Pharmacological Manipulation of the RAR/RXR Signaling Pathway Maintains the Repopulating Capacity of Hematopoietic Stem Cells in Culture

Abstract

The retinoid X receptor (RXR) contributes to the regulation of diverse biological pathways via its role as a heterodimeric partner of several nuclear receptors. However, RXR has no established role in the regulation of hematopoietic stem cell (HSC) fate. In this study, we sought to determine whether direct modulation of RXR signaling could impact human HSC self-renewal or differentiation. Treatment of human CD34⁺CD38⁻lin⁻ cells with LG1506, a selective RXR modulator, inhibited the differentiation of HSCs in culture and maintained long-term repopulating HSCs in culture that were otherwise lost in response to cytokine treatment. Further studies revealed that LG1506 had a distinct mechanism of action in that it facilitated the recruitment of corepressors to the retinoic acid receptor (RAR)/RXR complex at target gene promoters, suggesting that this molecule was functioning as an inverse agonist in the context of this heterodimer. Interestingly, using combinatorial peptide phage display, we identified unique surfaces presented on RXR when occupied by LG1506 and demonstrated that other modulators that exhibited these properties functioned similarly at both a mechanistic and biological level. These data indicate that the RAR/RXR heterodimer is a critical regulator of human HSC differentiation, and pharmacological modulation of RXR signaling prevents the loss of human HSCs that otherwise occurs in short-term culture.

The RAR/RXR heterodimer is a critical regulator of human HSC differentiation and pharmacologic modulation of RXR signaling sustains human HSCs in culture despite cytokine-induced proliferation.

Characterization of the intrinsic and extrinsic pathways that regulate hematopoietic stem cell (HSC) self-renewal and differentiation continues to evolve. Several pathways that regulate HSC fate determinations have recently been identified (1,2,3,4), and overexpression of Notch, HoxB4, and β-catenin in murine HSCs has been associated, in each case, with enhanced self-renewal capacity (1,2,3). Clinical methods to expand human HSCs based upon these discoveries are currently being tested (4,5). However, gene silencing studies have demonstrated that none of these pathways are required for HSC maintenance or reconstitution in vivo (6,7,8). These data suggest that HSC fate determinations are governed by a complexity of signals and provide impetus for further studies to identify additional HSC regulatory pathways.

Human HSCs are known to be maximally enriched within the CD34⁺CD38⁻lin⁻ subset, with a frequency of one long-term repopulating stem cell per 617 CD34⁺CD38⁻lin⁻ cells (9). Several studies have shown that ex vivo culture of human CD34⁺CD38⁻lin⁻ HSC-enriched populations with proliferation-inducing cytokines results in the predictable loss of HSCs within 7–14 d of culture (10,11,12,13,14). Recently, we demonstrated that the enzyme aldehyde dehydrogenase 1 (ALDH1), which is highly expressed in human HSCs (15,16), is an intrinsic regulator of HSC differentiation (17). Pharmacological inhibition of ALDH1 with diethylaminobenzaldehyde in short-term culture of human cord blood and bone marrow (BM) CD34⁺CD38⁻lin⁻ HSCs inhibited stem cell differentiation in vitro and promoted the expansion of primitive cells capable of repopulating nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice [SCID-repopulating cells (SRCs)] (17). Because ALDH1 is required for the intracellular conversion of retinaldehydes to retinoic acids, we hypothesized that inhibition of ALDH1 activity promoted HSC self-renewal via blockade of retinoid signaling in HSCs. As preliminary evidence to support this, we also demonstrated that the expression of CCAAT/enhancer binding protein-ε, a retinoic acid receptor (RAR)-dependent transcription factor, was down-modulated in HSCs in the presence of diethylaminobenzaldehyde (17).

To determine whether retinoid signaling plays a primary role in human HSC self-renewal and differentiation, we tested here whether direct pharmacological modulation of retinoid and rexinoid signaling could alter the self-renewal and differentiation potential of human HSCs. Activation of RARα with all-trans retinoic acid (ATRA) has been shown to inhibit the proliferation of both human embryonic hematopoietic progenitor cells and adult cobblestone-area forming cells in culture and promote the apoptosis of human CD34⁺ cells (18,19,20). Conversely, ATRA stimulates the proliferation of myeloid progenitors [colony-forming unit (CFU)-granulocyte-macrophage (GM)] (17) and induces the granulocytic differentiation of myeloid progenitors (21), and introduction of a dominant-negative RARα construct into a hematopoietic progenitor cell line suppresses neutrophil and monocyte development (22). It has also been shown that RARα is not required for granulocytic differentiation to occur (23). Interestingly, activation of RARα has been associated with enhanced maintenance of murine HSCs in vitro (24), and silencing of RARγ has been associated with a reduction of HSC content in vivo (25). Taken together, these data reveal a complex role for RAR in hematopoiesis. In contrast, the contribution of retinoid X receptor (RXR) signaling in hematopoiesis is less well defined (26,27). Treatment of murine or human BM progenitor cells with 9-cis-retinoic acid, an RXR agonist (rexinoid), stimulates myeloid progenitor proliferation although inhibiting erythroid progenitor cell production (18,26,27). Conditional knockout of RXRα expression in hematopoietic cells in mice was not associated with any alteration in BM cellularity or phenotype, suggesting that RXRα is dispensable for normal hematopoiesis and may be compensated for by up-regulation of RXRβ (28). To date, the contribution of RXR signaling in human HSC self-renewal and differentiation has not been established. Because RXR can function as a homodimer or as a heterodimer with RAR and other nuclear receptors (NR), it is unclear whether the previously described effects of rexinoids and retinoids on hematopoietic function occurred through the same pathway or whether they involve the regulation of parallel pathways. In this study, we demonstrate that pharmacological modulation of RXR signaling inhibits the differentiation of human HSCs and sustains hematopoietic repopulating cells in culture that are otherwise lost in response to cytokines. Furthermore, these studies demonstrate that the RXR-RAR heterodimer is a unique target for small molecules that facilitate the amplification of human HSCs.

Results

Modulation of RXR signaling inhibits the differentiation of human HSCs in culture

We recently showed that pharmacological inhibition of the enzyme ALDH1 impedes the differentiation of human HSCs in culture and promotes the amplification of primitive SRCs in response to cytokine treatment (17). Because ALDH1 is required for the intracellular production of retinoic acids, we sought to determine whether direct modulation of retinoid signaling could result in comparable effects. Human BM CD34⁺CD38⁻lin⁻ HSC-enriched cells were placed in culture with cytokines alone [20 ng/ml thrombopoietin, 120 ng/ml stem cell factor, and 50 ng/ml fms-like tyrosine kinase 3 ligand (Flt-3 ligand) (TSF)] or TSF plus ATRA, a canonical agonist of RAR. Treatment with 1 μm ATRA plus TSF caused an increase in total cells at d 7 compared with TSF alone (12.7- vs. 6.1-fold, P = 0.005) but was associated with the accelerated differentiation of HSCs as evidenced by a significant decline in the percentage of CD34⁺ progenitors (mean 24.8 vs. 82.6%, P = 2.8 × 10⁻⁶; Table 1) and a complete loss of CD34⁺CD38⁻ cells at d 7 compared with TSF alone (P = 0.001; Fig. 1A). Although the addition of ATRA to TSF did not cause an increase in total colony-forming cell production (CFU-total) compared with TSF alone, ATRA treatment caused a complete shift to myeloid differentiation (CFU-GM) in the absence of burst-forming units-erythroid (BFU-E) or CFU-mix production (data not shown). Using the same assay, we screened a series of chemically distinct RXR, RAR, and peroxisome proliferator-activated receptor (PPAR) modulators for their ability to inhibit HSC differentiation. In this manner, we found a single compound, LG1506, a selective RXR modulator (29,30), that inhibited HSC differentiation in culture. Culture of CD34⁺CD38⁻lin⁻ cells for 7 d with LG1506 plus TSF supported a 4.3-fold expansion of total cells (n =3; P = 0.1 vs. TSF alone), with a significantly increased percentage of CD34⁺CD38⁻ cells in culture compared with TSF alone (mean 86.4 vs. 34.1%, P = 1.4 × 10⁻⁵; Table 1). This led to a significant increase in the number of CD34⁺CD38⁻ cells in culture at d 7 in the LG1506-treated cultures as compared with TSF alone (3.7- vs. 2.1-fold, P = 0.02; Table 1). Treatment with LG1506 plus TSF also significantly reduced colony-forming cell (CFC) production from HSCs in culture as compared with TSF alone (Fig. 1B). The progeny of TSF alone contained 104.0 ± 11.1 CFU-total (× 10³) compared with 65.3 ± 15.1 CFU-total in the TSF- plus LG1506-treated group (Fig. 1B; P = 0.01). These data suggested that treatment with LG1506 inhibited HSC differentiation in culture in response to cytokines. Table 1 summarizes the effects of treatment with TSF alone, ATRA, and LG1506 on the phenotype and CFC content of human CD34⁺CD38⁻lin⁻ cells in vitro. Of note, treatment of BM CD34⁺CD38⁻lin⁻ cells with 1 μm LG1506 alone (without TSF) yielded no viable cells at d +7, similar to the results with media alone (data not shown).

Table 1.

Effects of NR ligands on human BM progenitor cell expansion

Condition	Total cells (×103)	% CD34+	Total CD34+ (×103)	% CD34+CD38−	Total CD34+CD38− (×103)	CFU-total/103 cells
d 0	4.0 ± 0	100.0 ± 0.0	4.0 ± 0.0	99.3 ± 0.0	4.0 ± 0.0	5.3 ± 1.5
d 7 TSF alone	24.3 ± 6.3	82.6 ± 1.3	20.2 ± 5.6	34.1 ± 2.7	8.21 ± 1.6	104.0 ± 11.1
d 7 TSF + ATRA	50.7 ± 7.1	24.8 ± 1.9	12.5 ± 1.1	0.0 ± 0.0	0.0 ± 0.0	101.4 ± 22.0
d 7 TSF + LG1506	17.2 ± 4.0	86.4 ± 3.2	14.7 ± 2.9	86.4 ± 3.1	14.7 ± 3.0	65.3 ± 15.1
d 7 TSF + Ro41-5253	17.2 ± 0.8	85.0 ± 2.8	14.6 ± 0.3	83.2 ± 2.6	14.3 ± 0.3	66.7 ± 8.3
d 7 TSF + rosiglitazone	8.0 ± 2.8	59.7 ± 2.5	4.8 ± 1.8	5.8 ± 1.0	0.5 ± 0.2	137.4 ± 19.0

Primary human BM CD34⁺CD38⁻lin⁻ cells (4 × 10³) were placed in culture with TSF with or without the addition of ATRA, LG1506, Ro41-5253, or rosiglitazone as described in Materials and Methods. Three to six replicate experiments were performed for each condition. 

Figure 1.

Antagonism of RXR maintains CD34⁺CD38⁻ cells and inhibits the production of committed CFCs from HSCs in culture. A, Representative FACS analyses of CD34 and CD38 expression are shown on human BM CD34⁺CD38⁻lin⁻ cells at d 0 and after culture with cytokines alone (TSF), TSF plus ATRA, TSF plus LG1506, TSF plus Ro41-5253 and TSF plus rosiglitazone. The addition of LG1506 or Ro41-5253 to cultures yielded high percentages of CD34⁺CD38⁻ cells compared with the other conditions. B, Human BM CD34⁺CD38⁻lin⁻ cells at d 0 or their progeny after culture with TSF alone, TSF plus ATRA, TSF plus LG1506, TSF plus Ro41-5253, or TSF plus rosiglitazone were placed in 14-d methylcellulose cultures to measure CFC content. As shown, d-0 FACS-sorted BM CD34⁺CD38⁻lin⁻ cells had little CFC content, whereas the progeny of this population after 7 d culture with TSF alone or TSF plus ATRA contained increased CFU-total compared with input. Conversely, treatment with either LG1506 or Ro41-5253 resulted in significant decreases in CFC production compared with TSF alone (*, P = 0.01 and P = 0.01, respectively). Treatment with rosiglitazone caused a significant increase in CFCs compared with TSF alone (^, P = 0.02). The gray, black, and white bars represent CFU-GM, CFU-mix, and BFU-E colonies, respectively.

To identify potential heterodimeric partners of RXR through which LG1506 was exerting these effects on HSCs (17,29,30), we also tested the effects of an RAR antagonist, Ro41-5253, and the PPARγ agonist rosiglitazone on the differentiation of primary HSCs in culture. The latter compound was included in the analysis because it has been demonstrated that LG1506, through its ability to modulate the RXR/PPARγ heterodimer, functioned as an insulin sensitizer in vivo (30). The addition of 1 μm Ro41-5253 to TSF increased the percentage of CD34⁺CD38⁻ cells at d 7 compared with TSF alone (mean 83.2 vs. 34.1%, respectively; P = 1.1 × 10⁻⁵; Fig. 1A) and inhibited CFC production compared with TSF alone (Fig. 1B and Table 1; P = 0.01). Conversely, treatment of HSCs with rosiglitazone plus TSF had no apparent effect on HSC phenotype at d 7 and promoted an increase in CFC production compared with TSF alone (Fig. 1B and Table 1; P = 0.02). These data indicated that activation of PPARγ did not account for the effects of LG1506 on human HSCs. Treatment of human CD34⁺CD38⁻lin⁻ cells with TSF plus clofibrate, a PPARα agonist, or carbaprostacyclin, a PPARδ agonist, also did not result in any significant expansion of total cells, CD34⁺ cells, or CD34⁺CD38⁻ cells compared with TSF alone (data not shown). Taken together, these data indicated that treatment of BM HSCs with either an RXR modulator, LG1506, or an RAR antagonist, Ro41-5253, produced comparable inhibitory effects on HSC differentiation.

Treatment with LG1506 sustains human SRCs in liquid suspension culture

Because treatment with LG1506 appeared to inhibit the differentiation and lineage commitment of BM CD34⁺CD38⁻lin⁻ cells in culture, we sought to determine whether treatment with LG1506 preserved long-term repopulating HSCs in culture. We transplanted a series of NOD/SCID mice with limiting doses of unmanipulated d-0 human BM CD34⁺CD38⁻lin⁻ cells or the progeny of the identical dose of BM CD34⁺CD38⁻lin⁻ cells after treatment with TSF alone or LG1506 plus TSF. BM collections from five healthy adult donors (Division of Cellular Therapy, Duke University, and Lonza, Walkersville, MD) were used as the source of BM for these studies. A total of 163 mice were transplanted with human BM CD34⁺CD38⁻lin⁻ cells or their progeny after 7 d culture. Over a range of doses from 1 × 10³ to 10 × 10³ cells transplanted per mouse, 15 of 59 (25%) mice transplanted with d-0 BM CD34⁺CD38⁻lin⁻ cells demonstrated human CD45⁺ cell engraftment at 8 wk (Fig. 2 and Table 2). Of note, the overall levels of human CD45⁺ cell engraftment in this study were lower than we have observed in previous studies (17). This is due to the use of BM cells in this study, which have one third the SRC content of cord blood cells (31) and perhaps also due to the immunomagnetic method used to lineage-deplete HSCs as opposed to FACS-based lineage (−) cell sorting. Mice transplanted with the progeny of these doses of BM CD34⁺CD38⁻lin⁻ cells after 7 d culture with cytokines alone (TSF) demonstrated human CD45⁺ cell engraftment in only one of 53 mice (2%) (Fig. 2, Table 2). Conversely, mice transplanted with the progeny of the identical doses of BM CD34⁺CD38⁻lin⁻ cells after culture with TSF plus LG1506 demonstrated engraftment in 15 of 58 mice (26%). Poisson statistical analysis demonstrated that the SRC frequency within the d-0 BM CD34⁺CD38⁻lin⁻ cells was one in 9435 [95% confidence interval (CI) = 1/5,737 to 1/15,516). Conversely, the SRC frequency within the progeny of TSF-cultured BM CD34⁺CD38⁻lin⁻ cells was one in 192,457 (CI = 1/27,412 to 1/1,351,210), indicating a pronounced decline in SRC content during culture with TSF alone. The SRC frequency within the progeny of BM CD34⁺CD38⁻lin⁻ cells after culture with TSF plus LG1506 was one in 13,481 (CI = 1/8,138 to 1/22,332) (Table 2). Taken together, these data demonstrate that manipulation of RXR signaling with LG1506 largely maintained human HSCs in cytokine culture, whereas HSCs otherwise dramatically declined in number in response to cytokine treatment (14-fold decline in TSF cultures compared with TSF plus LG1506). Detailed flow cytometric analysis revealed the presence of CD34⁺ progenitor cells, CD19⁺ B lymphoid cells, and CD33⁺ myeloid cells in mice transplanted with LG1506- plus TSF-cultured cells, demonstrating that a multipotent progenitor cell was sustained during culture with LG1506 (Table 3). The proportion of CD34⁺, CD19⁺ lymphoid, and CD33⁺ myeloid cells detectable at 8 wk was highly comparable between mice transplanted with LG1506-treated progeny and d-0 BM CD34⁺CD38⁻lin⁻ cells, indicating that LG1506 treatment did not alter the normal differentiation capacity of HSCs in culture (Table 3).

Figure 2.

Treatment with LG1506 preserves SRCs in culture. The scatter plot represents human CD45⁺ cell engraftment in the BM of NOD/SCID mice at 8 wk after transplant (n =163 mice total). Human BM CD34⁺CD38⁻lin⁻ cells (1 × 10³ to 10 × 10³) or their progeny after 7 d culture with TSF alone or TSF plus LG1506 were transplanted via tail vein infusion into sublethally irradiated NOD/SCID mice. The progeny of TSF plus LG1506 cultures contained 14-fold higher frequency of SRCs compared with the progeny of TSF alone. Mean human CD45⁺ cell engraftment levels in each group are represented by horizontal bars.

Table 2.

Human BM cell engraftment in NOD/SCID mice and SRC frequency

BM source	Cell dose	No. engrafted	SRC estimate	95% CI
d 0	1,000	0 of 10	1 in 9,435	1/5,737 to 1/15,516
BM CD34+CD38−lin−	2,500	0 of 17
	5,000	7 of 13
	10,000	8 of 9
d 7 TSF	1,000	0 of 10	1 in 192,457	1/27,412 to 1/1,351,210
	2,500	0 of 23
	5,000	0 of 14
	10,000	1 of 6
d 7 TSF + LG1506	1,000	0 of 10	1 in 13,481	1/8,138 to 1/22,332
	2,500	2 of 18
	5,000	6 of 19
	10,000	7 of 11

Limiting dilutions (1 × 10³ to 10 × 10³) of human BM CD34⁺CD38⁻lin⁻ cells or their progeny after 7 d culture with TSF alone or TSF plus LG1506 were transplanted into sublethally irradiated NOD/SCID mice as described in Materials and Methods. At 8 wk after transplant, mice were killed, and human CD45⁺ cell engraftment was measured in the BM of all mice. Positive engraftment was defined as at least 0.1% human CD45⁺ cells in the mouse BM, as previously described (41). 

Table 3.

Multilineage engraftment of human hematopoietic cells in NOD/SCID mice

Culture condition	Percentage of engrafted human CD45+ cell expressing:
	CD34	CD19	CD33
d 0 BM CD34+CD38−lin− cells	8.9 ± 5.0	93.8 ± 3.3	7.1 ± 6.8
d 7 LG1506 + TSF	5.5 ± 5.5	91.8 ± 3.8	7.7 ± 2.7

NOD/SCID mice received transplants from 2.5 × 10³ to 5 × 10³ BM CD34⁺CD38⁻lin⁻ cells or their progeny after 7 d culture with LG1506 plus TSF, as described in Materials and Methods. The numbers shown under the CD34, CD19, and CD33 columns indicate the mean percentage of engrafted human CD45⁺ cell that expressed that particular differentiation antigen (n = 5 mice analyzed per group). 

LG1506 inhibits cell cycle progression of HSCs

Because LG1506 preferentially maintained human BM HSCs in culture, we examined potential mechanisms through which this might have occurred. First, we examined whether LG1506 altered cell cycle progression of HSCs in response to cytokines because cell cycle progression has been associated with a loss of long-term repopulating HSCs in culture (11). Day-zero BM CD34⁺CD38⁻lin⁻ cells were primarily in G₀ (mean 97%; Fig. 3A), whereas only 5.0% of CD34⁺CD38⁻ cells remained in G₀ after 3 d culture with TSF alone. BM CD34⁺CD38⁻ cells in culture with TSF plus LG1506 had a significantly higher percentage of cells in G₀ at d +3 (P = 0.0002) and d +7 (P = 0.006) compared with the progeny of TSF alone (Fig. 3A).

Figure 3.

Treatment with LG1506 inhibits cell cycle progression of human HSCs in culture but has no effect on apoptosis or lineage differentiation. A, Bar graphs are shown representing the mean levels of BM CD34⁺CD38⁻lin⁻ cells in G₀, G₁, or G₂/S/M phase at d 0, +3, and +7 after culture with either TSF alone or TSF plus LG1506 (n = 3 experiments). *, P = 0.0002 and P = 0.006 for difference in percentage of cells in G₀ between TSF and TSF plus LG1506 group at d +3 and +7, respectively; ^, P = 0.002 for difference in percentage of cells in G₁ between TSF and TSF plus LG1506 groups. B, Bar graphs represent mean percentages of apoptotic cells (annexin V⁺7AAD⁻) and necrotic cells (annexin V⁺7AAD⁺) within BM CD34⁺CD38⁻lin⁻ cells at d 0, +3, and +7 during culture with TSF alone or TSF plus LG1506. C, The mean percent expression of the differentiation antigens CD3, CD19, CD33, and CD13 on CD34⁺CD38⁻ cells at d 0 and at d +3 and +7 of culture with TSF (black bars) and TSF plus LG1506 (gray bars). *, P = 0.002 for difference between the percent expression of CD33 on CD34⁺CD38⁻ cells in culture with TSF vs. culture with TSF plus LG1506.

We also examined whether treatment with LG1506 prevented HSC cell death during culture using annexin V/7AAD analysis. We found no significant differences in the percentage of apoptotic cells (annexin V⁺7AAD⁻) or necrotic cells (annexin V⁺7AAD⁺) in cultures with TSF plus LG1506 as compared with TSF alone (Fig. 3B). Lastly, we examined whether differences in lineage differentiation markers could be observed between the CD34⁺CD38⁻ cells remaining in culture with TSF plus LG1506 as compared with cultures with TSF alone. Few of the CD34⁺CD38⁻ cells in either condition expressed B lymphoid (CD19) of T cell (CD3) markers, and the majority (>90%) of CD34⁺CD38⁻ cells in either TSF cultures or TSF plus LG1506 cultures expressed the myeloid differentiation antigen CD13 (Fig. 3C). A small, but significant increase in CD33 expression was noted on CD34⁺CD38⁻ cells in TSF plus LG1506 cultures compared with TSF alone (P = 0.002). The broad expression of the myeloid differentiation antigen CD13 on CD34⁺CD38⁻ cells in culture emphasizes that the CD34⁺CD38⁻ phenotype may not be indicative of HSC content after culture.

LG1506 functions as an antagonist in the context of an RXR-RAR heterodimer

Because treatment with the RXR-specific ligand LG1506 prevented the differentiation of HSCs in culture in response to cytokines, we sought to determine which heterodimeric partner of RXR was involved in this process. We hypothesized that the RXR-RAR heterodimer was the target based upon the comparable effects that LG1506 and Ro41-5253 had on HSC phenotype and CFC production in vitro. To test this hypothesis, we assessed the ability of LG1506 to inhibit RAR transcriptional activity using transient transfection assays. Specifically, the classical RAR synthetic response element luciferase reporter construct (3xDR5 tk-luc) and RARα expression vectors were transfected into HepG2 cells in the absence or presence of different ligands. As expected, increasing concentrations of ATRA resulted in strong activation of 3xDR5 luciferase construct (Fig. 4A). Interestingly, both LG1506 and the RARα antagonist Ro41-5253, were able to repress the basal activity and inhibit ATRA-mediated transcriptional activity. Moreover, increasing concentrations of the RAR agonist ATRA reversed the repression manifested by these compounds. To verify that this effect was not limited specifically to the simple DR5 response element, we repeated this experiment using a known RAR-regulated promoter, that of the human RARβ gene, and obtained similar results (Fig. 4B). RAR has previously been shown to bind to hormone response elements that contain a palindrome repeat of the sequence AGGTCA (HREpal). LG1506 and Ro41-5253 were also able to repress the basal and ATRA-mediated transcriptional activity of this promoter (Fig. 4C). However, LG815, a compound that manifests RAR partial agonist activity, was unable to repress the basal activity on any of the tested promoters (Fig. 4, A–C). Collectively, these data suggest that inhibition of RAR transcriptional activity could be achieved by targeting either RAR or RXR and that RXR is not a silent partner in the RAR-RXR heterodimer. Interestingly, in these assays, LG1506 is functioning as an inverse agonist that, through its actions on RXR, is able to inhibit the activity of RAR below that observed in the absence of added ligand.

Figure 4.

LG1506 and Ro41-5253 function analogously to repress basal and retinoic acid-induced transcriptional activity. A, HepG2 cells were cotransfected with an RARα expression vector and a 5xDR5-tk-luciferase reporter construct. Twenty hours after transfection, HepG2 cells were treated with either LG1506 or Ro41-5253 in the presence and absence of the indicated concentrations of ATRA. Both LG1506 and the RARα antagonist Ro41-5253 were able to repress the basal activity and inhibit ATRA-mediated transcriptional activity. B, LG1506 and Ro41-5253 also repressed basal activity and ATRA-mediated transcriptional activity in HepG2 cells transfected with an RARα promoter-luciferase reporter construct. C, LG1506 and Ro41-5253 induced comparable repression of basal and ATRA-induced activity in HepG2 cells transfected with the HREpal-tk-luciferase reporter construct. All ligands were used at 1 μm concentration for these experiments.

To address the mechanism through which LG1506 achieves this inverse agonist effect on RAR signaling, we used the HL-60 human myeloid leukemia cell line, which has been used extensively as a model to study the differentiating effects of retinoic acid. In this cell line, differentiation in response to retinoic acid appears to be dependent on the expression of RARα receptor (32). We asked whether 1 μm LG1506 could block endogenous retinoic acid-induced gene expression in HL-60 promyelocytic leukemia cells. RARα has been shown to bind the promoter and activate transcription of the CD38 gene. Furthermore, the RARα antagonist Ro41-5253 completely suppresses the retinoic acid-induced expression of CD38 mRNA transcript and protein in HL-60 cells (33,34). As expected, HL-60 cells treated with ATRA induced the expression of CD38 mRNA in a concentration-dependent fashion (Fig. 5A). In this context, LG1506 as well as 1 μm Ro41-5253 were able to down-regulate the basal expression and to suppress retinoic acid-induced expression of CD38 mRNA (Fig. 5A). LG815 (1 μm) appeared to activate the expression of CD38, confirming the partial RAR agonist activity of this compound. In a similar manner, LG1506 was able to repress retinoic acid-induced expression of RARβ (data not shown). Using CD11b as a myelomonocytic differentiation marker (35), we demonstrated that LG1506 inhibited retinoic acid-mediated expression of CD11b (Fig. 5B). Importantly, LG1506 also suppressed CD38 mRNA expression in purified human CD34⁺CD38⁻lin⁻ HSCs during culture with TSF (Fig. 5C). These data suggest that inhibition of the basal endogenous (or culture-induced) RXR-RAR signaling by the specific RXR ligand LG1506 underlies its ability to maintain primitive hematopoietic progenitor cells.

Figure 5.

LG1506 and Ro41-5253 suppress basal and retinoic acid-induced gene expression. HL-60 cells were treated for 3 d with either 1 or 10 μm LG1506 or 1 μm Ro41-5253 or 1 μm LG815 in the presence and absence of the indicated concentration of ATRA. A, Analysis of their mRNA expression by real-time RT-PCR for CD38; B, analysis for expression of CD11b; C, analysis for mRNA expression of CD38 in primary CD34⁺CD38⁻lin⁻ cells at d 0 and in response to treatment with TSF alone, TSF plus ATRA, and TSF plus LG1506.

LG1506 induces a unique conformational change in RXRα

The above experiments imply that inhibition of RXR-RAR signaling impedes the differentiation of human HSCs that occurs in response to cytokines and that, in the presence of LG1506, RXR is actively involved in this inhibitory process. This active role of RXR in RAR antagonism has not been described previously. However, because of the importance of this activity with respect to HSC differentiation, we performed additional studies to determine how this compound distinguishes itself from other RXR ligands. In previous studies directed toward the estrogen receptor, we determined that the pharmacological activities of different estrogens and anti-estrogens were determined by their impact on receptor structure (36). Furthermore, it was shown that compounds that induced similar alterations in estrogen receptor structure usually have very similar pharmacological activities. Building on these observations, we reasoned that LG1506 must endow upon the receptor a specific structural alteration that determines its activity. To test this hypothesis, we purified recombinant RXRα and used it to capture M13 bacteriophage expressing short 19-mer peptides capable of interacting with the receptor in the presence of LG1506. The peptides identified in this manner were then used to develop a cell-based two-hybrid assay to survey RXRα conformation in the presence of different ligands. Specifically, full-length RXRα was expressed as a fusion protein with the VP16 acidic activation domain, and the peptides were produced as fusion proteins with the yeast Gal4-DBD. Using this assay, we were able to assess the interaction of the peptide probes with RXRα in the presence of different ligands by assaying their ability to recruit the RXRα-VP16 fusion to a 5xGAL4RE-Luc3 reporter construct. The results of this analysis, using four of the most informative peptides, are shown in Fig. 6. These results indicate that LG1506 allows the presentation of a surface in common with other RXR ligands (peptides RL101 and RL8) and at least one surface whose presentation in this assay is highly selective for this compound (peptides RL115 and RL135; Fig. 6A). Taken together, these data suggest that LG1506 distinguishes itself from other RXR ligands by virtue of its ability to alter RXRα conformation in a unique manner. If this specific conformation is that which determines the ability of RXR to inhibit retinoic acid signaling and therefore provides activity in HSC culture assays, then it stands to reason that other compounds that facilitate similar structural alterations in RXRα would function analogously.

Figure 6.

LG1506 induces a conformational change in RXRα that correlates with its inhibition of RXR-RAR signaling. A, LG1506 and LG1612 induce a closely related conformational change within the surface of RXRα. Affinity selection of LG1506-RXRα binding peptide was identified using phage display technology. The specific LG1506-RXRα peptide RL135 did not bind in the presence of other RXR ligands, LG268 (agonist) and LG208 (antagonist), or 9-cis-retinoic acid (putative endogenous ligand for RXR). However, RL135 did bind with high affinity in the presence of LG1612, which induced a closely related conformational change in RXRα to LG1506. B, LG1612 represses basal and retinoic acid-induced transcriptional activity in a manner similar to LG1506. HepG2 cells were cotransfected with an RARα expression vector and HREpal-tk-luciferase reporter construct. Twenty hours after transfection, HepG2 cells were treated with either LG1506 or LG1612 in the presence and absence of the indicated concentration of ATRA. C, LG1612 also suppresses the basal and retinoic acid-induced CD38 mRNA expression in HL-60 cells, analogous to LG1506. HL-60 cells were treated for 3 d with either LG1506 or LG1612 in the presence and absence of the indicated concentrations of ATRA and analyzed for mRNA expression by real-time RT-PCR. Note that all ligands used in these experiments were used at 1 μm concentration.

Using the cell-based two-hybrid assay, we screened for additional ligands that would allow the interaction of RXR with the LG1506-selective peptides. In this manner, we identified one additional compound, LG1612, that, within the limitations of the sensitivity of this assay, allowed RXRα to adopt a similar conformation to that observed in the presence of LG1506 (peptide RL135). Interestingly, LG1612, as previously shown for LG1506, was able to repress the basal activity and the retinoic acid-induced transcriptional activity on all the RAR-responsive reporter assays tested (Fig. 6, B and C). In contrast, neither LG208 (an RXR antagonist) nor LG268 (an RXR agonist), compounds that induced distinctly different RXR conformations, were able to repress RAR transcriptional activity or antagonize retinoic acid-induced gene expression in HL-60 cells (data not shown). Taken together, these results indicate that RXR is indeed the target of these inverse agonists in HSCs and that a firm relationship exists between structure and function of this receptor in HSCs.

LG1506 has a distinct mechanism of action

It was observed that LG1506 was able to inhibit RXR-RAR target gene transcription below the basal activity observed in the absence of an activating ligand. This suggested that rather than functioning as a classical antagonist, this compound may actually be an inverse agonist facilitating the interaction of corepressors with the apo-RXR-RAR heterodimer. To examine this hypothesis, we used a mammalian two-hybrid assay to assess the ability of full-length VP16-RARα to interact with Gal4-DBD-peptide fusions corresponding to the corepressor interaction domain within SMRT (SMRT ID) in the absence or presence of LG1506 or Ro41-5253 and increasing concentrations of ATRA. For comparative purposes, the interaction with a GAL4 fusion encoding the LXXLL-containing NR box from SRC1 was also evaluated in the same manner. This served as a probe that enabled us to assess the conformation of the AF-2 coactivator pocket. HepG2 cells express sufficient levels of RXRs for these experiments (Fig. 4, A–C, and data not shown). As expected, addition of increasing concentrations of ATRA resulted in enhanced RAR-SRC-1 NR box interaction. Moreover, addition of the RARα antagonist Ro41-5253 dramatically abolished the basal interaction of SRC-1 NR box with RARα and only at higher concentrations was ATRA able to induce SRC1 NR box recruitment to RAR. In this context, LG1506 has only a minor effect on SRC1 NR box-RAR interaction (Fig. 7A). To determine whether binding of LG1506 to RXR can promote the binding between RAR and corepressors, we examined the interaction of VP16-RARα with the Gal4-DBD SMRT ID. As shown in Fig. 7B, increasing concentrations of ATRA led to a dissociation of the SMRT ID from RARα, whereas addition of LG1506 enhanced the interaction of RAR-SMRT ID with apo-RAR although reducing the ability of ATRA to dissociate this complex. Interestingly, no increase in corepressor peptide binding to RARα was observed when it was occupied by Ro41-5253. Together, these data suggest that LG1506 (binding to RXR) represses RXR-RAR heterodimer transcriptional activity by promoting the association of the corepressors with this heterodimeric complex. On the other hand, although Ro41-5253 does not induce the RAR-SMRT interaction, it is able to inhibit transcription by disrupting the function of the AF-2 coactivator binding pocket within RAR. Thus, LG1506, through RXR, and Ro41-5253, by direct antagonism of RARα, can both repress RXR-RAR-mediated transcription via distinct mechanisms. Considering these functional data, we believe that it may be reasonable to use both drugs in combination to effect maximal HSC expansion in circumstances where a complete retinoid block is desired.

Figure 7.

LG1506 has a distinct mechanism of action compared with Ro41-5253. A, Ro41-5253 inhibits the interaction of RAR with an LXXLL-containing peptide. HepG2 cells were cotransfected with constructs expressing a VP16-RARα fusion protein, the Gal4-DBD-receptor interaction domain of SRC-1 (LXXLL-containing peptide), and a luciferase reporter with five Gal4-binding sites. Cells were cultured in the absence or presence of 1 μm LG1506 or 1 μm Ro41-5253 with increasing concentration of ATRA. B, LG1506 promotes the interaction of SMRT with RAR. HepG2 cells were cotransfected with a construct expressing a VP16-RARα fusion protein, a Gal4-DBD-receptor interaction domain of SMRT, and a luciferase reporter with five Gal4-binding sites. Cells were cultured in the absence or in the presence of 1 μm LG1506 or 1 μm Ro41-5253 with increasing concentrations of ATRA. In these experiments, endogenous RXR provides the dimerization partner for transfected VP16-RARα.

Chromatin immunoprecipitation (ChIP) assays demonstrate action of LG1506 on retinoid response gene CD38

We next evaluated whether the enhanced interaction of corepressors with the RAR-RXR complex observed in the presence of LG1506 could be observed at relevant target gene promoters. Specifically, we used a ChIP assay to access the interaction of SMRT with the promoter of the retinoid response gene CD38 in HL-60 cells. Given the limited number of cells available to us, it was not possible to perform an analogous experiment in primary HSCs. As expected, the RXR-RAR heterodimer was shown to interact with the corepressor SMRT in the absence of ligand. Upon addition of ATRA, however, SMRT was released from the complex (Fig. 8A, white bars). Importantly in this context, treatment with LG1506 enhanced the recruitment of SMRT to the retinoic acid response element (RARE) of CD38 gene although reducing the ability of ATRA to dissociate the complex (Fig. 8A, black bars). In addition, in agreement with the two-hybrid studies described above (Fig. 7B), Ro41-5253 also enhanced recruitment of SMRT to the CD38 gene (Fig. 8A, gray bars), albeit to a lesser degree. The functional consequences of this enhanced corepressor binding was next evaluated using ChIP to assess histone H4 acetylation at the CD38 promoter in cells treated as above. In this manner, we observed a high constitutive level of histone H4 acetylation at this promoter that was further enhanced by ATRA treatment (Fig. 8B). A significant reduction in both basal and ATRA-induced (0.1 μm) H4 acetylation was observed in cells treated with LG1506. However, as expected, this could be reversed when higher concentrations of ATRA were added. Interestingly, in this context, Ro41-5253 did not alter the basal H4 acetylation level, and its ability to reverse ATRA induced acetylation reflected its decreased ability to recruit SMRT when compared with LG1506. These data are in agreement with our hypothesis that LG1506, through its actions on RXR, is able to inhibit the transcriptional activity of the RAR-RXR complex and highlights the fundamental mechanistic differences between LG1506 and Ro41-5253. Figure 9 provides a schematic overview of the actions of LG1506 and Ro41-5253 on the RXR-RAR heterodimer. NR ligands and putative targets are listed in Table 4.

Figure 8.

LG1506 facilitates the recruitment of the corepressor SMRT to the RAR enhancer within the CD38 promoter. A, Recruitment of SMRT to the CD38 promoter was assessed using ChIP in HL-60 cells treated with vehicle, LG1506, or Ro41-5253 in the presence and absence of the indicated concentrations of ATRA. B, Acetylation of histone H4 was evaluated under the same conditions as in A. For both experiments, cells were harvested after cross-linking and subjected to immunoprecipitation with rabbit IgG control (IgG), SMRT antibody (SMRT), or anti-acetyl-histone H4 (Act-H4).

Figure 9.

Schematic overview of RXR-RAR biology in human HSCs. In the presence of a canonical agonist (ATRA), coactivators are recruited to the RAR-RXR heterodimer, resulting in transcriptional activation. Conversely, RAR antagonists such as Ro41-5253 recruit corepressors to the RAR-RXR complex, suppressing transcription. RXR-RAR inverse agonists such as LG1506 more potently promote the recruitment of corepressors to RXR-RAR, resulting in a super-repressed state of transcription.

Table 4.

NR ligands and putative targets

Compound	Function	Sequence
LG1506	RXR modulator
Ro41–5253	RARα antagonist
LG815	RAR partial agonist
LG1612	RXR modulator
LG208	RXR antagonist
LG268	RXR agonist
ATRA	RAR full agonist
LXXLL peptide
RL8		VGEWWSSLPLLLEQKTVGN
RL115		GLQSEHGLFRLLSQDPTTI
RL101		GDAKFPILRGLLIGESRDV
RL135		NWTGVFRLPQLLPHGQMEH

Discussion

The RXR is a member of the nuclear hormone receptor family and is involved in mediating several signaling cascades that have profound effects on organ homeostasis and metabolism (37,38). This receptor has the unique ability to heterodimerize with many different partners, including RAR, and also homodimerize with itself (37,38). Although RAR signaling in normal and malignant hematopoiesis has been extensively studied (18,19,20,21,22,23,24,25), much less is known regarding the contribution of RXR signaling in hematopoiesis and, in particular, the regulation of HSC fate (26,27,28,39). Previous studies have demonstrated that ligand activation of RXR causes an inhibition of BFU-E production from both murine and human progenitor cells in culture (27). Tocci et al. (18) reported that treatment of murine embryonic/fetal liver hematopoietic progenitor cells with 9-cis-retinoic acid, which can signal through RXR or RAR, caused a shift toward myelopoiesis at the expense of erythroid progenitor production. Treatment with 9-cis-retinoic acid also reduced the number of fetal liver-derived high-proliferative-potential CFCs and abolished their replating potential (18). However, before this study, the function of RXR in regulating adult human HSC fate determinations has not been described. We demonstrate that human hematopoietic repopulating cells can be maintained in culture via treatment with a mechanistically unique RXR ligand that functions as an inverse agonist in the context of the RXR-RAR heterodimer. Although the level of human hematopoietic cell engraftment was modestly higher in mice transplanted with d-0 BM CD34⁺CD38⁻lin⁻ cells compared with mice transplanted with their progeny cultured with TSF plus LG1506, the progeny of TSF plus LG1506 had 14-fold more HSC content than the progeny of TSF alone. These results demonstrate that RXR signaling plays an important role in the HSC differentiation program. This observation is also striking given multiple demonstrations that human HSCs characteristically decline rapidly in number in cytokine-treated liquid suspension cultures (40,41) absent cell contact with surrogate niches (42,43), microenvironmental ligands (e.g. Delta-1) (44,45), or genetic modification of HSCs (46). Our results indicate that RXR signaling is centrally involved in the control of human HSC differentiation, and modulation of RXR signaling can be achieved via translatable, pharmacological means.

Because the addition of LG1506 to cytokine cultures caused a preferential maintenance of HSCs in culture as compared with cytokines alone, we examined whether LG1506 had an effect on HSC cell cycling, apoptosis, or lineage differentiation in culture as a basis to explain these effects. We found that the addition of LG1506 had no significant effects on the percentage of apoptotic or necrotic cells in HSC cultures over time, nor did it significantly affect the overall expression of lineage markers on the progeny of HSCs in culture. However, treatment with LG1506 did delay entry into cell cycle in HSCs in culture, and this persisted through 7 d. Because progression through the cell cycle has been associated with a loss of repopulating capacity in HSCs (11,41), these effects of LG1506 may explain, at least in part, the mechanism through which LG1506 preserves HSCs in cytokine-treated culture.

It has been demonstrated that the actions of rexinoids can be mediated through either an RXR-RXR homodimer (37,38) or via heterodimerization with one of several partners (47,48,49). However, because the RXR-RXR antagonist LG208 had no effects on HSCs in culture, we explored the possibility that LG1506 was acting via inhibition of an RXR-RAR heterodimer. In support of this hypothesis, we observed that the addition of Ro41-5253, an RAR antagonist, to HSC cultures supported the preservation of CD34⁺CD38⁻ cells and inhibited the production of committed CFCs at a level comparable to the RXR modulator LG1506. We also found that, in the presence of ATRA, the response of RARα-expressing HepG2 cells was significantly inhibited by both LG1506 and Ro41-5253. Additionally, both LG1506 and Ro41-5253 significantly reduced the expression of the RAR-dependent gene CD38 and the myeloid differentiation marker CD11b in HL-60 cells in response to ATRA. Furthermore, we demonstrated that LG1506 induced a conformational change in RXR that provided a mechanistic explanation for its unique pharmacological activities. Lastly, we showed via ChIP assay that LG1506 induces the recruitment of SMRT corepressor to the promoter of the RAR target gene CD38 and inhibits both basal and ATRA-induced H4 acetylation of this target as well. Taken together, these studies indicate that LG1506 exerts its inhibitory effects on HSC differentiation via inhibition of the RXR-RAR heterodimer. These studies also suggest that direct inhibition of RAR may have a similar or additive effect to that of LG1506 on preserving long-term repopulating HSCs in culture. We have initiated studies to test whether this combined antagonism of RXR and RAR can provide clinically relevant expansion of human HSCs.

We have chosen to focus on the role of RXR in human hematopoiesis because the function of this NR in regulating human HSC fate has not been previously shown. Our results suggest that RXR, in concert with RAR, regulates the differentiation program of human HSCs in response to cytokines, and inhibition of RXR signaling facilitates the maintenance of HSCs in culture. Of note, Ricote et al. (28) recently reported that conditional silencing of RXRα in the murine hematopoietic system using interferon-inducible MxCre mice did not result in any significant defects in BM cellularity, mature blood counts or stem or progenitor cell function in vitro. However, the authors noted persistent RXRβ activity in hematopoietic cells that may have compensated for the loss of RXRα expression in these mice (28). Conversely, in this study, treatment of human HSCs with LG1506, which inhibits both RXRα and RXRβ, caused significant inhibition of HSC differentiation and the preservation of SRCs in culture, despite exposure to proliferation-inducing cytokines. Although we did not perform secondary transplant studies to confirm whether RXR antagonism supported an expansion of long-term repopulating HSCs, the sustainment of hematopoietic progenitors capable of NOD/SCID engraftment at 8 wk after transplant verifies that RXR antagonism indeed prevented the loss of human repopulating cells in culture. In addition, although RXRα may not be required for normal hematopoiesis to occur, we have shown that antagonism of RXRα signaling can significantly alter HSC fate determinations in response to proliferative signals. Lastly, RXR may have distinct activities in human HSCs as compared with murine hematopoietic progenitors.

Significant progress has been made in the identification of pathways that regulate HSC self-renewal and differentiation (1,2,3,4,50). Recent studies also indicate that although certain pathways, such as Notch- and Wnt-signaling pathways (1,3,50), contribute critically to HSC self-renewal events, these pathways may not be essential for normal hematopoiesis to occur (8,9). As importantly, HSC self-renewal likely involves a complex interplay of intrinsic and extrinsic signaling pathways (50,51,52). Our studies suggest a critical role for RXR in the regulation of human HSC differentiation and provide a pharmacological approach to modulate RXR signaling as part of an overall strategy to amplify human HSCs.

Materials and Methods

Isolation and purification of primary human HSCs

Whole, unprocessed human BM was purchased (Lonza, Gaithersburg, MD) and diluted more than 5-fold with Dulbecco’s PBS (D-PBS; Invitrogen, Carlsbad, CA) containing 10% heat-inactivated fetal bovine serum (FBS; Hyclone, Logan, UT) and 1% penicillin streptomycin (pcn/strp) (Invitrogen). Mononuclear cells were isolated by density centrifugation using Lymphoprep (Axis-Shield, Oslo, Norway) before lineage depletion, as previously described (17).

Lineage depletion was conducted using the Human Progenitor Enrichment Cocktail (Stem Cell Technologies, Vancouver, British Columbia, Canada), according to the manufacturer’s suggested protocol. Briefly, BM mononuclear cells were resuspended at 5 × 10⁷ to 8 × 10⁷ cells/ml in D-PBS/10% FBS/1% pcn/strp, and incubated with 100 μl/ml antibody cocktail for 30 min, followed by incubation with 60 μl/ml magnetic colloid for 30 min. Cells were then negatively selected using the manufacturer’s recommended procedure. Lin⁻ cells were washed twice, quantified by manual hemacytometer count, and cryopreserved in 90% FBS and 10% dimethylsulfoxide (Sigma-Aldrich, St. Louis, MO).

Lin⁻ BM cells were thawed rapidly, washed in Iscove’s modified Dulbecco’s medium (Invitrogen) containing 10% FBS and 1% pcn/strp and resuspended at 5 × 10⁶ to 1 × 10⁷/ml. Immunofluorescence staining was conducted using antihuman CD34-fluorescein isothiocyanate (FITC) and antihuman CD38-phycoerythrin (PE) monoclonal antibodies (Becton Dickinson, San Jose, CA). Stained cells were washed twice and resuspended in D-PBS/10% FBS/1% pcn/strp. Analysis and cell sorting was conducted using a FACSvantage cytometer (Becton Dickinson) to isolate CD34⁺CD38⁻ and CD34⁺CD38⁺ subsets. The CD34⁺CD38⁻ sort gate was set to collect only those CD34⁺ events falling in the lower 50% of the CD38⁻ quadrant, as determined by staining with isotype controls (Becton Dickinson) (13,14,17).

In vitro hematopoietic assays

Purified BM CD34⁺CD38⁻lin⁻ cells were seeded at 0.25 × 10⁴ to 1 × 10⁴ per well in 24-well plates (Becton Dickinson) containing Iscove’s modified Dulbecco’s medium with 10% FBS, 1% pcn/strp, and TSF (R&D Systems, Minneapolis, MN) with or without LG1506 (1 μm; Ligand Pharmaceuticals, San Diego, CA), Ro41-3532 (1 μm; BIOMOL International, Plymouth Meeting, PA), ATRA (1 μm; Sigma-Aldrich), clofibrate (1 μm; Sigma-Aldrich), rosiglitazone (1 μm; Cayman Chemicals, Ann Arbor, MI), and carbaprostacyclin (1 μm; Cayman). Cultures were maintained in a 37 C, 5% CO₂ atmosphere for 7 d. Cultured cells were collected at d 7, and viable cell counts were performed using trypan blue dye exclusion. Flow cytometric analysis using anti-CD34-FITC and anti-CD38-PE and the appropriate isotype control antibodies was performed on all samples. Immunofluorescence data were acquired using a BD FACScalibur instrument (Becton Dickinson).

CFC assays were performed in triplicate in 35-mm plates (Nunc, Rochester, NY) using 0.5 × 10³ to 1 × 10³ cells per dish in MethoCult GF-H4434 medium (Stem Cell Technologies). After 14 d, dishes were scored for BFU-E, CFU-GM, and CFU-mix colony (>50 cells) formation.

SRC assay

Eight-week-old NOD/SCID mice (Jackson Laboratory, Bar Harbor, ME) received 325 cGy total body irradiation from a ¹³⁷Cs source and then transplanted, via tail vein injection, with either d-0 BM CD34⁺CD38⁻lin⁻ cells (0.25 ×10⁴ to 2 ×10⁴) or their progeny after 7 d culture. Each mouse also received coinjection of 2 × 10⁴ BM CD34⁺CD38⁺lin⁻ accessory cells to facilitate engraftment, as previously described (17). Eight weeks after transplantation, mice were killed and BM was collected from bilateral femurs as previously described (17). Flow cytometric analysis was performed to determine human hematopoietic cell engraftment using monoclonal antibodies against human leukocyte antigens to identify engrafted human leukocytes and discriminate their hematopoietic lineages (13,14,17). Mice were scored as positively engrafted if the BM displayed at least 0.1% human CD45⁺ cell chimerism via flow cytometry analysis, consistent with previously published criteria for human cell repopulation in NOD/SCID mice (12,14). SRC frequency in each cell source was calculated using the maximum likelihood estimator as described previously by Taswell (53) for the single-hit Poisson model (31).

Analysis of HSC cell cycle, apoptosis, and lineage differentiation

Cell cycle status of d-0 input CD34⁺CD38⁻lin⁻ cells and the d +3 and +7 cultured progeny was determine using surface intracellular DNA analysis, as previously described, using anti-Ki-67-FITC, anti-CD38-PE, anti-CD34-allophycocyanin (APC), and 7AAD (Becton Dickinson) (54). BM CD34⁺CD38⁻lin⁻ cells and their progeny at d +3 and +7 of culture were stained with annexin V (Becton Dickinson) and 7AAD to measure the percentage of apoptotic cells (annexin V⁺7AAD⁻) and necrotic cells (annexin V⁺/7AAD⁺) in culture, as we have previously described (55). Postculture analysis for differentiation markers was conducted by staining with anti-CD34-FITC, anti-CD38-PE, anti-CD3-FITC, anti-CD19-APC, anti-CD33-FITC, and anti-CD13-APC or the appropriate isotype control antibodies for 30 min on ice. Immunofluorescence data were acquired using a BD FACScalibur with Cell Quest software (Becton Dickinson).

Expression constructs and reagents

Human RXRα-VP16, LG815, LG268, LG1506, LG1612, and LG208 were provided by Ligand Pharmaceuticals (San Diego, CA). RSV expression plasmid for the human RARα was a gift of Dr. R. Evans (Salk Institute, San Diego, CA). ATRA and TTNPB were purchased from Sigma-Aldrich. Ro41-5253 was purchased from BIOMOL.

Cell culture, transfection, and reporter gene assays

HepG2 cells were maintained and transfected as described previously (56). All transfections were performed in triplicate. The results presented were typical of several independent transfection experiments. HL-60 cells were obtained from the American Type Culture Collection (Rockville, MD). Cells were grown in Iscove’s modified Dulbecco’s medium (Invitrogen) supplemented with 20% FBS. HL-60 cells were treated with different compounds for 3 d before RNA isolation.

RNA isolation and real-time RT-PCR

Total RNA was isolated from HL-60 cells according to the manufacturer’s instructions for the Total RNA Mini-Kit (Bio-Rad, Hercules, CA). Total RNA (1μg) was reverse transcribed to cDNA using iScript cDNA synthesis kit (Bio-Rad). For HSCs, RNA isolation and real-time PCR analysis was performed as previously described (17). Gene-specific primers were purchased from Integrated DNA Technologies (Coralville, IA).

Affinity selection of RXR-bound LG1506 interacting peptides

Human RXRα was cloned into the EcoRI sites of a baculovirus shuttle vector pDW464 to make an in-frame fusion of the RXRα with the biotin acceptor peptide (Science Reagents, El Cajon, CA). Recombinant RXRα baculovirus was generated using Bac-To-BacR Baculovirus Expression System (Invitrogen) following the protocol provided by the manufacturer. RXRα recombinant protein was produced in Sf-9 insect cells after infection with recombinant baculovirus particles. A soluble extract of infected Sf-9 cells was used to affinity purify biotinylated RXRα fusion protein with monomeric avidin resin (Promega Corp., Madison, WI), as described for LRH-1 protein (56). To select for RXRα-LG1506 binding peptides, we used a protocol that was previously described (57), with minor modifications. Construction of the LXXLL M13-phage library was described previously (58).

ChIP assays

HL-60 cells (2 × 10⁷), treated as indicated in the figure legend, were harvested and cross-linked using 1% formaldehyde for 15 min at room temperature. Subsequently, the reaction was stopped by an additional 15-min incubation with 250 mm glycine. Cells were then washed twice with ice-cold PBS. After centrifugation, the cell pellets were lysed for 15 min at 4 C in 1 ml RIPA buffer [50 mm Tris-HCl (pH 7.5), 150 mm NaCl, 0.1% Nonidet P-40, 0.5% sodium deoxycholate, 0.05% SDS, 1 mm EDTA, and 1× protease inhibitor mixture] and sonicated 15 times for 15 sec each at setting 10 (Misonix Microson Ultrasonic Cell Disruptor XL). Supernatants were collected by centrifugation at 4 C for 15 min at 10,000 × g, diluted in RIPA buffer, and immunocleared in 100 μl protein A/G-Plus agarose beads [50% slurry in 10 mm Tris-HCl (pH 8.0), 1 mm EDTA, 200 μg/ml sonicated salmon sperm DNA, and 500 μg/ml BSA] for 30 min at 4 C. Immunoprecipitation was performed overnight at 4 C with either 10 μg anti-SMRTe antibody (06-891; Upstate, Charlottesville, VA), 2.5 μg anti-acetyl-histone H4 (06-866; Upstate) or 10 μg rabbit IgG control (Santa Cruz Biotechnology, Santa Cruz, CA). After immunoprecipitation, 100 μl protein A/G-Plus-agarose beads [50% slurry in 10 mm Tris-HCl (pH 8.0), 1 mm EDTA, 200 μg/ml sonicated salmon sperm DNA, and 500 μg/ml BSA] was added and allowed to incubate for 4 h at 4 C. Precipitates were washed twice sequentially for 5 min each with the following: buffer A [50 mm HEPES (pH 7.8), 140 mm NaCl, 1 mm EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, and 1× protease inhibitor], buffer B [50 mm HEPES (pH 7.8), 500 mm NaCl, 1 mm EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, and 2× protease inhibitor], buffer C [20 mm Tris-HCl (pH 8.0), 1 mm EDTA, 250 mm LiCl, 0.5% Nonidet P-40, 0.5% sodium deoxycholate, and 1× protease inhibitor], and Tris-EDTA (10 mm Tris-HCl and 1 mm EDTA). Precipitates were eluted twice in 50 mm Tris-HCl (pH 8.0), 1 mm EDTA, and 1% SDS by incubation at 65 C for 10 min, and then the cross-linking was reversed by addition of 230 mm (final concentration) NaCl and incubation at 65 C overnight. DNA was isolated using a QIAquick PCR purification kit (QIAGEN, Valencia, CA). Quantitative RT-PCR was performed with 1 μl immunoprecipitated DNA. Data are normalized to the input for the immunoprecipitation.