V. EMBRYONIC STEM CELLS AS CELLULAR MODELS IN DEVELOPMENTAL BIOLOGY AND PATHOLOGY
Experiments designed to understand gene function in the context of an organism require genetic strategies. Enhancer and promoter traps (129), gene traps, random activation of gene expression (RAGE), and genome-wide cell-based knockout (GECKO) represent genome-wide strategies to identify, isolate, or determine gene function (for information on RAGE and GECKO go to http://www.athersys.com/). Because of gene-targeting techniques, transgenic mice have also proven critical to the creation and evaluation of some models of human disease.
A. Gene Trapping
Gene trapping is the most commonly employed insertional mutagenesis strategy, and it has been extensively reviewed elsewhere (102, 140, 264); however, this technique is likely to prove very important for the study of human development, i.e., through the exploitation of hES cells in vitro. Essentially, when gene traps are introduced into ES cells, they integrate randomly in the genome (102, 331, 332, 376). Antibiotic-resistant ES cell colonies are easily selected and expanded in vitro, and clonal cells can be isolated for injection into mouse blastocysts or differentiation in vitro. Expression of the gene trap is assayed for reporter gene expression (e.g., -galactosidase activity), and staining is indicative of an insertion event. The transgene is only activated when it integrates correctly within an active transcriptional unit; however, some translational fusions (frame shifts) inactivate the reporter activity or may target the translated proteins into subcellular locations where reporter activity is not easily detectable. Gene trapping therefore selects for integration events in functional genes, and it is especially useful for the analysis of mammalian cells that have complex genomic organizations that consist of promoters and exons that are separated by introns (Fig. 7).
In vivo gene trap screens in mice have permitted the identification of many developmentally regulated genes that are expressed within specific tissues in a spatiotemporal pattern, including novel RA responsive (120), neuronal, glial, chondrocytic, myocytic (23), and hematopoietic (65, 157, 342) genes. A disadvantage of this approach is that it requires the production of a large number of mice from ES cell clones to identify a limited number of developmental genes. To limit the number of noninformative mice, Bonaldo et al. (45) employed gene trapping together with the differentiation potential of ES cells. By selecting for the activation of the reporter gene in tissue culture, the rate of gene disruption in recovered clones approaches 100%, and the random insertion of exogenous DNA into single sites in the mammalian genome (gene trapping) provides a genome-wide strategy for functional genomics. ES cell cultures thus provide a simple model system for studying the genetic pathways that regulate embryonic tissue development and permit high-throughput screening of clones for tissue-restricted gene trap expression (45).
In the postgenomic era, bioinformatic-based approaches have accelerated the evaluation of mutant clones (originating from gene traps, RAGE, and GECKO) leading to the rapid identification of informative cell lines on an unprecedented scale. When combined with computational approaches, expression profiling with DNA microarrays, and in situ hybridization analyses, the results often suggest an association with a specific biological process or disease state, which must be tested. For example, a sequence-tagged gene-trap library of >270,000 mouse ES cell clones has recently been developed that has been employed together with a functional screen of knock-out mice to identify genes regulating blood pressure (412). Efforts are also underway to make ES cell lines with gene traps freely available for researchers so that transgenic mice containing a potential gene of interest can be made to further understand the role of specific genes in development and disease (e.g., http://baygenomics.ucsf.edu/overview/welcome.html).
Finally, a new reporter system has been described that permits real-time monitoring of live cells. In this system, a -lactamase tagged library can be used to clone genes (387). Use of the nontoxic fluorescent substrate of -lactamase, CCF2-AM, enables real-time and sensitive monitoring of transcription in live cells (417). In theory, the monitoring of live cells can be used to identify cell clones with genes that are induced or repressed by different agents, including receptor ligands, drug candidates, or viruses (286).
B. In Vitro Models to Study Embryonic Lethality
As pointed out earlier, genetic modifications of ES cells can lead to embryonic lethality, certain aspects of which can be overcome through the use of conditional targeting. Alternatively, embryonic lethal models can also be studied in vitro with ES cells containing targeted mutations on chromosome pairs. In the case of X-linked genes like HPRT or GATA-1, a functional knockout of a gene in ES cells can be generated from a single targeting event (36). In the majority of cases, however, a knock-out ES cell line must be generated either by sequential targeting of chromosomal pairs in vitro or through an intermediate step involving the generation of homozygous mice lacking a functional allele. The generation (either from knock-out mice or by sequential targeting of chromosomal pairs) and analysis of knockout ES cell lines can be labor intensive and is neither practical nor useful for many studies. At times, the use of targeted chromosomal pairs in ES cells has, however, proven indispensable to the elucidation of gene function. Mitsui et al. (233), for example, targeted Nanog on chromosomal pairs to show that it was required for the maintenance of embryonic stem cell pluripotency and identity (see sect. II). When knockout cells have been coupled with random transgenesis, it has also been possible to rescue phenotypes (36), determine how the timing or duration of signaling determines cell fate (262), and develop new developmental paradigms (35, 36).
Targeted chromosomal pairs coupled with in vitro differentiation have also been used to elucidate the underlying mechanisms of embryonic lethality in mice. For example, the ryanodine receptor (RyR2), which serves as the major sarcoplasmic reticulum calcium release channel in heart to mediate a rapid rise of cytosolic free Ca2+, is normally expressed early in developing myocardium. A functional knockout of this gene causes the developing mice to die at approximately E10 day post coitum, but the mechanism responsible for this embryonic lethality was unclear (355). Examination of cardiomyocytes derived from RyR2-deficient ES cells showed that RyR2 was essential to increase the spontaneous beating rate in immature cardiomyocytes (408) (Fig. 5). When the heart rate slowed in the transgenic mouse model due to the absence of RyR2, the sphincter mechanism normally utilized in valveless embryonic heart was lost, and blood perfusion proved inadequate. Embryonic lethality in the RyR2–/– mice was therefore postulated to be due to functional incompetence of the contracting myocardium, a finding that was achieved only through complementary studies between in vivo and in vitro systems. Similarly, the use of ES cells homozygously deficient for 1-integrin (see Refs. 113, 145, 307) and desmin (385) genes allowed a detailed loss-of-function analysis and description of affected cell types in vitro, because animals deficient for these genes died early during embryogenesis.
C. Developmental and Disease Models
As described in section IIIA, the production of heterogeneous cell populations in vitro has constrained the analysis of ES cell-derived progeny. The use of transgenesis and gene targeting has overcome many of these limitations, and relatively pure homogeneous populations of ES cell-derived progeny have now been isolated. Genetic approaches involving transgenic mice have also greatly advanced our knowledge of development and disease. This has been accomplished primarily through 1) the isolation and cultivation of ES cells, which retain the ability to colonize all tissues of a host embryo including its germ line; 2) the resiliency of mammalian embryos/blastocysts to tolerate the addition or loss of embryonic cells; and 3) gene inactivation by homologous recombination or overexpression of transgenes to assess gene function and genetic labeling of precursor cells to determine cell lineages.
The earliest use of targeted animal models for gene therapy emphasized mouse models that simulated inherited disease, but these often proved disappointing (393). Subsequent studies have identified many useful mouse models for the study of human disease; however, the utility of these transgenic models frequently depends on the impact of environment and genetic background. A good example is seen with mouse models of cystic fibrosis (CF), where the CF transmembrane conductance regulator (CTFR) gene was interrupted or mutated. The initial CFTR-deficient mice did not develop pulmonary pathologies before death; however, subsequent genetic and environmental modifications have increased its usefulness to model CF (393).
Currently over 1,200 papers with transgenic mice can be found on-line (PubMed), and some 7,000 mutant mice have been described. While not all of these models have proven useful, some have been critical for determining promoter and gene function, functional gene redundancy, spatial distributions of expression, and lineage tracing. Numerous papers have also documented the use of gene targeting for mouse models of development and disease. Some mouse and ES cell studies have shed light on factors indispensable for hematopoiesis (256), while a number of knockout mouse models have been employed to reveal the critical roles for transcription factors (Ets family members) in guiding hematopoiesis, vasculo/angiogenesis, and other cellular differentiation processes (27). Many reviews have already been published showing how gene targeting has been employed to study cardiovascular (77), pancreatic (153), or renal (172) systems, while still others have employed Cre/loxP systems for conditional regulation (313).
Animal models of human diseases are critical to the early development and evaluation of gene- and cell-based therapies; therefore, studies with mouse ES cells in the context of transgenic models form the foundation for current and future work with human ES cells and their derivatives for studies in human. Experimentally, it remains unclear whether human ES cells will be as versatile as mouse ES cells with respect to self-renewal, genetic manipulation, or developmental capacity, but the ability to test these cells in disease models, transgenic or otherwise, will be critical to this evaluation.
D. Recent Advances
1. Extrachromosomal expression
As stated earlier, integration-dependent events can adversely affect gene expression in ES cells. Similar to retroviral sequences, transgenes randomly introduced into ES cell lines tend to be progressively silenced, resulting in mosaic expression, heterogeneous phenotypes, or complete silencing. Extrachromosomal plasmid replication avoids the problem of gene silencing and represents a powerful technique to overexpress genes without disrupting the pluripotentiality or differentiation capacity of ES cells.
One system of extrachromosomal replication exploits the replicative biology of polyoma virus (63, 134). Polyoma virus mutants, which either lack the entire large T intron or lack the splice sites employed to form properly processed middle T and small T mRNAs, are unable to transform cells. As long as large T transcripts are present, mutant polyoma virus DNA can replicate as free, unintegrated mini-chromosomes in infected mouse cells (189), and ES cells that stably express the polyoma large T protein efficiently support the episomal maintenance of plasmids containing the polyoma origin of replication. Gassmann et al. (134) developed a self-replicating vector system (pMGD20neo) for ES cells that contains the polyoma origin of replication with a mutated enhancer, a modified polyoma early region that encodes the large T antigen, and a neomycin resistance cassette (134). The utility of this system was recently demonstrated by Aubert et al. (14), who employed a variant of this extrachromosomal replication system to uniformly express a secreted frizzled related protein (SFRP2) transgene in ES cells (see sect. IVA and Ref. 14). They showed in puromycin-resistant cells that both undifferentiated ES cells and their descendents express transgenes more uniformly and stably than that normally achieved with transgenes inserted randomly into the genome. Stable transfectants were established at a frequency of 1–5% compared with integration events. Importantly, expression of polyoma large T protein at levels sufficient to support episomal replication appears to have no effect on ES cell self-renewal or pluripotency. The use of extrachromosomal vectors thus overcomes some of the major technical problems associated with random integration events: silencing, mosaicism, and/or interference with endogenous genes.
Both random transgenesis and homologous recombination have been limited by the time and site limitations associated with DNA engineering in Escherichia coli, particularly if conditional knockout models are being developed. The construction of targeting vectors often utilizes large regions of genomic DNA, and their construction can be labor intensive and complex. For example, restriction enzyme sites are not always conveniently located, and mutations must be made in the genomic sequences to introduce selection cassettes or loxP sites. Recent innovations use homologous recombination to construct targeting vectors in a process termed recombineering (85, 242). This form of chromosome engineering greatly shortens the time it takes to make a targeting vector and makes it possible to introduce selectable markers anywhere in a gene. An example of the approach was the generation of knock-in constructs for Olig-2, a transcription factor first expressed in ventral progenitor cells that gives rise to oligodendrocytes and motor neurons, but in spinal cord is only present in oligodendrocytes (400). With the use of a mouse genomic BAC library, the Olig-2 gene was isolated and a targeting construct was generated by homologous recombination in yeast. Following recombination, the construct was shuttled back into E. coli, modified, and introduced into ES cells. G418-resistant colonies were selected and differentiated in vitro. GFP-positive cells were found to be consistent with cells of the oligodendrocyte lineage that could be separated by fluorescence-activated cell sorting and cultured as pure populations. Although originally pioneered in yeast, recombineering explicitly refers to the use of homologous recombination in E. coli to manipulate genomic sequences. Specifically homologous recombination in E. coli is facilitated by the use of bacteriophage-based homologous recombination systems, which permit linear double-stranded DNA fragments (i.e., those carrying loxP sites and selection markers) that have short regions of homology with the target sequences at their ends to be inserted into virtually any large target DNA (plasmids, BACS, or PACs). The utility of this system in ES cells was recently described by Testa et al. (359), but it has yet to be shown applicable for use with hES cells.
3. RNA interference
RNA interference (RNAi) is a process whereby double stranded (ds) RNA induces targeted degradation of RNA molecules with homologous sequences. It has become a valuable tool for the analysis of gene function through suppresssion of specific gene products, and it has been extensively employed in Caenorhabditis elegans and plants (117). More recently, RNAi has proven useful in the study of mammalian systems (21). The major obstacle for the use of short interfering (si) RNA has been in the efficient and sustained delivery of dsRNA to mammalian cells; however, when introduced into these cells, the hallmark of RNAi is its specificity; dsRNA triggers specific degradation of homologous mRNA only within the region of identity of the dsRNA (413).
Sequence-specific RNAi has been demonstrated in the preimplantation mouse embryo and in oocytes by direct injection of dsRNA (354, 388). When introduced into mouse zygotes, dsRNAi proved effective at repressing GFP throughout the blastocyst stage up until E6.5. Recent results demonstrate that ES cells maintained in an undifferentiated or in a differentiated state can also respond to dsRNA for gene silencing (409, 418). In the latter case, the authors employed dsRNA to suppress the expression of PU.1 and C/EBP in CD34+ EB cells. As a consequence, the level of expression of the M-CSF receptor (CD115), a downstream target of PU.1, and C/EBP were both decreased within 2–3 days after transient transfection. With the success of this approach to knock-down genes in ES cells and recent improvements to the delivery of siRNAs to mammalian cells, RNAi may be an effective approach to the study of ES cell differentiation and as a gene therapy approach (68).