilized by a nongenomic function of full-length Nur77 in WT mice and by Nur77(1-117) in classical Nur77-KO mice, but not stabilized in complete Nur77-KO mice. Moreover, the lower level of HIF-1␣ protein in complete Nur77-KO mice was accompanied by decreased expression of several established HIF-1 target genes in the bone marrow, whereas the expression of these same genes was unchanged between WT and classical Nur77-KO mice (Fig. 3D). Finally, we found that gene expression of two chemokine receptors important for the induction of HSPC mobilization, Cxcr4 and Csf3r (21,22), was significantly higher in bone marrow of both classical Nur77-KO and complete Nur77-KO mice compared with WT mice (Fig. 3E), revealing a Nur77-mediated genomic function by suppression of these genes. On the other hand, gene expression of Egfr and Vcam1, two factors that promote HSPC retention in the bone marrow (23,24), was significantly lower in both classical Nur77-KO and complete Nur77-KO mice compared with WT mice (Fig. 3F).
These results show that Nur77(1-117) can stabilize and activate HIF-1␣ both in vitro and in vivo. Combined with the altered expression of several genes that regulate bone marrow mobilization in both classical Nur77-KO and complete Nur77-KO mice when compared with WT mice, we speculate that the bone marrow, liver, and spleen phenotypes observed exclusively in classical Nur77-KO mice are caused by an imbalance between the genomic action of Nur77 as a transcription factor versus its nongenomic action as a stabilizer of HIF-1␣.

A transcript variant of human NUR77 encoding solely for its N-terminal domain (TR3␤) has similar functions as the Nur77(1-117) protein expressed in classical Nur77-KO mice
Our findings for Nur77(1-117) may have implications beyond the transgenic mouse models described here, because a transcript variant of human NUR77 encoding for the entire NUR77 N-terminal domain was previously detected and deposited in the GenBank TM database (GenBank TM accession number D85245; Ensembl ENST00000548232). To date, no further characterization of this NUR77 variant has been reported. The transcript variant, designated TR3␤, shares a large part of NR4A1 exon 2 with full-length Nur77, but has a distinctive transcriptional start site, exon structure, and 3Ј-UTR ( Fig. 4A and Fig. S4A). Using a primer set targeting these unique features, TR3␤ mRNA expression was detected in various human tissues and cells (Fig. 4B). Furthermore, we found that TR3␤ mRNA expression, like full-length NUR77 mRNA, is induced by LPS stimulation in human monocyte-derived macrophages ( Fig. 4B and Fig. S4B) and the THP-1 monocyte cell line (Fig. S4C).
Similar to the mouse Nur77(1-117) protein, the TR3␤ transcript variant lacks all nuclear localization signals and nuclear export signals present in full-length NUR77. Consistently, overexpressed TR3␤ protein, like Nur77(1-117), localized throughout the cell (Fig. 4C). Using a similar Nur77-specific antibody IP and Western blotting approach that we used to detect Nur77(1-117) protein, we identified a band of 77 kDa (full-length NUR77) and a band of ϳ34 kDa in LPS-stimulated K562 bone marrow cells (Fig. 4D). This latter band closely corresponds with the predicted molecular mass of 33.8 kDa for TR3␤ based on its primary amino acid sequence. Furthermore, TR3␤ overexpression resulted in both increased hypoxia-induced HIF-1␣ protein expression (Fig. 4E) and higher HIF-1 transcriptional activity (Fig. 4F). Finally, both TR3␤ and Nur77(1-117) protein reduced the activation of a Nur77 response element (NurRE)-driven luciferase reporter construct by full-length Nur77, indicating that these two truncated Nur77 proteins can reduce the transcriptional activity of full-length Nur77 (Fig. 4G). Taken together, these results show that the human Nur77 transcript variant TR3␤ has cellular localization and HIF-1␣-activating properties similar to those of the murine Nur77(1-117) protein and also inhibits the transcriptional activity of full-length human Nur77.

The human NUR77 transcript variant TR3␤ is highly expressed in bone marrow and associated cancers
Because TR3␤ has similar effects as the mouse Nur77(1-117) protein, and classical Nur77-KO mice exhibit splenic and bone marrow abnormalities, we next investigated whether TR3␤ is expressed differentially in human bone marrow pathologies. To do so, we explored publicly available gene expression data sets. Interestingly, the Affymetrix Human Genome U133 Plus 2.0 Array contains multiple probes for products of the NR4A1 gene locus, one of which specifically targets the unique 3Ј-UTR of TR3␤ (probe ID: 210226_at; Fig. 4A). This allowed us to perform a meta-analysis of all publicly available microarray data sets using the Affymetrix Human Genome U133 Plus 2.0 Array with the NEBION Genevestigator software suite. From this meta-analysis, we found that full-length NUR77 mRNA is most highly expressed in various tissues in which it is known to play an important role, such as heart (25), lung (26), and adrenal gland (27) (Fig. 5A, left). When looking at the types of cancers with the highest expression of NUR77, a mix of endocrine tumors and lymphomas were found (Fig. 5A, right). On the other hand, TR3␤ mRNA is most highly expressed in a set of tissues distinct from full-length Nur77, which most notably includes a number of bone marrow cell types (Fig. 5B, left). Moreover, 4 of the 10 cancer types with the highest average expression of TR3␤ were various subtypes of AML (Fig. 5B,  right). Consistent with this latter observation, pathway analysis on a set of 800 genes whose expression most strongly correlated with that of TR3␤ showed enrichment of genes known to be involved in AML, such as RUNX1T1, SOS2, PIK3CB, and KRAS ( Fig. 5D; genes indicated by red dots). On the other hand, genes that strongly correlated in expression with full-length Nur77 were predominantly associated with coagulation, infection, and breast cancer pathways, but not AML (Fig. 5C). Together, these results indicate that the human NUR77 transcript variant TR3␤ is strongly associated with high bone marrow expression and AML.

Discussion
Nur77 regulates various aspects of hematopoiesis and related pathologies, such as monocyte differentiation (6) and AML (7). Although its actions as a transcription factor are established in

Nur77 variants regulate HIF-1␣ and bone marrow homeostasis
these processes, relatively little is known about the role of nongenomic Nur77 activity in the bone marrow. In that regard, the fact that the classical Nur77-KO mouse model still expresses part of the Nur77 N-terminal domain gave us insight into the nongenomic (N-terminal domain-mediated) and genomic actions of Nur77 in the regulation of bone marrow homeostasis.
Although the production of aberrant mRNAs and truncated proteins in transgenic mice due to suboptimal gene targeting strategies are an established phenomenon (10,11), the resulting mRNAs are not always translated and may be degraded by the cell's internal quality-control mechanisms, such as nonsensemediated RNA decay (NMD) (30). The NMD pathway is strongly induced by the presence of premature termination codons (PTCs): translational stop signals that precede the final exon-exon junction in the nascent mRNA that are bound by exon junction complexes. Although the Nur77(1-117)encoding mRNA produced in classical Nur77-KO mice contains an in-frame PTC directly following Nur77 exon 2, this mRNA may escape NMD for two reasons: 1) there are no intronic sequences, and therefore no exon-exon junctions or exon junction complexes, between Nur77 exon 2 and the PTC; 2) the Neo R cassette contains a polyadenylation sequence (9), which is a strong trigger for transcriptional termination (31). In fact, this second point provides a plausible explanation as to why the mRNA produced in classical Nur77-KO mice cannot be detected by primers targeting Nur77 exons 3 and 4, namely that these latter exons are simply not transcribed. Moreover, the presence of strong polyadenylation sequences derived from the Neo R cassette in the mRNA produced in classical Nur77-KO mice may make it extremely stable, which is consistent with the observation that the expression of this aberrant mRNA stays elevated for longer periods of time following LPS stimulation of classical Nur77-KO BMDMs (Fig. 2C).
Based on the data presented here, we propose that the hematopoietic phenotypes observed exclusively in aged classical Nur77-KO mice (namely increased LSK cells in the bone mar-row, immune cell infiltrates in the liver, loss of splenic architecture, and increased numbers of HSPC-derived colonies in the spleen) are caused by an imbalance between the genomic action of Nur77 as a regulator of genes involved in bone marrow mobilization (which is absent in these mice) and its nongenomic action as a stabilizer of HIF-1␣ (which is still present). This imbalance in Nur77 and HIF-1 signaling is not present in complete Nur77-KO mice, which may explain their apparent lack of overt hematopoietic phenotypes compared with classical Nur77-KO mice (Fig. 6). The phenotype of classical Nur77-KO mice may be further aggravated by the increased stability of the Nur77(1-117)-encoding mRNA, leading to higher Nur77(1-117) expression for longer periods of time (as was observed in LPS-stimulated macrophages).
Nur77 has previously been shown to stabilize HIF-1␣ protein by inhibiting the expression of the E3 ligase MDM2 and thereby its subsequent degradation of HIF-1 ␣. The N-terminal domain of Nur77 by itself is shown to be sufficient for this stabilization to occur, and the extracellular signal-regulated kinase (ERK) pathway is critically involved (20). We propose that Nur77(1-117) may stabilize HIF-1␣ in a similar fashion via ERK and MDM2.
A case could also be made that the phenotypes observed in classical Nur77-KO mice are a less severe form of the rapidonset AML phenotype observed in Nur77 and NOR-1 doubleknockout mice, which display splenomegaly, skewing of HSPCs toward myeloid progeny, and severe bone marrow defects within 21 days (7). We observed that classical Nur77-KO bone marrow gives rise to more macrophages than WT or complete Nur77-KO mouse bone marrow (Fig. S1D). We also occasionally observe severe splenomegaly in older classical Nur77-KO mice, with spleens of up to 2 g in 8.5-month-old mice (data not shown). Finally, the hematopoietic phenotypes in classical Nur77-KO mice described here only develop with old age and are seemingly absent in younger mice. It could therefore be speculated that the proposed imbalance in bone marrow Nur77

Nur77 variants regulate HIF-1␣ and bone marrow homeostasis
and HIF-1 signaling does not necessarily trigger the phenotypes by itself, but instead lowers the threshold for exaggerated reactions to other stimuli that can induce HSPC mobilization and proliferation, such as pathogenic infections or aging itself (29). Whereas we primarily focus on differences in hematopoietic biology between classical Nur77-KO and complete Nur77-KO mice and the regulation of HIF-1␣ signaling, the N-terminal domain of Nur77 interacts with a variety of other transcription factors, coregulators, and kinases (5). It may therefore be worthwhile to revisit other previously reported phenotypes of classical Nur77-KO mice. In that regard, we already show here that both Nur77-KO mouse models display decreased circulating Ly6C lo monocytes (Fig. S3A), which seems to be dependent on Nur77 transcriptional activity.
For this study, the existence of a microarray probe specific for TR3␤ mRNA and a plethora of publicly available expression data sets were crucial for determining the differences in tissue and tumor expression patterns between full-length Nur77 and TR3␤, especially because a natural form of TR3␤ does not exist in mice. These observations lend credence to the idea that the human TR3␤ transcript variant exists and that it has a biological function distinct from that of full-length Nur77. Moreover, our analyses suggest a role for TR3␤ in AML and KRAS signaling (Fig. 5, B and D). Future research could therefore focus on a possible interplay between TR3␤, HIF-1␣, and KRAS signaling and its functional consequences in AML, especially because cross-talk between HIF-1␣ and KRAS signaling has previously been reported to play a role in colon cancer (32). We also show that TR3␤ can reduce the transcriptional activity of full-length Nur77 (Fig. 4G). It could therefore be speculated that high expression of this transcript variant may inhibit the tumorsuppressive and homeostasis-promoting functions of fulllength Nur77, thereby predisposing bone marrow cells to pathologies such as AML.

Mice
All animal breeding, care, and procedures were approved by the institutional animal ethics committee of the University of Amsterdam (reference no. DBC102827), in accordance with both institutional and European directive 2010/ 63/EU guidelines.
B6;129S2-Nr4a1 tm1Jmi /J mice were purchased from the Jackson Laboratory (JAX stock no. 006187). These transgenic mice were originally generated by insertion of a phosphoglycerate kinase promoter-driven neomycin resistance (Neo R ) cassette into the coding region of exon 2 of the Nr4a1 gene (9). These mice are referred to as classical Nur77-KO mice in this work.
Mice in which exons 2-4 of the Nr4a1 locus are flanked by loxP sites (Nur77 fl/fl ) were generated in collaboration with H. Ichinose (Tokyo Institute of Technology) and were described elsewhere (33). C57Bl6/J mice overexpressing Cre recombinase under control of the human cytomegalovirus minimal promoter (CMV-cre mice) were purchased from the Jackson Laboratory (stock no. 006054). Nur77 fl/fl mice and CMV-cre mice were crossed to generate CMV-cre:Nur77 fl/fl mice. These mice were subsequently bred to lose the CMV-cre allele while remaining deficient for Nur77. These mice are referred to as complete Nur77-KO mice in this work.
For all experiments, male 6 -8-week-old, 3-month-old, or 6.5-month-old mice were used. Mice were euthanized by intraperitoneal injection of a lethal dose of ketamine (238 mg/kg) plus xylazine (102 mg/kg) and subsequent orbital bleeding. Before tissue harvest, mice were perfused with ice-cold PBS.
All animal studies have been performed and analyzed in a blinded manner

BMDM generation and culture
Bone marrow cells were flushed from femurs and tibias of 6 -8-week-old WT, classical Nur77-KO, and complete Nur77-KO mice and differentiated to macrophages by culturing in RPMI medium supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM L-glutamine, and 15% L929-conditioned medium for 7 days. In all experiments with LPS stimulation, a final concentration of 100 ng/ml LPS (Fluka) was used.

Cell line culture
The RAW264.7 mouse macrophage cell line was cultured in Dulbecco's modified Eagle's medium high-glucose medium with 10% FCS, 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine (all from Gibco). HEK293T and HeLa human cell lines were cultured in Dulbecco's modified Eagle's medium high-glucose medium with 10% FCS, 100 units/ml penicillin, and 100 g/ml streptomycin. The THP-1 and K562 cell lines were cultured in RMPI 1640 medium with 10% FCS, 100 units/ml penicillin, and 100 g/ml streptomycin.

Reverse transcription-quantitative PCR
Total RNA from primary single cell suspensions or cultured cells was isolated using TRIreagent (Sigma) according to the manufacturer's instructions. RT-PCR was performed on 1 g of RNA using the iScript cDNA synthesis kit (Bio-Rad). Quantita-

Nur77 variants regulate HIF-1␣ and bone marrow homeostasis
tive PCR was carried out using the SensiFAST SYBR No-ROX kit (Bioline) on a LightCycler 480 II PCR platform (Roche Applied Science). Quantification cycles, primer set amplification efficiencies, and transcript starting concentrations were calculated using the LinRegPCR software package (34). Target gene expression was normalized by dividing by the geometric mean of the gene expression of Actb and Rplp0 for mice or ACTB and RPLP0 for humans. Sequences of the primers used for RT-quantitative PCR are listed in Table S1.

Cloning and vector construction
The pCMV-myc-Nur77 vector has been described previously (18). For the pCMV-myc-Nur77(1-117) vector, the N-terminal domain of Nur77 (residues 1-115 plus 2 residues from the Neo R cassette) was amplified from genomic DNA of classical Nur77-KO mice by PCR using the primers 5Ј-CCCCT-GTATTCAAGCTCAATATGGA-3Ј (forward) and 5Ј-TTA-ATTCGATGGGTAGCAGCCGTA-3Ј (reverse) and a PCR program of 30-s denaturing at 94°C, 30-s annealing at 63°C, and 30-s extension at 72°C for 35 cycles. The amplification product was subsequently ligated into the pGEM-T Easy vector (Promega) and transferred into the pCMV-myc vector (Clontech) using EcoRI digestion to generate the construct. For the pCMV-myc-TR3␤ vector, the TR3␤ coding domain sequence was amplified from genomic DNA of human SMCs by PCR using the primers 5Ј-CTAGCGAATTCCACCCTGTATC-CAAGCCCAAT-3Ј (forward) and 5Ј-GATCGCTCGAGG-ACAGGACTTAGCTAGGAGGAGAA-3Ј (reverse), with a touch-down PCR program of 30-s denaturing at 95°C, 45-s annealing starting at 67°C and decreasing by 1°C per cycle to a minimum of 55°C, followed by a 2 min 30-s extension at 68°C for a total of 32 cycles. The amplification product was digested with EcoRI and XhoI and ligated into the pCMV-myc vector.

Luciferase assays
RAW264.7, HeLa, or HEK293T cells were transiently transfected with 4xNurRE-luc or 5xHRE-luc firefly luciferase reporter constructs and the pRluc-N3 Renilla luciferase reporter construct as an internal control using Lipofectamine LTX (RAW264.7) or Lipofectamine 2000 (HeLa and HEK293T) according to the manufacturer's instructions. After 24 h, cells were stimulated with 100 M cobalt chloride (CoCl 2 ), 100 M deferoxamine (DFO) or incubated at 1% O 2 in a H35 Hypoxystation (Don Whitley Scientific) for the times indicated. Lucif-erase assays were performed using the Dual-Luciferase reporter assay system on a Glomax multidetection system (both from Promega) according to the manufacturer's instructions.

Nur77 immunoprecipitations
For detection of endogenous Nur77 proteins, BMDMs were lysed in IP lysis buffer (50 mM Tris-HCl (pH 7.4), 100 mM NaCl, 10% (w/v) glycerol, 1% (w/v) IGEPAL CA-630) supplemented with phosphatase inhibitors (10 mM NaF, 1 mM Na 3 VO 4 ) and complete protease inhibitor mixture (Roche Applied Science). Lysates were precleared for 1 h at 4°C with protein A/Gagarose beads (Alpha Diagnostic), followed by incubation overnight with 2 g of rabbit-anti-Nur77 antibody (Santa Cruz Biosciences catalog no. sc-5569). The next day, complexes were precipitated using protein A/G-agarose beads, washed three times in IP lysis buffer, and eluted by boiling in Laemmli sample buffer. Immunoprecipitated Nur77 protein was subsequently detected by Western blotting.

Western blotting
Protein samples were separated on 10 or 12% SDS-polyacrylamide gels and transferred onto nitrocellulose membranes (Bio-Rad). Membranes were blocked using 5% nonfat milk for 1 h and subsequently incubated with rabbit anti-HIF-1␣ (Novus Biologicals catalog no. NB100-449) or mouse-anti-Nur77 (Pharmingen catalog no. 554088) primary antibodies overnight. Protein bands were detected by incubating membranes with appropriate HRP-conjugated secondary antibodies for 1 h at room temperature, followed by visualization using Supersignal West Femto ECL (Thermo Scientific) and detection using an ImageQuant LAS 4000 imager (GE Healthcare). Equal loading of samples was confirmed by reprobing blots with antibodies against mouse-anti-tubulin (Cedarlane catalog no. CLT9003) or rabbit anti-␤-actin (Cell Signaling catalog no. 4970).

Colony-forming unit assay
Mouse bone marrow and splenic single cell suspensions were seeded in MethoCult GF M3434 medium (STEMCELL Technologies) according to the manufacturer's instructions. For each mouse, 30,000 total bone marrow cells or 200,000 total splenic cells were seeded in duplicate wells. The total number of myeloid colonies per mouse was counted using an inverted light microscope, on day 8 for bone marrow and day 12 for spleen cells.

Data analysis
Statistical analyses were performed using GraphPad Prism software. Normal distribution of data was tested with the Kolmogorov-Smirnov normality test. Data are presented as mean Ϯ S.E. p values were calculated using Student's t test, one-way ANOVA, or two-way ANOVA with either Bonferroni or Tukey post hoc correction as indicated in the figure legends. A p value Ͻ0.05 was considered statistically significant. Flow cytometry data were analyzed using FlowJo software. Metaanalysis of publicly available microarray data sets using the Affymetrix Human Genome U133 Plus 2.0 Array was performed using NEBION Genevestigator software (28).