The potential of embryonic stem cells to differentiate to all cell types makes them an attractive model for development and a potential source of cells for transplantation therapies. Candidate approaches have identified individual genes and proteins that promote the differentiation of embryonic stem cells to desired fates. Here, we describe a rapid large-scale screening strategy for the identification of genes that influence the pluripotency and differentiation of embryonic stem cells to specific fates, and we use this approach to identify genes that induce neuron formation. The power of the strategy is validated by the fact that, of the 15 genes that resulted in the largest increase in neuron number, 8 have previously been implicated in neuronal differentiation or survival, whereas 7 represent novel genes or known genes not previously implicated in neuronal development. This is a simple, fast, and generally applicable strategy for the identification of genes promoting the formation of any specific cell type from embryonic stem cells.
Disclosure of potential conflicts of interest is found at the end of this article.
The pluripotency and unlimited growth of embryonic stem (ES) cells make them an attractive source of differentiated cells for cell therapy in a variety of human diseases. There are, however, several hurdles for the realization of this prospect, with one of the greatest being to find ways to differentiate ES cells to specific cell fates.
Our rapidly increasing understanding of the extracellular signals that direct embryonic development has aided in the improvement of protocols to differentiate ES cells to, for example, motor neurons . However, in most cases, our knowledge is incomplete regarding the molecular cascades that govern the differentiation of specific cell fates, or they may potentially be too complex to reproduce in vitro.
The orchestration of extracellular signals imposes the expression of a complement of transcriptional regulators, which direct differentiation. In the case of motor neurons, for example, positional cues induce the expression of the homeodomain transcription factor Mnr2/Hb9, which is necessary and sufficient to direct the differentiation of spinal progenitors to motor neurons . Intrinsic determinants necessary and sufficient for the generation of dopaminergic neurons have similarly been identified . Thus, in at least some cases, single genes play a pivotal role in the induction of specific cell types and can be used to drive the formation of a desired cell type. Interestingly, there are several protocols that promote the generation of dopaminergic neurons through the addition of extracellular factors [4, 5], but none are nearly as efficient as those acquired by directly expressing a master regulator gene . Identifying such genes may shed light on embryonic development and allow the efficient generation of cells for therapy.
The often complex nature of the extracellular milieu, both when it comes to the number of factors and their temporal regulation, makes the strategy of identifying individual or small numbers of key genes an attractive route for the directed differentiation of ES cells. Unintended genetic manipulation of cells for transplantation is undesirable due to the risk of cellular transformation and tumor development. However, transient gene expression or even administration of recombinant intracellular proteins coupled to cell penetrable molecules may circumvent this problem. In addition, other promising gene therapy approaches include human artificial chromosomes that can be stably maintained without adversely affecting the host cell  and targeted replacement of loci .
We report a method for the efficient gene transfer of an expression library into monolayer murine ES cells carrying a cell fate-specific fluorescent reporter. A functional screening scheme was designed to identify genes which, when overexpressed, increase the proportion of neurons in a culture after 4 days of differentiation (Fig. 1). Unique cDNAs from an arrayed expression library were combined into pools that were used to transfect low-density cultures of ES cells carrying a fluorescent marker (enhanced green fluorescent protein [eGFP]) under the control of the Tα1 early neuronal promoter. Following differentiation, cultures were dissociated, and the frequency of eGFP-positive cells was analyzed by flow cytometry. This method allows efficient high throughput identification of genes that promote cell differentiation to specific fates.
Materials and Methods
Culture Conditions and Creation of Reporter Cell Lines
E14 ES cells were maintained without feeder cells, on 0.1% gelatin (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) coated plates (Costar, Lowell, MA, http:www.corning.com), in Glasgow Minimum Essential Medium (Sigma) supplemented with 5% fetal bovine serum (FBS; HyClone, Logan, UT, http://www.hyclone.com), 5% KnockOut Serum Replacement (KSR; Gibco, Grand Island, NY, http://www.invitrogen.com), 2 mM l-glutamine (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), 0.1 mM nonessential amino acids (Invitrogen), 1 mM sodium pyruvate (Invitrogen), 0.1 mM 2-mercaptoethanol (Sigma), and 1,000 units of leukemia inhibitory factor (LIF; Chemicon, Temecula, CA, http://www.chemicon.com). For minimal differentiation, the ES cells were cultured in a 50/50 medium of Dulbecco's modified Eagle's medium/F12 and neurobasal complemented with N2, B27 (containing transferrin, selenium, hormones, and vitamins important for neural cell survival  [Invitrogen]), 20 μg/ml human insulin (Invitrogen), and 150 μg/ml bovine serum albumin (Sigma). Tα1-eGFP ES cells were established with reporter construct Tα1-eGFP (kindly given by Oliver Brüstle, Institute of Reconstructive Neurobiology, Life and Brain Center, University of Bonn and Hertie Foundation, Bonn, Germany). Tα1 α-tubulin expression in early neurons has been well documented , and the 1.1 kilobases (kb) of the 5′ flanking region direct expression of eGFP in developing neurons. Following differentiation of picked ES cell colonies in N2/B27 medium, cells were analyzed for eGFP fluorescence and immunoreactivity against Nestin (Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww), βIII-tubulin (Berkeley Antibody, Princeton, NJ, http://www.crpinc.com), tyrosine hydroxylase and peripherin (both Chemicon), glial fibrillary acidic protein (GFAP) (Sigma), O4 (Chemicon), and Oct-4 (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com).
Transfection of ES Cells
ES cells were plated on gelatinized 6-well plates (Costar) at a density of 1.75 × 105 cells per well. Twenty hours later, cells were washed with Opti-MEM (Invitrogen) and transfected with a mixture of 5 μg of DNA and 10 μl of Lipofectamine 2000 (Invitrogen) diluted in a total volume of 1,000 μl of Opti-MEM per well. The transfection was allowed to continue for 4 hours before the transfection mixture was removed, and ES cell medium containing 10% FBS LIF KSR was added to the cells. Sixteen hours later, cells were changed to N2/B27 medium, and medium was changed every day. Cells were usually differentiated for 4 days in the N2/B27 medium before flow analysis was performed on a BD Aria (BD Biosciences, San Diego, http://www.bdbiosciences.com). The expression of the transfected genes was controlled by either of the general promoters of cytomegalovirus (CMV) or CMV-enhanced β-actin promoter. pEGFP-N1 (BD Biosciences), which contains the same promoter (CMV) and termination signals as library cDNAs, was used to analyze transfection efficiency. All cell fate-inducing activity of the introduced cDNA was compared with the lineage choice of CMV-β-galactosidase transfected cells.
Plates 1–84 were acquired from Geneservice (Cambridge, U.K., http://www.geneservice.co.uk) and maintained as glycerol stocks in −80°C. For pooling, a multichannel pipettor was used to scrape and transfer Escherichia coli to 2 ml deep-well 96-well plates containing 1.5 ml of Luria-Bertani medium and carbenicillin. Plates were grown shaking for 20–24 hours at 37 degrees, and 32 wells were pooled for DNA extraction. DNA was isolated using the PureYield Endotoxin Free Midiprep System according to manufacturer's directions (Promega, Madison, WI, http://www.promega.com) and analyzed by spectrophotometry for purity.
Fluorescence-Activated Cell Sorting Analysis of Cell Fate
Cells were trypsinized after 4 days in minimal medium, and eGFP-positive cells were detected by flow cytometry using BD FACSCalibur or FACSAria. Gates for the positive population were established by placing square gates directly above the singlet population (based on side-scatter width) defined as negative in β-galactosidase control transfections. Listed genes were confirmed by cycle sequencing. One clone, 58H05, was found by BLAST analysis to be oxysterol binding protein 6, not RIKEN cDNA 2810475A17 gene, as seen in supplemental online data 2. Statistical analyses were performed with GraphPad Prism and StatsDirect using one-way analysis of variance and Dunnett's post-test.
Polymerase Chain Reaction
Following 0–5 days in N2/B27 minimal differentiation medium, ES cells were lysed in TRIzol (Invitrogen), and total RNA was prepared according to directions. Random primed cDNA was synthesized with SuperScript II (Invitrogen). Intron-spaced primer sequences are available upon request.
Monolayer Culture Is Permissive for Diverse Cell Fates
To establish a screening procedure for genes affecting ES cell fate (Fig. 1), we first examined the differentiation potential of mouse ES cells under minimally instructive conditions and differentiated low-density cultures on a gelatin-coated surface in medium lacking LIF and serum but in the presence of B27, N2, and insulin [8, 10]. Initially, all cells expressed Oct-4 (Fig. 2A), a marker of pluripotency . ES cells cultured in defined medium at low density in an adherent monolayer culture readily adopt neural characteristics [10, 12, –14]. After 2 days, Nestin-immunoreactive neural precursors were found widely distributed throughout the culture (Fig. 2A). Over the course of 4 days, a small percentage (1%–2%) of the ES cells differentiated to βIII-tubulin, a specific marker of neurons. A subset of these neurons was immunoreactive to markers associated with mature cells, such as tyrosine hydroxylase, a marker of dopaminergic neurons (Fig. 2A).
We assessed by reverse transcription-polymerase chain reaction the expression of markers of the three different germ layers over 5 days in the minimal differentiation medium (Fig. 2B). Otx2, a marker of anterior neurectoderm, increased with time in culture. Brachyury, a mesoderm-specific marker, also increased over 5 days, whereas α-fetoprotein, a marker of visceral endoderm , was undetectable. These results show that monolayer differentiation produces heterogeneous cell types over the course of a few days.
In order to better define the composition of monolayer cultures, we examined Oct-4 and βIII-tubulin localization by immunohistochemistry. Examination of cultures over 4 days showed accumulation of Oct-4-immunoreactive cells near the center of ES cell clusters, whereas neurons appeared near the periphery (Fig. 2C). The reduction of Oct-4-positive cells was asynchronous in the culture and, following 4 days in differentiation medium, there were still clusters of cells stained with Oct-4 (Fig. 2C). Cells immunoreactive to antibodies directed against peripherin (Fig. 2C), 5-hydroxytryptamine (serotonin), and GABA (data not shown) were also observed near the border of cell clusters, indicating diverse and mature neuronal cell fates. Continued culture of the ES cells in the minimal differentiation medium for up to 14 days favored the formation of cells immunopositive for the astroglial cell marker GFAP and the oligodendrocyte marker O4 (Fig. 2C).
Next, we examined if minimal conditions promoted neuroectodermal cell fates at the expense of mesodermal-derived tissues. In monolayer cultures differentiated for 10 days, we detected beating cells, indicating the presence of cardiomyocytes. Differentiation of an ES cell line containing a GFP insertion in the Nkx2.5 locus, a transcription factor specifically expressed in developing cardiomyocytes , confirmed the presence of nascent heart tissue in the beating cells (Fig. 2C and supplemental online data 1). These results show that, whereas diverse neural markers are present under minimal culture conditions, mesodermal fates are also possible. Thus, differentiation of monolayer cultures under minimal conditions is permissive for the formation of ectodermal and mesodermal cell fates.
Efficient Gene Delivery to ES Cells
Heterogeneous culture maturation over several days provides an opportunity to influence several different steps in the differentiation process from ES cells to mature, post-mitotic cells. In order to examine if a particular cell fate could be promoted by overexpression of genes, we procured a Mouse Genome Committee rearrayed IRAV library of approximately 8,000 full-length mouse cDNAs in an expression vector under the general CMV promoter. The prospect of introducing a library of full-length cDNAs into ES cells confronted us with the requirement for an efficient and easy gene delivery method. Using a modified liposome-based transfection procedure (Materials and Methods), we were able to reliably achieve greater than 80% transfection efficiency of monolayer cultures as detected by flow cytometry for CMV-eGFP expression (Fig. 3A, 3B). After 4 days, eGFP expression decreased to approximately 40% of all cells and decreased dramatically thereafter (Fig. 3A, 3B), consistent with a loss of transient plasmids. This high rate of transfection suggests that differentiating cells will be accessible over the course of at least 4 days to exogenously expressed genes from plasmid DNA that may influence their fate.
Measuring the differentiative capacity of each gene individually is laborious; therefore, we sought to refine our method to include pooled plasmids at a frequency that still allows for the detection of the effect of individual genes. Dilutions (1:100, 1:50, 1:32, 1:8) of the CMV-eGFP plasmid among other plasmids resulted in a preferable pool size of 32 full-length cDNAs, based on observed eGFP expression. Using this dilution, we were able to demonstrate that a single gene of a transfected pool will be expressed by approximately 70% of cells, as indicated by CMV-eGFP expression (Fig. 3C). The fluorescence intensity was decreased for individual cells; however, the majority showed hundredfold higher expression than background (Fig. 3C). This indicates that many genes per pool are likely to be expressed in the same cell. Although most genes are not expected to contribute directly to signaling pathways that influence cell fate in vitro, these results show that the expression of a single gene in a pool is abundant in post-transfected ES cell cultures, making it possible to assay its effect on lineage choice.
Neuronal-Specific ES Reporter Cell Line
In order to more readily detect specifically differentiated cell types, we created an ES cell line containing a cell fate-specific promoter driving the expression of a fluorescent reporter. This allows for a live cell analysis and isolation of differentiated cells in a quantitative manner by flow cytometry. For assaying neuronal differentiation, we created an ES cell line expressing eGFP under the control of the early neuronal promoter Tα1-tubulin [9, 17]. Neuron-specific eGFP expression was validated by antibodies directed against another neuronal marker, βIII-tubulin, in differentiated cells (Fig. 4A). Despite the apparent heterogeneity of differentiating monolayer cultures, the reporter cell line displayed a consistent frequency of neurons following 4 days in differentiation monolayer cultures, indicating that neurogenesis occurs similarly among cultures plated at the same time. This reporter ES cell line therefore provided us a simple tool to evaluate lineage choice in differentiated living ES cells.
To test the ability of a single gene to promote lineage choice in mouse ES cells, we introduced Mash1 into the Tα1-eGFP ES cell line by transient transfection. After 4 days of monolayer differentiation, cells were trypsinized, and eGFP expression in living cells was monitored by flow cytometry. A substantial increase in neuron numbers was observed, from 1.3%–3.0% of the total cells in cultures overexpressing Mash1 compared with controls (Fig. 4B, 4C). To directly test if Mash1 was effective in larger pools, we compared the effect of Mash1 on monolayer neurogenesis at a dilution factor of 1:32 total CMV-containing plasmid DNA transfected. Under these conditions, 1:32 Mash1 approximately doubled the number of neurons in the differentiating ES cell cultures compared with the control, indicating that reducing the 1:32 gene ratio further would diminish the sensitivity of detection (Fig. 4B, 4C). Marker gene expression confirmed βIII-tubulin immunoreactivity in eGFP positively-sorted cells following flow cytometry (Fig. 4C). These results indicate that a potent differentiation-promoting effect of a single gene is sufficiently visible in a larger pool of other genes.
An Efficient Screen for Neuronal Inducers
Based on the above results, a schema was devised to test for genes that could, upon overexpression, increase the proportion of neurons after 4 days in monolayer culture (Fig. 1). Wells of E. coli containing unique IRAV cDNAs were grown individually to confluency then combined into pools of 32, and plasmid DNA was extracted. The pooled cDNA was then used to transfect low-density cultures of Tα1-eGFP ES cells in duplicate. After 4 days in minimal medium, cultures were trypsinized, and the frequency of eGFP-positive cells was analyzed by flow cytometry.
In total, 252 pools comprising approximately 6,500 unique Unigene clusters (gene size 240–6,584 base pairs) were analyzed, and eGFP-positive values were recorded using gates established in Figure 4 (supplemental online data 2). Normalization of percent eGFP-positive cells to an average of transfections performed at the same time gave an indexed value for comparison of all pools (Fig. 5A, mean set to 1). In order to measure the reliability of individual pools, a score was devised based on the standard deviation/index average (Fig. 5A, lower value is more reliable). Seven pools gave approximately 50% more neurons than the indexed average and were selected for retransfection along with 18 pools scoring highly in their transfection group (in red, Fig. 5A). A subset of eight pools with at least 40% more neurons upon retransfection compared with β-galactosidase transfected controls was chosen for reduced pool transfections (4 cDNAs per pool) and individual gene transfections into Tα1-eGFP ES cells. From these, 15 genes were chosen for single gene transfections in triplicate, resulting in 28%–97% greater eGFP-positive cells compared with β-galactosidase controls (Fig. 5B). In contrast, three genes chosen from the library at random failed to increase the percentage. Eight of the fifteen genes have been implicated in nervous system development or maintenance, suggesting they are bona fide neural effectors.
In this study, we describe a sensitive method to uncover developmentally regulated inducers of cell fate choice and therapeutically relevant genes whose transient application yields increases in specific cell types. Monolayer differentiation presents several unique advantages in discovering cDNAs that are able to influence one or more steps of differentiation, from ES to mature postmitotic cell. Under minimal differentiation conditions, monolayers are highly transfectable, differentiate reproducibly to diverse cell types, and only transiently alter cellular DNA content. Culturing ES cells at a low density during differentiation likely increases the effect of the introduced genes, as it limits instructive signals from other cells in close contact. This allows for subtle increases in eGFP positive cells to be detected from several steps in a lineage. For example, cells near the center of ES clusters may be susceptible to genes promoting neural precursors, whereas neural precursors that form early at cluster peripheries may be responsive to those directing neuronal fates. Altogether, an overexpressed gene that promotes a particular lineage is likely to result in a distinct increase over background differentiation levels. The detection method we employed, flow cytometry, sensitively detects even slight differences among samples. Experimental variability in neuron number was further limited by transfecting library pools in duplicate and individual genes in triplicate.
In this report we have used an arrayed library of full-length cDNAs and a cell fate-specific promoter for sensitive detection of neurons. However, this method is also applicable for cells derived from nonectodermal cell types. Permissive medium conditions allow for the production of complex or developmentally late cell types such as cardiomyocytes, peripheral nervous system cells, and glia (Fig. 2C). Previously, these fates have been predominantly attainable through the usage of embryoid bodies. Endoderm is rarely formed from ES cells, consistent with our culture results, but the presence of both mesodermal and ectodermal cells suggests that conditions are favorable for detecting endoderm formation, and the screen could easily be adapted for the detection of genes inducing endodermal derivatives. In order to examine several distinct fates, combining additional spectrally separable promoter-fluorescent markers to the same cell line would further allow for simultaneous transfection and analysis.
The results of our analysis implicate several genes in a neuroprotective capacity, an unsurprising discovery since apoptosis is an important regulatory mechanism of neuronal numbers during development, and minimal differentiation medium lacks important vitamins and growth factors. Whereas genes such as the fragile X mental retardation gene 1, ring finger protein 130, and nuclease sensitive element binding protein 1 (Ybx1) have established roles during neural development [18, , , –22, 27, 28], Ybx1 also plays a role in stress response , indicating both functions may contribute to its repeated discovery in high scoring pools from separate plates (data not shown). Mice mutant for heat shock factor 2 display abnormally large ventricles and reduced areas of neurogenesis [30, 31], suggesting newly born neurons may be particularly sensitive to stresses inherent to both development and in vitro culture. This hypothesis is also supported in our analysis by the identification of ferritin light chain 1, whose function has been closely tied to neuroferritinopathy in humans .
One potential caveat to this study is the potential for plasmid integration into the genome where endogenous gene expression may be affected. An analysis of the frequency of integration events per well found approximately 300–500 integrants per initial 175,000 cells transfected. Although these integration events may contribute to the total number of neurons formed after 4 days, no significant differences were found between positive and negative samples tested (data not shown). We also cannot exclude a bias toward integration in loci containing homologous sequences found in library plasmids or in loci that affect neuronal differentiation; however, the CMV-promoter is also present in the control β-galactosidase vector and would be equally subject to silencing or positional effects upon integration .
Genetic screens such as the one described here are often biased for or against a class of molecules. For transfections, larger plasmids have a decreased chance of entry into lipoplexes than smaller plasmids, making it less likely to discover those clones. Under screening conditions, an 8–10-fold difference in transfection efficiency can be expected between the largest and smallest plasmids (, data not shown). However, analysis of library plasmids revealed that 88.6% of the constructs in the library were below 7.5 kb, the transfection of which is approximately 50% less efficient than that of the smallest plasmid (data not shown). Hence, a twofold difference in transfection efficiency exists for the large majority of the genes transfected. This effect is partially mitigated by the inclusion of fewer genes per pool. Although a bias exists against large insert plasmids, one of the final 15 isolated, flk1 kinase insert domain protein receptor, is 9,860 bases, showing that, despite this bias, large genes may be isolated from the library under these conditions.
Gene overexpression studies can circumvent loss-of-function complexities arising from gene families that share common function [37, 38]. Although in some cases canceling effects of expressing multiple genes may limit detection, synergies are also possible and can be identified through a clone repooling strategy. Interestingly, many pools resulted in potent decreases in neuron number (Fig. 5A), suggesting they promote formation of other tissues or processes that antagonize neuron formation. These pools effectively double the breadth of our analysis and are available along with other untested high scoring pools to the scientific community (supplemental online data 2).
Embryonic stem cell differentiation research holds promise for cell therapies that depend on a source of large numbers of specialized cell types. The technique described here addresses two significant problems: the establishment of an efficient and transient gene delivery method to embryonic stem cells and the discovery of genes contributing to lineage choice. Although generation of large absolute numbers of a mature cell type was not a goal of this study (but rather to use conditions sensitive for the detection of neuronal inducers), rich medium conditions and gene delivery to ES cells of molecules identified here and in similar studies may significantly increase purification of specialized cell types for cell therapy.
Disclosures of Potential Conflicts of Interest
J.F. owns stock in, has acted as a consultant to, served as an officer or member of the Board for, and has a financial interest in NeuroNova.
We are grateful to V. Wirta and J. Lundeberg for procuring the cDNA library, M. Toro and K. Hamrin for help with flow cytometry, C. K. Hidaka and T. Morisaki for generously providing the Nkx2.5-GFP ES cell line, S. Nyström for generous technical assistance, and O. Hermanson and A. Simon for critical reading of the manuscript. This project was supported by grants from the Swedish Research Council, the Karolinska Institute, the Swedish Cancer Society, the Tobias Foundation, the Foundation for Strategic Research, KOSEF, and by the European Commission through the FP6 project STEMS (LHSB-CT-2006-037328).