The retinoid signaling pathway inhibits hematopoiesis and uncouples from the Hox genes during hematopoietic development

Abstract

Retinoic acid (RA) is a well established inducer of Hox genes during development of neurectoderm, however effects of RA on Hox expression are poorly defined in mesoderm and not defined in the hematopoietic compartment. Both Hox genes and retinoid signaling have been suggested to modulate HSC self-renewal, supporting the notion that RA signaling might drive HSC self-renewal through Hox gene induction. Here we investigate this possibility by comprehensively evaluating Hox gene expression using mouse embryonic stem (ES) cells differentiated in vitro. In unspecified mesoderm, we find that RA coordinately upregulates anterior 3’ Hox genes from clusters A, B, and C, and downregulates posterior 5’ Hox genes from clusters A-D. However, hematopoietic development of mesoderm was inhibited by RA, and we find further that retinoids are entirely dispensable for hematopoiesis in vitro. More surprisingly, in fully specified hematopoietic progenitors, Hox genes are refractory to regulation by RA, although other RA targets are normally regulated. Pulses of RA exposure demonstrate that the Hox complexes are decoupled from RA regulation progressively in lateral plate mesoderm as it undergoes hematopoietic specification. Thus, Hox genes are targets of the RA pathway only in selected cell types, and are clearly not regulated by RA in the earliest hematopoietic progenitors. We propose that the developmental uncoupling of the Hox complexes protects the Hox code from potential RA signaling centers as hematopoietic stem cells migrate or circulate during development.

Introduction

Hox gene products play a key role in establishing positional identity along the anterior-posterior (AP) axis [1]. Mammalian Hox genes are organized in four genomic clusters (A, B, C, and D) each comprising from 9 to 11 genes arranged in a homologous array [2]. Hox gene expression is colinear with the AP axis in very early development: the 3’ Hox genes are expressed in anterior anatomical structures, while 5’ Hox genes are expressed in posterior structures [3, 4]. Later in development, Hox genes participate in organogenesis, and numerous observations suggest that Hox genes modulate different stages of hematopoietic development. For example ectopic overexpression of various Hox genes (i.e. HOXB4, HOXA10) in adult HSCs conferred a proliferation advantage allowing these genes to be used as tools to efficiently expand HSCs ex vivo [5-7]. In addition we have previously shown that ectopic over-expression of HOXB4 in ES cell-derived hematopoietic cells enhanced their proliferation capacity, and conferred long-term repopulating potential on these cells [8]. Of the many factors known to regulate Hox gene expression in vivo, retinoic acid (RA) has been shown to play an especially important role during early embryonic development [9]. In several cases, RA directly induces the expression of anterior Hox genes, and RA response elements (RAREs) have been identified in regulatory domains of several anterior Hox genes [10-12]. Most of these studies focus on the embryonic development of neurectoderm, while RA-dependent regulation of Hox genes in mesoderm and particularly within the hematopoietic compartment has been assumed but never tested comprehensively. It is well known that RA drives terminal differentiation of granulocytes from myeloid progenitors [13], and recent observations indicate that RA also shapes immune reactivity of various lymphocyte populations [14, 15]. In addition to effects on differentiated cells, RA has also been reported to delay differentiation of progenitors [16] and to stimulate the ex vivo expansion and long-term repopulating activity of adult HSCs [17]. Superficial evidence thus exists for similar phenotypic effects on HSC self-renewal downstream of both RA signaling and the Hox pathway. Since RA is a well established inducer of Hox gene expression during embryonic neurectodermal development, we reasoned that it might also regulate these genes during hematopoietic development, and thus that pharmacological activation of RA receptors (RARs) might enable the expansion of ES-derived hematopoietic stem/progenitor cells, as we have seen for Hoxb4 [8, 18]. Consistent with this hypothesis, when we evaluated unfractionated total embryoid bodies (EBs), we observed upregulation of 3’ Hox genes, including Hoxb4, upon RA treatment. However, in spite of the induction of these genes, RA-instructed EBs showed an impaired hematopoietic potential. We therefore performed a comprehensive evaluation of the effect of RA on the Hox gene family, in different cell fractions at various points in development. We find that the RA signaling pathway is functional in the first hematopoietic progenitors, but it is decoupled from regulation of the Hox genes.

Materials and Methods

ES cell Culture and in vitro differentiation

Mouse ES cells were maintained on irradiated MEFs (mouse embryonic fibroblast) in DMEM/15% FBS (fetal bovine serum). Except where indicated, the E14 ES cell line was used. For EB differentiation, ES cells were harvested and MEFs removed by 40 min adherence to gelatinized dishes. EBs were differentiated as described previously [8]. RA treatment started at day 4 and EBs were harvested at day 6 or otherwise indicated. To study the potential role of endogenous retinoids, EBs were cultured in medium where FBS was replaced with charcoal-stripped FBS (Sigma). Some of these samples were also treated with 10 ng/ml hBMP4 (bone morphogenetic protein 4; Peprotech).

Hematopoietic progenitor/endothelial cell culture

E14, iHoxB4 and iHoxA4 cell lines were used to produce undifferentiated hematopoietic progenitors: 20,000 c-Kit/CD41 double-positive iHoxB4 or iHoxA4 cells were FACS sorted from day 6 EBs (disaggregated with trypsin) and plated on OP9 monolayers (20,000 OP9 cells per well in 24-well dishes) in IMDM supplemented with 10% FBS, 5 ng/ml mVEGF (vascular endothelial growth factor), 40 ng/ml hTPO (thrombopoietin), 40 ng/ml mFlt-3 (Fms-like tyrosine kinase 3) ligand (all cytokines from Peprotech). Cells were cultured for 5 or 6 days, thereafter 100,000 cells were passaged to fresh OP9 monolayers (80,000 OP9 cells per well in 6 well dishes) and further cultured for 3 days. In addition, 600,000 Flk-1/VE-cadherin double-positive E14 derived cells were FACS sorted from day 6 EBs and plated on OP9 monolayers in IMDM supplemented with the same cytokines as described above. These cells were cultured for 6 days.

Inducible ES cell lines

The construction of inducible HoxB4 and HoxA4 ES cell lines was described previously [8, 18]. The cDNA for mouse Raldh2 was obtained from Open Biosystems.

Ligands

All-trans retinoic acid (RA), all-trans retinol (retinol) AM580, and CD437 were obtained from Sigma-Aldrich. Diethylaminobenzaldehyde (DEAB) was obtained from ScienceLab.com, Inc. Ligands were redissolved in 50% DMSO/ethanol and this solvent was used as vehicle control, Cells were treated with a 1000x stock solution of RA or other synthetic compounds. Doxycycline (Sigma-Aldrich) was dissolved in water.

Real-time quantitative PCR

RNA was extracted with Trizol (Invitrogen); reverse transcription was performed with a high-capacity cDNA reverse transcriptase kit (Applied Biosystems). Quantitative PCR was performed using real-time PCR (7500 or 7900 Real-Time PCR Systems; Applied Biosystems) as described [19]. In brief, 40 cycles at 95 °C for 12 s and 60 °C for 30 s using TaqMan Gene Expression Assay primer-probe sets (Applied Biosystems). The Taqman assay IDs are in Table S1. Quantitative analysis of gene expression was conducted using the comparative cycle threshold (Ct) method and normalized to Gapdh. All PCR reactions were done in triplicate with one control reaction containing cDNA that was reverse transcribed without RT enzyme.

Western-blot analysis

10 μg of protein from whole-cell extracts was separated by electrophoresis on a 10% polyacrylamide gel and transferred to polyvinylidene difluoride membrane (Pall Corporation). Membranes were probed with supernatant from I12 (hybridoma secreting rat anti-mouse Hoxb4 antibody; Developmental Studies Hybridoma Bank, Iowa City, IA) and then stripped and reprobed with anti-β-actin (Genscript).

FACS

For FACS analysis 20,000 events were collected. To eliminate dead cells or cell debris, live cells were gated based on FSC/SSC profile, and negativity for PI (propidium iodine) staining. The following antibodies were used: c-Kit-APC, CD41-PE, CD45-PE, Flk-1-APC, Flk1-PE, PDGFRα-PE, PSA-N-cam-APC and VE-cadherin-alexa647 (all from eBiosciences). Cells were analyzed and/or sorted on a FACS Aria II(BD Biosciences).

Colony assays

For each colony assay, 6 day EBs were disaggregated and 50,000 cells were plated into 1.5 ml methylcellulose medium (M3434, StemCell Technologies). Primitive erythroid colonies were counted after 5 days, other colonies after 10 days.

Detection of apoptosis

Apoptosis was detected with Vybrant apoptosis assay kit (Invitrogen). 6d EB derived disaggregated cells were stained with annexin V and propidium iodine as recommended by the manufacturer.

Statistical analyses

Values are expressed as mean±SD of the mean. Significant differences between mean values were evaluated using two-tailed, unpaired Student’s t test.

Results

Coordinated positive/negative regulation of anterior vs. posterior Hox genes by RA in unspecified mesoderm

To explore the effects of retinoid signaling on mesodermal/early hematopoietic development, an ES/EB in vitro differentiation model was used. The EB is a three dimensional cell aggregate containing cells from all three germ layers, although under conditions of 15% serum supplementation used here, the primary germ layer represented is mesoderm. Our flow cytometric analysis indicated that 70-80% of cells were positive for the early mesodermal markers Flk-1 and/or PDGFRα after 4 days of EB development. Production of mesoderm is robustly initiated at about day 3 and the first hematopoietic precursors are detected at day 4-5 in this model system [20]. To test the effects of retinoid signaling during early hematopoietic development, day 4 EBs were cultured in the presence of retinoic acid (RA), harvested at day 6, and analyzed for gene expression by quantitative RT-PCR. First we measured the transcript level of a Rarβ a well-established generic RA target [21, 22]. A dose-dependent induction of Rarβ was detected, demonstrating that day 4 EB cells are responsive to RA treatment (Fig. 1A). We have recently shown that all members of Hox paralog group 4 (Hoxa4, b4, c4 and d4) have a similar stimulatory effect on early hematopoietic differentiation [18], therefore we measured the expression of these genes. Hoxa4, b4 and c4 showed a dose-dependent up-regulation by RA treatment, with the fold induction of Hoxa4 being the highest. Hoxd4 was non-responsive (Fig. 1B). We also assessed the protein level of Hoxb4 by immunoblot and consistent with the RNA expression pattern, an elevated level of Hoxb4, which runs as a doublet with a faint upper band, was detected in RA treated EBs (Fig. 1C). The clustered Hox genes are among the most important RA-regulated transcription factors with the ability to modulate blood cell development; several members of the A, B and C clusters are expressed in hematopoietic cells and several Hox genes of the A and B clusters have demonstrated effects on hematopoiesis [23]. We therefore expanded our analysis to include all 39 clustered Hox genes. The A, B and C Hox clusters showed a very similar RA response: the anterior genes, Hoxa1-5, b1-5 c4-5 were up-regulated in EBs cultured with RA (although the Hoxb3 was poorly detected), while the remaining Hox genes in these three clusters were downregulated or some cases they were barely detectable (Hoxb13, Hoxc11-13) (Fig. 2). The anterior Hoxd genes showed a complex regulation: low concentrations of RA inhibited while higher concentrations had no effect. However the posterior genes (Hoxd8-13) were consistently downregulated as in the A, B and C clusters. This expression profile analysis indicated that RA indirectly or directly, but coordinately, regulates the expression of clustered Hox genes in EB-derived mesoderm. Similar up-regulation of 3’ Hox genes was observed in EBs generated from other ES cell lines (R1 and CCE, Fig. S1). Together these results indicated that RA administration reprogrammed the Hox code, enhanced the expression of anterior Hox genes in clusters A, B and C and suppressed posterior Hox gene expression in all clusters (A-D).

Figure 1.

RA dependent regulation of Hox genes in EBs. (A) Rarβ transcript level in the presence of various concentrations of retinoic acid (RA): vehicle, 10, 100, 1000 nM. (B) Transcript levels of Hox paralog group 4. RA concentration was as in part A. (C) Immuno-blot analysis of Hoxb4 protein. The identity of specific bands was confirmed by co-migration with a band (around 35 kD) seen in the extract of Hoxb4 over-expressing cells (iHoxB4 + doxycycline). β-actin was used as a loading control.

Figure 2.

Gene expression profile of Hox genes in EBs. Transcript levels of all clustered Hox genes. RA concentration was as in Figure 1 part A.

Hox gene expression is regulated mainly via the activation of Rarα receptor

RA is the active form of vitamin A (retinol) which regulates gene expression via the activation of nuclear hormone receptors. Retinoid signaling is transduced by RA receptors (Rarα, β, and γ) [24]. During the early stages of EB development (day 2-4) the mRNA level of Rarα and Rarγ showed the highest expression and Rarβ was barely detected, while at later stages (day 4-6) Rarα showed the highest expression (Fig. S2A), suggesting that Rarα would be primarily responsible for regulation Hox gene expression in EB-derived hematopoietic progenitors, which arise between days 4 and 6. Our qRTPCR assays do not distinguish between the two known isoforms of Rarα. To further investigate the contribution of various receptors to Hox gene regulation we pharmacologically probed the RARs using α- and γ-selective agonists. RA, and RARα-specific agonist (AM580) elicited a profound up-regulation of Rarβ and 3’ Hox genes, however the RARγ specific ligand was much less potent (Fig. S2B) suggesting again that the most relevant nuclear receptor is Rarα. However, since Rarβ is upregulated by RA (Fig. 1A), extended ligand treatments may result in Rarβ also participating in transduction of the RA signal.

Pulse treatment reveals that kinetics of Hox induction are distinct in early vs. late EBs

The studies described above were performed in EBs treated with RA for 2 days. To investigate potential direct effects of RA signaling, we analyzed cells immediately after 2- or 6-hour pulses with 1 μM RA at various stages of EB development (Fig. S3). These shorter treatments generally resulted in weaker Hox gene induction compared to EBs treated for 48 hours (Fig. 2). Interestingly, while a 2 hour pulse upregulated Hoxa1 and Hoxa3 at all stages tested, Hoxa4 was induced rapidly only in early EBs, but not in day 4 or later EBs suggesting that Hoxa1 and Hoxa3 are probably directly regulated at all time points, while regulation of Hoxa4 incorporates an aspect of indirect or delayed regulation at least at later time points. In addition, besides the length of RA treatment, the time-point of the RA administration was critical. EBs treated at day 2 or 3 showed a much more robust induction compared to 4-6 day EBs (Fig. S3). In contrast to the Hox genes, the classic generic retinoid response gene, Rarβ, did not show a diminished retinoid response at late time points.

We also evaluated the kinetics of downregulation of negatively regulated Hox genes (Hoxa11 and a13). Without RA treatment, the posterior Hox genes are expressed at significant and increasing levels during EB differentiation (Fig. S3 and data not shown). We observed that 6 hour RA treatment profoundly decreased the expression of both Hoxa11 and a13. Considering that the mRNA must be degraded to detect downregulation, this is a remarkably rapid response.

Early activation of the retinoid pathway promotes neurectoderm at the expense of mesoderm differentiation

The data above suggested that it might be more effective to treat the EB at the early time points to achieve maximal up-regulation of Hox genes, if the objective is to enhance hematopoiesis through paralog group 4-like effects. We therefore compared EBs treated with RA from day 2-6 vs. day 4-6. We indeed obtained a much greater induction of Hoxa3 and Hoxa4 by day 6 with lower concentrations of RA (10 or 100 nM) when we induced from day 2 vs. from day 4 (Fig. 3A). However, EBs treated with RA at day 2 showed impaired development; they were much smaller even in the presence of 10 nM RA (data not shown); more importantly most of these cells were Flk-1 and PDGFRα negative (Fig. 3B), suggesting that mesoderm development was inhibited. Consistent with these findings we detected severe reductions of mesodermal markers (Brachyury, Mesp1), moreover, these RA-instructed EBs expressed markers of neurectoderm (Pax6 and Neurog2; Fig. 3C). We then evaluated N-cam together with Flk-1 and PDGFRα by flow cytometry. In concert with neurectodermal markers, N-cam levels dramatically increased upon RA administration (Fig. 3D) suggesting that RA treatment shifted EB development from mesoderm toward neurectoderm. Therefore, although a more dramatic Hox gene induction is induced by RA early in development, early activation of the RA signaling pathway profoundly blocks mesoderm, preventing the subsequent generation of hematopoietic progenitors. This finding is consistent with previous reports that characterized the effect of early RA treatment on murine EB differentiation models [25, 26]. In the following experiments, we focus on the effects of RAR activation at later stages of EB differentiation, when mesoderm differentiation has already been initiated.

Figure 3.

Effects of RA administration on early stage of EB development. (A) Transcript profile of Hoxa3 and a4 in the presence of various concentrations of RA (control, 10, 100, 1000 nM). (B) Flow cytometric profile of Flk-1 and PDGFRα expression of 4d EBs in the presence of the indicated concentration of RA (ligand treatment was started at day 2). (C) RNA profile of Pax6, Neurog2, Brachyury and Mesp1 (RA treatment was as in figure part B). (D) Flow cytometric profile of Flk-1 and PDGFRα expression combined with N-cam. Day 2 EBs were treated with the indicated concentration of RA, EBs were harvested and analyzed at day 4.

Later activation of RA receptors interferes with the hematopoietic development of mesoderm

Our results indicated that pharmacological concentrations of RA enhanced the expression of several 3’ Hox genes in developing EBs. Previously we have described that over-expression some of these genes (Hoxa4-d4) increased the number of hematopoietic cells and stimulated the expansion of hematopoietic progenitors on OP9 stromal cells [18]. Our transcript analyses above indicated that Hoxa4-c4 showed the highest Hox expression suggesting that the upregulation of these transcription factors might enhance the development of hematopoietic progenitors. However, when we quantified hematopoiesis, we found the opposite effect. We obtained fewer undifferentiated hematopoietic progenitors (c-Kit/CD41 double positive cells) in RA-instructed EBs (Fig. 4A and B). CD41 is a panhematopoietic marker in early embryonic development, making c-Kit/CD41 the most inclusive phenotype for the detection of blood cell progenitors in this system [27, 28]. We also detected significantly fewer c-Kit⁺/CD45⁺ cells in the presence of high concentration of RA. To determine the frequency of hematopoietic progenitors in RA-instructed EBs, we measured colony-forming activity in methylcellulose colony assays. Consistent with the flow cytometric results, we detected fewer hematopoietic colonies from RA-treated EBs (Fig. 4C). In addition we also assessed the expression of several markers of blood cell development (Gata1, Runx1) and found diminished expression upon RA treatment (Fig. 4D). In some systems, RA can elicit apoptosis [29], however we did not detect increased annexin V positivity in day 6 EBs (Fig. S4) nor did we observe increased annexin V staining on the c-Kit/CD41 double positive fraction upon RA treatment (data not shown) suggesting that the enhanced apoptosis is not responsible for inhibition of blood cell generation. These findings imply that the pharmacological activation of RAR strongly compromised hematopoietic differentiation of mesoderm.

Figure 4.

RA administration negatively modulates hematopoietic development. (A) Flow cytometric profile of day 6 EBs in the presence of the indicated concentration of RA for detection of c-Kit, CD41 and CD45 expression. (B) Statistical analysis of double positive cell percentage: NS p>0.05, * p<0.05, ** p<0.01, *** p<0.001. (C) Colony forming unit (CFU) activity. Primitive erythroid (Ery P), definitive erythroid (Ery D), granulocyte erythrocyte-macrophage-megakaryocyte (GEMM) and granulocyte-macrophage (GM) colonies. (D) Transcript profile of Gata1 and Runx1 in the presence of various concentrations of RA. Statistical analysis of PCR data: NS p>0.05, * p<0.05, ** p<0.01, *** p<0.001. (E) Flow cytometric profile of endothelial progenitors (Flk-1⁺ / VE-cad⁺ cells) sorted from day 6 EBs and cultured on OP9 for 6 days. Under control conditions, this cell fraction produces both hematopoietic and endothelial derivatives assayed by c-Kit/CD45, and Flk-1/VE-cadherin, respectively. Addition of RA prevented the production of c-Kit⁺/CD45⁺ derivatives, but not Flk-1⁺/VE-cadherin⁺ derivatives. (F) The effect of RA on expansion of OP9 co-cultured hematopoietic progenitors. c-Kit/CD41 positive cells were sorted from day 6 EBs and cultured on OP9 stromal cells for 5-6 days, thereafter cells were treated with 1 μM RA for an additional 3 days. EBs and the sorted progenitors were derived from iHoxB4 or iHoxA4 ES cells (to activate these Hox genes, cells were treated with 1 μg/ml doxycycline).

Next we assessed the effect of RA on endothelial cells. We sorted Flk-1/VE-cadherin double positive cells from untreated day 6 EBs and cultured on OP9 stromal cells for 6 days. This cell fraction produces both endothelial and hematopoietic derivatives, which we assayed by staining for c-Kit/CD45, and Flk-1/VE-cadherin, respectively. RA administration had no significant effect on the frequency of Flk-1/VE-cadherin double positive cells, but inhibited the production of c-Kit⁺/CD45⁺ cells, similar to the effect seen when RA was applied to EBs (Fig. 4E). This data suggests that RA does not interfere with endothelial cell development.

We also tested the effect of RA on self-renewing hematopoietic progenitors. Because wild-type (E14) ES-derived hematopoietic progenitors rapidly differentiate on OP9 stomal cells, we used modified ES cell lines in which expression of Hoxa4 or Hoxb4 can be induced by doxycycline [18]. Expression of either Hoxa4 or Hoxb4 (iHoxA4 and iHoxB4) allows undifferentiated hematopoietic progenitors to be produced in large quantity, and the effects of RA on these cells determined. c-Kit⁺/CD41⁺ cells were sorted from day 6 EBs initiated from these cell lines and cultured on OP9 stromal cells with doxycycline. After 5-6 days of expansion, hematopoietic cells were treated with RA for an additional 3 days. Interestingly we obtained with both cell lines a strong inhibition of cell expansion suggesting that RA might directly modulate the cell proliferation/survival of these hematopoietic progenitors (Fig. 4F). Together these results indicated that RA treatment has a general negative influence on development of ES cell-derived hematopoietic progenitors.

Endogenous RA signaling does not enhance, and is not necessary for hematopoietic development of mesoderm

In this study we used natural or synthetic agonists of the RA receptors to manipulate Hox gene expression. An important issue is whether endogenous RA can modulate this process. To enhance endogenous RA production, we constructed a tet-inducible ES cell line expressing retinaldehyde dehydrogenase 2 (Raldh2, alternative name Aldh1a2), a key enzyme for the generation of RA during the early stages of embryonic development [30]. We observed elevated expression of the canonical RA target, Rarβ, upon administration of doxycycline in these EBs (Fig. 5A) suggesting that the induction of Raldh2 enhanced RA production from serum-derived retinol present in the medium. More importantly we also observed increased mRNA expression of Hoxa4 suggesting that endogenous RA production can elicit Hox gene up-regulation. However, this level of Hox gene expression was much lower compared to that seen in RA-treated cells. To further prime the endogenous retinoid pathway in these cells, EB culture medium was supplemented with 1 μM retinol. Retinol is first converted to retinaldehyde and this compound is oxidized to RA by Raldhs. We obtained much greater induction of Hox genes by doxycycline with supplementation of retinol, consistent with enhanced RA production (Fig. 5A). In addition this up-regulation was partially reverted by administration of Raldh inhibitor (DEAB). Some of the most responsive genes (Hoxa4 and Rarβ) were slightly increased by retinol alone (without doxycycline-mediated expression of Raldh2), suggesting a low level of endogenous Raldh activity. Control EBs (derived from E14 ES cells) did indeed express low levels of Raldh2, and expression increased slightly during EB development (Fig. 5B). These results prompted us to investigate the effect of retinol and RA synthesis inhibitor in wild-type EBs. Interestingly, a slightly elevated expression of Rarβ was detected in the presence of high concentrations of retinol; in addition DEAB treatment suppressed the baseline expression level of Rarβ (Fig. 5A). We also investigated the effects of these compounds on Hox gene expression: Hoxa4 expression was very low and we failed to detect a consistent regulation, however when we assessed expression of posterior Hox genes (eg, Hoxa13) which are expressed at significant levels and downregulated by RA (Fig. 2), the retinol treatment behaved as RA, causing downregulation. In contrast, the Raldh inhibitor elicited elevated expression of Hoxa13 (Fig. 5A). These findings raised the possibility that at least in the late stage of EB development, a low concentration of RA might be generated which has a weak effect on Hox gene expression.

Figure 5.

Activation of endogenous RA signaling by over-expression of Raldh2. (A) RNA profile of Hoxa4, Hoxa13, RARβ and Raldh2 in the presence of the following compounds: 1 μg/ml doxycycline (dox), 1 μM retinol (rol) or 10 μM diethylaminobenzaldehyde (DEAB). EBs were generated from iRaldh2 or E14 (wild-type) ES cells. Statistical analysis of PCR data: NS p>0.05, * p<0.05, ** p<0.01, *** p<0.001. (B) RNA profile of Raldh2 during EB differentiation, EBs were derived from E14 ES cells. (C) FACS profile of cells from 6 day EBs developed in tissue culture medium containing 15% charcoal-stripped FBS, without/with supplementation of 10 nM retinol and/or 10 ng/ml BMP4.

Although our results indicate that pharmacological concentrations of RA interfere with hematopoietic development, they do not exclude that a low concentration of RA might still be required for blood cell formation as suggested by some reports [31, 32]. To address this issue, we cultured EBs in retinol/RA free medium: regular FBS was replaced with charcoal-stripped FBS. In charcoal-stripped serum all of the hydrophobic molecules are eliminated, including retinoids. Unexpectedly we obtained virtually no c-Kit⁺/CD41⁺ cells in this medium. We attribute this to a failure mesoderm development due to the stripping of mesoderm-inducing factors from the serum. By supplementing charcoal-stripped serum with BMP4, an important mesoderm-inducing factor, [33] EB development and hematopoiesis were rescued (Fig. 5C), indicating that charcoal-stripped serum indeed lacks mesoderm-inducing factors. Adding retinol alone had no effect (Fig. 5C) and administration of low concentration RA (10 nM) was inhibitory with or without BMP4 (data not shown). The fact that hematopoiesis can occur in retinoid-deficient charcoal-stripped medium clearly shows that RA and retinol are not essential for hematopoiesis in vitro. Although a low concentration of retinol/RA may be present during EB differentiation using regular FBS, these molecules are neither sufficient nor necessary for mesoderm formation nor hematopoiesis.

The Hox complex becomes refractory to RA regulation in specified hematopoietic progenitors

Data presented above indicated that activation of the RA signaling pathway during mesoderm commitment negatively regulated the development of ES-derived blood cell precursors, in spite of the upregulation of numerous members of Hox gene family, including the most strongly-upregulated members, from paralog group 4. Since this paralog group positively regulates early hematopoietic development [18], we reasoned that the levels of Hox expression are not sufficient, are occurring in the wrong cell types, or are accompanied by regulation of additional RA targets which override the beneficial effects of Hox up-regulation. To test whether Hox gene expression was in fact induced and to what extent within the hematopoietic progenitor population under study, we fractionated cells from day 6 EBs (RA-treated from day 4-6) based on c-Kit and CD41 expression (Fig. 6A), and evaluated Hox gene expression levels in each fraction. Hox gene induction was very impaired in the c-Kit/CD41 double positive (and also the CD41 single-positive) cell fraction while remaining robust in the non-hematopoietic fractions. This was not due to differential Gapdh levels (used for normalization) as a very similar pattern was obtained if PCR data was normalized to β-actin (data not shown). Furthermore, in contrast to the Hox genes, the classical retinoid target (Rarβ) displayed strong upregulation at least in the double positive fraction (Fig. 6B). This demonstrates that the impaired Hox induction is not due to a lack of RA signaling capacity, but rather an uncoupling of the Hox loci from the RA pathway in the c-Kit/CD41 double positive cells.

Figure 6.

Hematopoietic and lateral plate progenitors show an impaired Hox gene response to RA. (A) FACS profile (c-Kit/CD41) of day 6 EB in the presence and absence of 1 μM RA (ligand treatment started at day 4). (B) RNA profile of Hoxa3, a4, a13 and Rarβ from day 6 EBs (total) and the sorted populations. (C) FACS profile (Flk1/PDGFRα) of day 5 EBs in the presence and absence of 1 μM RA (ligand treatment started at day 4). (D) RNA profile of Hoxa3, a4, a13 and Rarβ from day 5 EBs (total) and from sorted populations.

To trace this uncoupling backwards in development, we sorted day 5 EB-derived cells according to Flk-1 and PDGFRα expression (Fig. 6C) and investigated the various subpopulations. Lateral plate mesoderm is Flk-1 single-positive, while paraxial mesoderm is PDGFRα single-positive [34, 35] and the double-positive cells correspond to undifferentiated mesodermal progenitors as well as cardiac mesoderm [34, 36], Anterior Hox gene up-regulation in response to RA was observed in the PDGFRα single-positive or Flk-1/PDGFRα double positive fractions, however the Flk-1 single positive cells showed an impaired upregulation of Hoxa3 and Hoxa4 by RA (Fig. 6D). These results indicate that different types of mesodermal cells have intrinsically different capacities to express/to regulate the Hox gene cluster with lateral plate mesoderm, from which the hematopoietic lineage arises, having both diminished Hox gene expression, and an impaired response to RA.

To characterize the RA response in fully committed, adult-repopulating hematopoietic progenitors, we also evaluated RA-dependent Hox gene regulation in Hoxa4 or Hoxb4 over-expressing cells expanded on OP9 stroma. We used purified c-Kit⁺/CD45⁺ progenitors from these cocultures for RNA analysis (approximately 50% of the cells were c-Kit/CD45 double positive, data not shown). We also determined the transcript level of doxycycline-induced Hoxb4 (and Hoxa4) in day 6 EBs, in order to establish the levels that drive the expansion of a long-term repopulating progenitor. As we expected, doxycycine induced very high expression of Hoxb4 and Hoxa4 in both EBs and OP9 cocultured cells in the respective inducible cell lines (Fig. 7). In contrast, RA induced negligible upregulation of 3’ Hox genes, consistent with our previous observation. Rarβ was induced, although to somewhat lower levels in OP9-cultured cells compared to EBs. However other RA targets (eg. Cyp26a1) [37] were robustly up-regulated demonstrating that an impaired retinoid signaling pathway is not the reason for lack of activation of the Hox genes, but rather again the changing specificity of target genes of this pathway. Together these results indicate that the Hox genes are targets of the RA pathway only in selected cell types, and that in early embryonic hematopoietic progenitors, Hox gene activation is uncoupled from RA signaling.

Figure 7.

OP9 cocultured hematopoietic progenitors display an impaired Hox gene response to RA. RNA profile of Hoxb4, Hoxa1, a3, a4, Rarβ and Cyp26a1 from day 6 EBs derived from cell lines with inducible Hoxb4 or Hoxa4 (with or without doxycycline treatment) and from OP9 cocultured sorted cells (c-Kit/CD45 cells). EBs were treated with 1 μM RA and/or 1 μg/ml doxycycline for 2 days (day 4-6) and RNA profile was determined at day 6. In addition c-Kit/CD41 double-positive cells were sorted from day 6 EBs and cultured on OP9 stromal cells for 5-6 days, thereafter cells were treated with 1 μM RA for an additional 3 days, and c-Kit/CD45 double positive cells were sorted (OP9 cocultured cells were always treated with 1 μg/ml doxycycline). EBs were generated from iHoxB4 or iHoxA4 ES cells.

Discussion

This study reports two key findings of relevance to the development of the hematopoietic system, both unexpected. The first is that retinoic acid is not necessary, and is in fact inhibitory to hematopoietic development. This is in spite of the coordinate upregulation in early mesoderm of numerous Hox genes known to promote hematopoietic progenitor self-renewal. The second is that during development of the hematopoietic lineage, although the RA signaling pathway is intact and functional, the well-studied and developmentally critical targets, the clustered Hox genes, are uncoupled from the pathway. While Hox genes are regulated by RA in other lineages, they are refractory to RA signaling in the earliest hematopoietic progenitors.

Although RA is a well established direct transcriptional regulator of several anterior Hox genes during hindbrain patterning [11], the connection between the retinoid signaling pathway and Hox gene expression was previously much less clear in the mesodermal/hematopoietic compartment. Within uncommitted mesoderm, we found coordinate up/down-regulation of 3’/5’ (respectively) members of the HoxA, B and C gene clusters by RA. The D cluster was distinct in that downregulation of 5’ Hox genes was observed, but not upregulation of 3’ genes. Although neurectoderm presents examples of upregulation, coordinate antipodal regulation across the Hox clusters by RA has not previously been reported. Whether this novel finding is unique to early mesoderm remains to be tested.

Considering the entire gene family, we observed that paralog group 4 displayed the highest expression in RA-instructed EBs. Ectopic over-expression of paralog group 4 genes is known to stimulate hematopoietic development in EBs, and profoundly enhances the in vitro expansion of hematopoietic cells co-cultured with OP9 cells [18] suggesting that RA might substitute for genetic modification to enable in vitro expansion of ES-derived hematopoietic progenitors. However, we observed the opposite: RA-treated EBs showed impaired hematopoietic development, moreover RA administration interfered with the expansion of hematopoietic cells cultured on OP9 stromal cells. Our in vitro EB differentiation model and the sorted progenitors we studied probably represent primitive (yolk sac) hematopoiesis. However Hoxb4 or Hoxa4 instructed hematopoietic progenitors demonstrate definitive hematopoietic characteristics, and pharmacological concentrations of RA negatively regulated the expansion of these cells, suggesting that RA has a generic negative influence on hematopoietic development. The literature contains conflicting results regarding the hematopoietic effects of RA. For example, RA treatment has been reported to enhance the ex vivo expansion and repopulating activity of adult murine HSCs [16, 17] and it seems that Rarγ has a function in blood stem cells since null animals have a reduced number of HSCs, and ex vivo cultured Rarγ-deficient KSL cells have an impaired repopulating capacity [38]. In contrast, recent work indicates that some of the Rarγ effects on hematopoietic cells are not cell-autonomous [39] and it has also been reported that RA synthesis inhibitor-treated HSCs had enhanced self renewal potential [40, 41]. Moreover, high concentrations of RA have been found to inhibit the proliferation of hematopoietic progenitors in vitro [42-44] and primitive hematopoiesis in vitro and in vivo [45]. Depending on the maturation state of the blood cell progenitors and the local environment, RA signaling may elicit different responses.

To obtain maximal Hox gene induction we used a pharmacological concentration of RA (100-1000 nM). In vivo, the RA concentration is probably lower [46] however if we activated the endogenous RA production by overexpressing Raldh2, elevated expression of key Hox genes was observed. Moreover, in wild-type EBs, the well-established retinoid target, Rarβ, was induced by administration of retinol and reduced by RA synthesis inhibitor, suggesting that a low concentration of RA might generated upon EB development. However by using charcoal-stripped serum supplemented with BMP4, we show that this low retinoid signaling is not necessary for hematopoietic differentiation. This finding was unexpected, based on a study suggesting that endogenous RA production positively regulates the development of yolk sac hemogenic endothelium and its hematopoietic potential [31] and a report that vitamin A-deficient (VAD) quails have impaired yolk sac hematopoiesis [32], and suggests that these reports may be measuring indirect effects. Furthermore, there are no reports describing embryonic hematopoietic abnormalities in RA receptor compound null mutant (Rarα/β-, Rarα/γ-, and Rarβ/γ) animals, although these mice show most of the malformations characteristic of the fetal VAD syndrome [47].

Our studies focused on RA-dependent regulation, however our data also indicated that the baseline level of several posterior Hox genes (Hoxa9-13) gradually increased during the course of EB development (Fig. 2 and data not shown). This is consistent with the previously described Hoxa9 expression pattern [48]. However, in the sorted hematopoietic compartment (c-Kit/CD41), we observed very low expression of posterior Hox genes (Fig. 6B and data not shown) which stands in contrast with a previous report that suggested EB-derived sorted hematopoietic cells have an elevated expression of Hox genes [49]. However in that report, bulk c-Kit positive cells were considered hematopoietic and analyzed, while our PCR results clearly demonstrate that the c-Kit single-positive cells that would predominate in this fraction express high levels of Hox genes. c-Kit single-positive cells of the day 6 EB are non-hematopoietic: they do not expand on OP9 (data not shown) and others have shown that the c-Kit/CD41 double-positive fraction contains all of the hematopoietic CFC activity [27].

More remarkable than the low baseline levels of expression, was the minimal RA-dependent Hox gene responsiveness observed in the c-Kit/CD41 double-positive cell population, and in OP9 co-cultured hematopoietic cells. In contrast to the Hox genes, the bona fide RAR targets (Rarβ or Cyp26a1) showed a marked up-regulation in these blood cell progenitors which demonstrates that the RA pathway is intact, but that Hox genes are no longer targets of the pathway. Upregulation of various Hox genes is common in acute myeloid leukemia, moreover uncontrolled over-expression of Hox genes in hematopoietic progenitors promotes the development of leukemia [23, 50]. Since hematopoietic stem cells migrate during embryogenesis, and undergo cycles of mobilization in the adult, it is probable that HSCs or other blood cell progenitors are periodically or episodically exposed to RA. In particular, the liver is a potential target organ for retinoid exposure since hematopoiesis occurs in this organ during fetal development, moreover hepatic stellate cells contain retinoid droplets [51, 52] and may release retinoids [53]. Interestingly hepatic stellate cells reside in a microenvironment similar to the endothelial bone marrow stem cell niche [54]. If RA exposure resulted in upregulation of Hox gene expression in hematopoietic cells, it could result in overproliferation at sites of RA synthesis and possibly a pro-leukemic effect in adults and an inhibition of differentiation in embryos. The developmental uncoupling of Hox genes from the RA signaling pathway would serve to preserve the Hox code from potential RA signaling centers as hematopoietic stem cells migrate or circulate during development.