• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Stem Cells. Author manuscript; available in PMC Jun 7, 2010.
Published in final edited form as:
PMCID: PMC2881625

Production of Green Fluorescent Protein Transgenic Embryonic Stem Cells Using the GENSAT Bacterial Artificial Chromosome Library


Transgenic green fluorescent protein (GFP) reporter embryonic stem (ES) cells are powerful tools for studying gene regulation and lineage choice during development. Here we present a rapid method for the generation of ES cells expressing GFP under the control of selected genes. Bacterial artificial chromosomes (BACs) from a previously constructed GFP transcriptional fusion library (Gene Expression Nervous System Atlas [GENSAT]) were modified for use in ES cells, and multiple BAC transgenic ES cell lines were generated. Specific GFP expression in transgenic cell lines was confirmed during neural differentiation marking neural stem cells, neuronal precursors, and glial progeny by Hes5, Dll1, and GFAP, respectively. GFP was dynamically regulated in ES cell progeny in response to soluble factors that inhibit Notch signaling and a factor that directs astroglial fate choice. Our protocols provide a simple and efficient strategy to utilize the whole GENSAT BAC library to create hundreds of novel fluorescent cell lines for use in ES cell biology.

Keywords: Neural differentiation, Embryonic stem cell, Bacterial artificial chromosome transgenesis, Fluorescent reporter


The creation of transgenic embryonic stem (ES) cells that express green fluorescent protein (GFP) under the control of specific promoters has emerged as a powerful method for monitoring gene expression in live stem cells or their differentiated progeny. Traditionally, transgenic stem cells were created by the site-specific targeting of GFP into a gene of interest or by randomly integrating DNA containing a specific promoter or enhancer elements adjacent to GFP [1]. There are strengths and weaknesses to each strategy. Targeted insertion usually results in correct transgene expression but is labor-intensive and results in a cell line with only one copy of the gene of interest, since GFP replaces the other endogenous copy. Random integration of promoter-driven GFP constructs requires mapping the promoter structure. After mapping, recombinant DNA techniques must be used to create truncated promoters that typically are smaller than 20 kilobases (kb) due to the practical limitations of conventional cloning methods [2]. These smaller, randomly integrated transgenes are frequently dysregulated [3-7].

One strategy to bypass these limitations is the use of bacterial artificial chromosome (BAC) transgenes. BACs are composed of large (up to approximately 350 kb) pieces of genomic DNA. In mice and zebrafish, BAC-based reporters properly regulate gene expression irrespective of integration site when introduced randomly into the genome as a transgene [3-7]. Therefore, BAC transgenesis combines the advantages of “knock-in” strategies using homologous recombination (specificity of expression) with that of transgenic approaches (multiple copies and ease of use). Moreover, unlike knock-in strategies that modify or eliminate the coding sequence, the genomic locus being studied is not modified. The increased transgene fidelity observed with BACs is likely due to the BAC’s large size, which insulates the integrated transgene from the effects of the surrounding chromatin structure. The larger size of BACs might also provide a more complete set of regulatory sequences [3-7]. However, while BACs have been used extensively for creating transgenic mice, little work has been done to develop BAC transgenesis for use in ES cells and in vitro differentiation assays.

Here, we show that BACs can be readily introduced into ES cells to produce cells that express GFP at defined stages of neural differentiation. Our work provides the tools sufficient to retrofit the entire GENSAT library, allowing stem cell researchers rapid access to more than 400 GFP-expressing BACs. In the postgenomic era, ES-cell-based GFP expression libraries should greatly facilitate efforts aimed at the understanding of gene function and could provide many cell lines useful in high throughput screens for developmental processes or drug discovery.

Materials and Methods


The GENSAT BAC for GFAP::GFP was constructed from RP24-331K9, Hes5::GFP from RP24-341I10, and Dll1::GFP from RP23-306J23 (see http://www.gensat.org for more information on each BAC). GENSAT BACs can be obtained from ATCC (http://www.atcc.org).

Retrofitting BACs for Mammalian Selection

The retrofitting protocol was adapted from Liu et al. [8] (see Fig. 1 for flowchart). Comprehensive protocols can be found in the supplemental materials online. Briefly, GENSAT BACs were moved from Escherichia coli strain DH10B (Fig. 1A) to the E. coli strain EL350 (Fig. 1B), a strain that inducibly expresses the Cre recombinase [3]. A selection cassette flanked by loxP sites was excised from the plasmid pL452 and gel-purified. This linear cassette was electroporated into Cre-expressing EL350 (Fig. 1C), and selection was performed on LB plates containing chloramphenicol and kanamycin (12.5 μg/ml each). Southern blots were performed to verify that the selection cassette had incorporated into the loxP site in the BAC backbone. During the course of these experiments, we discovered that low-level Cre expression in EL350 led to a small population of BACs that had excised the selection cassette despite continued kanamycin selection (Fig. 1). Therefore, we transferred the modified BACs from EL350 back to DH10B (Fig. 1E) before preparing large-scale DNA preparations for ES cell electroporation.

Figure 1
Schematic representation of the bacterial artificial chromosome (BAC) retrofitting protocol. GENSAT BAC DNA is isolated from the bacterial strain DH10B (A) and is transformed into electrocompetent EL350 (B). Electrocompetent EL350 that express Cre and ...

Preparing Retrofitted BAC DNA to Electroporate ES Cells

To produce purified DNA to electroporate into ES cells, we used the Psi Clone Big BAC DNA kit (Princeton Separations, Adelphia, PA, http://www.prinsep.com, PP-121) and followed the manufacturer’s protocol. Approximately 100 μg of BAC DNA was electroporated into 5.6 × 106 CJ7 or E14 ES cells. Electroporation was performed as described [9] with a few modifications. Briefly, 7 × 106 ES cells were resuspended in 1 ml of room-temperature phosphate-buffered saline (PBS), and 800 μl of this cell suspension was added to the DNA preparation. Electroporation was performed in a 4-mm cuvette at 250 V, 500 μF, and cells were left at room temperature for 10–20 minutes before plating onto two 10-cm dishes of Neo-resistant mouse embryonic fibroblasts (MEFs) (Primary Mouse Embryo Fibroblasts, PMEF-N; Chemicon/Specialty Media, Phillipsburg, NJ, http://www.specialtymedia.com). Selection in 500 μg/ml G418 was started 2 days after electroporation, and colonies were manually picked and grown clonally after 7–10 days of selection. For experiments using polyclonal pools of ES cells, all of the colonies from an electroporation were passaged together as a single line. Each polyclonal line was derived from the following number of clones: Dll1::GFP, 115; GFAP::GFP, 123; and Hes5::GFP, 281.

ES Cell Propagation

We used low-passage (between 10 and 30) CJ7, R1, and E14 ES cells in this study and maintained them primarily on Neo-resistant MEFs (Specialty Media). Transgenic ES cells were maintained in 500 μg/ml G418 on Neo-resistant MEFs. Prior to differentiation, BAC transgenic ES cells were passaged onto gelatin-coated tissue culture dishes for 1 day to remove MEFs, and selection was removed during neural induction.

Neural Induction

ES cells were directed to neural cell fates using two different protocols [10, 11]. For the GFAP::GFP specificity experiments, we used a modified serum-free floating culture of embryoid-body-like aggregates (SFEB) protocol [11] to cause neural induction, then triturated and plated single cells in N2 medium with or without 50 μg/ml ciliary neurotrophic factor (CNTF) in tissue culture dishes for 14 days. The medium used for the SFEB protocol was a 1:1 mixture of SRM and N2 (components listed in [10,12]). The medium was changed at day 3, and spheres were used at day 6. The SFEB protocol was also used to test Dll1::GFP and Hes5::GFP transgene sensitivity to the Notch inhibitor N-[N-(3,5-difluorophenacetyl-l-alanyl]-S-phenylglycine tert-butyl ester (DAPT) (10 μM; EMD Biosciences, Darmstadt, Germany, http://www.emdbiosciences.com, catalog number 565770). The MS5 stromal feeder protocol [10] was used to count the number of cells coexpressing markers for Dll1::GFP and Hes5::GFP and to test the responses of 24 independent Dll1::GFP and 24 Hes5::GFP clones (Fig. 2B) to DAPT.All data are derived from at least three independent experiments unless noted.

Figure 2
Polyclonal GENSAT bacterial artificial chromosome (BAC) transgenic ES cells regulate GFP expression in response to soluble factors known to regulate the endogenous gene. (A): Neural progeny from Dll1::GFP and Hes5::GFP polyclonal ES cells are differentially ...

Indirect Immunofluorescence

Cultured cells were washed once with PBS before fixation with 4% paraformaldehyde/0.15% picric acid for 15 minutes at room temperature. Cells were washed three times with PBS before permeabilization with wash buffer for at least 2 minutes (0.3% triton X-100, 1.0% bovine serum albumin in PBS). Primary antibody (diluted in wash buffer) was added to the cells for 3 hours at room temperature. Cells were washed three times with wash buffer before the addition of secondary antibody (diluted in wash buffer) for 1 hour. Cells were washed three times with wash buffer and once with PBS and were stored at 4°C in 50% PBS and 50% glycerol before acquiring images.

Antibodies for Immunofluorescence

The rabbit polyclonal antibody against GFP was purchased from Molecular Probes (Eugene, OR, http://probes.invitrogen.com) (A11122, 1:600). A monoclonal antibody against glial fibrillary acidic protein (GFAP) was purchased from Chemicon (Temecula, CA, http://www.chemicon.com) (MAB 3580 1:1,000). The TuJ1 antibody was purchased from Covance (Princeton, NJ, http://www.covance.com) (MMS-435P). Goat anti-mouse Alexa Fluor 568 (A11004) and goat anti-rabbit Alexa Fluor 488 (A11008) secondary antibodies were purchased from Molecular Probes (1:500).

Fluorescence-Activated Cell Sorting for Quantification

For the quantification experiments in Figures 1A, 1B, and and3C,3C, Dll1::GFP and Hes5::GFP SFEBs were expanded for 3 days before the addition of 10 μM DAPT to half of the samples from days 3 to 6. On day 6, control and DAPT-exposed neural spheres were dissociated in 0.05% trypsin-EDTA for 7 minutes before fluorescence-activated cell sorting (FACS) into GFP-positive and -negative populations on a MoFlo Sorter (DakoCytomation, Glostrup, Denmark, http://www.dakocytomation.com).

Figure 3
Analysis of clonally derived Notch pathway transgenic lines Dll1::GFP clone 10 and Hes5::GFP clone 1. (A): Dll1::GFP clone 10 expresses GFP primarily in TuJ1-positive cells and not Nestin-positive cells. (B): Hes5::GFP clone 1 expresses GFP primarily ...

Quantitative Polymerase Chain Reaction to Determine Specificity of Transgene Expression

As above, Dll1::GFP and Hes5::GFP day 6 SFEB-derived neural progenitors were sorted into GFP-positive and -negative populations on a MoFlo Sorter (DakoCytomation) before RNA isolation. RNA was isolated using a QIAshredder (Qiagen, Hilden, Germany, http://www1.qiagen.com, catalog number 79654) and the RNeasy mini kit (Qiagen, catalog number 74104). RNase-free DNase (Qiagen, catalog number 79254) was added to the columns to eliminate genomic DNA. Reverse transcription was performed using random primers (Invitrogen, Carlsbad, CA, http://www.invitrogen.com, number 48190-011) and SuperScript RTII (18064-014). The amounts of RNA used for reverse transcription were 700 ng (Dll1::GFP) and 1,500 ng (Hes5::GFP). Of the resultant cDNA, 20 ng was used in a quantitative-polymerase chain reaction (PCR) using an iCycler (Bio-Rad, Hercules, CA, http://www.bio-rad.com) and predesigned TaqMan Gene Expression Assays for Hes5 (Mm00439311_g1, ABI), Dll1 (Mm00432841_m, ABI), HPRT (Mm00446968_m1), and β2-microglobulin (mCG11606). Amplification was carried out for 40 cycles (95°C for 15 seconds, 60°C for 1 minute). Triplicate threshold cycle values were averaged. Dll1 relative amounts were normalized to Hprt and Hes5 to B2M. Fold changes were calculated using the equation 2−DCT where DCT = (CT[gene of interest] – CT[normalization control]) [13].


Integration Site Effects on GENSAT BAC Transgene Expression

In mice, BACs have been shown to direct specific transgene expression independent of the BAC’s integration site [3-7]. However, transgenic mice are created by injection of BAC DNA directly into fertilized mouse eggs. Here, we modified GENSAT BACs by introducing a selection cassette followed by electroporation of the modified BAC into ES cells, drug selection, and isolation of transgenic stem cells. To examine possible effects of the integration site on transgene expression in ES cells, we electroporated ES cells with three separate BACs and passaged all drug-resistant ES cells as polyclonal cell lines. In this way, we could assay hundreds of BAC transgene integration sites by differentiating these pooled, polyclonal cell lines derived from each BAC.

We first examined the expression of two Notch-related genes after neural induction. It is thought that Notch ligands are expressed on the surface of neuroblasts where they interact with the Notch receptors to regulate the pool of neural stem cells [14-17]. To watch Notch signaling in differentiating neural cells, we created transgenic ES cells with delta-like1::GFP (Dll1— one of the Notch ligands) and Hes5::GFP (activated by Notch signaling). Neural induction of transgenic Dll1::GFP and Hes5::GFP ES cells using a serum-free culture system [11] resulted in spherical aggregates with unique patterns of GFP expression within spheres (Figs. (Figs.2A,2A, ,4).4). In Hes5::GFP spheres, the foci of GFP expression were usually located on the outside of neural spheres and in regions where two spheres contacted each other. Dll1::GFP progeny showed scattered expression of GFP throughout the sphere (Figs. (Figs.2A,2A, ,4).4). Inhibiting Notch signaling should reduce downstream Notch gene expression (Hes5) and increase Notch ligand expression (Dll1). Notch inhibition should cause Hes5-expressing neural stem cells to commit to a neuronal fate, activating Dll1 expression [15-17]. Therefore, we predicted that the Notch inhibitor DAPT should reduce GFP expression in Hes5::GFP spheres and enhance GFP expression in Dll1::GFP spheres if the BAC transgenes were correctly regulated. To test this prediction, we differentiated Dll1::GFP and Hes5::GFP ES cells into neural spheres. From days 3 to 6, half of each culture was exposed to DAPT. Direct microscopic observation revealed increased fluorescence in Dll1::GFP spheres that received DAPT when compared with control cultures without DAPT (Fig. 2A). Conversely, Hes5::GFP spheres exposed to DAPT showed decreased fluorescence (Fig. 2A). To quantitate this observation, neural spheres were dissociated on day 6, and FACS analysis was performed. The number of GFP+ cells derived from Dll1::GFP spheres in control cultures was 5.5 ± 2.1% and increased to 9.9 ± 2.0% in cultures exposed to DAPT. In Hes5::GFP control spheres, 5.0 ± 2.0% expressed GFP compared with 0.3 ± 0.2% when exposed to DAPT.

Figure 4
GFP expression patterns in Hes5::GFP clone 1 and Dll1::GFP clone 10 neural spheres. Neural spheres were cultured for 7 days in vitro before live imaging of GFP fluorescence. Five representative spheres of each cell line are shown. Scale bar = 400 μ ...

The low numbers of cells expressing GFP suggested that many G418-resistant clones did not express GFP. To verify this, and to confirm that the polyclonal population was providing us with an accurate representation of the behavior of individual clonal lines, we isolated and characterized 24 independent clones for the Dll1::GFP and Hes5::GFP BACs. Each of the individual BAC transgenic lines were cocultured in duplicate with MS5 cells to cause neural induction, and transgenic cells differentiated for 9 days before addition of DAPT for 2 days. For each BAC, 10 out of 24 lines expressed GFP (Fig. 2B). All 10 Dll1::GFP clones increased GFP expression compared with controls, while all 10 Hes5::GFP clones decreased GFP expression in response to DAPT. Note that a feeder layer was used to cause neural induction in this experiment; therefore, the percentage of GFP-expressing cells for each cell line can only be compared to itself (in the presence or absence of DAPT). The reason for this is that there are fluctuations in the number of cells initially seeded to begin the differentiation between the 48 different clones. These data demonstrated that approximately 40% of the clones expressed GFP after neural induction and that each BAC increased or decreased the expression of GFP in response to the Notch inhibitor DAPT. These data corroborate the observations made with the polyclonal ES cell lines.

In contrast to the Notch-related transgenic cell lines, the GFAP::GFP polyclonal cell lines did not fluoresce after neural induction. GFAP::GFP should be expressed in astrocytes, a neural cell type that appears later in development and after extended culture during in vitro differentiation of ES-derived or primary neural precursors. Astrocyte production can be further enhanced by exposure to the growth factor CNTF [12]. In agreement with these predictions, extended culture of GFAP::GFP ES-derived neural precursors led to the appearance of cells that were immunoreactive for both GFP and GFAP, and the addition of CNTF further increased the number of doublepositive cells (Fig. 2C). Given the polyclonal nature of the ES cells used in these experiments, our data indicate that BAC transgenes are regulated correctly in ES cells without obvious influence from the BAC integration site.

Specificity of Clonal BAC Transgenic Cell Lines

To further assess ES cell transgene specificity, we isolated and characterized single-cell-derived BAC transgenic ES cell clones. The percentage of drug-resistant clones that expressed GFP after neural induction varied depending on the BAC selected but was at least 10% (for example, see Fig. 2B, both Dll1::GFP and Hes5::GFP yielded 10 fluorescent colonies out of 24 total colonies, or approximately 40%). While GFP expression patterns appeared to be identical among multiple lines for each BAC examined, there were clear differences in the strength of GFP expression noted between lines. For each BAC, one representative cell line was studied in detail and is presented below.

Dll1::GFP Clone 10

Similar to the results obtained with mixed clones, Dll1::GFP clone 10 showed widespread scattered expression within differentiating neural colonies (Fig. 4). Immunofluorescent analysis of attached cells (Fig. 3A) demonstrated that 87.6 ± 2.3% of the GFP-positive cells were immunopositive for TuJ1, while only a minority of the cells expressed GFP only (9.7 ± 1.9%) or TuJ1 alone (2.7 ± 0.5%). Few of the GFP-positive cells coexpressed Nestin (7.5 ± 1.2%%+ - Fig. 3). Time-lapse microscopy revealed that Dll1::GFP-positive cells were motile and divided frequently during neural differentiation (supplemental online Video 1). Immunofluorescence analysis with antibodies against proliferating cell nuclear antigen and GFP confirmed our observation that many of the Dll1::GFP-positive cells were dividing (data not shown). This suggests that Dll1 is expressed in neuroblasts, dividing precursor cells that adopt postmitotic neuronal fates. Upon further differentiation we observed GFP-positive cells with mature neuronal morphologies (supplemental online Video 2).

To verify that GFP expression increases when inhibiting Notch, Dll1::GFP clone 10 neural spheres were exposed to DAPT as outlined above. Fluorescence microscopy revealed increased fluorescence after exposure to DAPT, and flow cytometry showed that the number of GFP-positive cells more than doubled to 45 ± 7.5% in cultures with DAPT when compared with control cultures (22.1 ± 3.7%, Fig. 3C). We also found that GFP expression identified cells expressing high levels of Dll1 transcripts. GFP-positive and -negative cells were separated by flow cytometry before performing quantitative PCR for Dll1 (Fig. 4D). We observed a 20.6-fold increase in Dll1 transcripts in Dll1::GFP-positive cells relative to Dll1::GFP-negative cells (see Materials and Methods).

Hes5::GFP Clone 1

Most Hes5::GFP-positive cells had a more flattened appearance and coexpressed Nestin (81.0 ± 2.5%, Fig. 3B). Fewer cells expressed Nestin only (16.1 ± 3.2%) or GFP only (2.9 ± 1.2%). Hes5::GFP clone 1 cells were mostly negative for the neuronal marker TuJ1 (11.2 ± 1.7%, Fig. 3B). Hes5::GFP neural spheres primarily had fluorescent cells clustered on the outside of spheres (Fig. 4), and incubation of Hes5::GFP neural spheres with DAPT decreased the proportion of GFP-positive cells from 33.0 ± 8.3% to 4.62 ± 2.2% (Fig. 3C). Separation of Hes5::GFP-positive and -negative cells by FACS followed by reverse transcription-polymerase chain reaction (RT-PCR) analysis demonstrated that Hes5 transcripts were highly enriched in the GFP-positive fraction (Fig. 3D). Quantitative PCR experiments showed that Hes5::GFP-positive cells contained 19.7-fold more Hes5 mRNA relative to the GFP-negative population (see Materials and Methods).

GFAP::GFP Clone 10

The ES cell line GFAP::GFP clone 10 only fluoresced after culture conditions that caused the appearance of astrocytes, similar to the GFAP::GFP mixed clones described above. Indirect immunofluorescent analysis using antibodies against GFP and GFAP (Fig. 5) demonstrated a striking degree of colocalization (89.5 ± 4.3%), although cells expressing only GFP (5.5 ± 1.6%) or GFAP (4.7 ± 2.9%) appeared in the cultures. These single-labeled cells were nearly always found near each other and the double-positive cells.

Figure 5
GFAP::GFP clone 10 expressed GFP in GFAP immunoreactive cells. Immunofluorescence demonstrated that most GFP-positive cells also express GFAP. The merged images include 4′,6-diamidino-2-phenylindole to demonstrate the specificity of the GFP and ...


We provide a method for the production of GFP BAC transgenic ES cells that obviates the need to map promoters or enhancers. Transgenic ES cell lines can be generated in less than 2 weeks, and multiple BACs can be modified in parallel. This method makes use of the previously constructed and characterized GENSAT BAC library, comprised of more than 400 BACs with GFP transcriptional fusions that have been used to make transgenic mice (http://www.gensat.org and [18, 19]). Our data demonstrate that such BACs can be modified for use in ES cells and that BAC ES lines express GFP in the correct cell type largely independent of integration site. Furthermore, we show that three different BACs can be used to make ES cell lines that dynamically regulate GFP in response to soluble factors that normally regulate the gene of interest.

Multiple lines of evidence suggest that the BAC transgenic ES cell lines function correctly. (a) Timing of expression: we did not find any ES cell clones that expressed appreciable amounts of GFP as ES cells (data not shown). As both Notch BAC ES cell lines differentiated into neural cells, they began to express GFP within the first week of differentiation when Notch components are detectable by RT-PCR (data not shown). However, the GFAP::GFP lines only began to express GFP into the third week of differentiation, when astrocytes begin to develop in the cultures [10]. (b) Patterns of GFP expression: each BAC has a characteristic pattern of GFP expression in differentiating ES-derived neural colonies. Hes5::GFP is expressed in continuous patches that are largely on the outside of neural spheres, whereas Dll1::GFP is scattered throughout the spheres (Fig. 4). GFAP::GFP is not expressed in neural spheres that contain neural progenitors and neurons but not glial cells (data not shown). As noted above, GFP-positive astrocytes (GFAP+) only appeared after extended culture. (c) Response to soluble factors: the Notch-related Hes5::GFP and Dll1::GFP neural spheres displayed opposing phenotypes after exposure to DAPT, a chemical inhibitor of the Notch pathway. DAPT exposure reduced Hes5::GFP expression while it increased Dll1::GFP expression as predicted. In GFAP::GFP lines, the addition of a growth factor that instructs astrocytic cell fate (GFAP+ cells) increased the number of GFP-positive astrocytes compared with control cultures that were not exposed to CNTF. (d) GFP expression in different cell types: immunofluorescent experiments (Figs. (Figs.3,3, ,5)5) demonstrated that Hes5::GFP and Dll1::GFP labeled different neural cell populations. Dll1::GFP showed a high degree of colocalization with TuJ1, consistent with neuronal labeling (Fig. 3A). Time-lapse microscopy revealed that many of these cells were actively dividing, consistent with the hypothesis that the Dll1::GFP-positive cells are mainly composed of neuroblasts and postmitotic neurons (supplemental Videos 1 and 2). Hes5::GFP, on the other hand, primarily labeled Nestin+ cells (Fig. 3B), and GFAP::GFP-positive cells were around 95% GFAP-positive astrocytes. (e) mRNA enrichment in GFP+ cells: Dll1::GFP and Hes5::GFP-positive neural cells contained approximately 20-fold higher levels of Dll1 and Hes5 mRNA compared with GFP-negative cells from Dll1::GFP and Hes5::GFP cell lines, respectively (Fig. 3D). Taken together, our data demonstrate that GENSAT BACs can be used to rapidly produce ES cell lines that fluoresce during the transcription of desired genes.

BAC transgenes have been shown to more reliably report on an endogenous gene in vivo when compared directly with smaller transgenes [3-7]. These studies showed that BAC transgenes are less susceptible to mosaic or position effect variegation, which is likely due to transgene silencing. It is also possible that the more extensive amount of sequence present in BACs is required for robust transgene expression, since even the inclusion of insulator elements on conventional transgenes only partially rescues mosaic expression [4]. We examined the effects of integration site in two ways: by producing polyclonal transgenic ES cell populations that consisted of all G418-resistant colonies passaged together, and by assaying 48 independently-derived clones for transgene activity (Fig. 2). In each case, we observed little influence of the integration site on transgene fidelity. For example, we did not observe GFP expression in the ES cell stage of any BAC examined, nor did we observe inappropriate GFAP::GFP expression after neural induction. While expression fidelity did not appear to be influenced by integration site, GFP expression levels did vary between clones. For the GFAP::GFP clone 10, we observed astrocytes (GFAP+) that did not express GFP (Fig. 5) and vice versa. There are a few possible reasons for this observation. One possibility is that the transgene is being silenced in a minority of the cells. Another possibility is that there are differences between promoter activity and protein stability. It is important to remember that all conventional transgenes express GFP during transcription of the gene of interest. Post-transcriptional regulation mechanisms, such as mRNA and protein stability, are not accounted for in such transgene strategies. Therefore, we believe that this is the likely explanation for the observation of single GFP- and GFAP-stained cells. Further experiments will be necessary to clarify this discrepancy.

The routine production of ES reporter cell lines will be crucial for many applications, even if some transgene silencing does occur. Such cell lines will allow the enrichment of specific neural cell types and developmental stages for preclinical applications in animal models of disease. BAC transgenic ES cells will also permit the identification and quantification of specific ES-derived neurons in response to extracellular stimuli that damage or protect particular neuron types. Such neuron types could include midbrain dopamine neurons or spinal motor neurons that can be readily obtained from mouse ES cells [20-22] and that are critical for the study of Parkinson disease and amyotrophic lateral sclerosis. Additional important applications will include genomic or proteomic studies of FACS-purified cell types within mixed neural cultures, the prospective isolation of ES-derived neural progeny for cell fate analysis and transplantation, and large-scale chemical, RNA interference, or cDNA screens to define molecules involved in stem cell fate specification.

The GENSAT BAC library was initially constructed to identify cell types during nervous system development [18, 19]. This project has provided hundreds of BAC transgenic mice that label defined classes of neural cells (http://www.gensat.org and [18, 19]). Our data show that this existing resource can be readily modified to study neural differentiation of ES cells in vitro. One advantage of using this existing BAC library is that the GFP transcriptional fusions have already been engineered, constructed, and tested in the production of transgenic mice that are available for distribution. The only modification necessary is to retrofit each BAC with a mammalian selection cassette. The efficiency of the technique presented here suggests that it should be possible to generate BAC transgenic ES cell lines for the whole GENSAT library. Access to GFP-expressing ES cell lines for more than 400 central nervous system related genes will provide an essential resource for the stem cell field.

Supplementary Material

Supplemental Movie 1

Supplemental Movie 2


We thank Nathaniel Heintz (Rockefeller University) for conversations, advice, and reagents to perform these experiments. In addition, we thank Neal Copeland and Nancy Jenkins (NCI) for providing the reagents necessary to retrofit the BACs. The entire Studer Lab provided helpful feedback, but Yechiel Elkabetz and Sabrina Desbordes deserve special thanks for constructive discussions. We also thank Jan Hendrikx and Cris Bare for help with flow cytometry and Agnes Viale and Juan Li for help with qPCR. Dennis Kunkel Microscopy, Inc. (Kailua, HI, http://www.denniskunkel.com) kindly provided the E. Coli image used in Figure 1.



The authors indicate no potential conflicts of interest.

See www.StemCells.com for supplemental material available online.


1. Hogan B, Beddington R, Costantini F, et al., editors. Manipulating the Mouse Embryo: A Laboratory Manual. Cold Spring Harbor Laboratory Press; Plainview, NY: 1994. Production of transgenic mice; pp. 219–225.
2. Griffiths AJ, Gelbart WM, Lewontin RC, et al. Modern Genetic Analysis: Integrating Genes and Genomes. W.H. Freeman and Company; New York: 2002. Methods of genetic manipulation; p. 225.
3. Lee EC, Yu D, Martinez de Velasco J, et al. A highly efficient Escherichia coli-based chromosome engineering system adapted for recombinogenic targeting and subcloning of BAC DNA. Genomics. 2001;73:56–65. [PubMed]
4. Reizis B, Leder P. The upstream enhancer is necessary and sufficient for the expression of the pre-T cell receptor α gene in immature T lymphocytes. J Exp Med. 2001;194:979–990. [PMC free article] [PubMed]
5. Jessen JR, Meng A, McFarlane RJ, et al. Modification of bacterial artificial chromosomes through Chi-stimulated homologous recombination and its application in zebrafish transgenics. Proc Natl Acad Sci U S A. 1998;95:5121–5126. [PMC free article] [PubMed]
6. Yang XW, Model P, Heintz N. Homologous recombination based modification in Escherichia coli and germline transmission in transgenic mice of a bacterial artificial chromosome. Nat Biotechnol. 1997;15:859–865. [PubMed]
7. Zhang XM, Ng AH, Tanner JA, et al. Highly restricted expression of Cre recombinase in cerebellar Purkinje cells. Genesis. 2004;40:45–51. [PubMed]
8. Liu P, Jenkins NA, Copeland NG. A Highly Efficient Recombineering-Based Method for Generating Conditional Knockout Mutations. Genome Research. 2003;13:476–484. [PMC free article] [PubMed]
9. Wurst W, Joyner AL. Production of targeted embryonic stem cell clones. In: Joyner AL, editor. Gene Targeting: A Practical Approach. Oxford University Press; Oxford: 1993. pp. 33–61.
10. Barberi T, Klivenyi P, Calingasan NY, et al. Neural subtype specification of fertilization and nuclear transfer embryonic stem cells and application in parkinsonian mice. Nat Biotechnol. 2003;21:1200–1207. [PubMed]
11. Watanabe K, Kamiya D, Nishiyama A, et al. Directed differentiation of telencephalic precursors from embryonic stem cells. Nat Neurosci. 2005;8:288–296. [PubMed]
12. Johe KK, Hazel TG, Muller T, et al. Single factors direct the differentiation of stem cells from the fetal and adult central nervous system. Genes Dev. 1996;10:3129–3140. [PubMed]
13. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2−DDCT method. Methods. 2001;25:402–408. [PubMed]
14. Henrique D, Adam J, Myat A, et al. Expression of a Delta homologue in prospective neurons in the chick. Nature. 1995;375:787–790. [PubMed]
15. de la Pompa JL, Wakeham A, Correia KM, et al. Conservation of the Notch signalling pathway in mammalian neurogenesis. Development. 1997;124:1139–1148. [PubMed]
16. Grandbarbe L, Bouissac J, Rand M, et al. Delta-Notch signaling controls the generation of neurons/glia from neural stem cells in a stepwise process. Development. 2003;130:1391–1402. [PubMed]
17. Shen Q, Goderie SK, Jin L, et al. Endothelial cells stimulate self-renewal and expand neurogenesis of neural stem cells. Science. 2004;304:1338–1340. [PubMed]
18. Gong S, Zheng C, Doughty ML, et al. A gene expression atlas of the central nervous system based on bacterial artificial chromosomes. Nature. 2003;425:917–925. [PubMed]
19. Heintz N. Gene expression nervous system atlas (GENSAT) Nat Neurosci. 2004;7:483. [PubMed]
20. Lee SH, Lumelsky N, Studer L, et al. Efficient generation of midbrain and hindbrain neurons from mouse embryonic stem cells. Nat Biotechnol. 2000;18:675–679. [PubMed]
21. Kawasaki H, Mizuseki K, Nishikawa S, et al. Induction of midbrain dopaminergic neurons from ES cells by stromal cell-derived inducing activity. Neuron. 2000;28:31–40. [PubMed]
22. Wichterle H, Lieberam I, Porter JA, et al. Directed differentiation of embryonic stem cells into motor neurons. Cell. 2002;110:385–397. [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...