First Published in R&D Systems' 2009 Catalog
The ability of egg cytoplasm to reprogram somatic nuclei has been known since the 1950s due to the experiments of John Gurdon and others.1 These experiments were performed to address the question of whether or not the genomes of differentiated cells undergo irreversible changes and could no longer support early stages of development. Gurdon showed that they had not, and that nuclei of differentiated cells from tadpoles could direct development of sexually mature adult frogs when transplanted into oocyte cytoplasm.2 Despite occurring 50 years ago, in the pre-recombinant DNA era, these early nuclear transfer, or cloning experiments, prompted speculations in the press about the possibility of cloning humans.1
Around the same time, McCullough and Till demonstrated the ability of a single, bone marrow derived cell to give rise to multiple hematopoietic cell types,3 a key discovery in the stem cell field. Twenty years later, embryonic stem (ES) cells, capable of giving rise to all tissue types of a mature mouse, were discovered.4, 5 These two fields (nuclear transfer and stem cells), each fascinating in its own right, collided in a truly remarkable way in 2006. Several key developments paved the way for this event. Nuclear transfer into mouse oocytes was shown to reprogram somatic nuclei to a pluripotent state, just as it did in amphibians.6 Tada et al. showed that not just oocyte cytoplasm, but ES cell cytoplasm could also reprogram a somatic nucleus in the context of cell fusion.7 Research in developmental biology then identified a set of genes with expression specifically in ES cells and other pluripotent cell types, genes that could thus be considered candidates for determinants of the pluripotent state.8
View Larger Image
|Figure 1. Nuclear Reprogramming
A. In the 1950s, developmental biologists were undertaking experiments to test whether the process of cellular differentiation involved permanent changes at the DNA level. If so, they reasoned, the nucleus from a differentiated cell type should not be able to direct early stages of embryonic development, even when placed in an early developmental context such as an egg. These experiments are also called somatic cell nuclear transfer (SCNT) or cloning. In fact, the nucleus of an intestinal cell from a tadpole, when inserted into an enucleated frog egg, was able to support development all the way to a mature frog. The process by which factors present in egg cytoplasm change the patterns of gene expression from those of a differentiated cell nucleus to those of an undifferentiated cell nucleus is called reprogramming. The changes involved are epigenetic, including changes in DNA methylation, histone modification, and chromatin structure.
B. Subsequently, cloned mice were produced by transfer of the nucleus of an adult cumulus cell into an enucleated mouse oocyte. As in the frog experiments, the cumulus cell nucleus was reprogrammed to a pluripotent state capable of supporting complete development.
C. Importantly, the capacity to reprogram somatic cell nuclei is not limited to egg cytoplasm, as demonstrated by experiments in which adult mouse thymocytes were fused with embryonic stem cells. Although reprogramming was not complete, an Oct-3/4 green fluorescent protein (GFP) transgene present in the somatic cell nucleus was expressed in the hybrid cells but not in unfused, parental thymocytes.
In 2006, Takahashi and Yamanaka reasoned that ES cell-specific gene products might be able to substitute for ES cell cytoplasm in reprogramming a somatic nucleus back to a pluripotent state.To test this hypothesis, they used retroviral expression constructs to overexpress 24 candidate genes in somatic cells.9 To ensure that rare reprogramming events could be detected, they developed an assay system in which activation of the ES cell/early embryo specific Fbx15 gene resulted in drug resistance. In the starting somatic cell population (mouse embryonic fibroblasts, MEFs), drug selection killed all the cells, since the Fbx15 gene was not expressed. They next transduced the MEFs with each of the 24 candidate gene constructs individually. Again, no drug resistant cells were obtained. However, when all 24 candidates were transduced together, drug resistant colonies could be isolated. Some of these colonies exhi-bited morphology and proliferation characteristics similar to those of ES cells. Further analysis demonstrated that they lost expression of fibroblast specific genes and gained expression of ES cell marker genes including Oct-3/4, Nanog, Cripto, Dax1, and FGF-4. By withdrawing each of the 24 factors individually, they identified ten that were essential for colony formation. A second round of withdrawal of single factors from the pool of ten narrowed the list of essential factors to four: Oct-3/4, KLF4, SOX2, and c-Myc. Takahashi and Yamanaka designated these cells iPS cells, for Induced Pluripotent Stem cells. To test whether the cells were functionally pluripotent as well, they injected the iPS cells subcutaneously into nude mice to generate teratomas. The ability of cells to form teratomas that give rise to tissues characteristic of all three germ layers (endoderm, ectoderm and mesoderm) is considered a reliable assay of pluripotency. The majority of iPS lines generated with the four factor combination displayed this type of pluripotency. However, the definitive test of pluripotency in the mouse system is formation of a chimera by contribution to all tissues of a normal embryo after injection of the cells into a mouse blastocyst. When these original iPS cells were tested in the chimera assay, embryos with contribution of the iPS cells to tissues of all three germ layers could be found at embryonic day 13. However, no live-born chimeric pups were recovered.
This landmark paper was met with great excitement, and raised some intriguing questions. For example, the authors concluded that iPS cells were similar, but not identical to, ES cells based on gene expression analysis and their behavior in the chimera assay. What was the nature of the relationship between the two cell types? Another question concerned the low frequency at which iPS clones were recovered, raising the possibility that rare, somatic stem cells are the only ones able to be induced to the more pluripotent state. Although they observed equivalent frequency of iPS derivation from cell populations enriched in somatic stem cells (like mouse bone marrow stroma), arguing against this model, the issue was not resolved in their publication.9
The key to generating iPS cells that more closely resembled ES cells turned out to be changing the gene used for selection. Three independent groups (including Yamanaka's) took this approach and all three generated iPS cells that, as assessed by global patterns of gene expression and performance in the chimera assay, were essentially identical to mouse ES cells.10, 11, 12 Okita et al. used activation of Nanog instead of Fbx15 as their selectable marker, reasoning that activation of a gene essential (Nanog), as opposed to dispensable (Fbx15) for pluripotency might be a more stringent criterion.10 Compared to iPS cells selected by Fbx15 expression, Nanog iPS cells expressed consistently higher levels of key ES cell marker genes including the endogenous Nanog gene, E-ras, Esg1, and Rex1. In addition, the retroviral constructs used to induce the iPS phenotype were silenced more effectively in the Nanog selected cells, suggesting they were more similar to ES cells, which are known to silence retroviral sequences. Other differences between the Fbx15 and Nanog selected cells included the frequency at which clones were recovered (10-fold lower with Nanog selection) and stability (higher with Nanog selection). Most importantly, in contrast to Fbx15 selected iPS cells, the Nanog selected cells contributed to chimeras that were born and survived to adulthood, even colonizing the germline and being transmitted to the next generation. However, this markedly improved performance in the chimera assay led to the discovery of one of the key drawbacks of the four factor reprogramming scheme – nearly 20% of the F1 mice developed tumors caused by reactivation of the c-Myc retrovirus.
Wernig et al. used activation of Nanog and also of Oct-3/4 as selectable markers, and obtained similar results to Okita et al.: cells phenotypically and functionally equivalent to mouse ES cells were generated at a frequency of 0.1%.11 The authors performed additional analysis of the epigenetic status of the iPS cells, showing that the pattern of histone modification more closely matched the pattern seen in ES cells as opposed to differentiated somatic cells. They also demonstrated that silencing of the retroviral transgenes occurred, speculating that this was necessary to allow the iPS cells to differentiate in vitro or in the context of a chimeric embryo. This group also demonstrated chimera contribution and germline transmission of the cells, but did not report any tumors.
Maherali et al. also used activation of Nanog for selection, showing X chromosome re-activation during reprogramming, and random X inactivation during differentiation of the iPS lines.12 This group also generated high percentage germline chimeras.
Two key results were established by the data of these three groups. First, expression of the four reprogramming factors is only needed transiently, since iPS cells silenced transgene expression but maintained the iPS phenotype. Apparently, this occurred by activation of the endogenous copies of the pluripotency associated genes. The second key result was that reprogramming is a gradual process with intermediate stages – for example, Fbx15 selected cells appear to be partially, but not fully, reprogrammed.
View Larger Image
|Figure 2. Generation of Pluripotent Stem Cells. In this figure, the classic technique for isolation of embryonic stem (ES) cell lines (right) is compared with the recently developed method for generating induced pluripotent stem (iPS) cells (left). ES cells are derived from the population of cells in the blastocyst (pre-implantation embryo) destined to give rise to all tissues of the embryo proper, the inner cell mass (ICM). Blastocysts are plated on a layer of mouse embryonic fibroblast feeder cells (MEF, not shown) and colonies of ICM-derived cells are picked and propagated, giving rise to an ES cell line.
In the case of iPS cells, somatic cells from skin or other tissues are cultured in vitro and transduced with expression vectors encoding transcription factors associated with pluripotency. For most cell types, four factors (c-Myc, Oct-3/4, SOX2, and KLF4) are used, although an alternate list (Oct-3/4, SOX2, Nanog, and LIN-28) or all six factors combined has also been used successfully. Expression of these exogenous factors triggers a gradual process of reprogramming whereby markers of the differentiated phenotype are silenced and markers of the pluripotent state are induced in some cells. The resulting colonies of iPS cells very closely resemble ES cells as assessed by gene expression, behavior in teratoma and chimera assays, and epigenetic status. Although questions remain about the exact relationship between iPS and ES cells, both are anticipated to be valuable tools for drug discovery and testing, as well as stem cell therapies and regenerative medicine.
The next big question, then, was whether the four factor combination could be used to coax human somatic cells to an ES-like, iPS phenotype. The obvious benefits of generating pluripotent human stem cells that could be made patient-specific without the use of human embryos, led to enormous interest in this question.
Several challenges faced researchers hoping to recapitulate the mouse results in human cells. The genetic tools used in the mouse system were not as readily available in the human system. If the efficiency was similarly low, investigators would be forced to search for iPS "needles" in the overall transduced population "haystack" without the aid of selection. Retroviral transduction rates in human cells are lower, and human ES cells have different growth condition requirements to maintain pluripotency. Nevertheless, in remarkably short order, human iPS cells were reported by three groups.13, 14, 15
Yamanaka's group used adult human dermal fibroblasts as a starting population, and optimized retroviral transduction efficiency by introducing a mouse retroviral receptor (Slc7a1) into the human cells.13 iPS colonies were isolated based on morphology and then shown to express hES cell markers including Oct-3/4, Nanog, SOX2, GDF3, Rex1, FGF-4, ESG1, DPPA2, DPPA4, and hTERT. Retroviral silencing occurred in the human iPS cells as well, allowing their pluripotency to be demonstratated by differentiation in vitro into cell types with markers characteristic of all three germ layers. The cells were also differentiated into dopaminergic neurons and cardiac muscle, as assessed by multiple markers for each cell type. The iPS cells showed tri-lineage differentiation in the context of teratomas. Efficiency of reprogramming for the human cells was low: 0.02%, as compared to 0.1% for mouse cells.
Thomson's lab took a slightly different route to human iPS cells.14 Seeking to avoid the potential tumorigenicity of transduction with c-Myc, his lab developed an alternate menu of reprogramming factors, Oct-3/4, SOX2, Nanog, and LIN-28. In addition, they engineered a selectable line of human ES cells in which the endogenous Oct-3/4 promoter controlled expression of a drug resistance gene. Using this line and a strategy analogous to Takahashi and Yamanaka's original approach in mice, they started with a list of 14 potential reprogramming factors which they narrowed down to four candidates. However, instead of in vivo-differentiated tissues containing the selectable marker, they had to use cells differentiated in vitro from the engineered ES cells, and use the four reprogramming factors to revert the differentiated ES cells back to an undifferentiated state. They then confirmed the ability of the four identified factors to reprogram primary human cells as well, using fetal and post-natal fibroblasts. Like the cells generated by Yamanaka's group, these iPS cells also expressed characteristic ES cell pluripotency markers and differentiated to cells from all three germ layers in the teratoma assay.
The third group to produce human iPS cells was led by Daley, and also started with Oct-3/4 based selection from genetically modified, then differentiated hES cells.15 ES-derived fibroblasts were retrovirally transduced with the original four factor cocktail, Oct-3/4, SOX2, KLF4, and c-Myc, and iPS cells were isolated with the relatively high efficiency of 0.1%. To confirm that primary cells could also be reprogrammed, two types of fetal fibroblasts were used and successfully generated iPS clones. However, no iPS lines could be isolated from more developmentally mature cells, including neonatal fibroblasts, adult mesenchymal stem cells, and adult fibroblasts, unless two additional factors, hTERT and SV40 large T antigen were added. Park et al. also examined methylation patterns in the human iPS cells, and showed that they resembled the patterns seen in hES cells more than patterns seen in the parental fibroblast.15
To address the question of whether some cell types may be reprogrammed more easily than others, subsequent analyses have focused on alternate starting populations, mostly in mouse. For example, Yamanaka's group tested epithelial cells derived from adult mouse liver and stomach.16 Both primary hepatocytes and gastric epithelial cells could give rise to iPS cells when transduced with Oct-3/4, SOX2, KLF4, and c-Myc and selected based on Fbx15 activation. These iPS lines also generated teratomas and contributed to adult chimeras including germline transmission for some. Interestingly, these results were in contrast to their previous work,9, 10 which showed that for embryonic or tail fibroblasts, selection with Nanog, but not with Fbx15, results in germline competent iPS cells. Tumorigenicity also varied between these endoderm-derived iPS cells as compared to the fibroblast derived iPS lines, with far fewer tumors resulting from the stomach and liver iPS cells. These differences are apparently due to a less prominent role for c-Myc in reprogramming of stomach and liver cells. Stadtfeld et al. used insulin-positive mouse pancreatic beta cells as a starting population and demonstrated reprogramming at efficiencies comparable to those observed with fibroblasts (0.1-0.2%).17 Thus, at least in mouse, a variety of fully differentiated cells derived from different organs can be successfully reprogrammed.
There is widespread agreement that current methods of producing iPS cells are not suitable for generating cells that could be used in clinical applications. Concerns center not only on c-Myc, but also on the numerous retroviral integration sites, which have the capacity to be mutagenic or oncogenic themselves. Thus, reprogramming without c-Myc, with fewer factors in general, or, ultimately, with small molecules that can substitute for the factors, is highly desirable. Toward this end, Nakagawa et al.18 showed that iPS cells can be generated from mouse and human fibroblasts without the use of c-Myc. A longer gap between transduction and application of the selective agent was necessary to allow such three factor iPS cells to be isolated, indicating that the reprogramming process is slower without c-Myc. The resulting cells contributed to adult chimeras but did not cause tumors. Human dermal fibroblasts could also be reprogrammed without c-Myc, though at significantly lower frequency than when c-Myc was included.
To minimize the number of factors that need to be supplied exogenously, Kim et al. sought starting populations that express higher endogenous levels of the reprogramming factors.19 For example, adult mouse neural stem cells (NSC) express high levels of endogenous Sox2 and c-Myc, and thus might require fewer exogenous factors. In fact, adult NSCs could generate iPS cells with the addition of just two factors, either Oct-3/4 and KLF4, or Oct-3/4 and c-Myc. These "two factor" iPS cells expressed characteristic ES cell markers and gave rise to all three germ layers in teratomas and chimeras. Furthermore, no tumors were seen in the chimeric mice. This approach will undoubtedly be tested in human NSC soon.
NSCs were also the starting population in a study that screened small molecules for their ability to substitute for one or more of the reprogramming factors. Shi et al. identified an inhibitor of G9a histone methyltransferase (BIX) that improved reprogramming efficiency of NSC transduced with two factors.20 In addition, the presence of BIX enabled generation of iPS cells from fetal mouse NSC in the absence of Oct-3/4 transduction if the other three factors were provided.
Many fascinating questions remain about iPS cells, in particular the mechanism(s) by which reprogramming occurs. Results from several labs have demonstrated that reprogramming is a slow process, and that expression of the exogenous reprogramming factors is only required initially, and is later dispensable. To allow for more controlled induction of reprogramming, two groups generated vectors with inducible copies of the four reprogramming factors.21, 22 Target fibroblasts were infected with the inducible construct and addition of doxycycline (Dox) was used to activate expression of the reprogramming factors. By temporally controlling expression of the reprogramming factors, these groups defined a sequence of events that occurs during the reprogramming process. They showed that at least eight days of expression of the four factors is required before iPS cells can maintain the pluripotent state independently. Their studies also demonstrated that one of the first events in reprogramming is the downregulation of markers characteristic of the differentiated starting fibroblasts, for example Thy1. Next, a subset of the Thy1 negative population gains expression of the pluripotency marker SSEA-1. Upregulation of endogenous SOX2, Nanog, and Oct-3/4 are late events in the reprogramming process. Wernig et al. took this approach one step further, using the Dox-induced iPS cells to generate chimeric mice.23 Primary cells isolated from different tissues and organs of the chimeras were then induced to undergo secondary reprogramming simply by adding doxycycline. By allowing isolation of intermediate populations, the inducible strategy should provide additional insight into the mechanism of reprogramming. A better understanding of the mechanism should in turn lead to the development of alternative, safer ways to generate iPS cells. But even if iPS cells are successfully generated without retroviral integration, questions and challenges remain before they can be considered practical for clinical use. Accumulated somatic mutations that are silent in differentiated dermal fibroblasts might not be silent in iPS cells or their targeted differentiated derivatives. There is some evidence that different hES lines vary in their propensity to differentiate along specific lineages,24 this may be true of iPS lines as well. The risk of teratoma formation by transplanted ES cell derivatives is also a concern for iPS derivatives.
Nevertheless, there are many benefits of iPS cells that can be realized immediately. For example, ten disease-specific iPS lines were recently generated by Park et al.25 Differentiation of such lines to the cell or tissue type affected by the disease in question can be used to produce cells for drug screening or to elucidate disease etiology. iPS cells may also provide an avenue to understand the nature of pluripotency, a long standing mystery of cell biology that is likely to have dramatic consequences for the future of regenerative medicine.