HOXB4 Can Enhance the Differentiation of Embryonic Stem Cells by Modulating the Hematopoietic Niche§



This article is corrected by:

  1. Errata: HOXB4 Can Enhance the Differentiation of Embryonic Stem Cells by Modulating the Hematopoietic Niche Volume 30, Issue 6, 1311–1312, Article first published online: 16 May 2012

  • Authors contributions: L.M.F., M.J., R.A.A., and J.M.B.: designed the research; M.J., R.A.A., A.H.T., J.A.W., S.A.M.G.-K., K.K., and H.S.: collected and assembled the data; M.J., L.M.F., H.S., and O.H.: analyzed the data; J.M.B. and N.H.: provided essential reagents and analytical tools; M.J. and L.M.F.: wrote the manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS November 14, 2011.


Hematopoietic differentiation of embryonic stem cells (ESCs) in vitro has been used as a model to study early hematopoietic development, and it is well documented that hematopoietic differentiation can be enhanced by overexpression of HOXB4. HOXB4 is expressed in hematopoietic progenitor cells (HPCs) where it promotes self-renewal, but it is also expressed in the primitive streak of the gastrulating embryo. This led us to hypothesize that HOXB4 might modulate gene expression in prehematopoietic mesoderm and that this property might contribute to its prohematopoietic effect in differentiating ESCs. To test our hypothesis, we developed a conditionally activated HOXB4 expression system using the mutant estrogen receptor (ERT2) and showed that a pulse of HOXB4 prior to HPC emergence in differentiating ESCs led to an increase in hematopoietic differentiation. Expression profiling revealed an increase in the expression of genes associated with paraxial mesoderm that gives rise to the hematopoietic niche. Therefore, we considered that HOXB4 might modulate the formation of the hematopoietic niche as well as the production of hematopoietic cells per se. Cell mixing experiments supported this hypothesis demonstrating that HOXB4 activation can generate a paracrine as well as a cell autonomous effect on hematopoietic differentiation. We provide evidence to demonstrate that this activity is partly mediated by the secreted protein FRZB. STEM CELLS 2012; 30:150–160.


Hematopoietic differentiation of mouse embryonic stem cells (ESCs) in vitro is a powerful tool to model early hematopoietic development and to generate cells for regenerative medicine. A number of strategies have been developed to increase the production of hematopoietic cells from ESCs and to enhance the functional properties of the hematopoietic progenitors that are generated. These include overexpression of the homeobox transcription factors, Hoxb4 (NM_010459) and Cdx4 and coculture on a variety of stromal cell lines [1–4]. The mechanism of action of these induction strategies is not fully understood but it seems reasonable to assume that stromal cells act by providing the appropriate microenvironment for the induction and proliferation of hematopoietic progenitors, and that transcription factors act by switching on the appropriate genetic program in hematopoietic precursors. Indeed, there is a body of evidence to indicate that overexpression of Hoxb4 exerts its effects by enhancing the proliferation and self-renewal of the hematopoietic progenitor cells (HPCs) [5–8].

We noted that Hoxb4 is expressed in the primitive streak mesoderm of the gastrulating embryo prior to the emergence of HPC [9]. Thus, we hypothesized that Hoxb4 might be able to promote hematopoietic differentiation of ESCs by altering expression of genes at the mesoderm stage of differentiation prior to the emergence of HPC.

Although differentiating embryoid bodies (EBs) do not have the complex cell movements and organization of the gastrulating embryo, they do exhibit a wave of brachyury expression [10, 11], and polarized regions of β catenin signaling have been described comparable to the primitive streak during gastrulation [12]. Differences in the expression of bone morphogenetic protein 4 (BMP4), and its soluble antagonists (e.g., Noggin), in the nascent mesoderm establish a BMP gradient that subsequently patterns the mesoderm into subtypes. High BMP4 levels have a posteriorizing effect, promoting lateral plate mesoderm [13] from which hematopoietic cells are derived [14–16] whereas in the more anterior streak, the inhibition of BMP4 promotes paraxial mesoderm that gives rise to somites and subsequently, bone, muscle, cartilage, and mesenchyme.

To test the hypothesis that HOXB4 could alter mesoderm fate during ESC differentiation, we generated ESCs with inducible HOXB4 activity. We show here that hematopoietic differentiation is increased when HOXB4 is activated early in differentiating ESCs prior to the emergence of HPC when the mesoderm is forming. Genes that are upregulated in response to HOXB4 activity have been identified from this stage of the differentiation process by microarray, and a novel subset of genes associated with paraxial mesoderm patterning was revealed. ESC mixing experiments demonstrated for the first time, to our knowledge, that HOXB4 can generate a paracrine effect on hematopoietic differentiation, and we provide evidence to show that this effect could be mediated in part by expression of Frzb that encodes a secreted protein previously associated with WNT signaling. Therefore, we propose a novel mechanism for HOXB4-mediated hematopoietic induction in ESCs.



Constructs were generated that used the CAG promoter to drive expression of the Hoxb4ERT2 fusion cDNA followed by an internal ribosomal entry site element and the puromycin resistance gene (pCAG Hoxb4ERT2IP). Frzb was TOPO (Invitrogen, http://www.invitrogen.com) cloned with a 3′ flagtag from an RT-PCR product of Hoxb4ERT2ESCs + 4-hydroxytamoxifen (4-OHT) and subcloned into pCAG/GW/IP using Gateway recombination (Invitrogen). Sequences were verified, and plasmids expressed proteins of the predicted size by Western blot (data not shown). Hoxb4VP16, fusion with the transactivation domain of the Herpes simplex VP16 [17, 18], was generated by PCR as two mutants: Hoxb4(N266P)VP16/Pcs2+ (Mut1) and Hoxb4(d266)VP16/Pcs2+ (Mut2) that had mutations in an asparagine (residue 266) in the third helix of the homeodomain, which functions in DNA recognition [19]. The previously described −2,128 to −204 2 kb Frzb promoter [20] was inserted into pGL3 Basic (Promega, http://www.promega.com) to generate Frzb-Luc. Frzb-specific shRNA was obtained from Sigma (http://www.sigmaaldrich.com, Mission shRNA SHCLNG-_NM_011356_1171s1c1).

Cell Culture

A subclone of CGR8 ESCs, known as CGR8.5, was routinely cultured in GMEM with 10% fetal calf serum (FCS) and LIF in 0.1% gelatinized plastics [21], electroporated with pCAGHoxb4ERT2IP and selected in 1 μg/ml puromycin for 10 days. Puromycin-resistant colonies were isolated and screened for HOXB4 expression by Western blot (data not shown). A subclone expressing HOXB4 and an untransfected subclone were selected for further experiments (Fig. 1). Hoxb4ERT2-expressing ESCs were stably transfected with plasmids to express Frzb-specific shRNA and selected in G418 (Supporting Information Fig. S6). These clones were screened for similarity to wild-type in self-renewal and differentiation assays. Cycloheximide was used at a final concentration of 100 μg/ml that has been used previously in ESCs without toxicity [22] at a dose sufficient to inhibit 95% protein synthesis [23]. Constitutively expressing eGFP ESCs [24] were used in cell mixing experiments. NIH-3T3 cells were cultured in Dulbecco's modified Eagle's medium and 10% FCS and transfected with pCAGFrzbFlagIP with Lipofectamine (Invitrogen) and selected with puromycin for 14 days. A pool of transfected cells were verified as expressing FRZB protein by Western blot using anti-FRZB antibody (AF592, R+D Systems, http://www.rndsystems.com; Supporting Information Fig. S4) and were irradiated at 50 Gy and used as a feeder layer for ESC coculture as described previously [25]. Human embryonic stem cells (RC7) were generated by Roslin Cells (http://www.roslincells.com) and cultured in StemPro medium (Invitrogen) plus bFgf (8 ng/ml; Invitrogen) on CELLstart matrix (Invitrogen)–coated six-well dishes (Costar, Corning, www.corning.com). pCAGHoxb4ERT2IP was linearized with PvuI, and single-cell suspensions of RC7, which had been treated with TripLE Select (Invitrogen) and 10 μM Rock Inhibitor (Calbiochem Y 27632), were electroporated using standard electroporation conditions [26]. Transfectants were selected with 0.3 μg/ml puromycin (Sigma P8833) for 10 days, and Hoxb4-expressing clones were isolated and screened by quantitative RT-PCR (qRT-PCR).

Figure 1.

Verification of the HOXB4-ERT2 induction system. Western blot analysis of cytoplasmic and nuclear protein lysates prepared from embryonic stem cells (ESCs) stably transfected with pCAG Hoxb4ERT2IP treated with 4-OHT for 0, 30, and 60 minutes (A) and their quantification using ImageJ; relative HOXB4 expression is represented as a ratio with β tubulin (cytoplasmic) and OCT4 (nuclear) (B). A representative Western blot of nuclear HOXB4 and Histone H3 expression 1, 2, and 3 days after removal and washing out of 4-OHT (C), and the quantification of three independent experiments using ImageJ (error bars represent standard deviation). (D): Induction of HOXB4 activity by addition of 800 nM 4-OHT from day 1 to day 6 of differentiation in an ESC line expressing the HOXB4-ERT2 fusion protein results in an induction in the number of hematopoietic colonies (E). Experiments were performed at least three times and error bars represent standard deviations. (*, p < .05 by Mann–Whitney U test). Abbreviations: CFU-GM, colony-forming unit–granulocyte macrophage; EBs, embryoid bodies; 4-OHT, 4-hydroxytamoxifen.

Hematopoietic Differentiation

EBs were formed by hanging drops in the presence of LIF for 2 days, washed in PBS, placed in suspension culture for 1 day in the absence of LIF, and then plated onto gelatinized flasks for a further 6 days. On day 6 of culture, EBs were disaggregated, and 1 × 104 and 5 × 104 cells were plated into methylcellulose culture (Stem cell Technologies, http://www.stemcell.com/) with the addition of stem cell factor (SCF), interleukin 3 (IL-3), erythropoietin, and IL-6 for 10 days before hematopoietic CFCs were scored by light microscopy. 4-OHT (H6278, http://www.sigmaaldrich.com) was dissolved in ethanol and added to cultures from day 1 at 200 nM when they were plated onto gelatin.

Western Blot

Nuclear and cytoplasmic extracts were isolated using the NE-PER kit (http://www.piercenet.com/) according to the manufacturer's instructions and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and Western blotting. Membranes were probed with rat anti-HOXB4 monoclonal antibody (I12 [27] obtained from the Developmental Studies Hybridoma Bank (DSHB) developed under the auspices of the National Institute for Child Health and Development and maintained by the University of Iowa, Department of Biology, Iowa City, IA). Membranes were incubated with goat anti-rat HPR secondary (Santa Cruz, http://www.scbt.com/antibody.html) antibody before adding ECL substrate and autoradiography. Expression was quantified by scanning under exposed films and densitometry using Image J (http://rsbweb.nih.gov/ij/). Loading controls used were β tubulin (E7, DSHB, http://dshb.biology.uiowa.edu/) for cytoplasmic extracts and Oct4 (C-10, Santa Cruz, http://www.insightbio.com) and Histone H3 (ab1791, Abcam, http://www.abcam.com) for nuclear extracts. Ratios of HOXB4:loading controls were used to compare expression levels.


RNA was extracted from controls and HOX4B ERT2-expressing day 3 EB using RNeasy Total RNA Isolation kit with DNase digestion (QIAGEN; http://www1.qiagen.com/). Biotinylated cRNA was synthesized with Perkin-Elmer's nucleotide analogs using the Ambion's MEGAScript T7 kit (http://www.ambion.com/). After fragmenting of the cRNA for target preparation using the standard Affymetrix protocol, 15 μg fragmented cRNA was hybridized for 16 hours at 45°C to Mouse Genome 430 Version 2 array. Arrays were washed and stained with streptavidin–phycoerythrin in the Affymetrix Fluidics Station 400 and further scanned using the AFFYMETRIX GeneChip Scanner 3000 7G (http://www.affymetrix.com). The image data were analyzed with GCOS 1.4 using Affymetrix default analysis settings and global scaling as normalization method as described [10]. Three replicate RNA preparations from the four conditions (controls or HOX4B ERT2 with or without 4-OHT treatment) were processed and hybridized onto 12 Mouse Genome 430 Version 2 Arrays (Affymetrix), which includes 45,101 probe sets. After robust microarray analysis normalization and exclusion of low or not expressed genes (maximum log(2) expression over all conditions and replicates <6), a parametric analysis of variance (F tests), a two-way analysis of variance (strain vs. treatment), and pairwise comparisons using the Student's t test was performed.


Genes that were upregulated in response to HOXB4 activity were verified using qRT-PCR which is described in detail in Supporting Information Protocol S4, and primer sequences are shown in Supporting Information Table S3.

Transient Transfection and Dual Luciferase Reporter Assays

A total of 1 × 105 ESCs were plated into gelatinized six-well dishes in GMEM +10% FCS +LIF. Cotransfections of effector plasmids, luciferase reporter plasmids, and an internal control plasmid (pEFBos/renilla luciferase) were performed as described by Brickman et al. [17]. Duplicate cultures were averaged, and ratios of firefly luciferase reporter:renilla luciferase control were compared with a reference sample to give fold induction. Effector plasmids were pCAGGscIP, pCAGFrzbFLAGIP, pCS2+ plasmids to express β catenin, Hoxb4, Hoxb4VP16, or Hoxb4 mutants Hoxb4(N266P)VP16, Hoxb(4delN266)VP16). Canonical Wnt/β-catenin/TCF reporters Topflash(3xTCF), Fopflash, super(8xTCF)Topflash, and superFopflash are previously described [28]. The −2,128 to −204 Frzb promoter/pGL3 (Frzb-luc) reporter plasmid was used as a readout of Frzb promoter activity.


Establishment of an Inducible HOXB4 System

We established a system to conditionally activate HOXB4 in mouse ESCs by generating a fusion protein between HOXB4 and the 4-OHT-sensitive mutant estrogen receptor ERT2 as described [25]. HOXB4-ERT2 is detected in the nucleus after 30 minutes (Fig. 1A, 1B), and a fivefold increase in the nuclear HOXB-4ERT2 is observed 5 hours after the addition of 4-OHT (Fig. 1C). When 4-OHT is washed out of the culture, nuclear HOXB4-ERT2 levels fall to baseline levels within 24 hours (Fig. 1C, 1D). Hematopoietic differentiation of HOXB4-ERT2-expressing ESCs was induced on addition of 4-OHT (Fig. 1E). Number of CFU-M and CFU-mix colonies were increased ninefold, CFU-GM and Ery/Mac colonies were increased eightfold, and there was a 11-fold increase in the number of secondary EBs with associated hematopoietic activity (burst EBs). No 4-OHT-dependent increase in hematopoietic differentiation was observed in cultures of control CGR8.5 ESCs.

Hematopoietic Progenitors First Appear at Day 5 in Control Differentiating Cultures

Using brachyury (T) gene expression by qRT-PCR, we defined the peak of mesoderm production to be day 4 in differentiating cultures of wild-type cells (Fig. 2A). CFU-mix colonies were not detected in the differentiating cultures until day 5 in both control and Hoxb4ERT2-expressing ESCs, and the appearance of these colonies correlated with expression of the progenitor markers VE cadherin (Cdh5) and c-Kit (Kit) (Fig. 2A). The CFU-mix colonies are the nearest in vitro equivalent to an HPC and provide a functional readout of progenitor activity. Taken together, our data suggest that there were no HPCs detectable in the culture system before day 5 of EB differentiation. However, we cannot exclude the possibility that some HPCs might be present that are not detected by this assay or are present in insufficient numbers to be detected.

Figure 2.

Activation of HOXB4 prior to HPC emergence induces hematopoietic differentiation. Quantitative RT-PCR analysis of expression of Brachyury (T), VE cadherin (Cdh5), c-Kit (Kit), and Flk-1 (Kdr) expression in wild-type ESCs from days 1–6 of differentiation compared with the appearance of multipotent, CFU-mix colonies (A). Hematopoietic colonies were assayed in day 6 HOXB4-ERT2-expressing EBs which had received a 1, 3, or 6-day pulse of 200 nM 4-OHT (shown by red arrow) (B, C). Experiments were performed three times and error bars are standard deviations. (*p, < .05 by Mann–Whitney U test). Abbreviations: CFU, colony-forming unit; EB, embryoid body; RT-PCR, reverse-transcription polymerase chain reaction; 4-OHT, 4-hydroxytamoxifen.

A Pulse of HOXB4 Activity Prior to HPC Generation Increases Hematopoiesis

We tested whether a pulse of HOXB4 activity administered before HPC arise could result in an increase in hematopoietic differentiation. 4-OHT (200 nM) was added at day 1 of differentiation and washed out on day 2 or day 4 of differentiation, thus providing a 1- or 3-day pulse, respectively (Fig. 2B). These pulses were both prior to the peak of brachyury gene expression. Washout experiments indicated that HOXB4 was no longer active in the system by the time HPCs (as defined by appearance of CFU-mix colonies in methylcellulose culture) were generated (Fig. 1A). These short pulses of 4-OHT resulted in a significant (p < .05) induction of hematopoietic colony formation that was subsequently assayed at day 6 of differentiation (Fig. 2C). A corresponding increase in Hbb-βh1 embryonic globin and Hbb-b1 adult globin expression was also observed in these cultures at day 6 after a short pulse of 4-OHT (data not shown).

Genes That Are Upregulated in Response to HOXB4 Include Genes Associated with Mesoderm Formation and Patterning

To identify genes that were differentially expressed on HOXB4 activation in our system, we performed Affymetrix microarray analysis on RNA isolated from HOXB4-ERT2-expressing and wild-type control ESCs that were differentiated as EBs for 3 days in the presence or absence of 4-OHT (days 1–3). This time point was chosen to capture the genes that are upregulated by HOXB4 in prehematopoietic cells. Four-way comparisons were made between the two cell lines and the two treatment conditions. Genes that were upregulated by 4-OHT in control cells were excluded from the analysis, as were duplicates and nonannotated genes. A total of 273 genes were upregulated ≥1.5-fold (p < .05) in Hoxb4ERT2 ESCs plus 4-OHT compared with control cells plus 4-OHT (Fig. 3A). Of these 273 genes, 23 were upregulated ≥1.5-fold (p < .05) in Hoxb4ERT2 ESCs plus 4-OHT compared with Hoxb4ERT2 ESCs minus 4-OHT. Given that the 23 were a subset of the 273 genes, we assume that the difference in number between the two comparisons likely reflects some leakiness in this tamoxifen-inducible expression system. Array data have been deposited in Array Express (http://www.ebi.ac.uk/microarray-as/ae/) number E-MTAB-547.

Figure 3.

Array data. A total of 273 genes that were regulated by HOXB4 activity were identified by microarray analysis of Hoxb4ERT2-expressing day 3 embryoid bodies (EBs) + 4-OHT compared with day 3 wild-type controls with 4-OHT. Of the 273 genes, 23 were also upregulated in Hoxb4ERT2 EBs +T compared with Hoxb4ERT2 EBs − 4-OHT (A). The expression of genes associated with the different mesoderm types are shown (B, C, D). The top 20 genes which were significantly changed ≥1.5-fold (p < .05 t test) in Hoxb4ERT2 ESCs plus 4-OHT compared with control cells plus 4-OHT and the fold increase as determined by quantitative RT-PCR is shown for some genes (E). Abbreviation: 4-OHT, 4-hydroxytamoxifen; RT-PCR, reverse-transcription polymerase chain reaction.

Gene enrichment analysis of gene ontology (GO) terms (http://www.geneontology.org) was performed for the 273 genes that were significantly upregulated in HOXB4-ERT2 ESCs compared with control cells in the presence of 4-OHT. Signaling pathway GO IDs which were enriched were three WNT signaling, one TGFβ signaling, three BMP signaling, and three Notch signaling. Interestingly, two gastrulation- and three mesoderm-related GO IDs were also enriched (Supporting Information Table S1). DAVID analysis (The Database for Annotation, Visualization and Integrated Discovery, http://david.abcc.ncifcrf.gov/) of the 273 genes induced by HOXB4 activity indicated enrichment of cadherins and transcription factors (Supporting Information Table S2).

Interestingly, the top five genes that were upregulated by HOXB4 in our analysis (Hoxb4-ERT2 ESCs compared with control cells in the presence of 4-OHT) are genes that are expressed in primitive streak mesoderm of the developing mouse embryo (Fig. 3E). The apelin receptor (Agtrl1 or APJ) is induced 13.5-fold and is expressed in the primitive streak and later in the vasculature of the developing embryo [29]. Homeobox gene Lhx1 (Lim1) is induced 10.7-fold and is involved in cell movements during gastrulation after the streak has formed [30, 31]. Frzb (Sfrp3) is induced 7.2-fold and is expressed in the streak and later in paraxial mesoderm [32]. Cyp26a1, the gene that encodes the retinoic acid-metabolizing enzyme is induced 7.2-fold and is expressed in primitive streak and later in endoderm and tail bud mesenchyme [33], and Cdh11 (bone cadherin) is expressed in primitive streak and paraxial mesoderm [34].

Genes involved in the formation of nascent mesoderm such as brachyury, Wnt3, Wnt3a, Mixl1, and Bmp4 were not upregulated as a result of HOXB4 activity, while early streak genes associated with cell movements after the mesoderm has formed (Mesp1, Evx1, Fgf8, Agtrl1, Lhx1, and Eomes) are all upregulated (3.7-, 2.4-, 1.5-, 13.5-, 10.5-, and 2.2-fold, respectively; Fig. 3B).

Given that we had observed an increase in hematopoietic differentiation, it was therefore surprising that genes expressed in lateral plate mesoderm which gives rise to the hematopoietic and endothelial lineages were unaffected by HOXB4 activation (Fig. 3D). However, we noted that genes associated with paraxial mesoderm (Tbx6, Frzb, Dll1, Dll3, Foxc1, Fst, and Noggin) were all significantly upregulated (1.7-, 7.2-, 1.5-, 1.7-, 2.0-, 2.0- and 2.0-fold, respectively; Fig. 3C). Therefore, we considered that HOXB4 might be inducing the expression of genes associated with the developing hematopoietic microenvironment. Consistent with this hypothesis, we noted an upregulation of secreted factors, SCF and VEGFA (2.8- and 1.9-fold respectively, data not shown), the ligands for the cell surface receptors c-KIT and FLK1, respectively, that are expressed on hematopoietic progenitors in the lateral plate mesoderm.

HOXB4 Can Act in a Paracrine Manner to Induce Hematopoietic Differentiation

If HOXB4 is able to alter the hematopoietic microenvironment in differentiating ESCs, we predicted that it could have an effect on hematopoietic differentiation in a paracrine manner. To test this hypothesis, we generated chimeric EBs by mixing HOXB4-ERT2 ESCs with an ESC line that constitutively expressed eGFP (Fig. 4A). In these experiments, we sought to determine whether activation of HOXB4 by 4-OHT could induce hematopoietic differentiation in a GFP+ neighboring cell that did not overexpress HOXB4. After 6 days of EB differentiation in the presence or absence of 4-OHT, the chimeric EBs were disaggregated and the hematopoietic activity of fluorescence-activated cell sorting isolated GFP+ cells was determined in methylcellulose colony-forming assays. (Sorting details are shown in Supporting Information Fig. S5.) In keeping with our baseline data for controls (Fig. 1), GFP+ cells taken from control or nonactivated Hoxb4 chimeric EBs generated ≤20 hematopoietic colonies (Fig. 4B). When HOXB4 was activated with 4-OHT, a fourfold increase in the number of hematopoietic colonies was generated from GFP+ ESCs compared with nonactivated EBs. These data suggest that HOXB4 can act in a paracrine, non-cell autonomous manner to increase hematopoietic activity in neighboring cells. These secreted factors could either act by inducing hematopoietic commitment or by facilitating the proliferation of existing hematopoietic progenitors and our data cannot formally distinguish between these two interpretations.

Figure 4.

HOXB4 can exert its inducting effect in a non-cell autonomous manner. Chimeric EBs were generated from embryonic stem cells (ESCs) constitutively expressing eGFP and either Hoxb4ERT2-expressing ESCs or CGR8.5 wild-type ESCs (A). These chimeric EBs were cultured for 6 days either + or −4-OHT before disaggregating at day 6 and fluorescence-activated cell sorting the GFP+ and GFP− cells. Sorted GFP+ (B) cells were plated into methylcellulose hematopoietic colony assay and scored on day 14 (n = 3). (Details of sorted cells are shown in Supporting Information Fig. S5.) ESCs constitutively expressing EGFP were cocultured on Frzb overexpressing NIH-3T3 cells before dissagregating and scoring GFP+ colonies in a hematopoietic colony assay (n = 2) (C). (Details of FRZB expression in of pCAGFrzbflagIP transfected NIH-3T3 cells is shown in Supporting Information Fig. S6.) Frzb expression in HOXB4 ERT2-expressing ESCs was blocked by stably transfecting a Frzb-specific shRNA (D). Induction of HOXB4 activity with 4-OHT and the resulting Frzb expression and hematopoietic differentiation are shown in D and E, respectively (n = 3). Differentiations were from day 1–6 ± 4-OHT as described in Materials and Methods section. Abbreviations: CFU, colony-forming unit; EBs, embryoid bodies; eGFP, enhanced green fluorescent protein; 4-OHT, 4-hydroxytamoxifen; shRNA, short hairpin RNA.

NIH-3T3 Cells Expressing FRZB Can Induce Hematopoietic Differentiation

One of the secreted factors that had the highest level of gene induction in our analysis was Frzb, a component of the WNT/βcatenin signaling pathway (Fig. 3C, 3E). To assess whether this factor alone could induce hematopoietic differentiation and thus potentially mediate the Hoxb4 activity, NIH-3T3 cells were stably transfected with a plasmid to express FRZB. A pool of Frzb transfectants was verified by Western blot and qRT-PCR (Supporting Information Fig. S6), grown to confluence in monolayers, irradiated and used as a feeder layer for coculture of differentiating ESCs. When EGFP-marked ESCs were cocultured on the FRZB-transfected 3T3, there was a marked increase in the number of multipotent progenitors produced compared with coculture on untransfected NIH-3T3 layers, indicating that FRZB expression could have a paracrine prohematopoietic effect on differentiating ESCs (Fig. 4C) and therefore a candidate as a mediator of the paracrine effects of HOXB4.

Knockdown of Frzb by shRNA Blocks Hematopoietic Differentiation

To investigate whether FRZB could be a mediator of the hematopoietic inductive effect of HOXB4, we transfected HOXB4-ERT2 ESCs with a prescreened Frzb shRNA construct (Supporting Information Fig. S7) and selected two transfected subclones (FrzbshRNA1 and 2) for further study. Unfortunately, HOXB4-ERT2 ESCs transfected with control shRNA could not generate stable ESC lines, and so comparisons were made with the parental HOXB4-ERT2 ESCs. We demonstrate that the expression of the Frzb shRNA significantly inhibited the expression of Frzb RNA on HOXB4 induction (Fig. 4D) and that this correlated with a significantly reduced hematopoietic activity (Fig. 4E). HOXB4-ERT2 and Frzb shRNA1 + 2-expressing ESC lines all differentiated into beating cardiomyocytes with no difference in numbers of beating EBs between ±4-OHT and no difference between the cell lines (data not shown). These data support our hypothesis that Frzb could be a mediator of the prohematopoietic non-cell autonomous activity of HOXB4 but in this context has no effect of differentiation per se.

Frzb Is Upregulated in Response to HOXB4

Quantitative RT-PCR confirmed that Frzb expression was significantly upregulated by the addition of 4-OHT in differentiating EBs generated from Hoxb4ERT2 but not control ESCs (Fig. 5A). A 14- and 120-fold induction of Frzb expression was observed after 24 and 48 hours of 4-OHT treatment, respectively, compared with noninduced cells on day 1 (Fig. 5A).

Figure 5.

Frzb is upregulated by HOXB4 and requires protein synthesis for its induction. (A): Expression of Frzb by quantitative RT-PCR (qRT-PCR) in HOXB4 ERT2-expressing ESCs and wild-type controls both without and with 200 nM 4-OHT for 24 and 48 hours (days 2 + 3 of differentiation, respectively). Frzb expression is shown as fold induction relative to no 4-OHT. Experiments were performed three times, and error bars represent standard deviations (*, p < .05 by Mann–Whitney U test) unless stated otherwise. (B): Plasmids to express HOXB4, FRZB, HOXB4VP16, or homeodomain mutants: Hoxb4(N266P)VP16 (Mut1) and Hoxb4(del266)VP16 (Mut2) were cotransfected into wild-type ESCs for a dual luciferase assay to detect activity of the Frzb promoter Frzb-luc. pGL-3-luc (pGL3 basic Promega) was used as a control reporter plasmid, and eGFP was used as a control effector plasmid. (C): Induction of Frzb after activating HOXB4 ± CHX is shown as fold compared with wild-type − 4-OHT. Control and Hoxb4ERT2 ESCs were cultured in monolayers either ± 200 mM 4-OHT and ± 100 μg/ml CHX protein synthesis inhibitor for 5 hours and Frzb expression was measured by qRT-PCR. Two replicate experiments were performed and error bars represent the range between the two experiments. (D): Expression of Hoxb4 by qRT-PCR in wild-type and three human RC7 Es cell clones stably transfected with a Hoxb4ERT2 transgene. Hoxb4 expression is shown as fold expression compared with wild type. (E):Frzb expression in the human Hoxb4-expressing ESC clones both + and −4-OHT. (F): Expression of Frzb and Hoxb4 by qRT-PCR are shown for mouse hematopoietic tissues and are expressed as fold expression compared with bone marrow. Abbreviations: AGM, aorta gonad mesonephros; CHX, cycloheximide; eGFP, enhanced green fluorescent protein; 4-OHT, 4-hydroxytamoxifen; PAS, para aortic splanchopleura; RT-PCR, reverse-transcription polymerase chain reaction.

Using a dual luciferase reporter assay, we demonstrated that HOXB4 is able to activate the −2,128 to −204 Frzb promoter (Frzb-Luc) when a Hoxb4-expressing plasmid was transiently cotransfected into wild-type ESCs (Fig. 5B). Luciferase activity was enhanced when ESCs were cotransfected with a plasmid encoding an “enhanced HOXB4” where HOXB4 is linked to the minimal VP16 transactivator domain [18]. In contrast, baseline levels of luciferase activity, comparable to the eGFP control transfections, were observed when constructs carried mutations (N266P and del266) in the third helix of the HOXB4 homeodomain (labeled Mut1 and Mut2), which functions in DNA recognition [35](Fig. 5B). These data strongly suggest that the homeodomain is required for HOXB4 activation of the Frzb promoter, although this interaction is not necessarily direct.

Frzb Induction Requires a Protein Intermediate

To establish whether HOXB4 could induce endogenous Frzb mRNA directly without synthesis of a protein intermediate, we induced HOXB4-ERT2 in monolayer cultures to ensure maximal penetration of the protein synthesis inhibitor, cycloheximide, used at 100 μg/ml (a dose that inhibits 95% protein synthesis [23]). A significant increase in Frzb expression was observed 3 hours after the addition of 4-OHT (data not shown) and after 5 hours, a 15-fold induction of Frzb was observed in HOXB4-ERT2-expressing ESCs (Fig. 5C). This induction was inhibited in the presence of cycloheximide suggesting that the synthesis of a labile cofactor or a protein intermediate is required for the activation of Frzb by HOXB4.

Frzb Is Upregulated by HOXB4 in Human ESCs

We have also shown that the expression of Frzb is increased when HOXB4-ERT2 is activated in human ESCs (Fig. 5D, 5E). Human ESC line RC7 (Roslin Cells) was transfected with the pCAG Hoxb4ERT2-IP construct, and puromycin-resistant transfectants were screened for expression of Hoxb4 by qRT-PCR (Fig. 5D). RC7 Hoxb4ERT2-transfected clones were stimulated with 4-OHT for 24 hours to induce HOXB4 activity. In three hESC clones tested, Frzb was induced by the addition of 4-OHT and the level of induction of Frzb correlates with the level of expression of Hoxb4 (Fig. 5E).

Hoxb4 and Frzb Are Expressed in Sites of Hematopoiesis in the Embryo

Expression of Hoxb4 and Frzb was analyzed in hematopoietic tissue from the mouse embryo to establish whether there was coexpression of these two genes in vivo. The highest level of expression of Frzb was observed in prehematopoietic para-aortic splanchnopleura at 9.5 dpc (Fig. 5F) and Hoxb4 and Frzb were also both expressed in the aorta gonad mesonephros (AGM) at day 11.5. Expression of these two genes was also detected, albeit at lower levels in bone marrow, yolk sac, and placenta (Fig. 5F).

HOXB4 Activation and Transient Transfection of Frzb Does Not Alter βcatenin Activity

Using β-catenin/TCF reporters SuperTopflash (8xTCF binding sites driving firefly luciferase) and SuperFopflash (containing mutant TCF sites) in Hoxb4ERT2 ESCs, we detected SuperTopflash but not SuperFopflash activity in dual luciferase assays reflecting activation of the canonical WNT/β catenin pathway in these cells. However, neither activation of HOXB4 with 4-OHT nor transient transfection of Frzb in wild-type ESCs resulted in detectable alterations in β catenin activity (Fig. 6A,B). This indicated that the endogenous Frzb induced as a result of Hoxb4 activity in Hoxb4ERT2 ESCs (Fig. 6A) or from a transiently transfected Frzb expression plasmid (Fig. 6B) did not activate the SuperTopflash reporter. Given the reported antagonistic effects of FRZB on the β-catenin pathway [36], a more surprising result was that activation of HOXB4 did not significantly inhibit WNT3a-mediated β-catenin (Fig. 6B).7

Figure 6.

Frzb does not inhibit β catenin activity. Wild-type embryonic stem cells (ESC) were transfected with plasmids to express Frzb, eGFP, or β catenin in addition to supertopflash and control renilla luciferase reporters before dual luciferase assay was performed (A). HOXB4 ERT2-expressing ESCs were transiently transfected with Topflash reporters and control renilla luciferase reporters and dual luciferase assay was performed either + or −4-OHT (B). Wnt3a was added at 6 ng/ml. Abbreviations: eGFP, enhanced green fluorescent protein; 4-OHT, 4-hydroxytamoxifen.

Figure 7.

Proposed model showing both the cell autonomous and paracrine effects of HOXB4. Overexpression of HOXB4 modulates mesoderm differentiation by upregulation of genes associated with paraxial mesoderm that gives rise to the hematopoietic microenvironment. Factors such as FRZB, SCF, and VEGF secreted by that microenvironment are also upregulated by HOXB4 and might serve to enhance the survival and proliferation of hematopoietic precursors. HOXB4 has also been shown to enhance the proliferation of hematopoietic precursors in a cell autonomous manner [5]. Consequences of HOXB4 overexpression are shown in red. Abbreviations: ES, embryonic stem; HPC, hematopoietic stem cell; SCF, stem cell factor; VEGF, vascular endothelial growth factor.


This study provides evidence to suggest that overexpression of HOXB4 can enhance hematopoietic differentiation during ESC differentiation in vitro by modulating the development of the intrinsic hematopoietic niche. This increased hematopoietic activity is preceded by increased expression of markers associated with paraxial mesoderm, a tissue that can provide signals to the developing embryonic hematopoietic niche (the AGM, derived from intermediate mesoderm) and which itself can develop into the adult hematopoietic niche. We show that HOXB4-expressing cells can confer an increase in hematopoietic activity in neighboring cells in chimeric embryoid bodies. Taken together, these data suggest that overexpression of HOXB4 enhances the hematopoietic differentiation of ESCs by supporting the formation of the hematopoietic niche as well as its known direct role in the expansion of the hematopoietic progenitor pool [4, 8].

HOXB4 Induces Paraxial Mesoderm and the Hematopoietic Niche

An early pulse of HOXB4 activity resulted in increased expression of genes associated with paraxial mesoderm such as Frzb, Tbx6, Cdh11, Mesp1, Fst, Dll3, FoxC1, Fgf8, and Noggin which supports our hypothesis that HOXB4 can induce the formation of the hematopoietic niche. In agreement with reports that expression of Hoxb4 increases the differentiation of human ESCs into bone [37], our preliminary data suggest that addition of tamoxifen to our HOXB4-ERT2-expressing ESCs enhanced the differentiation of mouse ESCs into the osteocyte lineage (data not shown). Genes encoding secreted factors that have been associated with the hematopoietic niche such as VEGFA and SCF were also induced by HOXB4.

Many of the genes we identified as being regulated by HOXB4 were also identified in other microarray analysis of differentiation conditions designed to promote paraxial mesoderm formation [38] [39]. For example, Frzb, Lhx1, Evx-1, Eomes, Mesp1, Cdh11, Gsc, and Cyp26a1 were identified in the PDGFRα+ subpopulation of paraxial mesoderm [38], and Tbx6, Evx1, Mesp1, Agtrl1, Sox9, and Dll3 were upregulated on addition of the BMP4 inhibitor Noggin to ESCs [39]. Noggin and Follistatin promote the formation of paraxial mesoderm by blocking BMP signaling [40], and we observed that the expression of both of these inhibitors was induced by HOXB4. The importance of paraxial mesoderm in hematopoietic development is highlighted in Zebrafish spadetail mutants that lack paraxial mesoderm due to a deficiency in Tbx16, the fish homolog of mammalian Tbx6; deficiencies in paraxial mesoderm signaling to intermediate mesoderm result in a profound reduction in hematopoietic cells [41]. One of the genes most highly upregulated by HOXB4 was the Apelin receptor, a G protein-coupled receptor recently identified as a marker for mesoderm-derived precursor for mesenchymal stem and endothelial cells [42] that are both implicated in the hematopoietic niche in vivo.

Frzb Is Upregulated by HOXB4 and Can Induce Hematopoietic Differentiation

Frzb has been shown to be upregulated by HOXB4 by our group and others [43]. Frzb induction occurred rapidly after the onset of HOXB4 activity and our mutational analysis demonstrated that the HOXB4 homeodomain is required for activation of the Frzb promoter. However HOXB4-mediated induction of Frzb was not observed when protein synthesis was inhibited indicating that either the interaction of HOXB4 on the Frzb promoter is indirect requiring the synthesis of a protein intermediate or that HOXB4 requires a labile protein cofactor for its activity. HOX proteins alone have a relatively low affinity for DNA binding in vitro [44] and several protein cofactor families, such as MEIS and PBX have been described [45]. Interestingly, Meis2 is also upregulated in Hoxb4-activated cells, and it is possible that its synthesis is required for HOXB4-mediated Frzb expression.

Hoxb4 and Frzb In Vivo

Hoxb4−/− and Frzb−/− mice display subtle hematopoietic defects indicating that, although these genes are not required for the generation of the hematopoietic program, they may play a role in the maintenance of an optimal adult hematopoietic system in vivo [46, 47]. It is possible that the subtle defects in adult hematopoiesis in these mutants are caused by the deficiencies in the establishment of the hematopoietic microenvironment during development. Coexpression of Hoxb4 and Frzb in the primitive streak and paraxial mesoderm [9, 32] and in sites of embryonic hematopoiesis (this study) supports their involvement but further work is required to define whether the development of the hematopoietic niche in vivo is dependent on interactions between these two genes.

HOXB4 and Inductive Environment

The paracrine prohematopoietic effect of Hoxb4 that we observed in cell mixing experiments was replicated in coculture of ESCs with stromal cells expressing ectopic Frzb. This demonstrates that Frzb can confer hematopoietic inductive activities to adjacent ESCs and could in part mediate some of HOXB4's activity. Additionally, HOXB4-mediated hematopoietic induction is reduced in cells where Frzb activity is knocked down using shRNA. This supports our idea that Hoxb4 could promote hematopoiesis in ESCs by inducing the differentiation of a niche or supportive cell type and that Frzb could be involved in mediating this activity. We previously demonstrated that when HOXB4-ERT2-expressing ESCs are cocultured on AGM-derived stromal cell lines, the HOXB4-mediated and stromal-mediated induction of hematopoiesis were not additive [25] and suggested that this could be due to shared mechanisms. The fact that HOXB4 is able to induce an inductive stromal microenvironment within the differentiating culture (this study) and that HOXB4-expressing ESCs have been reported to differentiate into mesenchymal stromal cells [48] could provide an explanation for this finding.

HOXB4, FRZB, and β-Catenin Activity

Frzb (Sfrp3) is a member of the secreted frizzled related protein family that has a signal sequence and a cysteine-rich domain that can bind to Wnts [36]. Although SFRPs are classically considered as inhibitors of β catenin signaling,, there are also studies indicating that FRZB can act as a WNT chaperone or can act to widen the domain of WNT signaling by altering the diffusion of WNTs [49–51]. Our data show clearly that Frzb induced by Hoxb4 activation does not alter canonical Wnt-3a-mediated β catenin signaling. This is in agreement with previous studies reporting that FRZB did not inhibit brachyury induction by Wnt-3a in differentiating ESCs [52]. It is still a possibility that FRZB could act via a noncanonical WNT pathway and could be having an effect on the diffusion of noncanonical Wnt ligands [51]. Interestingly, noncanonical Wnt signals, activated by Wnt11, has been shown to control mesoderm specification in human ESCs whereas Wnt3a promotes proliferation of hESCs-derived HPCs via the canonical pathway [53]. Our finding that Frzb expression is induced in hESCs upon HOXB4 induction might suggest that there are similarities in the mechanism of action of HOXB4-mediated hematopoietic induction between mouse and human ESCs.


In summary, we show for the first time that HOXB4 activity during early ESC differentiation can bias differentiation toward paraxial mesoderm and can induce hematopoietic differentiation of ESCs via a paracrine mechanism. We provide evidence that this paracrine, prohematopoietic activity of Hoxb4 could be mediated in part by FRZB. We propose that this novel paracrine action of HOXB4 would augment the cell autonomous role of Hoxb4 in expanding hematopoietic progenitors (Fig. 7).


This work was supported by Leukemia and Lymphoma Research, European Commission Framework 6 Fungenes Consortium, The Wellcome Trust, and University of Edinburgh College of Medicine and Veterinary Medicine. We thank Dr. Valerie Wilson, Institute for Stem Cell Research, Centre for Regenerative Medicine, Edinburgh, for helpful discussions and comments on the manuscript and Fiona Rossi and Shonna Johnston, Centre for Inflammation Research, University of Edinburgh for Flow Cytometry.


The authors indicate no potential conflicts of interest.