Gene Targeting Study Reveals Unexpected Expression of Brain-expressed X-linked 2 in Endocrine and Tissue Stem/Progenitor Cells in Mice*

Background: The role and precise expression pattern of individual brain-expressed X-linked genes in vivo were unknown. Results: Bex2-EGFP knock-in–knock-out mice were viable and fertile. Outside the brain, EGFP was expressed in specific cell populations. Conclusion: Bex2 plays redundant roles in vivo but is specifically expressed in endocrine and stem/progenitor cells. Significance: Bex2 is a novel marker for endocrine and stem/progenitor cells. Identification of genes specifically expressed in stem/progenitor cells is an important issue in developmental and stem cell biology. Genome-wide gene expression analyses in liver cells performed in this study have revealed a strong expression of X-linked genes that include members of the brain-expressed X-linked (Bex) gene family in stem/progenitor cells. Bex family genes are expressed abundantly in the neural cells and have been suggested to play important roles in the development of nervous tissues. However, the physiological role of its individual members and the precise expression pattern outside the nervous system remain largely unknown. Here, we focused on Bex2 and examined its role and expression pattern by generating knock-in mice; the enhanced green fluorescence protein (EGFP) was inserted into the Bex2 locus. Bex2-deficient mice were viable and fertile under laboratory growth conditions showing no obvious phenotypic abnormalities. Through an immunohistochemical analysis and flow cytometry-based approach, we observed unique EGFP reporter expression patterns in endocrine and stem/progenitor cells of the liver, pyloric stomach, and hematopoietic system. Although Bex2 seems to play redundant roles in vivo, these results suggest the significance and potential applications of Bex2 in studies of endocrine and stem/progenitor cells.


Identification of genes specifically expressed in stem/progenitor cells is an important issue in developmental and stem cell biology. Genome-wide gene expression analyses in liver cells performed in this study have revealed a strong expres-
sion of X-linked genes that include members of the brainexpressed X-linked (Bex) gene family in stem/progenitor cells. Bex family genes are expressed abundantly in the neural cells and have been suggested to play important roles in the development of nervous tissues. However, the physiological role of its individual members and the precise expression pattern outside the nervous system remain largely unknown. Here, we focused on Bex2 and examined its role and expression pattern by generating knock-in mice; the enhanced green fluorescence protein (EGFP) was inserted into the Bex2 locus. Bex2-deficient mice were viable and fertile under laboratory growth conditions showing no obvious phenotypic abnormalities. Through an immunohistochemical analysis and flow cytometry-based approach, we observed unique EGFP reporter expression patterns in endocrine and stem/ progenitor cells of the liver, pyloric stomach, and hematopoietic system. Although Bex2 seems to play redundant roles in vivo, these results suggest the significance and potential applications of Bex2 in studies of endocrine and stem/progenitor cells.
Studies on chromosomal assignment of the genes expressed during hematopoiesis have shown that stem cell-specific genes are located on the X chromosome more frequently than are differentiated cell-specific genes (1). The X chromosome is similar to autosomal chromosomes in structure, but the two types differ in their genomic constituents. For instance, in addition to its role in gametogenesis (2), the X chromosome also plays an important role in neural differentiation, as indicated by X-linked mental retardation syndromes (3). These findings alluded to the significance and potential application of X-linked genes for studies on both neuronal lineages and tissue stem cells.
Brain-expressed X-linked (BEX) genes are a family of genes that reside on the mammalian X chromosome and show sequence similarity with transcription elongation factor A (SII)-like (TCEAL) genes. Although the molecular properties of TCEAL genes are obscure, a closely related protein, TFIIS/ TCEA, maintains transcriptional fidelity by regulating the 3Ј-endoribonuclease activity of RNA polymerase II to bypass transcriptional pauses during the elongation process (4,5). At least five members of the BEX family have been identified to date, including BEX1, BEX2, BEX3, BEX4, and BEX5, but BEX5 appears to be lost in rodents (6). Members of this gene family were identified by screening of genes that are highly expressed in parthenogenic blastocysts (7) and account for more than 12% of the expressed sequence tags in the rat brain (6). In addition, these genes interact with olfactory marker proteins and are suggested to play an important role in olfactory neuronal development (8 -10). Deregulated expression of this gene family has been related to human cancers; for instance, BEX1 is a marker of neuroendocrine tumors (11). BEX2 is also highly expressed in a subset of primary breast cancer cells (12) and gliomas (13) and regulates cell proliferation and survival by mediating nuclear factor-B and c-Jun activity. In addition, BEX2 is a marker for acute myeloid leukemia with a chromosomal translocation at the mixed lineage leukemia gene locus (14). Moreover, members of this gene family mediate nerve growth factor signaling (15)(16)(17) and possess a nuclear localization signal for their translocation to the nucleus (6,15). Based on these reports, the BEX family genes are thought to function not only in cancer cells but also in developmental processes linking extracellular signaling to nuclear transcription events.
Although there are few studies on the physiological role of this gene family in vivo, recent studies using Bex1 knock-out mice revealed that Bex1 is involved in the regeneration of skeletal muscle (18) and neurons (19). Nonetheless, these mice displayed normal development and fertility, suggesting that other Bex genes play a redundant or major role in development. However, the functions of the other Bex family genes in vivo are not well known. In addition, although the expression of these genes has been examined through a screen of a cDNA library panel of bulk tissue samples (6), detailed analyses of their expression patterns at the cellular level have been difficult because of the challenges associated in raising specific antibodies against individual Bex family proteins. In this study, we investigated the expression of the Bex family genes in various tissues during the embryonic and adult stages. The results clearly showed that Bex2 expression highly correlates with the development of hepatic progenitor cells. To determine the physiological function and the expression pattern of Bex2 at the cellular level, we generated Bex2-deficient mice using a gene-targeting strategy by replacing the entire open reading frame (ORF) with enhanced green fluorescent protein (EGFP). 2 Molecular and cellular analyses of these mice revealed the significance and potential application of Bex2 for future studies of endocrine and tissue stem/progenitor cells.
Preparation of Mouse Embryonic Fibroblasts (MEFs)-At embryonic day (E) 12.5, ICR mouse embryos were dissected, and the head and internal organs were completely removed.
The torso was minced and dissociated in 0.05% trypsin/EDTA for 30 min. After washing, cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin/Lglutamine. MEFs were treated with mitomycin C at 37°C for 2 h and used as feeder cells.
Embryonic Stem (ES) Cell Cultures and Gene Targeting-EGR-101 cells, ES cells derived from the C57BL/6 NCr mouse strain, were cultured on MEFs in M15G medium. M15G medium is a mixture of knock-out DMEM (Invitrogen) supplemented with 15% FBS, 1% penicillin/streptomycin/L-glutamine, ␤-mercaptoethanol (100 M), and 1000 units/ml leukemia inhibitory factor (LIF; Chemicon, Temecula, CA). For gene targeting, plasmids carrying an EGFP-PGK-Neo-DTA cassette were used. Both the 7.8-kb region upstream of the third exon of Bex2 (5Ј-homology arm) and the 2.8-kb region downstream of the third exon of Bex2 (3Ј-homology arm) were cloned from BAC vectors containing a region that covered this genomic locus (clone Rp23-149K3; Geno Techs, Japan). The fragments were subcloned into the targeting vector. Purified plasmids were linearized with the AscI restriction enzyme and subsequently used for electroporation. One day after electroporation, transfected ES cells were selected with G418 (300 ng/ml) in culture. G418-resistant clones were expanded and genotyped over the short arm to detect the correct recombination by PCR. Selected clones were assayed again for correct recombination using Southern hybridization. For Southern hybridization probes, short fragments of Bex2 genome DNA were amplified using PCR and subcloned to pGEM-T easy vector System 1 (Promega, Madison, WI). PCR primers for mouse genotyping and the primers used for Southern hybridization probe generation are shown in Table 1.
Messenger RNA (mRNA) Detection by Reverse Transcription (RT)-PCR-Total RNA was extracted from primary and cultured cells using the RNeasy micro kit (Qiagen, Venlo, The Netherlands) or TRIzol (Invitrogen), according to the manufacturer's instructions. First-strand cDNA was synthesized using the High Capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA) and used as a template for quantitative PCR. The cDNA samples were normalized by expression of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), using the TaqMan probe (Applied Biosystems). Quantitative analyses of target mRNA expression levels were performed using the Universal Probe Library System (Roche Diagnostics). Primers and probes for both nonquantitative PCR and quantitative PCR are shown in Table 2. Isolation and Analysis of Foregut Endodermal Cells-The foregut endodermal cells can be identified based on Foxa3 expression in the foregut endoderm at around E8.5-9.0 (20 -22). Transgenic mice expressing Venus fluorescent protein under the control of a Foxa3-enhancer promoter sequence were generated by pronuclear injection. Induced pluripotent stem cell lines were established from fibroblasts derived from the Foxa3-promoter-Venus transgenic mice using a Dox-inducible Oct3/4-Klf4-Sox2-expressing lentivirus (23). These cell lines were injected into tetraploid embryos and subsequently implanted into a pseudopregnant mouse. Somite numbers (somite pairs 6 -15) were counted to validate the developmental stages identical to E8.5-9.0. Foxa3positive regions were detected in these mouse embryos using fluorescence microscopy ( Fig. 2A). These embryos were soaked in 0.25% trypsin/EDTA solution for 10 min at 37°C and then dissociated by pipetting. Cells were collected in a staining buffer (phosphate-buffered saline (PBS) with 3% FBS) and subsequently stained with anti-CD45-PE-Cy7, anti-Ter119-PE-Cy7 (eBioscience, San Diego), and anti-EpCAM-Alexa647 (clone G8.8, Santa Cruz Biotechnology, Dallas, TX) antibodies for 30 min. After washing out the remaining antibody with PBS, cells were resuspended in staining medium containing propidium iodide and then sorted using a MoFlo TM fluorescence-activated cell sorter (FACS) (DAKO, Glostrup, Denmark). Specificity of the EpCAM antibody was validated by comparing the staining pattern between Alexa647 conjugated EpCAM antibody and its rat-IgG2a isotype control (eBioscience) on WT E9.0 embryos (Fig. 2B).
Hepatoblast Isolation and Culture-Hepatoblast isolation was performed as described previously (24). Minced embryonic liver tissues from E13.5 mice were dissociated with a 0.05% collagenase solution. Dissociated liver cells were washed with a staining buffer and then incubated with antibodies against cell surface markers (shown in Table 3) for 60 min at 4°C. After staining the dead cells with propidium iodide, the cells were analyzed and sorted using a MoFlo TM cell sorter. CD45 Ϫ Ter119 Ϫ CD71 Ϫ Dlk1 ϩ CD133 ϩ cells derived from E13.5 livers were purified as hepatoblasts in this study. For microarray analyses, CD45 Ϫ Ter119 Ϫ c-Kit Ϫ Dlk1 ϩ CD133 ϩ cells were used as hepatoblasts. MEFs were plated in culture dishes at a density of 2 ϫ 10 5 cells/well (12-well plates) or

Gene
Forward (  3.2 ϫ 10 6 cells/dish (100-mm dishes) prior to the day of hepatoblast isolation. For gene expression analysis, hepatoblasts (2 ϫ 10 4 cells) purified from GFP-transgenic mouse embryonic livers were plated onto mitomycin C-treated MEFs in a 100-mm culture dish. After 5 days of culture, GFP ϩ hepatoblasts were purified using a flow cytometer and used for RNA preparation. For colony formation assays, hepatoblasts isolated from both wild-type and Bex2-EGFP knock-in mouse embryos were purified and plated onto MEF feeder cells in 12-well tissue culture dishes (200 cells/well). Cells were cultured for 7 days in the standard culture medium in the presence of 40 ng/ml hepatocyte growth factor and 20 ng/ml epidermal growth factor (PeproTech; Rocky Hill, NJ). The standard culture medium used was a 1:1 mixture of hepatocyte colony formation medium (DMEM/Ham's F-12 half-medium supplemented with 1ϫ insulin/transferrin/selenium X, 10 mM nicotinamide, 10 Ϫ7 M dexamethasone, 2.5 mM HEPES, 1ϫ penicillin/streptomycin/L-glutamine, and 1ϫ nonessential amino acid solution) and fresh DMEM supplemented with 10% FBS. To count colonies derived from individual single cells, we used the ArrayScan VTI HCS Reader (Thermo Scientific, Waltham, MA) after immunostaining. Isolation of Hepatic, Endothelial, Mesenchymal, and Hematopoietic Cells from E13.5 Fetal Liver by Flow Cytometry-Fetal liver cell fractionation was performed according to the method we have described previously (24). Briefly, minced embryonic liver tissue from E13.5 mice was dissociated with a 0.05% collagenase solution. Dissociated liver cells were washed with a staining buffer and incubated with antibodies against suitable cell surface markers (Table  3) for 60 min at 4°C. After incubation with propidium iodide (for staining dead cells), the cells were analyzed and sorted using a MoFlo TM cell sorter. CD45 Ϫ Ter119 Ϫ CD71 Ϫ Dlk1 ϩ PDGFR␣ Ϫ cells were purified as hepatoblasts. CD45 Ϫ Ter119 Ϫ CD71 Ϫ Dlk1 mid PDGFR␣ ϩ cells were purified as mesenchymal cells. CD45 Ϫ Ter119 Ϫ CD71 Ϫ Dlk1 mid PCLP1 ϩ cells were purified as mesothelial cells. CD45 Ϫ Ter119 Ϫ CD71 Ϫ Dlk1 mid Flk1 ϩ cells were purified as endothelial cells. CD45 ϩ or Ter119 ϩ or CD71 ϩ cells were purified as hematopoietic cells.
Adult Hepatocyte and Bile Duct Epithelial Cell Isolation-Hepatocytes were isolated from 7-week-old livers following a two-step collagenase digestion (25). Briefly, perfused liver tissues were subsequently dissociated with a 0.05% collagenase solution. The mature hepatocyte fraction was separated from nonparenchymal cells with three low-speed centrifugations (50 ϫ g, 1 min). Dead cell debris was removed by centrifugation in 50% Percoll solution (GE Healthcare).
Bile duct epithelial cells were purified from a nonparenchymal cell fraction by collecting the supernatant of three lowspeed centrifugations (50 ϫ g, 1 min). Dead cell debris was removed by centrifugation in 25% Percoll solution (GE Healthcare). The cells were then stained for antibodies against CD45, Ter119, and EpCAM. CD45 Ϫ Ter119 Ϫ EpCAM ϩ cells were sorted as bile duct epithelial cells using a MoFlo TM cell sorter. Antibodies used in this step are shown in Table 3.
Immunostaining-Cultured cells were fixed in 4% paraformaldehyde in PBS, washed three times with PBS, and permeabilized with 0.5% Triton/PBS for 10 min. After washing with PBS, the cells were incubated with 5% donkey serum (Millipore, Bedford, MA) in PBS for 1 h at room temperature and then with diluted primary antibodies (Table 4) overnight at 4°C. After washing with PBS, cells were incubated for 1 h at room temperature with an anti-rabbit IgG-Alexa488 antibody, an anti-rabbit IgG-Alexa546 antibody, or an anti-goat IgG-Al-exa546 antibody (Invitrogen). Cell nuclei were then stained with DAPI.
For immunohistochemistry, paraffin-embedded sections were washed three times with xylene for 5 min and subsequently dehydrated by soaking the sections sequentially into 100, 90, 80, and 70% ethanol for 3 min. These samples were then autoclaved in Tris/EDTA solution (pH 9.0) for 15 min at 105°C for antigen retrieval. The samples were then cooled at room temperature for more than 30 min. Then samples were washed with PBS, incubated with 5% donkey serum in PBS for 1 h at room temperature, and incubated with diluted primary antibodies (Table 4) overnight at 4°C. After washing with PBS, sections were incubated for 1 h at room temperature with either anti-rat IgG-Alexa546 antibody, anti-rat IgG-Alexa647 antibody, anti-rabbit IgG-Alexa488 antibody, anti-rabbit IgG-Alexa546 antibody, or anti-goat IgG-Alexa546 antibody (Invitrogen). Cell nuclei were then stained with DAPI. Fluorescent images were obtained with the Axio Observer Z1 fluorescent microscope (Zeiss), BX51 fluorescent microscope (Olympus Corp., Tokyo, Japan), and BZ-9000 fluorescent microscope (Keyence, Japan).
Flow Cytometric Analysis of Hematopoietic Cells Derived from Bex2 EGFP/Y Mice-For bone marrow analysis, the femurs and tibias of 8 -10-week-old male mice were prepared. The bone marrow was flushed out, and the cell number was counted. Cell suspensions were immunostained with the antibodies listed in Table 3. Briefly, cells were stained with biotinylated antibodies against lineage markers (CD3, CD4, CD8, CD127, B220, Ter119, and Gr-1) for 30 min on ice. After washing with PBS, the cells were then stained with fluorescently labeled antibodies against markers such as CD34, CD117 (c-Kit), and Ly6A/E (Sca-1). At the same time, lineage markers were stained with streptavidin-APC-Cy7. Cells were purified using either a MoFlo TM cell sorter or a FACS Aria II system (BD Biosciences).
For gene expression analysis of the Bex family genes across the hematopoietic lineages, cells derived from the bone marrow, thymus, and spleen were prepared by surgically dissecting the tissues from wild-type 8-to 10-week-old C57BL/6NCr  mice. Cells were stained with the antibodies listed in Table 3. Cells expressing each lineage marker or cells in the primitive fraction were sorted by a MoFlo TM cell sorter. Flow Cytometric Analysis of Stomach Samples from Bex2 EGFP/Y Mice-For analyses of EGFP-positive cells in the pyloric stomach, stomach samples from 8-to 10-week-old Bex2 EGFP/Y mice were used. Stomachs of the Bex2 EGFP/Y mice were opened along the greater curvature, and the pyloric region was surgically isolated. After removal of the muscle layer by dissection, the epithelial layer was used for further analyses. The epithelial layer was minced in buffer containing EGTA for 3-5 h on ice and then dissociated by pipetting with a 10-ml pipette. The samples were then treated with 0.25% trypsin/ EDTA (Sigma) for 30 min and then filtered through 45-m mesh filters for flow cytometry. Cells were sorted by a MoFlo TM cell sorter, and fractionated samples were collected and used for gene expression analysis. The primers used in the assay are listed in Table 2.
Transcription Profile Analysis Using Microarray-CD45 Ϫ Ter119 Ϫ CD133 ϩ Dlk ϩ hepatoblasts derived from E13.5 livers and adult hepatocytes were purified as described above. Total RNA was purified from these cells using the RNeasy micro kit, according to the manufacturer's instructions. Transcription profiles were analyzed using the Agilent Whole Mouse Genome Microarray 4 ϫ 44K. The original data are available from the Gene Expression Omnibus (accession number GSE56734). . Generation of Bex2 knock-out mice expressing an EGFP reporter by gene targeting. A, targeting strategy for the Bex2 genomic locus. Both the 7.8-kb upstream and 2.8-kb downstream fragments of the third exon of Bex2 were subcloned into the targeting vector as the 5Ј-and 3Ј-homology arms, respectively. Restriction enzymes used for Southern blotting are depicted (EcoRV and NotI for 5Ј-arm detection and SpeI for 3Ј-arm detection). Probes used for Southern blotting are also shown as the thick bar. Using these probes and restriction enzymes, the targeted allele can be detected in a 12-kb (5Ј-arm) and 6.2-kb (3Ј-arm) band, whereas the wild-type allele is detected at 16 kb (5Ј-arm) and 7.7 kb (3Ј-arm), respectively. Arrows represent the primers designed for mouse genotyping. B, Southern blotting of the 5Ј-arm and the 3Ј-arm. The correct homologous recombination was confirmed in several ES clones (1A2, 1B3, 2A5, and 2B3). The parental ES cell line (WT) was used as a control. C, detection of Bex1, -2, and -3 transcripts in the targeted ES cells. The parental ES cell line (WT) was used as a control. GAPDH was the internal control. Bex2 expression was undetectable in the targeted ES cells. D, examples of the genotyping bands observed with genomic PCR of tissue samples from offspring born by mating Bex2 EGFP/X female mice with wild-type C57BL/6NCr male mice. Distilled water (DW) was used as a negative control. E, numbers and the resulting genotypes of the offspring from Bex2 knock-out mice.
Expression data were analyzed using Gene Springs 12.6. Datasets were normalized, and genes differentially expressed by more than 400-fold between the average of two groups (hepatoblasts versus adult hepatocytes) were extracted and represented as a heat map.
Statistical Analysis-Microsoft Excel 2007 (Microsoft, Redmond, WA) was used to calculate standard deviations (S.D.), and statistically significant differences between samples were determined by using a Student's two-tailed t test.

Bex2 Is Expressed in Stem/Progenitor Cell Populations during
Liver Development-To identify genes that are specifically expressed in stem/progenitor cells, we performed comparative gene expression analysis between hepatoblasts (fetal hepatic stem/progenitor cells) and adult hepatocytes (Fig. 1A). Among the list of 30 genes that were more highly expressed in hepatoblasts by more than 400-fold compared with adult hepatocytes, the immature hepatocyte marker ␣-fetoprotein (Afp) as well as several genomic imprinted genes (Dlk1, Igf2, H19, Grb10, Rian, and Meg3) were identified. In addition to these genes, seven X-linked genes, including all the members of the Bex family genes, were also identified in the same list. Based on these data, we focused specifically on Bex family genes.
To gain insight into Bex family gene expression patterns in both embryonic and adult mice, we generated a tissue-specific cDNA panel from E13.5 and 10-week-old male mice. Subsequent RT-PCR analysis revealed the broad tissue expression of Bex family genes in the stage of embryonic development (Fig.  1B). Expression of these genes persisted at a high level in the brain and testis of adult mice. However, this expression was decreased in other tissues such as the endodermal organs and kidneys (Fig. 1C). Given that Bex1, -2, and -3 are highly expressed in proliferating neural precursor cells and are downregulated as the cells differentiate (15), we hypothesized that Bex family members would be specifically expressed in embryonic precursor cells of other somatic tissues as well. To confirm this hypothesis, we examined liver development using ES cells, endodermal progenitor cells isolated from Foxa3-promoter-Venus transgenic embryos (Fig. 2, A-D), fetal hepatoblasts, adult hepatocytes, and bile duct epithelial cells. Both Foxa2 and Foxa3 are known to be endodermal marker genes in mouse embryos. However, Foxa2 is also expressed in the cells of the notochord and the floor plate of the neural tube as well (26 -28). In contrast, Foxa3 is mainly expressed in the endodermal cells at around E8.0-E.9.5 (21,22,29). We therefore have selected Foxa3 as an alternative marker to specifically detect the endo- . Normal phenotypes of Bex2 knock-out mice. A, loss of Bex2 mRNA in the brain and testis of Bex2 knock-out mice. Total RNA was purified from the brain and testis of both wild-type (Bex2 X/Y ) and knock-out (Bex2 EGFP/Y ) mice. Bex family gene expression was analyzed using RT-PCR. GAPDH was the internal control. B, representative images of knock-out (Bex2 EGFP/Y ) and littermate wild-type (Bex2 X/Y ) mice. Bex2 EGFP/Y mice grew into adults with normal appearances. C, hematoxylin and eosin staining of tissue sections from both wild-type (Bex2 X/Y ) and knock-out (Bex2 EGFP/Y ) mice. SI, small intestine. dermal progenitor cells in the embryo. After confirming the expression pattern of reference genes (Nanog, Sox2, and Prdm14 for ES cells; Sox17 and Foxa3 for endoderm; Afp for hepatoblasts, Albumin, Cyp7a1, and HNF4␣1 for hepatocytes; and Sox9, Cytokeratin19, and Osteopontin for ductal cells) (Fig.  1D), we found that Bex gene family expression increased during the ES cell and hepatoblast stages, but eventually it became undetectable in adult hepatocytes. Additionally, although the expression of Bex1, -3, and -4 was detectable at high levels, expression of Bex2 was below detectable levels in the bile duct epithelial cells (Fig. 1E).
In addition, we analyzed the expression of Bex family genes during the in vitro culture of fetal hepatoblasts. Recently, we established a colony formation culture using mesenchymal cells as feeder cells (30), which expands and differentiates fetal hepatoblasts into cells positive for albumin, a hepatocytic marker, or cytokeratin 19, a cholangiocytic and progenitor marker. Using this culture system, we found that among the Bex family genes, only Bex2 was markedly decreased during this culture period, suggesting a correlation between the primary hepatic stem/progenitor-like state and Bex2 expression (Fig. 1F). These results suggest FIGURE 5. Images of Bex2 mutant fetuses at E13.5. A, offspring born by mating of Bex2 EGFP/X female mice with wild-type male mice. Sry was used to confirm male mice. Bex2 EGFP/Y males and Bex2 X/Y males were born in an expected Mendelian ratio. B, EGFP immunofluorescence in the embryos of wild-type male (WT, Bex2 X/Y ), knock-out male (Bex2 EGFP/Y ), and heterozygous female (Bex2 EGFP/X ) at E13.5. Cell nuclei were stained with DAPI. Scale bar ϭ 300 m. C, magnified view of the immunofluorescence of EGFP in the intestine, kidney, and genital ridge (testis and ovary) of wild-type male (WT, Bex2 X/Y ), knock-out male (Bex2 EGFP/Y ), and heterozygous female (Bex2 EGFP/X ) mouse embryos at E13.5. Scale bar, 100 m.
that Bex2 is a novel marker for stem/progenitor cells during liver development.
Generation of Bex2 Knock-out Mice with the EGFP Reporter-To examine the physiological functions of Bex2 and monitor the transcriptional activity of the Bex2 genomic locus in vivo, we used a gene-targeting strategy to replace the 3rd exonic sequence encoding the entire ORF with an EGFP sequence and confirmed the correct recombination by Southern hybridization (Fig. 3, A and B). We also confirmed the absence of Bex2 mRNA transcripts in these ES cells using RT-PCR (Fig. 3C). Although disrupting an X-linked gene in a male ES cell line could have a potentially negative effect on development, because it directly generates a nullizygous situation, Bex2-targeted ES cells exhibited normal proliferation in vitro. After confirming the normal karyotype, we injected the targeted cells into ICR blastocysts and obtained chimeric mice. We then collected sperm from these chimeras and performed intracytoplasmic sperm-head injections FIGURE 6. Analyses of E13.5 hepatoblasts in Bex2 knock-out mice with the EGFP reporter. A, representative immunohistochemical images of E13.5 embryonic livers derived from wild-type male mice (WT, Bex2 X/Y ), knock-out male mice (Bex2 EGFP/Y ), and heterozygous female mice (Bex2 EGFP/X ). EGFP reporter expression controlled by the Bex2 locus is shown. Cell nuclei were stained with DAPI. Scale bar, 100 m. B, flow cytometry analysis of hepatoblast populations derived from both wild-type (Bex2 X/Y ) and Bex2 EGFP/Y mice. The proportions of CD133 ϩ Dlk1 ϩ hepatoblasts in the nonhematopoietic cell fraction derived from E13.5 fetal livers are shown. C, EGFP expression in the CD133 ϩ Dlk1 ϩ hepatoblast fraction from both wild-type and knock-out mice. Gates used for flow cytometry analyses are shown in B. D, analyses of the colony forming ability of hepatoblasts derived from wild-type (Bex2 X/Y ) and Bex2 EGFP/Y mouse livers. Left, representative immunocytochemical staining of the colonies derived from CD133 ϩ Dlk1 ϩ hepatoblasts. Sorted cells were cultured for 7 days and stained with anti-albumin and cytokeratin 19 (CK19) antibodies. Note that the cells sorted from both genotypes formed colonies with albumin expression located in the central part of the colonies and cytokeratin 19 expression at the periphery. Scale bars, 100 m. Right, numbers and sizes of the albumin-positive colonies formed in culture. The colony forming efficiency did not differ between the two groups. Representative data of colony formation in triplicate wells are shown. N.S. means not significant. OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 into host oocytes. Several female pups with the Bex2-targeted allele were born, suggesting that Bex2-deficient ES cells can contribute to the germ line and transmit their genome to the next generation.

Restricted Activation of Bex2 Locus in Mice
Female mice with the targeted allele (Bex2 EGFP/X ) were crossed with wild-type males. Healthy offspring carrying a targeted allele, both male (Bex2 EGFP/Y ) and female (Bex2 EGFP/X ), were born as shown in Fig. 3, D and E. Male offspring with no Bex2 expression (Bex2 EGFP/Y ) grew into adulthood (Fig. 4, A and B). Histological analysis of various tissues revealed that there were no differences between Bex2 EGFP/Y adult males and wild-type Bex2 X/Y littermates, at least at the macroscopic level (Fig. 4C).

Bex2 Serves as a Novel Marker of Hepatoblasts in the Midfetal Liver, but Its Loss Does Not Affect Proliferation and
Differentiation-We crossed Bex2 EGFP/X female mice with wild-type male mice and obtained two types of Bex2 knock-in FIGURE 8. Usage of Bex2-EGFP as a novel fetal liver hepatic progenitor marker. A, staining pattern of hepatoblasts and mesenchymal cells derived from wild-type E13.5 fetal livers. CD45 Ϫ , Ter119 Ϫ , and CD71 Ϫ negative cell fractions were stained with anti-Dlk1, CD133, and PDGFR␣ antibodies or their isotype control antibodies. B, flow cytometry analyses of Bex2 EGFP/Y fetal liver cells. EGFP expression in the nonhematopoietic cell fraction is mainly consisting of two fractions, the EGFP high fraction (gated in red) and the EGFP mid fraction (gated in blue). In the right panels, both EGFP high and EGFP mid cell fractions gated in the left panel were analyzed using Dlk1, CD133, and PDGFR␣ antibodies. EGFP high fraction cells are generally Dlk1 ϩ CD133 ϩ , although EGFP mid fraction cells are mainly Dlk1 mid PDGFR␣ ϩ . C, expression of Bex2 transcript in fractionated fetal liver cells. Fetal liver cells from E13.5 livers were fractionated into cells enriched for hematopoietic cells (CD45 ϩ or Ter119 ϩ or CD71 ϩ ) or nonhematopoietic cells (CD45 Ϫ Ter119 Ϫ CD71 Ϫ gated) such as hepatoblasts (Dlk1 ϩ PDGFR␣ Ϫ ),  OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 embryos (Bex2 EGFP/Y males and Bex2 EGFP/X females) and control littermates (Fig. 5, A and B). Bex2 knock-in mice had specific EGFP signals in various organs such as intestine, genital ridge (testis for male and ovary for female), kidney (Fig. 5C), and liver (Fig. 6A). In particular, cells expressing EGFP at high levels were scattered in clusters in the fetal liver, resembling the structural localization of hepatoblasts. In all the tissues examined, the number of EGFP-positive cells in the Bex2 EGFP/X female embryos was smaller than that of Bex2 EGFP/Y male embryos in these tissues. The underlying mechanism to this phenomenon, if not all, might be the X chromosome inactivation occurring in the individual cells at the early pre-implantation embryo stage, which makes the female heterozygous Bex2 EGFP/X mice a mosaic of cells that have either the activated or inactivated X chromosome. Based on our initial observation that Bex2 is highly expressed in fetal hepatoblasts and becomes down-regulated upon hepatic differentiation, we specifically focused on the phenotypes of Bex2-deficient hepatoblasts using an in vitro colony assay. Bex2-deficient fetal livers from Bex2 EGFP/Y males showed normal cellular numbers and percentages of hepatoblasts (the CD133 ϩ Dlk1 ϩ cells in the CD45 Ϫ Ter119 Ϫ CD71 Ϫ nonhematopoietic cell fraction) compared with their wild-type Bex2 X/Y male littermates (Fig. 6B). Hepatoblasts derived from Bex2 EGFP/Y fetal livers were strongly positive for EGFP (Fig. 6C), indicating that the transcriptional activity of the Bex2 locus is high in hepatoblasts. In addition, the colony forming ability of Bex2 EGFP/Y hepatoblasts was similar to that of wild-type hepatoblasts (Fig. 6D).

Restricted Activation of Bex2 Locus in Mice
Bex2 EGFP/X mice have both Bex2 ϩ and Bex2 Ϫ cells because of X chromosome inactivation at the early embryonic stage. If the targeted allele is on the inactivated chromosome, these cells are theoretically wild-type. However, if the targeted allele is on the active X chromosome, cells with active Bex2 locus are EGFPpositive but Bex2-deficient. Taking advantage of this competitive situation, we analyzed the hepatoblasts in Bex2 EGFP/X female fetal livers. The proportions of EGFP-positive cells and EGFP-negative cells in the hepatoblast fraction derived from Bex2 EGFP/X livers were nearly equal (Fig. 7A), suggesting that loss of Bex2 does not affect cell proliferation, even in a competitive situation during the mid-fetal stages. In addition, hepatoblasts isolated from either the EGFP-positive or EGFP-negative fraction formed colonies at similar levels (Fig. 7B).
In this culture system, hepatoblasts can be distinguished and isolated from MEFs by the expression of the epithelial cell marker E-cadherin (Fig. 7C). Therefore, we compared EGFP expression between freshly isolated hepatoblasts and 7-daycultured hepatoblasts. Nearly a 10-fold (1-log) decrease in the EGFP intensity was detected in cultured hepatoblasts derived from Bex2 EGFP/Y mice (Fig. 7D). These data confirm the mRNA expression data shown in Fig. 1D. These results suggested that the defect of Bex2 in hepatoblasts is compensable both in vitro and in vivo, and the transcriptional activity of the Bex2 locus in hepatoblasts can be monitored by EGFP fluorescence during liver development.
Finally, we assessed whether the activity of the EGFP reporter is useful for the purification of hepatoblasts in mouse fetal livers. E13.5 fetal liver contains various types of cells such as hepatoblasts, mesenchymal progenitor cells, mesothelial cells, endo-thelial cells, and hematopoietic cells. We have previously reported that hepatoblasts can be enriched by the expression of CD133 and Dlk1 (31), while mesenchymal cells can be sorted by the expression of lower levels of Dlk1 and a prominent expression of PDGFR␣ (Fig. 8A) (24). Using flow cytometry, we analyzed the EGFP reporter-active cells in the fetal liver of Bex2 EGFP/Y mice. The higher EGFP reporter activity (Bex2-EGFP high ) was evident in the presumptive hepatoblast fraction expressing both CD133 and Dlk1 (Fig. 8B, red cells). However, the lower EGFP reporter activity (Bex2-EGFP mid ) was generally detected in the mesenchymal cell fraction expressing PDGFR␣ (Fig. 8B, blue cells). We confirmed these results by analyzing the expression levels of Bex2 mRNA transcripts in wild-type E13.5 fetal liver cells using quantitative PCR. Expression levels of Bex2 transcripts were the highest in the hepatoblast fraction, although lower expression levels of Bex2 were also detected in both mesenchymal and mesothelial cell fractions (Fig. 8C).
We also confirmed that Bex2-EGFP high cells but not Bex2-EGFP mid cells are hepatic progenitor cells that are capable of forming hepatic (albumin-positive) colonies in vitro (Fig. 8D). Interestingly, we compared the hepatic colony-forming efficiency between Bex2-EGFP high cells and Dlk1 ϩ CD133 ϩ cells and found that Bex2-EGFP high cells give higher hepatic colonyforming efficiency. Dlk1 is known to be expressed in the mesenchymal cells (24,32), and CD133 is expressed in a minor subset of mesenchymal cells similar to hepatoblasts. Therefore, Dlk1 ϩ CD133 ϩ cells purified using flow cytometry technically harbor risks of contamination of mesenchymal cells at a certain frequency and thus result in lower sorting efficiency of hepatoblasts. These results suggest the potential utility of Bex2 as a novel specific marker for fetal hepatoblasts.
Expression Profiling of Bex2 Genomic Locus Transcriptional Activity in Adult Tissues-As mentioned above, we could not detect obvious phenotypes in Bex2-deficient mice, particularly FIGURE 9. Macroscopic view of tissues derived from Bex2 EGFP/Y and wildtype littermate mice at 9 weeks old. Macroscopic views of the brain, eye, pancreas, intestine, testis, liver, kidney, heart, lung, and spleen are shown. The bright field and EGFP fluorescence of both wild-type (left) and Bex2 EGFP/Y (right) tissues are shown.
during liver development. However, these mice are useful for labeling hepatoblasts in the fetal liver using EGFP fluorescence. In addition, the EGFP intensity in adult mice was most evident in the brain and testis, although it was undetectable in the liver and kidney at macroscopic levels (Fig. 9). These results were consistent with the RT-PCR results using wild-type mouse tissues. This suggested that Bex2 deficiency did not affect Bex2 locus transcriptional activity, as monitored by EGFP. Thus, we aimed to obtain comprehensive information on the precise Bex2 expression pattern in adult tissues at the cellular level using EGFP fluorescence as readout of endogenous Bex2 genomic locus transcriptional activity.
Bex2-EGFP Reporter Is Active in the Central and Peripheral Neuronal Cells-Because the Bex2 locus was significantly active in the brain and eye (Fig. 9), we analyzed the EGFP fluorescence pattern of eye and brain tissue section samples from Bex2 EGFP/Y mice. EGFP fluorescence was positive in the entire neural retina and in the brain, suggesting that the Bex2 locus is active in most of the regions in the retina and brain (Fig. 10, A and B). These results are consistent with previous reports (10). In addition,  OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 29905
EGFP-positive neuronal cells were observed on the surface of gastrointestinal tissues in a reticulated structure (Fig. 10C), suggesting that Bex2 is active not only in the central nervous system but also in the peripheral nervous system.
Bex2-EGFP Reporter Is Active in Endocrine Cells-Endocrine cells of the digestive tract consist of pancreatic cells, clustered in the islets of Langerhans (33), and scattered cells, distributed throughout the digestive epithelium from the stomach to the colon (34), the latter of which are known as enteroendocrine cells. In the pancreas, basal EGFP intensity was observed, but the most evident EGFP signal appeared as a dotted structure (Fig. 11A). By immunohistochemistry, we confirmed that the bright EGFP intensity was derived from cells producing somatostatin (␣-cells), insulin (␤-cells), and glucagon (␥-cells) (Fig.  11B). Therefore, in the pancreas, Bex2 seems to be a common marker for all endocrine cells. We next evaluated whether a Bex2 signal could be observed in the endocrine cells of gastrointestinal tissues. Interestingly, EGFP was detected in cells expressing Chromogranin A (ChgA), a secretory protein generally produced by enteroendocrine cells in both the small intestine and colon (Fig. 11C, panel a). We also observed a strong EGFP-positive area at the submucosal region of the intestinal tube, a structure representing the neural lineage in the intestinal tissue (Fig. 11C, panel b). Therefore, we found that Bex2 is not only a marker for the neural cell lineage but also for the endocrine cell lineage in gastrointestinal tissues. In the testis, the Sertoli cell is a major somatic cell type that plays a central role combining endocrine action and spermatogenesis (35,36). In the Bex2 EGFP/Y mice, EGFP expression was evident in the periphery of the seminiferous tubules. Both Sertoli cells and germ cells (spermatogonia) are enriched in the peripheral region of the testis. Therefore, we stained these cells using specific antibodies against Sertoli and germ cell markers. EGFP expression was detected in cells positive for both Sox9 and WT-1, Sertoli cell markers (Fig. 11D), but not in cells positive for MVH, Tra98, and PLZF, germ cell markers (Fig. 11E). These results suggest that the Bex2 locus is specifically active in the Sertoli cells but not germ cells in the testis.

Bex2-EGFP Reporter Is Active in Stem/Progenitor Cells in the Pyloric Stomach and in Hematopoietic Lineages-Recent
reports have shown that molecules such as Lgr5 and Pw1 are specifically expressed in stem cells among various tissues (37,38). As shown above, Bex2 expression correlated with stem/ progenitor cell phenotypes in the fetal liver. Therefore, we hypothesized that Bex2 might also be useful for marking stem/ progenitor cells in other tissues. Through immunohistochemical analyses of various tissues, we found that EGFP-expressing cells were enriched at the bases of glands in the pyloric stomach, where rapidly proliferating Lgr5 ϩ stem cells were recently identified (39). As described above, a subset of these EGFP-positive cells were ChgA-expressing endocrine cells. However, many ChgA Ϫ EGFP ϩ cells were detected in this region (Fig. 12A). EGFP-positive cells were absent in the forestomach and were very rare in the corpus region. We also analyzed whether pre-viously reported stem cell markers (39) are expressed in the EGFP-positive cell fraction isolated from the pyloric epithelium. In addition to ChgA, expression of Lgr5 and Sox9, markers of pyloric epithelial stem cells, was enriched in the EGFP-positive cell fraction (Fig. 12, B and C), suggesting that Bex2-positive cells in the pyloric epithelium contained both a stem/progenitor cell fraction and an endocrine cell fraction.
Next, the EGFP-positive cell distribution along the hematopoietic cell lineage was analyzed. Using flow cytometry, we found that the c-Kit ϩ Sca-1 ϩ Lineage Ϫ (KSL) cell fraction, which is known to be rich in hematopoietic stem/progenitor cells (40), contains EGFP-positive cells at a higher frequency than the more committed progenitor cells (Lineage Ϫ fraction). In addition, further enrichment for hematopoietic stem cells within the KSL fraction using CD34 expression (41) revealed that EGFP-positive cells were found in greater abundance in the CD34 Ϫ KSL stem cell fraction compared with the CD34 ϩ KSL progenitor cell fraction (Fig. 13, A and B). To confirm that this was not an artificial observation in the genetically modified Bex2-deficient mice, Bex family gene expression was analyzed by quantitative PCR in several hematopoietic cell fractions derived from wild-type mice (Fig. 13C). The Bex family genes were most strongly expressed in CD34 Ϫ KSL cells. In contrast, lower levels of gene expression were detected in progenitor  OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 cells (both CD34 ϩ KSL and Lineage Ϫ cells). In particular, Bex1 and Bex2 were significantly expressed in CD34 Ϫ KSL cells, although the expression was sharply down-regulated or undetected in committed hematopoietic cells. Collectively, these results suggest that Bex2 can be a marker of stem/progenitor cells, not only for the hepatic lineages but also for the pyloric stomach and hematopoietic tissues.

DISCUSSION
To determine the physiological role of Bex2 and monitor the endogenous transcriptional activity of the Bex2 genomic locus, we utilized a homologous recombination strategy. Bex2 mRNA expression was detected in many fetal and adult tissues. However, Bex2-deficient mice were viable, with no apparent differences when compared with their wild-type littermates under physiological conditions. In addition, Bex2-deficient hepatoblasts proliferated and differentiated normally both in vivo and in vitro. Most Bex family genes had a similar expression pattern in both fetal and adult tissues. Of particular note, Bex1 and Bex2 exhibited ϳ90% sequence similarity and shared an extremely similar expression pattern in vivo (6). These findings suggest that the lack of phenotypes in Bex2-deficient mice may be due to the redundancy of Bex family genes. Therefore, generation and analyses of Bex1 and Bex2 double knock-out mice are of interest for future studies. In addition, because Bex1-deficient mice display specific phenotypes that cannot be compensated for by Bex2 in regeneration models after injury, there may be Bex2-specific functions in various disease and injured processes (e.g. carcinogenesis, inflammation, and regeneration) as well. The role of Bex2 in these conditions remains an intriguing topic to be explored in the near future.
Using our Bex2 knock-out mice with the EGFP reporter, we were able to monitor the activation of the Bex2 locus in vivo. Although we cannot completely exclude the possibility that Bex2 is involved in its own expression, our present results showed that Bex2 locus activation in Bex2 knock-out mice was restricted to the cells or tissues that were positive for Bex2 transcripts in wild-type mice. For example, the brain and testis tissues had high EGFP expression in the Bex2 reporter mice, and these tissues also showed strong expression of Bex2 mRNA in wild-type mice. Many researchers have attempted to identify markers that are specifically expressed in a particular cell type, e.g. stem cells or cells of a specific differentiated lineage. Using these mice, we found that the Bex2 locus is not only active in the nervous system but is also specifically active in very rare cell types such as endocrine cells and several tissue stem/progenitor cells.
Although the endocrine cells of the gastrointestinal and neural systems are derived from different origins (42,43), they express common marker genes involved in the biosynthesis of neurotransmitters and have similar ultrastructural properties (44). Moreover, it has recently been shown that the differentiation of gut endocrine cells requires the basic helix-loop-helix transcription factors MATH1, Neurogenin 3, and NeuroD as well as several Sry-box-related transcription factors (45,46), as in the differentiation of nervous cells. Therefore, Bex2 expression in endocrine cells might be regulated by these transcription factors, which are commonly involved in nervous cell dif-ferentiation (47). We also demonstrated that Bex2 expression is highly correlated with the development of hepatic progenitor cells. Using several hepatic progenitor cell surface markers, such as Dlk1, CD133, CD13, and Liv2, we found that almost all of the CD133 ϩ Dlk1 ϩ cells in E13.5 fetal livers expressed EGFP in the reporter mice. Conversely, cells expressing Bex2-EGFP at the highest levels were all positive for the hepatoblast markers and were able to form hepatic colonies in culture at very high efficiency, thus confirming the utility of this EGFP reporter as a novel hepatic progenitor marker in the fetal liver. In addition, Bex2 also marked stem/progenitor cells of the adult pyloric epithelium and the hematopoietic system in these mice. Comparison of gene expression patterns between Bex2-EGFP-positive cells and Bex2-EGFP-negative cells in various tissues in this Bex2-EGFP reporter mouse may identify transcription factors and signaling molecules that are specific to these cell types and thus may lead to further understanding of the regulatory mechanisms of endocrine cell differentiation as well as the self-renewal and differentiation processes of stem/progenitor cells in future studies.
Our comparative gene expression analysis between hepatoblasts and adult hepatocytes revealed a close association of many stem/progenitor cell-specific genes and the X chromosome, similarly to the hematopoietic system (1). In addition, we found that a number of imprinted genes were specifically expressed in the stem/progenitor cells of the liver. The tight regulation of the imprinted gene network in stem cells has been documented previously in the lung (48), brain (49), and hematopoietic (50) systems. In addition, using a transgenic reporter mouse line, paternally expressed gene 3 (Peg3/Pw1), an imprinted gene located on chromosome 7, was recently shown to mark stem/progenitor cells in various tissues (38). Considering the close association between various tissue stem/progenitor cells and these X-linked and imprinted genes, the selective activation of X-linked and imprinted genes appears to be a common feature of stem/progenitor cells of various tissues. Exploring the mechanisms of how these genes, including Bex2, are transcribed may reveal the nature of tissue stem cells.
Collectively, by generating Bex2-EGFP knock-in mice, we have provided one example of using genes on the X chromosome to study neuronal/endocrine cells as well as stem/progenitor cells. This could be the basis for generating additional animal models with genetic modifications on the X chromosome for use in research in neuronal and stem cell biology.