Similar to the beauty we experience the first time we see a good magic act, the day the seminal description of reprogramming by Takahashi and Yamanaka was published in Cell , most people in the field thought: this must be magic. But as with any magic trick, this “act” of nuclear reprogramming was the result of a series of elegant and rigorous experiments performed to convince an incredulous audience that this phenomenon was real and not magic.
In this review, I will attempt to give a brief overview of the different methodologies for the generation of induced pluripotent stem cells (iPSCs), that have emerged since that seminal publication, reviewing each methods' properties, advantages, and disadvantages. The overall goal is to provide an understanding of the rapid evolution of the reprogramming field that has taken place in the last 5 years and to give a glimpse of where we are heading in the future.
In their original manuscript, Yamanaka and colleagues chose to use gammaretroviruses (also known generally as simple retroviruses) derived from the Moloney murine leukemia virus to introduce 24 individual transcription factors and ultimately the famous four “Yamanaka factors,” namely Oct3/4, Klf4, Sox2, and cMyc (OKSM). It is important to note that in this study , Yamanaka and colleagues were able to obtain iPSC by selecting for Fbx15-driven antibiotic resistance in ESC culture conditions. Fbx15, however, is not essential for the maintenance of the pluripotent state, which, together with the timing of drug selection may explain why the first iPSC lines failed to generate adult chimeric mice and exhibited a global gene-expression profile that was not identical to that of ESC. Indeed, by changing the selection method and culture conditions, the follow-up studies that also relied on the use of gammaretroviruses were able to derive germline-competent iPSC from mouse fibroblasts [2–4].
The successful use of gammaretroviruses was no coincidence, and I will even dare to say that the experiment would have failed if it were not for the use of retroviruses. Retroviral vectors are by far the most well studied and used vectors for gene transfer into mammalian cells, due in part to their ability to integrate their genomes into the host chromosomes, which enables efficient and long-term gene expression. Integrating viruses have evolved over million of years to optimally use the host transcription machinery for expression of the viral transduced transgenes. However, retroviral vectors are prone to epigenetic silencing [5, 6] and herein lies the basis for my second assertion. Had Yamanaka chosen a different methodology such as lentiviral vectors whose constitutive promoters are less sensitive to silencing he would have likely failed to obtain iPSC able to properly differentiate, form mouse chimeras, and contribute to mouse germline. This would have made it difficult to convince himself and the rest of the stem cell world that his reprogrammed cells were truly pluripotent. One can speculate as to why he did not use a more transient or nonintegrating methodology and, as I will explain in the next few pages that would have also fallen short of his expectations. In summary, by using retroviruses, Yamanaka found a good balance of sufficient amounts of gene expression, that last just long enough (due to silencing) to allow for the emergence of the “magic” iPSC.
But a good balance is not a perfect one. The low efficiency associated with premature silencing and the use of multiple individual retroviruses, together with the desire for developing more “user-friendly” approaches, prompted investigators including myself to look for alternative reprogramming methodologies  (Fig. 1). Moreover, the fact that these vectors do integrate, while convenient in terms of appropriate levels of gene expression, posed a critical issue in terms of safety, specifically when aberrant expression of cMyc and probably also of the other reprogramming transgenes is known to induce oncogenic transformation .
The first to use lentiviruses for iPSC generation, a close cousin of gammaretroviruses, was the laboratory of Ramalho Santos . In contrast to gammaretroviruses, lentiviruses can transduce nondividing cells and they are capable of transducing human cells more efficiently than gammaretroviruses. However, in that original study the efficiency of reprogramming was not much improved and it was not clear as to whether using a lentivirus with a constitutive promoter allowed for the generation of fully pluripotent cells, capable of differentiating appropriately in the presence of constitutive overexpression of the reprogramming factors. Generation of individual lentiviral vectors carrying inducible (tetracyclin responsive) promoters was a clear improvement to this technology [9, 10]. The ability to shut down expression of exogenous transgenes by simply removing doxycycline from the culture media allowed confirmation that nuclear reprogramming was achieved through activation of the endogenous stem cell transcription machinery and also proved to be a valuable tool for the development of secondary iPSC systems and the study of the dynamics and molecular mechanisms underlying nuclear reprogramming.
Remarkably, the fact that shortly after the original Yamanaka's report, again Yamanaka's laboratory and teams led by James Thomson in Wisconsin and George Daley in Boston were able to produce iPSCs from human fibroblasts using a similar experimental design [11–13] served to confirm the robustness of Yamanaka's findings and to give a major push forward to the use of iPSC in regenerative medicine. The resulting human iPSCs were strikingly similar to human ESCs, judged by morphology, surface marker expression, methylation status in the promoter regions of pluripotency-associated genes, in vitro differentiation, and teratoma formation. Following these first studies, retroviruses were used to reprogram somatic cells from patients with a variety of diseases [14, 15] including amyotrophic lateral sclerosis, Parkinson's disease, type 1 diabetes mellitus, Huntington's disease, and Down syndrome, providing an unprecedented opportunity for disease modeling and drug screening.
An important step to bring this technology closer to the clinics was achieved with the design of polycystronic vectors expressing all factors from a single construct. Different versions were published approximately at the same time, either as a transfectable plasmid by Yamanaka himself , a platform that suffered from very low efficiency, or as lentiviruses by our laboratory at Boston University and Rudolf Jaenisch at Massachusetts Institute of Technology [17, 18]. While the concept was the same, significant differences in terms of efficiency of reprogramming were found among these different vectors (ranging from 0.0001% to 0.5%), mostly due to the specific engineering design dependent on either the combination of 2A peptides with an internal ribosomal entry site (IRES) element versus the use of tandem 2A peptides alone. In our hands, the latter suffered from inefficient “cleavage” in the downstream 2A peptides affecting the overall production of the multiple protein products (unpublished), which could impart a disadvantage compared to the other design that relies on 2A peptides and IRES. Furthermore, the specific order of the genes within the polycystronic cassette, which allows for a characteristic stoichiometry of protein expression, likely played a role in the efficiency of reprogramming . Indeed, the specific design of our STEMCCA vector has allowed us and others to consistently achieve reprogramming of both mouse and human cells to obtain iPSC clones containing a single vector integration [18–20]. An obvious immediate consequence and application was the addition of loxP sites to make the polycystronic cassette excisable on Cre exposure, a design demonstrated again by a few different laboratories [20, 21]. In contrast to multiple vectors, the use of a polycystronic vector appears to achieve a more reliable reprogramming to generate iPSC that are transcriptionally closer to ESC . It must be noticed, however, that when using excisable lentiviruses, even after removal of most exogenous sequences, a residual inactive long terminal repeat (LTR) (∼200 bp) remains integrated within the host chromosome. Its potential safety threat by insertional mutagenesis, while still present, could be minimized by further sequencing of the proviral integration site. Importantly for its future application in the clinical arena, there is to date no published data supporting an oncogenic risk based on the presence of an integrated inactive LTR. Quite the opposite, the use of inactive LTRs has been shown to significantly diminish those risks . Furthermore, in a seminal but often forgotten study by the laboratory of Verma and colleagues, the oncogenic potential of virally mediated integration was 100% correlated with the transduced transgene (the common receptor γ chain in that specific study) and absent when the same integrating control lentiviral backbone was used . There will be a need in the future to carefully weigh the benefits of using integrating methodologies against their potential risks when attempting to move iPSC technology forward to the bedside. In this regard, having a small residual genetic tag (such as the inactive LTR) could serve a beneficial purpose and could be welcomed by regulating agencies, as it will allow investigators to more rigorously define the contribution, distribution, and in vivo function of the pluripotent derived cells (see letter by Ellis et al. ).
A major breakthrough based on a modification of this approach was the use of excisable transposon elements, such as piggyBac transposons, expressing all reprogramming factors also from a polycystronic message [26, 27]. Transient exposure of the cells to a specific transposase achieves seamless excision of the reprogramming cassette and the generation of genetically unmodified iPSC. This approach was received with much excitement by the stem cell community, however, the overall low efficiency together with the need for a laborious screening before and after transposon removal and the potential genomic toxicity mediated by transposase activity have so far limited its generalized applicability.
Significant efforts have been devoted to develop approaches to induce nuclear reprogramming using nonintegrating methodologies. These include nonintegrating viral vectors such as Adenovirus and Sendai virus, as well as direct transfection of plasmids, RNA, proteins, and finally, the use of chemicals and small molecules aimed at recapitulating the reprogramming role of the OKSM factors. The first methodology to demonstrate that genomic integration was not necessary for reprogramming to occur was the study by Stadtfeld, Hochedlinger and coworkers using individual Adenoviruses , followed almost at the same time by Yamanaka's group using transfection of DNA plasmids . Both methods, however, were limited by orders of magnitude of lower efficiencies due to the transient nature of gene expression. Adenoviral vectors have been shown to be able to reprogram human cells as well . The use of nonintegrating RNA Sendai viruses appeared to improve these methodologies, by achieving both the generation of genetically unmodified iPSC and a relatively high efficiency of reprogramming [30, 31]. The latter might well be the result of the very high number of viral copies obtained in each infected cell and the availability of viral “in sense” RNA ready-to-be translated into the reprogramming proteins. Time will serve to confirm the benefits of this approach as more publications reporting the use of this methodology appear in the near future. Other integration-free vectors have been described, including self-replicating selectable episomes  and minicircle vectors , the former requiring the use of additional factors such as SV40LT. Recently, a long sought-after method was reported by the laboratory of Rossi and coworkers , in which a sophisticated use of modified RNAs encoding OKSM factors achieved high efficiency of reprogramming while minimizing the adverse effects of interferon mediated anti-RNA responses. While some of the highly stringent technical aspects of this methodology may prevent its widespread use by laboratories in the field, it holds great promise for its use in clinical applications.
The use of purified proteins to achieve reprogramming has been hailed by many as the ultimate method to give the iPSC field its chance to have an impact at industrial scale. Indeed, for biotechnology and pharmaceutical companies the use of purified proteins represents a more appealing way to scale up this technology for commercialization. So far, only a few laboratories have succeeded in demonstrating the feasibility of obtaining reprogramming of mouse and human cells using either Tat-mediated transfection of recombinant purified OKSM proteins , or using whole cell extracts from cells expressing the reprogramming factors [36, 37], albeit at such a low efficiency that its practical implications and general applicability for now appear far from reality.
The low efficiency obtained with most nonintegrating methods has prompted investigators to screen for chemical compounds and small molecules that promote reprogramming. Indeed some molecules were found to increase the efficiency of reprogramming in the context of OKSM overexpression and even to replace individual factors, giving rise to the tempting idea of generating iPSC solely with chemicals (reviewed in [7, 38, 39]). But once again, low efficiency of reprogramming limited the use of chemicals, and so far, reprogramming solely with chemicals has been unsuccessful. Importantly, it must be cautioned that most of these molecules are potent modifiers of DNA and chromatin, and therefore, they may introduce undesired epigenetic abnormalities in the resultant iPSC. Finally, a most recent study by Morrisey and coworkers demonstrated a novel and different mechanism for nuclear reprogramming based on expression of a single miRNA cluster . In this study, lentiviral-mediated expression of miR302/367 was shown to induce reprogramming of both mouse and human cells at higher efficiency and more rapidly than compared to other methodologies. The exact mechanisms underlying miRNA-mediated reprogramming is still unknown but appears to be mediated in part by direct activation of OCT4 expression and suppression of Hdac2 .