Molecular Functions of the LIM-Homeobox Transcription Factor Lhx2 in Hematopoietic Progenitor Cells Derived from Mouse Embryonic Stem Cells


  • Kenji Kitajima,

    Corresponding author
    1. Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan
    • Correspondence: Kenji Kitajima, Ph.D., Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, 2-1-6 Kamikitazawa, Setagaya-ku, Tokyo 156-8506, Japan. Telephone: 81-3-5316-3130; Fax: 81-3-5316-3226; e-mail: Correspondence: Takahiko Hara, Ph.D., Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, 2-1-6 Kamikitazawa, Setagaya-ku, Tokyo 156–8506, Japan. Telephone: 81-3-5316-3130; Fax: +81-3-5316-3226; e-mail:

    Search for more papers by this author
    • Author contributions: K.K.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; M.Ka.: collection and assembly of data; M.I., M.Ky.: provision of study material; T.H.: conception and design and manuscript writing.

  • Manami Kawaguchi,

    1. Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan
    Search for more papers by this author
  • Michelina Iacovino,

    1. Lillehei Heart Institute, Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota, USA
    Search for more papers by this author
  • Michael Kyba,

    1. Lillehei Heart Institute, Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota, USA
    Search for more papers by this author
  • Takahiko Hara

    Corresponding author
    1. Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan
    • Correspondence: Kenji Kitajima, Ph.D., Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, 2-1-6 Kamikitazawa, Setagaya-ku, Tokyo 156-8506, Japan. Telephone: 81-3-5316-3130; Fax: 81-3-5316-3226; e-mail: Correspondence: Takahiko Hara, Ph.D., Stem Cell Project Group, Tokyo Metropolitan Institute of Medical Science, 2-1-6 Kamikitazawa, Setagaya-ku, Tokyo 156–8506, Japan. Telephone: 81-3-5316-3130; Fax: +81-3-5316-3226; e-mail:

    Search for more papers by this author


We previously demonstrated that hematopoietic stem cell (HSC)-like cells are robustly expanded from mouse embryonic stem cells (ESCs) by enforced expression of Lhx2, a LIM-homeobox domain (LIM-HD) transcription factor. In this study, we analyzed the functions of Lhx2 in that process using an ESC line harboring an inducible Lhx2 gene cassette. When ESCs are cultured on OP9 stromal cells, hematopoietic progenitor cells (HPCs) are differentiated and these HPCs are prone to undergo rapid differentiation into mature hematopoietic cells. Lhx2 inhibited differentiation of HPCs into mature hematopoietic cells and this effect would lead to accumulation of HSC-like cells. LIM-HD factors interact with LIM domain binding (Ldb) protein and this interaction abrogates binding of LIM-only (Lmo) protein to Ldb. We found that one of Lmo protein, Lmo2, was unstable due to dissociation of Lmo2 from Ldb1 in the presence of Lhx2. This effect of Lhx2 on the amount of Lmo2 contributed into accumulation of HSC-like cells, since enforced expression of Lmo2 into HSC-like cells inhibited their self-renewal. Expression of Gata3 and Tal1/Scl was increased in HSC-like cells and enforced expression of Lmo2 reduced expression of Gata3 but not Tal1/Scl. Enforced expression of Gata3 into HPCs inhibited mature hematopoietic cell differentiation, whereas Gata3-knockdown abrogated the Lhx2-mediated expansion of HPCs. We propose that multiple transcription factors/cofactors are involved in the Lhx2-mediated expansion of HSC-like cells from ESCs. Lhx2 appears to fine-tune the balance between self-renewal and differentiation of HSC-like cells. Stem Cells 2013;31:2680–2689


Hematopoiesis is a tissue stem cell-based process by which hematopoietic stem cells (HSCs) differentiate into more than 10 types of mature blood cells [1]. HSCs have multilineage differentiation and self-renewal capabilities. During differentiation, HSCs exit G0 phase, proliferate, and differentiate into hematopoietic progenitor cells (HPCs) that have lineage-specific differentiation potentials and give rise to terminally differentiated mature blood cells. The differentiation of HSCs is elegantly controlled by environmental cues and intrinsic genetic programs. The bone marrow niche is important for regulating the self-renewal, survival, and differentiation of HSCs [2]. Conversely, intrinsic genetic programs that maintain the characteristics of HSCs are controlled by several transcription factors and epigenetic modifiers [3, 4]. Transcriptional networks comprising these factors cooperatively and antagonistically control the fate of HSCs.

Homeobox transcription factors play crucial roles in embryogenesis. The homeobox domain (HD) is a highly conserved protein motif that can bind DNA. Among which, members of the LIM-HD transcription factor protein family possess a LIM domain comprised of two zinc finger-like structures in their N-terminal region. The LIM domain recognizes a variety of transcriptional cofactors. LIM domain binding protein (Ldb) 1 and 2 modulate the molecular and biological functions of LIM-HD proteins [5, 6].

Lhx2 (also known as LH2) is a LIM-HD factor that has been originally identified in pituitary cells and in pre-B-cell lines [7, 8]. Lhx2 expression is essential in a wide variety of progenitor/stem cell populations. Lhx2-null mice revealed that Lhx2 has indispensable roles in the brain, the eye, and definitive hematopoiesis [9]. Additionally, Lhx2 acts on stem cells in hair follicles to preserve the postnatal bulge compartment [10]. Lhx2-null mouse embryos die in utero with severe anemia, suggesting that Lhx2 has a critical role in hematopoiesis [9]. Lhx2 is aberrantly expressed in several cases of human chronic myelogenous leukemia [11], suggesting that Lhx2 stimulates the growth of immature hematopoietic cells. The effects of Lhx2 on hematopoiesis have been analyzed by enforced expression of Lhx2 in mouse HSCs isolated from adult bone marrow [12], which resulted in the ex vivo expansion of engraftable HSC-like cells. In addition, when Lhx2 is introduced into mouse embryonic stem cells (ESCs) that are subsequently induced to differentiate via the embryoid body formation method, multipotent HPCs continuously proliferate [13]. These HPCs are mainly composed of c-Kit+/Sca1/lineage (KL) cells, but not c-Kit+/Sca1+/lineage (KSL) cells.

We previously explored the consequences of enforced expression of Lhx2 during hematopoietic differentiation of mouse ESCs in vitro [14]. When the OP9 coculture method is used to induce hematopoietic differentiation of mouse ESCs, expression of Lhx2 results in the expansion of KSL/KL cells. These KSL/KL cells are also amplified from mouse-induced pluripotent stem (iPS) cells. Furthermore, Lhx2-induced KSL/KL cells display long-term hematopoiesis-repopulating (LTR) activity. When transplanted into lethally irradiated congenic mice, Lhx2-induced KSL/KL cells differentiate in vivo over 4 months into multilineage hematopoietic cells such as myeloid cells, erythroid cells, megakaryocytes/platelets, and B cells. Thus, KSL/KL cells induced by retroviral transduction of Lhx2 displayed HSC-like property. However, T-cell repopulation is hardly detected, suggesting that Lhx2 inhibits T-cell differentiation [14].

Understanding the mechanisms that underlie the expansion of mouse ES-derived KSL/KL cells by Lhx2 would provide invaluable information for the future therapeutic applications of human iPS cells and human cord blood. However, the molecular mechanisms responsible for the dramatic effects of Lhx2 remain unclear. In this study, we demonstrate that overexpression of Lhx2 decreases the amount of Lmo2 (also known as rbtn2) and upregulates Gata3 expression, both of which are expressed in newly emerged HSCs in the aorta/gonad/mesonephros region of mouse embryos [15]. These changes underlie the accumulation of KSL/KL cells in vitro by overexpression of Lhx2.

Materials and Methods

Cell Culture

Mouse ESCs (RENKA, E14tg2a, Gata1-KO [16], A2Lox-cre [17]) were maintained on mitomycin C-treated mouse embryonic fibroblasts in Dulbecco's modified Eagle's medium (DMEM) (Sigma, St. Louis, MO, supplemented with 15% knockout serum replacement, 1% fetal bovine serum (FBS, Gibco-Invitrogen, Carlsbad, CA,, penicillin/streptomycin (Pen/Str), nonessential amino acids (Gibco-Invitrogen), sodium pyruvate (Gibco-Invitrogen), 2-mercaptoethanol (Sigma), and 103 U/ml ESGro (Gibco-Invitrogen). ESC lines with inducible Lhx2 (iLhx2-ESCs) and Gata3 (iGata3-ESCs) expression were generated by transfecting p2Lox-FLAG-Lhx2 or p2Lox-Gata3 into A2Lox-cre ESCs, respectively. Cre/loxP-mediated integration of FLAG-Lhx2 and Gata3 into the Hprt locus was confirmed by genomic PCR. The expression of Lhx2 and Gata3 was confirmed by immunofluorescent staining, Western blotting, and reverse transcription (RT)-PCR.

OP9 stromal cells were maintained in alfa-MEM (Gibco-Invitrogen) supplemented with 20% FBS and Pen/Str [18]. Mouse δ-like 1 (Dl1) cDNA was cloned from mouse thymus by RT-PCR. OP9-DL1 cells were generated by retroviral transduction of pMY-Dl1-IRES-EGFP followed by sorting of EGFP+ cells. Dl1 expression was confirmed by fluorescence-activated cell sorting (FACS) analysis using an anti-Dl1 antibody (BioLegend, San Diego, CA, 293T cells and Plat-E packaging cells were maintained in DMEM supplemented with 10% FBS and Pen/Str. Retrovirus production was performed by transfecting Plat-E cells with pMY-IRES-EGFP, pMY-Lhx2-IRES-EGFP, pMY-IRES-EYFP, pMY-Lmo2-IRES-EYFP, or pMY-Ldb1-IRES-EYFP using FuGene HD reagent (Promega, Madison, WI, The culture supernatants were collected and used for gene delivery [14]. Lentiviral vectors carrying EGFP and short hairpin RNAs (shRNAs) against Gata3 mRNA were transfected into 293T cells with pMD2G and pCMVR8.91 vectors [19] and supernatants were used for stable introduction of shRNA.

OP9 Coculture of ESCs

Mouse ESC lines (3 × 104 cells) were plated onto confluent OP9 cells in a 60 mm dish on day 0. Half of the culture medium was replaced on day 2 or 3. Cells were detached by treatment with 0.25% trypsin/10 mM EDTA (Sigma) on day 5, and 3 × 105 cells were plated onto fresh confluent OP9 cells or OP9-Dl1 cells in a 60 mm dish. For expansion of HSC-like cells, interleukin (IL)-6 (10 ng/ml, Peprotech, Rocky Hill, NJ, http://www.peprotech.como) and stem cell factor (SCF) (50 ng/ml, Kirin, Takasaki, Japan, were added to the culture medium from day 5. Expression of Lhx2 or Gata3 was induced in iLhx2-ESCs or iGata3-ESCs by the addition of 1 µg/ml doxycycline (dox). Hematopoietic cells induced from mouse ESCs were collected by mild pipetting, stained with biotin-lineage cocktail (Miltenyi Biotec, Auburn, CA,, phycoerythrin (PE)/Cy7-conjugated anti-c-Kit antibody (BioLegend), and allophycocyanin-conjugated anti-Sca1 antibody (BioLegend), followed by PE-conjugated streptavidin (BioLegend) or PE/Cy5-conjugated streptavidin (BioLegend), and analyzed with a FACSAria cell sorter (BD Pharmingen, Franklin Lakes, NJ, FACS analyses and sorting were carried out as previously described [14]. Retroviral transduction was carried out by spin infection as previously described [14].

Hematopoietic lineage-directed differentiation of ESCs was induced by the addition of cytokines from day 5: 10 ng/ml granulocyte macrophage colony-stimulating factor (GM-CSF) (Peprotech) for myeloid differentiation; 50 ng/ml fms-related tyrosine kinase 3 ligand (Flt3L) (Peprotech), 50 ng/ml SCF (Kirin), and 10 ng/ml IL-7 (Peprotech) for B- and T-lymphoid differentiation. T-cell induction was induced by coculture with OP9-Dl1 cells instead of OP9 cells. The differentiated cells were subjected to May-Grünwald Giemsa staining and labeled with biotin-conjugated anti-Mac-1 (BioLegend), PE-conjugated anti-CD19 (BioLegend), biotin-conjugated anti-CD8 (BioLegend), and PE-conjugated anti-CD4 (BioLegend) antibodies. Biotinylated antibodies were visualized with allophycocyanin-conjugated streptavidin (BioLegend).

Gene Expression Analysis and Reporter Assays

RT-PCR was performed as previously described [16]. Briefly, total RNA was isolated using the RNeasy mini kit (QIAGEN, Valencia, CA, For RT-PCR, cDNA was synthesized from 500 ng RNA using the PrimeScript reverse transcription kit (Takara, Shiga, Japan, and amplified with Taq DNA polymerase (Takara) according to the manufacturer's recommendations. Primer sequences are listed in Supporting Information Table S1. Real-time PCR was carried out on Light Cycler 480 (Roche Diagnostics, Indianapolis, IN, according to the manufacturer's instruction. For microarray analysis, total RNA was extracted from Lhx2-expressing KSL/KL cells induced from iLhx2-ESCs by the addition of dox (+dox) and these cells were cultured for a further 3 days without dox (−dox). The 3D-gene Mouse Oligo chip 24k (Toray, Tokyo, Japan, was used for microarray analysis. RNA was labeled and hybridized to the array chips.

The Lhx2 deletion mutants were synthesized by deleting HD region for Lhx2ΔHD (nucleotides 1–720) and deleting LIM domain for Lhx2ΔLIM (nucleotides 601–1,221). FLAG sequence was inserted into the end of N termini. To construct the LIM domain point mutations, histidine residues (amino acid number 74 and 137, respectively) of N- and C- fingers of the LIM domains of FLAG.Lhx2 were replaced to glycine residues. These Lhx2 mutants were subcloned into pMY-IRES-EGFP retroviral vector.

A genomic fragment around the human CGA promoter was cloned from 293T cells and inserted into the pGL3-Basic reporter plasmid (Promega) to generate pGL3-hCGA. pGL3-hCGA was cotransfected with pRL-TK and pMY-Lhx2-IRES-EGFP or its derivatives into 293T cells using the CaPO4 method. Luciferase activity was measured using the dual luciferase reporter assay system (Promega). The transcription efficiency was calculated according to the activity of pRL-TK. The transfection of pGL3-hCGA and pRL-TK into iLhx2-ESCs was carried out by FuGene HD transfection reagent (Promega).

Western Blot Analysis and Immunofluorescent Staining

293T cells were cotransfected with pCMV-Lhx2, pCMV-FLAG-Lmo2, and pCMV-HA-Ldb1 using the CaPO4 method. Western blotting was performed as previously described [20]. Briefly, 2 × 105 cells were collected 2 days after transfection, suspended in 100 µl of SDS-lysis buffer, and sonicated. Twenty microliters of the cell lysate was loaded onto a 10% polyacrylamide gel and transferred onto a polyvinylidene difluoride membrane. The membranes were stained with an HRP-conjugated anti-FLAG antibody (Sigma), a rat anti-HA antibody (Sigma), or a goat anti-Lhx2 antibody (Santa Cruz Biotech, Santa Cruz, CA, followed by an HRP-conjugated anti-rat IgG antibody (GE healthcare, Chalfont St. Giles, Buckinghamshire, U.K.,, or an HRP-conjugated anti-goat IgG antibody (Abcam, Cambridge, MA, A mouse anti-β-tubulin antibody (Sigma) was used as a loading control in combination with an HRP-conjugated anti-mouse IgG antibody (GE healthcare). 293T cells transfected with pCMV-FLAG-Lmo2 and pCMV-HA-Ldb1 with/without pCMV-Lhx2 were treated with 10 µM MG132 (Sigma) for 8 hours. Coimmunoprecipitation assays were carried out as previously described [20]. Quantitative analyses were carried out by LAS3000 imaging system (FUJIFILM, Tokyo, Japan,

For immunofluorescent staining of ESCs, a goat anti-Lhx2 antibody was used with Alexa564-conjugated anti-goat IgG antibody (Invitrogen). Fluorescent image was captured using IX71 microscope (Olympus, Tokyo, Japan,


Conditional Expression of Lhx2 During Mouse ESC Differentiation

The dox-mediated gene expression system (inducible cassette exchange) was used to conditionally express Lhx2 in mouse ESCs (Fig. 1A) [17]. This inducible gene expression system uses Cre/LoxP recombination to introduce a transgene into the host genome. Therefore, clonal variation due to random transgene integration into the host genome is excluded. Using this system, we established a mouse ESC line with inducible FLAG-Lhx2 expression (iLhx2-ESCs). The expression of FLAG-Lhx2 was dependent on the addition of dox as revealed by Western blotting and immunofluorescent staining using anti-FLAG and anti-Lhx2 antibodies, respectively (Fig. 1B, 1C).

Figure 1.

Generation of iLhx2-embryonic stem cells (ESCs). (A): The inducible cassette exchange system. Genes of interest can be targeted to a specific conditionally regulated locus. The transactivator rtTA and a tetracycline responsive element were introduced at Rosa26 and Hprt loci, respectively. (B): iLhx2-ESCs were cultured in the presence or absence of dox for 1 day. Lhx2 expression was detected by Western blotting with an anti-FLAG antibody. (C): iLhx2-ESCs were cultured in the presence or absence of dox for 1 day and then stained with an anti-Lhx2 antibody and an Alexa564-conjugated anti-rabbit antibody. Scale bar = 100 µm.

When ESCs were put onto OP9 stromal cells without leukemia inhibitory factor, ESCs were differentiated into mesodermal cells on day 5 of the differentiation induction. We previously showed that retroviral transduction of Lhx2 into ESC-derived differentiated cells on day 5 results in the accumulation of KSL/KL cells [14]. Therefore, we investigated whether dox-mediated conditional expression of Lhx2 induced the same phenotypes. iLhx2-ESCs were cocultured with OP9 cells for 5 days in the absence of dox. Then, cells were reseeded onto fresh OP9 cells with IL-6 and SCF in the presence or absence of dox. In the absence of dox, approximately 50% of cells differentiated into lineage-positive (Lin+) cells by day 14 (Fig. 2A). In contrast, almost all cells were lineage-negative (Lin) cells in the presence of dox. These cells contained KSL and KL cells. These data are consistent with our previous data using retroviral transduction of Lhx2 [14].

Figure 2.

Induction of c-Kit+/Sca1+/lineage (KSL)/c-Kit+/Sca1/lineage (KL) cells from iLhx2-embryonic stem cells (ESCs). (A): Effects of dox-induced Lhx2 expression on KSL/KL cell formation. iLhx2-ESCs were differentiated on OP9 cells without dox for 5 days, and then cells were recovered and reseeded on OP9 cells with IL-6/SCF in the absence (Off) or presence (D5On) of dox. Fluorescence-activated cell sorting analysis was performed on day 14. (B): Lhx2 promotes the accumulation of Lin cells. Lhx2 expression was started on day 5 and cells were analyzed on days 7 and 9. (C): c-Kit/Sca1 staining of Lhx2-expressing cells on day 9. (D, E): Effects of Lhx2 on lineage-directed differentiation. Induction of Lhx2 was started on day 5 in the presence of granulocyte macrophage colony-stimulating factor on OP9 cells (D), or in the presence of IL-7/SCF/Flt3L on OP9 cells (E, left) or OP9-Dl1 cells (E, right). Scale bar = 50 µm.

Accumulation of KSL/KL Cells by Inducible Expression of Lhx2

We next analyzed the time course of KSL/KL cell production after the induction of Lhx2 expression. Lhx2 expression was induced from day 5, and analyzed on days 7 and 9. Most cells were Lin on day 5 and in the absence of Lhx2 induction, Lin+ cells were gradually differentiated by days 7 and 9 (Fig. 2B). By contrast, Lhx2 induction inhibited the lineage differentiation, thereby Lin cells accumulated. These Lin cells on day 9 contained KSL/KL cells (Fig. 2C). However, we previously reported that Lhx2-transduced KSL/KL cells can differentiate into mature hematopoietic cells in vivo and in vitro [14]. Therefore, the effects of dox-induced Lhx2 expression on hematopoietic terminal differentiation were reinvestigated using hematopoietic cytokines. Cytokines were added to the culture medium and Lhx2 expression was induced on day 5. Myeloid cells and B cells were generated when GM-CSF and Flt3L/IL-7/SCF were added, respectively (Fig. 2D, 2E). Thus, Lhx2 did not block cytokine-induced differentiation into myeloid cells and B cells, although the efficiency of myeloid cell differentiation was reduced when Lhx2 expression was induced. Conversely, T-cell induction was severely impaired by Lhx2 in the presence of OP9-Dl1 cells (Fig. 2E), consistent with our previous in vivo data [14]. Thus, Lhx2-induced KSL/KL cells preferentially undergo self-renewal in the presence of IL-6/SCF, but retain multilineage differentiation potentials.

Lhx2-induced hematopoietic cells expanded well on OP9 stromal cells by day 20 (Fig. 3A), and the number of hematopoietic cells continued to increase throughout the time course (Fig. 3B). On day 28, KSL and KL cells were still present and were morphologically blastic (Fig. 3C). We next examined whether continuous expression of Lhx2 is required to maintain induced KSL/KL cells. Lhx2-induced KSL/KL cells were cultured with or without dox for 7 more days. In the presence of dox, most cells were KSL or KL (Fig. 3D). By contrast, all cells were Lin+ in the absence of Lhx2 (Fig. 3D). Thus, continuous expression of Lhx2 is required for the expansion of KSL/KL cells. Lhx2 expression was hardly detected in the absence of dox (Fig. 3E).

Figure 3.

Continuous proliferation of c-Kit+/Sca1+/lineage (KSL)/c-Kit+/Sca1/lineage (KL) cells in the presence of Lhx2. (A): Gross appearance of hematopoietic cells on day 20 in the absence (Off) or presence (D5On) of Lhx2 expression. Expression of Lhx2 was started on day 5. Scale bar = 100 µm. (B): Number of hematopoietic cells in the absence or presence of Lhx2. Mean values are plotted (n = 5) and the error bars show SDs. (C): Lineage/c-Kit/Sca1 and May-Grünwald Giemsa staining of Lhx2-induced cells on day 28. Scale bar = 100 µm. (D, E): Effects of Lhx2 withdrawal on KSL/KL cells. Lhx2-induced KSL/KL cells were cultured in the presence (On/On) or absence (On/Off) of Lhx2 from day 28 for 7 days. Fluorescence-activated cell sorting (D) and semiquantitative RT-PCR (E) analyses are shown. cDNAs were serially diluted (1:1, 1:3, and 1:9) and used for semiquantitative RT-PCR.

Identification of Lhx2-Target Genes

Microarray analysis was performed to identify candidate genes that are regulated by Lhx2. Lhx2-induced KSL/KL cells (+dox) were cultured in the absence of Lhx2 expression for 3 days (−dox), and the gene expression profiles of these two groups of cells were compared (accession number: GSE44778). In the presence of Lhx2, 954 genes were upregulated (more than twofold) and 1,311 genes were downregulated (less than twofold). The top 50 genes that Lhx2 upregulated and downregulated are shown in Supporting Information Table S2A, S2B, respectively. Several HSC marker genes were upregulated by Lhx2 (Supporting Information Table S2C). Of note, components of Gata transcription factor complexes (Gata3 and Tal1/Scl) and Hox transcription factors (HoxA5, HoxD4, and HoxD8) were upregulated by Lhx2 (Table 1). The downregulated genes were mainly differentiation-associated genes (Supporting Information Table S2B). Another ESC line with inducible Lhx2 expression was previously established and candidate Lhx2 target genes in HPCs were investigated by microarray analysis [21]. In this case [21], ESC-derived HPCs induced by Lhx2 were analyzed at 36, 72, and 96 hours after dox removal. According to the report, 170 genes were upregulated in the presence of Lhx2. When we compared it with our data, 26 genes were commonly upregulated by Lhx2 (Supporting Information Fig. S1).

Table 1. Transcription factors upregulated by Lhx2
GeneDescriptionRefSeq_idFold increase
  1. Abbreviation: ESC, embryonic stem cell.

Hesx1Homeobox gene Expressed in ESCsNM_01042011.9
Tcf3Transcription factor 3NM_0093329.3
Gata3GATA binding protein 3NM_0080916.4
Sox4SRY-box containing gene 4NM_0092383.9
Ebf2Early B-cell factor 2NM_0100953.6
Tal1T-cell acute Lymphocytic leukemia 1NM_0115273.0
Stat4Signal transducer and activator of transcription 4NM_0114873.6
Stat3Signal transducer and activator of transcription 32.9
Hoxa5Homeobox A5NM_0104532.8
Foxo1Forkhead box O1NM_0197392.8
Egr1Early growth response 1NM_0079132.6
Meis1Myeloid ecotropic viral integration site 12.6
Hoxd4Homeobox D4NM_0104692.4
Smad1MAD homolog 1 (Drosophila)NM_0085392.3
Evi1Ecotropic viral integration site 1NM_0079632.3
Gfi1bGrowth factor independent 1BNM_0081142.3
Hoxd8Homeobox D8XM_3553382.3
NfyaNuclear transcription factor-Y alpha2.0

Molecular Mechanisms Underlying Lhx2-Induced Expansion of HSC-Like Cells

To elucidate the molecular function of Lhx2, we made deletion mutants (Fig. 4A) lacking various regions of Lhx2 and introduced each mutant into differentiating ESCs by retrovirus-mediated gene transfer. Lhx2 mutant lacking the HD (ΔHD) or LIM (ΔLIM) domain did not amplify c-Kit+ cells (Fig. 4B), indicating that both domains are required for the expansion of KSL/KL cells. Next, we evaluated point mutants of the LIM domain. The N-finger mutant (mtLIM-N) marginally reduced c-Kit+ cell expansion, whereas the C-finger mutant (mtLIM-C) and the double mutant (mtLIM-NC) markedly reduced c-Kit+ cell expansion (Fig. 4C). The expression levels of Gata3, a candidate Lhx2 target gene identified by our microarray analysis, were positively correlated with the frequency of c-Kit+ cells in HPCs transduced with these Lhx2 mutants (Fig. 4D). However, it remains determined whether or not Lhx2 directly upregulates Gata3 mRNA via direct transcriptional activation.

Figure 4.

Analysis of Lhx2 mutants. (A): Lhx2 mutant constructs. (B, C): c-Kit staining of cells transduced with FLAG-Lhx2 or Lhx2 mutants. EGFP+ cells are shown (B). E14tg2a embryonic stem cells (ESCs) were differentiated on OP9 cells, and various retroviral vectors harboring Lhx2 or mutated Lhx2 were transduced on day 5. The transduced cells were analyzed on day 11. Numbers denote the percentage of EGFP+ cells that were c-Kit+ (C). (D): RT-PCR analysis of Gata3 expression after transduction of Lhx2 and Lhx2 mutants. (E): Activation of hCGA promoter in ESCs. iLhx2-ESCs were transfected with hCGA-luc and pRL-TK and cultured with/without dox. (F): Reporter assays of Lhx2 LIM domain point mutants. Lhx2 or Lhx2 mutants were cotransfected with the reporter plasmid hCGA-luc and pRL-TK into 293T cells. (E, F): Bars show mean values (n = 5) and error bars show SD.

To examine the capacity of each Lhx2 mutant as a transcriptional activator, we used the human CGA promoter that is used for monitoring the transcription-enhancing activity of LIM-HD transcription factors [22]. The reporter activity was significantly increased by dox addition in iLhx2-ESCs (Fig. 4E), verifying that the human CGA promoter works in mouse ESCs. Subsequent reporter assays revealed that mtLIM-N, mtLIM-C, and mtLIM-NC possessed similar transcription-enhancing activities (Fig. 4F). These data suggest that the expansion of c-Kit+ cells and the transcription-enhancing capacity of Lhx2 are not associated with each other. Therefore, we sought to identify a function of Lhx2 that is independent of the transcriptional regulation.

Degradation of Lmo2 in the Presence of Lhx2

It has been reported that LIM-HD proteins affect the stability of Lmo proteins (Fig. 5A) [23]. Therefore, we investigated the status of hematopoietic Lmo protein, Lmo2. FLAG-Lmo2 was cotransfected with increasing amounts of Lhx2 into 293T cells, and the amount of FLAG-Lmo2 was quantified by Western blotting. Quantity of FLAG-Lmo2 protein was gradually decreased when the amount of Lhx2 was increased (Fig. 5B). Addition of the proteasome inhibitor MG132 inhibited this Lhx2-induced reduction of Lmo2 protein expression (Fig. 5C), suggesting that it is a ubiquitin/proteasome-dependent protein degradation. Next, the reduction of Lmo2 protein expression in the Lhx2-transfected cells was rescued by cotransfection of Ldb1 expression vector (Fig. 5C). Differences in the amount of Lmo2 protein in each experiment were statistically significant (p < .05 by Student's t test) when normalized with the levels of β-tubulin. These data indicated that Lhx2 disrupts the Lmo2:Ldb1 complex and released Lmo2 became unstable. Presumably, overexpression of Ldb1 blocks Lmo2 degradation by promoting the Lmo2:Ldb1 complex formation, as shown in a previous study [23]. To confirm this possibility, we carried out coimmunoprecipitation assays in the presence of MG132 to prevent Lmo2 degradation. When FLAG-Lmo2 and HA-Ldb1 were cotransfected, HA-Ldb1 was successfully recovered in the anti-FLAG (Lmo2) immunoprecipitate (Fig. 5D). In the presence of Lhx2, the amount of HA-Ldb1 bound to FLAG-Lmo2 was decreased to approximately one-third of all (Fig. 5D).

Figure 5.

Lhx2 promotes degradation of Lmo2. (A): Schematic of the Lmo degradation pathway in the presence of LIM-HD. In the presence of excess LIM-HD, Ldb dissociated from Lmo and the LIM-HD:Ldb complex forms. The dissociated Lmo becomes unstable and is degraded by the E3 ubiquitin ligase Rlim. (B): Effects of Lhx2 on the level of Lmo2. FLAG-Lmo2 expression vector (4 µg) was cotransduced with 0 (−), 1 (+), 2 (++), and 4 µg (+++) of Lhx2 expression vector into 293T cells. (C): Effects of MG132 and Ldb1 on Lhx2-induced decrease of Lmo2. (D): Coimmunoprecipitation assay. Lmo2 and Ldb1 were cotransfected into 293T cells with/without Lhx2. These cells were cultured in MG132 and Lmo2 complex was immunoprecipitated by anti-FLAG antibody and the amount of Ldb1 was analyzed. (E): Effects of Lhx2 withdrawal on the level of endogenous Lmo2 in induced KSL/KL cells. Lhx2-induced KSL/KL cells were cultured in the presence (On/On) or absence (On/Off) of Lhx2 for 3 days, and the level of Lmo2 protein and mRNA was analyzed by Western blotting and real-time PCR, respectively. (F): Effects of Lmo2 on KSL/KL cell accumulation. Empty, Lmo2 and Ldb1 vectors harboring EYFP were cotransduced with the Lhx2 vector harboring EGFP on day 5 of differentiation induction of embryonic stem cells and analyzed on day 19. Lineage staining of the EGFP+/EYFP+ cell fraction was shown. (G): Effects of Lmo2 cotransduction on gene expression. cDNAs were serially diluted (1:1, 1:3, and 1:9) and analyzed by semiquantitative RT-PCR. (H): The protein expression of Lhx2 and Lmo2 in Lhx2/Lmo2 cotransduced cells. (B-E, H): Quantified analyzes were shown as bars. Error bars indicated SD (n = 3). (B, C, E, H): The amount of Lmo2 protein was normalized by β-tubulin. (D): The amount of immunoprecipitated Ldb1 was normalized by input Ldb1. (E): Lmo2 mRNA expression was normalized by β-actin mRNA. Abbreviation: LIM-HD, LIM-homeobox domain.

Next, we examined endogenous Lmo2 protein levels in the ESCs. KSL/KL cells were generated from iLhx2-ESCs by dox addition and subcultured in the absence or presence of dox for 3 days. Lower amount of Lmo2 protein was detected in the continuous presence of dox when compared with the cells without dox in the last 3 days (Fig. 5E), again demonstrating the destructive action of Lhx2 against Lmo2 protein. Conversely, Lmo2 mRNA was moderately reduced (approximately 0.7-fold) in the absence of Lhx2 based on real-time PCR analysis (Fig. 5E). This could be a negative feed-back regulation.

To clarify whether the level of Lmo2 affects the Lhx2-induced generation of KSL/KL cells, we investigated the effect of Lmo2 overexpression. ESC-derived mesodermal cells were coinfected with empty, Lmo2, or Ldb1 retroviral vectors harboring IRES-EYFP in combination with Lhx2-IRES-EGFP. Expression of Lhx2 and Lmo2 was monitored by EGFP and EYFP fluorescence, respectively. When Lhx2 and Lmo2 were cotransduced, the proportion of Lin+ cells in the EGFP+/EYFP+ cell fraction increased (Fig. 5F). Conversely, EGFP+/EYFP cells, namely Lhx2+/Lmo2 cells, were mostly KSL/KL cells in all three samples (Supporting Information Fig. S2A). Cotransduction of Ldb1 slightly increased the proportion of Lin cells (Fig. 5F). This might be due to the enhancement of Lhx2 activity by Ldb1, since Lhx2 and Ldb1 work together. Thus, the level of Lmo2 is crucial for the Lhx2-induced expansion of KSL/KL cells. To clarify whether Lmo2 affects the initial emergence of KSL/KL cells, or whether Lmo2 disrupts the self-renewal of KSL/KL cells, Lmo2 was introduced into cells after transduction of Lhx2. Lhx2 was first introduced on day 5 and Lmo2 was subsequently transduced on day 9 or 12. The proportion of Lin+ cells still increased in both cases (Supporting Information Fig. S2B), although percentage of the Lin+ population steeply decreased when Lmo2 was transduced on day 12. These data suggest that Lmo2 inhibits the Lhx2-mediated expansion of KSL/KL cells when present in an early-stage of the population establishment. The Lin+ cells generated by cotransduction of Lhx2 and Lmo2 were mainly Mac-1+ (Supporting Information Fig. S2C).

Inhibition of Mature Hematopoietic Cell Differentiation by Gata3

Microarray analysis revealed that Gata3 and Tal1/Scl mRNAs were upregulated by Lhx2 overexpression. These data were confirmed by semiquantitative RT-PCR (Fig. 3E). Gata3 is expressed in adult HSCs and in the aorta/gonad/mesonephros regions in which HSCs emerge [15, 24]. Tal1/Scl is an interaction partner of Lmo2, and is required for the HSC development in mouse embryos [25, 26]. Therefore, we focused on these molecules. When Lmo2 and Lhx2 were cotransduced, the expression of Gata3 but not Tal1/Scl was reduced (Fig. 5G). We confirmed the increase of Lmo2 protein in these experiments (Fig. 5H). Thus, Gata3 expression may be more related to the Lhx2-induced expansion of KSL/KL cells. We newly generated an ESC line carrying an inducible Gata3 expression cassette (iGata3-ESCs) and investigated the effects of Gata3 overexpression. When Gata3 was expressed from day 5, hematopoietic cell differentiation was accelerated (Fig. 6A). Induction of Gata3 expression resulted in the accumulation of Lin cells (Fig. 6B). However, only a small number of KSL/KL cells were generated in this case. These data indicate that overexpression of Gata3 inhibits the hematopoietic differentiation (Lin to Lin+), but is not sufficient for the expansion of KSL/KL cells. Next, we performed Gata3-knockdown experiments using a specific shRNA (Supporting Information Fig. S3A). First, lentiviral vectors carrying EGFP in combination with Gata3 shRNA or control shRNA (Supporting Information Fig. S3B) were transduced into iGata3-ESCs and Gata3 expression was induced by dox. Gata3 expression was decreased by Gata3 shRNA, but not by control shRNA (Supporting Information Fig. S3C). Next, KSL/KL cells generated from iLhx2-ESCs were infected with these lentiviral vectors. After the transduction, numbers of empty vector- and control shRNA transduced EGFP+ cells were increased (Fig. 6C). In contrast, Gata3 shRNA-transduced cells were poorly proliferated (Fig. 6C). We confirmed that Gata3 mRNA was suppressed by Gata3 shRNA but not by control shRNA (Fig. 6D). In addition, generation of KSL/KL cells in EGFP+ cells was inhibited by Gata3 shRNA, but not by control shRNA (Fig. 6E). Thus, Gata3 is required for self-renewal of the Lhx2-induced KSL/KL cells.

Figure 6.

Gata3 function on Lhx2-induced KSL/KL cells. (A): Gross morphology of immature hematopoietic colonies on day 7. iGata3-ESCs were differentiated and Gata3 expression was started on day 5. Scale bar = 50 µm. (B): Inhibition of Lin+ cell differentiation on day 13 by Gata3 expression. (C): Cell number of Gata3 shRNA-transduced KSL/KL cells induced by Lhx2. Lentiviral vectors carrying EGFP and Gata3 shRNA or control shRNA were transduced into the Lhx2-induced KSL/KL cells. Dots show mean values (n = 5) and error bars show SD. (D): RT-PCR analysis of Gata3 expression in shRNA-transduced cells. (E): Decrease of KSL cells by Gata3 shRNA. (F): Schematic illustration of the proposed molecular functions of Lhx2 in the Lhx2-mediated expansion of ES-derived KSL/KL cells. Abbreviation: ESC, embryonic stem cell.


We previously showed that when Lhx2 is introduced into mesodermal cells derived from mouse ESCs/iPS cells, HSC-like cells are robustly expanded [14]. These HSC-like cells acquire in vitro self-renewal ability in the presence of IL-6/SCF and OP9 cells without losing their multipotential differentiation ability. In this study, we focused on the role of Lhx2 in the self-renewal of KSL/KL cells. Here, we revealed that Lhx2 inhibits the hematopoietic differentiation of Lin cells into Lin+ cells, and this activity is tightly associated with the expansion of KSL/KL cells. This inhibition of hematopoietic differentiation was in part caused by a decrease in the level of Lmo2 protein, which may occur after disruption of the Lmo2:Ldb1 complex by Lhx2 overexpression (Fig. 6F).

In the presence of a higher amount of Lhx2, several transcription factors were upregulated in ES-derived KSL/KL cells. Among them, Gata3 was involved in the inhibition of hematopoietic differentiation. It remains unknown whether the upregulation of Gata3 is mediated through direct transactivation by Lhx2. However, Lhx2 N-finger point mutants (mtLIM-C, mtLIM-NC) exhibited a weaker activity for the expansion of c-Kit+ cells despite having a normal transcription factor activity. Therefore, it is likely that the upregulation of Gata3 is an indirect event. Since our data suggest that Lmo2 can regulate Gata3 expression, the degradation of Lmo2 might result in the upregulation of Gata3. In fact, overexpression of Lmo2 attenuated the Lhx2-mediated upregulation of Gata3 mRNA (Fig. 5G). These findings uncovered a novel molecular pathway underlying the Lhx2-induced expansion of KSL/KL cells. Notably, microarray data revealed that hprt1/hprt was upregulated by Lhx2 (Supporting Information Table S2A). In iLhx2-ESCs, dox-inducible Lhx2 expression cassette was located in the hprt1/hprt locus. Hence, it might be possible that dox addition would activate the adjacent hprt1/hprt gene. However, in ESCs, it was not the case (Supporting Information Fig. S4A). Therefore, expression of hprt1/hprt is not influenced by tetracycline responsive elements itself. In contrast, in the Lhx2-induced KSL/KL cells, hprt1/hprt expression was decreased by withdrawal of dox (Supporting Information Fig. S4B). Thus, hprt1/hprt could be one of the Lhx2-target genes in these cell population.

In the mouse embryonic brain, Lhx2 is expressed in cortical precursor cells and operates as a selector gene to maintain cell identity [27]. This data revealed that Lhx2 maintains KSL/KL cell identity, and it is likely that these two effects of Lhx2 are mediated by a common mechanism. Lhx2 interacts with MRG1/Cited2 [28], which associates with the histone acetyltransferase CBP/p300. Thus, the ability of Lhx2 to maintain the identities of KSL/KL cells and cortical precursor cells might be dependent on epigenetic control mediated by these nuclear factors.

Several LIM-HD and Lmo proteins competitively associate with Ldb, and these interactions appear to regulate a variety of biological events in vertebrates and invertebrates. In Drosophila, the Lhx2 homolog Apterous is required to specify dorsal cell fate during wing disc formation, and the activity of Apterous in this compartment depends on Chip, an Ldb protein in Drosophila [29]. Apterous activity is regulated by the Lmo protein, dLmo [30]. In chick limb development, inhibition of Lhx2 activity arrests limb outgrowth, and CLIM2/Ldb1 inhibition results in the same phenotype [22, 31, 32]. Thus, regulation of transcription factor complexes comprising LIM-HD, Ldb, and Lmo seems to be important for embryonic development of various organisms. Degradation of Lmo2 is mediated by the E3 ubiquitin ligase Rlim/Rnf12 [23]. Rlim/Rnf12 also induces Ldb1 degradation [23], suggesting that Rlim/Rnf12 regulates the abundance of Lmo2 and Ldb1, and that the maintenance of these cofactors at appropriate levels is important for coordinated developmental processes. Recently, it was demonstrated that Rlim/Rnf12 controls neural induction and mesoderm formation from ESCs [33], suggesting this molecule has indispensable roles in embryonic development.

A transcription factor complex comprising Lmo2, Tal1/Scl, Gata1, Ldb1, and E47 regulates erythroid-specific gene expression [34]. Lmo2-null mice die in utero due to failure of yolk sac primitive erythropoiesis. This suggests that Lmo2 plays an important role in the induction of erythroid differentiation. Moreover, Lmo2 is also required for definitive hematopoiesis [35, 36]. Lmo2-null ESCs do not contribute to any lineages of adult hematopoietic cells in chimeric mice in vivo [36], suggesting that Lmo2 is required for lineage differentiation of HSCs/HPCs. It remains to be determined whether Lmo2 is also involved in the self-renewal and/or survival of HSCs. Interestingly, Lmo2 overexpression increased the frequency of Lin+ cells in the Lhx2-induced KSL/KL cells when we used Gata1KO ESCs (Supporting Information Fig. S2D). This suggests that Gata1 is dispensable for Lmo2 function in this context.

Gata2 and Gata3 can associate with Lmo2 [23, 37]. Therefore, Lmo2 may function in HSCs by associating with Gata2 and/or Gata3. Interestingly, an association partner of Lmo2, Tal1/Scl, is required for yolk sac primitive erythropoiesis and definitive hematopoiesis [25, 26]. The phenotypes of Lmo2-null and Tal1/Scl-null embryos are indistinguishable. Therefore, Lmo2 and Tal1/Scl may work together in HSCs. Surprisingly, when the Tal1/Scl gene was conditionally deleted in adult HSCs using the Cre/LoxP system, the HSCs retained LTR activity [38]. Thus, Tal1/Scl is required for HSC development but not for the maintenance of HSCs in adulthood. Presumably, Lmo2 may also be dispensable in the adult HSC compartment.

When Tal1/Scl is transduced into adult HSCs, the HSCs preferentially differentiate into myeloid cells [39]. Our data revealed that transduction of Lmo2 into the Lhx2-induced KSL/KL cells resulted in differentiation into Mac-1+ myeloid cells. It is likely that this is due to an excess amount of the transcriptional complex comprising Lmo2, Tal1/Scl, and possibly Ldb1.

Our previous transplantation assay and present in vitro differentiation experiments suggest that Lhx2 inhibits T-cell differentiation [14]. Aberrant Lmo2 expression due to chromosomal translocation is one of the major causes of acute T-cell leukemia [40]. Although the role of Lmo2 and other Lmo family members in normal T-cell development remains obscure, Lhx2-mediated inhibition of T-cell differentiation might be explained by the decreased availability of Lmo2 and/or other Lmo proteins.


Using a conditional gene expression system in ESCs, we revealed that robust production of KSL/KL cells from ESCs by the enforced expression of Lhx2 was caused by quantitative changes in the levels of transcription factors and cofactors critical for embryonic hematopoiesis. Our findings are helpful for the application of Lhx2, and Lhx2-interacting molecules identified in this study, in the induction of HSCs/HPCs from human iPS cells.


This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (24591415 to K.K. and 23390256 to K.K. and T.H.), and the SENSHIN Medical Research Foundation.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.