III. GENETIC MANIPULATION OF EMBRYONIC STEM CELLS
Cell biology-based techniques have proven critical to the early isolation of ES cells and the subsequent delineation of differentiation protocols (see sect. IV). Except for neurogenesis, in vitro differentiation has required an initial aggregation step with formation of EBs before specialized cell types form in vitro. Two impediments initially prevented the full potential of the in vitro ES cell model from being realized. 1) We knew relatively little about differentiation pathways in culture and how these pathways compared with those in the developing embryo, and 2) differentiation protocols resulted in the simultaneous production of heterogeneous cell populations, thus constraining studies on selected subsets of cells. To overcome these limitations, genetic tools have proven indispensable to the study of ES cells and their progeny, both in vitro and in vivo. The capacity of ES cells to be clonally expanded permits the identification of independent and stable integration events (301), and a number of technologies have been developed to rapidly generate stably transfected ES cell clones and transgenic mouse models.
DNA can be introduced into ES cells by conventional infection, transfection, or electroporation protocols (66, 67). Random insertion events have been employed to overexpress, mutate, and tag genes in phenotype-driven screens, and the discovery that DNA (cloned or genomic) introduced into ES cell lines can undergo homologous recombination at specific chromosomal loci has revolutionized our ability to study gene function. The ability to introduce virtually any mutation into the genome following gene targeting in mouse ES cells provides a powerful approach for elucidating gene function both in vitro and in the whole animal. ES cell progeny can therefore be biased into a desired cell lineage by exposure to appropriate differentiation factors and by genetic manipulations of key developmental genes. Recent advances have shown that hES cells are also amenable to genetic manipulation, thus opening the door to genetic analysis of human development and disease in vitro (104, 202, 419).
A. Random Transgenesis
Random transgenesis results in the indiscriminate incorporation of DNA within the genome. The use of sequences that confer antibiotic resistance (e.g., neomycin, puromycin, hygromycin, and herpes simplex virus thymidine kinase) for clonal selection or of reporter genes [e.g., green fluorescent protein (GFP/EGFP), LacZ (–galactosidase)] to identify specific cell lineages has greatly facilitated this approach both in vitro and in vivo (140). Additional constructs have been designed to overexpress transcription factors (e.g., GATA4, Twist), signaling molecules (e.g., insulin-like growth factor II, Cripto), or cellular proteins in differentiated phenotypes of myogenic (95, 278, 308), erythroid (150), pancreatic (38), and cardiomyocytic (262) cell lineages. Promoters of either viral or mammalian origin have, however, often proven inconsistent in the formation of stably expressing ES cell clones.
Retroviral vectors have been used for the delivery of genetic material into cells for over 20 years. The advantage of a retroviral system is that genetic sequences can easily, efficiently, and permanently be introduced into target cells. In fact, the first successful reports of genetic manipulation of ES cells involved retroviral vectors. These early experiments demonstrated that integrated viruses (provirus) could be transmitted through the germ line (300, 348); however, sustained transgene expression from integrated proviruses proved difficult to achieve. ES cells have high de novo cytosine methylation at CpG dinucleotides, which effectively represses gene expression regulated from viral long-terminal repeats (LTRs) (28, 171, 348). In addition, provirus gene silencing is mediated by trans-acting factors that bind to the LTRs of some viral promoters (76, 260). The lack of significant provirus transcription in ES cells and ES cell progeny have effectively limited the use of simple retroviral vectors in experiments of random transgenesis (300).
The development of more complex lentiviral vectors, based on the human, feline, equine, or simian immunodeficiency viruses (246, 255, 274, 317), offer several advantages over other retroviruses (for review of vectors, see Ref. 282). Lentiviruses infect both dividing and nondividing cells, and transgene expression is not silenced in ES cells. Pfeifer et al. (271) furthermore demonstrated that lentiviral vectors could efficiently transduce human ES cells, and subsequent analyses have shown that lentivirus infections are highly effective for the delivery of functional transgenes into human ES cells (143, 214). Importantly, transgene expression is not "shut off" during differentiation in vitro (EBs) or in vivo (teratomas), and functional transgenes can be successfully passed through the germ line without loss of expression (271). These proof-of-principle experiments, with reporter constructs, demonstrate that lentiviruses are capable of foreign gene transfer to hES cells. This is particularly important, because electroporation, which has served as the main method for the introduction of foreign DNA into murine ES cells (331, 360), adversely affects the survival of hES cells (104). Lipofection-based transfection techniques, similarly, show transfer efficiency rates in hES cells that are generally 104). Lentiviral delivery of foreign DNA to hES cells therefore has significant relevance for the isolation of stably transfected hES cell clones and for the future development of gene- and cell-based therapies.
Random integration of DNA plasmid constructs containing tissue-restricted promoters has been used extensively to purify or mark cells, including neurons (210), pancreatic -cells (338), cardiomyocytes (192), and endothelial cells (220, 281); however, data from these studies should be interpreted with care. In vitro expression is not always consistent with in vivo analyses. For example, vimentin, which is usually restricted to mesenchymal cells in vivo (84, 125), is expressed in most cell types in vitro (126). The myosin light chain 2v (Mlc2v) promoter has also been used to identify ventricular chamber myocytes derived from differentiating ES cells in vitro (230), but this "specific" expression is only apt for adult rodent heart. During development, this gene is expressed in the anterior (atrial and atrio-ventricular) portions of the heart tube, and at later stages, in the caval myocardium (81, 123, 124). Since ES cell-derived cardiomyocytes are not typical of adult myocardium, the Mlc2v promoter probably cannot be used to identify purely ventricular myocytes. It is therefore essential that in vitro results be analyzed in conjunction with developmental models before deciding which ES cell progeny are most useful for cellular therapeutics. Finally, integration-dependent events can adversely affect gene expression in ES cells. As with pronuclear injection, the location of integration and the number of copies of integrated DNA can affect transgene expression. In particular, transgenes randomly introduced into ES cell lines tend to be progressively silenced, resulting in mosaic expression, heterogeneous phenotypes, or complete silencing. These limitations have restricted the use of random transgenesis in functional studies of ES cells and their progeny.
B. Gene Targeting
Targeting approaches that selectively modify endogenous genes have generally proven more powerful than random transgenesis in generating mutations in endogenous mouse genes. In 1987, Thomas and Capecchi (360) first showed that transfected DNA could integrate into the mES cell genome via homologous recombination. In 1989, the first report of germ-line transmission of a targeted allele was published (361), demonstrating that genetically modified ES cells could contribute in the developing mouse embryo to produce viable chimeras. Today the production of germ-line chimeras is a standard procedure for many laboratories, and the topic has been extensively reviewed in the literature (47, 179).
The ability to produce mice that carry altered genomic DNA has greatly facilitated the study of many biological processes; however, not all biological processes can be studied by gene inactivation. Gene-targeting that results in developmental arrest or embryonic lethality in vivo reflects the earliest nonredundant role of a gene and precludes analysis of function at later stages. Additionally, some genes have functions during embryogenesis that may differ from those in the adult [e.g., LIF (18, 19) and vimentin (84)]. Inactivation of these genes may lead to adaptations that preclude their functional analysis at later stages. To address these problems, a number of modifications to the original gene-targeting strategies have been developed.
Embryonic lethality can be overcome by generating conditional knock-out or knock-in ES cells and mice, which can be used to activate or inactivate a gene both spatially and temporally (243). Typically, a conditionally targeted allele is made by inserting loxP or frt sites into two introns or at the opposite ends of a gene. Expression of P1 bacteriophage-derived Cre or yeast-derived Flp recombinases in mice carrying the conditional allele catalyzes recombination (insertions, deletions, inversions, duplications) between the loxP/frt sites, respectively, to inactivate (or activate) the gene (209). By expressing Cre recombinase from an endogenous or tissue-specific promoter, the conditional allele can be recombined in a restricted lineage or cell type. The timing of recombinase expression can also be controlled using inducible expression systems (313) or viral delivery systems such as adenovirus or lentivirus (270, 328), which makes it possible to inactivate a gene in a temporal-specific fashion. This technique has been widely used in the analysis of mice, and its use in adult mice overcomes a major limitation associated with standard transgenics, i.e., the developmental consequences of inactivated genes (209). The system has also been adapted for ES cell lines, both for in vitro studies and the generation of new mouse models [e.g., allele replacement by double loxP recombination (2, 395); Fig. 5]. The use of site-specific recombination events (insertions, deletions, inversions, or duplications) can also be extended to the engineering of long-range modifications in the ES cell genome (416).