The arrival of iPS cells.
In the first report of defined factor reprogramming (10
), Kazutoshi Takahashi and Shinya Yamanaka reprogrammed mouse fibroblasts through retroviral transduction with 24 transcription factors highly expressed in ES cells. This cadre of genes was gradually reduced to four that encode the transcription factors octamer 3/4 (Oct4), SRY box–containing gene 2 (Sox2), Kruppel-like factor 4 (Klf4), and c-Myc (10
). The resulting iPS cells were selected based on their ability to express the gene F-box protein 15 (Fbx15
), which is specifically expressed in mouse ES cells and early embryos. Although the selected cells were similar to ES cells in morphology, growth properties, and ability to form teratomas (neoplasmic tumors characterized by the presence of cells corresponding to all three embryonic germ layers) in immunodeficient mice, they differed in terms of global gene expression profiles and certain DNA methylation patterns. Perhaps the most important difference was that these cells failed to produce adult chimeric mice. This might be because although Fbx15
is specifically expressed in mouse ES cells and embryos, it
is dispensable for maintaining pluripotency and mouse development (11
). In subsequent studies (12
), when improved end points for the reprogramming process were selected, such as the expression of Nanog
, the resulting iPS cells were even more similar to ES cells, could contribute to adult chimeras, and were transmitted through the germ line.
Shortly after these reprogramming successes in the mouse, Yamanaka used the human orthologs of the four transcription factor–encoding genes to generate iPS cells from human fibroblasts (2
). Concurrently, two other groups achieved similar reprogramming of human somatic cells using slightly different combinations of genes that also included OCT4
(see Table for details) (16
). Within months, it had been proven that it was possible to derive iPS cells from patients suffering from the neurodegenerative disease amyotrophic lateral sclerosis (ALS) (18
) as well as patients with other diseases, including juvenile-onset type 1 diabetes mellitus, Parkinson disease (PD) (19
), and spinal muscular atrophy (SMA) (20
Mouse and human iPS cells have been generated in a variety of ways
Mechanism of reprogramming.
Given that all cells within an organism have the same genome, the functional characteristics of different cell types are defined by specific patterns of gene expression. Epigenetic molecular mechanisms control gene transcription by inducing stable changes in gene expression. These changes favor the formation of either an accessible or inaccessible chromatin state without directly affecting the DNA sequence (21
Developmental programming establishes gene expression patterns that are set and maintained via histone modifications and DNA methylation (22
). This is a one-way process (reversed only in germ cells) that gradually leads to somatic cell types with reduced pluripotency (for more detailed reviews, see refs. 22
) (Figure ). Transcription factors are the key regulators of this process (23
), and there can be no better evidence for their imperial role in determining the functional characteristics of a cell than their ability to reprogram differentiated adult cells to a pluripotent state (2
). Different combinations of just six transcription factor–encoding genes, Oct4
, and Lin28
, have been used to achieve reprogramming in mouse and human cells (2
). Of these, Oct4
), and Nanog
) were known to encode proteins functionally important in maintaining pluripotency in embryos and ES cells; the rest encode proteins with functions in a wider variety of cell types. Determining the exact functional role of each of the proteins encoded by the six genes in the reprogramming process is an active area of investigation, although a recent genome-wide analysis of promoter occupancy at various stages of iPS cell generation has shown that they co-occupy regulatory regions of a number of genes associated with pluripotency (28
The concepts of developmental programming and reprogramming.
Initial protocols for making iPS cells utilized viral vectors to exogenously express the four transgenes, which integrated in the genome of somatic cells. However, it is now possible to induce iPS cells with non-integrating methods in mouse (29
) and human cells (32
), and although these methods are not yet efficient, they have clearly demonstrated that random insertional mutagenesis is not an essential mechanistic requirement for reprogramming. Furthermore, successful reprogramming of terminally differentiated postmitotic cells such as B lymphocytes (34
) has definitively shown that reprogramming does not require multipotent target cells that may exist within somatic cell populations.
ES versus iPS cells.
A number of studies have clearly demonstrated that iPS cells are highly similar to ES cells (12
). Most of the iPS cell lines that have been generated have a normal karyotype and possess telomeres with characteristics of those in ES cells (38
). In terms of the epigenetic state of iPS cell lines, the promoters of pluripotency genes have no DNA methylation, while female mouse iPS cells reactivate the inactive X chromosome and exhibit random inactivation upon differentiation (12
). Furthermore, the promoters occupied by the reprogramming proteins in iPS cells are highly congruous to those occupied by the endogenous proteins in ES cells, and this correlates with closely related global gene expression profiles (28
However, small differences in gene expression patterns between iPS and ES cells exist (39
). The nature of these differences is not yet well understood, although genomic integration of transgenes has been shown to be a contributing factor. Soldner et al. recently partially addressed the issue of integration by demonstrating that upon excision of the reprogramming transgenes (but not the viral long-terminal repeats [LTRs]), the genome-wide expression profile of iPS cells generated from human fibroblasts correlated more closely with that of ES cells than it did prior to excision: only 48 genes remained substantially different in their expression in contrast to 271 before excision (36
). Similarly, when Yu et al. generated human iPS cells without any detectable genomic integration, only a small set of genes remained differentially expressed when compared with human ES cells (33
The reason for the remaining differences in the global transcriptome between ES and iPS cells remains unclear, although the unique genetic makeup of each line may contribute to a certain degree, as has been shown for human lines (40
) and mouse ES cell lines (41
). It is noteworthy that a recent report indicates that iPS cells (generated by retroviral genomic integration) exhibit a unique gene and miRNA expression profile that gradually becomes more similar to that of ES cells with extended in vitro culturing (39
). Moreover, this iPS cell–specific expression signature is at least partially conserved among iPS cells generated in independent reprogramming experiments and even in integration-free iPS cell lines (39
). These issues can be best resolved by comparing the expression profiles of an ES line and an iPS cell line of the same genetic background.
While the value of comparing iPS and ES cell lines is unquestionable, the functional ramifications of small differences between iPS and ES cells in terms of global gene expression levels and epigenetic state remain unclear. Moreover, it remains unknown whether the observed differences will result in functional variation that interferes with the utility of certain iPS cell lines.
The ability of mouse iPS cells to generate an entire mouse, as was recently shown via tetraploid complementation assays (a technique in which iPS cells are injected into tetraploid blastocysts) (42
), and of human iPS cells to form teratomas in vivo (19
) indicates in the most stringent tests that they are pluripotent cells and suggests that the defined factor reprogramming approach produces cells with a developmental potential similar to that of ES cells. To our knowledge, no conspicuous differences in the efficiency of differentiation between iPS and ES cells toward specific lineages have been reported. Although these experiments are difficult to perform due to the heterogeneity of the resulting cell populations, they will be a key indicator of whether iPS cells can be readily used for a variety of practical applications, including disease modeling.
What is the best way of making iPS cells?
First-generation iPS cells were generated by retroviral transduction (2
). Since then, the technique has been optimized and reproduced in a number of different ways (Table and references therein). The most important variables include choice of cell type to reprogram, choice of the cocktail of reprogramming genes, and method for gene transfer (Figure ). Nimet Maherali and Konrad Hochedlinger recently wrote an excellent review of protocols, highlighting the details of different methodologies to make iPS cells, and we therefore cover this area only briefly (45
Embryonic fibroblasts (MEFs) and tail-tip fibroblasts (TTFs) in the mouse and dermal fibroblasts in the human have been the most widely used cell types to reprogram, largely due to their availability and ease of accessibility. In addition, various other cell types have also been reprogrammed, including hepatocytes (29
), stomach cells (46
), B lymphocytes (34
), pancreatic β cells (47
), and neural stem cells (48
) in the mouse; keratinocytes (49
), mesenchymal cells (19
), peripheral blood cells (50
), and adipose stem cells (51
) in the human; and melanocytes in both species (52
). Variable efficiencies and kinetics of the process have been described, while the in vitro age of the cell type (passage number) correlates inversely with the efficiency of reprogramming (53
). The same inverse correlation has been effectively demonstrated for the differentiation stage of target cells, with mouse hematopoietic stem and progenitor cells being more efficiently reprogrammed than terminally differentiated B and T lymphocytes (54
A recent study addressed the variability in the teratoma-forming propensity of cells differentiated from iPS cell lines generated from a number of mouse tissues (55
). Miura et al. found that neurospheres (nonadherent spherical clusters of neural stem and progenitor cells) derived from iPS cells generated from MEFs formed teratomas with efficiency similar to that of neurospheres derived from ES cells, while neurospheres derived from iPS cells generated from either TTFs or hepatocytes formed teratomas more readily. These results were associated with the presence of greater numbers of residual undifferentiated iPS cells in neurospheres derived from iPS cells generated from either TTFs or hepatocytes, and they can be used to indirectly assess the variation in differentiation efficiency that might arise as a result of the target cell that is reprogrammed.
Of the original four transcription factor–encoding genes, c-Myc has been shown to be dispensable for reprogramming in the mouse and human, while Klf4 and Sox2 have been shown to be dispensable in reprogramming strategies utilizing cell types that endogenously express them (see Table ). Oct4 is the only factor that cannot be replaced by other family members and the only one that has been required in every reprogramming strategy in either mouse or human cells.
Importantly, small molecules, including DNA methyltransferase inhibitors (35
), the histone deacetylase inhibitor valproic acid (VPA) (56
), and the histone methyltransferase inhibitor BIX-01294 (58
), have been shown to substantially improve the efficiency of reprogramming, even in cases without inclusion of exogenous Klf4
(for VPA) or exogenous Sox2
(for BIX-01294). In addition, the compound kenpaullone can compensate for the reprogramming factor
a combination of BIX-02194 and BayK8644 can compensate for Sox2
), even in cells that do not endogenously express it, although the mechanism(s) by which these small molecules function remain unclear. We have also recently identified a small molecule inhibitor of TGF-β signaling that replaces Sox2
by inducing Nanog
). Of note, it remains to be determined whether full reprogramming through the exclusive use of chemicals is possible.
iPS cells have been generated using a number of different gene transfer methods, including retro-, lenti-, and adenoviral vectors and nonviral plasmids, and recently by direct recombinant protein delivery. The use of genome-integrating viral vectors such as retroviruses and lentiviruses results in iPS cells that would be inappropriate for therapeutic use, as even a single insertional mutation (62
) and the potential for reactivation of viral transgenes substantially increase the risk that transplanted cells would become transformed (14
). The major advantage of retroviral vectors is that they are known to undergo progressive silencing in ES cells, while lentiviral systems remain active. However, in certain cases, probably due to the site of genomic integration, retroviral vector expression is maintained (18
). Recent studies have reported on the generation of iPS cells using genomic integration systems such as lentiviruses (36
), plasmids (66
), and transposons (67
), all of which allow for subsequent transgene removal through the Cre-lox system or transposases.
More importantly, iPS cell generation has now been achieved without genomic integration. This has been done using adenoviruses (29
), repeated plasmid transfection (30
), and recombinant proteins (31
) in the mouse and via the use of episomal vectors (33
) and recombinant proteins (32
) in the human. If the ultimate use of the iPS cells requires cells free of genomic integrations, then these methods are likely to be preferred over the classic retroviral and lentiviral transduction systems.
Much of the focus of recent research has understandably been on the generation of clinically applicable iPS cells free of viruses and transgenic integrations. We believe that it is now critically important that iPS cells generated by distinct methods are carefully assessed for their variability, stability, and differentiation potential as well as the quality and long-term survival of differentiated cells derived from them. Ultimately, iPS cells generated by each method will need to be examined in detail at the genomic, epigenomic, and functional level in order to determine which reprogramming methods are safe for clinical cell therapy.
Although iPS cells generated using one or two factors rather than four or using recombinant proteins rather than viral expression systems may be clinically safer, it has yet to be demonstrated that this is the case. In addition, the logistical, financial, and practical aspects of each technique will need to be taken into account. It is likely that how one reprograms cells will be determined by the intended usage of the reprogrammed cells. As things currently stand, for example, the fastest and most efficient method to create human iPS cell lines is through lenti- and retroviral transduction. If it is shown that transgene integration sites do not substantially affect the differentiation potential and status of cells types relevant for disease, then these methods might be the preferred approaches for generating iPS cells for use in large-scale drug-screening programs and disease modeling.
An important and controversial issue regarding iPS cell derivation is the standard by which their pluripotent potential is evaluated (45
). We believe that the intended application of the iPS cells should determine the evaluation method. It may be that a neurobiologist who wants to generate motor neurons from patient-specific iPS (PS-iPS) cells should be concerned more with determining the ability of the iPS cells to generate cells that are functionally equivalent to motor neurons found in vivo than whether they are truly pluripotent. The issue of pluripotency becomes more important for studies of the functional mechanism of reprogramming. However, a reproducible, inexpensive, and rapid method to determine the quality of newly established iPS cell lines is urgently required. If it cannot be found, the cost of fully characterizing every new cell line may substantially slow progress. Perhaps an epigenetic and gene expression signature that selectively defines fully reprogrammed iPS and ES cell lines can be identified. This then might enable high-throughput screening of newly derived cell lines.